Evaluation of blood–brain barrier thiamine efflux using the in situ rat brain perfusion method


Address correspondence and reprint requests to David D. Allen, PhD, Department of Pharmaceutical Sciences, Texas Tech University HSC, 1300 So. Coulter Dr, Amarillo, Texas, 79106-1712, USA. E-mail: dallen@cortex.ama.ttuhsc.edu


Thiamine is an essential, positively charged (under physiologic conditions), water-soluble vitamin requiring transport into brain. Brain thiamine deficiency has been linked to neurodegenerative disease by subsequent impairment of thiamine-dependent enzymes used in brain glucose/energy metabolism. In this report, we evaluate brain uptake and efflux of [3H]thiamine using the in situ rat brain perfusion technique. To confirm brain distribution was not related to blood–brain barrier endothelial cell uptake, we compared parenchymal and cell distribution of [3H]thiamine using capillary depletion. Our work supports previous literature findings suggesting blood–brain barrier thiamine uptake is via a carrier-mediated transport mechanism, yet extends the literature by redefining the kinetics with more sensitive methodology. Significantly, [3H]thiamine brain accumulation was influenced by a considerable efflux rate. Evaluation of the efflux mechanism demonstrated increased stimulation by the presence of increased vascular thiamine. The influx transport mechanism and efflux rate were each comparable throughout brain regions despite documented differences in glucose and thiamine metabolism. The observation that [3H]thiamine blood–brain barrier influx and efflux is regionally homogenous may have significant relevance to neurodegenerative disease linked to thiamine deficiency.

Abbreviations used

cerebrovascular permeability surface-area product

Thiamine is an essential, water-soluble vitamin required for normal brain glucose utilization. Dietary thiamine exists and is converted in vivo to three primary states: thiamine, thiamine monophosphate and thiamine diphosphate (active form). Thiamine and its monophosphate form are present both intra- and extracellularly, whereas the diphosphate form exists only within the cell. The active diphosphate functions as an essential enzymatic cofactor for the cytosolic and mitochondrial tricarboxylic acid cycle enzymes transketolase, pyruvate dehydrogenase and α-ketoglutarate dehydrogenase. The relationship of thiamine conversion to its active form consists of intracellular thiamine phosphorylation to the diphosphate form. Thiamine monophosphate must be hydrolyzed intracellularly to thiamine before being converted to the diphosphate form (Singleton and Martin 2001).

Wernicke–Korsakoff syndrome is the preeminent disease process illustrating CNS thiamine deficiency. This neurologic disorder is seen in a susceptible population suffering from chronic alcoholism. Characteristics of the disease include memory loss and selective neuronal degeneration (Victor et al. 1971). The etiology is presumed to be related to poor nutrition and impaired intestinal thiamine absorption, both of which lead to inadequate plasma thiamine levels and subsequent diminished brain distribution (Thomson 2000). Furthermore, inadequate concentrations of free brain thiamine (and its diphosphate) are thought to play a role in other neurodegenerative diseases such as progressive supranuclear palsy, Alzheimer's, Parkinson's and Huntington's diseases (Gibson and Zhang 2002). The central link of neurodegnerative disease to thiamine deficiency is hypothesized to be the impairment of thiamine-dependent enzymes and subsequent reduction in cerebral glucose utilization (Hakim and Pappius 1981).

The transport mechanism of thiamine and thiamine monophosphate into and out of the cell is critical for thiamine diphosphate concentrations and has been evaluated extensively (review; Singleton and Martin 2001). Similarly, transport of thiamine into brain from plasma is as fundamental, given extracellular brain thiamine concentrations drive thiamine diphosphate concentrations in neuron and glia (Gibson and Zhang 2002). For thiamine to penetrate the CNS, it must cross the blood–brain barrier. The blood–brain barrier is comprised of a continuous layer of endothelial cells connected by tight junctions circumferentially surrounding the cell margin (Butt et al. 1990). These junctions endow the barrier with properties similar to a cell membrane with regard to transcellular solute movement. Specifically, lipid soluble molecules move rapidly across the blood–brain barrier by diffusion across lipoid endothelial cell membranes, whereas hydrophilic compounds show restricted permeation unless transported by a carrier exchange mechanism (Smith 1996). Because thiamine is a physiologically charged monovalent cation (Komai and Shindo 1974), the blood–brain barrier attenuates passive brain entry but allows brain penetration through a carrier-mediated transporter (Greenwood et al. 1982; Patrini et al. 1988).

