An analysis of risk factors for spontaneously occurring Type 2 Diabetes Mellitus in rhesus macaques (Macaca mulatta)

Abstract

Background:

Type 2 Diabetes Mellitus (T2D) is a chronic disease with a high prevalence worldwide. Human literature suggests factors beyond well-known risk factors (e.g., age, body mass index) for T2D: cytomegalovirus serostatus, season of birth, maternal age, birth weight, and depression. Nothing is known, however, about whether these variables are influential in primate models of T2D.

Methods:

Using a retrospective methodology, we identified 22 cases of spontaneously occurring T2D among rhesus monkeys at our facility. A control sample of n=1199 were identified.

Results:

Animals born to mothers that were ≤ 5.5 years of age, and animals that showed heightened activity and emotionality in response to brief separation in infancy, had a greater risk for development of T2D in adulthood.

Conclusions:

Knowledge of additional risk factors for T2D could help colony managers better identify at-risk animals and enable diabetes researchers to select animals that might be more responsive to their manipulations.

Introduction

Type 2 Diabetes Mellitus (T2D) is a chronic disease characterized by insulin resistance and hyperglycemia. A recent review estimated that, worldwide, 529 million people were living with diabetes, with a 6.1% prevalence; in 2021, approximately 96% of these cases were T2D [1]. A variety of risk factors have been identified for humans, including presence of prediabetes conditions, high body mass index (BMI), a close relative with T2D, and race [2].

Animal models have been of great value in understanding the onset, pathogenesis, and molecular mechanisms associated with T2D [3], and primate models have been of special value, inasmuch as T2D can occur spontaneously in both New and Old World monkeys (Harwood et al., 2012). In these species, T2D is associated, as in humans, with obesity and increasing age, among other similarities [3–5]. In addition, nonhuman primate models, because of the many similarities with humans [6] are excellent candidates for studies involving induction of T2D.

While obesity and aging have been identified as risk factors for spontaneous development of T2D in both humans and Old World monkeys, a variety of other risk factors have been proposed in human studies that have not been examined in primate models. Identification of additional risk factors may provide additional opportunities for study of mechanisms of pathogenesis of T2D, and may also be useful for colony managers to identify at-risk animals that could be given special treatment to potentially forestall the development of this disease – this would be valuable to both the animals and the veterinary and animal care staff, as treatment usually involves daily administration of drugs by technicians, and often, removal from a group setting (often from less-expensive housing outdoors) to (more expensive) individual or pair housing indoors.

Cytomegalovirus (CMV) status.

There has been growing interest in the role of viral infections in triggering both Type 1 and Type 2 diabetes [7]. Recently, cytomegalovirus has been shown to be related to T2D onset, although some evidence suggests this may not be the case for all human populations [8]. In one study, 576 adult patients with CMV disease were followed for five years along with 2880 controls; neither group had evidence of T2D at the outset. Over the five year follow-up, new-onset cases were significantly higher in the group with CMV disease compared to case controls [9]. Other studies have also shown relationships between seropositivity for CMV and T2D [10, 11] and with pre-diabetes incidence [12], even in the absence of CMV disease per se. Because studies of CMV pathogenesis is a significant research focus at our institution [13], we have serology on hundreds of animals that can be used to address this question.

Season of birth.

Season of birth has been examined for its association with T2D, with the idea that seasonality may be a proxy for adverse prenatal experiences. In a prospective cohort study of nearly half a million people from diverse areas in China, individuals born in Summer (June, July, August) had a 9% lower risk of T2D than individuals born in other seasons of the year [14]. Results from studies of European populations have been mixed, however, leading Si et al. [14] to speculate that latitude might be a mediating factor in the conflicting results. (Davis, California is within the range of latitudes of the Si et al. study.) Mechanistically, the authors hypothesize that seasonal differences in fetal nutrition may be involved inasmuch as impaired fetal nutrition has been linked to T2D through changes in pancreatic islet cells and tissue sensitivity to insulin [15].

Maternal age.

