Brain-Derived Major Glycoproteins Are Possible Biomarkers for Altered Metabolism of Cerebrospinal Fluid in Neurological Diseases

Cerebrospinal fluid (CSF) plays an important role in the homeostasis of the brain. We previously reported that CSF major glycoproteins are biosynthesized in the brain, i.e., lipocalin-type prostaglandin D2 synthase (L-PGDS) and transferrin isoforms carrying unique glycans. Although these glycoproteins are secreted from distinct cell types, their CSF levels have been found to be highly correlated with each other in cases of neurodegenerative disorders. The aim of this study was to examine these marker levels and their correlations in other neurological diseases, such as depression and schizophrenia, and disorders featuring abnormal CSF metabolism, including spontaneous intracranial hypotension (SIH) and idiopathic normal pressure hydrocephalus (iNPH). Brain-derived marker levels were found to be highly correlated with each other in the CSF of depression and schizophrenia patients. SIH is caused by CSF leakage, which is suspected to induce hypovolemia and a compensatory increase in CSF production. In SIH, the brain-derived markers were 2–3-fold higher than in other diseases, and, regardless of their diverse levels, they were found to be correlated with each other. Another abnormality of the CSF metabolism, iNPH, is possibly caused by the reduced absorption of CSF, which secondarily induces CSF accumulation in the ventricle; the excess CSF compresses the brain’s parenchyma to induce dementia. One potential treatment is a “shunt operation” to bypass excess CSF from the ventricles to the peritoneal cavity, leading to the attenuation of dementia. After the shunt operation, marker levels began to increase within a week and then further increased by 2–2.5-fold at three, six, and twelve months post-operation, at which point symptoms had gradually attenuated. Notably, the marker levels were found to be correlated with each other in the post-operative period. In conclusion, the brain-derived major glycoprotein markers were highly correlated in the CSF of patients with different neurological diseases, and their correlations were maintained even after surgical intervention. These results suggest that brain-derived proteins could be biomarkers of CSF production.


