Permeability and route of entry for lipid-insoluble molecules across brain barriers in developing Monodelphis domestica

Abstract

1. We have studied the permeability of blood-brain barriers to small molecules such as [¹⁴C]sucrose, [³H]inulin, [¹⁴C]l-glucose and [³H]glycerol from early stages of development (postnatal day 6, P6) in South American opossums (Monodelphis domestica), using a litter-based method for estimating steady-state cerebrospinal fluid (CSF)/plasma and brain/plasma ratios of markers that were injected i.p..

2. Steady-state ratios for l-glucose, sucrose and inulin all showed progressive decreases during development. The rate of uptake of l-glucose into the brain and CSF, in short time course experiments (7–24 min) when age-related differences in CSF production can be considered negligible also decreased during development. These results indicate that there is a significant decrease in the permeability of brain barriers to small lipid-insoluble molecules during brain development.

3. The steady-state blood/CSF ratio for 3000 Da lysine-fixable biotin-dextran following i.p. injection was shown to be consistent with diffusion from blood to CSF. It was therefore used to visualise the route of penetration for small lipid-insoluble molecules across brain barriers at P 0–30. The proportion of biotin-dextran-positive cells in the choroid plexuses declined in parallel with the age-related decline in permeability to the small-molecular-weight markers; the paracellular (tight junction) pathway for biotin-dextran appeared to be blocked, but biotin-dextran was easily detectable in the CSF. A transcellular route from blood to CSF was suggested by the finding that some choroid plexus epithelial cells contained biotin-dextran.

4. Biotin-dextran was also taken up by cerebral endothelial cells in the youngest brains studied (P0), but in contrast to the CSF, could not be detected in the brain extracellular space (i.e. a significant blood-brain barrier to small-sized lipid-insoluble compounds was already present). However, in immature brains (P0–13) biotin-dextran was taken up by some cells in the brain. These cells generally had contact with the CSF, suggesting that it is likely to have been the 2source of their biotin-dextran. Since the quantitative permeability data suggest that biotin-dextran behaves similarly to the radiolabelled markers used in this study, it is suggested that these markers in the more immature brains were also present intracellularly. Thus, brain/plasma ratios may be a misleading indicator of blood-brain barrier permeability in very immature animals.

5. The immunocytochemical staining for biotin-dextran in the CSF, in contrast to the lack of staining in the brain extracellular space, together with the quantitative permeability data showing that the radiolabelled markers penetrated more rapidly and to a much higher steady-state level in CSF than in the brain, suggests that lipid-insoluble molecules such as sucrose and inulin reach the immature brain predominantly via the CSF rather than directly across the very few blood vessels that are present at that time.

The term blood-brain barrier is used to describe a series of structural and functional mechanisms that control the internal environment of the central nervous system (CNS), which in the adult is distinct and remarkably stable compared with that of the rest of the body. These mechanisms operate across three main interfaces between the blood and the CNS: the cerebral blood vessels (blood-brain barrier), the choroid plexuses and the pia-arachnoid membranes (each of these latter two constitute separate blood-CSF barriers). Evidence has accumulated over the last 30 years to show that the apparent permeability of the blood-brain and blood-CSF barriers to low molecular weight (MW) lipid-insoluble compounds is greater in younger compared with older animals (see Saunders, 1992; Habgood et al. 1993; Saunders & Dziegielewska, 1997; Saunders et al. 1999). There is a widespread belief that this greater apparent permeability in the immature animal is simply a reflection of a less-well-developed barrier system (e.g. Ganong, 1999; Timbrell, 2000). Although this is probably true for some mechanisms, for others, not only are they well developed early in brain development, but the immature brain possesses additional specialised mechanisms that are not present in the adult (Fossan et al. 1985; Møllgård et al. 1987; Dziegielewska et al. 2000). Quantitative estimates of age-related changes in permeability are quite variable, even within the same species, probably because some studies were not carried out under steady-state conditions (cf. Ferguson & Woodbury, 1969; Habgood et al. 1993). Even when such conditions have been met, there have been quite different interpretations of the greater apparent permeability in younger animals. Some authors (e.g. Bass & Lundborg, 1973) have suggested that the age-related decline in blood-CSF exchange is due primarily to an increase in CSF turnover (the sink effect, see Davson & Segal, 1996), whereas others have interpreted it as a real decline in the intrinsic permeability of these interfaces (see Saunders et al. 2000 for discussion). The rate of CSF turnover in the immature brain is technically difficult to measure. Some data have been obtained for fetal sheep (Evans et al. 1974), and from these data it has been calculated that, in this species at least, the sink effect in the maturing brain actually decreases if account is taken of the total volume of CSF and the size of the brain as it grows (Saunders, 1992).

We have investigated blood-CSF and blood-brain barrier permeability in very young (postnatal days (P)6–17) and older (P37–65) South American opossums (Monodelphis domestica). This is the earliest stage of brain development at which such studies have been carried out. Just how early in development this is can be illustrated by comparing the age at which the choroid plexuses appear. In fetal sheep, the choroid plexus appears in the fourth ventricle at embryonic day (E)18–21, in the lateral ventricle at E21–24, and in the third ventricle at E30–36 (Jacobsen et al. 1983). In the rat, the choroid plexuses appear at E12 (fourth ventricle), E13–14 (lateral ventricle) and E16 (third ventricle; Chamberlain, 1973). The earliest age at which permeability studies with inulin and sucrose have been carried out in these species is E40–60 in fetal sheep (Evans et al. 1974; Dziegielewska et al. 1979; Cavanagh et al. 1983) and E18 to P2 in rats (Ferguson & Woodbury, 1969; Habgood et al. 1993). In contrast, marsupials are born with only rudimentary lateral ventricular choroid plexuses, no third ventricular plexus, and only a small fourth ventricular choroid plexus (Dziegielewska et al. 2001). In addition, we have studied the rate of blood-brain and blood-CSF uptake in short-term experiments, where any age-related differences in the effect of CSF turnover on ratios are minimal. By comparing the initial rate of uptake of a lipid-insoluble (l-glucose) and a more lipid-soluble (glycerol) compound in such short-term experiments, we have been able to demonstrate a decline in barrier permeability with increasing age, a phenomenon that is independent of any change in CSF turnover (CSF sink).

