Main

There is growing evidence that maintenance of the high level of nervous system DHA is necessary for optimal development and function(1–6). For example, rats fed low levels of n-3 fatty acids have poorer acquisition of maze(7) and brightness discrimination(8) tasks and delayed electroretinographic responses(9) compared with those fed diets containing higher levels of n-3 fatty acids. Rhesus monkeys exhibited a loss in visual acuity when fed a diet low inn- 3 fatty acids relative to those fed an n-3-adequate diet; in this study, the lower level of DHA was directly verified by brain and retinal lipid analysis(10). Premature infants fed formula lacking LC-PUFA have lower levels of plasma and erythrocyte DHA, exhibit increased rod electroretinogram thresholds(11), and have lower visual acuity measured behaviorally(12, 13) or electrophysiologically(13) compared with infants fed formula containing marine oil. Lower cognitive scores due to vegetable oil-based formula feeding have also been reported(14). In addition, intellectual performance of school-age children who had been formula-fed is lower relative to those breast-fed(15), an observation that was confirmed in subsequent studies(16, 17). With regard to the latter, it should be noted, however, that there may be other intervening variables responsible for this effect in these nonrandomized trials in addition to nutritional status.

Many of these studies have been performed with formulas or diets using corn or safflower oil as the fat source, with very high LA to LNA ratios and no 20-22 carbon n-6 or n-3 LC-PUFA such as AA or DHA. However, Birch et al.(13) observed that a soy-based formula containing an LA/LNA ratio of 7.7:1 and no LC-PUFA did not support visually evoked potential responses as well as did human milk or formula with added DHA. A recent clinical trial has demonstrated improved visually evoked potential acuity in term infants receiving formula supplemented with marine oil compared with those receiving conventional formula(18). Most infant formulas today lack the LC-PUFA present in breast milk and contain LA and LNA as the only polyunsaturated fatty acids. The above studies suggest that DHA must be added to formulas to support optimal neural development.

However, few studies have examined the possibility of lowering the dietary LA/LNA ratio without the addition of DHA or other n-3 LC-PUFA, to support the proper accretion of DHA during development. This approach has practical significance to the formula industry, if it can be shown that the addition of LC-PUFA is not necessary. Arbuckle and Innis(19) reported that brain and retinal DHA accretion requirements in the piglet can be met without the supply of dietary LC-PUFA, when LNA is supplied as 1.7 energy% (en%) and the LA/LNA ratio is 4:1. Clarket al.(20) observed that lowering the LA/LNA ratio in formula lacking LC-PUFA resulted in increased n-3 LC-PUFA including DHA, in the erythrocytes of term infants. However, the levels of erythrocyte DHA in these formula-fed subjects remained significantly lower than those of their breast-fed counterparts.

Questions regarding the possible essentiality of DHA for normal human brain development may have particular significance for premature infants. LC-PUFAs such as DHA accrue rapidly during the periods of brain development which comprise the third trimester and early postnatal period in human term infants(21). In very premature infants, on the other hand, this period occurs postnatally, when all nutrients must be provided in the formula. Therefore, the precise profile of LC-PUFA in a formula that will produce optimal brain fatty acyl composition and function requires empirical knowledge of the relationship between dietary fatty acid and LC-PUFA deposition in the developing brain.

To examine the CNS effect of altering the dietary LA/LNA ratio without the use of LC-PUFA, the present study employs artificial rearing during rat early development. In the AR model, as in the infant primate or pig models, the infant nutrient source can be directly and precisely manipulated without the confounding influence of LC-PUFA, such as DHA, present in the maternal diet and mammary processes. However, unlike the case of the newborn monkey or, to a lesser extent, the piglet, the brain of the newborn rat is at a stage of development that is roughly analogous to that of the premature infant(22). Furthermore, retinal development in the rat is occurring primarily during the early postnatal period(23, 24). Although the AR method has been used extensively in the study of early nutrition and metabolism(25–29) including lipid nutrition(30), it has not been applied to the questions of LC-PUFA and brain development in the very premature infant Brain growth(22) and retinal development(23, 24, 31) in the rat occur to a great degree in the postnatal period before the age of weaning. The model permits the evaluation of dietary influences on development without the confounding influence of maternal diet and mammary processes, which includes the presence of DHA and other LC-PUFA. In addition, the model approximates the extremely premature infant who is exposed to the unnatural stress of maternal separation, nasogastric formula feeding, testing, and surgical procedures.

