Abstract
Background:
Progranulin is implicated in frontotemporal dementia (FTD), but its role in other neurodegenerative disorders is unknown.
Objective:
To investigate the levels of progranulin (PGRN) in cerebrospinal fluid (CSF) in different neurodegenerative dementias and their correlation with levels in plasma in cognitively normal subjects.
Methods:
We measured PGRN in CSF in 229 patients with amnestic mild cognitive impairment, Alzheimer’s disease dementia, sporadic FTD, dementia with Lewy bodies, corticobasal syndrome, or progressive supranuclear palsy. We also measured PGRN in CSF and plasma in 74 cognitively normal individuals. We examined the correlation between PGRN levels in CSF and diagnosis, cortical thickness, genetic factors and other CSF biomarkers. We also investigated the correlation between plasma and CSF levels of PGRN in cognitively normal individuals.
Results:
CSF levels did not differ across diagnoses or correlate with cortical thickness. Polymorphism rs5848 in GRN influenced CSF PGRN levels, but APOE ɛ4 allele did not. Amyloid-β42, t-tau, p-tau, and YKL-40 levels correlated weakly with PGRN in CSF. We found a weak correlation (r = 0.362) between plasma and CSF PGRN levels in cognitively normal individuals.
Conclusions:
Our findings do not support a diagnostic value of CSF PGRN in neurodegenerative diseases. Our data confirm that levels of PGRN in plasma do not reflect accurately levels in CSF in cognitively normal controls. These data should be considered in clinical trials aiming to increase PGRN.
INTRODUCTION
Progranulin protein (PGRN) is encoded by the granulin gene (GRN) and expressed in many tissues and cell types, where it is involved in angiogenesis, wound repair, cell proliferation, and inflammation [1–3]. In the adult brain, PGRN is expressed in neurons and microglia and it regulates neurite outgrowth and survival [4]. Heterozygous mutations in the GRN gene cause frontotemporal lobar degeneration with TAR-DNA-binding protein 43 inclusions (FTLD-TDP) [5–8]. Pathogenic GRN mutations are typically nonsense and splice-site mutations resulting in haploinsufficiency and low plasma PGRN levels in mutation carriers [9]. Measurement of plasma and serum PGRN levels has thus proven to be a reliable biomarker to identify symptomatic and asymptomatic carriers of GRN mutations [10–12].
In addition to a link between PGRN and frontotemporal dementia (FTD), findings to date suggest a relationship between PGRN expression in the brain and other neurodegenerative diseases. Functional genetic variants that modulate GRN expression, such as the rs5848 polymorphism, have been shown to increase the risk of Alzheimer’s disease (AD) independently of the APOE ɛ4 allele [13–18]. Low brain PGRN expression has been shown to enhance amyloid-b (Aβ) and tau pathologies in transgenic mouse models of AD [19, 20]. These data suggest that PGRN may play a role in the pathophysiology of neurodegenerative dementias other than FTD.
PGRN levels can be measured in plasma and cerebrospinal fluid (CSF) but the relationship between plasma and CSF levels in subjects without GRN mutations has only been addressed in one study [21]. The authors found a weak correlation between plasma and CSF PGRN levels in 272 subjects, including 49 subjects with mild cognitive impairment (MCI) or AD dementia. The levels of PGRN in CSF in primary neurodegenerative dementias have not been systematically investigated. It is important to determine if PGRN is altered in the different neurodegenerative disorders as well as it relationship with other biomarkers such as cortical thickness or AD biomarkers in CSF. Finally, since PGRN is involved in the inflammatory response, it may influence other inflammatory proteins, such as the astroglial protein YKL-40 in CSF [22].
The objectives of this study were to evaluate how CSF PGRN levels differ across diagnosis in patients with primary neurodegenerative dementias without GRN mutations, how plasma PGRN levels correlate with CSF PGRN levels, and how these levels relate to genetic factors (rs5848 polymorphism in GRN gene and APOE ɛ4 allele), cortical thickness, core AD biomarkers and YKL-40 levels in CSF.