METHODS

Animals

South American opossum (Monodelphis domestica) pups were obtained from the colony established at the Central Animal House, University of Tasmania. Like all marsupials, this species is born at an earlier stage of brain development than any eutherian mammal, after a gestational period of only 14 days (Saunders et al. 1989). To avoid the risk of cannibalism, pups were detached from their mothers before injection of permeability markers and kept in a humidicrib (60–70 % relative humidity), for up to 4.5 h, at an air temperature of 28.5–29.5 °C. Normal adult body temperature in Monodelphis is 32 °C (Saunders et al. 1992). All experiments were carried out in accordance with National Health and Medical Research Council guidelines and with the approval of the University of Tasmania Ethics (Animals) Committee.

Permeability studies

Injection of radioactive markers

The radioactively labelled compounds used in this study, [¹⁴C]sucrose (CFB-146; MW 362), [³H]inulin (TRA-324; MW ∼5200), [¹⁴C]l-glucose (CFA-328; MW 182) and [³H]glycerol (TRA-244; MW 92), were obtained from Amersham International.

All radioactive markers injected were prepared in sterile isotonic NaCl solution (154 mm) and the volume injected i.p.. was standardised at 6 μl (g body mass)⁻¹. The final activities of injected markers were 4.2–5.0 kBq μl⁻¹[³H] and 0.4–0.6 kBq μl⁻¹[¹⁴C]. The activity of each injection solution was always measured before and after each experiment to ensure reproducibility between the experiments. Inulin, sucrose and l-glucose were chosen as small lipid-insoluble test molecules of permeability since they are not metabolised in the body. Glycerol was used as a moderately lipid-soluble molecule of similar size to l-glucose. Thin-layer chromatography (TLC) was used to check the integrity of [³H]glycerol 30 min after an i.p. injection, which was the longest time that glycerol was present in an animal. Five microlitres of plasma, CSF or injectate was run in an ascending fashion on TLC silica gel plates (Merck, G60), which were then developed in chloroform-methanol (1:2). A glycerol standard was run in parallel to identify the position of glycerol on the gel. At the end of the run, each lane was divided into equal fractions, the silica was removed carefully into scintillation vials, and the radioactivity in each vial was determined by liquid scintillation counting (Dziegielewska et al. 1980). Thirty minutes after the i.p. injection, 70–80 % of the radioactivity was still attached to glycerol in both the plasma and CSF (this compares with about 95 % in the injectate). The remaining 20–30 % of the counts were distributed evenly in all of the fractions below the glycerol position, indicating that these counts were attached to water and/or other molecules that were more hydrophilic than glycerol (see also Discussion). No significant counts were detected in the fractions above glycerol.

The relative lipid solubility of each compound used in the permeability studies was estimated by measuring its partition between equal volumes of a phosphate buffer (pH 7.4) and 2-octonol pre-equilibrated with the buffer (Rapoport, 1976).

Sampling and radioactivity counting

For these and all other experiments, animals were killed by inhalation of halothane (Zeneca). Blood was then collected in heparinised glass micropipettes by gentle mouth suction. In older animals (> P10) the rib cage was opened and blood was collected directly from the left ventricle of the heart. In young animals (< P10), blood was collected by drainage from the left subclavian artery. Blood samples were transferred to plastic tubes and kept on ice until the plasma was separated by centrifugation (5 min, 5000 r.p.m.). CSF samples were collected from the cisterna magna of the hindbrain by gentle mouth suction using a glass micropipette. Any tube with visible blood contamination was discarded (Habgood et al. 1993). The brain was then dissected out, weighed and left in 0.5 ml of tissue solubiliser overnight (Soluene-350, Packard) at 40 °C. Aliquots of plasma and CSF were transferred to scintillation tubes, weighed, and then 5 ml of scintillation fluid (Packard) was added. The activity of markers in each sample was determined by liquid scintillation counting (Beckman LS3801) with window settings to allow the determination of both [¹⁴C] and [³H] activity. Background activity was estimated from tubes containing only scintillation fluid, or for brain samples from tubes containing scintillation fluid and 0.5 ml Soluene. This was subtracted from the total activity of each sample tube. A greater activity of [³H] relative to [¹⁴C] (10:1 ratio) was used to minimise the effect of spill-over of [¹⁴C] counts into the [³H] window (Dziegielewska et al. 1979). The activity of each brain tissue sample was corrected for its residual vascular space, which was estimated from the initial distribution space (less than 5 min) of inulin in P15 pups (n = 5; used to correct values in animals between P6 and P17) and in P40 pups (n = 5; used to correct values in animals between P32 and P65). Isotope activity in each sample was calculated as Bq (g of tissue)⁻¹.

Determination of steady-state CSF/plasma ratios

Preliminary experiments to determine the time required to approach steady state were performed in nephrectomised P37 animals (n = 12). Nephrectomy was used in order to maintain a steady plasma level for each marker molecule over several hours (Habgood et al. 1993). Nephrectomy was performed on a heated operating table under inhaled isoflurane anaesthesia (2.5–3.5 % in oxygen; Abbott). The kidneys were exposed through skin and muscle incisions and the renal blood vessels were ligated with silk thread. The wound was closed and skin sealed with cyanoacrylate adhesive. Blood, CSF and brain samples were collected at various time points after an i.p.. injection of isotope-containing solution (1, 2, 3, or 4 h). The small amount of blood available in young opossums made it impossible to take serial blood samples from one animal. In order to monitor the plasma concentration of markers over time, a litter-based model was used (Habgood et al. 1993).