Introduction
Cerebrospinal fluid (CSF) plays a pivotal role in the homeostasis of the central nervous system by distributing growth factors, removing metabolic waste, controlling pH, and maintaining intracranial pressure (ICP) [1]. Cushing hypothesized that the CSF is produced mainly by the choroid plexus of the lateral ventricles of the brain, and is absorbed in the arachnoid granulations (bulk flow theory) [2]. In this way, CSF in the lateral ventricles passes through the interventricular foramina to the third ventricle, then from the cerebral aqueduct to the fourth ventricle. The fluid then passes from the fourth ventricle into the subarachnoid space through the foramina of Magendie and Luschka. CSF in the subarachnoid space bathes the surface of the brain and is subsequently absorbed by arachnoid granulations near the parietal bone. The bulk flow theory is accepted particularly by neurosurgeons as it adequately explains the mechanism underlying obstructive hydrocephalus; i.e., aqueduct stenosis blocks CSF flow, causing upstream compartments to become enlarged due to CSF accumulation. In contrast, Jessen et al., hypothesized that the "glymphatic system" plays a major role in generating the CSF flow; i.e., body fluid in para-arterial spaces enters the brain parenchyma, it collects waste, and returns to lymphatic systems via para-venous spaces [3,4]. The mechanism of CSF production and absorption thus remains contentious.
The composition of major proteins in the CSF is similar to that of blood [5,6]. For instance, CSF and serum contain albumin as a major component, constituting 50-60% of their total protein. Albumin is biosynthesized only in the liver, indicating that it is secreted from the liver into blood and then translocated to the CSF. Notably, the protein concentration in plasma/serum is approximately 60-80 mg/mL, which is 50-100 times higher than in the CSF [1,7], suggesting that a small amount of serum protein diffuses into the CSF possibly down its concentration gradient. In this way, the "blood-brain barrier" is leaky even under physiological conditions. Iron plays many important roles in the brain, and has involvement in myelination, neurotransmission and electron transfer in the respiratory chain [8]. However, as free iron results in cellular toxicity by generating harmful reactive oxygen species (ROS) [9], a physiological response has evolved whereby iron binds to specific carrier proteins, such as transferrin (Tf), in the extracellular fluid and blood [10]. Like albumin, Tf that is glycosylated with sialylated glycans (Sia-Tf) is biosynthesized in the liver, secreted into the blood, and then translocated to the CSF. We also found that isoforms of Tf that carry unique glycans were produced in the brain [11]: these include Tf that carry N-acetylglucosamineterminated glycans (GlcNAc-Tf) and Tf that carry mannose-terminated glycans (Man-Tf). These Tf isoforms were undetectable in the CSF of hydranencephaly patients, who congenitally lack most of the cerebrum [12], indicating that they are produced by the cerebrum. To identify Tf-producing cells in the cerebrum, we analyzed the expression of Tf mRNA by in situ hybridization [11]. The most intense signals were detected in the choroid plexus epithelium stained with anti-Tf antibody, as well as a GlcNAc-binding probe, recombinant Psathyrella velutina lectin (rPVL) [13]. Co-localized signals were found in the epithelium, suggesting that GlcNAc-Tf is produced in the choroid plexus epithelium. In situ hybridization also revealed that Tf mRNA was highly expressed in most neurons in the cerebral cortex, hippocampus, and basal ganglia. To identify glycan isoforms expressed in the cerebral cortex, Tf was purified from the cortex and its glycan content was analyzed by mass spectrometry [11]. Cortical Tf mainly carried mannose-terminated glycans (~90%), suggesting that Man-Tf is secreted by neurons. We speculated, therefore, that Man-Tf could be a biomarker for neurons and that its expression might be altered in neurodegenerative disorders. Man-Tf levels were quantified in the CSF of patients with Alzheimer's disease analyzed statistically, avoiding the potential confounding effects of aging. In depression and schizophrenia patients, the levels of all markers were slightly higher than those of non-SIH patients, but the differences were not significant. SIH patients showed very high and variable levels of all markers compared to non-SIH patients. Man-Tf, GlcNAc-Tf, L-PGDS, Sia-Tf, and total protein levels were increased by 2.9-, 2.2-, 2.0-, 1.7-, and 1.6fold, respectively, indicating that both brain-and blood-derived proteins were markedly increased in SIH. In iNPH, GlcNAc-Tf and L-PGDS levels were significantly higher than the levels in non-iNPH patients.
was mainly composed of blood-derived proteins (~80%) such as albumin, IgG, and serum Tf [1,6]. Serum Tf carries sialic acid-terminated glycans (Sia-Tf). Tf isoforms were quantified by a lectin/antibody-sandwich ELISA; the glycan-specific lectins used were Sambucus sieboldiana agglutinin (SSA) for Sia-Tf, recombinant Burkholderia cenocepaciar (rBC2L-A) for Man-Tf, and rPVL for GlcNAc-Tf. Patients were classified into two groups based on the peak ages of onset: one group included depression, schizophrenia, SIH, and non-SIH patients with peak ages in the thirties and forties; another group included iNPH and non-iNPH patients with peak ages in the seventies. The two groups were analyzed statistically, avoiding the potential confounding effects of aging. In depression and schizophrenia patients, the levels of all markers were slightly higher than those of non-SIH patients, but the differences were not significant. SIH patients showed very high and variable levels of all markers compared to non-SIH patients. Man-Tf, GlcNAc-Tf, L-PGDS, Sia-Tf, and total protein levels were increased by 2.9-, 2.2-, 2.0-, 1.7-, and 1.6-fold, respectively, indicating that both brain-and blood-derived proteins were markedly increased in SIH. In iNPH, GlcNAc-Tf and L-PGDS levels were significantly higher than the levels in non-iNPH patients. Figure 1. CSF marker levels in neurological diseases. CSF marker levels were analyzed in depression, schizophrenia, SIH, non-SIH, iNPH, and non-iNPH patients. Tf isoforms such as Man-Tf, Glc-NAc-Tf, and Sia-Tf were quantified by a lectin/antibody-sandwich ELISA; the glycan-specific probes used were rBC2L-A, rPVL, SSA lectin, respectively. L-PGDS levels were quantified with a sandwich ELISA kit. Total protein was quantified by the micro-BCA method. Ages of patients in each disease are listed in boxes (mean ± S.D.). Data normality for each marker is indicated with a plus (+) or minus (−) sign. Multiple comparisons among depression, schizophrenia, SIH, and non-SIH patients were assessed by the Kruskal-Wallis method, followed by Bonferroni correction. Comparisons of iNPH and non-iNPH were assessed by a Mann-Whitney U test. All combinations of clinical groups were assessed, and combinations showing significant differences are indicated with horizontal bars. Probability values: * p < 0.05, ** p < 0.01, *** p < 0.001. n.s.: not significant. Figure 1. CSF marker levels in neurological diseases. CSF marker levels were analyzed in depression, schizophrenia, SIH, non-SIH, iNPH, and non-iNPH patients. Tf isoforms such as Man-Tf, GlcNAc-Tf, and Sia-Tf were quantified by a lectin/antibody-sandwich ELISA; the glycan-specific probes used were rBC2L-A, rPVL, SSA lectin, respectively. L-PGDS levels were quantified with a sandwich ELISA kit. Total protein was quantified by the micro-BCA method. Ages of patients in each disease are listed in boxes (mean ± S.D.). Data normality for each marker is indicated with a plus (+) or minus (−) sign. Multiple comparisons among depression, schizophrenia, SIH, and non-SIH patients were assessed by the Kruskal-Wallis method, followed by Bonferroni correction. Comparisons of iNPH and non-iNPH were assessed by a Mann-Whitney U test. All combinations of clinical groups were assessed, and combinations showing significant differences are indicated with horizontal bars. Probability values: * p < 0.05, ** p < 0.01, *** p < 0.001. n.s.: not significant.

