Address correspondence and reprint requests to Dr. N. Bernoud at U 352 INSERM, Biochimie et Pharmacologie, INSA-Lyon, 20 Ave. A. Einstein, 69621 Villeurbanne, France.
Abstract : The passage of either unesterified docosahexaenoic acid (DHA) or lysophosphatidylcholine-containing DHA (lysoPC-DHA) through an in vitro model of the blood-brain barrier was investigated. The model was constituted by a brain capillary endothelial cell monolayer set over the medium of an astrocyte culture. Cells were incubated for 4 h with a medium devoid of serum, then the endothelial cell medium was replaced by the same medium containing labeled DHA or lysoPC-DHA and incubations were performed for 2 h. DHA uptake by cells and its transfer to the lower medium (astrocyte medium when they were present) were measured. When the lower medium from preincubation and astrocytes were maintained during incubation, the passage of lysoPC-DHA was higher than that of unesterified DHA. The passage of both forms decreased when astrocytes were removed. The preference for lysoPC-DHA was not seen when the lower medium from preincubation was replaced by fresh medium, and was reversed when albumin was added to the lower medium. A preferential lysoPC-DHA passage also occurred after 2 h with brain endothelial cells cultured without astrocytes but not with aortic endothelial cells cultured and incubated under the same conditions. Altogether, these results suggest that the blood-brain barrier cells released components favoring the DHA transfer and exhibit a preference for lysoPC-DHA.
We have previously shown that the developing rat brain takes up preferentially DHA esterified in lysophosphatidylcholine (lysoPC), compared with the unesterified form, both being delivered bound to albumin (Thiès et al., 1994). This preferential uptake was specific to the brain, as it was not observed in the liver, heart, or kidney. It was also specific to unsaturated fatty acids, as no difference between the two forms was seen for saturated fatty acids. These results suggest that lysoPC could be an efficient delivery form of unsaturated fatty acids to the developing rat brain, and that a recognition system of lysoPC species might exist, particularly at the level of blood-brain barrier (BBB), which controls the nutrient passage to the brain (Pardridge, 1991). In vivo experiments are unable to detail the role of the different cells involved in the lysoPC metabolism and transport of fatty acids carried by lysoPC to the brain. Dehouck et al. (1992) have developed an in vitro model of BBB by growing brain capillary endothelial cells on the upper side of a filter insert lying on a culture medium covering an astrocyte culture. Under these conditions, the in vitro BBB mimics the in vivo BBB reasonably well.
The present study was designed to study the transport of unesterified DHA and lysoPC-DHA through the in vitro BBB. Our study shows that lysoPC-DHA is preferentially transferred and taken up by astrocytes over the unesterified DHA, this transfer being possibly facilitated by molecules released by the in vitro BBB.
MATERIALS AND METHODS
[4,5-3H]DHA (46.5 Ci/mmol), [9, 10-3H]palmitic acid (16.6 Ci/mmol), and [2-palmitoyl-9, 10-3H]phosphatidylcholine ([3H]PC 89.0 Ci/mmol) were purchased from NEN (Du Pont, France). sn-1-Palmitoyl-2-lysoPC from egg yolk, DHA, palmitic acid, and lyophilized delipidated bovine serum albumin were purchased from Sigma Chemical (St. Louis, MO, U.S.A.).
Bovine brain capillary endothelial cells (BBCECs) were isolated and characterized as described by Méresse et al. (1989), and newborn rat astrocytes were prepared by the method of Booher and Sensenbrenner (1972). The coculture of the two cell types was performed as previously described (Bénistant et al., 1995), using the coculture device reported in Fig. 1. In brief, BBCECs were cultured in a medium containing 10% fetal calf serum, 10% horse serum (Hyclone, Logan, UT, U.S.A.), 2 mM glutamine, and 50 μg/ml gentamicin in Dulbecco's modified Eagle's medium (DMEM). Basic fibroblast growth factor (1 ng/ml) was added every other day to the cultures. For the experiments, cells were used between passages 4 and 7.
Astrocytes were plated onto six-well plates and astrocyte cultures stabilized 3 weeks after seeding. Endothelial cells were seeded onto cell culture inserts (Transwell, 0.4 μM, 75 mm diameter, nucleopore polycarbonate membrane from Costar) coated on the upper side by rat-tail collagen. These inserts were lying on the medium of wells containing astrocyte cultures. Experiments were performed 5 days after confluence of BBCECs. Experiments with BBCECs not cultured with astrocytes were performed using the same culture device.
Bovine aortic endothelial cells (BAECs) were obtained as previously described (Bénistant et al., 1993). They were grown under the same conditions as BBCECs and used at passage 4.
