By continuing to browse this site you agree to us using cookies as described in About Cookies
Wiley Online Library is migrating to a new platform powered by Atypon, the leading provider of scholarly publishing platforms. The new Wiley Online Library will be migrated over the weekend of February 24 & 25 and will be live on February 26, 2018. For more information, please visit our migration page: http://www.wileyactual.com/WOLMigration/
Roxana O. Carare, Anatomical Sciences, School of Medicine, University of Southampton, Boldrewood, Bassett Crescent East, Southampton, SO16 7PX, UK. E-mail: firstname.lastname@example.org; Tel: +44-2380594461, Fax: +44-2380594433
Elimination of interstitial fluid and solutes plays a role in homeostasis in the brain, but the pathways are unclear. Previous work suggests that interstitial fluid drains along the walls of arteries. Aims: to define the pathways within the walls of capillaries and arteries for drainage of fluid and solutes out of the brain. Methods: Fluorescent soluble tracers, dextran (3 kDa) and ovalbumin (40 kDa), and particulate fluospheres (0.02 μm and 1.0 μm in diameter) were injected into the corpus striatum of mice. Brains were examined from 5 min to 7 days by immunocytochemistry and confocal microscopy. Results: soluble tracers initially spread diffusely through brain parenchyma and then drain out of the brain along basement membranes of capillaries and arteries. Some tracer is taken up by vascular smooth muscle cells and by perivascular macrophages. No perivascular drainage was observed when dextran was injected into mouse brains following cardiac arrest. Fluospheres expand perivascular spaces between vessel walls and surrounding brain, are ingested by perivascular macrophages but do not appear to leave the brain even following an inflammatory challenge with lipopolysaccharide or kainate. Conclusions: capillary and artery basement membranes act as ‘lymphatics of the brain’ for drainage of fluid and solutes; such drainage appears to require continued cardiac output as it ceases following cardiac arrest. This drainage pathway does not permit migration of cells from brain parenchyma to the periphery. Amyloid-β is deposited in basement membrane drainage pathways in cerebral amyloid angiopathy, and may impede elimination of amyloid-β and interstitial fluid from the brain in Alzheimer's disease. Soluble antigens, but not cells, drain from the brain by perivascular pathways. This atypical pattern of drainage may contribute to partial immune privilege of the brain and play a role in neuroimmunological diseases such as multiple sclerosis.
Extracellular fluid associated with the brain consists of blood, cerebrospinal fluid (CSF) and interstitial fluid (ISF) . In the human brain, it is estimated that there is 280 ml of ISF, and that CSF comprises 140 ml of which 30 ml are in the ventricles and 110 ml in the subarachnoid spaces . In humans, CSF is produced by the choroid plexus in the ventricles, flows into the subarachnoid spaces and drains into the blood largely through arachnoid villi and granulations within the walls of venous sinuses . In rodents and sheep, a large proportion of the CSF drains via the cribriform plate and nasal lymphatics to cervical lymph nodes [4–6].
The pathways by which ISF and solutes are eliminated from the brain are less well characterized than CSF drainage pathways. Interstitial fluid is generated from the blood and from metabolic activity within brain tissue at an estimated rate of 0.1–0.3 μl/min/g in the rat . Following the injection of soluble radio-iodinated albumin tracer into the brains of rats, radioactivity was detected in the walls of leptomeningeal arteries and in cervical lymph nodes . Such observations suggest that ISF drains out of the brain along blood vessel walls.
Although the elimination of solutes from the brain is associated with blood vessel walls, the drainage route within the vessel wall has not been fully elucidated. There is some indication from the distribution of amyloid-β (Aβ) in blood vessel walls in cerebral amyloid angiopathy (CAA) that the drainage pathway may be along the basement membranes around capillaries and in the tunica media of arteries. Amyloid-β in vessel walls in CAA in elderly humans, in Alzheimer's disease and in transgenic mouse models of Alzheimer's disease [8–12] is deposited in the basement membranes of leptomeningeal and cortical arteries and less frequently in capillary basement membranes [8–10,13]. The Aβ that is overproduced in transgenic mice is derived solely from the brain , so the presence of Aβ in the basement membranes of cerebral arteries is a strong indication that it is draining out of the brain along artery walls.
In this study, we test the hypothesis that solutes in the ISF are eliminated from the brain along basement membranes in the walls of capillaries and arteries. Defining the pathways for the drainage of ISF and solutes from the brain may allow them to be manipulated in the future for the treatment and management of Alzheimer's disease and neuroimmunological disorders such as multiple sclerosis.
