Different Inflammatory Signatures in Alzheimer’s Disease and Frontotemporal Dementia Cerebrospinal Fluid

INTRODUCTION

Alzheimer’s disease (AD) and frontotemporal de-mentia (FTD) are two of our most common neurodegenerative disorders. Whereas both diseases are characterized by neuroinflammatory processes, we have limited knowledge of how these are mirrored in the cerebrospinal fluid (CSF).

AD is characterized by cerebral accumulation of amyloid-β (Aβ) into plaques, aggregation of hyperphosphorylated tau into neurofibrillary tangles, and neuroinflammation associated with activation of mic-roglia and astrocytes [1]. During the disease development, patients pass through a stage of mild cognitive impairment (MCI), which may occur several years before the onset of dementia [2]. However, only about 50% of all MCI cases progress to AD dementia, whereas the other half remains cognitively stable even after several years of follow-up [3]. It is today largely unknown what molecular features, other than Aβ and tau mismetabolism, that distinguish MCI patients with incipient AD from non-progressors. Among different theories, neuroinflammation has been hypothesized to be such a disease determinant. More specifically, temporary inflammatory reactions have been proposed to be beneficial in early AD stages but may contribute to an accelerated neurode-generative process if becoming chronic [4]. Furthermore, many AD risk gene variants (e.g., TREM2, CD33, and APOE) have been associated with mic-roglia dysfunction, which in turn may be implicated in chronic inflammation [4]. However, current knowledge of inflammatory protein involvement in AD development is limited.

Several studies have evaluated neuroinflammatory features in AD CSF or plasma, with diverse findings such as decreased levels of IL-8 or increased levels of GFAP [5, 6]. A recent review described increases in eleven inflammatory proteins in AD in at least two CSF studies, whereas only one was decreased [7]. Of the studies reviewed, only two included MCI patients, none of which made distinction between MCI due to AD and stable MCI [8, 9].

As the second most common neurodegenerative disorder before the age of 65 [10], FTD includes the behavioral variant of frontotemporal dementia (bvFTD) as well as forms characterized by various degrees of language dysfunction, e.g., primary prog-ressive aphasia (PPA) and its subtypes nonfluent variant (nfvPPA), semantic variant (svPPA), and log-openic variant (lvPPA) [11, 12]. Similar to cellular changes in the AD brain, microgliosis and astrocytosis are typically seen also in the FTD brain [13]. Mixed results have been presented regarding CSF levels of inflammatory proteins in FTD. Some studies have shown increased CSF levels of the pro-inflammatory proteins TNF, MCP-1, IL-8, IL-11, and YKL-40 in FTD compared with healthy controls, as well as neuroinflammatory signaling on TSPO-PET [13–18]. In contrast, decreased CSF levels of pro-inflammatory IL-12 and increased levels of anti-inflammatory transforming growth factor-β (TGFβ) have also been reported in FTD [18, 19].

The primary aim of this study was to investigate the CSF signature of inflammatory proteins in AD and FTD, using a multiplex proximity extension assay (PEA). To that end, we quantified 92 such proteins, preselected by the manufacturer. A secondary aim was to assess if any of these proteins were altered already at an early stage of AD.

METHODS

RESULTS

The 184 study participants consisted of 90 women (48.9%) and 94 men (51.1%) with a mean±SD age of 69.5±9.1 years (Table 1). Women were slightly overrepresented in the AD and MCI/AD group (59.5% and 58.6%, respectively), while the FTD group consisted of more men (69.0%) with a slightly lower mean±SD age of 64.6±9.0 years (Table 1). Amyloid-β₄₂ (Aβ₄₂) CSF levels (mean ng/L±SD) were roughly 50% lower in the AD and MCI/AD group compared with the controls (AD: 415±82, MCI/AD: 439±108, Controls: 899±289), while the total-tau (t-tau) levels (mean ng/L±SD) were approximately 55–85% higher (AD: 725±296, MCI/AD: 608±235, Controls: 392±233, Table 1). Compared with the controls, the FTD group had similar Aβ₄₂ and t-tau levels (FTD versus Controls, Aβ₄₂: 1025±508 versus 899±289, t-tau: 346±156 versus 392±233, Table 1), while the stable MCI group had similar Aβ₄₂ levels but roughly one third lower t-tau levels (MCI versus Controls, Aβ₄₂: 826±203 versus 899±392; t-tau: 265±86 versus 392±233, Table 1). Of the 92 proteins included in the assay, 56 (61%) were detected in at least 75% of the CSF samples. All proteins, their abbreviations, and their detectability are available in Supplementary Table 1, while the type and function of the detected proteins are listed in Table 2. Two of the AD samples deviated more than 0.3 NPX from the median of the internal quality control samples (supplied by the manufacturer) and were therefore excluded from further analyses.