Lammi et al. [16] found a U-shaped relationship between maternal age and T2D. Maternal ages ranged from 15 to 39 years of age in their study, and their analyses considered maternal age as a continuous variable. They hypothesized that this relationship may be a reflection of low birthweight, however, and not maternal age per se. We have maternal age at offspring birth for all animals in our colony.

Birth weight.

While the evidence is strong that obesity in adulthood is a risk factor for T2D, there is evidence that birth weight may also be a risk factor. Specifically, low birth weight has been linked to T2D and impaired glucose tolerance (see review in [15]). This has been taken as evidence of the “thrifty phenotype” hypothesis [17], which suggests that poor maternal nutrition could program the fetus’s metabolism to promote storage of fat, which could become a problem in a nutritionally rich postnatal environment. In addition, a meta-analysis showed that high birth weight is also associated with T2D, suggesting a U-shaped distribution [18]. Harder et al. [18] speculate that high birth weight may be a proxy for “undiagnosed and nontreated maternal hyperglycemia during pregnancy and/or maternal overweight during pregnancy (p. 856).” While we do not have birth weight data on our animals, we do have weights of infants 3-4 months of age, collected in a standardized fashion during an infant BioBehavioral Assessment (BBA; see below), on a substantial subset of animals.

Depression.

A number of psychological factors have been linked to risk for T2D, among them depression. In a large prospective cohort study involving nearly half a million participants, Meng et al. [19] reported that individuals with a major depressive episode in the past year had a 32% increased risk of T2D. In addition, those with depressive symptoms, but who did not meet the definition of a major depressive episode, had a 19% increased risk of T2D. These results echo those from a number of other studies, including a large meta-analysis [20]. While we do not have adult data on depressive behavior in our animals, we do have data on the BBA subset of animals for two variables, activity and emotionality, that were positively related to depressive behavior in adulthood [21]. These were the measures used to test this hypothesis.

We extracted relevant information from a variety of sources to test these hypotheses: our colony database, a serology database from our Pathogen Detection Laboratory, and the archive from the BBA program. Our focus was on individuals born between 2001 and 2019.

Materials and Methods

Humane Care Guidelines.

The Institutional Animal Care and Use Committee of the University of California, Davis approved the animal use protocols for the work reported in this paper. UC Davis is an AAALAC-accredited institution. The authors adhered to the ethical policies of the journal, as noted on the journal’s author guidelines page. Animals were maintained in accordance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals and the US Public Health Service’s Policy on Humane Care and Use of Laboratory Animals.

Subjects.

A total of 14,614 live rhesus monkey (Macaca mulatta) births were recorded in colony records between 2001 and 2019, inclusive. Of these, n=62 cases of Type 2 diabetes were identified via flags in their records. Careful examination of project history and medical records, however, revealed that most of these cases were either the result of experimental manipulation (e.g., dietary interventions to induce T2D) or of administration of substances that had diabetes as a side effect (e.g., tacrolimus). A total of n=22 cases of spontaneously occurring diabetes were retained. Mean age of these cases was 15.6 years, with a range from 8.7 to 20.7 years. We restricted our analyses to all animals that were 8.7 years of age or older at the time of our analysis. This resulted in a total sample size of n=1221, 22 of whom were diagnosed as diabetic, and 1199 of whom served as controls.

Subjects were born and reared in any of the locations throughout our colony. These include the outdoor, half-acre field corrals, which can hold up to 200 animals of all age/sex classes; smaller outdoor enclosures (i.e., corncribs) that can house up to 30 animals of all age/sex classes; and indoors, either with no mother but with an age-matched peer (nursery); or mother-only, or with mother plus daily access to a second mother and her infant. Rearing location was determined for colony animals based on their location at 106 days of age, which is the mean age at which BBA testing occurs. All animals in our colony are fed a standard diet (Lab Diet 5047) twice daily with fresh fruit and vegetables provided once or twice weekly as supplements. Animals diagnosed with T2D typically receive a fiber-balanced diet (Lab Diet 5000), and supplements comprise only vegetables, no fruit. Table 1 shows breakdowns for the main sample for sex and rearing, as well as summaries for the outcome measures in the main sample and BBA subsample.

Table 1:

Characteristics of main sample and BBA subsample.

