The Presence of Caffeic Acid in Cerebrospinal Fluid: Evidence That Dietary Polyphenols Can Cross the Blood-Brain Barrier in Humans

Epidemiological data indicate that a diet rich in plant polyphenols has a positive effect on brain functions, improving memory and cognition in humans. Direct activity of ingested phenolics on brain neurons may be one of plausible mechanisms explaining these data. This also suggests that some phenolics can cross the blood-brain barrier and be present in the brain or cerebrospinal fluid. We measured 12 phenolics (a combination of the solid-phase extraction technique with high-performance liquid chromatography) in cerebrospinal fluid and matched plasma samples from 28 patients undergoing diagnostic lumbar puncture due to neurological disorders. Homovanillic acid, 3-hydroxyphenyl acetic acid and caffeic acid were detectable in cerebrospinal fluid reaching concentrations (median; interquartile range) 0.18; 0.14 µmol/L, 4.35; 7.36 µmol/L and 0.02; 0.01 µmol/L, respectively. Plasma concentrations of caffeic acid (0.03; 0.01 µmol/L) did not correlate with those in cerebrospinal fluid (ρ = −0.109, p = 0.58). Because food (fruits and vegetables) is the only source of caffeic acid in human body fluids, our results indicate that the same dietary phenolics can cross blood-brain barrier in humans, and that transportation of caffeic acid through this barrier is not the result of simple or facilitated diffusion.


Cerebrospinal Fluid Collection
One and a half of the milliliter cerebrospinal fluid samples (CSF) were collected by lumbar puncture in polypropylene tubes. Half a milliliter thereof was placed into a separate Eppendorf tube, centrifuged (10 min, 1500× g, 4 • C) and the supernatant was stored at −80 • C until the phenolics measurement, but not for longer than 3 months. The rest was used for the determination of routine parameters. CSF samples with red blood cells count >500/µL were recognized as bloody [27] and excluded, in order to avoid false positive results of phenolics determination.

Matched Plasma Samples
Four mL fasting venous blood was collected into vacutainer tubes with EDTA (Becton Dickinson, Franklin Lakes, NJ, USA), within 30 min before lumbar puncture. Blood was centrifuged (10 min, 1500× g, 4 • C) and obtained plasma was stored under the same conditions as CSF samples.

Determination of Selected Phenolics in CSF and Plasma Samples
We determined 12 phenolics in CSF and plasma specimens with a combination of solid-phase extraction technique with high-performance liquid chromatography, with electrochemical (HPLC-ECD) or ultraviolet-visible (HPLC-UV-Vis, diode array detector wavelengths 210, 280, and 325 nm) detection, as previously described [28,29]. Vanillic acid, dihydrocaffeic acid, caffeic acid (CA), and HVA were determined by HPLC-ECD, while 3-hydroxyphenyl acetic acid (3HPAA), hippuric acid, 3-hydroxyhippuric, 4-hydroxyhippuric, 3,4-dihydroxybenzoic acid, chlorogenic acid, ellagic acid, and urolithin A were determined by HPLC-UV-Vis. All procedures related to polyphenols determination were the same for CSF and plasma samples. The instrumentation of HPLC-ECD and HPLC-UV-Vis and chromatographic conditions and parameters were described in detail elsewhere [28,29]. All chemicals, buffers, and water used throughout the study were of high-performance liquid chromatography (HPLC) grade.

Solid Phase Extraction Technique
Plasma or CSF samples (0.5 mL) were mixed with 50 µL of 0.78 mol/L acetate buffer (pH 5.2), 25 µL of 20% ascorbic acid solution, 10 µL of 2 mg/L fisetyn solution, 10 µL of β-glucuronidase from bovine liver (type B-3) solution in 0.1 mol/L acetate buffer (4000 U/mL, pH 5.2), and 10 µL of sulphatase from Helix pomatia (type H-1) solution in 0.1 mol/L acetate buffer (1028 U/mL, pH 5.2; Sigma-Aldrich Chemical, St. Louis, MO, USA), and incubated for 60 min at 37 • C (to hydrolyze phenolic glucuronides and to remove sulfate group from phenolics). Then, 0.5 mL of 1 mol/L phosphoric acid solution was added and the sample was incubated again for 10 min at 37 • C and then mixed with 0.5 mL water and poured into a Speedisk Column H2OPhobic DVB (3 mL solid phase extraction column, 50 mg per column, 15 µm particle diameter, J.T. Baker, Phillipsburg, NJ, USA), preconditioned with 1 mL of methanol and 0.5 mL of water. The column was washed three times with 0.5 mL of water, dried under a vacuum, and then eluted with 0.4 mL of methanol. The methanol eluate containing polyphenols was used for further analyses [28,29].