METHODS
Study participants
This study included 303 subjects attended at the Memory Unit at Hospital Sant Pau between January 2009 and October 2014. All subjects were evaluated by neurologists with expertise in neurodegenerative diseases, and all underwent formal cognitive evaluation using a previously published neuropsychological battery [23]. Participants had the following diagnoses: amnestic mild cognitive impairment (aMCI, n = 90), AD dementia (n = 73), FTD (n = 32), progressive supranuclear palsy (PSP, n = 3), corticobasal syndrome (CBS, n = 8), and dementia with Lewy bodies (DLB, n = 23). Patients with FTD included behavioral variant (bvFTD, n = 24), semantic variant of primary progressive aphasia (svPPA, n = 4), and non-fluent variant of primary progressive aphasia (nfvPPA, n = 4). Patients with MCI and AD dementia met the recent NIA-AA criteria [24, 25], and patients with FTD met the new international consensus criteria for bvFTD [26, 27], svPPA and nfvPPA [28], PSP [29], and CBS [30]. Mutations in the GRN gene were excluded by direct sequencing of the coding region in FTD patients with one or more affected first-degree family member as previously described [31]. Cognitively normal controls were volunteers with a normal neuropsychological evaluation for age and education, normal levels of core AD biomarkers in CSF, and no cognitive complaints. All subjects signed the informed consent form to participate in the study, and all study protocols were approved by the local ethics committee at Hospital Sant Pau.
CSF and plasma collection and analysis
CSF was obtained by lumbar puncture as described [22, 32]. Levels of core AD biomarkers (Aβ1 - 42, total tau, and phosphorylated tau) and YKL-40 in CSF were measured using commercially available kits from Fujirebio-Innogenetics (InnotestTM) and Quidel, respectively, as previously described [22, 32]. Plasma was drawn on the day of the lumbar puncture in all subjects after 6 h of fasting [31]. Our laboratory has extensive experience in determining CSF biomarkers and participates in the Alzheimer’s Association external quality control program for CSF biomarkers [33].
PGRN ELISA assay
PGRN levels were measured in CSF in all 303 subjects using the Human Progranulin ELISA Kit (Adipogen, Inc., Seoul, Korea) and in plasma in 74 cognitively normal subjects, as previously described [31]. Plasma and CSF samples were diluted prior to analysis at 1:200 and 1:5, respectively. Intra-and inter-assay coefficients of variation were 3.4% and 14.4% for CSF and 3.2% and 9.1% for plasma, respectively.
Magnetic resonance imaging (MRI) analysis
Routine imaging was performed in all subjects included in the study and 97 had consented to undergo a research MRI with a harmonized protocol for neurodegenerative diseases. Twelve of these 97 participants were excluded because of segmentation errors, and 85 were finally analyzed (Supplementary Table 1). There were not significant differences in age, sex, and PGRN levels in CSF between subjects with and without MRI in each clinical diagnostic group (data not shown). MRIs were acquired on a 3T MRI scanner (Philips 3.T x series Achieva). A high-resolution three-dimensional structural dataset was acquired with the following parameters: T1- weighted magnetization-prepared rapid gradient-echo, repetition time 8.1 ms, echo time 3.7 ms, 160 slices, matrix size 240×234; slice thickness 1 mm, voxel size 0.94×0.94×1 mm. Cortical thickness analyses were performed as previously described [34]. We studied the correlation between levels of CSF PGRN and cortical thickness. Only results that survived family-wise error correction at p < 0.05 were considered.
APOE and GRN genotyping
DNA was extracted using standard procedures and APOE was genotyped according to previously described methods [35]. Genotyping of GRN rs5848 was performed using Taqman technology with a rs5848-specific genotyping assay (Life Technologies) and carried out in a 7900-HT Fast Real-Time PCR System (Applied Biosystems).
