High Correlation among Brain-Derived Major Protein Levels in Cerebrospinal Fluid: Implication for Amyloid-Beta and Tau Protein Changes in Alzheimer’s Disease

The cerebrospinal fluid (CSF) plays an important role in homeostasis of the brain. We previously demonstrated that major CSF proteins such as lipocalin-type prostaglandin D2 synthase (L-PGDS) and transferrin (Tf) that are biosynthesized in the brain could be biomarkers of altered CSF production. Here we report that the levels of these brain-derived CSF proteins correlated well with each other across various neurodegenerative diseases, including Alzheimer’s disease (AD). In addition, protein levels tended to be increased in the CSF samples of AD patients compared with the other diseases. Patients at memory clinics were classified into three categories, consisting of AD (n = 61), mild cognitive impairment (MCI) (n = 42), and cognitively normal (CN) (n = 23), with MMSE scores of 20.4 ± 4.2, 26.9 ± 1.7, and 29.0 ± 1.6, respectively. In each category, CSF protein levels were highly correlated with each other. In CN subjects, increased CSF protein levels correlated well with those of AD markers, including amyloid-β and tau protein, whereas in MCI and AD subjects, correlations declined with AD markers except p-tau. Future follow-up on each clinical subject may provide a clue that the CSF proteins would be AD-related biomarkers.


Introduction
Alzheimer's disease (AD) is the most common cause of dementia [1], but the link between pathophysiological processes and the emergence of dementia has not been fully elucidated. Sperling et al. hypothesized that the AD pathology begins many years prior to the onset of dementia, and that amyloid-β (Aβ) peptide accumulation is a key event in the initial pathophysiology [2]. Aβ is produced by the sequential proteolytic cleavage of amyloid precursor protein (APP), a membrane-bound protein that is initially cleaved extracellularly by β-secretase, generating a soluble APP fragment and a stub (C99) in the membrane [3]. The C99 stub is further cleaved by β-secretase, generating extracellular Aβ and intracellular peptide fragments. In this process, several isoforms of Aβ, the most abundant of which is the 40-amino acid isoform (Aβ40), can be produced and subsequently diffuse to the CSF. Although a minor component, a second 42-amino acid isoform (Aβ42) is preferentially deposited in amyloid plaques due to its hydrophobic properties. As such, the Aβ42 isoform content in the CSF begins to decrease early in AD progression, making this trait a marker for AD diagnosis. The Aβ42/Aβ40 ratio shows better diagnostic accuracy than Aβ42 alone because the change in Aβ42 can be normalized to the major Aβ40 isoform [2]. Amyloid plaques can be visualized, even in living individuals, by amyloid positron emission tomography (PET) [4]. With this technique, the PET signal intensity and the cerebrocortical distribution of amyloid support the diagnosis of AD. In addition, the PET signal shows good concordance with CSF biomarkers such as Aβ42 and the Aβ42/Aβ40 ratio [5]. A longitudinal study revealed that some cognitively normal (CN) and mild cognitive impairment (MCI) subjects converted from PET-negative to PET-positive during follow-up [6]. Prior to follow-up, Aβ deposits may be insufficient to show PET positivity, suggesting that Aβ changes in CN and MCI subjects may take place over time prior to PET positivity. Thus, the initial pathophysiological changes in AD remain difficult to detect. In later stages of the disease, the formation of amyloid plaques is followed by the abnormal metabolism of tau. Tau is an intracellular microtubule-associated protein that is associated with axonal transport. In AD pathology, tau is hyperphosphorylated (p-tau) [2], detaches from the microtubules, and is deposited in intracellular neurofibrillary tangles. Leakage of intracellular tau and p-tau into the CSF is associated with neuronal injury and death, the progression of which is manifested in the form of dementia. As CSF tau and p-tau are also representative diagnostic markers for AD, they are collectively referred to, along with Aβ42, as "AD core biomarkers" [7].
The cerebrospinal fluid (CSF) plays a crucial role in homeostasis of nutrients and pH, etc. We previously demonstrated that major CSF proteins could serve as biomarkers of altered CSF production in spontaneous intracranial hypotension (SIH) and idiopathic normal pressure hydrocephalus (iNPH) [8][9][10][11]. Patients with SIH show cranial hypotension (<60 mm H 2 O) due to CSF leakage [10]. Myelography often reveals the location and extent of a CSF leak. Radioisotope (RI) scintigraphy reveals rapid disappearance of RI from subarachnoid space, providing quantitative data on CSF leakage. The leakage and flowing cranial hypotension were speculated to induce a compensatory increase of CSF production [10]. Concomitantly, major CSF proteins secreted from choroid plexus were increased; i.e., lipocalin-type prostaglandin D2 synthetase (L-PGDS) and brain-derived transferrin (Tf). In addition, these proteins were inversely correlated with levels of cranial hypotension and RI disappearance [11]. In contrast to SIH, these markers were decreased in the CSF of patients with iNPH, in which the production of CSF is thought to decrease [11][12][13]. INPH is treated by shunt operation, which bypasses excess CSF from ventricle to peritoneal cavity, leading to normalization of CSF production. Within 3 months after the shunt operation, the levels of brain-derived Tf increased to control levels [14], suggesting that the brainderived Tf could be a marker of altered CSF production. Tf glycan-isoform analysis also reveals that different brain-derived Tf isoforms, but not blood-derived isoforms, could be markers of CSF production. CSF Tf is composed of three glycan isoforms: Tf with sialic acid-terminated glycans from blood (Sia-Tf), Tf with N-acetylglucosamine (GlcNAc)terminated glycans, possibly from the CSF-producing choroid plexus [8,12,13], and Tf with mannose-terminated glycans from neurons (Man-Tf) [15]. Indeed, Man-Tf and GlcNAc-Tf are not detected in the CSF of congenital hydranencephaly patients, in which the cerebrum is mostly lacking [15]. Thus, brain-derived Tf levels could be markers of altered CSF production, but blood-derived Sia-Tf is unlikely to be such a marker.
In the present study we analyzed CSF levels of brain-derived Tf, L-PGDS, and AD markers in neurodegenerative diseases and examined their correlations.