Net thiamine movement into brain has been reported to be the difference between total influx less proposed, but not demonstrated, efflux. Influx at the blood–brain barrier is mediated predominantly by a saturable carrier-mediated transport mechanism (approximately 90%) and a non-saturable mechanism, presumably passive diffusion. A possible efflux mechanism at the blood–brain barrier has also been suggested (Greenwood et al. 1986) but not fully explored. Thiamine efflux has been demonstrated at the membranes of neuroblastoma cells (Bettendorf and Wins 1994; Bettendorf 1995), rat liver mitochondria (Barile et al. 1990), Escherichia coli (Nishimune et al. 1981), and Schizosaccharomyces pombe (Hirose 2000).

Current literature has implicated thiamine deficiency as an etiology for many significant neurodegenerative diseases and current blood–brain barrier kinetic methodologies are more sensitive than previous methods, therefore we evaluated thiamine kinetics at the blood–brain barrier, considering both efflux and influx parameters. Further, the net movement of thiamine into brain is discussed in its potential relationship to thiamine deficiency-related neurodegenerative disease.

Materials and methods

Uptake of [3H]thiamine into brain was assessed using the in situ rat brain perfusion technique of Takasato et al. (1984) with modifications described (Smith 1996; Allen et al. 1997). In this study, short perfusions of 5–135 s were used to determine initial brain uptake. Saturation kinetics of the transporter were completed at 15 s to minimize [3H]thiamine brain efflux. All studies were approved by the Institutional Animal Care and Use Committee and were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.


High specific activity [3H]thiamine (10 Ci/mmol, > 98% purity) was obtained from American Radiolabeled Chemicals (St. Louis, MO, USA). The [14C]sucrose (4.75 mCi/mmol) was obtained from Dupont-New England Nuclear (Boston, MA, USA). In each experiment, [3H]thiamine was dried prior to being dissolved in perfusion buffer, to remove volatile tritium contaminants including [3H]H2O.

Perfusion procedure

Male Fischer-344 rats (220–330 g; Charles River Laboratories, Kingston, NY, USA) were anesthetized with sodium pentobarbital (50 mg/kg intraperitoneal). A PE-60 catheter filled with heparinized saline (100 U/mL) was placed into the left common carotid artery after ligation of the left external carotid, occipital and common carotid arteries. Common carotid artery ligation was accomplished caudal to the catheter implantation site. The pterygopalatine artery was left open during the experiments (Allen et al. 1997). Rat rectal temperature was monitored and maintained at 37°C by a heating pad connected to a feedback device (YSI Indicating Controller, Yellow Springs, OH, USA). The catheter to the left common carotid artery was connected to a syringe containing buffered physiologic perfusion fluid [containing (in mm): NaCl 128, NaPO3 2.4, NaHCO3 29.0, KCl 4.2, CaCl 1.5, MgCl2 0.9, and d-glucose 9] with 0.75 µCi/mL [3H]thiamine (final thiamine concentration ∼ 75 nm) and 0.3 µCi/mL [14C]sucrose (to determine vascular volume). Perfusion fluid was filtered and warmed to 37°C and gassed with 95% O2 and 5% CO2. The pH and osmolarity of this solution were ≈ 7.35 and 290 mOsm, respectively, immediately prior to perfusion. The perfusion fluid was infused into the left carotid artery with an infusion pump for periods of 5–135 s at 10 mL/min (Harvard Apparatus, South Natick, MA, USA). This perfusion rate was selected to maintain a carotid artery pressure of ∼ 120 mmHg (Takasato et al. 1984).