In order to test the hypothesis, soluble lysine-fixable fluorescein labelled 3-kDa dextran and FITC-labelled ovalbumin (40 kDa) were injected separately into the grey matter of the corpus striatum in mouse brains. The distribution of the tracers was defined at 5 min, 30 min, 3 h and 24 h. Dextran is an established tracer used for lymphatic transport studies, and the dextran tracer used in this study is similar in molecular weight to Aβ in its monomeric soluble form .
Previous studies to define perivascular drainage pathways have used particulate matter such as Indian ink as a tracer . We therefore injected fluospheres 1.0 μm and 0.02 μm in diameter, into mouse striata to compare the distribution and potential drainage pathways for fluospheres with the distribution and drainage of soluble tracers. As it has been shown in peripheral tissues that local inflammation may enhance the migration of antigen-presenting cells into the lympatics , we co-injected inflammatory agents, lipopolysaccharide (LPS) and kainic acid, which provoke inflammation in the brain with or without accompanying neuronal degeneration [17,18].
Materials and methods
Injection of tracers into the brain
Intracerebral injections of tracers were performed on a total of 112 male 6-to 8-week-old MF1 mice (in-house breeding colony). Mice were anaesthetized by intraperitoneal injection of 0.1 ml per 5 g body weight Avertin (2,2,2 tribromoethanol in tertiary amyl alcohol). The scalp was shaved and local anaesthetic (Lignocaine 5% from Biorex Laboratories Ltd, Enfield, UK) was placed in the external auditory meati as mice were positioned in a stereotaxic frame (Kopf Instruments, Tujunga, CA, USA). All tracers were injected stereotactically into the striatum (bregma 1 mm anterior, lateral + 1.5 mm, 2.5 mm deep) or lateral ventricle (bregma 1 mm posterior, lateral + 1.5 mm, 1.5 mm deep).
Critical to the analysis of drainage pathways, for either soluble or particulate matter, from the brain is that the volume of the injected material and the speed at which it is delivered must be taken into account . It has previously been shown that small volumes of less than 1 μl of heat-killed bacteria  or virus  can be micro-injected into the parenchyma of the rodent brain without evoking an immune response. In contrast, injection of the same volume into the ventricles leads to a rapid immune reaction indicating that the antigens or particulate microorganisms rapidly find their way to the periphery. We used these observations to guide our preliminary studies in which we injected small volumes of tracers into the corpus striatum without involving spread of the tracers to the ventricles (data not shown). We chose in all subsequent studies to inject a volume no larger than 0.5 μl over a period of no less than 2 min. Leaving the pipette in situ for a further 2 min prevented the reflux of the dye or fluospheres along the injection track.
Biotin and fluorescein labelled 3-kDa soluble lysine fixable dextran (Microemerald) and FITC-labelled ovalbumin (Invitrogen, UK) were injected into the grey matter of the corpus striatum (caudate putamen) in the mouse brain (Table 1) at a concentration of 1 μg/μl in a volume of 0.5 μl over a period of 2 min through a glass micropipette with the injecting tip <50 μm (Sigma); this technique results in minimal damage at the injection site . After each injection, the micropipette was left in situ for 2 min to minimize reflux back along the injection tract. The striatum was selected for the site of injection as pilot experiments showed significant leakage of tracer into the ventricles when it was injected into other sites such as the hippocampus.
Table 1. Distribution of dextran and ovalbumin following injection into the mouse striatum related to time after injection
Tracer (n = number of mice)
Location of tracer
Dextran (n = 8)
Diffuse in brain parenchyma
Basement membranes of capillaries
Basement membranes of arteries
Smooth muscle cells
Dextran (n = 3)
Diffuse in brain parenchyma
Basement membranes of arteries
Dextran (n = 13)
Dextran (n = 5)
Dextran (n = 5)
Diffuse in brain at site of injection
Ovalbumin (n = 5)
Diffuse in brain parenchyma
Basement membranes of capillaries
Basement membranes of arteries
Ovalbumin (n = 3)
Diffuse in brain parenchyma
Basement membranes of arteries
Ovalbumin (n = 5)
Microemerald, with a fluorescence absorption maximum at 494 nm and a fluorescence emission maximum at 521 nm (Molecular Probes, Oregon, USA), had been selected as the dextran tracer from the results of pilot experiments comparing three labelled dextran tracers at different concentrations and amounts. A concentration of 1 μg/μl Microemerald in a volume of 0.5 μl proved to be optimal.