LysoPC preparation was conducted as previously described (Thiès et al., 1994). In brief, PC species with DHA esterified at the sn-2 position were synthesized by acylation of sn-1-acyl-2-lysoPC. The acyl moiety at the sn-1 position was then removed by enzymatic hydrolysis, by the Rhizoppus arrhizus lipase.
Incubations of lysoPC or free fatty acid (FFA)
Culture media were replaced by DMEM 4 h before incubation with labeled substrates. Lyophilized delipidated bovine serum albumin was coated by the labeled lysoPC or the labeled FFA in the soap form (Thiès et al., 1994) and finally dissolved in DMEM. The final concentration of FFA or lysoPC, bound to albumin with a ratio of 0.5, was 5 μM. The upper medium was replaced by DMEM containing the labeled substrate (0.1 μCi) and the incubations were performed for 2 h or an appropriate period of time for kinetic studies at 37°C. At the beginning of incubations, the lower medium was either maintained or replaced by fresh DMEM or replaced by DMEM + 0.2% albumin. In some experiments, astrocytes were removed before the incubations. At the end of incubations, media were removed, and endothelial cells and astrocytes were scraped off.
Lipids from media were extracted by the method of Bligh and Dyer (1959) and those from cells by the method of Folch et al. (1957). The radioactivity of aliquots from total lipids or the lipid extract aqueous phase was determined by liquid scintillation counting. Lipid classes were separated by HPLC under isocratic conditions as described by Bernhard et al. (1994).
Statistical analysis was performed by using Student's t test as appropriate and analysis of variance and Duncan's test. Differences were judged significant at p < 0.05.
The in vivo composition of the interstitial fluid between endothelial cells and astrocytes is unknown. Therefore, we have checked different lower medium conditions to mimic the in vivo passage of unesterified DHA and lysoPC-DHA. A 4-h preincubation period with a medium devoid of serum was used. No change in the BBB permeability was observed as evidenced by unaffected glucose or insulin permeability, regardless of the conditions used (not shown).
After incubation with unesterified DHA, the radioactivity was recovered in the FFA fraction only, in both lower and upper medium (not shown). After incubation with lysoPC-DHA, 10-15% of the lipid radioactivity recovered in the lower medium was located in free DHA and 10% of that recovered in the upper medium was also present in the free form (not shown). This indicates a spontaneous hydrolysis of lysoPC (as observed by incubation without cells), a small part of which is due to cell metabolism. This suggests that the major part of lysoPC-DHA was taken up by endothelial cells and transferred to the lower medium in the intact form. It is noteworthy that the aqueous phase radioactivity after lipid extraction of all compartments studied was very low regardless of the delivery form of DHA used.
Figure 2I shows the percentage of the initial radioactivity recovered in lipids in the different lower media after incubation of unesterified [3H]DHA or [3H]lysoPC-DHA bound to albumin. When the preincubation lower medium was replaced by fresh DMEM at the beginning of incubation, the percentage of radioactivity recovered from lysoPC-DHA was higher than for unesterified DHA (Fig. 2I-A). In contrast, no difference was observed when astrocytes were removed (Fig. 2I-D). When the preincubation lower medium was maintained, the percentage of radioactivity from both delivery forms was enhanced compared with the first situation and the preference for lysoPC-DHA was maintained (Fig. 2I-B), even after removing astrocytes (Fig. 2I-E). When the lower medium was replaced by fresh DMEM containing 0.2% albumin (which binds both unesterified DHA and lysoPC-DHA), the percentage of radioactivity recovered from both delivery forms was greater compared with the other conditions. However, the radioactivity from unesterified DHA was higher than that from lysoPC-DHA (Fig. 2I-C). A preferential recovery of unesterified form was also observed by addition of serum to the lower medium, or by using either the usual culture medium containing serum as lower medium or a preconditioned astrocyte medium containing lipoprotein-poor serum (not shown). Altogether these results show that the preferential recovery of lysoPC-DHA in the lower medium was associated with the presence of astrocytes or that of the preincubation lower medium.
