Background

Down syndrome (DS) has been conceptualized by the International Working Group for Alzheimer’s disease (AD) as a genetically determined form of AD [1] due to the triplication of the gene encoding amyloid precursor protein (APP), among others [2]. AD in DS has a long preclinical phase and a predictable sequence of changes in biomarkers that is strikingly similar (both in order and temporality) to that described in autosomal dominant AD [25]. All adults with DS included in the study had confirmed trisomy of chromosome 21.

CSF collection and biomarker assessment

CSF samples were collected following international consensus recommendations [26] as previously described [27]. Samples had been previously stored at − 80 °C and had not been thawed prior to analysis. Commercially available automated immunoassays were used to determine levels of CSF Aβ1–42 (Lumipulse Aβ1–42, Fujirebio-Europe, Belgium) total tau (Lumipulse G Total Tau, Fujirebio-Europe) and tau phosphorylated at threonine residue 181 (Lumipulse G PTAU 181, Fujirebio-Europe) in the DABNI cohort and SPIN controls [17]. An in-house enzyme-linked immunosorbent assay targeting the N-terminal fragment encoding amino acid 1 to 201 of NPTX2 was performed in the Baltimore Laboratory as previously described [12].

Targeted liquid chromatography mass spectrometry

We monitored a set of 3 proteotypic peptides corresponding to GluA4 by Selected Reaction Monitoring (SRM) as previously described [28]. Briefly, we digested individual CSF samples overnight and spiked isotopically-labeled peptides (Pepotech SRM custom peptides, grade 2, Thermo Fisher Scientific) into each sample. We analyzed an equivalent of 5ul of each sample in a randomized order over a 120-min gradient (0–35% ACN + 0.1%FA) in SRM mode using a triple quadrupole-Qtrap mass spectrometer (5500 QTrap, Sciex, Massachussetts) coupled to a nano-LC chromatography column (300 ul/min, 25-cm C18 column, 75 um I.d., 2 um particle size). We ran BSA technical controls between each sample. We visualized and analyzed transitions using Skyline 3.5. Injection of a pool of all the samples over the duration of the mass spectrometric measurements and monitoring the peak area of the standard peptides confirmed the stability of all peptides over the course of the experiment. We processed the SRM transitions using the dataProcess function of MSstats v3.5 package in R [29] and removed transitions with between run interference (betweenRunInterferenceScore< 0.8). Censored missing values were samples with log base-2 endogenous intensities under the cut-off designated by the MSstats package (10.35). We used the EqualizeMedians function to normalize the transitions and Tukey’s Median Polish to generate a summarized value.

Magnetic resonance imaging (MRI) acquisition and analysis

A high-resolution three-dimensional structural dataset was acquired in a 3 T MRI scanner (Philips 3.T X Series ACHIEVA) with the following parameters: T1-weighted magnetization-prepared rapid gradient-echo, repetition time 8.1 msec, echo time 3.7 msec, 160 slices, matrix size 240 × 234; slice thickness 1 mm, voxel size 0.94 × 0.94 × 1 mm. Cortical thickness reconstruction was performed with Freesurfer package v6 (http://surfer.nmr.mgh.harvard.edu) using a procedure that has been described in detail elsewhere [30], as previously reported [28, 29]. A Gaussian kernel of 15 mm full width at half maximum (FWHM) was applied to the subjects’ cortical thickness maps before further analyses as it is customary in surface-based analyses. From the 75 adults with DS who had available MRI, 14 were excluded due to suboptimal image quality, which included subtle movement artifacts, poor signal to noise ratio, and gradient artifacts. The breakdown of diagnosis in these individuals was as follows: aDS n = 38, pDS n = 15, dDS n = 8.