Main Sample (n=1221)1:		Diabetic (n=22)	Non-Diabetic (n=1199)
Sex:	Male	5	243
	Female	17	956
Rearing:	Field Corral	17	859
	Corncrib	2	99
	Indoor	3	241
CMV (serology):	Positive	6	355
	Negative	3	151
Season of Birth:	Summer	3	128
	Other seasons	19	1071
Maternal Age:	<= 5.5 years	13	429
	> 5.5 years	9	770
BBA Sample (n=655)2:		Diabetic (n=13)	Non-Diabetic (n=642)
Infant Weight (kg)		0.943 (0.846, 1.040)	0.998 (0.985, 1.010)
Activity:	Day 1	0.351 (−0.181, 0.884)	−0.058 (−0.133, 0.018)
	Day 2	0.518 (−0.020, 1.056)	−0.017 (−0.094, 0.059)
Emotionality:	Day 1	0.386 (−0.156, 0.928)	−0.044 (−0.121, 0.033)
	Day 2	0.376 (−0.163, 0.914)	−0.035 (−0.111, 0.042)

¹Values shown are number of cases.

²Values are mean z-scores and 95% Confidence Intervals

Diabetes diagnosis.

Diagnosis of diabetes was made by CNPRC veterinary staff. The indoor-housed CNPRC geriatric population (>19 years) is screened every 6-9 months for the development of T2D. During their semi-annual physical exams, blood is collected to assay for glycosylated hemoglobin (GHB). Abnormal elevated results are repeated for pre-diabetic animals (GHB between 5.6% to 7.0%) or animals are admitted into the diabetes monitoring and treatment list if their GHB is consistently higher than 7.0%. Outdoor and younger animals are diagnosed based on clinical signs (obesity, polydipsia, polyphagia), abnormal chemistry results (hyperglycemia, increased fasted triglycerides, glycosuria) and a definitive diagnosis is ascertained with an elevated GHB. Because of the different pathways used to diagnose T2D, a date of diagnosis is not always included in our colony records. In our study we defined the “date of diagnosis” by the first administration of one of several medications that are routinely used to treat diabetes. Theses medication in our colony include metformin, pioglitazone, glipizide, insulin degludec, and insulin glargine.

CMV status.

CNPRC has a specific pathogen-free (SPF) program, in which particular viruses are eliminated from subsets of animals. CMV is one of these pathogens, and its presence or absence is routinely and repeatedly tested for many animals by CNPRC’s Pathogen Detection Laboratory using serologic and PCR assays [22]. Cases were identified that never had a positive result, and these were designated CMV−, while those with at least one positive result were considered CMV+. Of the n=1221 animals in our cohort, n=154 animals were CMV− and n=361 animals were CMV+. This contrast was our first analysis for the possible effects of CMV in T2D. Because virtually 100% of animals born in a CMV+ cage are themselves positive by approximately 1 year of age [23], however, we performed a second analysis, in which the missing data from our cohort were assumed to be CMV+ by the time of their T2D diagnosis. In this analysis, n=154 were CMV− and n=1067 were CMV+.

Season of birth.

Following [14], we contrasted animals born in Summer months (June, July, or August) with animals born at other times of the year. We also examined the distribution of births month by month to see if the number of diabetes cases occurred at a greater-than-expected frequency for any given month.

Maternal age.

Parentage at CNPRC is assessed via a panel of microsatellite markers for every animal born at CNPRC [24, 25]. The difference between the mother’s and the offspring’s dates of birth was used to identify maternal age at birth.

Infant weight.

As mentioned earlier, we did not have information on weight at birth. Of our n=1221 cases, however, n=655 animals participated in CNPRC’s BioBehavioral Assessment (BBA) program, described below. Briefly, animals that are approximately 90-120 days of age are removed from their mothers for a 25-hr period for behavioral and physiological testing. Once animals are relocated to the BBA testing area, they are weighed and their weights recorded. These are the data used for this analysis.

Depression (proxy), and brief description of the BBA program.