The proportion of radioactivity recovered in astrocytes was parallel to that recovered in the lower medium (Fig. 2II-A and II-B). It is interesting that the preference for lysoPC-DHA was further enhanced [lysoPC-DHA/DHA ratio : 2.43 ± 1.83 (Fig. 2II-A), vs. 1.86 ± 1.33 (Fig. 2I-A) ; 4.84 ± 2.23 (Fig. 2II-B) vs. 2.18 ± 1.88 (Fig. 2I-B)], suggesting that astrocytes can take up to lysoPC-DHA easier than unesterified DHA. Such a preference for lysoPC-DHA was never observed in endothelial cells, the percentage of radioactivity recovered in cell lipids being similar (8 ± 1.5%), regardless of the conditions used for both delivery forms (not shown). This supports the idea of a selective exportation of lysoPC-DHA to the lower medium. When the lower medium contained 0.2% albumin (Fig. 2II-C), more radioactivity from unesterified DHA than from lysoPC-DHA was recovered in astrocyte lipids. Moreover, the radioactivity from lysoPC-DHA was significantly lower than that observed in the absence of albumin (Fig. 2II-C vs. II-A and II-B). The lysoPC-DHA concentration in the lower medium was the highest in the presence of albumin (Fig. 2I-C), and this suggests that albumin is not the best carrier to deliver lysoPC-DHA to astrocytes.
Figure 2III shows the total passage of the molecules through the brain endothelial cell monolayer (transfer to the lower medium and incorporation into astrocytes). As the astrocyte lipid radioactivity was parallel to that observed in the lower medium, the differences in lipid passage on the experiment conditions were analogous to that observed in the lower medium. However, the preference for lysoPC-DHA is accentuated. When the lower medium was fresh DMEM, and astrocytes were removed, the passage of lysoPC-DHA was the lowest compared with the other medium conditions, and not different from that of unesterified DHA (Fig. 2III-D). Therefore, it can be assumed that endothelial cells cocultured with astrocytes but incubated alone did not facilitate the passage of any specific form of DHA. When astrocytes were maintained, the passage of lysoPC-DHA was unchanged, whereas that of unesterified DHA was decreased (Fig. 2III-A vs. III-D), showing a greater passage of lysoPC-DHA in the presence of astrocytes. When the lower medium from preincubation was maintained and astrocytes were removed (Fig. 2III-E), the passage of both delivery forms was enhanced (Fig. 2III-E vs. III-A), and the preferential passage of lysoPC-DHA was also observed. This shows that the preference for lysoPC-DHA was associated with the lower medium as well as with the astrocytes themselves. This suggests that some substances present in the lower medium and released during the preincubation period could participate in the passage of both lipid forms and could be responsible for the lysoPC-DHA preference. When astrocytes were present and the lower medium from preincubation was maintained, the passage was higher for both delivery forms (Fig. 2III-B vs. III-A, III-D, or III-E) and the preference for lysoPC-DHA was maintained (the same comparisons for unesterified DHA were only significant for 2III-B vs. III-A). Conversely, when the preincubation lower medium contained albumin (or serum) and astrocytes were maintained (Fig. 2III-C), the passage of unesterified DHA was higher than that of lysoPC-DHA, although the passage of lysoPC-DHA was the highest in comparison with all other conditions. These results show that the presence of an acceptor such as albumin (or lipoproteins and albumin when the medium contained serum) in the lower medium can modulate the passage of both delivery forms and supports the idea that acceptor(s) for lysoPC, likely different from albumin, are present in the lower medium after preincubation.
Figure 3 shows the radioactivity distribution among the highest labeled lipid classes from endothelial cells and from astrocytes when the preincubation lower medium was maintained. In endothelial cells (Fig. 3A), PC was the most labeled, regardless of the delivery forms. Some radioactivity was also observed in the fraction containing neutral lipids (NL) + FFAs and, to a lesser extent, in ethanolamine phosphoglyceride (EPG). Noticeable radioactivity was also observed in lysoPC after incubation with lysoPC-DHA. In astrocytes (Fig. 3B), PC was also the most labeled, but a higher percentage of radioactivity was recovered from lysoPC-DHA than from unesterified DHA, suggesting a direct acylation of lysoPC. The opposite was observed in EPG where more radioactivity from unesterified DHA than from lysoPC-DHA was recovered, both being higher, however, than that observed in endothelial cells. In contrast to endothelial cells, no significant labeling was detected in lysoPC of astrocytes after incubation with lysoPC-DHA, suggesting that lysoPC radioactivity in endothelial cells could mainly represent the lysoPC trafficking through the cells.
