In vivo conversion of 18- and 20-C essential fatty acids in rats using the multiple simultaneous stable isotope method.

An important question for mammalian nutrition is the relative efficiency of C18 versus C20 essential fatty acids (EFAs) for supporting the tissue composition of n-3 and n-6 pathway end products. One specific question is whether C22 EFAs are made available to tissues more effectively by dietary alpha-linolenic acid (18:3n-3) and linoleic acid (18:2n-6) or by dietary eicosapentaenoic acid (20:5n-3) and dihomo-gamma-linolenic acid (20:3n-6). To address this question in a direct manner, four stable isotope compounds were given simultaneously in a novel paradigm. A single oral dose of a mixture of 2H5-18:3n-3, 13C-U-20:5n-3, 13C-U-18:2n-6, and 2H5-20:3n-6 was administered to rats given a defined diet. There was a preferential in vivo conversion of arachidonic acid (20:4n-6) to docosatetraenoic acid (22:4n-6) and of 22:4n-6 to n-6 docosapentaenoic acid (22:5n-6) when the substrates originated from the C18 precursors. However, when the end products docosahexaenoic acid (22:6n-3) or 22:5n-6 were expressed as the total amount in the plasma compartment divided by the dosage, this parameter was 11-fold greater for 20:5n-3 than for 18:3n-3 and 14-fold greater for 20:3n-6 than for 18:2n-6. Thus, on a per dosage basis, the total amounts of n-3 and n-6 end products accreted in plasma were considerably greater for C20 EFA precursors relative to C18.

Thus, the suggestion has been made that it is desirable to bypass this initial desaturase step in supplying precursor fatty acids in the diet to support the composition of the metabolic end products of the n-3 and n-6 pathways. This may be particularly important during early development, as, for example, 18:3n-6 or 20:3n-6 have been added to infant formula or artificial rat milk in place of 20:4n-6 (21,22).
The purpose of this study was to evaluate the in vivo metabolism of dietary 18:3n-3 compared with that of 20:5n-3 and of dietary 18:2n-6 compared with that of 20:3n-6. We used stable isotope markers as pioneered by Emken and colleagues (23,24) and used by several other groups for human (25)(26)(27)(28)(29)(30) or nonhuman primate (31) stable isotope tracer studies of 18:2n-6 and 18:3n-3 metabolism. Here, we demonstrate that the direct longitudinal comparison of four isotopes is possible when given simultaneously to the same individual. This analytical technique has been described and termed multiple simultaneous stable isotopes (MESSI) (32). It uses several stable isotopic fatty acids given simultaneously and labeled with either deuterium or carbon-13 to yield isotopomers of differing mass values, which can then be independently measured with a highly sensitive, negative chemical ionization gas chromatography-mass spectrometry measurement technique. This approach has the advantage that identical experimental conditions are used for each isotopic fatty acid to be compared as they are metabolized in the same animals at the same time.

Animals and diet
This study was performed after a protocol was approved by the National Institute on Alcohol Abuse and Alcoholism Animal Care and Use Committee, National Institutes of Health. Three week old, weaning male Sprague-Dawley rats (murine pathogen-free; Taconic, Germantown, NY) were purchased and group housed in our animal facility under conventional conditions. The animal holding facility had a 12 h (7 AM to 7 PM) light cycle, a temperature of 22 Њ C, and a relative humidity of ‫ف‬ 55%. Rats were fed pelleted food and water ad libitum with a custom diet (custom diet 100509; Dyets, Bethlehem, PA) based on the AIN-93G (33) standard as detailed in Table 1 . The diet was modified with respect to its fat components to yield a fatty acyl distribution of 40% saturated fat, 42% monounsaturates, 15% 18:2n-6, and 3% 18:3n-3; no longer chain (C20 and C22) EFAs were added to the fat mixture.
After maintenance on this diet for 9 weeks, average body weights were 418 Ϯ 37 g. One day before the start of the experiment, seven rats were singly housed, fasted for 12 h (2 PM to 2 AM), and then allowed their day's ration of food over a 5 h period. A mean weight of 9.3 Ϯ 1.7 g of food (containing ‫ف‬ 28 mg of 18:3n-3 and 140 mg of 18:2n-6) was consumed during this time. Rats were then dosed with an ethyl ester mixture of 32 mol of 2 H 5 -18:3n-3, 2.9 mol of 13 C-20:5n-3, 155 mol of 13 C-18:2n-6, and 2.1 mol of 2 H 5 -20:3n-6 in olive oil in a total of 0.4 ml and returned to their home cages. The dosing was done by gavage using a 16 gauge, 3 inch long animal feeding needle attached to a 1 ml syringe. An aliquot of the isotope mixture was saved for quantitative fatty acid analysis. Animals were then given access to food 4 h after dosing and maintained for 30 days with ad libitum access to the custom diet and water.
All solvents were of gas chromatographic or higher grade except acetonitrile, which was HPLC grade. Other chemicals were of analytical-reagent grade. They were commercially purchased and used without further purification.

