Blood-brain barrier permeability in response to caffeine challenge

Abstract

Purpose:

Caffeine is known to alter brain perfusion by acting as an adenosine antagonist, but its effect on blood-brain barrier (BBB) permeability is not fully elucidated. This study aimed to dynamically monitor BBB permeability to water after a single dose of caffeine tablet using a non-contrast MRI technique.

Methods:

Ten young healthy volunteers who were not regular coffee drinkers were studied. The experiment began with a pre-caffeine measurement, followed by four measurements at the post-caffeine stage. Water-extraction-with-phase-contrast-arterial-spin-tagging (WEPCAST) MRI was used to assess the time dependence of BBB permeability to water following the ingestion of 200mg caffeine. Other cerebral physiological parameters including cerebral blood flow (CBF), venous oxygenation (Y_v), and cerebral metabolic rate of oxygen (CMRO₂) were also examined. The relationships between cerebral physiological parameters and time were studied with mixed-effect models.

Results:

It was found that after caffeine ingestion, CBF and Y_v showed a time-dependent decrease (p<0.001), while CMRO₂ did not change significantly. The fraction of arterial water crossing the BBB (E) showed a significant increase (p<0.001). In contrast, the permeability-surface-area product (PS), i.e. BBB permeability to water, remained constant (p=0.94). Additionally, it was observed that changes in physiological parameters were non-linear with regards to time and occurred at as early as 9 minutes after caffeine tablet ingestion.

Conclusion:

These results suggest an unchanged BBB permeability despite alterations in perfusion during a vasoconstrictive caffeine challenge.

Introduction

Caffeine, as the most widely used neurostimulant, is known to have a multifaceted effect on the human brain, especially through its role as an antagonist of adenosine (1). By binding to the adenosine receptors (mainly A₁, A_2A and A_2B) of the neurons, caffeine can block the inhibitory effect of adenosine on synaptic vesicle release and in turn increase neural activity and improve vigilance (2). On the other hand, caffeine can also constrict cerebral blood vessels by binding to the A_2A and A_2B receptors on the cerebrovascular smooth muscle, which leads to a profound change in brain hemodynamic parameters (3,4). Previous studies have reported that caffeine can cause a reduction in cerebral blood flow (CBF) and venous oxygenation (Y_v) (4–9). Through its vascular effects, caffeine has also been shown to alter temporal dynamics of BOLD response and lead to a widespread decrease in functional connectivity (10–14).

Blood-brain barrier (BBB), formed by endothelial tight junctions, astrocyte and pericytes, is a selective barrier that regulates the blood-brain exchange of substances and maintains constant microenvironment (15). All types of adenosine receptors were demonstrated to be expressed on brain capillary endothelial cells, but whether and how the adenosine receptor activation regulates BBB permeability remain unknown (16–18). As a broad-spectrum adenosine antagonist, it is useful to understand the impact of caffeine ingestion on BBB. Several prior studies have investigated the chronic effect of caffeine uptake (19,20) on BBB but not its acute effect. Furthermore, the previous studies were performed in animal models and focused on BBB permeability to large molecules such as Evan’s blue dye (19,20). From a normal physiological point-of-view, it is also interesting to examine whether BBB permeability can remain stable when brain perfusion is altered by factors in daily life including breathing pattern, perspiration, or consumption of beverage containing vasoactive substances such as caffeine.

In humans, BBB permeability to gadolinium (21) can be measured with contrast agent based MRI methods, such as dynamic contrast-enhanced (DCE). Recently, there has been an increasing interest in examining the BBB permeability to water, a small molecule that may be more sensitive to subtle changes in BBB. Several non-contrast MRI techniques have been proposed by separating intravascular and extravascular spins based on their difference in spatial location, T₁, T₂ or diffusion properties (22–29). Water-extraction-with-phase-contrast-arterial-spin-tagging (WEPCAST) MRI is one of such techniques which aims to measure global water extraction fraction (E) and BBB permeability-surface-area product (PS) by selectively measuring arterial spin labeling (ASL) signals in the venous system (24,25). This method can provide an estimation of BBB permeability at around 5 min with a good test-retest reproducibility (24). Therefore, in this study, we aimed to utilize WEPCAST MRI to investigate dynamic changes in BBB permeability following a single-dose of caffeine ingestion in young healthy adults.