Injections of dextran [0.5 μl of Microemerald (1 μg/μl)] into a lateral ventricle was performed in nine mice in order to determine whether the tracer would be detectable in cervical lymph nodes.
Dextran was injected into the striatum of five dead mice at the onset of cardiac arrest induced by terminal anaesthesia with sodium pentobarbitone (250–300 μl intraperitoneally) in order to assess whether distribution of tracer within the brain parenchyma was affected by vital forces, such as vascular pulsations.
Fluospheres of 1-μm and 0.02-μm diameter (fluorescence excitation wave length 580 nm, fluorescence emission 605 nm) were injected as a 2% aqueous suspension (Molecular Probes, Eugene, OR, USA) in a volume of 0.5 μl into the grey matter of the striatum in a total of 24 mice. Prior to the injection, sodium azide was removed by dialysis. Subsequent histological examination showed that tissue damage was confined to the injection site, with minimal inflammation. Fluospheres of 1-μm diameter were injected through a glass micropipette (Sigma, St Louis, MO); the tip of the capillary pipette was <50 μm. Fluospheres of 0.02-μm diameter were injected using a Hamilton syringe; the needle tip measured 330 μm. Results from pilot experiments indicated that fluospheres of 0.02-μm diameter were not able to pass through the glass micropipette, possibly due to the formation of aggregates, but were injected successfully through a Hamilton syringe.
Salmonella abortus equi LPS (Sigma) was used to induce acute focal inflammation . Fluospheres of 1 μm together with LPS (1 μg) were injected into the striatum (n = 10). One nanomolar kainic acid (Sigma) was injected into the striatum together with 1-μm fluospheres (n = 10), in order to induce a brain excitotoxic lesion .
Fluospheres were also injected into one lateral ventricle as a 2% aqueous suspension in a volume of 0.5 μl in a total of 21 mice.
Following surgery, mice were placed in a heated recovery chamber at 37°C and re-housed in groups. All procedures were carried out under UK Home Office Licence and in accordance with the Animals (Scientific Procedures) Act, 1986.
Preparation of tissue
Mice were killed at various time intervals between 5 min and 7 days by terminal anaesthesia with sodium pentobarbitone (250–300 μl intraperitoneally) and transcardially perfused with heparinized 0.9% saline followed by 4% paraformaldehyde in 0.1 M phosphate buffer pH 7.4. Brains were removed and further fixed by immersion in 4% paraformaldehyde for 4–6 h and placed in 30% sucrose for 48 h for cryoprotection. The brains were trimmed to form coronal blocks 7 mm thick with the injection site in the centre. Blocks were then frozen in Tissue Tek OCT (Sakura Finetek Europe B. V., Zoeterwoude, the Netherlands) and sectioned in a coronal plane (10 μm thick) on a cryostat (Leica 17–20). Every tenth section was collected on a gel-coated slide for histological examination, quantification studies and immunocytochemistry. Brains from dead animals that had been injected with dextran following cardiac arrest were prepared in the same way as the mice that were injected while alive.
Two deep cervical lymph nodes and blocks of spleen and liver were taken from all animals after perfusion. Ten serial cryostat sections from each deep cervical lymph node and 16 sections from each liver and spleen were collected on gelatine-coated slides, air-dried and examined under a fluorescence microscope for the presence of fluorescent soluble tracer or fluospheres.
Immunocytochemistry and Lectin staining
Serial cryostat sections of brain, 10 μm thick, were stained by immunocytochemistry to identify tissue elements in the brains following injection of tracers. Vascular endothelial cells were identified by Isolectin IB4 (Molecular Probes, Eugene, OR, USA) at a dilution of 1:100. The laminin component of vascular basement membranes was located by immunocytochemistry using a pan-laminin: polyclonal antibody (Sigma) – 1:500 dilution; smooth muscle cells of the tunica media of arteries were identified by α-smooth muscle actin: monoclonal antibody (Sigma) – 1:4000 dilution; activated macrophages by F4/80 monoclonal antibody (Serotec -MCA 497) – 1:500 dilution, and HB1 (University Department of Pharmacology, Oxford) – 1:20 dilution. Astrocytes were identified by glial fibrillary acidic protein (GFAP) polyclonal antibody (DAKO, Glostrup, Denmark) – 1:400 dilution; and neurones by NeuN (monoclonal antibody – Chemicon International, Temecula, CA, USA), 1:100 dilution. Cell reactions in acute inflammation were evaluated by immunocytochemistry using MBS1 (in-house antibody ) to assess neutrophil recruitment and co-localization of fluospheres with neutrophils.