The control group and the MCI/AD group had a very similar distribution pattern of protein levels in the PCA (Fig. 1). The MCI and FTD groups deviated most from the control group, while the AD group displayed a pattern between those of the control and FTD groups. The PCA did not reveal any major differences between the three medical centers, or between the plate IDs (Fig. 1).

Fig. 1

Principal component analysis (PCA) of cerebrospinal fluid protein levels. The PCA is divided into (A) patient groups, (B) medical centers, and (C) number of plate sent for analysis. Each ellipse represents 95% of respective group’s samples. D) The detected proteins’ contribution to principal component 1 (PC1) and 2 (PC2). AD, Alzheimer’s disease dementia; C, healthy controls; FTD, frontotemporal dementia; MCI, mild cognitive impairment, cognitively stable at the MCI level; MCI/AD, mild cognitive impairment due to Alzheimer’s disease.

When assessing the individual analytes separately, matrix metalloproteinase-10 (MMP-10) was significantly increased in AD compared with controls (FC = 1.32, 95% confidence interval [CI] 1.14–1.53, q = 0.018, Fig. 2A), as well as between MCI/AD patients and controls (FC = 1.53, 95% CI 1.20–1.94, q = 0.045, Fig. 2A). It was also significantly higher in the MCI/AD group compared with the MCI group (FC = 1.38, 95% CI 1.16–1.65, q = 0.011, Fig. 2B), as well as in the FTD group compared with controls (FC = 1.42, 95% CI 1.10–1.83, q = 0.020, Fig. 2C). Upon stratification by patient group and medical center a trend toward higher levels of MMP-10 was observed in all three neurodegenerative disease groups (AD, MCI/AD and FTD), but with varying distributions among the different centers and a higher variability in the FTD group (Fig. 3).

Fig. 2

Volcano plots of group differences in cerebrospinal fluid proteins. The plots show fold change and p-values of inflammatory proteins in cerebrospinal fluid from (A) patients with AD and MCI/AD compared with healthy controls, (B) patients with MCI/AD compared with patients with MCI, and (C) patients with FTD compared with healthy controls. The proteins were analyzed using ANOVA F-tests, adjusting for age, sex, plate ID, and study site (A, C), and age, sex, and plate ID (B), respectively. Proteins with a false discovery rate below 0.05 are labeled. AD, Alzheimer’s disease dementia; C, healthy controls; FTD, frontotemporal dementia; MCI, mild cognitive impairment, cognitively stable at the MCI level; MCI/AD, mild cognitive impairment due to Alzheimer’s disease.

Fig. 3

Box plot of MMP-10 levels stratified by patient group and medical center. AD, Alzheimer’s disease dementia; C, healthy controls; FTD, frontotemporal dementia; MCI, mild cognitive impairment, cognitively stable at the MCI level; MCI/AD, mild cognitive impairment due to Alzheimer’s disease; MMP-10, matrix metalloproteinase-10; NPX, normalized protein expression.

We found eleven additional proteins to be significantly increased in MCI/AD compared with MCI (q < 0.05, Table 3, Fig. 2B), six of which were growth factors; tumor necrosis factor ligand superfamily member 12 (TNFSF12, also known as TWEAK; 46% higher), hepatocyte growth factor or scatter factor (HGF/SF, 36% higher), programmed death-ligand 1 (PD-L1, 38% higher), vascular endothelial growth factor A (VEGFA, 34% higher), nerve growth factor beta (NGF, 23% higher), stem cell factor (SCF, 37% higher), fibroblast growth factor 5 (FGF5, 34% higher), leukemia inhibitory factor receptor (LIFR, 27% higher), CUB domain-containing protein 1 (CDCP1, 29% higher), macrophage colony stimulating factor 1 (CSF-1, 22% higher), and sirtuin 2 (SIRT2, 34% higher).

Thirty-six of the proteins were decreased in FTD compared with controls (q < 0.05, Table 3, Fig. 2C), including all the eleven proteins increased in MCI/AD compared with MCI, except SIRT2, whereas only MMP-10 was increased. Of the 36 decreased proteins, eleven were growth factors, ten chemokines, and five related to TNF-signaling. In addition, several of the decreased proteins were related to T cell function, including T cell receptors, T cell receptor ligands, growth factors and chemokines (Table 2).

DISCUSSION

In this study, we have compared CSF levels of 56 different inflammation-related proteins between groups of patients with AD, MCI/AD, MCI, FTD, and healthy controls. Several findings emerge. Firstly, levels of MMP-10 were found to be increased in AD, MCI/AD, and FTD patients, compared with healthy controls. Secondly, MMP-10 and eleven other inflammatory proteins were elevated in CSF from patients with MCI/AD, as compared with MCI subjects that remained cognitively stable after several years of follow-up. Notably, none of the proteins analyzed were found at decreased levels in the AD or the MCI/AD group. Thirdly, and in stark contrast to the pattern in the AD and MCI/AD groups, 36 of the inflammatory proteins were found at decreased levels in the FTD group, compared with healthy controls. Apart from MMP-10, none of the markers were found at increased levels in the FTD patients.