Newborn opossum model

The very small size of newborn opossum pups means that it is technically impossible to achieve constant plasma concentrations of markers, either by constant infusion or by nephrectomy (Bradbury & Davson, 1965; Habgood et al. 1993; Keep et al. 1995). Instead, the final CSF and brain concentrations for each marker were related to the integral of the plasma concentration throughout the experiment (plasma profile) after a single injection in non-nephrectomised unoperated animals. Ratios were calculated as:

(1)

(2)

where C and T are concentration and time, respectively. The plasma concentration profiles for [¹⁴C]sucrose and [³H]inulin in unoperated (non-nephrectomised) animals were determined in groups of opossums at P9–10 (n = 15), P15 (n = 15) and P37 (n = 12), and for [¹⁴C]l-glucose and [³H]glycerol in animals at P17–18 (n = 11) and P38–39 (n = 10). Animals were given a single standardised i.p.. injection (6 μl (g body mass)⁻¹) of isotopes. Between 20 min and 4 h after the injection, each animal was terminally anaesthetised (inhaled halothane) and samples of blood, CSF and brain tissue were collected. The plasma concentration of each radioactive marker was plotted against time after injection, and curves were computer fitted to the data points (non-linear Bézier function). The plasma concentration curves thus constructed are representative of all littermates. In order to compare litters of animals from different experiments, the plasma concentration curve at each age was standardised relative to the concentration of the marker in the injection solution using eqn (3), as shown in Fig. 1:

(3)

where C is the concentration of radioactive marker (in Bq μl⁻¹).

Figure 1. CSF and plasma concentration ratios for sucrose and inulin in nephrectomised and non-nephrectomised opossums.

The left ordinate in A and C shows the plasma/injectate (▪) and the CSF/injectate concentration ratios (•) for [¹⁴C]sucrose (A) and [³H]inulin (C) following a single i.p. injection in nephrectomised opossums at P37. The plasma and CSF concentrations have been presented as ratios normalised to the original injection solution in order to compare directly between different animals and experiments. The right ordinates in A and C show the CSF/plasma concentration ratios (×) for [¹⁴C]sucrose (A) and [³H]inulin (C) in nephrectomised opossums at P37. The plasma/injectate and CSF/injectate concentration ratios for [¹⁴C]sucrose (A) and [³H]inulin (C) approached a steady level by 3 h after injection. CSF/plasma concentration ratios also approached a steady level by 3 h after injection for both markers. B and D show the time course of plasma/injectate concentration ratios for [¹⁴C]sucrose (B) and [³H]inulin (D) after a single i.p. injection in non-nephrectomised animals at P9–10 (dotted line, ♦), P15 (continuous line, ▿) and P37 (dashed line, ♦) The curves shown in B and D were computer fitted to data points for each age and used to calculate mean plasma concentrations (see Methods). The plasma/injectate concentration ratios in these non-nephrectomised animals all reach a peak and then gradually decline over time. The curves for P9–10 and P15 animals were very similar for both markers, whereas the peak for P37 animals occurred somewhat later. The data points shown are means (n = 3–4) and the error bars indicate ±s.e.m.. Where no error bars are visible, they are obscured by the symbols.

Steady-state experiments

Neonatal opossum litters were injected i.p.. with a mixture of [¹⁴C]sucrose and [³H]inulin at P5–7 (n = 9), P9–13 (n = 15), P15–17 (n = 5), P32–37 (n = 10) or P65 (n = 3), or with [¹⁴C]l-glucose at P18 (n = 4) and P38 (n = 4). All animals were terminally anaesthetised between 3 and 3.5 h after injection (inhaled halothane) and samples of blood, CSF and brain tissue were collected. Steady-state CSF/plasma and brain/plasma ratios in animals were estimated as described above. The mean plasma levels for [¹⁴C]sucrose and [³H]inulin of animals between P6 and P10 were estimated from the plasma concentration curves for P9–10 animals, and for animals between P13 and P17 from the plasma curve for P15 animals; finally, for animals between P32 and P37 the plasma concentration curve from P37 animals was used. Similarly, the mean plasma concentrations of [¹⁴C]l-glucose at P18 and P38 were determined from the plasma concentration curves for P17–18 and P38–39 animals, respectively. Animals at P65 were nephrectomised prior to injection, but were otherwise treated in the same manner, and ratios were calculated assuming a steady plasma level. Ratios for each age are expressed as means ±s.e.m.

Initial uptake rates

Opossum pups at P17–18 (n = 9) and P38–39 (n = 13) were injected i.p.. with a mixture of [³H]glycerol and [¹⁴C]l-glucose. Animals were killed by inhaled halothane anaesthesia and samples of cisternal CSF were collected between 7 and 24 min after the injection. Animals were then decapitated and samples of blood and then brain tissue were collected. The mean (±s.e.m..) time difference between CSF collection and blood sampling was 26 ± 1.8 s in P17–18 animals (n = 8) and 18 ± 1.6 s in P38–39 animals (n = 10). The time of brain sampling was recorded as the decapitation time. CSF/plasma and brain/plasma ratios were calculated and plotted against the blood collection times (Fig. 4).

Figure 4. CSF/plasma ratios of different markers against diffusion coefficients in opossums at different ages.

Plot of steady-state CSF/plasma concentration ratios (ordinate, note logarithmic scale) against the diffusion coefficients (D₃₂) for molecules of various molecular sizes (abscissa) in opossums at P5–7 (•), P10–13 (○), P15–17 (▪), P32–37 (□) and P65 (▴). Values of D were calculated for 32 °C, as this is the normal body temperature for Monodelphis. Data for succinylated albumin (s-Albumin) are from Knott et al. (1997). Values shown are means (n = 3–10) and the error bars are ±s.e.m.. Where no error bars are visible, they are obscured by the symbols. There was a parallel decrease in the curves of CSF/plasma ratios against D₃₂ with increasing postnatal age. The star ([ww2]) indicates the CSF/plasma concentration ratio for D-3308 at P16. The ratio falls close to the line between inulin and sucrose, which indicates that the handling of D-3308 by the brain barriers is similar to that of both sucrose and inulin.

