Docosahexaenoic acid [DHA, 22:6(n-3)]⁵ and arachidonic acid [ARA, 20:4(n-6)] are the predominant (n-3) and (n-6) polyunsaturated fatty acids (PUFA), respectively, in the mammalian central nervous system (Sastry 1985). DHA is particularly concentrated in the synaptic membranes and the microsomes of grey matter, as well as in the membranes of the rod outer segments of the retina (Tinoco 1982). DHA and ARA are derived from their parent essential fatty acid (EFA) compounds, -linolenic acid [LNA, 18:3(n-3)] and linoleic acid [LA, 18:2(n-6)] through a series of metabolic steps involving desaturation and subsequent chain elongation (Cook 1991). PUFA are integral components of membrane phospholipids, and the 20-carbon compounds are the precursors of the eicosanoids, biologically active compounds that mediate various cellular processes (Kinsella et al. 1990). Because the (n-3) and (n-6) fatty acids (FA) compete for the same desaturase enzymes, the dietary (n-6):(n-3) ratio plays an important role in their metabolism (Arbuckle et al. 1994).

Dietary deficiencies in animals of (n-3) fatty acids during development are associated with reductions of levels of DHA in the brain and retina, with a reciprocal increase in levels of 22:5 (n-6) (e.g., Bourre et al. 1989, Galli et al. 1971, Wainwright et al. 1994a and 1994b). These changes are accompanied by effects on various measures of visual function in rodents and primates (reviewed in Connor et al. 1992). While some behavioral studies have reported deficits in performance on learning tasks in (n-3)-deficient rodents, it is not clear whether these reflect direct effects on learning, or associated changes in sensory or motor capacity, or in motivational factors (reviewed in Wainwright 1992).

The effects of high levels of (n-3) FA have been less frequently studied than those of (n-3) deficiency. Studies of feeding diets varying in (n-3) levels to weanling rats suggest that there is a limit on the incorporation of (n-3) FA into the brain (Anding and Hwang 1986, Mohrhauer and Holman 1963). We previously conducted a dose-response study, feeding diets that varied widely in their (n-6):(n-3) FA ratio, to pregnant and lactating mice, and measured the growth and brain FA composition in their 12-d-old pups (Huang et al. 1992, Wainwright et al. 1992). In this study, we used fish oil concentrate containing eicosapentaenoic acid [EPA, 20:5(n-3)] and DHA, to provide long-chain (n-3). We found that the (n-6):(n-3) ratio of the dam's milk showed a strong linear relationship with that of the diet. However, levels of DHA in brain phosphatidylethanolamine tended to plateau beyond an (n-6):(n-3) ratio of 4.0. While there were no significant effects on body weight, decreasing the (n-6):(n-3) ratio was associated with slightly smaller brains, relative to body weight. In this, and a further study (Huang et al. 1993), we showed that the decline in tissue ARA levels associated with fish-oil feeding could be ameliorated by providing a portion of the (n-6) FA as the post-6-desaturation product, -linolenic acid [GLA, 18:3(n-6)]. However, in contrast to this, in a study in human term infants, addition of GLA to a fish oil-based formula did not compensate for the reduced ARA levels in erythrocytes (Makrides et al. 1995).

Accretion of DHA and ARA in the developing human brain and retina occurs mainly during the third trimester and early postnatal period (Martinez 1992). A study in which the formula of preterm infants was supplemented with marine oil containing high levels of both DHA and EPA showed that although supplementation improved visual acuity (Carlson et al. 1992 and 1993b), it decreased first year growth; poorer growth (regardless of group) was also associated with lower scores of psychomotor development (Carlson et al. 1994). The supplemented group also showed lower blood levels of ARA (Carlson et al. 1993a), suggesting that the effects on growth and development might have been attributable to the reduced availability of ARA as a result of competitive inhibition by the high levels of EPA in the marine oil. This is supported by the results of a subsequent study in which the oil used was high in DHA but low in EPA, and in which growth effects were minor (Carlson et al. 1996). Although plasma ARA concentration was unaffected by diet in this second study, the DHA-supplemented infants did have a lower ratio of ARA to DHA. Length and head circumference were unaffected by DHA supplementation, but the DHA-supplemented infants had a significantly lower weight-for-length than controls at most ages during the first year of life. Thus, if (n-3) FA are provided partially as EPA and/or DHA, it may also be necessary to provide some of the (n-6) FA as ARA (Carlson et al. 1993a). This is supported by a recently published study in rats (Innis et al. 1995) showing that feeding freshwater fish oil, which, like marine oil, contained DHA but had higher ARA and less EPA, raised the levels of ARA in brain lipids to those of the soybean oil-fed controls.