FDG-PET acquisition and analysis

From the remaining 61 adults with DS with MRI, 42 underwent [18F]-fluorodeoxyglucose (FDG) positron emission tomography (PET) scan on an integrated PET-TC system (Philips Gemini TF). The breakdown of diagnosis in these individuals was as follows: aDS n = 28, pDS n = 10, dDS n = 4. Overnight fasting subjects’ plasma glucose was measured and confirmed to be < 180 mg/dL prior to starting the scanning procedure. Data were acquired 50 min after the injection of 259 MBq/ml (7 mCi) of [18F]-FDG. After computerized tomography data was obtained, brain PET was acquired. The reconstruction method was iterative (LOR RAMBLA, 3 iterations and 33 subsets) with a 128 × 128 image size, 2 mm pixel size and 2 mm pixel slice thickness. [18F]-FDG-PET images were intensity-scaled by the reference pons-vermis region [31], spatially normalized to the Freesurfer anatomical space using a PET surfer approach [32] and then smoothed with a 10 mm FWHM kernel in order to obtain similar relative smoothing FDG maps as compared with cortical thickness maps [33, 34]. With this surface-based approach, the reliability of PET effects improves and the inter-individual variance is reduced [35]. All resulting images were visually inspected to check for possible registration errors.

Statistical analysis

Statistical analyses were performed in R version 3.4.3 [36]. Group differences were compared using One-way Analysis of Variance (ANOVA) with post hoc Tukey tests adjusted for multiple testing using the Bonferonni-Hochberg method. Robust logistic regressions were performed using the lm function included in the “robustbase” package that uses MM-type regression estimators to compute robustness weights in the fitting process. All reported r2 values were adjusted for the number of predictors in the model. Where residuals deviated from a Gaussian distribution (Shapiro-Wilk p < 0.05), square root transformed values were used. Outliers were excluded using the 3 × IQR rule. For the neuroimaging analyses we performed Monte-Carlo simulation with 10,000 repeats (family wise error [FWE] correction at p < 0.05). Due to the small size of each clinical group with available neuroimaging, analyses were performed on the complete DS dataset only. The figures show only those results that survived FWE correction. Alpha was set at 0.05 for global analyses and 0.01 for post-hoc analyses.

Results

The Table 1 shows the demographic and clinical data for the samples included in the study. The mean age was 52 years (standard deviation 15) across the whole study (n = 168). Compared to controls the mean age was higher in the sporadic AD group (+ 19 years, 95% CI 13 to 25, adj.p < 0.0001), lower in the aDS group (− 14 years, 95% CI 8 to 19; adj.p < 0.0001) and comparable in pDS and dDS (both adj.p > 0.9). As would be expected for an age-related disease such as AD, mean age was lower in aDS than pDS (− 14 years, 95% CI − 7 to − 21; adj.p < 0.0001) and dDS (− 15 years, 95% CI − 9 to − 21; adj.p < 0.0001). The proportion of males:females included in the study was 46:54 and was comparable across diagnostic groups (p = 0.09). The proportion of individuals with severe/profound intellectual disability was comparable across AD stages (p = 0.7). All cognitive measures were lower in sAD compared to controls (all adj.p < 0.0001) and in dDS compared to aDS (all adj.p < 0.0001). CSF biomarker levels were as previously reported [17]. Briefly, compared to controls mean Aβ1–42 levels were lower in all groups (all adj.p < 0.0001) and p-tau and t-tau levels were higher in sAD, pDS and dDS (all adj.p < 0.0001).

Table 1 Demographics and clinical data for study participants
Full size table

CSF NPTX2 levels are similarly reduced in DS and sporadic AD

Mean CSF NPTX2 levels were associated with clinical diagnosis (Fig. 1a; n = 168, p < 0.0001). Specifically, post hoc analyses showed that compared to controls, mean NPTX2 levels were 0.5-fold in sporadic AD (− 491 pg/ml, 95% CI −280 to − 702; adj.p < 0.0001), 0.6-fold in aDS (− 400 pg/ml, 95% CI − 196 to − 604; adj.p < 0.0001), 0.5-fold in pDS (− 472, 95% CI − 208 to − 736; adj.p < 0.0001) and 0.3-fold in dDS (− 595, 95% CI − 361 to − 828; adj.p < 0.0001). Mean NPTX2 levels were comparable between sporadic AD and DS (all AD stages adj.p > 0.3) and across AD stages in DS (all adj.p > 0.07).