Morphological studies

Biotin-dextran experiments

Litters of opossum pups at P0, P5, P9, P13 and P30 were removed from the mother and then injected i.p.. with 3000 Da, lysine-fixable, biotin-dextran (BDA-3000, Molecular Probes; 700 μg (g body mass)⁻¹) in a sterile 154 mm NaCl solution. Five to ninety minutes after injection, animals were terminally anaesthetised with inhaled halothane and their brains were immediately fixed. For younger animals (< P15), brains were dissected out and fixed by immersion in Bouin's fixative. Older animals (> P15) were first perfuse-fixed with paraformaldehyde and then the brains were dissected out and fixed in Bouin's fixative. After 24 h in Bouin's fixative, these brains were dehydrated in serial solutions of ethanol (70 % to 100 %) and then cleared in chloroform. After embedding in paraffin wax, serial sagittal or coronal sections of brain tissue (5 μm) were cut.

Sections were dewaxed in xylene and rehydrated in decreasing concentrations of ethanol (100 % down to 70 %). This was followed by incubation in peroxidase blocker (Dako) followed by protein blocker (Dako) for 30 min each at room temperature. After washing in phosphate-buffered saline (PBS) containing 0.2 % Tween 20 (Tween 20 PBS; Sigma), an avidin-horseradish peroxidase (HRP) complex was used to detect biotin with a Vectastain Elite ABC kit (Vector). Sections were then processed with the diaminobenzidine tetrahydrochloride (DAB kit, Dako) reaction for about 5–15 min. After a 10 min rinse in tap water, the sections were dehydrated and mounted in DPX mounting medium (Aldrich, Australia). Control staining was performed on sections obtained from animals not given a BDA-3000 injection.

Biotin-dextran and albumin double labelling

Sections for double labelling of endogenous albumin and biotin-dextran were dewaxed in xylene, rehydrated in serial ethanol solutions (100 % to 70 %) and incubated for 30 min in protein blocker (Dako). After washing in 0.2 % Tween 20 PBS, sections were incubated with rabbit antibodies against human albumin diluted to 1:100 (Dako), overnight at 4 °C. These antibodies have shown a high degree of cross-reactivity with endogenous Monodelphis albumin (Knott et al. 1997). Sections were washed and incubated in a mixture of 1:30 dilution of swine anti-rabbit fluorescein isothiocyanate (FITC)-conjugated antibodies (Dako) and a 1:70 dilution of streptavidin-Texas Red (Vector) overnight at 4 °C. Sections were mounted in an aqueous mounting medium (Dako) and examined with the aid of an Olympus BX50 fluorescence microscope. Antibody dilutions were made in PBS with 2 % fish gelatine solution (Sigma), and incubation of sections was carried out in a humid chamber.

Validation of BDA-3000 as a marker for transfer of lipid-insoluble molecules

To confirm that BDA-3000 is a suitable marker for small lipid-insoluble molecules, CSF/plasma ratios were measured for a structurally similar lysine-fixable 3000 MW rhodamine dextran (D-3308, Molecular Probes). D-3308 was used instead of BDA-3000 since it is fluorescent and can therefore be measured quantitatively. A group of opossum pups at P16 (n = 9) were injected i.p.. with D-3308 (500–1000 μg (g body mass)⁻¹ in sterile 154 mm NaCl solution), and samples of plasma and CSF were collected 3 h later. The amount of fluorescence in each sample was measured with an Olympus BX50 fluorescence microscope attached to a PM30 photomicrograph unit. Samples of plasma and CSF were transferred to 5 μl glass capillaries and mounted on glass slides. At ×10 magnification, the exposure time (ET) was recorded for the centre of the glass capillary. CSF/plasma ratios were calculated as ET_plasma/ET_CSF (exposure times being inversely proportional to concentration). The accuracy of measuring ratios this way was confirmed by comparisons with ratios obtained using absorbance spectrophotometry. The absorbance of serial dilutions of D-3308 in saline was determined by a spectrophotometer at 518 nm. Ratios calculated using absorbance spectrophotometry between the different solutions were almost identical to ratios calculated from the ET readings. The influence of media on ET readings was also examined; similar concentrations of D-3308 in saline, CSF or plasma all gave similar ET readings. This method allowed the measurement of very small sample volumes (1–2 μl), undiluted in a glass capillary.

In order to confirm that the brain uptake of D-3308 and BDA-3000 was similar, their distributions in brain sections were compared in opossums of the same ages. P8 and P16 opossums were injected i.p.. with either D-3308 (0.7 mg (g body mass)⁻¹) or BDA-3000 (0.7 mg (g body mass)⁻¹), the brain fixed in Bouin's fixative after 30–40 min and processed as described above for histochemistry. BDA-3000 was visualised as described above for double labelling using streptavidin FITC, and D-3308 was detected directly using filters for rhodamine.

RESULTS

Permeability experiments

Figure 1A and C shows the plasma and CSF concentration curves for [¹⁴C]sucrose and [³H]inulin in nephrectomised P37 animals after a standardised i.p.. injection (6 μl (g body mass)⁻¹). The plasma concentration increased rapidly after the injection and reached a plateau within 3 h for inulin and earlier for sucrose. This shows that nephrectomy is a method of choice for achieving a steady plasma concentration of a substance when constant infusion is not practical. CSF/plasma ratios in these animals also approached a plateau by 3 h after injection (Fig. 1A and C).