Methods

Participants

Ten young healthy volunteers were recruited (age 29.1±9.3 years old, 6 females and 4 males). The participants were carefully screened and did not report any health issues. None of the participants were regular coffee drinkers. The subjects reported an average of 1.1 cups of caffeine-containing drink per week. All experiments were performed on a 3T Siemens MRI system (Prisma, Siemens Healthcare, Erlangen, Germany). The experimental procedures in this study have been approved by the Institutional Review Board of the Johns Hopkins University School of Medicine. Each participant gave informed written consent before participating in the study.

Experimental Procedures

The subject was instructed to avoid any caffeine uptake for 4 hours prior to the study. The study design was: 1) baseline measurement of WEPCAST, 4 phase contrast (PC), T₂-Relaxation-Under-Spin-Tagging (TRUST) (10 min) was performed; 2) participant was taken out of the scanner, sat up on the table, ingested a 200mg caffeine tablet (equivalent to 2 cups of regular coffee) and was quickly put back into the scanner; 3) WEPCAST/PC/TRUST measurement was performed immediately after caffeine ingestion (10 min); 4) T1-weighted MPRAGE structural scan (4 min); 5) Three further repetitions of WEPCAST/PC/TRUST (10 min each). A diagram of these procedures is shown in Supporting Information Figure S1.

MRI protocols

WEPCAST MRI was performed to estimate water extraction fraction and BBB permeability index, PS. Details of WEPCAST MRI can be found in Lin et al (24,25). Briefly, a pseudo-continuous ASL module is used to label the incoming arterial blood at the cervical region. A fraction of the labeled blood is exchanged into tissue at capillary-tissue interface through BBB while the non-extracted water is drained into the venous system. By selectively measuring the ASL signal in the main draining veins of the brain, e.g. superior sagittal sinus (SSS), using a phase-contrast-encoded acquisition, a global water extraction fraction, E, can be estimated. Together with CBF (f) measurement, PS can be obtained (30):

$$PS= -\ln \left(1-E\right)\cdot f.$$ [1]

WEPCAST MRI was conducted in mid-sagittal plane with a labeling duration (τ) of 4000ms and a post-labeling delay (PLD) of 3000ms. Other imaging parameters were as follows: single-shot gradient echo planar imaging (EPI) readout, field of view (FOV) = 200×200mm², single slice, matrix = 64×64, voxel size = 3.13×3.13mm², slice thickness = 10mm, GRAPPA factor = 3, flip angle = 90°, repetition time (TR) = 9200ms, echo time (TE) = 9.5ms, encoding velocity (V_enc) = 20cm/s, number of control/label pairs = 10 and scan duration = 6min54s. An additional M₀ image with same TE and V_enc and a long TR = 10s was also acquired for normalization.

Global CBF was estimated with PC MRI. PC MRI was performed at four major feeding arteries (left/right internal carotid arteries and left/right vertebral arteries) to quantify global CBF. The following parameters were used (31): TR=16.0ms, TE=10.2ms, flip angle=15°, FOV=200×200×5mm³, voxel size=0.5×0.5×5mm³, single slice, V_enc=40cm/s and scan duration=13s.

We measured global venous oxygenation, Y_v, using T₂-Relaxation-Under-Spin-Tagging (TRUST) MRI (32,33). The following parameters were used: TR = 3000 ms, TE = 3.9 ms, TI=1022 ms, flip angle = 90°, FOV = 220 × 220 × 5 mm³, voxel size = 3.4 × 3.4 × 5 mm³, four effective TEs (eTE = 0, 40, 80, and 160 ms) with a τ_CPMG of 10 ms, labeling thickness = 100 mm, and scan duration = 1min27s.

A 3D T₁-weighted magnetization-prepared-rapid-acquisition-of-gradient-echo (MPRAGE) scan was acquired with the following parameters: TR=8.8ms, TE=3.8ms, shot interval=2100ms, inversion time (TI)=1100ms, flip angle=12°, FOV=208×256×160mm³, voxel size=1×1×1mm³, number of slices=160, sagittal orientation, and scan duration=4min3s.