METHODS

Institutional review. This study was performed in conformity with the regulations of the Animal Care and Use Committee of the National Institute on Alcohol Abuse and Alcoholism, National Institutes of Health.

Fat sources. Saturated fat was supplied primarily as medium chain triglycerides (MCT oil, Mead Johnson, Evansville, IN). Monounsaturated fat (and some longer chain saturated fat) was supplied as triglyceride from macadamia nut oil (Sigma Chemical Co., St. Louis, MO). Polyunsaturated fat was supplied from a mixture of triglycerides and ethyl ester oils depending on the LA/LNA ratio desired (Table 1). Corn oil (ICN Biochemicals, Irvine CA) and LA ethyl ester (NuCheck Prep Inc., Elysian, MN) were the primary sources of LA. Linseed oil (ICN Biochemicals) and LNA ethyl ester (NuCheck Prep Inc.) were the primary sources of LNA. The saturate to monounsaturate to polyunsaturate ratio was held constant between AR diets at 60:30:30 g of fatty acid/L of diet.

Table 1 Rat milk substitute dietary components
Full size table

Rat milk substitute preparation. Dietary protein was supplied from a bovine whey powder containing sodium caseinate (27/73, wt/wt, bovine whey/sodium caseinate, clinical product SW 9112, Ross Laboratories, Columbus, OH). It was provided as a gracious gift from Dr. John Edmond. This protein source contained 27.8 g of fat/kg of powder, of which <3.0% of the total fatty acids were essential fatty acid, mainly LA, with no detectable LC-PUFA present. A premilk base was formed by dissolving 134 g of SW 9112 powder in 1 L of deionized water at 40 °C. The mixture was then refrigerated at 5°C overnight. The carbohydrate, vitamin, mineral, and micronutrient components listed in Table 1 were then added to 900 mL of this premilk base the following day in the order listed. The dietary oils were individually weighed and premixed to achieve LA/LNA ratios of 10:1, 1:1, and 1:12 (Table 1). The oils were then added last (120 g) to the RMS mixture. The complete RMS mixture was then homogenized using a Polytron, separated into 50-100-mL glass vials, degassed, sealed under nitrogen gas, and stored at -20 °C until the day of use. No detectable amounts of 20 or 22 carbon PUFA were found in the complete RMS diet. Although some of the unsaturated lipid components were added in the form of the ethyl esters, it has previously been shown that their effect on lipid composition is similar in the esterified or triglyceride forms when fed over an extended period of time(32, 33).

Animal care. Timed-pregnant Sprague-Dawley rats (Taconic Farms, Germantown, NY) were obtained at 10-14 d of gestation where they had been fed NIH31 chow. Dams were housed individually in polycarbonate cages with hardwood chip bedding and were fed NIH31 (not autoclaved) rat chow ad libitum. The nutrient composition of this chow has been previously described(34). This chow contains fish meal as a protein source; its fatty acyl composition (wt%) was as follows: 16:0 (14.4), 18:0(3.2), 18:1n-9 (20.7), 18:2n-6 (43.6), 18:3n-3(4.4), 20:4n-6 (0.2), 20:5n-3 (2.3), 22:5n-3(0.4), 22:6n-3 (2.8). All rat pups were dam-reared for the first 4 d of life. On the 5th day, the pups were separated, matched according to litter and sex, and randomized to one of the four groups studied. Dam-reared litters were culled to 10-12 pups per dam.

Artificial rearing procedure. A method of AR described by Diazet al.(25) was modified for our use. Five-day-old neonatal Sprague-Dawley rat pups underwent AR until 19 d of age. Pups randomized to AR underwent a gastrostomy procedure as described by Westet al.(35) with inhaled isofluorane anesthesia. AR pups were then fed by intermittent infusion of RMS by an automated syringe pump. Feedings were given as 10-min infusions each hour over a 24-h period. The total volume of formula given in milliliters was one-third the pup weight in grams, updated daily. Many of the procedures used in AR and for the manipulation of the fatty acid content of RMS have been previously described by Ward et al.(36).