In nephrectomised animals, the final plasma concentration can be used as an accurate estimate of the mean plasma concentration up to that time point. In non-nephrectomised animals, the final plasma concentration does not reflect the mean plasma concentration, but this can be estimated by dividing the integral of the plasma concentration curve by the time from the injection until plasma was collected (see Methods, eqns (1) and (2)). Plasma concentration curves determined at P9–10, P15 and P37 for sucrose and inulin are shown in Fig. 1B and D. Plasma concentration curves for l-glucose were determined in opossums at P18 and P38 (data not illustrated).

A comparison of steady-state results from nephrectomised and control (non-nephrectomised) animals at P37 revealed no significant differences for CSF/plasma or brain/ plasma ratios of both sucrose and inulin (see Table 1). Thus, steady-state ratios can be estimated accurately from animals without prior nephrectomy. Similar ratios for nephrectomised and non-nephrectomised animals were also observed at earlier and later time points (data not shown).

Table 1.

Comparison of ratios of nephrectomised and intact animals

	Sucrose		Inulin
	CSF/plasma (%)	Brain/plasma (%)	CSF/plasma (%)	Brain/plasma (%)
Nephrectomised (n = 8)	15.0 ± 1.8	4.4 ± 0.8	6.9 ± 0.6	3.5 ± 0.9
Intact (n = 5)	14.2 ± 2.0	4.3 ± 0.5	7.3 ± 0.7	3.4 ± 0.4

Steady-state cerebrospinal fluid (CSF)/plasma and brain/plasma ratios for [¹⁴C]sucrose and [³H]inulin in nephrectomised and intact (non-nephrectomised) P37 opossums. CSF concentrations were measured in samples collected from the cisterna magna. In nephrectomised animals, the CSF and plasma concentrations both approached steady state by 3 h after i.p. injection, whereas in intact (nonnephrectomised) animals the plasma concentration peaked at 30–60 min after the injection and then gradually fell. In these intact animals, the mean plasma concentration was determined from the integral of the plasma concentration curve over the entire time course of each experiment and ratios were calculated as CSF/mean plasma concentration or brain/mean plasma concentration (see Methods for details). These two different methods of determining steady-state ratios gave similar results. Values are presented as means ±s.e.m.

Permeability of markers during development

The changes in steady-state CSF/plasma and brain/plasma ratios for inulin and sucrose with age are shown in Fig. 2. The CSF/plasma ratios fell about 10-fold between P6 and P65, and brain/plasma ratios fell 4- to 5-fold during the same period. Brain/plasma ratios were consistently much lower than CSF/plasma ratios at all ages. Ratios for the smaller sucrose molecule were consistently higher than for inulin at all ages.

Figure 2. CSF/plasma and brain/plasma ratios during postnatal development in opossums.

The ordinates are steady-state CSF/plasma (upper panel) and brain/plasma (lower panel) concentration ratios for [¹⁴C]sucrose (▪), [³H]inulin (○), and [¹⁴C]l-glucose (▵) the abscissae are postnatal age. All ratios were measured in non-nephrectomised animals, except those at P65, which were measured in animals that had been nephrectomised. Both the CSF/plasma and brain/plasma concentration ratios declined markedly with increasing postnatal age. Brain/plasma concentration ratios were also consistently lower than CSF/plasma concentration ratios at all ages. Values shown are means (n = 3–8) and the error bars indicate ±s.e.m.. Where no error bars are visible, they are obscured by the symbols.

Steady-state ratios for l-glucose at P18 and P38 are also shown in Fig. 2. The CSF/plasma ratios for l-glucose declined significantly between P18 and P38 (P < 0.01, two-tailed t test), as did brain/plasma ratios (P < 0.01).

Short time course experiments

CSF/plasma and brain/plasma ratios measured in short time course experiments for both glycerol and l-glucose are shown in Fig. 3. During this short initial period (7–24 min) the plasma concentrations were relatively stable and therefore changes in CSF/plasma and brain/plasma concentration ratios with time were related directly to the rate of uptake into the CSF and brain. Analysis of the data points showed that uptake into the CSF and brain was linear over this time period and the slope of fitted linear regression lines could be used to represent the initial uptake rate into either the CSF or brain. Differences in the slope of the regression lines were compared using Student's t test (unpaired). l-Glucose showed a much faster initial rate of uptake into CSF (Fig. 3A) and brain (Fig. 3C) at P18 compared to P38. The slope of the regression line for CSF uptake of l-glucose in P18 animals was significantly steeper than that of P38 animals (P = 0.01). In addition, the rate of uptake into CSF was greater than into brain, at both ages. Because of the greater scatter of the data, the difference in uptake into the brain between the two ages failed to reach statistical significance (P = 0.07). The rate of uptake of glycerol into CSF and brain appeared similar at the two developmental ages (see Fig. 3B and D). The slopes of the regression lines for CSF uptake of glycerol and for brain uptake of glycerol were not significantly different between the two ages (P = 0.97 and P = 0.33, respectively).

Figure 3. Initial uptake rates for l-glucose and glycerol in P18 and P38 opossums.

The ordinates are CSF/plasma (A and B) and brain/plasma (C and D) concentration ratios for [¹⁴C]l-glucose (A and C) and [³H]glycerol (B and D), measured in short time course experiments after a single i.p.. injection in opossums at P18 (▪) and P38 (○). The abscissae are the time after injection at which samples were collected. Linear regression lines have been fitted to the data points. The slopes of the regression lines (the rate of uptake) for P18 and P38 animals were compared using Student's t test. The initial rate of uptake of [¹⁴C]l-glucose into the CSF in P18 animals was significantly faster than in P38 animals (P = 0.01). The initial rate of uptake of [¹⁴C]l-glucose into the brain at P18 was also faster than at P38, but this did not reach statistical significance (P = 0.07). There was no significant difference in the initial rate of [³H]glycerol uptake into the CSF or brain between P18 and P38 (P = 0.33 and P =−.97, respectively).