HPLC-ECD Separation and Detection
Fifty µL samples of the eluate were mixed with 450 µL of mobile phase and 20 µL samples of these solutions were subjected to isocratic analysis with HPLC-ECD. The flow rate of the mobile phase was kept constant at 0.20 mL/min and the total run time was 90 min. The mobile phase was prepared by addition of 0.75 g KH 2 PO 4 , 40 mg EDTA, 3 mL 1 mol/L H 3 PO 4 , and 45 mL of methanol to 440 mL of water, and then the pH was adjusted to 2.80 with concentrated CH 3 COOH, and the volume was made up to 500 mL with water. Separation was achieved with Hypersil BDS C18 column (150 × 2.1 mm ID, 3-µm particles) and detection was executed with an electrochemical detector, with a flow-through detection cell equipped with a glass carbon electrode and an Ag/AgCl reference one set at +0.82 V; working temperature 32 • C (Decade, Antec Leyden, The Netherlands) [28,29].

HPLC Data Collection and Elaboration
Chromeleon software (Dionex, Sunnyvale, CA, USA) was used for chromatographic data collection and the calculation of phenolics concentrations. Identification of particular compounds was conducted on the basis of their spectral characteristics and retention times in comparison to corresponding standard substances. All standards came from Sigma-Aldrich Chemie GmbH (Steinheim, Germany) or from Fluka, Sigma-Aldrich (Buchs, Steinheim, Germany). Individual results were obtained as means from triplicate measurements and expressed in micromoles per liter. The detection and determination limits and phenolics recovery from solid phase extraction columns were described elsewhere [29].

Statistical Analyses
Results are expressed as a mean (SD) and median (Me) and interquartile range (IQR). Differences between plasma and CSF concentrations of selected phenolics were evaluated, alongside differences between patient subgroups, such as sclerosis multiplex (SM) vs. other neurological disorders (OND) subgroup. They were estimated with the U Mann-Whitney, due to the non-parametric distribution of data evaluated with Shapiro-Wilk's W test (STATISTICA 13.1 StatSoft software). Spearman's rank correlation coefficient (ρ) was used for analysis associations between CSF phenolics levels and other parameters. A p value < 0.05 was considered significant.  Table 2). Dihydrocaffeic acid, vanillic acid and hippuric acid were detected in about half of studied plasma samples, while CSF was always negative for these compounds. The remaining phenolics (3-hydroxyhippuric, 4-hydroxyhippuric, chlorogenic acid, ellagic acid, and urolithin A) were detected neither in CSF, nor in plasma samples ( Table 2).

Results
Nutrients 2020, 12, x FOR PEER REVIEW 5 of 15

Statistical Analyses
Results are expressed as a mean (SD) and median (Me) and interquartile range (IQR). Differences between plasma and CSF concentrations of selected phenolics were evaluated, alongside differences between patient subgroups, such as sclerosis multiplex (SM) vs. other neurological disorders (OND) subgroup. They were estimated with the U Mann-Whitney, due to the nonparametric distribution of data evaluated with Shapiro-Wilk's W test (STATISTICA 13.1 StatSoft software). Spearman's rank correlation coefficient (ρ) was used for analysis associations between CSF phenolics levels and other parameters. A p value < 0.05 was considered significant.

Results
Figures 1 and 2 illustrate representative chromatograms of the CSF sample obtained from a single patient. From 12 measured phenolics, only three (CA, HVA and 3HPAA) were detected in both CSF and corresponding plasma samples. However, the number of positive plasma samples for 3HPAA was approximately 5-times lower than the number of positive readings of this compound in CSF samples (5 vs. 27, Table 2). Dihydrocaffeic acid, vanillic acid and hippuric acid were detected in about half of studied plasma samples, while CSF was always negative for these compounds. The remaining phenolics (3-hydroxyhippuric, 4-hydroxyhippuric, chlorogenic acid, ellagic acid, and urolithin A) were detected neither in CSF, nor in plasma samples ( Table 2).  Table S5). The bottom chromatogram illustrates the elution of standards: A-dihydrocaffeic acid, B-vanillic acid, C-caffeic acid, D-homovanillic acid. Numbers over the peaks represent their area.  Table S5). The bottom chromatogram illustrates the elution of standards: A-3-hydroxyphenyl acetic acid. Numbers over the peaks represent their area.  Table S5). The bottom chromatogram illustrates the elution of standards: A-dihydrocaffeic acid, B-vanillic acid, C-caffeic acid, D-homovanillic acid. Numbers over the peaks represent their area.