Increased CSF levels of MMP-10 have previously been reported in AD and have also been found to correlate with increased levels of tau and p-tau [26]. A recently published study, using the same panel, also found increased levels of MMP-10 in CSF samples of AD and MCI/AD patients compared with healthy controls [27]. To our knowledge, the current study is the first to show increased CSF levels of MMP-10 in FTD patients. The MMPs comprise a group of enzymes that help to increase cell motility by degrading parts of the extracellular matrix. Increased activity of these enzymes has previously been associated with the disruption of the blood-brain barrier (BBB) [26]. The levels of MMPs are regulated by tissue inhibitors of metalloproteinases (TIMPs), which may in turn be reduced by cerebral microbleeds [26]. Cerebral microbleeds are common in neurocognitive disorders and have been found to increase the risk of dementia, including AD, possibly by activating a neuroinflammatory and neurodegenerative process in microglia through fibrinogen-induced spine elimination [28]. Previously, a combination of eight proteins, including TIMP-1, MMP-9, and MMP-2, but not MMP-10, was shown to discriminate patients with subcortical vascular disease from patients with AD, with a sensitivity of 89% and a specificity of 90% [29]. MMP-10 was also increased in both of these patient groups compared with controls [29]. Thus, the increased CSF levels of MMP-10 may be a result of a disrupted BBB in the AD, MCI/AD, and FTD brain. Disruption of the BBB may lead to the influx of plasma proteins into CNS, which in turn may increase the level of neurotoxicity and accelerate the neurodegenerative process [30].

As for the increase of the eleven other inflam-mation-related proteins in MCI/AD compared with stable MCI subjects, several have previously been implicated in AD or in other neurodegenerative diseases. After reviewing the putative functions of these eleven proteins, we have divided them into pro- or anti-inflammatory proteins (Table 2).

The pro-inflammatory proteins include tumor necrosis factor (TNF)-related weak inducer of apoptosis (TWEAK), vascular endothelial growth factor A (VEGFA, or VEGF), fibroblast growth factor 5 (FGF5), CUB domain-containing protein 1 (CDCP1, also known as CD318) and macrophage colony stimulating factor (M-CSF, also known as CSF-1). Many of these are involved in regulation of blood vessel permeability, induction of angiogenesis and/or mobilization or activation of pro-inflammatory cells or cytokines. In models of multiple sclerosis and amyotrophic lateral sclerosis, endothelial cells respond to stimulation by TWEAK through neuroinflammation and BBB-disruption by upregulating MMP-9, increasing the level of pro-inflammatory cytokines and attracting monocytes [31]. Increased levels of VEGFA have been associated with neurogenesis, learning, and neuroprotective effects, but may also increase vascular permeability and be neurotoxic in high concentrations [32]. FGF5 may be involved in reactive astrocytes, as FGF5 knock-out mice have been shown to have a reduced number of GFAP-positive astrocytes in midbrain tegmentum [33]. Blockage of the T cell receptor CD6 and its ligands, such as CDCP1, have shown protection from disease in mouse models of psoriasis and multiple sclerosis [34]. M-CSF, or CSF-1, binds to a receptor (CSF-1R) that is located on microglia. In a mouse model, blocking the CSF-1R has led to a blockage of microglial activation and a shift from pro-inflammatory to anti-inflammatory microglia [35].

The remaining six anti-inflammatory proteins with increased CSF levels in MCI/AD compared with MCI include hepatocyte growth factor (HGF), programmed death-ligand 1 (PD-L1), nerve growth factor (NGF), stem cell factor (SCF), leukemia inhibi-tory factor (LIF), and sirtuin 2 (SIRT2). Many of these proteins are involved in tissue regeneration, immune cell inhibition, or alterations of immune cells to anti-inflammatory phenotypes (Table 2). Previous studies have found increased levels of HGF in astrocytes and microglia surrounding senile plaques in the AD brain [36]. Blockage of PD-L1, or its receptor (PD-1), has been used in cancer therapy to activate the T cell mediated immune response, which in a mouse model of tauopathy has led to an increased recruitment of circulating monocytes to the brain and the rescue of neurons [37]. Previous studies have implicated a mismatch in NGF metabolism in AD, where higher levels of MMPs have been associated with an increased degradation of NGF and an increase in loss of cholinergic neurons [38]. Microglia exposed for SCF have shown an anti-inflammatory phenotype in vitro. Intracellular signaling cascades in neurons and oligodendrocytes, caused by LIF, leads to neurosurvival. It also alters immune cells to anti-inflammatory phenotypes [40]. Several studies have observed a positive effect of blocking SIRT2, both in models of Parkinson’s disease and AD, which stabilizes microtubules and increases the autophagic breakdown of accumulated proteins [41]. Which of these mechanisms that are relevant in the pathogenesis of early AD and their internal interactions remain to be determined. It is noticeable that more proteins differed when comparing MCI/AD with MCI than when comparing MCI/AD with controls. These results seem to be partly caused by a general decrease in CSF inflammatory proteins in stable MCI patients. However, as the MCI groups were rather small, it may be a coincidental observation and larger studies are thus needed to confirm these data.