Statistical analyses
The normality of the variables was assessed by the Kolmogorov-Smirnov test. PGRN levels in plasma did not follow a normal distribution and were transformed on the natural logarithm (LOG) scale. Nevertheless, since results and levels of significance were similar we showed untransformed values when referring to plasma PGRN. We explored the association between PGRN levels and sex using a T-test and between PGRN levels and age using a Pearson correlation coefficient (r). We studied the association between plasma and CSF PGRN levels using a bivariate linear regression model in the group of cognitively normal controls. The relationships between clinical diagnosis, APOE genotype and CSF PGRN levels were explored by ANCOVA test, covariated by variables that had a significant association with PGRN levels in the single variable analyses (sex). The association between levels of PGRN and levels of Aβ1 - 42, total-tau, phospho-tau, and YKL-40 in CSF was investigated by a Pearson correlation coefficient (r). The level of significance was set at 5% (α= 0.05) and results were corrected for multiple comparisons (Bonferroni). All statistical analyses were performed using SPSS software version 21.0 for Windows.
RESULTS
Table 1 shows the demographic, clinical, and CSF biomarker data for all diagnostic groups.
Influence of sex, age and genetic factors on PGRN levels in CSF and plasma
We first investigated the influence of age, sex, and genetic factors (APOE ɛ4 and GRN rs5848 genotypes) on PGRN levels in CSF (n = 303) and plasma (n = 74). We did not find any correlation between age and PGRN levels in CSF (p = 0.235) or plasma (p = 0.701). Males showed higher levels of PGRN in CSF than females (4.99 and 4.66 ng/ml respectively, p = 0.019). However, these differences were not significant in plasma. APOE ɛ4 did not influence PGRN levels in CSF or plasma (p = 0.585 and p = 0.438, respectively). PGRN levels in CSF were influenced by the GRN rs5848 genotype. Subjects with the TT genotype had lower CSF PGRN levels than the group with CC and CT genotypes (3.94 and 4.87 ng/ml, respectively, p < 0.001). No significant effects of the rs5848 genotype on plasma PGRN levels were found.
PGRN levels in CSF across diagnoses and their correlation with clinical and neuroimaging variables
Next, we evaluated how levels of PGRN in CSF differed across a subset of common neurodegenerative diseases. Sex-adjusted analyses did not reveal any differences between groups (Fig. 1). Levels of PGRN in CSF did not correlate with MMSE scores in the whole cohort or within clinical groups. We did not find any correlation between cortical thickness and PGRN levels in CSF in the whole group of subjects with available MRI (Supplementary Table 1), or in the exploratory analysis in each clinical group.
Relationship between PGRN and other biomarkers measured in CSF
We also investigated the relationship between the PGRN protein and other AD-related biomarkers in CSF. We found a weak correlation between PGRN and Aβ1 - 42 levels in CSF (p = 0.003, r = 0.169), t-tau (p = 0.001, r = 0.192), and p-tau (p < 0.001, r = 0.201) in the whole cohort (Fig. 2A–C). We also examined the relationship between PGRN with the astrocytic marker YKL-40 in CSF and found a positive correlation between the two (p < 0.001, r = 0.269, Fig. 2D) in the entire cohort.
Relationship between PGRN levels in plasma and CSF in healthy controls
To assess whether plasma PGRN can be used as a surrogate marker of PGRN in the CSF we compared PGRN levels in paired samples of plasma and CSF from 74 cognitively normal subjects in a simple regression model. The levels of PGRN in CSF were lower than those in plasma. We found a weak but significant correlation between plasma and CSF levels of PGRN (p = 0.002, r = 0.362, Fig. 3).
DISCUSSION
In this study we found that PGRN levels in CSF did not differ across a variety of primary neurodegenerative dementias. We also observed that CSF PGRN levels were largely independent of plasma PGRN levels and correlated weakly with core AD and YKL-40 biomarkers.
We first investigated the levels of PGRN in CSF across different neurodegenerative conditions without GRN mutations. The lack of differences in CSF PGRN levels between clinical groups indicates a low diagnostic value in neurodegenerative dementias. In agreement with these results, we did not find any association with MMSE scores or with cortical atrophy maps in the subgroup of subjects with available brain MRI. These results suggest that PGRN levels are tightly regulated in the CSF and are not disturbed by an underlying neurodegenerative process. This is in contrast with what has been observed in inflammatory disorders such as multiple sclerosis, in which elevated CSF PGRN levels have been described [36]. PGRN concentrations in CSF were also elevated in a group of patients with viral encephalitis. This suggests that an acute high-grade inflammation may be required to increase PGRN in CSF [36] in contrast with the chronic low-grade inflammation observed in neurodegenerative disorders.