The BBA program has been described in detail elsewhere [26, 27]. Briefly, once animals are relocated to the BBA testing suite for the 25-hr period, they experience several behavioral assessments, as well as blood samples for assay of plasma cortisol. The first assessment the animals experience is a five-minute per animal focal behavioral observation (conducted on each animal in a predetermined random order) in their individual housing cage. Using several hundred cases from the BBA archive, exploratory and confirmatory factor analyses identified an underlying latent structure to the data involving two factors: Activity (comprising six behaviors: proportion of time spent locomoting; proportion of time NOT hanging from the side of the cage; rate of environmental exploration; and whether the animal ate food, drank water, or crouched) and Emotionality (comprising five behaviors: rate of coo vocalization; rate of bark vocalization; and whether the animal scratched, displayed threats, or lipsmacked). The initial observations were made beginning 15 min after relocation to the test area, and the measures from this set of observations were labelled Day 1 Activity and Day 1 Emotionality. The focal observations were repeated the next morning beginning at 0700 hr (22 hrs after relocation), and the resulting scales from these observations were labelled Day 2 Activity and Day 2 Emotionality. For each year, scores on the computed scales are z-scored. As described earlier, higher scores on the Activity and Emotionality measures were significantly associated with the display of depressive behavior following relocation as adults [21]. See [28] for details of the analysis and definitions of behaviors.

Statistical analyses.

Continuously distributed variables (maternal age, birth weight, and the Activity and Emotionality measures) were analyzed using analysis of variance, with diabetic status (1 = diabetic, 0 = no diabetic diagnosis) as the independent variable. Other variables (CMV status, season of birth) were analyzed using chi-square tests. Significant variables were combined into a final omnibus logistic regression (with diabetic status as the outcome measure).

We conducted preliminary analyses to insure there were no significant variables that might be considered covariates in our formal analyses. Chi-square analysis showed there was no significant effect of sex (p = .776) or rearing environment (p = .754) on diabetes status (see Table 1). In addition, participation in the BBA program itself did not predict diabetes status (p = .605).

Results

CMV status.

There was no evidence of a relationship between CMV status and T2D. The initial analysis used only those n=515 cases for which serology data were available, and the chi-square was not significant (p = .821). The second analysis used all cases, under the assumption that by one year of age in our colony all animals are CMV+, except for those animals that were purpose-bred for negative serostatus. Again, the chi-square was not significant (p = .884).

Season of birth.

There was no evidence that birth timing was associated with diabetes status. The initial analysis contrasted animals born in June, July, or August (n=131) with those born at any other time (n=1090). The chi-square was not significant (p = .657). We followed up this analysis by a more detailed analysis using chi-square looking at diabetes status by month of birth, with births in September and later in the year combined into a single category. Again, the chi-square was not significant (p = .991). Inspection of the adjusted standardized residuals for each cell revealed that no residual was larger than |.9|.

Maternal age.

We found evidence that diabetic animals were born to mothers of younger age. This was indicated by trend in the analysis of variance: F(1,1219) = 2.824, p = .093. Mean age of mother for diabetic and non-diabetic animals was 6.3 and 7.6 years, respectively. We examined the age distribution of mothers of diabetic animals, and classified maternal age into two categories: 5.5 years of age or younger versus >5.5 years of age. This classification reflected a natural split in the distribution of maternal age of the diabetic animals. In this analysis chi-square was significant: Chisquare(1) = 5.083, p = .024. There were significantly more diabetic cases (n=13) than expected in animals whose mothers were 5.5 years of age or less. Only nine diabetic animals had mothers that were more than 5.5 years of age (Table 1). Because only three diabetic cases were born to mothers 10 years of age or older, we could not examine the impact of older mothers on T2D.

Infant weight.

As described above, we did not have data on weight at birth, but for the BBA subsample (n=655, with n=13 diabetic cases) we did have weight at the time of BBA testing. We found no effect, as indicated by a non-significant analysis of variance (p = .225). Because the range of ages for BBA testing was relatively wide (90 – 130 days of age), and because there is a significant correlation between weight and age at testing (r = .407, p < .001), we repeated the analysis with age as a covariate. Again, the result was not significant (p = .199).

Depression (proxy).