A percutaneous gastrostomy was placed, and artificial rearing was accomplished according to the method of Messer as modified by Hall(37) and Diaz et al.(25). The RMS was delivered to each rat pup in the following manner. A 12-mL syringe(Monoject no. 8881-512910, Sherwood Medical, St. Louis, MO) containing the specified daily volume of RMS was placed on a multisyringe infusion pump (no. 55-4899, Harvard apparatus, Southnatick, Mass). The pump was cycled with an electronic timer (Chron Trol XT-4, Chron Trol Corp., San Diego, CA). The syringe was capped with a 22-gauge syringe needle (Precision Glide no. 305155, Becton Dickinson, Parsippany, NJ) to which a 3-ft length of polyethylene tubing (0.58 mm inside diameter, PE-50, Becton Dickinson) was attached. The other end of the 3-ft long tube was connected to 0.28-mm inside diameter, 0.61-mm outside diameter (PE-10, Becton Dickinson) gastrostomy tubing (10 cm in length) implanted in the rat pup. Between d 5 and 12 the AR rat pups were housed in round plastic containers with hardwood chip bedding. Two 12-inch diameter containers, each housing six pups separated by nylon screening, floated in a water bath (no. W-1 ICS warmer, Thermocare Inc., Incline Village, NV) maintained at 36-38 °C. This arrangement allowed the pups to smell and contact each other in a limited fashion without tangling feeding lines. On d 12, the pups were housed individually in 1000-mL glass beakers (Corning Glass Inc., Corning, NY) with hardwood bedding, with the containers floating in a temperature-controlled water bath. Survival was similar in all of the AR groups (61%).

Tissue harvesting. The pups were killed at 19-20 d of age, corresponding to the date of natural weaning. Each pup was given inhaled isofluorane anesthesia until unconscious and respiratory efforts had ceased. This was followed immediately by blood sampling via left ventricular needle aspiration, followed by left ventricular saline perfusion and decapitation. The brains and eyeballs were removed, weighed, and placed in ice-cold, buffered saline (DPBS, Advanced Biotechnologies, Columbia, MD). The retinas were then dissected, and the two retinas from each animal were pooled and stored in methanol under a nitrogen atmosphere at -80 °C within 24 h. Lipid extraction was completed within 1 wk.

Lipid analysis. Brain or retina samples were homogenized in 1 mL of methanol containing 50 μg of butyl-hydroxy toluene. A total lipid extract was obtained using the method of Bligh and Dyer(38). A triglyceride internal standard (Tritricosanoin, NuCheck Prep Inc.) was added in a known amount to each sample before total lipid extract formation. A portion of the total lipid extract collected was then transmethylated using 14% wt/vol BF3(39). The fatty acid methyl esters were then separated and analyzed by capillary gas chromatography using a 30-m length, 0.25-mm inside diameter capillary column(DB-FFAP no. 122-3232, J&W Scientific) on an HP 5890 gas chromatograph fitted with a flame ionization detector. The injector and detector temperatures were set at 240 °C, and the initial oven temperature was 130°C and was programmed to rise at 4 °C/min to 175 °C, and then at 1°C/min to 215 °C. Hydrogen was used as the carrier gas at a linear velocity of 50 cm/s. Fatty acid methyl esters were identified according to retention times compared with mixtures of authentic fatty acid methyl esters(e.g. standard no. 411, NuCheck Prep Inc.). The percent of total fatty acids and absolute fatty acid concentration (micrograms/g of tissue) of the tissue analyzed were then calculated.

Statistical analysis. The number of animals per dietary group(n) was a minimum of 12. Differences in the mean level of each fatty acid between dietary groups were examined using one-way analysis of variance using Statview 4.01 for the Macintosh (Abacus Concepts, Berkeley, CA). Fisher's least significant difference was used with a Bonferroni correction. Differences with p < 0.005 after correction were considered significant.