These observations support the importance of understanding the nature of the metabolic relationship between dietary long-chain (n-6) and (n-3) FA during development, and subsequent effects not only on growth and tissue lipid composition, but also on behavioral outcomes. Because marine oils contain high levels of both EPA and DHA, it is difficult to identify which is responsible for the effects on ARA levels. The recent availability of single-cell microbial oils, in which only one particular FA predominates, e.g., DHASCO and ARASCO (Martek Biosciences, Columbia, MD) provides the opportunity to address the unique effects of DHA and ARA, respectively. We therefore used these oils in the present study to assess the effects of a wide range of (n-6):(n-3) ratios on growth, brain fatty acid composition and behavior in mice. In this study, the (n-6) FA were provided either entirely as LA, or as LA in combination with ARA, and the (n-6):(n-3) ratios were adjusted by partial replacement of the (n-6) FA with DHA. The diets were fed to pregnant and lactating mice, and to their offspring for the duration of the study. The levels varied from a very high (n-6):(n-3) ratio of 49 [(n-3) deficient] through a normal ratio of 4.0 to a low (n-6):(n-3) ratio of 0.32. The hypotheses were that spatial learning ability would be impaired by a high (n-6):(n-3) ratio [(n-3) deficiency], that high levels of DHA in a diet with a low (n-6):(n-3) ratio would be associated with growth retardation and adverse behavioral outcomes, and that these latter effects would be prevented by concomitant supplementation with ARA. Outcome measures included preweaning growth and the fatty acid composition of the brain both at weaning and in adult offspring. In addition, adult females were tested on learning performance in the Morris water-maze and on measures of activity and habituation in the open field.

MATERIALS AND METHODS

Table 1. Dietary formulation [View Table]

Table 2. Composition of dietary oils1 [View Table]

The first day was a cued test, using a visible platform. This had black edges and was placed 1 cm above the surface of the water, with a black rubber stopper suspended from a string above it. A white curtain surrounding the tank eliminated all extra-maze cues. Each mouse was given four trials to find the platform, each beginning from the inner wall of the quadrant opposite that housing the platform. The purpose of this test was to assess whether the groups differed in the motor or proximal visual skills necessary to locate and escape to the platform.

The next 4 d constituted the acquisition learning phase, in which a white platform was submerged below the surface of the water in a quadrant different from that used in the cue trial. For each trial, the mouse was placed facing the inner wall of the tank in one of the four quadrants and allowed to swim for a maximum of 45 s. Once the mouse found the platform, it remained there for 10 s before beginning the next trial. If the mouse did not find the goal within that time, it was placed on the platform and allowed to remain there for 10 s. Mice were given four trials per day, starting from each of the four quadrants. The order of quadrant entries was randomized for each mouse on each day of testing. After a 2-d break, the sixth and seventh days of testing were "reminders," conducted as described for the learning trials. The final 3 d of testing consisted of the reversal learning phase, in which the goal platform was placed in the quadrant opposite to the target quadrant in the learning phase, and the animals were trained according to the same procedure as that used for learning trials.

The behavior of the mice was filmed with a video camera and recorded by a computer (software provided by Videomex-V, Columbus Instruments, Columbus, OH), thereby eliminating any possibility of observer bias. The distance travelled and latency to reach the goal were recorded and the average of the four trials per day was used for analysis. The swimming speed was calculated by dividing the distance travelled by the latency. All behavioral testing was conducted during the dark cycle, which is when mice are normally most active (Wainwright et al. 1996).

In the week following the completion of the Morris-maze testing, the mice in the first cohort were tested in a spatial version of the open field, adapted from work done previously in mice (Roullet and Lasalle 1990). This provided information on general levels of activity in a novel setting. The apparatus used was a circular metal chamber (diameter 67 cm × height 27 cm). The floor and wall were painted white. Three square zones (object zones) were demarcated using the video tracking system (Videomex V) and were monitored independently of the rest of the field. Three different objects (L × D × H, 3 × 3 × 4 cm) were fixed in the center of these zones. The mice were tested once daily, for four consecutive days. On d 1-3, the objects were arranged in a "V" configuration; on d 4, they were rearranged in a line. The behavior of the mouse was monitored continuously for 3 min, and measures included total distance travelled in the field and inside the object zones, total moving time in the field, number of entries to the object zones, and total time spent in the object zones.

Further post-hoc analyses comparing all groups with each other were conducted using Tukeys t test (Cody and Smith 1991). The  level was set at 0.05. The number of litters weaned in each group was as follows: LA-(n-3) DEF, 24; ARA-DHA, 22; LA-DHA, 21; ARA-high DHA, 22; LA-high DHA, 19. Discrepancies between these numbers and those reported in the analyses reflect missing data. The relationship between brain weight and brain fatty acid composition was assessed using stepwise multiple regression.

