|Home | About | Journals | Submit | Contact Us | Français|
Metabolic brain imaging is widely used to evaluate brain function and disease, and quantitative assays require local retention of compounds used to register changes in cellular activity. As labeled metabolites of [1- and 6-14C]glucose are rapidly released in large quantities during brain activation, this study evaluated release of metabolites and proteins through perivascular fluid flow, a pathway that carries solutes from brain to peripheral lymphatic drainage sites. Assays with [3,4-14C]glucose ruled out local oxidation of glucose-derived lactate as a major contributor of label loss. Brief infusion of [1-14C]glucose and -[14C]lactate into the inferior colliculus of conscious rats during acoustic stimulation labeled the meninges, consistent with perivascular clearance of [14C]metabolites from interstitial fluid. Microinfusion of Evans blue albumin and amyloid-β1−40 (Aβ) caused perivascular labeling in the inferior colliculus, labeled the surrounding meninges, and Aβ-labeled-specific blood vessels in the caudate and olfactory bulb and was deposited in cervical lymph nodes. Efflux of extracellular glucose, lactate, and Aβ into perivascular fluid pathways is a normal route for clearance of material from the inferior colliculus that contributes to underestimates of brain energetics. Convergence of ‘watershed' drainage to common pathways may facilitate perivascular amyloid plaque formation and pathway obstruction in Alzheimer's disease.
Labeled glucose and glucose analogs are widely used in autoradiographic, positron emission tomographic, and magnetic resonance spectroscopic studies of brain function, astrocyte–neuron interactions, and neurologic diseases. However, incomplete product trapping because of rapid release from activated tissue of labeled metabolites derived from [1- or 6-14C]glucose (Figure 1A) causes underestimation (≥50%) of functional activation in different brain structures under normal stimulatory conditions compared with parallel studies with [14C]deoxyglucose, which is metabolized mainly to [14C]deoxyglucose-6-phosphate that is trapped intracellularly and not further metabolized by the glycolytic pathway (Sokoloff et al, 1977); undetected losses of labeled glucose metabolites also affect interpretation of results obtained under pathophysiologic conditions, for example, when excitatory activity is not reflected by label accumulation (Collins et al, 1987; Ackermann and Lear, 1989; Adachi et al, 1995; Cruz et al, 1999, 2007). Discordant results obtained with [6-14C]glucose and [14C]deoxyglucose were hypothesized to arise from upregulation of glycolysis, with increased [14C]lactate production and release from activated structures (Figure 1A) (Collins et al, 1987; Ackermann and Lear, 1989; Lear and Ackermann, 1989; Lear, 1990), a concept supported by rapid efflux of [14C]lactate to blood during spreading cortical depression (Adachi et al, 1995; Cruz et al, 1999). Spreading and release of [14C]glucose-derived label during acoustic activation of the inferior colliculus is reduced by inhibition of lactate transporters and astrocytic gap junctions (Cruz et al, 2007), supporting the importance of lactate release and implicating astrocytes in metabolite dispersal. Astrocytic syncytia in the inferior colliculus are comprised of thousands of cells and are capable of long-distance metabolite trafficking (Ball et al, 2007), which is selective with respect to glycolytic intermediates and involves lactate (Gandhi et al, 2009a, 2009b). We hypothesize that labeled lactate and other metabolites are rapidly dispersed from activated cells, in part, through astrocytic syncytia, with release from astrocytic endfeet to perivascular fluid, followed by distribution within brain and release to blood and the lymphatic system (Dienel and Cruz, 2003, 2004, 2008) (Figure 1B).
The fluid space around cerebral blood vessels, such as Virchow-Robin or perivascular space (Figure 1B), is a continuous pathway from arterioles to capillaries to venules. Perivascular fluid communicates with interstitial and cerebrospinal fluids and serves as a conduit for clearance of proteins and other material to arachnoid villi and cervical and spinal lymph nodes (Bradbury et al, 1981; Wagner et al, 1983; Ichimura et al, 1991; Stoodley et al, 1996). In rodents, the arachnoid pathway accounts for about half the clearance of cerebrospinal fluid and solutes from brain, with lymphatic pathways the other half (Koh et al, 2005); lymphatic drainage is considered to be part of the brain's immune system (Cserr et al, 1992). Aortic pulsations power perivascular-mediated spreading of marker molecules, and, within 5 to 10mins, tracer compounds injected into brain distribute widely throughout brain, with slower delivery to cervical and spinal lymph nodes (Wagner et al, 1974; Bradbury and Cserr, 1985; Rennels et al, 1985). Impairment of perivascular fluid flow is likely in patients with cardiovascular disease, diabetes, and Alzheimer's disease, in which basement membrane thickening and amyloid plaque deposits in perivascular space may reduce normal fluid flow and solute trafficking (Carare et al, 2008; Weller et al, 2008).