Major Proteins in Serum and Cerebrospinal Fluid
Blood serum contains albumin, immunoglobulins, and Tf as major proteins ( Figure 1A). These proteins are also major components of the CSF, in addition to L-PGDS and TTR. SDS-PAGE analysis reveals that Tfs were detected around 70-73 kDa area as two bands, one of which migrates slightly slower than the other ( Figure 1A). Both bands can be seen on immunoblots to react anti-Tf antibody ( Figure 1B). We previously reported that the slower-migrating band includes Tf carrying sialic acid terminated-glycan (Sia-Tf) [13]. The faster-migrating band includes Tf carrying N-acetylglucosamine terminated-glycan (GlcNAc-Tf) and mannose terminated-glycan (Man-Tf) [15]. Broad bands around the 23-24 kDa area and a sharp band around 14 kDa react with anti-L-PGDS and anti-TTR antibodies, respectively ( Figure 1A,B). The migration positions of recombinant TTR and L-PGDS were shifted due to the inclusion of a His-tag motif. Purified standards of GlcNAc-Tf and Man-Tf were obtained from CSF as previously described, reprinted with permission from ref. [15], 2021 [15], Yasuhiro Hashimoto. The presence of TTR and L-PGDS, in addition to Man-Tf and GlcNAc-Tf, in the CSF was confirmed by immunoblotting using specific antibodies (B). Recombinant His-tagged TTR and L-PGDS show a mobility shift.