For visualization under a fluorescence microscope, secondary antibodies or streptavidin (in the cases where avidin–biotin complex was used) were labelled with Alexa Fluor 488 fluorochrome (Molecular Probes). The immunocytochemistry was controlled by omitting the primary antibody; all controls were negative. Mounted sections were coverslipped using Mowiol.
Microscopy of the brains and imaging were performed using a Zeiss Axioskop attached to a PC-running Imaging Associates image analysis software. Dual localization of tracer and immunocytochemical staining was achieved using a Leica SP2 confocal laser scanning microscope. The co-localization programme compares the labelling intensity from the two data sets, voxel by voxel; if there is no labelling in either corresponding voxel, it appears black; labelling in one but not the other voxel appears green or red. Whenever a high intensity of labelling in both channels (indicating co-localization) is detected, the programme converts the point to a chosen colour (for images in this study, blue or yellow). To minimize false positives due to fluorescence detected in both channels (‘bleed through’), sequential rather than simultaneous acquisition of data sets for each fluorochrome was obtained.
Following transcardial perfusion with paraformaldehyde, small blocks of brain adjacent to the injection sites of 0.02-and 1.0-μm fluospheres were immersed in 3% glutaraldehyde in 0.1 M cacodylate buffer plus 2 mM CaCl2 at pH 7.2 for 1 h. The specimens were then rinsed in 0.1 M cacodylate buffer plus 0.23 M sucrose and 2 mM CaCl2 at pH 7.2, post-fixed in 2% osmium tetroxide in 0.1 M cacodylate buffer for 1 h, rinsed in buffer, block-stained with 2% aqueous uranyl actetate, dehydrated and embedded in Spurr resin. Semithin and Ultrathin sections were cut on a Leica OMU 3 ultramicrotome, stained with toluidine blue or Reynolds lead stain, and viewed by light microscopy or in a Philips 201 transmission electron microscope.
Quantification and statistical analysis
Assessment of the distribution of dextran, ovalbumin and fluospheres in a coronal plane in the brain was performed using a Zeiss Axioskop attached to a PC equipped with Imaging Associates image analysis software (KS-400), 1300 × 1030 format and images were obtained at ×10 magnification. The radial extensions of dextran, ovalbumin and fluospheres from the site of injection were calculated. Five values were obtained from each animal at each time point. Statistical analysis was performed using Statview. As the data are normally distributed, anova analysis of variance was performed and Bonferroni adjustment was applied. Results with P < 0.01 were considered to be significant.
Classification of blood vessels
Blood vessels were classified according to the diameter of each vessel and by the presence or absence of smooth muscle cells in the vessel wall, as revealed by immunocytochemistry for α-smooth muscle actin as follows:
Capillaries Blood vessels less than 10 μm in diameter that lacked smooth muscle cells in their walls.
Arteries and arterioles Blood vessels larger than 10 μm in diameter with smooth muscle cells in the wall depicting the tunica media. Most intracerebral arteries are arterioles as they lack internal elastic laminae.
Veins Blood vessels larger than 10 μm in diameter that lacked smooth muscle cells in their walls. Venules are smaller and may be difficult to distinguish from capillaries unless they are surrounded by inflammatory cells.
Distribution of dextran tracer in the brain (Table 1)
Five minutes after injection Five minutes after injection, soluble dextran had spread diffusely through the brain parenchyma and was located within the walls of blood vessels (Figure 1A). The maximum diffuse spread of dextran at this time was up to 2500 μm in the antero-posterior plane from the injection site in the striatum as estimated by the number of serial sections containing tracer.
Dextran was located within capillary walls at 5 min after injection, and was co-localized with laminin in the basement membranes (Figure 1B). Dextran in artery walls was also co-localized with laminin in the basement membranes surrounding smooth muscle cells in the tunica media as depicted by a dark blue line in Figure 1C. Basement membranes of the endothelium and those on the outer aspect of the artery wall contained no dextran at this time, and thus stained red for laminin alone. Dextran was diffusely distributed in the cytoplasm of selected smooth muscle cells (Figure 1C). No tracer was seen in the walls of veins.