Our studies of metabolite trafficking and brain imaging use the inferior colliculus as a model structure because (i) it has the highest capillary density and rates of blood flow and glucose usage in brain (Sokoloff et al, 1977; Gross et al, 1987), (ii) it is readily activated by acoustic stimulation of conscious subjects, and (iii) Alzheimer patients have amyloid plaques (Sinha et al, 1993) and reduced cytochrome oxidase levels (Gonzalez-Lima et al, 1997) in the inferior colliculus, as well as abnormal auditory-evoked potentials and latency responses associated with midbrain–brainstem processing of acoustic signals (O'Mahony et al, 1994). As metabolite removal through the perivascular system could influence brain energetics, function, and imaging in normal subjects and Alzheimer's patients, the major objective of this study was to determine whether the meninges and the lymphatic drainage system serve as common routes for release of glucose, lactate, and amyloid-β (Aβ) after their insertion into interstitial fluid of the inferior colliculus in normal adult rats (Figure 1B). We report that these compounds are quickly distributed to meninges and perivascular pathways deliver Aβ to cervical lymph nodes.
-[3,4-14C]glucose (54.5mCi/mmol) and -[1-14C]glucose (54.5mCi/mmol) were purchased from PerkinElmer (Boston, MA, USA), -[1-14C]lactate (55mCi/mmol) from American Radiolabeled Chemicals (St Louis, MO, USA), halothane from Halocarbon Laboratories (River Edge, NJ, USA), pentobarbital from Abbott Laboratories (North Chicago, IL, USA), artificial cerebrospinal fluid (aCSF) from Harvard Apparatus (Holliston, MA, USA), and Evans blue and bovine serum albumin (Cat. No. A9647, molecular weight, ~66kDa) from Sigma (St Louis, MO, USA). Guide cannulae (MD-2250, outer diameter=725μm), stylets, combination microdialysis–microinfusion probes (MD-2262, 2mm-long dialysis membrane containing a microinfusion probe that exits at the tip, outer diameter=320μm), microinfusion probes (MD-2252, 2mm long, outer diameter 105μm, inner diameter 40μm), gas tight syringes, tubing, connectors, syringe drives (MD-1001), and controller (MD-1020) were from Bioanalytical Systems (West Lafayette, IN, USA). Cy5-labeled human Aβ1−40 (Cy5–Aβ) (4,968Da, catalog no. FC5-018-01) was a gift from Dr Jaw Kang Chang, Phoenix Pharmaceuticals (Belmont, CA,USA). Unlabeled human Aβ1−40 (4,329.8Da, product no. A1075) was purchased from Sigma. All Aβ samples were treated with hexafluoro-2-propanol according to the procedure of Klein (2002) to monomerize Aβ; samples were lyophilized and stored at −80°C until the day of the experiment when the peptide was dissolved in phosphate-buffered normal saline (0.1 to 4μg/μL) immediately before use.
Male Wistar Hanover rats (250 to 400g) were purchased from Taconic Farms (Germantown, NY, USA). On the day before the experiment, nonfasted rats were anesthetized with halothane and a guide cannula plugged with a stylet was stereotaxically (David Kopf Instruments, Tujunga, CA, USA) implanted into the inferior colliculus (coordinates in relation to bregma were anterior–posterior, −8.2mm; lateral–medial, −1.7mm; dorsal–ventral, −1.5mm; Paxinos and Watson, 1998) and secured to small screws placed in the skull with dental acrylic. The next day, nonfasted rats were anesthetized and a femoral vein cannula was implanted for intravenous injection of pentobarbital at the end of the experiment; then, the stylet was removed and a microinfusion or combination microinfusion–microdialysis probe was inserted into the guide cannula. The dialysis membrane and microinfusion probe extend into a previously undisturbed tissue a distance 2mm beyond the end of the guide cannula. All rats were restrained by means of a hind limb plaster cast and allowed to recover for 3h; rectal temperature was monitored and maintained at 37°C. Physiologic variables were measured in rats infused with [3,4-14C]glucose and [1-14C]glucose. Mean arterial blood pressure was measured with a calibrated Micro-Med Analyzer (Louisville, KY, USA). Arterial blood was drawn for assay of PCO2, PO2, and pH (CIBA-Corning Model 248 pH/Blood Gas Analyzer; Siemens, Norwood, MA, USA), hematocrit, and plasma glucose and lactate levels (Yellow Springs Instruments, Model 2700 Dual Channel Analyzer; Yellow Springs, OH, USA). All animal use procedures were in strict accordance with the NIH Guide for Care and Use of Laboratory Animals, and were reviewed and approved by the local animal care and use committee.
Three types of experiments were performed in conscious, nonfasted rats approximately 3h after insertion of the infusion or combination probe (Figure 1B), (i) microinfuse [3,4-14C]glucose (0.75μCi/ml) for 40mins and capture extracellular-labeled compounds through microdialysis over a 60min interval, (ii) microinfuse [1-14C]glucose (145μCi/ml) or -[1-14C]lactate (25μCi/ml) for 5mins and analyze regional labeling of grossly dissected brain samples, and (iii) microinfuse or inject Cy5-labeled Aβ, unlabeled Aβ, or Evans blue bound to serum albumin to assess perivascular localization of protein infused for 5mins or up to 90mins. Radiolabeled compounds were dried, dissolved in aCSF, pH 7.4, and microinfused at 0.1μL/min (drainage of interstitial fluid from various rat brain regions ranges from 0.18 to 0.29μL/g/min; Szentistványi et al, 1984). The microinfusion rate of each syringe pump drive was verified after each experiment by counting timed samples of solutions containing known concentrations of 14C-labeled compounds with a Packard Model 2550 liquid scintillation counter (PerkinElmer, Shelton, CT, USA).