Data Analysis

All MRI data were processed using in-house MATLAB (version R2016a, MathWorks, Natick, MA) scripts. Details of WEPCAST processing can be found in Lin et al. (24). Briefly, pairwise subtraction of the velocity-encoded images for control and label conditions yields arterially labeled venous signal (∆M):

$$\Delta M=2\alpha \left(1-E\right)M_0{e}^{-\frac{{\delta }_v}{T_{1b}}},$$ [2]

where E is water extraction fraction, α is labeling efficiency (assume to be 86%) (34), M₀ is the equilibrium magnetization and was measured from the M₀ scan mentioned above, T_1b is venous blood T₁ (assumed to be 1584ms) (35), δ_v is the bolus arrival time to the vein and, for the location where ΔM/M₀ reaches the peak, it is $PLD+\frac{\tau }{2}$.

For PC-MRI, region of interest (ROI) was drawn manually on the complex difference images to trace the targeted arteries. Integration of the velocity, after accounting for phase foldover, within the ROI yielded flow in the units of mL/min. The summation of flux across all feeding arteries yielded total blood flow. The MPRAGE images were segmented using an automatic processing tool, MRICloud (www.MRICloud.org, Johns Hopkins University, MD) for total brain volume quantification (36). Then the global CBF was calculated as total blood flow divided by total brain volume. BBB permeability to water, i.e. PS value, was then obtained using Eq. [1].

For TRUST MRI, subtraction between control and label images yields pure labeled blood signal at SSS, which was mono-exponentially fitted as a function of eTEs to generate venous blood T₂. Blood T₂ was then converted to Y_v using an established calibration curve (37). Cerebral metabolic rate of oxygen (CMRO₂) was also calculated based on Fick’s principle (38):

$$CMR{O}_2=CBF\cdot \left(Y_a-Y_v\right)\cdot C_h,$$ [3]

where Y_a is arterial oxygenation (assumed to be 98%), C_h represents the oxygen carrying capability of hemoglobin and calculated as C_h=20.4*hematocrit based on previous literature. Here the hematocrit value was assumed to be 42% for male and 40% for female (39).

The time dependence of the physiological parameters (i.e. E, PS, CBF, Y_v and CMRO₂) was studied with a mixed-effect model. The model started with a linear term of time to examine whether the parameter changed with time. A quadratic term, time², was added to the model to investigate whether time-change was non-linear. Additionally, we compared each post-caffeine time point to the value during baseline with paired t-tests. A Bonferroni corrected p<0.05 was considered statistically significant.

Results

Figure 1 shows representative WEPCAST control, label and difference images at baseline and difference images at different time points after caffeine ingestion. As can be seen, due to the phase-contrast flow-encoded acquisition scheme, tissue signals are successfully suppressed, and vessel signals are selectively measured. At baseline, prominent signal can be seen at the SSS, representing the labeled water spins that were not extracted by the tissue and drained directly to the venous system. After caffeine uptake, WEPCAST signal decreased gradually with time, indicating that a larger fraction of water was extracted by the tissue. Peak signal along the SSS was then used to calculate water extraction fraction. Representative images of other physiological MRI sequences, PC MRI and TRUST MRI, are shown in Supporting Information Figure S2 and S3.

Figure 1:

Representative WEPCAST dataset in a volunteer. (A) Control, label and difference WEPCAST images at baseline. (B) WEPCAST difference images as a function of time after caffeine ingestion.

Time courses of the water extraction fraction, BBB permeability and other brain physiological parameters were shown in Figure 2. Table 1 summarizes results from linear mixed model analyses. When studying the linear relationship between physiological parameters and time, it was found that caffeine ingestion had a time-dependent effect Y_v, CBF, and E, but not on CMRO₂ or PS. Upon further adding a quadratic term, time², to the model, Y_v, CBF, and E, but not CMRO₂ or PS, showed a significant effect (Table 1), suggesting that these physiological changes are non-linear with time. At the last time point of our measurements (approximately 45 minutes after caffeine ingestion), Y_v was found to decrease by 19%, CBF decreased by 29%, E increased by 6%. Results of paired t-tests between post-caffeine and baseline (i.e. pre-caffeine) values for physiological parameters are shown in Table 2. It can be seen that changes in physiological parameters, i.e. E, CBF and Y_v, occurred as early as 9 minutes after caffeine tablet ingestion.