Correlation of CSF Markers in Patients with Depression
We analyzed correlations between CSF markers in CSF samples from patients with depression. Scatter diagrams and Spearman's rank correlation coefficients are indicated for each marker combination ( Figure 2, Table 1). Brain-derived markers were highly correlated with each other: Man-Tf vs. GlcNAc-Tf (rs = 0.87), Man-Tf vs. L-PGDS (rs = 0.77), and GlcNAc-Tf vs. L-PGDS (rs = 0.70). In contrast, Man-Tf was not correlated with Sia-Tf (rs = 0.20) or total protein (rs = 0.16). Nevertheless, Sia-Tf and total protein were well correlated with each other (rs = 0.65).
We analyzed correlations between CSF markers in CSF samples from patients with depression. Scatter diagrams and Spearman's rank correlation coefficients are indicated for each marker combination ( Figure 2, Table 1). Brain-derived markers were highly correlated with each other: Man-Tf vs. GlcNAc-Tf (rs = 0.87), Man-Tf vs. L-PGDS (rs = 0.77), and GlcNAc-Tf vs. L-PGDS (rs = 0.70). In contrast, Man-Tf was not correlated with Sia-Tf (rs = 0.20) or total protein (rs = 0.16). Nevertheless, Sia-Tf and total protein were well correlated with each other (rs = 0.65).

Correlations between CSF Markers in Schizophrenia Patients
Correlations between CSF markers were analyzed in schizophrenia patients. Scatter diagrams and Spearman's rank correlation coefficients are indicated in Figure 3 and Table  2 for each marker combination. The following brain-derived markers showed correlations with each other: Man-Tf vs. GlcNAc-Tf (rs = 0.76), Man-Tf vs. L-PGDS (rs = 0.68), and GlcNAc-Tf vs. L-PGDS (rs = 0.55). In contrast, Man-Tf was not correlated with Sia-Tf (rs = 0.13) or total protein (rs = 0.18).

Correlations between CSF Markers in Schizophrenia Patients
Correlations between CSF markers were analyzed in schizophrenia patients. Scatter diagrams and Spearman's rank correlation coefficients are indicated in Figure 3 and Table 2 for each marker combination. The following brain-derived markers showed correlations with each other: Man-Tf vs. GlcNAc-Tf (rs = 0.76), Man-Tf vs. L-PGDS (rs = 0.68), and GlcNAc-Tf vs. L-PGDS (rs = 0.55). In contrast, Man-Tf was not correlated with Sia-Tf (rs = 0.13) or total protein (rs = 0.18).   Spearman's rank correlation coefficients (correl. coeff.) among Man-Tf, GlcNAc-Tf, and L-PGDS are highlighted in yellow, while that of Sia-Tf vs. total protein is highlighted in pink. Data normality (Nor.) is shown by a plus (+) or minus (−) sign. Probability level: ** p < 0.01.