RESULTS

Table 3. Selected fatty acids in brain of B6D2F2 mice at weaning fed diets differing in fatty acid composition and (n-6):(n-3) ratios1,2 [View Table]

Table 4. Selected fatty acids in brain of adult female B6D2F2 mice fed diets differing in fatty acid composition and (n-6):(n-3) ratios1,2 [View Table]

Although dietary ARA was associated with higher levels of brain ARA, levels of 18:2 (n-6) and 20:3 (n-6) were lower. At weaning LA-high DHA mice had slightly more of these latter two FA than LA-DHA mice, and the direction of these differences was reversed in adults. Levels in LA-(n-3) DEF mice were lower than those in LA-DHA mice. Dietary ARA was associated with lower levels of 20:5 (n-3) and 22:5 (n-3). Compared with LA-DHA mice, levels of 20:5 (n-3) and 22:5 (n-3) were higher in LA-high DHA mice and lower in LA-(n-3) DEF mice. As expected, levels of 22:5 (n-6) were high only in LA-(n-3) DEF mice. At weaning, levels of 18:1(n-9) were higher in LA-high DHA mice than in LA-DHA and ARA-high DHA mice. The dietary treatments did not affect the overall amount of total lipid or that of the saturated fats, 16:0 and 18:0. The (n-6):(n-3) ratio was lower in LA-high DHA mice than in LA-DHA and ARA-high DHA mice. It was lower in LA-DHA mice than in ARA-DHA mice in adults, but not at weaning. It was higher in LA-(n-3) DEF mice than in LA-DHA mice.

Stepwise multiple regression of the (n-6) and (n-3) PUFA on brain weight indicated a two-variable model which included statistically independent effects of both ARA and DHA (R² = 0.16). As shown in Figure 6, these relationships were in opposite directions, with brain weight increasing as ARA increased, but decreasing as DHA increased.

DISCUSSION

In this study, the very low (n-6):(n-3) ratio (high levels of dietary DHA) had no effect on litter size at birth or pup birth weight. During the preweaning period, however, the pups from both of the low (n-6):(n-3) ratio groups did not gain weight as fast as those from the litters in which the dams were fed diets with an (n-6):(n-3) ratio of 4.0 or 49.0. This weight differential was reduced substantially after the pups were weaned and began to consume the diets directly. Only the group fed the low (n-6):(n-3) ratio diet without ARA continued to show slightly lower body and brain wet weights than those fed the normal (n-6):(n-3) ratio. This recovery of body growth after weaning suggests that the low (n-6):(n-3) ratio during lactation may have impaired the dams' ability to provide sufficient nutrition for the growth of the developing pups. Interestingly, a recent study in rats has shown that (n-6) FA are necessary during pregnancy and lactation for mammary gland development (Ollivier-Bousquet et al. 1993). In a previous study (Wainwright et al. 1994a), the same population of mice was fed until weaning a low EFA diet with levels of (n-6) FA similar to those in the low (n-6):(n-3) ratio groups in the present study. In this previous study, there were no effects on body weight; thus the growth retardation in the present study cannot be attributed to a simple (n-6) dietary deficiency. This suggests that the problem may be that of a metabolic (n-6) deficiency resulting from high (n-3) relative to (n-6) levels, such that, due to inhibition of 6-desaturase, the conversion of LA to ARA is impeded, and ARA becomes unavailable for growth. Such a feedback inhibition is a well-known consequence of (n-3) FA supplementation (Kinsella et al. 1990), and, in this study, may be due to either DHA itself, or to the elevated levels of EPA generated by retroconversion of the dietary DHA. Although this explanation leaves unanswered the question of why the very high DHA group, which was supplemented with ARA, was also growth retarded before weaning, it may explain the recovery in growth in this group after weaning.

There were no effects of (n-3) deficiency on growth, which is consistent with our previous findings in this mouse population (Wainwright et al. 1994a and 1994b). It should be noted that, apart from growth retardation, at no time during the study did the animals show obvious signs of physical anomalies, despite changes of the (n-6):(n-3) ratio in the diet of more than 100-fold (0.32 to 49). This was somewhat surprising in view of the large amount of DHA being consumed by the pregnant dams (5.5% of dietary energy, or ~3 g DHA/ (kg·d). This would be roughly equivalent to an average person (75 kg) consuming about 2.5 kg of fish oil per day. The dietary levels of ARA in the highest dose group correspond to about 1.8 g ARA/(kg·d). Similar high dose studies in rats have shown little effect on growth and development (Boswell et al. 1996). Indeed, Atkinson and Meckling-Gill (in press) showed that when DHA was provided as 10% of the dietary energy (twice the highest levels in this study) to weanling rats, there was no effect on growth over a 60-d period. One reason for this may be that species differ in their vulnerability to long-chain FA deficiency due to differences in fatty acid metabolism. Comparison of the fatty acid profiles of plasma, red blood cell and liver phospholipids in rats and mice fed the same semipurified diet indicates that ARA levels are considerably higher in rats (Horrobin et al. 1984). Thus, the enhanced ability in the rats to convert LA to ARA may also render them less vulnerable to metabolic inhibition through increased (n-3) levels.