To examine whether the preference for lysoPC-DHA was specifically associated with astrocytes, we have compared the passage kinetics of unesterified DHA and lysoPC-DHA across BBCECs cocultured (Fig. 4A) or not (Fig. 4B) with astrocytes, the preincubation lower medium being maintained. Moreover, to show a possible specificity for the cerebral endothelium, we have studied the passage of the same molecules through a large vessel endothelium by using BAEC cultures (Fig. 4C). When astrocytes were present, the passage was estimated as the sum of lipid radioactivity in lower medium + astrocytes. In this condition, more lysoPC-DHA than unesterified DHA crossed the BBCEC monolayer and the difference increased with time (Fig. 4A). The lysoPC-DHA transfer still increased at 4 h, whereas that of unesterified DHA reached a plateau at 2 h (not shown). Within the first hour, the transfer of lysoPC-DHA and unesterified DHA across the BBCEC cultured alone (Fig. 4B) was similar and slightly higher than that observed when they were cocultured (Fig. 4A). After 2-h incubation, the preferential passage of lysoPC-DHA was also observed (Fig. 4B). It is noteworthy that this was not the case when BBCECs were cocultured with astrocytes and incubations performed after removing the preincubation lower medium and astrocytes (Fig. 2III-D). The passage through the BAEC monolayer (Fig. 4C) of both lipid forms was lower than that observed across BBCECs. Moreover, the two delivery forms were transferred at the same rate regardless of the incubation time. Also, we found that more radioactivity from lysoPC-DHA than from unesterified DHA was recovered in cellular lipids at each time, whereas BBCECs (cocultured or not) incorporated both delivery forms at the same rate and to a lower extent (Fig. 5).
To check the possibility of a specific utilization of lysoPC-DHA, we have studied the transfer of a saturated fatty acid, palmitic acid, esterified or not in a lysoPC, through the BBCECs in coculture when the lower medium from preincubation was maintained. After 4-h incubation, both lysoPC-DHA and lysoPC-16:0 were more transferred than their corresponding unesterified fatty acids (Fig. 6). Moreover, the transfer of lysoPC-DHA was significantly higher than that of lysoPC-16:0. The preferential transfer of esterified fatty acids in lysoPC versus their unesterified form was also apparent at a shorter incubation time. However, until 2 h, no difference was observed between lysoPC-DHA lysoPC-16:0 (not shown).
In the present study, we investigated the passage of unesterified DHA and DHA esterified in lysoPC through an in vitro model of BBB to assess the conditions that mimic the in vivo passage of the two lipid forms.
The concentration of lysoPC-DHA used (5 μM) in these experiments was within the physiological concentration range (Thiès et al., 1994 ; Brossard et al., 1997). Moreover, to facilitate the lipid binding onto albumin and to ensure that there was not free lysoPC in the medium, we used a lipid/albumin ratio of 0.5, although the physiological ratio is 1 (Klopfenstein, 1969). A recent study (Wong et al., 1998) shows that lysoPC/albumin molecular ratios of > 1 with 50 μM final concentration are required to observe cell signaling events (without cell alteration). Therefore, in our experimental conditions, we may exclude any deleterious effect of lysoPC on the BBB cells. For example, with BAECs, or BBCECs cocultured with astrocytes but incubated without these astrocytes, and with fresh DMEM as the lower medium, the passage of lysoPC-DHA was the same as that of unesterified DHA. Therefore, the preferential transfer of lysoPC observed under other conditions (i.e., presence of astrocytes and/or preincubation medium) is likely the result of cell activity in the BBB.
We show that the reconstituted BBB favors the passage of DHA esterified in lysoPC over unesterified DHA when the lower medium (mimicking the intercellular fluid between BBCECs and astrocytes, of which the in vivo composition is unknown) was devoid of serum. Under our conditions, the passage of lysoPC-DHA decreased when astrocytes or the preincubation lower medium was removed. Moreover, the preference was reversed to the advantage of the unesterified form when albumin (or serum) was used as a lipid acceptor in the lower medium. It is noteworthy that both nonspecific transport and transcytosis of albumin are very low in our BBB model (Plateel et al., 1997). In contrast, the differences in the astrocyte uptake of lysoPC-DHA versus unesterified DHA were more pronounced than those in concentration between the two lipid forms in the lower medium. Altogether, these observations strongly suggest that the passage of both delivery forms and the preference for lysoPC resulted from both a substance release (likely carriers with particular affinity for lysoPC) and a preferential uptake of lysoPC by astrocytes. These results suggest the existence of liquid carriers within the intercellular spaces in the brain. Several lymphatic systems are present in the brain (Foldi, 1996), but no information is available on their composition except for those in relation to the CSF. CSF contains very low concentrations of albumin and lipoproteins (Koudinov et al., 1996 ; Knott et al., 1997), which may derive from plasma via the choroid plexus (Knott et al., 1997) and from the blood (Pitas et al., 1987b ; Borghini et al., 1995), respectively. Recent studies (Borghini et al., 1995 ; Koudinov et al., 1996 ; LaDu et al., 1998) have shown that CSF essentially contains lipoproteins in the size range of small low-density lipoproteins or high-density lipoproteins with specific composition, supporting the idea of a specific lipoprotein synthesis and release into the brain as suggested by Pitas et al. (1987b). Astrocytes are able to synthesize different apolipoproteins (Pitas et al., 1987a ; Han et al., 1994 ; Fukagawa et al., 1995 ; Mouchel et al., 1995 ; Patel et al., 1995) and can release nascent lipoprotein particles in their culture medium (LaDu et al., 1998). We therefore suggest that these cells could be responsible for the release of lipid carriers having a special affinity for lysoPC. This agrees with the similar passage of both lipid forms in fresh DMEM as lower medium and in the absence of astrocytes. However, when BBCECs were cultured alone, the passage of both molecules was higher than that observed in coculture. This can be easily explained by the greater permeability of the BBCEC monolayer cultured without astrocytes (Dehouck et al., 1992). Under these conditions, after 2 h, the passage of lysoPC-DHA significantly exceeded that of unesterified DHA (Fig. 4B), suggesting that under coculture conditions, both cell types of the BBB could participate in the release of carriers that favor the passage of lysoPC into the lower medium.