Rats were decapitated and cerebral samples obtained as previously described (Allen and Smith 2001). Briefly, the brain was removed from the skull, and the perfused cerebral hemisphere dissected on ice after removal of the arachnoid membrane and meningeal vessels. Brain regions were placed in scintillation vials and weighed. In addition, two 50 µL aliquots of the perfusion fluid were transferred to a scintillation vial and weighed. The brain and perfusion fluid samples were then digested overnight at 50°C in 1 mL of 1 m piperidine. Ten milliliters of Fisher Chemical scintillation cocktail (Beckman, Fullerton, CA, USA) was added to each vial and the tracer contents assessed by dual-label liquid scintillation counting. Dual-label scintillation counting of brain and perfusate samples were accomplished with correction for quench, background and efficiency.

For wash out studies, a 45 s perfusion with [3H]thiamine was completed followed by a tracer-free buffered physiologic perfusion fluid infusion for 5–60 s. Vascular correction was accounted for by subtraction of [14C]sucrose vascular volume. Cerebral samples were obtained and evaluated similarly as described (Allen and Smith 2001).

To evaluate the distribution of [3H]thiamine into brain capillary endothelial cells, capillary depletion was accomplished using the method of Triguero et al. (1990). Briefly, the brain was removed after a 45 s in situ perfusion containing both [3H]thiamine and [14C]sucrose. The brain was then homogenized in physiologic buffer (4°C) in a glass homogenizer (6 strokes). After homogenization was complete, ice-cold dextran (34% w/v) was added to the homogenate, and further homogenization (4°C) was completed (4 strokes). The homogenate was then centrifuged at 5400 × g for 15 min at 4°C. The subsequent supernatant (brain parenchyma) was then carefully separated from the pellet (brain microvasculature) and both were prepared for dual-label scintillation counting.

Kinetic analysis

Concentrations of tracer in brain and perfusion fluid are expressed as dpm/g brain or dpm/mL perfusion fluid, respectively. Blood–brain barrier [3H]thiamine transport into brain was determined by perfusion with [3H]thiamine (75 nm) for a 5–135 s period as described previously (Takasato et al. 1984; Smith 1996). Given an apparent non-linear uptake pattern was observed, a calculated uptake transfer constant (Kin) and a brain efflux rate coefficient (kout) was estimated from the following relationship as described (Smith 1996):


where Q* is the quantity of 3H-tracer in brain (dpm/g) at the end of perfusion, C* is the perfusion fluid concentration of [3H]thiamine (dpm/mL) and T is the perfusion time. Tracer trapped in the vascular space was accounted for by the subtraction of [14C]sucrose vascular volume.

Calculation of the effective volume of distribution at steady state is by the relationship:


For determination of the saturable kinetics of [3H]thiamine transport, a perfusion time of 15 s was chosen that allowed an adequate amount of tracer to pass into brain and minimized apparent efflux. We then calculated Kin in single-point uptake experiments in each brain region from the following relationship as described (Smith 1996):


where Qtot = Qbr + Qvas represents the sum of the amount of thiamine remaining in the perfusate in the blood–brain vessels and the amount of thiamine that has penetrated into brain. Cerebral perfusion flow rate (F) was determined in separate experiments as previously described (Momma et al. 1987). Cpf is the perfusion fluid concentration of tracer thiamine and T is the net perfusion time. To ascertain points below the Km, we decreased perfusion fluid [3H]thiamine concentration to 37.5 nm by dilution of the perfusion fluid. This necessitated removal and aggregation of the entire cortical regions for statistically valid dual-label scintillation counting.

Kin values were converted to apparent cerebrovascular permeability surface-area products (PA) using the Crone–Renkin equation (Smith 1996),


where F is the cerebral perfusion flow determined from the uptake of [3H]diazepam according to the method of Momma et al. (1987). Regional perfusion flow was used for regional PA determination to account for regional flow variations. In all instances, PA differed by < 2% from Kin, because F exceeded Kin by > 40-fold. Concentration dependent [3H]thiamine brain uptake was evaluated as a single saturable and non-saturable process where:


Statistical analysis

Data presented are from the frontal cerebral cortex unless otherwise specified. [3H]Thiamine brain uptake and PA reduction over time were fit with non-linear regression using least-squares analysis. Washout regression lines were calculated with least-squares linear regression. One-way anova analysis followed by a Bonferoni's multiple comparison test were used for evaluation of regional brain uptake and efflux of [3H]thiamine. For all data, errors are reported as standard error of the mean unless otherwise indicated. Differences were considered statistically significant a priori at the p < 0.05 level (GraphPad Prism version 3.00 for Windows, GraphPad Software, San Diego, CA, USA).