In the first experiment, [3,4-14C]glucose was microinfused into the inferior colliculus of five rats implanted with combination probes for a total of 40mins, comprising an initial 20mins period of ‘rest' (quiet environment, no stimulus) followed by 20mins of acoustic stimulation; label was not infused during a 20mins of recovery from stimulation but dialysate samples were collected and analyzed. The acoustic stimulus (broadband click, 40Hz to 8kHz tone at an intensity setting of 95dB) was presented using a Grass Instruments (West Warwick, RI, USA) S10CTCM Click-Tone Module and two 10H2S headphones. Dialysis probes were perfused with aCSF (2.5μL/min) and effluent samples were collected over 10min intervals in NaOH (0.1mol/L) to trap 14CO2. The sample-containing tubes were then placed in closed containers containing hanging wells with filter paper saturated with NaOH (1mol/L), the sample acidified by injecting HCl (5mol/L) through the rubber stopper, the 14CO2 collected overnight, and counted. After removal of 14CO2, labeled acidic compounds were isolated by Dowex-1-formate (Bio-Rad, Hercules, CA, USA) anion exchange column chromatography. Samples were adjusted to pH 7 to 8 and applied to the columns, washed with water to remove [14C]glucose, neutral amino acids, and other neutral or cationic compounds; the retained acidic compounds were eluted with 2mol/L HCl plus 2mol/L NaCl and counted (Cruz et al, 1999). The timed Dowex-1 column effluent fractions from one animal were also analyzed by Dowex-50-H+ cation exchange column chromatography; virtually all of the label was recovered in the neutral, glucose-containing Dowex-50 column effluent fractions. As the amounts of 14C-labeled neutral amino acids (e.g., serine, glycine, or alanine; see Figure 1A) were negligible, fractions were not separated from the other samples.
In the second experiment, -[1-14C]glucose (n=6) or nonmetabolizable -[1-14C]lactate (n=4) was microinfused for 5mins into the inferior colliculus during acoustic stimulation, as described above. Then rats were euthanized (pentobarbital, 200mg/kg, intravenous), their brains rapidly removed, meningeal membranes quickly dissected from dorsal and ventral surfaces of the brain and inferior colliculus, immediately weighed in ice-cold tubes, and frozen in dry ice. Brains were then grossly dissected into major regions from each hemisphere, samples were weighed, dissolved in 1mL of NaOH (1mol/L), and their 14C contents determined. Membrane samples were extracted with 65% ethanol and analyzed as described in detail by Cruz et al (2005). In brief, extracts were lyophilized, dissolved in deionized water, a portion counted, and labeled compounds were separated by anion exchange high pressure liquid chromatography (HPLC) using a Dionex (Sunnyvale, CA, USA) DX-500 chromatography system, an IonPac AS11-HC analytical column, and a NaOH gradient elution program, and 14C contents of timed fractions of the column effluent and column wash determined; recoveries of 14C typically exceed 95%. Anion standards indicated that lactate eluted at ~11 to 13mins, slightly earlier than previously obtained (Cruz et al, 2005). An additional four rats were infused with -[1-14C]glucose as described above, and label distribution within brain was evaluated by autoradiography in 20μm-thick serial coronal sections (Sokoloff et al, 1977).
The third experiment used fluorescent or unlabeled proteins of different molecular weights, that is, Evans blue albumin, Cy5-labeled Aβ1−40, or unlabeled Aβ1−40, to track the perivascular pathways for clearance of material from interstitial fluid of the inferior colliculus. Evans blue is fluorescent when protein bound, and it is completely bound when used in the ratio of ≤1% (wt/vol) Evans blue plus 5% albumin (Ichimura et al, 1991). Groups of conscious (n=4) or euthanized (n=3) rats were microinfused (0.1μL/min) for 5mins with Evans blue albumin dissolved in aCSF, and tracer diffusion was allowed to continue for an additional 5mins; then the rats were euthanized and brains sampled. A third group of conscious rats was infused (0.1μL/mins) for 5mins with Evans blue albumin (n=4) or Cy5–Aβ (n=4) and their brains were sampled immediately. A fourth group was used to visualize transfer of material from the inferior colliculus to cervical lymph nodes and employed longer duration infusions of high doses of the monomerized amyloid peptide, either unlabeled Aβ1−40 (n=10) or Cy5-labeled Aβ1−40 (n=8). Unlabeled Aβ (0.1μg/μL) was microinfused into the inferior colliculus of conscious rats at a rate of 0.1μL/min for 90mins (total dose 0.9μg). Additional rats were anesthetized with halothane, a burr hole was drilled through the skull, and Aβ was stereotaxically injected into the inferior colliculus of anesthetized rats with a Hamilton syringe (Hamilton Company, Reno, NV, USA). Cy5-labeled Aβ (0.25 to 0.5μg/μL) was administered in a volume 2μL (total dose 0.5 or 1μg) over a 2min interval and unlabeled Aβ (2.0 or 4.0μg/μL) was injected in a volume of 2 or 4μL (total dose 4, 8, or 16μg) over a 2 to 7min interval. The syringe was retained in place for 10mins after the injection, then slowly withdrawn, and the burr hole sealed with bone wax; rats were euthanized 60mins after the injection. Control rats (n=3) were not injected. At the end of the experimental interval, rats were euthanized and brain and cervical lymph node tissue samples were obtained. Cy5–Aβ-labeled rats were decapitated and their brains frozen in isopentane at about −40°C. Rats administered unlabeled Aβ were deeply anesthetized and perfused transcardially with ice-cold saline (0.32mol/L) to remove blood from the vasculature followed by flash freezing, or they were perfused with ice-cold 0.1mol/L phosphate-buffered 0.9% saline, pH 7.4 followed by perfusion with 4% paraformaldehyde in phosphate-buffered saline for ~10mins; brains were fixed or post fixed in 4% paraformaldehyde overnight. Fixed brains were cryo-preserved in 30% sucrose at 4°C. Brains were stored at −80°C until serial coronal sections were cut (40μm thick for fixed tissue; 7 or 20μm thick for fresh-frozen tissue) at −20°C using a Leica CM 1850 cryostat.