Lipid solubility of markers

Measurements of the relative lipid solubility of the markers were made at pH 7.4 and 25 °C and are referred to as log K_octanol,25 (where K is the partition coefficient). The measured log K_octanol,25 values for markers used in this study are listed in Table 2. The similar log K_octanol,25 values for sucrose and inulin mean that the greater CSF and brain uptake of sucrose at each age cannot be attributed to differences in their lipid solubility.

Table 2.

LogK values

Marker	LogK	n
Inulin	−3.51 ± 0.009	5
Sucrose	−3.47 ± 0.005	5
l-Glucose	−2.68 & −2.68	2
Glycerol	−1.94 ± 0.005	5

LogK_octanol,25 values (where K is the partition coefficient) for each marker calculated from their partitioning between equal volumes of phosphate buffer, pH 7.4, and 2-octanol at 25°C. Values shown are means ±s.e.m. Since only two measurements were made for l-glucose, both are listed here. Inulin and sucrose have very similar values, indicating that any difference in permeability for these markers cannot be attributed to differences in lipid solubility.

Uptake of D-3308 into the CSF

Dziegielewska et al. (1979) showed that CSF/plasma ratios for low molecular weight lipid-insoluble markers are related to diffusion coefficients (D) calculated from the Einstein-Stokes radii. Values of D for inulin, sucrose and l-glucose were taken from Normand et al. (1971) and corrected from 25 °C (i.e. D₂₅) to 32 °C (i.e. D₃₂; body temperature in Monodelphis, see Saunders et al. 1992) by a factor of 1.1924. The D value for D-3308 was estimated from its molecular weight. A linear relationship (r = 0.997, least squares linear regression) was found between MW^−1/2 and the D₂₅ values for a range of molecules given by Normand et al. (1971), and from this the D₂₅ value for D-3308 could be estimated and then corrected to 32 °C. Figure 4 shows the steady-state CSF/plasma ratios for albumin, inulin, sucrose and l-glucose plotted against D₃₂ at various ages. The 3 h ratio for D-3308 at P16 was approximately 16 %, which is similar to the steady-state ratio for inulin (14 %), but is substantially less than the ratios for the much smaller molecules sucrose (35 %) and l-glucose (36 %). In order to justify D-3308 as a quantifiable substitute for BDA-3000, the distribution of the two markers was compared in brain sections from P8 and P16 opossums, 30–40 min after an i.p.. injection. The two markers showed very similar distribution patterns with staining in a small proportion of the choroid plexus cells, the proportion being less at P16 compared to P8. Both markers showed intracellular staining in some cells in the ventricular zone and in the hippocampal area (not illustrated). The rest of the brain seemed to lack any staining except for the blood vessel lumen and endothelial cells, which were strongly stained.

Histochemistry of BDA-3000

Examination of brain sections processed for BDA-3000 revealed that in the brain, most of the reaction product was in endothelial cells and within the lumen of the blood vessels (see Fig. 5). Occasional cells in the ventricular zone and the hippocampal area showed intracellular staining. No staining was detectable outside the blood vessel wall (Fig. 5F). The surfaces of both the ventricular neuroependyma and the choroid plexuses stained positive where CSF had precipitated on these surfaces. In general, the extent of staining was more prominent in the brains of younger animals, where intracellular staining in the basal forebrain and diencephalon was also found. There was no staining in control sections.

Figure 5. Localisation of BDA-3000 in the developing opossum brain.

BDA-3000 in coronal sections of the opossum forebrain at P0 (A–C) and P8 (D–G) 30–45 min after an i.p.. injection. The micrograph shown in F was viewed with Nomarski optics, and all other photomicrographs were viewed with bright field microscopy. BDA-3000 was detected using a Vectastain Elite ABC kit followed by the DAB reaction. B, a magnified area of the developing neocortex (boxed region in A). C, a high-power micrograph of a blood vessel (boxed region in B) from a P0 animal. Strong staining can be seen in precipitated CSF (A and B). At this age, very few blood vessels were present within the brain (A and B). Note in B that some cells in the ventricular zone had taken up BDA-3000 from the CSF. D, low-power view of a P8 opossum brain showing strong staining in the stroma of the choroid plexus and in blood vessels. A few cells in the hippocampal area have taken up BDA-3000. Note the larger number of blood vessels in the brain compared to P0 (A). Magnified areas from D are illustrated in E and F. The stroma and a few epithelial cells (arrows) were stained in the choroid plexus (E; also see Fig. 6). A few cells around the ventricles were also stained, but not as many as at P0 (compare E and B). Note that BDA-3000 seems to have penetrated the pia on the dorsal surface of the brain, and the marginal zone (mz) was stained. However, very little reaction product was found in the cortical plate (cp) or further into the brain. vz, ventricular zone. F, high-power photomicrograph of a vessel showing that red blood cells and endothelial cells were strongly stained, but the extracellular space around the vessel lacked any reaction product. G, high-power photomicrograph of BDA-3000 distribution in the lateral ventricular choroid plexus of a P8 opossum. BDA-3000 appeared to penetrate in between the epithelial cells from the blood side of the choroid plexus, but was stopped along the way and did not reach the apical side (CSF side). Scale bars: 500 μm, D; 300 μm, A; 50 μm, B and E; and 10 μm, C, F and G.

BDA-3000 localisation in the choroid plexuses

In the choroid plexuses the stroma and connective tissue were strongly stained along with a minority of the epithelial cells. The intracellular reaction product had a granular appearance and was absent from the nuclei of the epithelial cells (Fig. 6).

Figure 6. Localisation of BDA-3000 labelling and co-localisation with endogenous albumin in the developing choroid plexus in opossums.