Correlation of CSF Markers in CSF of SIH Patients
SIH is caused by CSF leakage, which induces intracranial hypotension and the dilatation of veins. An increase in CSF production may occur to compensate for this leakage. All marker levels were elevated and varied widely in CSF samples from SIH patients (Figure 4). In spite of the extreme variability of values, brain-derived marker levels were significantly correlated with each other (Table 3)

CSF Markers in SIH 2.3.1. Correlation of CSF Markers in CSF of SIH Patients
SIH is caused by CSF leakage, which induces intracranial hypotension and the dilatation of veins. An increase in CSF production may occur to compensate for this leakage. All marker levels were elevated and varied widely in CSF samples from SIH patients ( Figure 4). In spite of the extreme variability of values, brain-derived marker levels were significantly correlated with each other (Table 3)    Spearman's rank correlation coefficients (correl. coeff.) among Man-Tf, GlcNAc-Tf, and L-PGDS are highlighted in yellow, while that of Sia-Tf vs. total protein is highlighted in pink. Correlation coefficients between blood-derived markers and GlcNAc-Tf or L-PGDS are highlighted in blue. Data normality (Nor.) is shown by a plus (+) or minus (−) sign. Probability level: ** p < 0.01.

Correlation of Markers in CSF from Non-SIH Patients
CSF marker correlations were analyzed in CSF samples from non-SIH patients. Scatter diagrams and Spearman's rank correlation coefficients are indicated for marker combinations ( Figure 5, Table 4). Except for GlcNAc-Tf vs. L-PGDS, the brain-derived markers correlated significantly with each other: Man-Tf vs. GlcNAc-Tf (rs = 0.79), Man-Tf vs. L-PGDS (rs = 0.55). In addition, Sia-Tf and total protein were moderately correlated with the following markers: Man-Tf (rs = 0.51~0.44), GlcNAc-Tf (rs = 0.46~0.50), and L-PGDS (rs = 0.51~0.67). Sia-Tf and total protein were also well correlated (rs = 0.73).  iNPH is caused by a reduced absorption of CSF, which induces the accumulation of CSF in the ventricles. MRI imaging reveals enlarged ventricles with excess CSF, which compress the brain parenchyma, leading to dementia, gait disturbance, and urinary in-  Spearman's rank correlation coefficients (correl. coeff.) among Man-Tf, GlcNAc-Tf, and L-PGDS are highlighted in yellow, while that of Sia-Tf vs. total protein is highlighted in pink. Correlation coefficients between brain-derived and blood-derived markers are highlighted in blue. Data normality (Nor.) is shown by a plus (+) or minus (−) sign. Probability level: * p < 0.05, ** p < 0.01.

CSF Markers in iNPH 2.4.1. Correlation of Markers in CSF of iNPH Patients
iNPH is caused by a reduced absorption of CSF, which induces the accumulation of CSF in the ventricles. MRI imaging reveals enlarged ventricles with excess CSF, which compress the brain parenchyma, leading to dementia, gait disturbance, and urinary incontinence. Marker correlations were analyzed in CSF of iNPH patients, with scatter diagrams and Spearman's rank correlation coefficients shown below for each marker combination ( Figure 6, Table 5). Brain-derived markers were highly correlated with each other: Man-Tf vs. GlcNAc-Tf (rs = 0.82), Man-Tf vs. L-PGDS (rs = 0.75), and GlcNAc-Tf vs. L-PGDS (rs = 0.73). In contrast, Man-Tf showed a low correlation with Sia-Tf (rs = 0.37) and total protein (rs = 0.29). Nevertheless, Sia-Tf and total protein were well correlated with each other (rs = 0.70).  Spearman's rank correlation coefficients (correl. coeff.) among Man-Tf, GlcNAc-Tf, and L-PGDS a highlighted in yellow, while that of Sia-Tf vs. total protein is highlighted in pink. Correlation coe ficients between brain-derived and blood-derived markers are highlighted in blue. Data normali (Nor.) is shown by a plus (+) or minus (−) sign. Probability level: * p < 0.05, ** p < 0.01.