[3H]Thiamine brain uptake

Figure 1 illustrates the time course of [3H]thiamine brain uptake measured using the in situ rat brain perfusion method. The brain/perfusion fluid concentration ratio of [3H]thiamine increased non-linearly over 5–135 s in the absence of unlabeled thiamine. Brain/perfusion fluid ratios (i.e. volume of distribution or ‘space’) were fit to eqn 1 (r2 = 0.978) (Fig. 1). The calculated Kin and estimated kout for [3H]thiamine uptake were 1.05 ± 0.2 × 10−3 mL/s/g and 2.7 ± 0.6 × 10−2 s−1, respectively. The calculated brain distribution volume at steady state (Kin/kout) equaled 0.039 ± 0.0033 mL/g. Of significance, the [3H]thiamine kout significantly influenced blood–brain barrier influx.

Figure 1.

Non-linear time course of [3H]thiamine brain uptake with perfusion of physiologic saline containing [3H]thiamine (75 nm). The line represents the least-squares fit of eqn 1. Data are mean ± SEM for frontal cortex; n = 3–5 for all points.

To ascertain if [3H]thiamine brain distribution seen in Fig. 1 is overestimated secondary to capillary endothelial association, we evaluated cellular localization of [3H]thiamine and [14C]sucrose (Table 1) using the capillary depletion method (Triguero et al. 1990). Rats were perfused with both tracers for 45 s. We observed a significant (p < 0.05) brain parenchyma (Dhomogenate) accumulation of [3H]thiamine (compared to [14C]sucrose) and a non-significant difference of the tracers associating with the pellet (Dpellet).

Table 1.  Capillary localization of [3H]thiamine and [14C]sucrose tracer after a 45 s perfusion uptake
 Total percentage of tracer
associated with pellet
  1. Cellular localization of [3H]thiamine and [14C]sucrose using capillary depletion. Values reported are for 45 s perfusions of [3H]thiamine and [14C]sucrose at tracer concentrations. Values shown are mean ± SEM, n = 3–5.

[3H]Thiamine0.4 ± 0.11%
[14C]Sucrose0.8 ± 0.21%

Transport kinetics for [3H]thiamine

To evaluate saturable transport kinetics a perfusion time of 15 s was chosen and evaluated as single time brain uptake experiments with data being fit to eqn 5 (r2 = 0.948) (Takasato et al. 1984; Smith 1996). Saturable kinetic parameters were evaluated with the addition of unlabeled thiamine to the perfusion fluid at concentrations of 25–10 000 nm. Figure 2 shows dose-dependent reduction of [3H]thiamine PA in the presence of unlabeled thiamine for the frontal cortical region. Saturable kinetic parameters observed were: Km = 96 ± 6 nm, Vmax = 8.7 ± 1.6 pmol/min/g, and Kd = 0.42 ± 0.04 × 10−3 mL/s/g. To evaluate points below the Km, the perfusate concentration was decreased to 37.5 nm. Given the subsequent decreased tracer penetrating brain it was necessary to evaluate the entire cortical region as a single region. The PA obtained in this experiment (data not shown) was consistent with the saturable kinetic data shown in Fig. 2. Table 2 lists the estimated saturable transport parameters in various brain regions. Significant differences for PA were seen in the occipital cortex, hippocampus and the caudate putamen region, whereas minimal regional differences were noted for Km and Vmax.

Figure 2.