The distribution of BDA-3000 labelling in the lateral choroid plexus of Monodelphis domestica after i.p.. injection at P0 (A) and P9 (B), and double staining of endogenous albumin (C) and BDA-3000 (D) at P8. BDA-3000 in A and B was detected using a Vectastain Elite ABC kit followed by the DAB reaction. The proportion of choroidal epithelial cells positive for BDA-3000 in the lateral ventricle decreased with age (compare A and B). Note the developmental change in choroid plexus morphology from a pseudostratified layer of epithelial cells at P0 (A) to the cuboidal plexus cells exhibiting a larger apical surface area at P9 (B). Albumin in C was detected using antibodies to human albumin and FITC-coupled secondary antibodies, and BDA-3000 in D was detected using streptavidin conjugated with Texas Red. The thick arrow indicates a double-labelled cell and the thin arrow indicates a single-labelled cell for BDA-3000. The number of cells positive for BDA-3000 was higher than for albumin. No cells were found to be labelled only for albumin. Scale bar is 25 μm.

The number of BDA-3000-positive cells in the choroid plexuses from all four ventricles decreased progressively with age (Fig. 6). At the time of birth (P0), a proportionally higher number of epithelial cells was positive in the choroid plexuses in the lateral ventricle (Fig. 6A), whereas the fourth ventricular plexus contained only occasional stained epithelial cells. At this age the choroid plexus cells in the lateral ventricles have an elongated shape, whereas those in the fourth ventricle have a more mature columnar form. At P9 and later, only very occasional stained epithelial cells were found in the fourth ventricular choroid plexus, and the proportion of stained epithelial cells in the lateral ventricles (Fig. 6B) was substantially less than at P0. At P9–13 the plexus in the third ventricle had a high proportion of stained epithelial cells. At P30 all choroid plexuses showed only very occasional stained epithelial cells with rather faint intracellular staining. The choroidal epithelial cells at this age have the cuboidal shape that is characteristic of the adult choroid plexus. At P8, and in older animals, the BDA-3000 seemed to be stopped by the tight junctions in between epithelial cells (see Fig. 5G). At earlier ages it was difficult to determine where the staining was present in between epithelial cells since the epithelial layer has a densely packed pseudostratified appearance, which is typical of immature choroid plexus epithelial cells (Shuangshoti & Netsky, 1966; Dziegielewska et al. 2001).

Double labelling for BDA-3000 and albumin

The occasional staining of choroidal epithelial cells for BDA-3000 was similar to the staining pattern for albumin reported by Knott et al. (1997), although the proportion of BDA-3000-positive cells was greater. The co-localisation of BDA-3000 and albumin was studied using double-labelling immunocytochemistry. In all sections studied, choroid plexus epithelial cells were either labelled for BDA-3000 alone, or were double labelled for both BDA-3000 and albumin. In no sections were there any epithelial cells labelled only for albumin (Fig. 6C and D).

DISCUSSION

Penetration of markers into the CNS

The present study not only presents quantitative permeability measurements for small lipid-insoluble molecules at much earlier stages of brain and choroid plexus development than has been reported previously, it also has investigated the route of penetration for such molecules between blood and brain, and between blood and CSF. In comparison to l-glucose, sucrose and inulin, biotin-dextran (BDA-3000) was shown to have the CSF/ plasma ratio that would be expected from its molecular size (see Fig. 4). A striking and unexpected result of these experiments was that BDA-3000 had an intracellular rather than an extracellular distribution in the immature brain. Although cerebral endothelial cells took up the BDA-3000, none was detectable in the brain extracellular space. However, it was easily detectable in the CSF, and the cells in the brain that stained for BDA-3000 seem likely to have taken it up from the CSF rather than via the cerebral blood vessels, since the BDA-3000-positive cells were generally in contact with CSF surfaces rather than with blood vessels. In addition, the permeability to the low molecular weight markers used was strikingly greater at the youngest ages studied than previously reported for other species at somewhat older ages. Thus at P6 in Monodelphis, the CSF ratio for sucrose was about 60 %, compared with about 30 % at P2 in rats (Habgood et al. 1993). The higher ratios reported by Ferguson & Woodbury (1969) for fetal rats are misleading, because they were estimated from single terminal samples of both blood and CSF and therefore did not estimate the true steady state (see Methods). Because the BDA-3000 observations in the youngest Monodelphis suggest that the low molecular weight compounds did not appear to cross the endothelial cells of the cerebral vessels to any great extent and were probably largely, if not entirely distributed within cells in the immature brain, rather than the extracellular space, brain/plasma ratios in these immature animals may not be a meaningful estimate of blood-brain barrier permeability for small lipid-insoluble molecules. If this interpretation is correct, then it suggests that entry into the brain via the CSF may be an important route in the developing brain. This is perhaps not so surprising when the paucity of vascularisation of the immature brain is considered. Very few blood vessels are apparent in the brains of newborn or even P10 opossums (Fig. 5). The main period of vascularisation occurs later in development.

Two main developmental changes could be considered as possible explanations for the age-related decrease in steady-state CSF/plasma and brain/plasma ratios that occurs for markers such as sucrose and inulin during development.

(i) Increased CSF turnover

As mentioned in the Introduction, a decrease in steady-state levels in the brain and CSF with increasing age has been suggested by some authors (e.g. Bass & Lundborg, 1973) to be a result of an increase in CSF turnover (the sink effect), thus flushing molecules out of the brain and CSF at a higher rate. A review of previous studies concluded that particularly if the size of the brain is taken into account, the sink effect probably declines rather than increases with age (Saunders, 1992). Evidence to support this conclusion comes from the present experiments involving short-term uptake and comparison of l-glucose and glycerol, in which age-related differences in the rate of initial uptake of l-glucose but not glycerol into the CSF and brain were observed under conditions in which differences in CSF secretion rate would have been negligible.