Correlation of CSF Markers in Non-iNPH Patients
Marker correlations were analyzed in CSF samples from non-iNPH patients (Figu 7, Table 6   Spearman's rank correlation coefficients (correl. coeff.) among Man-Tf, GlcNAc-Tf, and L-PGDS are highlighted in yellow, while that of Sia-Tf vs. total protein is highlighted in pink. Correlation coefficients between brain-derived and blood-derived markers are highlighted in blue. Data normality (Nor.) is shown by a plus (+) or minus (−) sign. Probability level: * p < 0.05, ** p < 0.01.

Correlation of CSF Markers in Non-iNPH Patients
Marker correlations were analyzed in CSF samples from non-iNPH patients (Figure 7, Table 6  Spearman's rank correlation coefficients (correl. coeff.) among Man-Tf, GlcNAc-Tf, and L-PGDS ar highlighted in yellow, while that of Sia-Tf vs. total protein is highlighted in pink. Correlation coef ficients between GlcNAc-Tf and blood-derived markers, and between L-PGDS and total protein ar highlighted in blue. Data normality (Nor.) is indicated by a plus (+) or minus (−) sign. Probabilit level: * p < 0.05, ** p < 0.01.

Diagnostic Accuracy of CSF Markers for SIH and iNPH
We next examined the diagnostic accuracy of CSF markers for SIH and iNPH, which showed significant differences in marker levels. The criteria for good markers are a sensi tivity and specificity of more than 85%, together with an area under the ROC curve (AUC greater than 0.9. SIH is mainly diagnosed on the basis of MRI imaging and the measure ment of intracranial hypotension. As biochemical markers for SIH are barely established CSF markers were examined for their diagnostic accuracy. Sensitivity values for Man-Tf GlcNAc-Tf, L-PGDS, Sia-Tf, and the total protein levels were 97%, 77%, 77%, 46%, and 62%, respectively, while the specificity values were 95%, 75%, 75%, 95%, and 85%, respec tively (Table 7). Man-Tf showed a 0.98 AUC, indicating that it would likely be a much better diagnostic marker than GlcNAc-Tf or L-PGDS.  Spearman's rank correlation coefficients (correl. coeff.) among Man-Tf, GlcNAc-Tf, and L-PGDS are highlighted in yellow, while that of Sia-Tf vs. total protein is highlighted in pink. Correlation coefficients between GlcNAc-Tf and blood-derived markers, and between L-PGDS and total protein are highlighted in blue. Data normality (Nor.) is indicated by a plus (+) or minus (−) sign. Probability level: * p < 0.05, ** p < 0.01.

Diagnostic Accuracy of CSF Markers for SIH and iNPH
We next examined the diagnostic accuracy of CSF markers for SIH and iNPH, which showed significant differences in marker levels. The criteria for good markers are a sensitivity and specificity of more than 85%, together with an area under the ROC curve (AUC) greater than 0.9. SIH is mainly diagnosed on the basis of MRI imaging and the measurement of intracranial hypotension. As biochemical markers for SIH are barely established, CSF markers were examined for their diagnostic accuracy. Sensitivity values for Man-Tf, GlcNAc-Tf, L-PGDS, Sia-Tf, and the total protein levels were 97%, 77%, 77%, 46%, and 62%, respectively, while the specificity values were 95%, 75%, 75%, 95%, and 85%, respectively (Table 7). Man-Tf showed a 0.98 AUC, indicating that it would likely be a much better diagnostic marker than GlcNAc-Tf or L-PGDS.
Levels of GlcNAc-Tf and L-PGDS in CSF samples from iNPH patients were significantly lower than those of non-iNPH patients. CSF markers were examined for their diagnostic accuracy to differentiate iNPH from non-iNPH. The sensitivity values for Man-Tf, GlcNAc-Tf, L-PGDS, Sia-Tf, and the total protein levels were 62%,72%, 82%, 81%, and 46%, respectively, whereas their respective specificity values were 38%, 28%, 57%, 19%, and 54% (Table 8). All markers showed low specificities, indicating that they are not applicable for the diagnosis of iNPH.  Patients with iNPH usually undergo a "shunt operation", in which excess CSF is bypassed from the ventricles to the peritoneal cavity. Following the operation, prior symptoms such as dementia and gait disturbance usually attenuate gradually (shunt responder), suggesting that the shunt system functions well and that CSF metabolism has normalized. Responders were examined for their marker levels with the expectation that the operation would change pre-operative marker levels. CSF specimens were withdrawn from the shunt valve over time post-operation, and marker levels were quantified and compared with pre-operative levels. Brain-derived markers slightly increased on postoperative days 1-3 and further increased on days 4-6 ( Figure 8). Man-Tf, GlcNAc-Tf, and L-PGDS eventually showed 1.6-, 1.6-and 1.8-fold increases, respectively (Table 9). Blood-derived markers showed only subtle changes during the same period. Table 9. CSF marker levels at post-operative days 1-3 and 4-6.