Repeated blood sampling from the rat caudal lateral vein
To habituate the rats to the Plexiglas restraint cylinder to be used for blood collection, animals were placed in the chamber for 2 min/day over a 10 day period before the start of the experiment. To achieve vasodilation to promote bleeding, animals were placed on a warm water bottle and their tails placed into water at 50 Њ C for ‫ف‬ 1 s before sampling, and then gentle massage was applied to the tail to aid in blood flow. Blood (0.1-0.2 ml/sample) was then collected using a heparinized 23 gauge, 1 inch long needle (without adaptor) via lateral vein puncture as described (34). Skilled personnel could withdraw 0.2 ml of blood in ‫ف‬ 1 min. Sampling started from the distal end of the tail and then gradually moved forward with alternative sampling of first the right and then the left lateral veins, so that the same puncture  11,14,18,23,27, and 30 days. Baseline samples were taken 1 day before dosing. One control animal gavaged with olive oil was periodically sampled to confirm that the olive oil vehicle did not alter the baseline for the various GC-MS signals to be measured. Blood samples were kept on melting ice and then centrifuged in 1.5 ml Eppendorf tubes for 10 min at 1,250 g . Plasma was removed, quickly frozen using dry ice, and then stored at Ϫ 80 Њ C until analysis.

Total lipid extraction, derivatization reactions, and instrumental conditions
Both labeled exogenous fatty acids and endogenous fatty acids were analyzed in each plasma sample at every time point by GC-MS in negative chemical ionization analysis (32,35,36) and gas chromatography-flame ionization detection (36,37), respectively, as described previously. Calibrations were performed for these mass spectral data as described in the accompanying paper (36).

Data analysis and calibration
The unlabeled fatty acid values in the samples were calculated by comparing the integrated areas of the fatty acid peaks on the gas chromatograms with that of the internal standard. Fatty acid weights were converted to moles using the respective molecular weights for each compound. Data are expressed as means Ϯ SEM (n ϭ 7) as nanomoles of fatty acid per milliliter of plasma.
To express mass spectral data in terms of nanomoles, the peak areas of each labeled compound as detected by selected ion monitoring were compared with that of the internal standard. The standard curves for the various compounds were able to calibrate a range of sample amount loaded onto the GC column from ‫ف‬ 0.003 to 1.7 pmol. This process is described in detail in the accompanying paper for 2 H 5 -18:3n-3, 13 C-U-18:2n-6, 13 C-U-20:5n-3, and 2 H 5 -20:3n-6 standard curves presented (36). The calibration curves for the in vivo metabolites from the four precursors were derived from those made with the isotopic precursors and unlabeled EFAs.
The maximal concentration (C max ) and time point of the maximal concentration (T max ) for each labeled fatty acid in plasma was obtained directly from the concentration-time course curves. The half-time corresponded to the time that the plasma concentration decreased to half its maximal value, C max, on the concentration-time course curve. The maximal percentage of dose (D max ) expresses the percentage of the oral isotopic dose observed in plasma, and it was calculated as follows: [C max ϫ total plasma volume ϫ dose Ϫ 1 ] ϫ 100. The maximal enrichments (E max ) were calculated at these time points by dividing the concentration of the stable isotope-labeled compound of interest by the concentration of the unlabeled endogenous compound. Area under the curve (AUC) was calculated using a trapezoidal method starting from 0 h, with the curve drawn between each succeeding time point until such time as the signal could no longer be reliably detected (38)(39)(40). A two-sided, pairwise Student's t -test was applied to detect significant differences ( P Ͻ 0.05) using Excel 2002 for Windows 2000 Professional (Microsoft, Seattle, WA).