Cy5-labeled Aβ was directly visualized by fluorescence microscopy. Unlabeled Aβ was detected by either fluorescence immunocytochemistry or the peroxidase antiperoxidase procedure (Sternberger, 1979). Sections were first washed three times in 0.1mol/L phosphate-buffered 0.9% saline, pH 7.4, containing 0.1% Triton-X (PTX). For immunofluorescence, samples processed were incubated for 30mins in 10% goat serum (Dako, Carpinteria, CA, USA) in PTX, then in a primary antibody for 48h at 4°C. The primary antibody was either a polyclonal rabbit antihuman Aβ1−40 (Sigma product no. A8326; diluted 1:250) or mouse monoclonal anti-Aβ (4G8, Signet Laboratories, Dedham, MA, USA, diluted 1:600). The primary antibody was omitted in control samples. Sections were then washed three times, incubated for 1h in either Texas-Red-labeled goat antirabbit or antimouse IgG (Invitrogen/Molecular Probes, Eugene, OR, USA) diluted 1:500 in 10% goat serum-PTX. Sections were washed three times and mounted on slides in Gelmount mounting medium (Biomeda Corp., Foster City, CA, USA). For the peroxidase antiperoxidase method, endogenous peroxidases were first blocked by incubation in a solution containing 50% methanol, 50% PTX, and 0.3% hydrogen peroxide followed by two washes in PTX, blocking in 10% normal swine serum (Dako) in PTX, and overnight incubation in polyclonal rabbit antihuman Aβ1−40 (Sigma, diluted 1:250) at 4°C; primary antibody was omitted in control samples. After three washes, samples were incubated for 1h in swine antirabbit IgG (Dako) diluted 1:500 in 10% swine serum, washed three times, and incubated for at least 45mins in rabbit peroxidase antiperoxidase (Dako) diluted 1:100 in phosphate-buffered saline. Sections were washed three times with PTX, incubated in the dark for 8 to 10mins in a 3, 3′-diaminobenzidine solution prepared from Sigma fast diaminobenzidine peroxidase substrate tablet set (Sigma) according to manufacturer's instructions, and mounted on slides with Permount (Fisher, Fair Lawn, NJ, USA). Sections were viewed with a Zeiss Axioskop 2 microscope (Carl Zeiss, Germany), imaged with an MTI CCD 72 camera (Dage-MTI, Michigan City, IN, USA), and processed with MCID Imaging Software (InterFocus Imaging Ltd., Cambridge, UK).
Our previous imaging studies with [1- and 6-14C]glucose during acoustic stimulation implicated [14C]lactate release as a prominent factor in label loss from activated inferior colliculus (Cruz et al, 2007) but did not rule out rapid release of 14CO2 because of local lactate oxidation in small, active metabolic compartments that do not mix with the large unlabeled amino acid pools; dilution of glucose-derived label into tricarboxylic acid (TCA) cycle-derived amino acid pools causes label trapping and is the basis for assays of glucose oxidation rates (Figure 1A). The possibility of rapid lactate oxidation was tested by microinfusion of tracer quantities of [3,4-14C]glucose into the inferior colliculus (Figure 1B) and analysis of relative levels of labeled CO2 and acidic metabolites in microdialysate. 14CO2 is released from [3,4-14C]glucose at the pyruvate dehydrogenase step (this eliminates labeling of TCA cycle compounds, Figure 1A), and if lactate were generated from glucose and quickly, locally oxidized, the quantity of 14CO2 in extracellular fluid would be similar to that of [14C]lactate. [3,4-14C]Glucose was infused into rats for 40mins during rest and acoustic stimulation and dialysate samples were collected over a 60min interval (Figure 2A). Physiologic variables of 14C-glucose-infused rats were similar during rest and acoustic activation (Table 1). Label was recovered in the dialysate after a lag from infusion onset and dialysate 14C was quite variable from animal to animal; mean values between 20 and 40mins ranged from about 25,000 to 35,000dpm, then fell after the infusion was stopped, indicating clearance of extracellular-labeled compounds (Figure 2A). After fractionation of dialysate samples, most 14C was recovered in the [14C]glucose fraction and the quantity in acidic compounds (mainly lactate) was three to four-fold greater than that of 14CO2 (Figure 2B). Thus, local oxidation of glucose-derived lactate in small compartments is not a major contributor to label loss from activated tissue.