Demonstration of concentration-dependent [3H]thiamine brain uptake. Shown is the cerebrovascular permeability surface-area product (PA) in relation to addition of unlabeled thiamine (concentrations added were 25–1000 nm). The PA at 75 nm represents [3H]thiamine tracer brain uptake. The line represents the least square fit of eqn 4: r2 = 0.984; where Km = 95.5 ± 6.4 nm, Vmax = 8.7 ± 1.7 pmol/min/g, and Kd = 0.43 + 0.04 × 10−3 mL/s/g. All data represent mean ± SEM for frontal cortex; 15 s perfusions and n = 3–5 for all points.

Table 2.  Calculated cerebrovascular permeability (PA) and saturable kinetic parameters for [3H]thiamine transport at the blood–brain barrier
Brain regionPAKm
(mL/s/g × 103)
  • Regional cerebrovascular PA (mL/s/g) × 104; calculated Vmax, Km and Kd values for blood–brain barrier [3H]thiamine transport into brain. Values are reported for 15 s perfusions of [3H]thiamine at tracer concentrations (75 nm).

  • *

    Differs significantly from mean value for frontal cortex (p < 0.05). Values shown are mean ± SEM, n = 3–5.

Frontal cortex10.8 ± 0.495.5 ± 6.48.7 ± 1.70.42 ± 0.04
Parietal cortex12.4 ± 0.583.7 ± 3.53.6 ± 0.90.56 ± 0.03*
Occipital cortex8.3 ± 2.1*97.0 ± 8.23.1 ± 0.60.39 ± 0.03
Hippocampus7.8 ± 0.8*128.9 ± 32.7*14.4 ± 4.90.31 ± 0.06*
Caudate/putamen7.8 ± 1.0*75.1 ± 4.11.4 ± 0.60.39 ± 0.03
Thalamus/hypothalamus11.2 ± 1.085.9 ± 4.45.2 ± 1.30.31 ± 0.04*
Cerebellum8.9 ± 0.888.5 ± 8.45.0 ± 1.90.28 ± 0.05*
Pons/medulla8.5 ± 0.285.8 ± 6.53.8 ± 0.80.26 ± 0.04*

Efflux of [3H]thiamine at the blood–brain barrier

Since we observed non-linear brain uptake of [3H]thiamine over time, we calculated brain efflux constants using a washout method. Briefly, we evaluated the tracer brain/perfusion ratios after 45 s loading of [3H]thiamine with a following ‘wash’ of thiamine-free saline for periods of 5–60 s. Figure 3 shows the brain/perfusion ratio in the frontal cortex dropped significantly over a 60 s period (0 s: 0.024 ± 0.003 mL/g; 60 s: 0.010 ± 0.001 mL/g). The efflux constant (1.3 ± 0.18 × 10−2 s−1) is calculated by the linear regressed slope (−2.2 ± 0.25 × 10−4 ml/g) where kout = ln2/t½. Table 3 shows efflux is regionally homogenous. The calculated efflux constant (kout) for each brain region was not significantly different among regions (p > 0.05).

Figure 3.

Time course of [3H]thiamine washout from brain (frontal cortex) after 45 s of [3H]thiamine brain perfusion. Wash consisted of tracer-free saline. The calculated kout (1.3 ± 0.18 × 10−2 s−1) is based upon the linear regressed slope. All data represent mean ± SEM for frontal cortex; n = 3–5 for all points.

Table 3.  Regional efflux constants for [3H]thiamine in the absence of vascular thiamine
Brain regionkout (× 102 s−1)
  1. Calculated regional efflux constants in the presence of 0 nm thiamine. Efflux constants were evaluated by a 45 s loading of [3H]thiamine with a following ‘wash’ of thiamine and tracer-free saline for periods of 5–60 s. The efflux constant is calculated by linear regression where kout = ln2/t½. The calculated efflux constant (kout) for each brain region was not significantly different from the frontal cortex (p > 0.05).