Increasing levels of DHA in the diet increased brain DHA and decreased ARA; mice fed high levels of DHA also had high levels of EPA, supporting a considerable amount of retroconversion in these groups. Addition of ARA to the diet increased ARA; it decreased DHA at high DHA levels, but had no effect on DHA at lower DHA levels. This finding, consistent at weaning and in adult brains, is interesting because the actual amount of ARA in the diet was much higher at the lower DHA level. The outcome of this was that the brains of the group fed the low (n-6):(n-3) ratio diet with very high levels of DHA, and supplemented with ARA, did not differ in terms of the amount of ARA and DHA from the group fed lower DHA levels with a normal (n-6):(n-3) ratio of 4.0. Although brain weight correlated positively with brain ARA levels and negatively with brain DHA levels, the functional significance of these findings remains to be demonstrated, because the magnitude of the changes was very small in relation to the overall variability of the data (R² = 0.06).

Differences in performance on the cued version of the Morris maze reflect differences in the animals' sensory capacity (can they see proximal cues?), motivation (do they want to escape from the water?) or motor capability (can they swim?). There were no significant differences among the dietary groups in terms of the distances they swam to escape, suggesting similar response to the water and visual cues. However, the mice fed the low (n-6):(n-3) ratio diet containing very high levels of DHA without ARA supplementation showed a significantly longer latency to escape, and this was due to their slower swimming speed. Similarly, in both the acquisition and reversal phases of the place version of the maze, diet did not affect the ability of the animals to learn the location of the hidden platform, but, again, those fed the low (n-6):(n-3) diet containing high DHA without ARA swam more slowly. Despite these consistent effects on swimming speed, there were no effects of diet on overall activity in the spatial open field, which is what one might expect if gross motor activity were impaired. One explanation for the effect on swimming, but not on locomotion, is that the smaller size of the mice was related to less physical strength when moving against the resistance of the water. Some evidence against this interpretation is provided by our previous work, in which mice fed an EFA-deficient diet after weaning had body weights 75% that of controls, and yet swam at comparable speeds (Wainwright et al. 1994a). Further studies might extend this investigation by the use of gait analysis, and other more sophisticated measures of motor capabilities.

There were no effects of dietary (n-3) deficiency on distance swum to find the platform in the Morris maze, which is consistent with our two previous studies using this measure in this mouse population (Wainwright et al. 1994a and 1994b). These findings differ from those reported by Nakashima et al. (1993), in which (n-3)-deficient diets in mice were associated with longer latencies on this task. Inspection of their data indicates, however, that the (n-3)-deficient mice were consistently slower, even when their latency to find the platform had reached a plateau. This suggests that differences in motivation to swim, or swimming speed, rather than differences in learning ability, may have accounted for their results. In contrast, in the present study, it was the LA-(n-3) DEF group that swam the fastest. Thus, in this study, despite a range in brain DHA levels of 76-120% of control values, there was no indication that this was related to learning ability as measured by distance covered to reach the platform. Of course, such a statement must be tempered with the awareness that the Morris maze, as used here, may not be capable of discriminating subtle dietary-induced differences in learning ability. The assessment of dietary effects on cognitive function in animals presents complex problems, and Strupp and Levitsky (1995) identify tests of cognitive flexibility as a potentially informative strategy in this regard. Our use of reversal learning in the water maze represents a simple version of such a test, in which again, we failed to show significant differences among the dietary groups.

In summary, in this study mice were fed diets with (n-6):(n-3) ratios ranging from 0.3 to 49, in which the (n-6) and (n-3) components were provided as preformed ARA and DHA, during both the pre- and postnatal periods, and continuing after weaning. Despite effects on brain fatty acid composition, there were remarkably few effects on gestation, growth and development, and learning ability. A small, but significant reduction in body and brain growth was observed in the low (n-6):(n-3) diet groups, and they swam more slowly in the water maze. Whether these findings can be attributed to the high DHA levels directly or are due to its retroconversion to EPA and subsequent effects on down regulating the conversion of LA to ARA remains to be determined. Some support for the latter interpretation is indicated by the fact that the effects in the low (n-6):(n-3) group were overcome by partial replacement of dietary LA with ARA.