Another main finding of this study is the general decrease of inflammatory markers in CSF of FTD brains, including almost all of the proteins found to be increased in MCI/AD compared with MCI. This finding indicates that the pathogenic processes involved in FTD may affect microglial function, inflammation and neurodegeneration in a different manner than in AD. An interesting observation is that several of the decreased proteins were related to T cell function, a research area that has not been much explored in FTD.

A combination of CSF t-tau and Aβ₄₂ has been shown to have a sensitivity of 95% and a specificity of 83% for detection of incipient AD in patients with MCI [42]. In addition, the ratio of t-tau/Aβ₄₂ gives a high sensitivity and specificity when comparing AD to FTD [43]. However, this ratio is not as useful in distinguishing FTD from healthy controls [43]. The use of inflammatory biomarkers in CSF can potentially aid clinicians in predicting which MCI cases will progress to AD dementia, as well as distinguishing patients with FTD from AD.

Both our results and previous studies indicate that elevated CSF levels of inflammatory proteins is an early feature in AD pathogenesis [27]. In postmortem studies as well as in studies of neuroinflammatory translocator protein (TSPO) positron emission tomography (PET), inflammation has been implicated at early AD stages [44, 45]. Moreover, it has been suggested to be involved in the propagation of tau pathology in AD, increasing linearly throughout the disease and reaching the highest levels in late stages [44, 45]. Some TSPO PET studies have noted an early inflammatory peak in MCI/AD, with a subsequent decline before the inflammation reaches even higher levels in AD dementia [45, 46]. In our study, there was a tendency toward higher levels of inflammatory proteins in MCI/AD than in AD dementia CSF (Fig. 2A), indicating that the different methods might measure different aspects of neuroinflammation. This contrast seems to be especially prominent in FTD, where increased levels of inflammation has been found with TSPO PET, while the levels of inflammatory proteins were dramatically lower in the present study [13]. As discussed in other studies, TSPO PET may not be able to differ between activated microglia of pro- or anti-inflammatory phenotype [45, 46]. It seems that more than one pro- or anti-inflammatory phenotype of microglia may exist and that the current measurements focusing on the morphology or the activity of microglia may miss to differentiate between those phenotypes.

A limitation of this study is the small sample sizes in the MCI/AD and MCI groups, which may have affected the precision of the results in these groups. Still, the sample sizes in the AD, FTD, and control groups were robust and, although sample sizes from the individual study sites were too small to be investigated separately, the general trends of increased MMP-10 levels in AD and FTD, as well as the dec-reased levels of inflammatory proteins in FTD, were similar between the study sites. Whereas the methods to diagnose AD and FTD were the same between the sites, the control groups were recruited differently, which affects comparability and generalizability of the results. However, all subjects in the control group were cognitively healthy and represent a group free of neurocognitive disease. As another limitation we did not have access to genetic information of the AD or FTD patients. Since FTD is a heterogeneous disease where 30–50% have a familial form, it would be of interest to see if the observed differences relate to mutations in specific genes, such as microtubule-associated protein tau (MAPT), progranulin (PGRN), or C9orf72 [47]. In general, the AD and FTD groups had higher levels of MMP-10, but the within-group variation was large, indicating that different phenotypes or subgroups may exist within these diseases, which could be attributable to disease specific mutations. Moreover, the proteins selected for analyses in this study were based on a commercially available pre-manufactured, instead of a custom-made protein panel, or on unbiased methods such as mass spectrometry. Although Olink Proteomics’ PEA technology gives a fast and accurate analysis of relative protein levels of many different and small proteins, using only a small volume sample, important inflammatory mediators may not be included in the panel.

The CSF levels of neuroinflammatory proteins likely correlate with the levels of the same proteins within the brain, but other factors, such as the thickness and structure of the CSF-brain barrier, probably also affect the protein concentrations. To further elucidate how neuroinflammatory proteins vary during the course of neurodegenerative diseases, postmortem studies of neuroinflammatory proteins from brains of patients at different disease stages would be of great interest.

In summary, our findings indicate that AD and FTD have fundamentally different inflammatory signatures in CSF. As the only exception, MMP-10 levels were found to be increased in CSF from patients with both disorders. However, additional studies on yet other cohorts are needed to confirm our results.