Diets and animals
To have a well-defined metabolic experiment involving fatty acids, the diet must be carefully controlled, particularly with respect to its EFA content. In this study, a semisynthetic diet was used based on the AIN-93G standard (33). The macronutrients of protein, carbohydrate, and fat represented ‫ف‬ 19, 56, and 23% of calories, respectively. The fat content was 10 wt% and was constructed to supply ‫ف‬ 15% of the fatty acids as 18:2n-6 and 3% as 18:3n-3 to produce a 5:1 ratio of n-6/n-3 fatty acids ( Table 1). The remainder was composed of approximately equal amounts of saturated and monounsaturated fatty acids. Animals were allowed to equilibrate their tissue lipid composition with this diet for 9 weeks before the experiment. Adult male rats 12 weeks of age with a mean body weight of 418 Ϯ 37 g were used. After dosing of the stable isotope mixture and frequent blood sampling, the mean body weights of the rats declined within the first 24 h by 4.7 Ϯ 0.4 g. Accurate sampling times were maintained throughout the study with a coefficient of variation of Ͻ 5% over the first 36 h and Ͻ 0.5% thereafter.

Multiple ion chromatograms
To demonstrate that the multiple isotope tracer technique used here is valid, selected ion chromatograms are presented for some of the principal metabolites ( Fig. 1 ). It is demonstrated that for each metabolite, the endogenous (unlabeled), deuterium-labeled, and carbon-13-labeled compounds of the same fatty acid can be independently measured in the same animal. For example, in Fig. 1A, the endogenous 22:5n-3 was detected at the M Ϫ PFB mass value of 329; this peak is enormous compared with the trace signals representing the stable isotope-labeled metabolites. The 13 C-22:5n-3 is not chromatographically resolved from the endogenous 22:5n-3 but has an m/z value of 349; thus, the mass spectrometer can easily detect this as a discrete signal unaffected by the large endogenous signal at m/z 329. The deuterium-labeled 22:5n-3 is baseline separated from the endogenous 22:5n-3 peak envelope because of the presence of five deuterium atoms (by 0.25 min); thus, its signal at m/z 334 is easily resolved from either of the other two 22:5n-3 species. Similarly, the three selected ion chromatograms are presented in Fig. 1B-F for the endogenous, 13 C-and 2 H 5 -derived signals for 22:6n-3, 24:5n-3, 24:6n-3, 22:4n-6, and 22:5n-6. From these plots, it can be observed that the signal abundances of C22 n-3 PU-FAs were ‫1ف‬ order of magnitude greater (1,000-3,000 counts; Fig. 1A, B) than those of the C22 n-6 PUFAs (100-400 counts; Fig. 1E, F) at the dosages given as a result of the larger isotopic dilutions of the n-6 precursors from the endogenous pools relative to those of the n-3 precursors.

Time course curves of the four precursors
A sufficient number of time points during sampling of the plasma were obtained to provide a good description of the in vivo absorption, decay, and elimination phases of the concentration-time course curves of the four precursors (Fig. 2). The four isotopic precursors and the symbols used to denote them both in the text and in the figures are as follows: 2 H 5 -18:3n-3 (open triangles), 13 C-U-20:5n-3 (open circles), 13 C-U-18:2n-6 (closed triangles ), and 2 H 5 -20:3n-6 (closed circles). Larger doses of the C18 isotopes were given relative to the C20 isotopes, so the C20 isotopes are plotted on a separate y axis scale appearing on the right. Three precursors, 13 C-U-20:5n-3, 13 C-U-18:2n-6, and 2 H 5 -20:3n-6, shared a similar kinetic appearance, reaching C max values of 7.7 Ϯ 1.5, 102 Ϯ 11, and 3.2 Ϯ 0.5 nmol/ml, respectively, at 6.45 Ϯ 0.05 h, as seen in Table 2. Continued metabolism then decreased the isotopomer concentrations to half of the C max value in 7, 8, and 9 h, respectively. The precursors were still detectable in plasma at 60 h after dosing. In contrast, the 2 H 5 -18:3n-3 reached its C max of 14 Ϯ 3 nmol/ml at a T max of 1.2 Ϯ 0.02 h. The isotopomer con-centration then decreased to half of the C max in ‫5ف‬ h, and it nearly disappeared from plasma by 36 h.
D max values, describing the maximal percentage of the initial dose in the plasma compartment, were as follows: 0.7 Ϯ 0.1% for 32 mol of 18:3n-3, 4.1 Ϯ 0.7% for 2.9 mol of 20:5n-3, 1.0 Ϯ 0.1% for 155 mol of 18:2n-6, and 2.3 Ϯ 0.3% for 2.1 mol of 20:3n-6 ( Table 2). Thus, the C20 precursors showed greater entry into rat plasma per milligram of administered dosage compared with the C18 precursors. Different dosages for the various precursors were administered in an attempt to compensate for likely differences in the endogenous pool sizes maintained by the combined interactions of a steady daily dietary input with the inherent metabolic dynamics. E max values for the four precursors appearing in plasma had a small range (10.4, 11.4, 18.5, and 19.2; Table 2), indicating a relatively successful choice of comparable dietary doses. Following the first metabolic step after the C20 precursors, Table 2 shows that E max values for the two isotopomers were similar for 20:4n-6 (0.169 and 0.225) in the n-6 pathway and for 22:5n-3 (2.93 and 3.51) in the n-3 pathway. However, the lower subsequent enrichment of n-6 metabolites compared with n-3 metabolites likely reflects the relatively large pools of tissue 20:4n-6 that are maintained by the steady daily dietary intake of ‫003ف‬ mg of 18:2n-6.