It is noteworthy that a synthesis of apolipoprotein E by cerebral endothelial cells has been reported (Wells et al., 1995). BBCECs not cocultured with astrocytes took up lysoPC-DHA and unesterified DHA to the same extent but exported lysoPC-DHA more effeciently in the lower medium after 2 h. This suggests that there is a control of the transfer. Conversely, although the uptake of lysoPC-DHA was greater than that of unesterified DHA in BAECs, the passage through the endothelial cell monolayer of the two DHA forms was similar, suggesting a paracellular passage.
The preferential transfer of lysoPC-DHA over that of unesterified DHA in the in vitro BBB model agrees with the results obtained by Thiès et al. (1994) in young rats. More generally, our results support the view of a BBB utilization of polyunsaturated fatty acids esterified in lysoPC bound to albumin and resulting from both the lecithin-cholesterol acyltransferase activity (Subbaiah et al., 1994) and a liver release (Mangiapane and Brindley, 1986 ; Graham et al., 1988 ; Brindley et al., 1993). In in vivo experiments of Thiès et al. (1994), the injection procedure combined with the short half-life of lysoPC in blood (2-3 min) suggested that the main part of lysoPC-DHA was taken up by the brain as 2-acyl-lysoPC. This is unlikely in the present in vitro experiments, as a fast isomerization of the initial 2-acyl-lysoPC occurs within the first 10 min in the buffered medium at pH 7.4, yielding ~90% 1-acyl-lysoPC and 10% 2-acyl-lysoPC. The isomerization might not affect the lysoPC uptake by BBCECs, as, in vivo, the brain takes up both isomers at the same rate (Thiès et al., 1992). However, it cannot be ruled out entirely that the isomerization could influence the lysoPC metabolism and its transfer through the barrier.
In vivo experiments also showed that the brain incorporates preferentially unsaturated lysoPC (Thiès et al., 1992). However, in our in vitro kinetics, the transfer of lysoPC-DHA was close to that of lysoPC-16:0 until 2 h incubation and was higher thereafter. These results suggest that the differences observed in vivo are delayed in vitro. This delay could be due to the long time necessary for BBB cells to adapt their metabolism to the absence of serum in the upper and lower media. Moreover, the absence of astrocyte feet, the lower surface exchange, and the large volume of the lower medium might also be parameters responsible for the delay. For both BBCEC and astrocytes, unesterified DHA and lysoPC-DHA were mainly recovered in PC. In astrocytes, more radioactivity was recovered from lysoPC-DHA than from unesterified DHA, in agreement with an efficient lysoPC acylation (Fisher and Rowe, 1980 ; Morash et al., 1989). Esterification of DHA from both unesterified DHA and lysoPC-DHA into EPG of both cells was low, although slightly higher in astrocytes. This suggests that the high degree of DHA esterification into the brain phosphatidylethanolamine from lysoPC-DHA and/or unesterified DHA observed in vivo (Thiès et al., 1994 ; Croset et al., 1996) is mainly related to neuron metabolism.
In conclusion, we have shown that the in vitro model of reconstituted BBB mimics the preferential passage to the brain of lysoPC-DHA over unesterified DHA, although several differences with the in vivo brain uptake are observed. The BBB model allowed showing the preference of astrocytes for lysoPC and the necessary presence of molecules to accept the lysoPC between BBCECs and astrocytes. The latter raises the question of the nature and cell origin of these molecules. This is now under investigation.