Correlation Coefficients of CSF Marker Levels at Postoperative Days 1-6
The correlation coefficients for each marker combination were examined at post-operative days 1-6 ( Table 10)

Marker Levels at Post-Operative Months 3, 6 and 12
CSF specimens were withdrawn at post-operative months 3, 6 and 12. The time course of the marker level changes was examined by comparing pre-and post-operative levels. The brain-derived marker levels increased significantly in the post-operative period, while the Sia-Tf and total protein levels increased only marginally (Figure 9, Table 11). Spearman's rank correlation coefficients (rs) for markers at 3 and 12 months and Pearson's correlation coefficients (r) at 6 months are indicated. Coefficients among Man-Tf, GlcNAc-Tf, and L-PGDS are highlighted in yellow, those of GlcNAc-Tf vs. blood-derived proteins are highlighted in blue, and those of Sia-Tf and total proteins (p < 0.05) are highlighted in pink. Data normality (Nor.) is indicated by a plus (+) or minus (−) sign. Probability level: * p < 0.05, ** p < 0.01.

Discussion
We previously reported that the CSF levels of brain-derived glycoproteins, such as Man-Tf, GlcNAc-Tf, and L-PGDS, are well correlated with each other in neurodegenerative disorders, including AD [6]. The present results revealed that, except for L-PGDS in non-SIH, these CSF markers were significantly correlated in other neurological diseases such as depression, schizophrenia, SIH and iNPH. Although the latter two diseases showed different levels of markers from other diseases, the levels were still correlated.
Concerning the mechanisms that underlie the elevation of brain-derived proteins in SIH, it is known that the disease is caused by CSF leakage, which induces intracranial hypotension and subsequent brain dislocation, leading to orthostatic headache and dizziness, among other symptoms [16,17,19]. SIH is usually treated by the "blood patch" method, in which a patient's own blood is injected into leakage sites, generating blood clots to stop the CSF leak. This therapy relieves the patient's symptoms, suggesting that the leakage and loss of CSF are the primary causes of the disease for which a compensatory increase in CSF production is presumed. The results of the present study revealed that brain-and blood-derived protein levels were markedly elevated in SIH. Among the markers, L-PGDS is biosynthesized in the choroid plexus and then secreted into the CSF [25]. It is assumed that L-PGDS secretion in SIH increases concomitantly with enhanced CSF production. This might also be the case for GlcNAc-Tf, which is secreted from the choroid plexus [13,26]. This hypothesis may, at least partly, account for the high correlation of these markers in SIH. In addition, the two markers were correlated with Man-Tf, a neuron-derived protein [11]. One possible explanation is that Tf gene expression is controlled by common transcriptional regulation in neurons and choroid plexus epithelium. After the translation, Tf proteins are post-translationally modified with different glycans: mannose-terminated glycans in neurons and GlcNAc-terminated glycans in choroid plexus epithelium. Taken together, we consider that Tf isoforms and PGDF are markers of the enhancement of CSF production in SIH.
Next, in relation to the mechanisms associated with the elevation of blood-derived proteins in SIH, our previous study demonstrated that total protein, albumin, immunoglobulin, and Sia-Tf markedly increased in CSF, but not in serum [7]. The pattern of protein expression in the CSF was similar to that of serum, except for the presence of brain-derived proteins, suggesting that each blood protein was translocated from the blood into CSF with similar efficacies. The present results showed that the total protein and Sia-Tf levels significantly increased, and were correlated with each other in the CSF. It is also noteworthy that the levels of these markers were also highly correlated in other neurological diseases and controls (rs = 0.70~0.83). These results suggest that, regardless of marker levels, small amounts of blood proteins are translocated from the blood into the CSF with similar efficiency. It is also reported that, unlike blood capillaries that form the blood-brain barrier, choroid plexus capillaries are fenestrated and have no tight junctions, suggesting that blood proteins could diffuse into the CSF in a concentration dependent-manner at this site [27]. The level of diffusion may, at least partly, depend on the intracranial pressure (ICP) of the affected individual. As included in the diagnostic criteria, the ICP in SIH (0~60 mm H 2 O) is much lower than that of the controls (70-180 mm H 2 O) [17]. Consequently, a decreased ICP induces the dilatation of blood vessels and a concomitant increase in the intracranial blood volume, which might facilitate the diffusion of blood proteins into the CSF. We therefore assume that the elevation of blood-derived marker levels in the CSF of SIH patients is mainly due to the presence of intracranial hypotension.