Concentration of Major Proteins in the CSF
We previously demonstrated that CSF Man-Tf levels are increased in MCI and AD [15]. Therefore, Man-Tf, together with L-PGDS, TTR, and Tf isoforms in the CSF, was quantified for CN, MCI, and AD subjects ( Figure 2) whose MMSE scores (mean ± S.D.) were 29.0 ± 1.6, 26.9 ± 1.7, and 20.4 ± 4.2, respectively. CN subjects recruited in the present study did not show evidence of dementia at baseline. However, this did not necessarily mean that CN subjects were free from AD pathology; only 28% of CN subjects had CSF tau and p-tau values that were lower than cutoff values. Indeed, five out of 23 CN subjects converted to MCI during the follow-up period ( Figure 2, red filled circle). Disease controls included patients with neurodegenerative diseases such as progressive supranuclear palsy (PSP), frontotemporal degeneration (FTD), Parkinson's disease (PD), and dementia with Lewy bodies (DLB). Five CN subjects who converted to MCI are indicated with red filled circles. Sia-Tf data from three conversion cases were not determined. Man-Tf data of disease control, MCI, AD is cited from our previous paper, reprinted with permission from ref. [15], 2021, Yasuhiro Hashimoto. Multiple comparisons were assessed by the Kruskal-Wallis method followed by Bonferroni correction. All combinations of clinical groups were assessed, and combinations showing significant difference were indicated with horizontal bars except that Man-Tf data between DC and AD (p = 0.07).
Compared to disease controls, the Man-Tf levels were significantly higher in CN and MCI, and showed a tendency to be higher (p = 0.07) in AD ( Figure 2A). Likewise, GlcNAc-Tf levels were significantly higher in all groups, indicating that Man-Tf and GlcNAc-Tf showed similarly incrementing profiles ( Figure 2B). L-PGDS levels were higher only in AD subjects (p < 0.001) ( Figure 2C), while TTR levels were slightly higher in MCI and lower in CN and AD compared to control, but differences were not significant ( Figure 2D). Neither Sia-Tf nor total protein levels showed significant differences for any groups ( Figure 2E,F). Taken together, with the exception of L-PGDS in MCI, CSF levels of brain-derived L-PGDS, Man-Tf, and GlcNAc-Tf showed a tendency to increase in the CSF of patients from all groups.

Correlations between Major Proteins in the CSF
We previously demonstrated that CSF levels of L-PGDS and brain-derived Tf were concomitantly altered in SIH and iNPH [8]. Therefore, we analyzed Man-Tf, GlcNAc-Tf, and L-PGDS levels in the CSF samples of all subjects included in the present study; i.e., AD, MCI, CN, PSP, FTD, PD, and DLB. We found that Man-Tf levels correlated well with GlcNAc-Tf levels (r = 0.80) and moderately with L-PGDS levels (r = 0.56) ( Figure 3A,B), while GlcNAc-Tf levels correlated well with L-PGDS (r = 0.70) ( Figure 3C). In contrast, Man-Tf levels did not correlate with Sia-Tf (r = 0.02), TTR (r = 0.03) or total protein levels (r = 0.09) ( Figure 3D-F). This result suggests that the levels of brain-derived proteins correlate well with each other regardless of the disease involved.