Next, spreading of labeled compounds from inferior colliculus to other brain regions was examined by microinfusion of 14C-labeled glucose or -lactate (Figure 1B) for 5mins (i.e., the duration of typical autoradiographic [14C]glucose usage assays) and determination of label distribution in grossly dissected structures. Less than half of the infused label derived from [1-14C]glucose (40.9±9.7% n=6) was recovered in brain tissue, presumably because of efflux from brain of labeled glucose through glucose transporters (and perivascular flow), which reflects the outbound aspect of concentration gradient-driven, bidirectional metabolite trafficking across the blood–brain barrier (Figure 1B). Of the 14C recovered in brain tissue, 97.2±3.2% was retained in the infused inferior colliculus. However, when total meninges was taken into account, a high but variable fraction (34±46%) of the recovered 14C was in these membranes, about 60% was in the ipsilateral inferior colliculus, a few percent in the ipsilateral superior colliculus, occipital cortex, and cerebellum (Figure 3A), and very low but detectable levels in other ipsilateral structures, with even lower levels in contralateral structures (not shown in Figure 3A). On a weight basis, the dpm/g recovered in meninges was 2.3 times of that in the inferior colliculus.
The regional distribution of nonmetabolizable -[14C]lactate delivered by microinfusion was also assessed, and recovery of 14C in brain was even lower than for [14C]glucose, that is, 18.9±7.9% (n=4) of the infused amount. Thus, washout of extracellular -lactate is faster than -glucose, presumably because of its lack of metabolism and it is being a poor transporter substrate. The ipsilateral inferior and superior colliculi accounted for about 70 and 17% of the recovered -[14C]lactate, respectively, with about 3% in the meninges, occipital cortex, cerebellum, and midbrain (Figure 3B). Of the meningeal labeling, about 75% was in membranes surrounding the ipsilateral inferior colliculus, 7 to 10% in meninges from the cerebellum and contralateral inferior colliculus, and smaller amounts in membranes from the ventral surface of the brain. On a weight basis, the dpm/g recovered in meninges above the inferior colliculus was 21 and 82% of that in the inferior and superior colliculus, respectively. Samples of arterial blood drawn at the end of the experimental intervals contained only a few dpm above background, which was negligible compared with total counts in the membranes (Figures 3A and 3B). Thus, label in meninges was derived primarily from the inferior colliculus, not from blood contained within the vasculature in the meninges. Presumably, the [14C]tracer released from brain to blood became highly diluted and was probably also taken up into body tissues.
HPLC analysis of membrane extracts from the [14C]glucose-infused rats revealed five major fractions that contained 85±9% of the total 14C (Figure 3C). Glucose was shown earlier to elute in fraction 1 (Cruz et al, 2005), which accounted for almost 40% of the label, and the lactate fraction contained ~18% of the label (Figure 3D). Compounds in the other fractions were not identified in this study, but we showed earlier that glutamine eluted at about 6mins (fraction 2), carbonate at about 25mins (fraction 4), and more strongly retained anions (e.g., glucose-6-phosphate) would be eluted in the column wash (fraction 5) (Cruz et al, 2005). One sample had quite low (5%) labeling of the glucose fraction, and fraction 5 of this and one other sample had ~50% of the total 14C, contrasting the 3 to 7% in other samples. Some labeled metabolites probably arose from meningeal metabolism of [14C]glucose during and after the 5mins in vivo assay procedure. For technical reasons (difficulty of dissecting meninges from funnel-frozen brain and no access to a microwave fixation unit) we were unable to eliminate postmortem changes, but all 14C in the meninges originated in the infused inferior colliculus.
Autoradiographic assays of brain at 5mins after microinfusion of [1-14C]glucose into the inferior colliculus during acoustic stimulation of conscious rats showed that most label was localized within the inferior colliculus with labeling of the meningeal membranes around the colliculus (n=4, data not shown), consistent with the results for dissected tissue (Figure 3) and our previous studies with microinfusion of 14C-labeled compounds (Cruz et al, 2007). In contrast, microinfusion of [1-14C]glucose into a lateral ventricle of one rat caused widespread periventricular labeling throughout brain, as reported for [14C]sucrose (Ghersi-Egea et al, 1996). Although some leakage of label from the infusion site along the microinfusion probe track to meninges and cerebrospinal fluid is likely to occur, the results support the conclusion that labeled lactate generated from glucose in the inferior colliculus is released to the surrounding meninges and that label spread to other brain regions through ventricular cerebrospinal fluid is low.