Metabolite concentration-time course curves
The various n-3 metabolites of the stable isotope-labeled precursors are presented in Fig. 3. In Fig. 3A, the concentration of the 13 C-U-20:5n-3 precursor is depicted along with the smaller amounts of 2 H 5 -20:5n-3 formed from 2 H 5 -18:3n-3. The concentrations of the two isotopomers are similar after one elongation step had occurred, with the production of 22:5n-3, with only a slightly greater peak concentration of the 13 C-labeled isotopomer formed from 13 C-20:5n-3 (Fig. 3B, Table 2). Much lower concentrations (C max ϭ 10-27 pmol/ml) of the C24 metabolites were observed in plasma relative to 22:5n-3 and 22:6n-3 (C max near 1,000 pmol/ml). Nevertheless, as n-3 metabolism proceeded by elongation to 24:5n-3, desaturation to 24:6n-3, and retroconversion to 22:6n-3, there was a continued slight predominance of the 13 C-labeled compounds relative to the 2 H-labeled compounds (Fig. 3C-E). The different turnover rates of isotopomers in the different metabolite pools can be described by the time to decrease peak concentrations by half (T C1/2 ). Such  (Table 2), even though the initial dietary dose was 10-fold greater.
The general temporal relationships of isotope appearance among the various fatty acids in plasma were as expected if the metabolic scheme of Voss et al. (41) is as-      Table 2).

Integrated time-concentration curves
One measure of the apparent accumulation of labeled fatty acids in plasma over the entire experimental period (up to 30 d) is the AUC. This parameter is influenced by both the accretion of the particular metabolite and the persistence of that metabolite in the plasma. There is a significant difference (P Ͻ 0.05) between the AUC for the two isotopomers of 20:5n-3, with that of the 13 C-20:5n-3 being greater (Fig. 6A). However, there was no significant difference between the two 22:5n-3 or 22:6n-3 isotopomer values. Both isotopomers of 22:6n-3 showed higher AUC values than those of 22:5n-3 because of their more extended time course curves, which result from greater persistence in the plasma. A pattern distinct from that for n-3 metabolism was observed for the n-6 PUFA isotopomers (Fig. 6B). For the n-6 family, there were no differences between the two isotopomers with respect to the AUCs for 20:3n-6, but 20:4n-6, 22:4n-6, and 22:5n-6 showed a much greater (P Ͻ 0.05) AUC for the 13 C-labeled substrates.