Statistical Analyses
Results are expressed as a mean (SD) and median (Me) and interquartile range (IQR). Differences between plasma and CSF concentrations of selected phenolics were evaluated, alongside differences between patient subgroups, such as sclerosis multiplex (SM) vs. other neurological disorders (OND) subgroup. They were estimated with the U Mann-Whitney, due to the nonparametric distribution of data evaluated with Shapiro-Wilk's W test (STATISTICA 13.1 StatSoft software). Spearman's rank correlation coefficient (ρ) was used for analysis associations between CSF phenolics levels and other parameters. A p value < 0.05 was considered significant.

Results
Figures 1 and 2 illustrate representative chromatograms of the CSF sample obtained from a single patient. From 12 measured phenolics, only three (CA, HVA and 3HPAA) were detected in both CSF and corresponding plasma samples. However, the number of positive plasma samples for 3HPAA was approximately 5-times lower than the number of positive readings of this compound in CSF samples (5 vs. 27, Table 2). Dihydrocaffeic acid, vanillic acid and hippuric acid were detected in about half of studied plasma samples, while CSF was always negative for these compounds. The remaining phenolics (3-hydroxyhippuric, 4-hydroxyhippuric, chlorogenic acid, ellagic acid, and urolithin A) were detected neither in CSF, nor in plasma samples ( Table 2).   Table S5). The bottom chromatogram illustrates the elution of standards: A-3-hydroxyphenyl acetic acid. Numbers over the peaks represent their area.  Table S5). The bottom chromatogram illustrates the elution of standards: A-3-hydroxyphenyl acetic acid. Numbers over the peaks represent their area. Other phenolics, such as 3-hydroxyhippuric, 4-hydroxyhippuric, chlorogenic acid, ellagic acid, and urolithin A, were not detected in CSF and plasma samples.

Correlations of Polyphenols in CSF with Other Measured Variables
CSF levels of CA and HVA acid did not correlate with their concentrations in plasma (Figures 4 and 5). Similarly, CSF concentrations of CA, HVA and 3HPAA did not correlate with other CSF parameters such as cell count, total protein and glucose (ρ ranged from −0.243 to 0.285, p > 0.05).

Discussion
We found three phenolics: HVA, 3HPAA and CA in almost all studied CSF samples (Supplementary  Tables S1 and S2). The presence of CA in human CSF is described for the first time to the best of our knowledge. These afore-mentioned, relatively simple phenolic acids are present in some fruits and vegetables and can also be produced by gut microflora from ingested, more complex polyphenols and then absorbed into the blood stream [30]. Their existence in CSF suggests that they may cross BBB or blood-CSF barrier and be proof that plant phenolics can exert a direct neuro-protective effect explaining epidemiological data on the beneficial influence of a polyphenol rich diet on brain function during ageing [11]. However, this idea has important limitations, because HVA and 3HPAA can be formed endogenously in the brain [31][32][33]. On the other hand, no metabolic pathways leading to endogenous production of CA were identified in the human body so far [34,35]. Thus, CA may confirm the hypothesis that some ingested plant phenolics can reach brain neurons in humans.

Homovanillic Acid in CSF
Certain fractions of catecholamines (adrenaline, noradrenaline, dopamine) released by brain neurons undergoes several reactions catalyzed by brain monoamine oxidase (MAO) and catechol-O-methyltransferase (COMT), resulting in the formation of 3,4-dihydroxyphenylacetic acid, vanillylmandelic acid and HVA [31,32]. Catecholamines produced outside of the central nervous system are metabolized in a similar way [31]. Thus, the afore-mentioned compounds were found in plasma, urine, and CSF samples [22,31,32]. In our study, we measured 12 phenolics, including HVA in CSF and matched plasma samples. Measurement of HVA served as a positive control of our analytical techniques. We found HVA in all samples of CSF and venous plasma. Higher concentration of HVA in CSF than in plasma corresponds to previous observations concerning higher levels of this compound in internal jugular venous plasma than in arterial plasma in humans [36]. Surprisingly, we did not observe significant correlation between CSF and plasma HVA concentrations in studied patients. Although HVA can leak from brain into systemic circulation, the brain derived HVA contributes only to about 10% of total body HVA production [36]. Moreover, dietary flavanols (e.g., quercetin) present in vegetables (e.g., tomato, onion) and beverages (e.g., tea or coffee) can increase the circulating pool of HVA [37,38]. This may explain the negative result of the afore-mentioned correlation analysis. On the other hand, it is interesting whether dietary intervention elevating the HVA circulatory level above that present in CSF can reverse HVA transportation through BBB or the blood-CSF barrier.