RESULTS

Growth. There were no significant differences in the mean weight of each experimental group at 5 d of age. The average weight of AR animals at 19 d of age was 37 ± 0.5 g, in agreement with previous reports on artificial rearing(26, 28, 40, 41). Overall, the survival of the AR animals was 61%; others report AR survival between 60 and 100%(28, 36, 42). Survival of dam-reared animals was 100%. There were no significant differences in weight between the various AR groups at 19 d of age. However, the average weight of dam-reared pups was significantly greater than the average weight of the AR pups at 19 d of age (48 ± 6 g versus 37 ± 0.5 g). The brain weight to body weight ratios were higher in all AR groups compared with the dam-reared group.

Lipid analysis. The fatty acyl composition of brain(Table 3) and retinal (Table 4) tissues in each of the dietary groups examined are given. The total fatty acid concentration in each experimental group was not significantly different. The percentage of DHA in the brain of the 10:1 and 1:1 AR groups was significantly lower than that in the dam reared group. However, the level of DHA in the 1:12 group did not differ significantly from that in the dam-reared group. The DHA levels increased significantly between each of the AR dietary groups as dietary LNA was increased, indicating a dose-response relationship. Also, the levels of the other n-3 fatty acids, LNA, 20:5n-3 and 22:5n-3, were elevated in the 1:12 group. The brain level of AA in the AR animals was the same as the dam-reared pups only in the 1:1 group. Interestingly, the level of brain AA was elevated in the 10:1 group. There was a fairly marked decline in brain AA in the 1:12 group. The levels of othern- 6 fatty acids including 18:2n-6, 20:2n-6, 22:4n-6, and 22:5n-6 were all significantly decreased in the 1:12 group with respect to the dam-reared values. Also, there was a significant increase in the monounsaturated fatty acids, particularly 18:1n-9 in this group. All of the 20 and 22-carbon n-6 fatty acids were elevated above the dam-reared levels in the 10:1 group.

Table 3 Comparison of brain fatty acid composition of rat pups fed varying LA to LNA ratios*†
Full size table
Table 4 Retinal fatty acid composition of rat pups fed varying LA to LNA ratios and dam-reared controls
Full size table

The DHA levels in the retina of all AR dietary groups were significantly lower than that of the dam-reared group, and there were no significant differences between any of the AR groups. Although the retinal DHA level was not supported at the dam-reared level in the highest LNA group (1:12), the levels of all of the other n-3 fatty acids in the retina were elevated by this dietary treatment. Conversely, all of the retinaln- 6 fatty acids declined in the 1:12 group, most of them significantly. There was an elevation in 18:1n-9 in this group as well. In the pups receiving the highest LA (10:1), there was a significant elevation of retinal 20:3n-6, 22:4n-6, and 22:5n-6. The 22:5n-6 elevations are of significance because they are a sensitive indicator of inadequate n-3 fatty acid supply(1–4, 36). In both the brain and retina, there was a significant elevation in the level of 22:5n-6 in the 10:1 and 1:1 diets. However, the high 18:3n-3 supply in the 1:12 diet was capable of decreasing the 22:5n-6 relative to the dam reared group in the brain; this was consistent with the maintenance of the 22:6n-3 in the brain but not the retina with this diet.

DISCUSSION

This study was designed to test the hypothesis that high levels of LNA could support the same neural levels of DHA as are found in dam-reared rat pups fed milk containing preformed DHA. When the LA/LNA ratio was lowered to the fairly extreme ratio of 1:1, the DHA levels in the brain and retina were significantly lower than the levels in dam-reared pups. Even at the very extreme LA/LNA ratio of 1:12, the retinal DHA was not maintained at a level comparable to that of dam-reared pups, and the levels of AA and othern- 6 LC-PUFA were lowered. This suggests that there is no level of LNA that is capable of supporting the level of neural DHA found in the dam-reared rat pups. This analysis was carried out at the end of the AR period when the pups were 19 d old, and it is possible that the levels of DHA may have “caught up” for the AR animals had the equilibrium levels been attained at adulthood. Nevertheless, the first 3 wk of life in the rat are very active periods of brain growth and neuronal formation(22–24, 31), and it is likely that a deficiency observed at this time point in one of the principal structural components of neuronal membranes will have adverse consequences.