Frontal cortex1.30 ± 0.2
Parietal cortex1.53 ± 0.4
Occipital cortex1.37 ± 0.6
Hippocampus1.29 ± 0.2
Caudate/putamen1.25 ± 0.6
Thalamus/hypothalamus1.46 ± 0.1
Cerebellum1.15 ± 0.2
Pons/medulla1.46 ± 0.3

To evaluate if efflux could be stimulated, such as a trans-stimulation mechanism, subsequent washout experiments were completed with unlabeled thiamine (300 nm) present in the wash. The brain/perfusion ratio of [3H]thiamine at 15 s was significantly reduced when 300 nm was added (thiamine-free saline: 0.022 + 0.0004 mL/g; 300 nm thiamine 0.011 + 0.0002 mL/g). Figure 4 shows the calculated efflux constants in the presence of varying vascular concentrations of thiamine. The efflux rate constant in the presence of 75 nm thiamine (Fig. 1; estimated during uptake using eqn 1) was approximately twofold higher than in the absence of thiamine and approximately half that of the presence of 300 nm thiamine (Fig. 4). This data suggested that the presence of vascular thiamine stimulated brain efflux. Regional examination of efflux constants at 300 nm were not significant from frontal cortical values (data not shown).

Figure 4.

Brain efflux [3H]thiamine constants at vascular thiamine concentrations of 0, 75 and 300 nm. Efflux constants were determined at 15 s. An * indicates p < 0.05 and ** indicates p < 0.01 when the brain/perfusion ratio is compared to 0 nm concentration. Data are mean ± SEM for frontal cortex; n = 3–5 for all data points.


The results of the studies presented herein confirm previous reports that thiamine influx at the blood–brain barrier is via a carrier-mediated transport mechanism, yet provide a new estimation of saturation kinetics. The importance of this report is twofold. First, net thiamine brain accumulation is significantly influenced by a rapid efflux mechanism that may be trans-stimulated. Second, the observation of homogenous influx and efflux of thiamine at the blood–brain barrier may explain regional depletion of thiamine-dependent enzymes in neurodegenerative disease linked to thiamine deficiency.

Initial experiments to determine brain [3H]thiamine distribution during perfusion time frames revealed two min distribution volumes to be consistent with previous work (Greenwood et al. 1982). However, in contrast to earlier work that suggested linear thiamine uptake in the initial period, we observed a non-linear [3H]thiamine brain uptake distribution pattern. The disparity is most explained by methodological considerations. In situ brain perfusions can accurately detect brain distribution of a high specific activity radiolabeled compound at 5 s (Takasato et al. 1984). The non-linear uptake suggested rapid uptake and subsequent efflux of [3H]thiamine from brain (Fig. 1). Similar non-linear uptake profiles are seen with drugs having significant brain efflux, such as cyclosporin A, vincristine, theophylline and iodoantipyrine (Murakami et al. 2000). Previous thiamine brain distribution literature has suggested an efflux mechanism may be present (Greenwood et al. 1986), but this mechanism has not been explored in detail.

To confirm initial [3H]thiamine distribution was not overestimated secondary to blood–brain barrier endothelial cell accumulation and/or association, we compared [3H]thiamine brain parenchyma and cell distribution after a 45 s perfusion. Table 1 shows significant [3H]thiamine brain parenchyma distribution (homogenate distribution compared to [14C]sucrose) and a non-significant difference of the tracers associating with the pellet (blood–brain barrier endothelial cells). This data suggests [3H]thiamine brain accumulation is related to penetration and unlikely to be related to significant endothelial cell thiamine uptake.

In order to evaluate saturable transport kinetics using in situ rat perfusions with significant efflux, it is necessary to perform single time uptake experiments with and without unlabeled compound for inhibition (Smith 1996). Given the non-linear time-dependent [3H]thiamine uptake, we chose a perfusion time based upon two factors: (i) it was long enough for [3H]thiamine detection, and (ii) it was short enough to minimize apparent efflux. Thus, a perfusion time of 15 s was chosen given [3H]thiamine brain uptake was approximately linear at this time. Figure 2 shows that addition of unlabeled thiamine (25–10 000 nm) to the perfusion fluid resulted in a dose-dependent reduction of PA, consistent with a saturable carrier-mediated process. Data points below the Km were evaluated by decreasing the perfusate tracer concentrations and evaluating the entire cortex as a single region. The PA obtained from this approach (data not shown) was consistent with the saturable kinetic data shown in Fig. 2.