Comparison of metabolic end products on a dosage basis
The amount of each fatty acid isotopomer was calculated for the whole plasma compartment and then expressed as a percentage of the initial dose of its respective precursor, so the time course curves allow a more direct comparison (Fig. 7). Although a much greater dosage of the C18 precursor was given, a much greater response was observed for the 13 C-20:5n-3-derived 22:6n-3 than for that derived from 2 H 5 -18:3n-3, once related to a per dosage basis (Fig. 7A). When the maximal values (D max ; Table 2) were compared for the 22:6n-3 derived from the two precursors, an 11-fold greater value (0.56) for the 20:5n-3-derived compound to that of the 18:3n-3-derived compound (0.05) was observed. Similarly, for the n-6 fatty acids, this type of plot indicated a much greater response for the C20-derived  metabolites with respect to those of the C18 precursors. The curves for both 20:4n-6 and 22:5n-6 were much greater for the 2 H 5 -20:3n-6-derived precursor when expressed in this manner. The D max values were 1.6 versus 0.03 for 20:4n-6 derived from 2 H 5 -20:3n-6 versus that from 13 C-18:2n-6 ( Table 2), a 48-fold difference. Also, 22:5n-6 derived from 2 H 5 -20:3n-6 had a 16-fold greater D max value than when derived from 13 C-18:2n-6 (0.0024 vs. 0.00015; Table 2).

DISCUSSION
The main finding in this stable isotope tracer study was that more labeled 22:6n-3 was produced from 13 C-U-20:5n-3 than from 2 H 5 -18:3n-3 when the concentration-time course curves were related to the initial precursor dosage, and more labeled 20:4n-6 and 22:5n-6 were produced from 2 H 5 -20:3n-6 than from 13 C-U-18:2n-6 when expressed in a similar manner. When the data are expressed in this way, the differences are an order of magnitude or more in favor of the C20 intermediates (Fig. 7). Pawlosky et al. (42,43) performed a physiological compartmental modeling analysis with deuterium-labeled 18:3n-3 in adults. Using this model, they calculated that only ‫%3.0ف‬ of the 18:3n-3 dose appeared in the plasma and only 0.2% of the total amount of labeled 18:3n-3 in plasma was converted to 20:5n-3. Thus, the overall efficiency of 18:3n-3 conversion to 20:5n-3 and longer chain n-3 metabolites is very low. However, they also calculated that ‫%36ف‬ of the plasma 20:5n-3 was accessible for the production of 22:5n-3 and 37% of the 22:5n-3 was available for the synthesis of 22:6n-3 (42). Those data support the present observations that 20:5n-3 and 20:3n-6 are more efficiently converted to 22:6n-3 and 20:4n-6, respectively, than the C18 precursors. Our data are also consistent with the greater efficacy of the C20 EFAs for altering the tissue highly unsaturated fatty acid composition compared with the C18 precursors. Marzo et al. (44) previously compared radioactive 18:3n-3 and 20:5n-3 metabolism with that of 22:6n-3 in cultured retinoblastoma cells.

Apparent preferential metabolism
The pair of n-3 isotopomers seemed to have fairly similar rates of entry to and exit from the various C22 n-3 pools, with the exception that the 13 C-20:5n-3 exits the plasma compartment with a half-time of 7 h and the 2 H-20:5n-3 has a half-time of 14 h (Fig. 3A). However, the entry and exit of the two n-6 20:3n-6 isotopomers differed appreciably: 2 H-20:3n-6 appeared rapidly in plasma and then disappeared with a several-fold faster half-life than 13 C-20:3n-6 produced from 13 C-18:2n-6 (9 vs. 31 h; Fig. 4A).
Although the simplest hypothesis may have been an expectation that the 2 H-and 13 C-tracers would mix in a similar manner within the same endogenous pools, the apparent differential metabolism indicates that carbon atoms entering n-6 metabolism as 18:2n-6 or 20:3n-6 in rat may have different compartmentalization of the precursors and their metabolites in different tissue lipids. This may be attributable to different turnover rates for the different types of plasma lipids or molecular species that carry the two isotopomers. Nevertheless, the general pattern of enrichment values over time ( Table 2) follows that expected for the flow of both isotopes through the intermediates along the recognized pathways. The greater progressive loss of enrichment in the n-6 pathway compared with that in the n-3 pathway may reflect larger intermediate pools of n-6 than n-3, which result from the greater proportions of n-6 precursors provided in the maintenance diet.
In addition to these kinetic differences between the C20 precursors and their isotopomers and the C18 precursors, differences were also seen in the successive elongated and desaturated metabolites, in particular in the n-6 family. Several interesting features were apparent upon observing the AUC data. When the first step of metabo- , and the n-6 intermediate 2 H 5 -20:4n-6 (closed circles) and 13 C 18 -20:4n-6 (closed triangles) (B). Time course curves over 11 days for the n-6 end products lism after the introduction of the C20 isotope was inspected (Fig. 6), there appeared to be a relatively greater AUC response for metabolites that originated from the C18 precursor. That is, there was a greater AUC for 13 C-20:5n-3 than for 2 H-20:5n-3 (P Ͻ 0.05), so a greater amount of the 13 C-22:5n-3 would be expected. However, an approximately equal AUC of the two 22:5n-3 isotopomers was produced, indicating a preference for that derived from the 2 H 5 -18:3n-3. Similarly, although the AUCs for the two 20:3n-6 isotopes were very similar, a larger response was evident for the AUC of the 20:4n-6 derived from the 13 Clabeled precursor. This effect may be attributed at least in part to the apparent inefficient use of the precursors during their initial absorption phase. That is, there appear to be fewer C20 molecules routed into the fatty acid anabolic pathways relative to newly synthesized 20:5n-3 or 20:3n-6 molecules, perhaps because of greater catabolism during absorption (45). It is interesting, however, that in the subsequent metabolism of 20:4n-6 to 22:4n-6, the 13 C compound was again formed preferentially with respect to the 2 H 5 compound (Fig. 4C). It appeared then that 20:4n-6 substrate was more efficiently elongated when it had a history of sequential elongation/desaturation (i.e., 13 C-20:4n-6 derived from 13 C-18:2n-6) than when it had more recently entered the metabolic pathway. This is consistent with previous reports of a channeling concept for coupled fatty acid elongation/desaturation systems (46), perhaps by a multienzyme complex (47).