As 14C labeling does not have high spatial resolution and labeled glucose metabolites may diffuse and obscure local distribution patterns, pathways for efflux from a point source in the inferior colliculus to the meninges were visualized with fluorescent or unlabeled proteins (Figure 1B). Microinfusion of Evans blue bound to albumin (molecular weight ~66kDa) for 5mins caused blue labeling of meningeal membranes around the inferior colliculus (Figures 4A and 4B), as well as perivascular labeling of blood vessels at the base of the brain (circle of Willis), under the olfactory bulbs, and along the middle cerebral artery that was more pronounced with longer infusion times (not shown). Coronal brain sections show labeling by Aβ (Figure 4C) and Evans blue albumin (Figures 4D to 4I) that is localized mainly within the inferior colliculus (Figures 4C to 4E). Perivascular labeling of blood vessels is visible in brightfield images (Figures 4D, 4F, and 4H), but Evans blue fluorescence shows a greater label distribution and higher labeling along blood vessels that extend from the inferior colliculus to the meninges and from meninges into cortex (Figures 4E, 4G, and 4I).
Pulsatile blood flow may help remove Evans blue albumin from perivascular fluid and adjacent tissue, thereby limiting its accumulation and detection. Protein-bound dye was, therefore, also microinfused for 5mins into the inferior colliculus of euthanized rats. Elimination of blood flow enhanced dye labeling within the inferior colliculus, particularly in tissue surrounding the vasculature and meninges. For example, the path of a blood vessel from the ventral region of the infused tissue (Figure 5A) to meninges (Figure 5B) is readily visible, there is greater dye penetration into the tissue along major blood vessels (Figures 5C and 5D), and spreading of dye from a vessel along the tissue–meningeal border is more readily detected in euthanized compared with conscious rats (compare Figures 5C to 5F with Figure 4). Together, these findings show that material in the interstitial fluid in the central, ventral, and lateral regions of the inferior colliculus can travel to the meninges along many blood vessels independent of any dye leakage along the cannula track. Some dye localized along the cannula track may have been drawn upward from the infusion site when the infusion probe and guide cannula were removed.
Exogenous Aβ inserted into the inferior colliculus (Figure 1B) labels the meninges as well as perivascular pathways to the cervical lymph nodes (Figures 6 and and7).7). Microinfusion of Cy5–Aβ (4,968Da), a protein smaller than albumin (~66kDa), into inferior colliculus of conscious rats for 5mins caused perivascular labeling of vessels that linked the central (Figures 6A to 6C) and dorsal (Figures 6D and 6E) zones of the inferior colliculus with the meningeal margin at sites distant from the infusion guide cannula track. When Cy5–Aβ was manually injected into the inferior colliculus and allowed to diffuse for 60mins, label spread throughout the entire structure, with high perivascular localization that is evident in cross-section views of blood vessels, particularly in the meninges (arrows, Figure 6F). Longitudinal views of vessels (Figures 6f and 6G) show diffuse labeling along the length of vessels and the fine structure of small vessels (arrow, Figure 6G). In more anterior sections of brain at the level of the caudate (Figure 7A), specific midline blood vessels located dorsal to the optic chiasm exhibited high perivascular labeling at 90mins. In addition, there was a low unilateral background red fluorescence in tissue near the anterior commissure ipsilateral to the infused inferior colliculus (Figures 7C, 7E, and 7e). Selective labeling by Cy5–Aβ was detected around vessels within meninges between frontal cortex and dorsal aspect of the olfactory bulbs and under the olfactory bulbs (Figures 7B, 7D, and 7F). Cervical lymph node labeling was heterogeneous (Figures 7G and 7H).
To rule out the possibility that the fluorescent Cy5–Aβ labeling pattern may have arisen, in part, from release of Cy5 from Aβ, separate groups of conscious rats were infused or injected with unlabeled Aβ and the protein was detected by immunohistochemistry. All of these assays yielded results (not shown) similar to those for Cy5–Aβ, that is, (i) highest labeling in the inferior colliculus near the injection site and in the meninges surrounding the midbrain and cerebral cortex, (ii) high, selective perivascular labeling of large and small blood vessels in serial coronal brain sections rostral to the inferior colliculus with very low staining of the brain tissue, and (iii) heterogeneous labeling of cervical lymph nodes. The major difference between Cy5–Aβ and unlabeled Aβ was background labeling. The Cy5-background fluorescence was very low, whereas background for immunolabeling was much higher, particularly in formalin-fixed compared with fresh-frozen tissue. Blood in the vasculature also gave a high background that was eliminated by transcardial perfusion; vascular luminal staining was absent in Cy5-labeled rats. Omission of the primary antibody eliminated perivascular and meningeal immunolabeling.
Our previous studies of acoustic activation of the inferior colliculus in conscious rats (Cruz et al, 2007) showed that local glucose usage rates (CMRglc) in tonotopic activation bands calculated from accumulated levels of labeled metabolites of [1- and 6-14C]glucose are underestimated by 43 to 65% because of (i) decarboxylation reactions in the pentose phosphate shunt and other pathways (Figure 1A), (ii) metabolite dispersal that is mediated, in part, by lactate transporters and gap junctions and includes spread through extracellular fluid and the astrocyte syncytium (Figure 1B), and (iii) rapid glucose-derived label release from activated tissue by unidentified routes. The major findings of this study extend this work by (i) ruling out substantial lactate oxidation and loss of 14CO2 in a small, active compartment, (ii) demonstrating that 14C-labeled glucose and -lactate infused into extracellular fluid during acoustic stimulation quickly spread to the meninges, and (iii) identifying perivascular pathways traversed by protein markers from the central zone of the inferior colliculus to meninges, other brain regions, and cervical lymph nodes. Discharge of Aβ from the inferior colliculus through lymphatic drainage pathways links diseases that may cause impairment of perivascular flow to brain energetics and imaging.