Several observations to date suggest a relationship between PGRN expression in the brain and neurodegenerative diseases other than FTD. Recent studies have shown that PGRN deficiency aggravates AD pathology in animal models of AD [19, 20]. In particular, PGRN deficiency increased Aβ load and reduced phagocytosis in transgenic AβPP mouse models of AD, and accelerated tau phosphorylation and intracellular accumulation in transgenic tau mouse models [19, 20]. However, whether PGRN plays a role in the pathogenesis of patients with AD remains unclear. In this study, we found a weak correlation between PGRN levels and different AD biomarkers in CSF that may support data from transgenic animal models. However, these correlations were weak, and they did not translate into differences in PGRN levels in CSF across clinical groups with different amounts of Aβ and tau pathologies. Although the highest correlation was found between PGRN and the inflammatory marker YKL-40 in CSF, whether this finding reflects the known modulatory effects of PGRN on inflammation remains to be investigated.
The use of plasma PGRN levels to detect GRN mutation carriers is one of the best examples of a reliable plasma biomarker in neurodegenerative diseases [10–12]. Another example is the triggering receptor expressed on myeloid cells 2 (also known as TREM2), a protein involved in phagocytosis. Heterozygous mutations in TREM2 cause a rare FTD-like disorder syndrome [37] and plasma soluble TREM2 levels are markedly reduced in mutation carriers [38]. However, whether these “peripheral” markers are useful in common sporadic neurodegenerative conditions is uncertain. PGRN plasma levels in FTD patients without GRN mutations are typically unchanged [31]. In the case of soluble TREM2, patients with AD or FTD had plasma levels similar to those of healthy controls [38]. This raises the question of whether a “peripheral” marker can be used to monitor relevant pathophysiological changes in the CNS in common neurodegenerative diseases. To address this issue, an important step is to establish the correlation between plasma and CSF compartments. A recent study suggested that PGRN levels in plasma and CSF were weakly correlated [21]. In our study of paired samples, we observed a higher correlation (partial r = 0.36) between CSF and plasma PGRN levels than in the previous study (partial r = 0.17, [21]). However, this correlation was still weak and plasma PGRN levels explained only 13.1% of the variability of CSF PGRN levels. This is in agreement with other plasma markers investigated in neurodegenerative conditions, such as tau or YKL-40, where no or little correlation was found between plasma and CSF compartments [39, 40]. This weak correlation precludes the use of plasma PGRN as a surrogate measure of central PGRN. The permeability of the blood-brain barrier, aging and factors influencing clearance in these compartments may play an important role in influencing protein levels in each compartment. These findings have important implications for the use of biomarkers in clinical trials aiming to increase PGRN in patients with FTD. In a clinical trial setting, the measurement of PGRN could potentially be used as an indicator of target engagement of the drug. Although our study did not include subjects with GRN mutations, our work suggest that such a trial should include CSF measures instead of plasma for monitoring the drug treatment response. In agreement with other studies, we confirmed the influence of sex and the rs5848 in CSF PGRN levels.
The strengths of this study are the inclusion of subjects who had a detailed clinical and neuropsychological evaluation, covering several neurodegenerative dementias, and the fact that all plasma and CSF samples were drawn at the same time point. The main limitations are the lack of cases with GRN mutations, the lack of neuropathological confirmation, and the small sample size in some groups.
In summary, our results indicate that PGRN levels in CSF are tightly regulated and have no diagnostic value in primary neurodegenerative dementias not associated with GRN mutations. Plasma PGRN levels do not accurately reflect CSF PGRN levels in cognitively normal individuals. It would be important to investigate whether plasma PGRN levels reflect those in CSF in subjects with GRN mutations. These data are relevant for disease-modifying trials aimed at increasing PGRN.