Scattered neurones and astrocytes in the brain parenchyma contained dextran at 5 min after injection; these cells were identified by co-localization of dextran and NeuN and GFAP staining, respectively (data not shown).
Thirty minutes after injection At this time point, fluorescent dextran was more widespread, but was still located in the interstitial spaces of the brain parenchyma as well as within basement membranes of arteries, but not in the walls of capillaries. Dextran was also detected within perivascular macrophages around capillaries and arteries.
Three and twenty-four hours after injection Dextran was no longer detectable in the extracelluar spaces of the striatum or in the basement membranes of capillaries and artery walls by 3 and 24 h. However, dextran was present in a punctate form in artery walls within the striatum, cerebral cortex and leptomeninges (Figure 1D–F). Immunocytochemistry for macrophage markers showed that the punctate deposits of dextran were associated with perivascular macrophages. (Figures 1D,F).
Distribution of dextran following intraventricular injection A small amount of dextran in a punctate form was seen in the ependyma of the lateral ventricle 3 and 24 h after injection. Despite extensive analysis of sections from peripheral tissues, no dextran was detectable within cervical lymph nodes, liver or spleen.
Dextran injections in dead animals Following injection into the striatum of dead mice, dextran was diffusely spread in the brain parenchyma a short distance from the injection site, but there was no selective accumulation of dextran in the basement membranes in the walls of capillaries or arteries.
Quantification of the distribution of dextran The distribution of dextran in relation to the site of injection differed with time, especially between 5 min and 3 h. By 5 min, dextran had spread diffusely in the brain parenchyma and was within the basement membranes of capillary and artery walls. By 3 h and 24 h, dextran was present in a punctate form within perivascular macrophages associated with capillaries and arteries. The tracer was significantly more widespread within the brain and leptomeninges at 3 h and 24 h when compared with 5 min after injection. At 5 min, there was a mean radius of dextran distribution of 785 μm (SD ± 129 μm) in the coronal plane, and dextran was not observed in the walls of leptomeningeal arteries. By 3 h, the mean had nearly doubled and was 1207 μm (SD ± 101 μm) and by 24 h, was 1323 μm (SD ± 26 μm). At both these time points, dextran was observed as punctate deposits in the walls of parenchymal and leptomeningeal arteries. The anova statistical analysis reveals a significant difference (P < 0.0001) in the radius of spread of dextran in the coronal plane between 5 min and 3–24 h. This suggests that the dextran had drained out of the brain from the site of injection to reach the leptomeningeal arteries. Although dextran could not be detected in the basement membranes of the leptomeningeal arteries at 5 or 30 min, it was detectable within the perivascular macrophages associated with the walls of leptomeningeal arteries at 3 and 24 h.
Distribution of ovalbumin tracer in the brain
The distribution of ovalbumin (40 kDa) after injection into grey matter of the mouse brain was very similar to that observed with fluorescent 3-kDa dextran (Table 1).
Five minutes after injection At 5 min after injection into the striatum, ovalbumin was diffusely distributed in the extracellular spaces of the brain parenchyma, and was also in the basement membranes in the walls of capillaries and arteries (Figure 2A). Co-localization of ovalbumin with laminin in the basement membranes of capillary and artery walls is depicted in yellow in Figure 2A.
Thirty minutes after injection Ovalbumin was still diffusely spread in the brain parenchyma at the site of injection at 30 min, and was co-localized with laminin in the walls of arteries; it was not detectable in capillary basement membranes.
Twenty-four hours after injection Ovalbumin was no longer detectable in the extracellular spaces of the brain or in the basement membranes of the capillary and artery walls by 24 h after injection. However, ovalbumin was detected as punctate deposits within cells in the walls of arteries within brain parenchyma and in the leptomeninges (Figure 2B,C). From their position in relation to the vessel walls and their positive staining with macrophage markers F4/80 and HB1, the cells containing ovalbumin were identified as perivascular macrophages.
Quantification of the distribution of ovalbumin Detectable ovalbumin was not as widespread as dextran, and showed no significant difference in the radius of spread between 5 min and 24 h. Ovalbumin had a mean maximum spread in the coronal plane of 617 μm (SD ± 532) at 5 min and 685 μm (SD ± 552) at 24 h.
Distribution of fluospheres in the brain
Fluospheres injected into the striatum remained mainly as a mass near the injection site, but did also extend along blood vessel walls away from the injection site. There was no significant difference in distribution of fluospheres between 5 min and 7 days, suggesting that the spread may have occurred at the time of injection. By 24 h, fluospheres had been taken up by cells of the macrophage lineage either in the perivascular spaces or in the brain parenchyma.
