Subjects. Study 1 was conducted on 64 adult animals (32 males, 32 nulliparous females) that originated from three different sources—two national primate research centers and one commercial vendor. These animals were assigned to ongoing projects assessing obesity propensity. They were singly housed and fed, in ad lib quantities, a commercial marmoset diet (Purina Mazuri gelled diet; primary ingredients: glucose, casein, ground wheat, corn flour, dehulled soybean meal, gelatin, porcine fat, dehydrated alfalfa meal, dried beet pulp, egg yolk solids, dried whey, soybean oil; 22% protein, 6% fat; 4.47 kcal/g dry weight) and 3–5 days per week were given one small slice of fruit or 1–3 raisins in the afternoon.
Data collection. After a habituation period of 4 months, weights, body composition, and blood samples were collected from each animal, within a 1-week period in August 2006. Animals were weighed, either through placing a scale into the animal's home cage or by removing the animal from its home cage in a catch box and weighing it in the catch box. All scales provided weights to the nearest gram and were calibrated every 6 months.
Estimates of lean and fat mass were obtained through quantitative magnetic resonance scans, using an EchoMRI unit (EchoMRI; Echo Medical Systems, Houston, TX) designed for marmosets, based on a previously designed rat systems. This system has been extensively validated for mice (34), and the detection methodology used in the marmoset system is identical to that used in the mouse system with only the volume of the homogenous magnetic region differing. Unsedated animals were placed in a plastic tube, which was then inserted into the magnetic chamber. Scans took <2 min, on average, for each animal.
Glucose metabolism (glucose, HbA1c) and lipid (high-density lipoprotein (HDL), low-density lipoprotein (LDL), very low–density lipoprotein (VLDL), and triglyceride) parameters were measured in blood samples collected into a heparinized syringe from unsedated animals that had been fasted overnight. At the time of blood draw, a drop was used to measure glucose using a FreeStyle glucometer and 10 µl of whole blood separated for use in assaying HbA1c. The remaining blood was processed for plasma collection. The plasma and the HbA1c samples were shipped to the laboratory of M. Paulik, GSK, Research Triangle, NC for analysis. All serum chemistry parameters were obtained utilizing the Olympus Au640 clinical chemistry analyzer (Olympus America Inc, Melville, NY) with the reactions run at 37 °C.
The relationship of morphometric parameters—body weight, lean weight, fat weight, and relative fat (total weight of fat based on magnetic resonance imaging divided by total body weight)—to sex and age were determined through a generalized linear model analysis, with sex as a main factor and age as a covariate.
In humans, obesity and other risk factors associated with increasing cardiovascular disease or diabetes risks are often defined by assessment in large-scale epidemiological studies or in consensus conferences of clinicians (35). Although information on actual body fat and its distribution is most desirable, for practical reasons human obesity has been variously defined relative to weight (>30% above average in relation to sex and age; BMI (weight/height2, with values >30 kg/m2 indicative of obesity), waist circumference (>102 cm in men or >88 cm in women), and only occasionally in relation to relative body fat (>25% in males and 30% in females representing obesity). To relate the findings from the marmoset population to assessments of human risk factors for metabolic syndrome, we used the following operational definitions for atypically high or low values of parameters identified as risk factors in various human metabolic syndrome models:
1. Total body fat or relative body fat >80th percentile (the top quintile of animals), defined separately for males (58.2 g or 14%) and females (73.4 g or 17%). The 80th percentile was chosen to differentiate animals that were 30% above the sex-specific mean relative body fat, making these results comparable to human studies and to operational definitions of fatness in other nonhuman primate studies (e.g., ref. 14).
2. Average HDL—1 s.d.: <42 mg/dl.
3. Average fasting glucose + 1 s.d: 219 mg/dl or average HbA1c + 1 s.d.: 5.5%.
4. Triglyceride concentration >400 mg/dl, representing the point at which the elongated portion of the high concentrations driving the non-normal distribution began.
HDL, glucose and triglyceride cutoff values were not sex specific because a comparison of the mean/median values for these parameters between males and females exceeding the 80th percentile of relative body fat revealed no significant sex differences.
Two analyses of the relations among these variables were conducted: (i) the mean or median concentration for metabolic and lipid parameters were compared in subjects exceeding vs. not exceeding the 80th percentile in body fat. Comparisons involving non-normally distributed variables (VLDL and triglyceride concentrations) were made using Mann–Whitney U tests while comparisons involving normally distributed variables (all others) were made using analyses of variance; (ii) the number and distribution of atypical factors displayed among the subjects was tallied and is presented in tabular form.
A set of longitudinal analyses were conducted on a second population consisting of offspring produced from a single marmoset breeding colony, begun in 1994 and continuously breeding since that time—see ref. 36 for details on the colony history and management. All animals were housed in typical marmoset family groups consisting of one breeding female, one breeding male and their offspring up to 2–4 years of age. This colony has been maintained on two base diets fed simultaneously—a purified gelled diet (Teklad Research Diets, Madison, WI; primary ingredients: lactalbumin, dextrin, sucrose, soybean oil; 15% protein, 5% fat, 4.0 kcal/g) and one of two commercial diets—either Purinary-Mazuri gelled diet (see Study 1) or ZuPreem canned diet (primary ingredients: cracked wheat, whole egg, soybean meal, sucrose, rice, vegetable oil, alfalfa meal; 22% protein, 6% fat; 4.84 kcal/g). This population also received supplements of fresh or dried fruit 3–5 times per week. The following data were available for >80% of individuals born into this colony who were reared to weaning: dam, sire, sex, birth date, litter size, birth weight, and average early adult weight (taken as the average of all weights between 17 and 24 months of age). Details on methods for weighing are provided in ref. 36.
A general linear model analysis was used to determine the relations of the following variables to early adult weight: sex, litter size, and birth year cohorts (as an estimator of secular trends). Dam identity was included as a random variable to correct for possible pseudo-replication effects inherent in individual dams contributing variable numbers of offspring to the population. The total population in this analysis was 210 individuals.
A subset of individuals from this population was chosen to further examine the relation of prepubescent growth in weight to early adult weight. The subset of subjects were selected as follows: (i) all litters with at least one individual who exceeded the 90th percentile of body weight as an early adult; (ii) each litter in the population immediately preceding or proceeding the birth of the above defined litter with sufficient early weight data to calculate early growth estimates. This selection process resulted in 49 animals from 29 litters born between 1998 and 2006: 10 male–female litters, 4 male–male litters, 6 female–female litters, four female singletons and five male singletons. Litter size refers, here to nursed litter size, not necessarily birth litter size. Weight data for each subject from birth to 9 months of age were used to calculate linear regression estimators. Marmoset growth is linear during this age period (36), reflected in an average r2 of 0.90 for the animals in this analysis. As compared to study 1, a more conservative approach was used to compare body sizes, given that only body weight was available. A generalized linear model was used to compare weights at 0, 2, 4, and 6 months for animals exceeding 90th percentile in weight vs. those under the 90th percentile by early adulthood. A more stringent weight criterion (i.e., 90th vs. 80th percentile) was chosen for this analysis, given that we had only weight measures in this population and could not, therefore, differentiate between fat and lean differences. In addition, paired t-tests were used to compare estimated 2-, 4-, and 6-month weights within litters.
All analyses were conducted using SPSS, version 15.0 (SPSS, Chicago, IL).