Figure 2:

Time course of cerebral blood flow (CBF), venous oxygenation (Y_v), cerebral metabolic rate of oxygen (CMRO₂), water extraction fraction (E), and BBB permeability-surface-area product (PS). t=0 indicates baseline measurement (i.e. before caffeine tablet ingestion). Other time points are labeled with the measurement start time (t=9min, 25min, 35min and 45min) after caffeine ingestion. Error bars denote standard errors across participants.

Table 1:

Summary of linear mixed effect model results for BBB permeability and cerebral physiology

	coefficient	std. error	p-values		coefficient	std. error	p-values
E
Time	0.0011	0.00018	<0.0001	Time	0.0026	0.00065	0.0003
				Time2	−0.000034	0.000014	0.024
PS
Time	0.0097	0.12	0.93	Time	0.044	0.46	0.93
				Time2	−0.00077	0.010	0.94
CBF
Time	−0.33	0.047	<0.0001	Time	−0.78	0.16	<0.0001
				Time2	0.010	0.0036	0.0076
Y v
Time	−0.26	0.030	<0.0001	Time	−0.60	0.10	<0.0001
				Time2	0.0076	0.0022	0.0016
CMRO 2
Time	−0.18	0.093	0.058	Time	−0.23	0.36	0.52
				Time2	0.0011	0.0077	0.89

Table 2:

Summary of BBB permeability and cerebral physiology before and after caffeine uptake

	E (%)		PS		CBF		Yv		CMRO2
	Mean±SD	p-value*	Mean±SD	p-value*	Mean±SD	p-value*	Mean±SD	p-value*	Mean±SD	p-value*
Baseline	91.0±3.5		133.6±16.9		54.9±10.1		63.8±3.9		155.0±27.1
Time (t) after caffeine ingestion
t=9min	93.5±3.2	0.026	128.6±14.0	0.77	45.9±7.8	0.078	57.4±6.5	0.021	151.5±13.3	0.98
t=25min	95.5±2.9	0.0090	136.9±18.3	0.99	42.2±7.1	0.018	53.7±7.7	0.0024	151.5±12.6	0.97
t=35min	96.1±2.3	0.0039	131.0±22.7	0.99	38.7±6.3	0.0043	51.6±7.9	0.00053	146.0±14.3	0.75
t=45min	96.2±3.0	0.018	132.3±21.0	1.00	39.0±7.4	0.0035	51.6±7.6	0.00043	147.0±15.5	0.74

^*p-values from paired t-test compared with baseline (Bonferroni corrected)

Discussion

Caffeine has a known effect on the neurovascular system, but its impact on the blood-brain barrier remains unclear. In this study, we dynamically measured the effect of caffeine on the BBB permeability to a small molecule, water, using a novel non-contrast technique, WEPCAST MRI. Our findings suggested that water extraction across the BBB increased after caffeine uptake. On the other hand, BBB permeability to water remained unchanged. We also found that CBF and venous oxygenation reduced after caffeine ingestion, consistent with its known vasoconstriction effect.

BBB prevents the entry of pathogens and neurotoxins into the brain parenchyma, thus an intact BBB function is crucial for the health of the nervous system (15). Abnormalities of BBB has been implicated in a number of brain diseases, such as Alzheimer’s disease (40), Parkinson’s disease (41) and multiple sclerosis (42). In order to evaluate the feasibility of BBB permeability as a disease-specific marker, it is important to examine, in the healthy brain, whether BBB permeability may change with physiological maneuvers that are known to alter hemodynamics. In the present report, we used caffeine, a substance contained in many common beverages, as a physiological challenge and observed its expected effects on CBF and venous oxygenation. However, we found that BBB permeability to water remained unchanged, indicating that the permeability of a healthy BBB is relatively stable. We did observe that the amount of water molecules that were extracted across BBB, E, was increased. Although we cannot rule out other possible factors such as change in neural activity, we postulate that the increase in E was primarily attributed to a reduced CBF (see Eq. [1]), thereby longer exchange time (43,44).

One implication of our findings is that patients or participants receiving BBB MRI do not need to refrain from caffeine consumption before the exam. In many physiological MRI exams such as CBF, CBV, venous oxygenation, cerebrovascular reactivity imaging, it is often necessary to instruct the subject not to take caffeine beverages several hours before the scan in order to avoid potential biases. It appears that this is not needed for BBB MRI. Furthermore, our findings suggest that CBF can alter without corresponding changes in BBB permeability. Technically, this characteristic may be exploited for improving the reliability of BBB MRI measurement. For example, it may be possible to use hypercapnia or acetazolamide to increase CBF and enhance the BBB MRI signal, without biasing the accuracy of the PS estimation.