Concerning the potential mechanisms that underlie marker changes in iNPH, it was previously demonstrated that affected patients show an accumulation of CSF in the ventricles, which is possibly due to a reduced absorption of CSF [28]. A secondary decline in CSF production is suspected. L-PGDS and GlcNAc-Tf levels in the CSF samples of iNPH patients were significantly lower than those of non-iNPH samples, but levels of other markers were not. In addition, the brain-derived marker levels were highly correlated with each other. L-PGDS and Man-Tf did not correlate with the blood-derived marker levels, whereas GlcNAc-Tf levels did correlate moderately. A shunt operation is the standard approach usually used to treat iNPH. In the present study, CSF marker levels in responders to the shunt operation were examined before and after patients underwent the procedure. In the shunt responders, the brain-derived marker levels in these patients tended to increase, even in post-operative days 1-6 (50~65%), compared to pre-operative levels, while the bloodderived markers scarcely increased (0~11%). The brain-derived marker levels were highly correlated with each other, but the blood-derived markers were not, suggesting that the former levels were controlled by a mechanism distinct from the latter, at least during this period. We also analyzed marker levels at post-operative months 3-12. The brain-derived marker levels were significantly higher than the pre-operative levels (49~113%), whereas the blood-derived markers increased only marginally (12~44%). The brain-derived markers were highly correlated with each other at all time points, with correlation coefficients at post-operative months 3, 6, and 12 found to be 0.59-0.64, 0.68-0.73, and 0.74-0.76, respectively. During this period, clinical symptoms were generally attenuated in responders, which suggests that CSF metabolism was enhanced in line with the increase in the markers. In contrast, blood-derived marker levels were only moderately correlated in this period. We therefore propose that brain-derived markers, but not blood-derived markers, could serve as indicators or biomarkers of altered CSF production.
Lastly, we discuss the diagnostic applications of the markers examined in our work. Mase et al., reported that L-PGDS levels were decreased in the CSF of NPH patients [29]. We also demonstrated a significant decrease in GlcNAc-Tf, in addition to L-PGDS. Given that L-PGDS showed a sensitivity of 82% and a specificity of 57%, while GlcNAc-Tf had a sensitivity of 96% and a specificity of 28%, our results suggest that these markers would not serve as good diagnostic markers for iNPH.
The present results revealed that all markers were elevated in SIH compared with the non-SIH CSF samples. Based on its sensitivity of 97%, specificity of 95%, and AUC of 0.98, our results suggest that Man-Tf could be a good diagnostic marker for SIH. In contrast, the other markers we examined would not be applicable for diagnostic purposes given their low sensitivity values (46~77%).
We previously demonstrated that Man-Tf derived from neurons could be used as a diagnostic marker for AD and MCI [11]. Indeed, Man-Tf levels in the CSF are highly correlated with those of phosphorylated-tau protein (p-tau), a marker for AD-specific neuronal death. A concomitant increase in Man-Tf and p-tau led the authors to develop an index, Man-Tf x p-tau, as a diagnostic marker. The index showed a sensitivity of 94% and a specificity of 89% for AD, and a sensitivity of 84% and a specificity of 90% for MCI. The present study revealed that Man-Tf was markedly increased in SIH, indicating that such an increase in Man-Tf is not valid as a specific marker for AD pathology. Nevertheless, measuring Man-Tf or the index would still be of diagnostic value for both diseases. SIH patients, who are usually less than 50 years old, mostly complain of orthostatic headaches, while AD and MCI patients, who are more likely to be in their sixties or older, tend to show signs of dementia. Together with the distinctive difference in symptoms and the peak age of onset, the Man-Tf measure should serve as a supportive diagnostic marker for AD and SIH. Tf captures ferric iron with very high affinity (Kd = 1 × 10 −20 ), which prevents Fe-dependent cellular toxicity. Indeed, Fe toxicity is suspected to induce neuronal death in AD pathology. Man-Tf, mainly produced by neurons, may play a role as a local protein that protects against iron toxicity, although the mechanisms underlying this have yet to be elucidated.