Lactate is assigned as the predominant acidic metabolite of [3,4-14C]glucose in extracellular fluid (Figure 2) based on intracellular trapping of phosphorylated glycolytic intermediates, the high lactate/pyruvate ratio in brain (~13; Siesjö, 1978), and elimination of labeling of TCA cycle compounds, glutamate, glutamine, and aspartate because of release of 14CO2 from pyruvate before its entry into the TCA cycle (Figure 1A). Efflux of [14C]lactate from cells to extracellular fluid (Figure 2) accompanies the rise in extracellular unlabeled lactate during acoustic stimulation (Cruz et al, 2007). As extracellular [14C]lactate levels exceed 14CO2 by a factor of 3 to 4 (Figure 2), most lactate released to extracellular fluid is not taken up into nearby cells and oxidized. Thus, lactate is not a major fuel during activation as proposed in the lactate shuttle model (Pellerin et al, 2007); instead, lactate is released, contributing to low registration of metabolic activation by [1- and 6-14C]glucose (Cruz et al, 2007; Figure 1). Transfer of labeled glucose and lactate from inferior colliculus interstitial fluid to meninges (Figure 3) implicates rapid removal of labeled metabolites from activated tissue, in part, through the egress routes revealed by Evans blue albumin and Aβ (Figures 4, ,5,5, ,6,6, ,7).7). These results are consistent with the pathways for penetration of Ruthenium Red through the glia limitans in cerebral cortex, that is, diffusion through intercellular clefts and passage along penetrating blood vessels (Wagner et al, 1983). The role of aortic pulsations in tracer distribution within brain (Rennels et al, 1985) is underscored by greater spreading of labeled protein into perivascular interstitial space in the absence of blood flow (Figure 5). The quantitative importance of the meningeal pathway for metabolite trafficking is inferred from its surprisingly high labeling by [14C]glucose (i.e., more than half of the 14C recovered in the inferior colliculus and twice that of inferior colliculus on a per gram tissue basis) and its more modest labeling by -[14C]lactate (3% of total label, 21% by weight) (Figure 3). Note that these values are only ‘static' measures that reflect relative 14C contents at tissue sampling; flow rates must be determined to evaluate flux through this pathway. The lower percentage recovery of -[14C]lactate in meninges compared with [14C]glucose and glucose-derived [14C]lactate probably reflects its lack of metabolism and faster washout.
Lactate has an important role in release of glucose-derived label from activated tissue because it has a high specific activity (half that of glucose; Dienel and Cruz, 2009) and is readily diffusible and transportable. During spreading cortical depression, the mean cortical lactate level rose to 8.5μmol/g and calculated CMRglc was underestimated by 45% when determined with [6-14C]glucose (Adachi et al, 1995). Efflux of [14C]lactate to blood sampled at the superior sagittal sinus was rapid, equivalent to 20% of net glucose uptake into brain, and accounted for about half of the CMRglc underestimate (Cruz et al, 1999). These findings suggest that perivascular delivery of labeled metabolites to blood downstream of the sampling site may account for the other half, that is, the ‘missing' 14C-metabolites that correspond to ~22% of CMRglc during spreading depression. During acoustic activation, lactate level rose to ~2μmol/g and the overall rate of transporter-mediated lactate efflux from brain across the blood–brain barrier would be lower than during spreading depression (MCT1 Km ~3.5mmol/L; Manning Fox et al, 2000). As underestimation of CMRglc by [6-14C]glucose was in the range of 45 to 65% in both studies, perivascular clearance of 14C during acoustic stimulation may exceed that during spreading depression. Perivascular flow-mediated metabolite release would not only interfere with brain imaging and spectroscopic studies, but also with global metabolic assays determined by arteriovenous difference because efflux of material from brain to sites downstream of the usual venous sampling locations (e.g., sagittal sinus or jugular vein) would cause overestimation of net uptake into brain and underestimation of net release from brain. Currently, quantification of the fraction of metabolite release from most brain regions through venous blood and perivascular fluid is not possible because of inaccessibility and complexity of drainage systems.