In view of the intricate role that BBB plays in cell signaling systems and cerebrovasculature, some animal studies have investigated the potential effect of caffeine on BBB in the context of its therapeutic effect in neurodegenerative diseases (19,20,28,45,46). For example, Chen et al. suggested that chronic ingestion of caffeine can protect against BBB damage, including mitigating the leakage to Evan’s blue dye and the decrease in tight junction protein level, in animal models of Alzheimer’s disease and Parkinson’s disease (19,20). It has also been suggested that caffeine can help with amyloid-β clearance across the BBB (46). It was hypothesized that caffeine can control BBB permeability through blockade of adenosine receptors, inhibition of cAMP phosphodiesterase activity, and by mediating the calcium release from intracellular spaces (45). Hurtado-Alvarado et al. reported a recovery of BBB permeability to dextrans and Evan’s blue after A_2A receptor antagonist administration in rat model of sleep deprivation (47). However, our study is different from these previous reports in several ways. First, our study mainly focused on the acute effect of caffeine challenge and all our participants were non-regular coffee drinkers, while the prior studies focused on chronic caffeine consumption. Acute ingestion of caffeine can inhibit intracellular levels of cAMP, while chronic exposure to caffeine can result in the up-regulation of adenosine receptors on endothelial cells, the activation of which can increase the cAMP level and protect BBB against disruption (45,48,49). Second, the previous studies measured the BBB permeability to large molecules, which may be different from the permeability to water. Third, the prior reports were conducted on conditions where BBB was already leaky due to disease while our study examined normal BBB function in response to caffeine ingestion. Finally, the earlier studies were all performed in animals whereas the present report was conducted in humans.

Recently there has been a surging interest in the measurement of BBB permeability to water using MRI and some have studied the effect of caffeine challenge (25–28,50,51). For example, Wengler et al. proposed an intrinsic diffusivity based method and reported a decrease of BBB permeability to water after caffeine challenge (28). However, it should be pointed out that the method of Wengler et al. is sensitive to both tight junction integrity in endothelia and abundance of aquaporin-4 (AQP4) on the astrocyte endfeet, since the water spins have to reach the interstitial space to experience a major reduction in apparent diffusion coefficient (that is, apparent diffusion coefficient in perivascular space is quite large). The WEPCAST technique used in the present study, on the other hand, measures ASL signal in large veins, thus can differentiate perivascular signal from vessel signal. Therefore, different MRI methods on BBB permeability to water may be probing different aspects of the water exchange process. Note that there are some reports that caffeine may regulate the expression and polarity of AQP4 and may explain the observations reported in Wengler et al. (52,53).

The current study has several limitations. First, although WEPCAST MRI can estimate BBB permeability to water efficiently, it is a global method and cannot provide regional information of BBB. Previous study showed that CBF decreased at different rate in different brain regions after caffeine ingestion (5), suggesting that the caffeine effect can be region-dependent. Further technical development on measuring regional BBB permeability can provide a better understanding of the spatial dependence of the caffeine effects. Second, we did not measure the caffeine concentration in blood and different people may react differently to caffeine. However, the measurement of CBF and Y_v can be an indicator of the caffeine effect, and we did observe significant decrease for all our participants.

Conclusion

In summary, using a non-contrast technique, WEPCAST MRI, we investigated the effect of caffeine ingestion on BBB permeability to water. Our findings suggested that despite a pronounced alteration in blood flow and oxygenation, BBB water permeability remains unchanged after caffeine uptake, indicating a stable BBB function in the presence of hemodynamic changes.

Supplementary Material

supinfo

Figure S1: A diagram illustrating the experimental procedure of the study.

Figure S2: Representative PC MRI dataset in a volunteer. (A) PC images of four feeding arteries (LICA, RICA, LVA and RVA) at baseline. (B) PC images of four feeding arteries as a function of time after caffeine ingestion.

Figure S3: Representative TRUST MRI dataset in a volunteer. (A) TRUST images (control, label and difference images at different eTEs) at baseline. (B) TRUST difference images at different eTEs as a function of time after caffeine ingestion.