3-hydroxyphenyl Acetic Acid in CSF
Notably, 3HPAA is one of the metabolites of rutin and quercetin formed by colonic microbiota in humans [39,40]. It can be absorbed from the colon into the blood stream, because its urinary excretion rises after the introduction of high polyphenol diet in humans [41]. Increased urinary excretion of this compound was also noted after ingestion of a single dose of soluble cocoa powder or dark chocolate [42,43]. The concentrations of 3HPAA in CSF are similar to those reported by other authors [33,44]. However, the authors recognized 3HPAA as a metabolite of oxidative deamination of p-tyramine [33], and not as a metabolite of plant polyphenols ingestion. p-tyramine belongs to "trace amines" and can act as a neurotransmitter via G-protein coupled receptors in the brain. It can be formed endogenously as a product of enzymatic decarboxylation of tyrosine [45], or ingested with food (e.g., fish, cheese, chocolate, certain vegetables) [46]. In vitro studies with the model of BBB suggest that dietary tyramine can reach the central nervous system [47]. In humans, the diet seems to be the main source of physiologically relevant tyramine concentrations [46]. The mean concentration of 3HPAA was many times higher than that in corresponding plasma. Moreover, the number of 3HPAA positive CSF samples was about 5.4-times greater than the number of positive plasma samples in studied patients. These suggest that CSF 3HPAA could be the result of the penetration of plasma tyramine into the brain, with its subsequent oxidative deamination. Thus the existence of 3HPAA in CSF cannot be recognized as a convincing evidence of any transport of plant phenolics through BBB in humans. Interestingly, increased urinary excretion of 3HPAA was found in children with phenyloketonuria [48] and autistic patients [49], as a consequence of the abnormal metabolism of phenylalanine. The overgrowth of intestinal microbiota such as Clostridium species may also lead to the enhanced formation of 3HPAA [49]. Because some brain dysfunctions are observed in autistic patients and those with untreated phenyloketonuria, one may speculate that circulating 3HPAA may cross BBB and be toxic for neurons. On the other hand, in vitro experiments revealed a protective effect of 3HPAA on the BBB model against oxidative stress [50]. Therefore, as to whether 3HPAA is a biomarker of the afore-mentioned diseases or rather a metabolite involved in the development of brain damage is still an open question.

Caffeic Acid in CSF
We described, for the first time, the existence of CA in human CSF. Food (e.g., olives, fruits, carrots, coffee beans) is the only source of CA in human body fluids. This compound is absorbed in the free form in the colon and undergoes methylation, sulphatation or glucuronidation that makes the molecule of CA more hydrophilic, facilitating its urinary elimination [51]. The concentration of CA in CSF was almost the same as that in plasma. However, there was no correlation between CSF and plasma levels of CA. These suggest that the transportation of CA through BBB could not to be the result of simple or facilitated diffusion. Intestinal cells actively absorb free CA using monocarboxylic acid transporters [51]. Various solute carriers, including monocarboxylic acid transporters, are present on the capillary endothelial cells of BBB [52]. Experiments on animals and on in vitro models of BBB revealed that transportation of various substances including phenolics is affected by their methylation, sulphatation or glucuronidation [53,54]. Methylation or glucuronidation of selected polyphenols increased their traversing through BBB models in vitro [54]. Sulphatation of dopamine caused its permeation through the BBB in rats [53]. Moreover, transient conjugation or de-conjugation of polyphenols can take place in close vicinity of BBB [54]. Since CA is present in the plasma in free and conjugated form, the process of its transportation through BBB may depend on these afore-mentioned factors. These may explain no association between CSF and plasma levels of CA. On the other hand, low concentrations of CA close to the limit of the assay sensitivity can be responsible for the negative results of the correlation analysis. Nevertheless, the presence of CA in CSF shows that at least some dietary plant phenolics after absorption to the blood can cross BBB in humans and directly affect the functions of brain cells. These are in agreement with previous experiments on laboratory animals, showing neuroprotective activity of CA [55,56] and plausible permeation of this phenolic acid, through BBB into brain in mice [57].