Considering the carrier kinetics (Table 2), the observed Vmax has somewhat large errors, which may be attributable to using [3H]thiamine levels (75 nm) close to the calculated Km value. Higher specific activity [3H]thiamine (not commercially available) would allow better characterization of the saturation kinetics (i.e. the points below Km) and provide less error. However, Vmax estimates obtained in this report [1.4–144 pmol/min/g are consistent with previous estimates reported at the blood–brain barrier [16.5–18.6 pmol/min/g (Greenwood et al. 1982) and in a neuroblastoma cell line (Bettendorf 1995).

The apparent Km values (∼ 75–130 nm) obtained are less than previous apparent Km estimates of 0.61, 0.78, 2.5 and 4.98 µm (Greenwood et al. 1982, 1986; Reggiani et al. 1984; Patrini et al. 1988). The current values may be more physiologically consistent with baseline plasma thiamine levels that range from ∼ 6–240 nm (Weber and Kewitz 1985; Bettendorff et al. 1986; Patrini et al. 1993; Tallaksen et al. 1997).

Differences amongst the kinetic values may be related to the experimental time frames and thiamine concentrations. The data herein were completed in time frames from 5 to 135 s with high specific activity [3H]thiamine (concentrations ∼ 75 nm with no contribution from plasma thiamine), whereas previous studies used initial time frames of 1–2 min and low specific activity [14C]thiamine (concentrations ∼ 280 nm; Greenwood et al. 1986) in addition to baseline plasma thiamine concentrations. Other differences may be from methodology, not simultaneously considering brain vascular space or not accounting for thiamine erythrocyte accumulation (Casirola et al. 1990). The in situ rat perfusion method using high specific activity compounds accounts for the described limitations (Smith 1996).

To investigate and accurately account for the reported blood–brain barrier thiamine efflux mechanism (Greenwood et al. 1986), we perfused the brain with [3H]thiamine for 45 s followed by a tracer and thiamine-free perfusion buffer for 5–60 s. Significant efflux from the brain was observed over the 1-min period in the frontal cortical region (Fig. 3). The calculated [3H]thiamine efflux constant confirmed the efflux observed during initial uptake experiments. Examination of regional [3H]thiamine efflux is shown in Table 3 and the estimated efflux constant for each brain region was not significantly different.

To determine if additional thiamine could stimulate efflux, we perfused the brain with [3H]thiamine for 45 s, but with 300 nm of thiamine added to the wash buffer (vascular thiamine concentration). Thiamine included in the wash at a concentration of 300 nm resulted in a significant increase in [3H]thiamine efflux from brain at 15 s. This data is compared to the efflux constant calculated in the presence of 75 nm (during initial uptake experiment) in Fig. 4. The calculated efflux constants showed a significant difference in the presence of varying concentrations of vascular thiamine. This evidence of efflux stimulation suggests a trans-stimulation or self-exchange mechanism may be present for thiamine transport at the blood–brain barrier. A similar mechanism of increased thiamine efflux was found in a neuroblastoma thiamine homeostasis study (Bettendorf 1995). In this study, neuroblastoma cells loaded with radiolabeled thiamine had significantly increased efflux in the presence of increasing amounts of external thiamine. The author suggested the high-affinity thiamine influx carrier is a self-exchange mechanism that is sensitive to external thiamine and intracellular thiamine diphosphate concentrations.

Explanation as to the physiologic rationale for thiamine efflux from brain has not been elucidated; however, we propose two possibilities. First it is possible that the presence of high levels of plasma thiamine quickly saturate the intracellular environment of the blood–brain barrier with free thiamine (to be converted to thiamine diphosphate), which in turn stimulates the blood–brain barrier efflux mechanism. Second, Bettendorf (1995) proposed thiamine efflux from neuroblastoma cells is simply related to the storage of excessive thiamine. This rationale is based on previous work (Sanemori and Kawasaki 1982) demonstrating enormous doses of thiamine resulted in only small increased brain concentrations but significantly augmented thiamine concentrations in liver and erythrocytes. The hypothesis is the latter two compartments store thiamine for periods of deficiency, allowing remaining tissues to efflux excessive thiamine for storage. Whether this mechanism is by self-exchange or by an alternate pathway warrants further evaluation. Future work may also explore the possible asymmetric contribution of the luminal and abluminal membranes in either influx or efflux of thiamine movement at the blood–brain barrier.