Fig. 1
figure 1

NPTX2 levels in control, sporadic AD and DS CSF. Box and whisker plots and individual points of the CSF levels of NPTX2 (pg/ml) are grouped (a) by clinical diagnosis and (b) by degree of intellectual disability (DS population only). P-values resulting from one-way ANOVA on square root transformed NPTX2 levels are shown in the top right corner. P-values resulting from Bonferroni-adjusted post-hoc Tukey test are shown where p < 0.05

Full size image

In adults with DS, CSF NPTX2 levels were not associated with level of intellectual disability (Fig. 1b; n = 94, p = 0.7) and were not associated with any measure of cognitive performance in DS (n = 57–64, all p > 0.07). Mean NPTX2 levels were not associated with age overall (p = 0.6) or at any AD stage (all p > 0.1).

Low CSF NPTX2 levels correlate with low CSF levels of a PV-interneuron specific receptor

We next sought to determine whether the PV-interneuron specific marker, GluA4, shows a coordinated reduction with NPTX2 in the CSF. While CSF GluA4 levels were not associated with clinical diagnosis (Fig. 2a; p = 0.2), low NPTX2 levels correlated with low GluA4 levels in controls (n = 27 r2 = 0.2, p = 0.003), sAD (n = 20, r2 = 0.4, p < 0.0001) and in DS (n = 81, r2 = 0.4, p < 0.0001): The association in DS was apparent at all AD stages; aDS (n = 36, r2 = 0.5, p = 0.04), pDS (n = 18, r2 = 0.5, p = 0.005) and dDS (n = 21, r2 = 0.4, p = 0.02).

Fig. 2
figure 2

GluA4 levels in control, sporadic AD and DS CSF. a Box and whisker plots and individual points of the CSF levels of GluA4 (log2 intensity) grouped by clinical diagnosis. P-value resulting from one-way ANOVA on square root transformed GluA4 levels is shown in the top right corner. b CSF NPTX2 levels are plotted against CSF GluA4 levels with associated linear regression lines shown for controls, sporadic AD and DS groups (see legend)

Full size image

Low CSF NPTX2 levels correlate with CSF markers of AD pathology in adults with DS

We next sought to determine the relationship between CSF NPTX2 levels and core CSF AD biomarkers in DS. Low NPTX2 levels were associated with low CSF Aβ1–42 levels (Fig. 3a) in all diagnostic groups; aDS (r2 = 0.5, p < 0.0001), pDS (r2 = 0.3, p = 0.0003) and dDS (r2 = 0.6, p = 0.006). Low NPTX2 levels were also associated with p-tau (Fig. 3b) in all diagnostic groups; aDS (r2 = 0.4, p < 0.0001), pDS (r2 = 0.3, p = 0.0004) and dDS (r2 = 0.3, p = 0.001). Similar correlations were observed for CSF t-tau (r2 = 0.3 to 0.4, p < 0.003).

Fig. 3
figure 3

Correlation between CSF NPTX2 levels and CSF markers of AD pathology. CSF NPTX2 levels are plotted against (a) CSF Aβ1–42 and (b) CSF p-tau with associated linear regression lines shown for each clinical diagnosis (see legend)

Full size image

Low CSF NPTX2 levels in adults with DS are associated with cortical atrophy and neuronal dysfunction in signature AD brain regions

We also analyzed the relationship between CSF NPTX2 levels and structural and functional brain changes in adults with DS. As shown in Fig. 4, low CSF NPTX2 levels were associated with reduced cortical thickness in the temporal and parietal cortices (top; n = 63, FWE < 0.05) and with reduced glucose metabolism in more widespread regions of the temporal and parietal cortices (bottom; n = 44, FWE < 0.05).