Limitations of the Study
Although exclusion criteria involved diseases (stroke, head trauma, acute liver failure, hypertensive encephalopathy, HIV infection) that destroy the integrity and function of BBB [58,59], other diseases that were an indication for lumbar puncture and CSF collection may also enhance BBB permeability. For instance, MS and diabetes mellitus were reported to increase BBB permeability [58]. Moreover, two patients suffered from meningitis and one from brain tumor, diseases which potentially induce BBB dysfunction. Thus one may suppose that our results concerning CA concentrations in CSF were higher than those in healthy subjects. On the other hand, 4 patients with mononeuropathy or polyneuropathy without coexisting diabetes mellitus and any pathology localized in the brain represented detectable CA in CSF. Systemic inflammation negatively affects BBB function [58]. The majority of studied patients (especially those with MS) had plasma CRP, a marker of intensity of inflammatory response within normal range. In addition, the lack of correlation between CSF and plasma concentrations of CA suggests that simple permeation was not the main mechanism of CA transportation through BBB. These indicate that our results were not significantly biased by the presence of neurological disorders in studied subjects.
We used relatively nonspecific analytical methods (HPLC-UV-Vis and HPLC-ECD) for the measurement of phenolics in human CSF and matched plasma. An application of HPLC with mass spectrometric detection would probably give an opportunity to find more dietary phenolics in CSF. However, despite this, we achieved our goal. We describe for the first time the presence of CA in human CSF. Plasma and CSF samples were collected under fasting conditions after 15 h from the last meal. Maximal increase in CA plasma levels was observed within 0.5-1 h after consumption of 200 mL of red wine containing 1.8 mg of CA [60]. Similarly, an ingestion of 200 mL of instant coffee resulted in the maximal increase in chlorogenic acid plasma levels (an ester of caffeic acid and quinic acid) after 1 h, with a half-life time of about 0.3 h [34]. These suggest that non-fasting plasma and CSF samples could have considerably increased CA concentrations. On the other hand, studies with ileostomy patients revealed that almost all the ingested CA dose was absorbed in the small intestine, while in the case of chlorogenic acid it was only one third of the dose [61]. This indicates that the majority of chlorogenic acid from food could reach the colon and be hydrolyzed by colon microbiota into free CA, with its subsequent absorption into the blood. Thus, the absorption of CA can occur even after several hours from the last meal and the conditions of blood and CSF samples collection can have a rather moderate effect on CA measurement results. However, the considerable differences between CSF concentrations of CA and the remaining two phenolics (HVA and 3HPAA) are in line with these data and support, to the same extent, the conclusion that the presence of caffeic acid in CSF results only from its passage through the blood-brain barrier. Other phenolic acids (e.g., ferulic acid, dihydrocaffeic acid, 3,4-dihydroxybenzoic acid) are also rapidly absorbed in the small intestine and reach maximal plasma concentration, within 1 to 2 h after ingestion. Their elimination lasts about a few hours [60]. Similarly, hippuric acids (hippuric acid, 3-hydroxyhippuric, 4-hydroxyhippuric), which are metabolites of various polyphenols, are rapidly excreted with the urine [60]. Therefore, it is possible that the lack of detectable concentrations of these compounds in plasma and CSF samples could result from the relatively long fasting period preceding their collection. This is in line with our previous observations [29]. In strawberry consumers (500 g of strawberry pulp daily, between 11:00 a.m. and 2:00 p.m., for 30 days) fasting plasma had no detectable ellegic acid and its main metabolite urolithin A, while HPLC analysis of spot morning urine revealed the elevated concentrations of this last compound [29]. Nevertheless, an analysis of post-meal CSF and plasma samples would clearly solve this issue.

Concluding Remarks
Screening of human CSF samples for 12 selected phenolics with HPLC-UV-Vis and HPLC-ECD revealed the presence of HVA, 3HPAA and CA. The source of the first two compounds in CSF is dual. They can originate from endogenous metabolic processes in the brain and can be absorbed into the blood as derivatives of ingested plant polyphenols produced by gut microflora. However, the only source of CA in human body fluids is food. This indicates that CA absorbed from ingested fruits and vegetables into the blood stream can cross BBB and reach brain cells. This is direct evidence that some plant phenolics can cross BBB in humans.