Location of fluospheres associated with blood vessel walls Following injection into the striatum, fluospheres spread along capillaries and arteries between the outer aspect of the vessel wall and the basement membrane of the glia limitans of the surrounding brain parenchyma. In Figure 2D, groups of 0.02-μm fluospheres are between vessel walls and surrounding brain, and have focally extended into the brain parenchyma. Fluorospheres around capillaries appear to have split the basement membrane formed by the endothelium and glia limitans. The fluospheres around arteries are located on the inner aspect of the glial basement membrane, and have not entered the tunica media. By 24 h, fluospheres had been phagocytosed by perivascular macrophages that accumulated in the perivascular spaces located between the outer membrane of the vessel wall and the basement membrane of the glia limitans (Figure 2E,F). The fluospheres separated the vascular and glial basement membrane components that are normally fused.
Quantification of spread of fluospheres in the coronal plane The maximum distance of spread of 0.02-μm and 1.0-μm fluospheres was 800–1000 μm from the site of injection with no significant difference between the two sizes of fluosphere.
Injection of 1.0-μm fluospheres and lipopolysaccharide or kainic acid into the striatum
Injection of LPS or kainic acid into the brain parenchyma results in a local inflammatory response with recruitment of polymorphonuclear leucocytes and monocyte/macrophages. Fluospheres of 1.0-μm diameter were co-injected with LPS into the striatum to test the hypotheses that the presence of acute inflammation would modify the distribution of fluospheres and that cells ingesting them would migrate out of the brain.
Five minutes after co-injection with LPS, fluospheres remained as a compact mass at the injection site, similar to the distribution 24 h after injection of fluospheres alone. However, by 24 h after coinjection with kainic acid or LPS, fluospheres were in aggregated masses at the injection sites and in groups of 2–10 in MBS1 and HB1 positively labelled cells in perivascular spaces (polymorphonuclear leucocytes and macrophages, respectively).
Quantification of spread of fluospheres in the coronal plane Spread of fluospheres within the brain was significantly greater following co-injection with kainic acid or LPS than in those animals injected with fluospheres alone. The distance that 1.0-μm fluospheres had spread in a coronal plane by 24 h after injection was 1184 μm (SD ± 64) for kainic acid co-injected animals, and 1224 μm (SD ± 42) by 24 h following co-injection of LPS. Both these measurements had a P-value of < 0.005 when compared with injection of 1.0-μm fluospheres alone (800 μm, SD ± 52).
Drainage of fluospheres from lateral ventricle to lymph nodes, liver and spleen Fluospheres, 0.02 μm and 1.0 μm in diameter, were detected in the cervical lymph nodes, liver and spleen following injection into the lateral ventricles. However, examination of large numbers of sections from cervical lymph nodes, liver and spleen revealed no fluospheres in these peripheral tissues following injection of fluospheres into the striatum, either with or without the co-injection of LPS or kainic acid.
The results of the present study suggest that the solutes, 3-kDa dextran and 40-kDa ovalbumin, drain from the brain in two phases; initially by diffusion through the extracellular spaces and then by perivascular drainage along the basement membranes of capillaries and arteries. Figure 3 summarizes the pattern of drainage of solutes observed. Tracers co-localized with laminin in the basement membranes of capillaries and arteries within 5 min of injection. As shown particularly well following injection of dextran, tracer was located in the basement membranes between the smooth muscle cells in the tunica media of arteries, but not in the basement membranes related to arterial endothelium or to the outer aspects of the tunica media. No tracer was observed in the basement membranes of leptomeningeal arteries in the present study. However, it seems likely that the concentration of tracer falls as it is diluted by ISF from neighbouring parts of the brain until the tracer in the basement membranes of leptomeningeal arteries is too dilute to be detectable. Nevertheless, dextran and ovalbumin were detected in perivascular macrophages by 3 and 24 h, not only around arteries in the brain parenchyma but also around arteries in the leptomeninges. This suggests first, that some tracer had passed radially through the artery walls to reach the periphery of the vessel and is taken up by perivascular macrophages, and second, that tracers had reached the leptomeningeal arteries on their passage out of the brain before being taken up by perivascular macrophages.