Fig. 4
figure 4

Structural and functional neuroimaging correlates of CSF NPTX2 levels in DS. Correlation of CSF NPTX2 levels and cortical thickness (top) and glucose metabolism (bottom) in all DS participants. Regions in red represent significant correlations (p < 0.05). Only the clusters that survived family-wise error correction p < 0.05 are shown. Scatterplots with linear regression line (black) and standard error (gray) show the correlation between CSF NPTX2 and neuroimaging biomarkers by diagnosis in a representative cluster that survived multiple comparisons, indicated with an asterisk. Regression lines are shown for each clinical group for informative purposes (see legend)

Full size image

Discussion

Neuronal Pentraxin Receptor 2 (NPTX2) has recently been identified as a novel CSF biomarker of inhibitory circuit dysfunction in sporadic AD [12, 15]. This finding is consistent with studies that report a similar reduction in NPTX2 expression in postmortem sporadic AD and DS brains [12] and indicates that the mechanism underlying these changes is common to both sporadic and genetic forms of AD and that these changes may be reflected in the CSF. As these changes were apparent across all AD stages, CSF NPTX2 may be a useful marker of AD-related changes in adults with DS even prior to symptom onset.

DS cohorts with available CSF are scarce and current biofluid markers for AD in the DS population are restricted to surrogate markers of amyloidosis (CSF Aβ1–42 or amyloid PET tracers) and tau-mediated neurodegeneration (CSF p-tau and t-tau) combined with neuropsychological assessment, which can be confounded by substantial inter-individual variation in intellectual disability. NPTX2 represents a new addition to the DS biomarker arsenal that may be used to detect early signs of AD in this relatively under-studied population. Furthermore, this work opens the door to future studies exploring the potential of CSF NPTX2 as a prognostic marker for late-onset myoclonic epilepsy in Down syndrome and sporadic AD.

The lack of association of NPTX2 levels with intellectual disability suggests that the reduction in NPTX2 levels is not neurodevelopmental. This is supported by the finding that reduced CSF NPTX2 levels were associated with increased cortical atrophy and reduced glucose metabolism in temporal and parietal regions of the brain, regions that are particularly susceptible to AD pathology. Thus, CSF NPTX2 levels inversely correlated with the extent of both neuronal loss and neuronal dysfunction in signature AD regions. Taken together, we propose that changes in CSF NPTX2 levels in adults with DS are related to AD pathophysiology and not to intellectual disability.

Low CSF NPTX2 levels in adults with DS were associated with low CSF GluA4 and low CSF Aβ1–42, a surrogate CSF marker of increased aggregation of brain Aβ1–42. This is the first evidence that a synergized increase in brain Aβ1–42 aggregates and reduced NPTX2 can be measured indirectly in the CSF of adults with DS. NPTX2 is expressed by pyramidal neurons and secreted by axon terminals to mediate activity-dependent strengthening of pyramidal neuron excitatory synapses on GABA-ergic PV interneurons, a mechanism that is dependent upon the AMPA receptor subunit, GluA4 [11]. Histological studies have shown reduced NPTX2 and GluA4 expression in signature AD regions of AD and DS brains [12], regions that also show a reduced number of PV interneurons [37]. This is consistent with the accumulating evidence for reduced inhibitory interneuronal activity in the AD brain [10]. Studies in mice [37,38,39] have shown that low NPTX2 expression at excitatory synapses of PV interneurons combined with brain amyloidosis is critical for inhibitory circuit dysfunction [12].Thus, a hypothesis emerges whereby a loss of NPTX2-GluA4 synapses in regions affected by amyloidosis may lead to inhibitory circuit dysfunction in AD. While extrapolation of data from mouse to human brain should be treated with caution, this could be one plausible explanation for the low CSF NPTX2 levels we report in sporadic AD patients and adults with DS.