Of major significance in this report is that [3H]thiamine brain uptake and efflux are to a large extent regionally homogenous. A hallmark characteristic of thiamine deficiency induced Wernicke–Korsakoff syndrome is the selective neuronal cell degeneration and/or death (Victor et al. 1971). Specifically, the vulnerable brain regions include: thalamic nuclei, inferior colliculus, inferior olivary, mammillary bodies and the lateral vestibular. Regions where this is not observed include the caudate nucleus, frontal-parietal cortex and the hippocampus (Todd and Butterworth 1999). This regional vulnerability has been positively correlated to the rate of thiamine neuronal uptake and metabolism in the specific brain regions. For example, Rindi et al. (1980) demonstrated that the cerebellum region has the greatest turnover and influx rate of thiamine, whereas the cerebral cortical region has the lowest. Given the regional differences in brain thiamine utilization, we initially hypothesized that there would be significant regional differences in blood–brain barrier thiamine transport. However, the characterization of regional PA, influx kinetics and efflux rates described herein demonstrate thiamine homeostasis at the blood–brain barrier is essentially consistent throughout brain regions. This uniformity, compared to regional thiamine utilization rates, may suggest vulnerable brain regions are depleted of thiamine (at a greater rate than non-vulnerable regions) under conditions of low brain delivery, i.e. low plasma levels secondary to chronic alcoholism.

A current hypothesis with regard to neurodegenerative disease linked to thiamine deficiency is that regional depletion (under conditions of low plasma levels) leads to regional production impairment of mitochondrial tricarboxylic acid cycle thiamine-dependent enzymes. Subsequently, the regional enzymatic decrease diminishes cellular cerebral glucose metabolism in areas of high glucose utilization (vulnerable brain regions) ultimately resulting in selective neuronal degeneration (Gibson and Zhang 2002). This argument can be associated with the data presented in this report in two ways: (i) regionally homogenous blood–brain barrier thiamine influx and efflux constants despite reports of regionally significant differences between thiamine/glucose utilization rates and (ii) the suggestion of rapid brain thiamine depletion as evidenced by significant blood–brain barrier thiamine efflux.

Also of significance in this report is the comparison of endothelial cell association in perfusion experiments between thiamine and choline, both physiologically charged cations. [3H]Choline non-specifically binds to blood–brain barrier endothelia during perfusion distribution experiments, as seen by post-perfusion washouts resulting in an immediate reduction in the brain/perfusion ratio (Allen and Smith 2001). This suggests cationic molecules may non-specifically bind to blood–brain barrier endothelia. However, the time course of [3H]thiamine during our washout study (Fig. 3) revealed an essentially linear decrease of the brain/perfusion ratio, suggestive of brain efflux with minimal endothelial disassociation.

In summary, this research confirms previous literature suggesting blood–brain barrier thiamine uptake is via a carrier-mediated transport mechanism, yet extends the literature by redefining the blood–brain barrier kinetics with a more sensitive methodology. With regard to blood–brain barrier thiamine homeostasis, accumulation was significantly influenced by considerable efflux rates. Further evaluation of the efflux rates demonstrated efflux stimulation in the presence of increased vascular thiamine concentrations. The observation of regionally homogenous thiamine homoeostasis at the blood–brain barrier may be of significant relevance to neurodegenerative disease linked to thiamine deficiency.


The authors gratefully acknowledge the Cancer Biology Research Center at the Texas Tech University Health Sciences Center School of Pharmacy, the American Federation for Aging Research: Glen AFAR Research Scholarship Project and an American Foundation for Pharmaceutical Education predoctoral fellowship for financial support.