Correlations between Major CSF Proteins in CSF of CN, MCI, and AD
Correlations of the CSF proteins were also analyzed for the CN, MCI, and AD clinical groups. In CN, Man-Tf levels correlated well with GlcNAc-Tf (rs = 0.85) and L-PGDS (rs = 0.61) levels (Table 1), while GlcNAc-Tf also correlated well with L-PGDS (rs = 0.72). In contrast, Man-Tf levels did not correlate with Sia-Tf or TTR levels. High correlations were thus found between Man-Tf, GlcNAc-Tf, and L-PGDS in the CN group. In MCI, Man-Tf levels correlated well with GlcNAc-Tf (rs = 0.71) and L-PGDS (rs = 0.64) levels, while GlcNAc-Tf correlated well with L-PGDS (rs = 0.68). Man-Tf levels were negatively, but weakly, correlated with TTR (rs = −0.36) levels, but not with Sia-Tf. Again, high correlations were observed among brain-derived proteins in the MCI group. In AD, Man-Tf levels correlated well with GlcNAc-Tf (rs = 0.78) and L-PGDS (rs = 0.64) levels, but not with Sia-Tf or TTR. On the other hand, GlcNAc-Tf levels correlated well with L-PGDS (rs = 0.69) levels. These results suggest that the brain-derived proteins showed high correlations with each other across all clinical groups. Spearman's rank correlation coefficients (rs) with values higher than 0.50 are highlighted with bold letters in yellow cells. Man-Tf data of MCI and AD are cited from our previous paper [15]. Probability levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Discussion
It has been established that L-PGDS is secreted from the choroid plexus and pia mater [8]. Man-Tf and GlcNAc-Tf are also biosynthesized in the brain, because these Tf isoforms are undetectable in CSF of hydranencephaly patients with congenital lack of cerebrum [15]. In the previous study, we reported that the CSF proteins derived from the brain could be a biomarker of altered CSF production. In the present study, we analyzed these brain-derived protein levels in CSF samples from CN, MCI, and AD patients as well as patients with other neurodegenerative diseases, which include other tauopathy (PSP and FTD) and synucleinopathy (DLB and PD). The levels of brain-derived proteins were well correlated with each other irrespective of the disease in question, suggesting that good correlation is a common phenotype at least in these diseases, but not specific to AD. In contrast, elevations of CSF proteins are demonstrated only in AD-related groups, suggesting that the elevations are related to AD pathophysiology, although the molecular mechanism has yet to be clarified. In addition, the elevations are correlated with AD markers. In the CN group, the brain-derived CSF protein levels correlated well with those of p-tau, tau, and Aβ40, and correlated moderately with Aβ42. Similar correlation patterns were observed for the MCI group, but their correlation coefficients tended to be lower. In the AD group, p-tau correlated with brain-derived CSF protein levels, whereas other AD markers did not, suggesting that the correlation mechanism is altered in AD progression.
Besides serving as possible markers of CSF production, the brain-derived CSF proteins have Aβ binding activity, which may potentially prevent AD progression by affecting Aβ metabolism. Raditsis et al. reported that Tf inhibited Aβ aggregation by binding to Aβ oligomers as well as monomers [16]. Another major protein, L-PGDS, was demonstrated to bind Aβ monomers and fibrils to thereby inhibit not only self-aggregation but also seed-dependent aggregation [17]. In addition, L-PGDS-deficient mice showed a 3.5-fold higher Aβ deposition than control mice after the intraventricular infusion of Aβ42 [18].
These results suggest that CSF proteins prevent the aggregation and deposition of Aβ in the brain. Further studies on potentially neuroprotective CSF proteins could lead to the development of new therapeutic strategies.
Quantifying AD markers in plasma will be promising laboratory tests for AD diagnosis [19,20]. Moscoso et al. reported that plasma p-tau181 is a marker for monitoring neurodegeneration and cognitive decline in AD [21]. Tatebe et al. reported, however, that the plasma p-tau concentration is only 0.1% of that of the CSF [19], requiring sensitive methods. A potential advantage of the brain-derived proteins is their high concentration in CSF; e.g., L-PGDS concentrations of CSF and serum are 5-15 µg/mL and 50-200 ng/mL, respectively, [22] which are measurable by a conventional ELISA kit. Indeed, Tsutsumi et al. reported that L-PGDS concentration in a skull vein is significantly higher than that of peripheral veins, suggesting that CSF L-PGDS could diffuse into a vein and provide information on L-PGDS in CSF [22]. This may be also the case with other CSF proteins, because Man-Tf and GlcNAc-Tf show high concentrations in CSF: 2-8 µg/mL for Man-Tf and 3-11 µg/mL for GlcNAc-Tf [15]. Quantifying the CSF proteins in plasma would give a clue to monitor CSF metabolism in neurological diseases.

Patients
Patients with AD, MCI, or other neurodegenerative disorders were consecutively recruited from Fukushima Medical University, Tohoku University, and Juntendo University. MMSE score criteria for AD, MCI, and CN were <25, 25~27 and >28, respectively. For each disease, the number of patients, age (mean ± S.D.), male/female distribution, and MMSE are listed in Table 3. Disease diagnosis was based on the following criteria: AD, "The diagnosis of dementia due to Alzheimer's disease: Recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease." [1]; MCI, "The diagnosis of mild cognitive impairment due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease" [23]; progressive supranuclear palsy (PSP), "Accuracy of clinical criteria for the diagnosis of progressive supranuclear palsy" [24]; frontotemporal degeneration (FTD), "Frontotemporal lobar degeneration: a consensus on clinical diagnostic criteria" [25]; dementia with Lewy bodies (DLB) "Diagnosis and management of dementia with Lewy bodies: Fourth consensus report of the DLB Consortium." [26]; Parkinson's disease (PD), "the Queen Square Brain Bank criteria for the diagnosis of Parkinson's disease" (https://www.mims.ie/news/queen-squarebrain-bank-qsbb-criteria-for-pd-diagnosis-02-04-2013/, accessed on 14 March 2022).

Statistical Analyses
Statistical analyses were performed using SPSS software (version 26). Data normality was examined by the Shapiro-Wilk test. Significant differences among multiple comparisons were assessed by the Kruskal-Wallis method followed by Bonferroni correction.