This analysis assumes that the levels of brain and retinal DHA in dam-reared pups are valid reference points. It must be noted that, due to the presence of fish meal protein in the chow that was fed to the dams, the levels of their milk n-3 LC-PUFAs were relatively high(Table 2). Human milk levels of DHA in North America and Europe are in the range of 0.1-0.4% of total fatty acids(43–45), but are 1.4% in Inuit women(46) and may be raised to 1.9% in American women after as little as 8 d of a high level of fish oil consumption(47). Thus, the brain and retinal levels of DHA in our rats may have been elevated over the levels associated with a more modest dietary DHA supply. In consideration of this point, it must first be pointed out that our chow was not unusual. For example, the n-3 LC-PUFA content in another recent study of rat milk fatty acyl content was over 5%(48). NIH07 chow also contains fish meal and has ann- 3 LC-PUFA content of about 5% of total fatty acids (N. Salem, unpublished results). Perhaps of greater importance is the similarity of neural DHA levels in our dam-reared pups to those in other studies when lowern- 3 LC-PUFA was fed. For example, Sanders et al.(49) found that a DHA level of 15.5% in forebrain phospholipids in 22-d-old rat pups when fed a butter/lard-based diet that had only traces of n-3 LC-PUFA and 0.7 wt% LNA coupled to a low LA/LNA ratio. Although this level was somewhat higher than our dam-reared level(14.1%), this was probably due to their sampling of forebrain rather than whole brain, as in our study, and the slightly greater age of their animals. Also, although they studied total phospholipids and we studied the total lipid extracts, the total lipid extract is very similar to the phospholipid fraction in the brain and retina, as other acylated lipid classes represent a small fraction of the total lipid. Although the relatively high milk DHA in our chow-fed dams may have influenced the nervous system DHA in our AR pups, the use of a 14 wt% level for weanling rat brain DHA as a reference point appears reasonable.

Table 2 Fatty acid composition* of artificial rat milk substitutes and rat breast milk†
Full size table

Still, it is possible that there was some elevation of the nervous system DHA levels due to the fish meal-derived lipid in the chow diet fed to our dams. It has been shown by Yeh et al.(50), for example, that a 20-wt% menhaden oil-based diet is able to elevate brain DHA levels relative to a 20-wt% corn oil-based diet. However, this is a much higher level of long chain n-3 PUFA than that contained in our chow diet, and the comparison is made to a vegetable oil-based diet devoid of DHA. At the very least, our data would indicate that a diet containing 1.1% DHA(stomach contents of dam-fed pups) is more efficient in supporting DHA deposition in the nervous system than a 25% LNA (1:12 AR diet); it is thus more than 20-fold more efficacious. This is consistent with the previous findings of Sinclair(51, 52) where radiolabeled LNA incorporation into rat brain DHA was compared with that of preformed DHA. This is consistent also with the classic study of Mohrhauer and Holman(53) of rat brain acyl composition when varying levels of a single essential fatty acid were fed in which the highest level of brain DHA was obtained when LNA was 0.32 en% and was not further elevated at 9.42% LNA. Furthermore, Anderson et al.(54), in a study of chicken development, observed that LNA was ineffective relative to eicosapentaenoic acid and DHA in restoring nervous system DHA aftern- 3 deficiency.

Our observation that, at the highest level of LNA, the brain DHA level was the same level as that of the dam-reared pups but that the retinal level was still lower in the AR pups indicates selectivity of this effect and supports the view that even this very high LNA intake is inadequate for the complete support of nervous system essential fatty acid accretion. The drop in both brain and retinal AA in the 1:12 LA/LNA group would also indicate that this is a suboptimal milk substitute lipid composition.