Patients and CSF Samples
Patients were consecutively recruited from Fukushima Medical University (FMU), Juntendo University (JTN), Sanno Hospital (SNO), and the National Center for Neuropsychiatric Disease (NCNP). The number of patients, age (mean ± S.D.), male/female distribution, and hospitals for CSF collection for each disease type are listed in Table 13. Depression and schizophrenia were diagnosed based on "Diagnostic and Statistical Manual of Mental Disorders-DSM-5". SIH was suspected by clinical presentation, particularly by orthostatic headaches, MRI and/or computed tomography (CT), and RI scintigraphy. Patients were diagnosed based on the International Classification of Headache Disorders, 3rd edition (beta-version) [18] and the diagnostic criteria by Schievink et al. [17]: (i) morphological evidence of CSF leakage, such as pachymeningeal enhancement, on cranial MRI and/or low CSF opening pressure (≤60 mm H 2 O), (ii) no recent history of dural puncture, and (iii) not attributable to another disorder. The iNPH was diagnosed according to clinical guidelines issued by the Japanese Society of Normal Pressure Hydrocephalus, Version 3 [30]. Briefly, iNPH was suspected based on clinical symptoms such as gait disturbance, dementia, and urinary incontinence. In addition, excess CSF in the ventricles (ventriculomegaly) was observed by CT or MRI. Patients with suspected iNPH underwent a tap test, which is a time-specific temporal removal of 30 mL of CSF by lumbar puncture. When the patients showed only a subtle improvement in their symptoms after CSF removal (tap test-negative), they were classified into the non-iNPH group. When the patients showed an evident improvement in their symptoms, they were classified into the probable iNPH group and underwent a shunt operation that bypassed excess CSF into the peritoneal cavity. Either a lumbar-peritoneal (LP) or a ventricle-peritoneal (VP) shunt was performed by using a Codman Hakim programmable valve with Siphonguard (#6244000, Integra Japan Co., Ltd., Tokyo, Japan) or a Medtronic Strata system (Medtronic Japan Co., Ltd., Tokyo, Japan).
Patients showing an improvement in their symptoms after surgery were diagnosed as responders or definite iNPH. Several responders underwent a time-course analysis after the shunt operation, in which specimens were withdrawn from the shunt valve. The iNPH condition is occasionally a comorbidity associated with AD. Comorbid patients were excluded based on elevated levels (>50 pg/mL) of p-tau protein, a specific marker of neuronal death in AD pathophysiology. Total Tf was quantified with a Human Transferrin ELISA Quantitation Set (E80-128-23, Bethyl Laboratories, Inc., Montgomery, TX, USA). The protein concentration was estimated using a Micro-BCA protein assay kit (23235, Thermo Fisher Scientific-JP, Tokyo, Japan).

Statistical Analyses
Statistical analyses were performed using SPSS software (version 26). Data normality was examined using the Shapiro-Wilk test. Spearman's rank correlation coefficients (rs) and Pearson's correlation coefficients (r) were analyzed with and without data normality, respectively. Significant differences among multiple comparisons were assessed by the Kruskal-Wallis method followed by Bonferroni correction. Non-parametric comparisons were assessed by a Mann-Whitney U test.

Conclusions
The present study reveals that CSF contains brain-derived glycoproteins such as L-PGDS, Man-Tf, and GlcNAc-Tf, and that their CSF levels were highly correlated with each other in several neurological diseases. INPH, disorders featuring abnormal CSF metabolism, was treated with shunt operation. After the surgical intervention, brainderived glycoproteins increased over a year and their CSF levels were highly correlated with each other. The present observational and interventional study reveals that the glycoproteins are novel biomarkers for CSF metabolism.