Different lipid classes
One possible explanation for these results would be the initial differential incorporation of labeled 18:2n-6 and 20:3n-6 or 18:3n-3 and 20:5n-3 or their metabolites into different lipid classes. A dietary study conducted by Gibson et al. (48) showed that triglycerides (TG) carried most of the 18:3n-3 and part of the 18:2n-6, whereas phospholipids (PL) and cholesteryl esters (CE) retained approximately half of the 18:2n-6. 20:5n-3 was preferentially carried in the CE fraction in rats regardless of whether it originated from endogenous metabolism or from dietary intake. Similarly, 20:4n-6 was predominant in the CE fraction irrespective of the dietary fat source. A study conducted by Morise et al. (20) showed that ‫%54ف‬ of the long-chain PUFAs, primarily of the n-6 family, were localized in PLs, and only a small amount of 18:3n-6 and 20:3n-6 accumulated (49)(50)(51).
In our study, pilot data of the lipid class distribution in rat plasma at 24 h after giving the four isotopic precursors were collected. Approximately 65% of 2 H 5 -18:3n-3 was recovered in TG; however, 13 C-20:5n-3 was found primarily in CE (65%). PL (40%) and CE (32%) were the main classes for both the precursor 13 C-18:2n-6 and 2 H-18:2n-6 retroconverted from 2 H-20:3n-6. However, 2 H-20:3n-6 was found in NEFA (31%) and PL (48%). Thus, the initial incorporation of the two n-3 and the two n-6 precursors reflects a different lipid class profile, with the precursor having a preference for the lipid classes indicated: 2 H 5 -18:3n-3 in TG, 13 C-U-20:5n-3 in CE, 13 C-18:2n-6 in PL and CE, and 2 H 5 -20:3n-6 in PL and NEFA. These studies suggest that differences in the initial lipid class distribution of the various precursors could play an important role in the transport and metabolism of a particular isotope.

Method advantages and additional applications
This study presents a new methodological approach for using the multiple stable isotope tracer technique for the in vivo study of EFA metabolism. It is demonstrated that four different isotopes may be studied simultaneously using this MESSI technique. An advantage of performing experiments by this method is that fewer experimental animals or human subjects can be used, because experimental conditions are identical for longitudinal comparisons of the metabolism of different metabolic precursors over time. This approach was used to interpret the flow of EFAs among metabolite pools in animals equilibrated and maintained with steady dietary inputs similar to those commonly used (and similar to typical dietary inputs in the United States). The method can have a wider application in studies of both EFA and NEFA metabolism and also in studies in which dietary variables are introduced. It should be particularly important where sampling must be limited, such as in studies of human infants.