Drainage of cerebrospinal, interstitial, and perivascular fluid from brain involves fluid release through the arachnoid villi and flow to the peripheral lymphatic system through routes common to many species; about half of the labeled tracers inserted into CSF or interstitial fluid are cleared through perivascular fluid and can be recovered in blood and in cervical and spinal lymph nodes (Bradbury and Cserr, 1985; Cserr et al, 1992; Johnston et al, 2004; Koh et al, 2005). Four properties of perivascular flow systems are relevant to metabolite release during brain activation. Bradbury and Westrop, (1983) showed the following: (i) efflux of perivascular fluid along the olfactory vasculature traverses through the cribriform plate and at this point the tight-junction cerebral vessels become fenestrated so that molecules <5,000Da (e.g., [14C]lactate) are released to the circulation, whereas larger molecules are delivered to the cervical lymph nodes and can be quantified; (ii) labeled albumin passes through the system more quickly when CSF flow and CSF pressure are increased (e.g., during intense exercise); (iii) postural alternations of head position affect the rate and direction of CSF drainage to the cervical compared with spinal cord lymphatic systems. In addition, (iv) perivascular flow is driven by arterial pulsations, with reduced tracer distribution after placement of an aortic cuff (Rennels et al, 1985; Hadaczek et al, 2006). The rate of bulk flow of perivascular fluid calculated from clearance of compounds from interstitial fluid in the rat, cat, and rabbit is very low (0.2 to 0.3μL/g/mins; Bradbury and Cserr, 1985) compared with cerebral blood flow (~1mL/g/mins). However, bulk flow carrying molecules restricted to extracellular space does not provide direct information about rates of trafficking of transportable, metabolizable compounds.
Proteins and membrane-impermeant tracers are restricted to travel within tortuous extracellular space, whereas transportable metabolites can enter and exit cells. Trafficking may be retarded or blocked by metabolism, but compounds taken up by or generated in astrocytes can diffuse within the syncytium to endfeet. Astrocytes exhibit selective gap junctional trafficking of glycolytic intermediates, and have a two to four-fold faster and higher capacity for lactate uptake from extracellular fluid and for lactate dispersal through the astrocytic syncytium compared with neuronal lactate uptake from extracellular fluid or shuttling of lactate to neurons from neighboring astrocytes (Gandhi et al, 2009a, 2009b). Astrocytes in the inferior colliculus are gap junction coupled to thousands of other astrocytes and quickly transfer Lucifer yellow dye to the colliculus-meninges boundary and to their endfeet (Ball et al, 2007); gap junction-coupled astrocytic endfeet form an extensive perivascular surface area containing various transporters. Our working model for metabolite clearance (Figure 1B) is that transportable compounds are taken up from interstitial fluid by astrocytes, dispersed within the extensive syncytium, and released through endfeet to perivascular fluid from which they can be transported directly into blood; once in perivascular fluid, metabolites and proteins can also be removed through lymphatic drainage pathways. Glucose and lactate diffusion and carrier-mediated transport are driven by local concentration gradients that govern the direction and magnitude of local fluxes. During brain activation with high glycolytic rates, local accumulation of lactate can be minimized by its rapid dilution into a large syncytial volume and discharge to perivascular space for clearance from tissue. Owing to anatomic, metabolic, and signaling relationships of astrocytes with each other, neurons, and the vasculature, astrocytes are poised for metabolite trafficking.
Little is known about how the perivascular fluid flow system responds to brain activation and how it is affected by neurodegenerative and cerebrovascular diseases. The small number of drainage pathways rostral to the inferior colliculus that are highly labeled by exogenous Cy5-labeled and unlabeled Aβ (Figure 7) suggests that continuous removal of endogenous Aβ from interstitial fluid derived from a substantial tissue volume (Figure 6) converges to common perivascular pathways, thereby delivering material, regardless of cellular origin, from a large ‘watershed' tissue volume to the perivascular space of specific vessels. In fact, proteins and other compounds injected into the caudate nucleus, midbrain, internal capsule, and cerebrospinal fluid also drain to the cervical lymph nodes through the Circle of Willis and cribriform plate; of considerable importance is the finding that the concentration of tracer in the Circle of Willis is typically 15 to 40-fold higher than that in CSF, suggesting slow mixing or compartmentation of interstitial and cerebrospinal fluids (Bradbury and Cserr, 1985). Accumulation of amyloid deposits around blood vessels in Alzheimer's disease led to the hypothesis that perivascular fluid flow is a normal clearance route for Aβ and that obstruction of perivascular pathways by amyloid deposits contributes to the pathophysiology of Alzheimer's disease (Carare et al, 2008; Weller et al, 2008). Our results support this hypothesis and indicate that perivascular amyloid deposits need not arise from local production of Aβ. Progressive accumulation of amyloid deposits in Alzheimer's disease and thickening of basement membranes in cardiovascular disease and diabetes could impair vascular elasticity, reduce the impact of arterial pulsations on bulk perivascular fluid flow, and influence the distribution within brain and clearance from brain of glucose, lactate, and other metabolites, and signaling molecules. Metabolite trafficking within and from brain is not well understood, but it has important roles in brain energetics and imaging, it contributes to underestimates of CMRglc determined with labeled glucose, and it is likely to be affected by pathophysiologic involvement of the perivascular space and lymphatic drainage systems; new approaches must be developed to measure metabolite fluxes in vivo.
This work was supported in part by grants from the Alzheimer's Association (IIRG-06-26022), NIH (NS36728, NS047546), and a pilot study award from the College of Medicine, University of Arkansas for Medical Sciences. The content of this work is solely the responsibility of the authors and does not necessarily represent the official views of The National Institute of Neurological Disorders and Stroke or The National Institutes of Health.
Disclosure/conflict of interest
The authors declare no conflict of interest.