Diffusion of solutes in brain parenchyma
Diffusion of solutes through the narrow extracellular spaces of grey matter in the brain is considered to be an important communication system between cells in the brain . Such diffusion depends upon a number of defined factors. The extracellular spaces are tortuous, and the rate of diffusion of solutes is related to the size and shape of the molecules, neuronal activity, pathological changes and the location in the brain [24,25]. In the present study, the slow injection of 0.5 μl over a period of 2 min resulted in a pool of tracer in the extracellular spaces in the striatum that took 30 min or more to clear. This suggests that there may be some restriction of the capacity to clear solutes from the extracellular spaces into the bulk flow channels of vascular basement membranes.
Perivascular drainage of solutes
Experimental studies have shown that tracers are cleared from brain parenchyma at the same rate and independent of molecular weight; this suggests that bulk (convection) flow is a major component of the clearance system . In the present study, tracers of two different molecular weights, namely 3-kDa dextran and 4-kDa ovalbumin, appeared to be cleared by the same bulk flow pathway.
The motive force for the drainage of solutes and fluid from the brain along artery walls in the opposite direction to the flow of blood may be related to the pulsations of arteries . Dextran tracer injected into the brains of mice following cardiac arrest in the present study did not selectively accumulate in the walls of blood vessels as it did in live animals. This again suggests that perivascular drainage is not by simple diffusion and may depend upon vascular pulsations. Mathematical models  suggest that the contrary (or reflection) wave that follows each main pulse wave, travels in the reverse direction to the pulse way. It is proposed that the contrary wave drives the solutes and ISF out of the brain along artery walls . However, the model suggests that transport of solutes in the reverse direction to the arterial pulse will only occur if some form of attachment of solutes or a valve-like action is present . The early co-localization of the tracers with laminin in the basement membrane indicates that the basement membrane may act as a potential site of attachment, or perhaps even as a filter as molecules enter perivascular bulk flow pathways.
Amyloid-β acts as a natural tracer for perivascular drainage pathways
The distribution of Aβ within the basement membranes of intracerebral capillaries and arteries, and in leptomeningeal arteies in CAA is identical to that of dextran and ovalbumin observed in the present study [8–10,13]. As with dextran, basement membranes related to the endothelium of arteries and the outer aspect of the media do not appear to contain Aβ in the early stages of CAA [13,27]. This could have implications for the preservation of vascular endothelia in CAA.
Ultrastructural studies of the early stages of deposition of Aβ in the walls of leptomeningeal arteries have shown that fibrils of insoluble Aβ accumulate in the lamina densa in the centre of the basement membranes between the smooth muscle cells . This suggests that the lamina densa may be the major conduit for perivascular drainage of solutes from the brain. Smooth muscle cells within cerebral artery walls produce Aβ in culture, and these cells may also contribute to the accumulation of Aβ in the walls of arteries in CAA [28–30].
Amyloid-β is deposited not only in artery walls within the brain, but also in the walls of leptomeningeal arteries extending to the base of the brain [31,32]. Biochemical analyses of human cerebral arteries show increasing amounts of Aβ in the walls of cerebral arteries with age and in Alzheimer's disease . One key observation in this biochemical study  was that levels of Aβ in the walls of internal carotid arteries in the neck were very low, despite the significant Aβ load in the walls of middle cerebral arteries at the base of the brain. Perivascular drainage appears to end at the base of the skull, but the fate of the solutes is not known. It is unclear whether solutes are dissipated within the perivascular tissues at the base of the skull, or whether they drain to deep cervical lymph nodes associated with the carotid sheath .
Relevance of perivascular drainage to neuroimmunology
Lymphatic drainage has been shown to play a role in immunological reactions in the brain . Radioactively labelled proteins drain from the striatum to lymph nodes within hours and appear to pass along blood vessel walls . Injection of antigens into the rat striatum results in the formation of antibodies in cervical lymph nodes, indicating a major role for lymphatic pathways in B lymphocyte-mediated immune reactions in the brain . Removal of cervical lymph nodes significantly reduces T lymphocyte-mediated autoimmune reactions in the rat brain , and there is some indication that lymphocytes from cervical lymph nodes target the brain .
Despite the physiological significance of lymphatic drainage of the brain, the complete anatomical route for drainage of antigens from brain parenchyma to cervical lymph nodes is unclear both in experimental animals and in humans. In the present study, neither dextran nor ovalbumin was detected in lymph nodes following injection into the striatum. Nor was dextran detected in lymph nodes following direct injection into ventricular CSF. The failure to detect these tracers in lymph nodes is likely to be a reflection of the concentration and the amount of tracer injected. However, lymphatic drainage of 0.02-μm and 1-μm fluospheres was demonstrated in the present study when the fluospheres were injected into ventricular CSF. Similar drainage of Indian ink particles from CSF to cervical lymph nodes occurs within one minute [5,15].