(ii) Permeability of blood-CNS interfaces

A developmental reduction in the overall rate of entry across the brain barriers would also reduce steady-state ratios. Such a change in brain and CSF uptake could come about due to a change in the surface area for exchange (brain-barrier surface area) relative to brain volume, or to a change in the intrinsic permeability of the exchange interfaces. Since the capillary density in rat brain has been shown to increase markedly during the postnatal period (Caley & Maxwell, 1970), the decline in steady-state ratios cannot be attributed to a reduced surface area for exchange, but is more likely to be due to a reduced permeability of the blood-brain and blood-CSF interfaces. In order to investigate this possibility, short time course (7–24 min) uptake experiments were carried out using l-glucose, which is small enough to enter the brain and CSF to a measurable extent without being affected by the sink effect.

The initial rate of uptake into the CSF for l-glucose was significantly less in older animals compared to younger animals (P = 0.01), although this was not the case for the brain (P = 0.07). This is consistent with a reduction in the permeability of the blood-CSF barrier interface, which is separate from any apparent decrease due to changes in CSF turnover. Glycerol, being moderately lipid soluble (Table 2), readily penetrates the cell membranes of endothelial and epithelial cells. The radioactive labelling of glycerol in the plasma and CSF was checked in samples 30 min after injection. As indicated in the Methods, 70–80 % of the label was still linked to glycerol. The rest was linked to water or water-soluble compounds. Any labelled water would have been rapidly and widely distributed throughout the whole body water and would thus have added only a small amount to both the numerator and denominator of the CSF/plasma and brain/plasma ratios. The lack of change in uptake rates for glycerol at different ages (Fig. 3B and D) indicates that neither the permeability nor the surface area for exchange for glycerol changes during development. In contrast to the transfer of lipid-soluble glycerol across the whole-cell membrane interface between blood and brain and blood and CSF, lipid-insoluble molecules such as l-glucose, sucrose, inulin and BDA-3000 would be expected to transfer via hydrophilic ‘pores’. The nature of these pores has never been clearly defined, although there is a widespread belief that they lie in the tight junctions themselves (e.g. Goodenough, 1999); it may be that pores correspond to aquaporins in the cell membranes. Given the preservation of the relationship between steady-state ratios and diffusion coefficients at different ages (Fig. 4), the nature of this change is likely to be a reduction in the total number of ‘pores’ relative to the size of the brain (see Dziegielewska et al. 1979). These ‘pores’ could be represented by an intracellular pathway, which is suggested by the observation of BDA-3000-positive epithelial cells, since the proportion of positive cells also declined with age (see Fig. 6) in a manner parallel to the decline in steady-state ratios for inulin, sucrose and l-glucose.

Visualising the route of transfer of small lipid-insoluble markers across the brain barriers

Dextrans are hydrophilic polysaccharides and have favourable characteristics for use as external markers, such as good water solubility, low toxicity and relative inertness and they are available in a variety of different molecular sizes. In order to assess the usefulness of BDA-3000 as a lipid-insoluble tracer molecule, the 3 h CSF/plasma ratio of a fluorescent dextran (D-3308) that is structurally similar to BDA-3000 was measured and compared to other lipid-insoluble markers. BDA-3000 reached a ratio similar to that of inulin, but lower than sucrose, as would be expected from its molecular size (Fig. 4). Light microscopically, D-3308 and BDA-3000 co-localised in the brain, indicating that the two markers behave in a similar manner. The use of fluorescently labelled dextran allowed the quantification of its concentration in CSF and plasma in very small samples. Either form of biotin-dextran can be used to visualise the pathways from blood to the CSF and brain (Fig. 5). Previous studies have used markers such as HRP (Van Deurs 1978a; Tauc et al. 1984), microperoxidase (Van Deurs 1978b) and lanthanum (Castel et al. 1974), but the physiological interpretation of these studies is difficult because it is not known whether the cerebral endothelial and choroid plexus epithelial handling of these tracers is similar to that for quantitative markers, such as inulin and sucrose, used in physiological experiments (e.g. Ferguson & Woodbury, 1969; Habgood et al. 1993).

At the light microscopic level, there was less BDA-3000 staining in older brains compared to younger brains. The marker seemed to be stopped by the tight junctions that exist between epithelial cells in the choroid plexus from an early developmental stage (Fig. 5G). Previous electron microscopic studies of the fetal rat choroid plexus have shown that these tight junctions are not penetrated by HRP (Tauc et al. 1984), but HRP is a much larger molecule (32000 Da) than the biotin-dextran (3000 Da) used in the present experiments. As has been described previously for the transfer of albumin from blood to CSF in the immature brain (Knott et al. 1997; Balslev et al. 1997), a small proportion of choroid plexus epithelial cells was positive for BDA-3000 in the present study, suggesting the existence of a transcellular pathway into the CSF. Furthermore, as with albumin, the proportion of choroid plexus epithelial cells that was positive for BDA-3000 also declined with age, in parallel with the decline in blood-to-CSF permeability (cf. Fig. 2 and Fig. 4). The highest proportion of BDA-3000-positive cells in the lateral ventricular choroid plexus was in the newborn animal and in the third ventricular choroid plexus shortly after it appeared on P5. The proportion of BDA-3000-positive cells fell subsequently (earlier in the lateral ventricular choroid plexus), corresponding to the decline in permeability to small lipid-insoluble markers. In order to resolve in more detail the intracellular and intercellular localisation of BDA-3000 in the choroid plexus epithelium, an electron microscope study would be required.

If the explanation for the greater apparent permeability of barriers between the blood and CSF (and brain) to low molecular weight lipid-insoluble compounds in the immature brain is that there is predominantly greater transfer across choroid plexus epithelial cells from the blood into the CSF, with subsequent uptake into brain cells, rather than directly across the endothelial cells of the few blood vessels present, this would suggest that the greater permeability of barriers in the developing brain represents a specialisation rather than being a reflection of immaturity.