Unlike NPTX2, GluA4 levels were not significantly reduced in CSF of adults with DS compared to controls, making NPTX2 a potentially more valuable CSF biomarker in DS. Based on these findings, CSF NPTX2 can be added to the growing list of biomarkers that show changes similar, both in direction and magnitude, to those described in sporadic or autosomal dominant AD [4, 20, 40]. The potential of CSF NPTX2 as a novel biomarker of interneuronal dysfunction can now be extended to DS, a genetic form of AD that targets the same cortical regions as the sporadic and autosomal dominant forms [5].

CSF Aβ1–42, t-tau and p-tau are excellent diagnostic markers in sporadic AD and in the DS population and we do not expect NPTX2 to improve upon those markers. Rather, this study supports the increasing literature that suggests CSF NPTX2 as a surrogate marker of underlying interneuronal circuit dysfunction in sporadic and, now genetic, forms of AD even at the preclinical stage. Further studies in longitudinal cohorts are needed to explore the full potential of CSF NPTX2. The capacity of NPTX2 as an early prognostic marker would be one potential use worth exploring for example. Alternatively, as tau and Aβ are common drug targets in AD clinical trials, a surrogate measure of synaptic dysfunction that does not directly measure the drug target levels would be a useful addition to the biomarker arsenal.

In previous studies in sporadic AD, CSF NPTX2 levels correlated better with cognitive performance than the core AD biomarkers [12]. Furthermore, in genetic frontotemporal dementia, low CSF NPTX2 levels predicted subsequent decline in phonemic verbal fluency and Clinical Dementia Rating scale [41]. In contrast, here we report that reduced CSF NPTX2 levels were not associated with any measure of cognitive performance in adults with DS. That being said, the high variability in intellectual disability in the DS population can complicate cognitive assessment and therefore these neuropsychological tests could only be performed in individuals with mild or moderate intellectual disability, which resulted in a limited sample size. Thus, the potential relationship between CSF NPTX2 levels and cognitive performance in the DS population warrants further study in a larger population. On the other hand, there is evidence that the temporal onset of cognitive decline may differ in DS compared to autosomal dominant AD [4], thus underscoring the importance of considering the neurodevelopmental differences in the DS population when interpreting the cognitive results.

The association of increased CSF NPTX2 levels with increased CSF p-tau levels in the DS population could indicate that widespread neuronal loss at later disease stages may lead to increased NPTX2 clearance into the CSF and may mask a reduction in NPTX2 synapses. Nevertheless, the inverse relationship we report between CSF NPTX2 levels and a neuroimaging marker of cortical atrophy, suggests that neurodegeneration has a discernable but limited confounding effect on CSF NPTX2 levels.

A limitation of this study is the cross-sectional design; longitudinal studies are needed to fully establish the prognostic value of NPTX2 in both sporadic AD and in DS. Furthermore, these findings warrant replication in an independent DS cohort. Exploration of the potential of NPTX2 as a blood biomarker would also be valuable to minimize the use of lumbar puncture in people with DS.

Conclusions

Here we report that CSF NPTX2 levels are reduced in adults with DS compared to cognitively normal, non-trisomic controls even prior to AD onset. Similar reductions were observed in the CSF of sporadic AD patients and have been previously reported in postmortem AD and DS brains [12, 15]. Substantial evidence from animal studies suggests that such changes may be associated with reduced inhibitory inter-neuronal activity. While further functional studies are needed, this could explain the correlation we report between low CSF NPTX2 and low CSF GluA4, the predominant AMPA receptor subunit in PV interneurons. In conclusion, this study shows for the first time that NPTX2 could be a valuable CSF biomarker of AD-related inhibitory neuronal circuit dysfunction in adults with DS even prior to symptom onset.