There appear to be two possible routes by which solutes drain from rodent brains to cervical lymph nodes, but neither has been totally substantiated. One possibility is that ISF and solutes draining along the walls of arteries join the CSF in the subarachnoid space and drain to cervical lymph nodes via the cribriform plate and nasal lymphatics [5,15]. The other is that ISF and solutes drain along artery walls to the base of the skull and then to deep cervical lymph nodes. Evidence for the perivascular route comes from the presence of radioactive tracers in the walls of leptomeningeal arteries following injection of the tracer into the brain and the drainage of such tracers to lymph nodes on the same side of the neck .
Fate of particulate matter injected into the brain
In contrast to the drainage of solutes out of the brain along the basement membranes in capillary and artery walls, particulate matter dilates perivascular spaces between vessel walls and the surrounding glia limitans of the brain. This was observed with fluorospheres in the present study as shown in the summary diagram (Figure 4), and in previous work using Indian ink as a tracer . Subsequently, particulate tracers were ingested by perivascular macrophages and remained in the perivascular spaces for 7 days in the case of fluorospheres (the longest time point in the present study) and for at least 2 years with Indian ink .
Fluospheres injected in the ventricles were readily located in the systemic lymphoid tissue and the liver, in line with recent data showing that antigen-presenting cells delivered to the CSF rapidly migrate to peripheral lymphoid tissue . In contrast, none of the fluospheres injected into the brain parenchyma was found peripherally. Despite the fact that Indian ink particles  and fluospheres were taken up by perivascular macrophages, these cells did not migrate from the brain and carry this material to the periphery. Similarly, antigen-presenting cells delivered to the parenchyma are unable to migrate to the periphery . It is known, however, that antigen-presenting cells in peripheral tissues will rapidly migrate to local lymph nodes from sites of inflammation (see ), thus attempts were made in the present study to induce migration of macrophages along perivascular spaces by the co-injection of LPS or kainic acid with the fluospheres. Injection of these agents into the mouse brain led to local inflammation [17,18] and some increase in the spread of the fluospheres within the brain, but did not result in detectable drainage of fluorospheres from the brain. This suggests that perivascular pathways are not routes for the traffic of particles, macrophages or other inflammatory cells from the brain. Recent analysis of drainage from the anterior chamber of the eye has shown that solutes rapidly find their way to lymph nodes, but cells labelled with fluospheres do not . It has been suggested that the differential capacity of soluble antigens but not antigen-presenting cells to migrate from the brain plays an important role in the relative immunological privilege of the brain [3,39].
The space that is created around arteries by the injection of fluospheres and Indian ink is not entirely artefactual as equivalent dilatation of perivascular spaces is observed particularly in the corpus striatum of middle-aged to elderly humans . In rare instances, the perivascular spaces in the corpus striatum, white matter or mid-brain become widely dilated and form cyst-like structures . Perivascular spaces in the human cerebral cortex do not dilate in the same way, probably because of the structural differences between cortical and striatal arteries [40,42].
Dilatation of perivascular spaces occurs in the white matter in Alzheimer's disease; this has been associated with the severity of amyloid angiopathy in leptomeningeal arteries . It is considered that amyloid in the walls of arteries may impede the drainage of ISF which then accumulates in the dilated perivascular spaces in the white matter .
The present study has shown that basement membranes in the walls of capillaries and arteries act as an important component of the pathway for drainage of solutes from the brain. This perivascular route is permissive for the drainage of soluble antigen, but not particulate matter or cells from the brain to peripheral lymph nodes; this has implications for neuroimmunology and may play a part in the relative immune privilege of the brain. Amyloid-β is deposited in the perivascular drainage pathways in CAA and acts as a natural tracer to outline such pathways in humans and in transgenic mice. Blockage of perivascular drainage in CAA may be in part mediated by changes in perivascular basement membranes, and may be a factor in the accumulation of Aβ in the brain in Alzheimer's disease. By elucidating the factors involved in perivascular transport of solutes from the brain, it may be possible to influence their elimination in the management of Alzheimer's disease and neuroimmunological disorders.
This study was supported by the Alzheimer Research Trust.