If it is accepted that increasing LNA and the LNA/LA ratio is not able to achieve a balanced, proper level of nervous system AA and DHA in the rat, it is of importance to consider the generalization of this finding to other mammalian species. A variety of experimental approaches has led to the general acceptance that the rat is one of the most capable mammals with respect to essential fatty acid metabolism. For example, Willis(55) has complied the liver microsomal δ-5 desaturase activity profiles of five species for AA formation. He concludes that the δ-5 desaturase is active in the mouse and rat but relatively slow in the guinea pig, rabbit, and human. Very recent studies of human infant essential fatty acid metabolismin vivo by Salem et al.(56) support the view that essential fatty acid human metabolism, although present, is relatively slow, as it is apparently inadequate to support neural AA and DHA accretion. Moreover, direct studies of increasing the LNA/LA ratio in infant formula have demonstrated that values of 3.2:1(20) or 4.8:1(57) do not produce the same levels of infant plasma or erythrocyte DHA as those in breast-fed infants although compromisingn- 6 LC-PUFA composition.

If it is accepted that human essential fatty acid metabolism is not more efficient than that of the rat, it follows that infant formulas designed like those in our study would not support proper human nervous system accretion of LC-PUFAs. Moreover, even the intermediate ratio of LA/LNA of 1:1 would not be permitted under the guidelines established by international bodies that have considered the issue for infant formula(58, 59). Very recently, an international society of lipid chemists have recommended an infant formula composition in which the n-6/n-3 ratio be within the range of 5:1 to 10:1 and contain AA and DHA(60). For practical purposes of infant formula design, it would follow that the inclusion of longer chain n-3 PUFA such as DHA is obligatory for the optimal support of neural DHA levels such as are obtained in the breast-fed case from a well nourished mother. Becausen- 3 fatty acids may, in some cases, antagonize AA composition(61), AA status appears linked to infant growth(62), and AA subserves important functions in nearly every peripheral organ system, it is suggested that AA also be added to infant formula along with DHA. The policy of the FAO/WHO expert committee that infant formula should mimic the composition of human milk(63) would thus be supported by our study.

Although only lipid compositional features of the nervous system were measured in the present study, it is relevant to consider whether the magnitude of the differences observed are of functional significance. The magnitude of the DHA reductions in the 1:1 diet were of the order of 8 and 11% in the brain and retina, respectively, and the 10:1 diet produced losses of about 16% in both organs. In many of the animal studies involvingn- 3 fatty acid deficiency, which typically involve more than one generation of n-3 dietary limitation, rather drastic losses in neural DHA were associated with poorer performance on tasks related to neural function (for reviews, seeRefs. 1–6). Paradoxically, it is the case of premature infants that provide some of the clearest evidence of functional losses in visual(11–13) and brain(13–18) performance-related measures associated with vegetable oil-based formulas. In separate studies, it has been demonstrated in autopsy studies of infants, who died of sudden infant death within the first 43-48 wk of life, that there is a loss in brain DHA associated with feeding infant formulas relative to a breast-fed reference point(64, 65). In the study of Farquharsonet al.(64), the average losses in brain DHA was 22-23% for two of the formulas studied. Makrides et al.(65) observed a 12% loss in brain cortex DHA in formula-fed infants. Carlson et al.(66), Uauyet al.(11), and Hoffman and Uauy(67) were able to demonstrate losses of DHA in the circulation in their premature infant studies. However, the DHA status of bloodstream components do not necessarily reflect that of the nervous system(68). Moreover, because the n-3-deficient formulas studied by Farquharson et al.(64) are similar in many respects to those used in the studies of infant nervous system function, it appears that 12-23% losses of DHA are sufficient to produce losses in function. It should also be noted that higher level neural functions that may be the most affected by DHA deficiency such as cognition, being more highly developed in the human, can in principal be assessed in a more sensitive manner than in rodents; this may help to explain the observations of functional effects of smaller DHA deficits in the human.

In summary, this study has demonstrated that very high levels of dietary LNA in the form of very low LA/LNA formulas cannot support the nervous system composition of DHA and AA found in a dam-reared reference pup during early development where the dam is fed a diet containing appreciable levels of long chain n-3 fatty acids. We have argued that, because human essential fatty acid metabolism is not faster than rat metabolism, and is believed to be markedly slower, that these data are relevant to the design of human formulas. Our data lend further support to the suggestion of the FAO/WHO(63) that the fatty acyl composition of infant formula be modeled after the composition of human milk and, specifically, that it contain comparable levels of AA and DHA.