Using the BBA subsample, we found evidence that animals with higher values for Activity and Emotionality were more likely to be diabetic. A two-way analysis of variance (day by diabetic status) showed a significant main effect for Activity (F(1,653) = 4.350, p = .037) and a near-significant main effect for Emotionality (F(1,653) = 3.353, p = .068). Interactions of diabetic status and day were non-significant (p = .687 and p = .951, respectively). See Figure 1.

Figure 1:

Animals showing higher levels of Activity and Emotionality as infants in response to separation and relocation are more likely to be diagnosed diabetic as adults. For the non-diabetic group, n = 642; for the diabetic group, n = 13. Values shown are means across Day 1 and Day 2 for each group, along with standard errors.

Omnibus logistic regression.

The logistic regression analysis, in which diabetic status was the outcome variable, and maternal age (dichotomized as 5.5 years or less versus greater than 5.5 years) and the means (across days) for Activity and Emotionality were the independent variables, was significant (Chisquare(3) = 10.01, p = .019). (Note that because the behavioral variables were included, this analysis used only the BBA subsample.) Nagelkaerke R² was .086. As shown in Table 2, the only variable with a significant coefficient was Activity (p = .032): the odds ratio indicates that animals with higher values of Activity in infancy have more than double the chances of being diagnosed with T2D later in life (odds ratio = 2.101).

Table 2:

Results of omnibus logistic regression. Outcome is diabetes diagnosis (yes=1, no=0).

							95% C.I.for EXP(B)
	B	S.E.	Wald	df	Sig.	Odds Ratio	Lower	Upper
Mean Activity	.742	.347	4.585	1	.032	2.101	1.065	4.144
Mean Emot.	.495	.311	2.533	1	.112	1.640	.892	3.017
Dam ≤ 5.5 yrs.	.962	.586	2.697	1	.101	2.618	.830	8.254
Constant	−4.608	.489	88.713	1	.000	.010

Discussion

Our goal was to examine risk factors for spontaneously occurring T2D in rhesus monkeys that had been identified in human studies. Our interest was in risk factors beyond the well-known ones (for both monkeys and humans) of age and weight. While we could not compare weights in our sample, our sample averaged 15.6 years of age, which, for this species, is advanced adulthood, though not yet “aged.” We found that animals with T2D were born to younger mothers, and showed a behavioral profile in infancy that was associated in adulthood with depressive behavior. We discuss each of our results next.

CMV status.

We found no evidence that CMV serostatus was associated with T2D. The reason for this may be that, unlike humans, the monkeys in our colony are virtually 100% positive for CMV by one year of age, the exception being those animals that were reared expressly to be negative for this virus [23]. Prevalence figures for humans range widely but typically do not reach 100% [29]. In our sample, we only had n=3 animals (out of a total of n=22) with T2D that were seronegative for CMV. While we found no relationship, we consider this an open question worth pursuing as larger numbers of animals with T2D accumulate in our colony.

Season of birth.

Again, we found no evidence that season of birth was associated with T2D. This negative result may, however, be a reflection of birth seasonality in this species. While there is evidence of birth seasonality in humans, humans give birth all year round, and seasonal birth patterns can differ by location, culture, etc. [e.g., 30]. In contrast, among rhesus monkeys, seasonality is more strict. In our colony, for example, 78.0% of all births between 2001 and 2019 occurred in the months of March, April, and May (data not shown). Moreover, as mentioned above, fetal nutrition is considered one possible mechanism for seasonal effects in humans. In captive colonies of primates, however, nutritional demands are controlled by the facility, not by natural cycles of weather, food availability, etc. Consequently, limitations of our small sample size notwithstanding, we consider rhesus monkeys poor models for exploring the relationship between birth seasonality and T2D.

Maternal age.

In the full sample, we did find a greater than expected number of cases of T2D in monkeys whose mothers were 5.5 years of age or younger (although we note that this measure did not survive the omnibus logistic regression using the smaller, BBA sample). In our colony, female monkeys often have their first offspring at 3 or 4 years of age (personal observation), so by the time they are 5.5 years of age, they may have given birth to up to three infants. (We note that an analysis of whether diabetic cases were disproportionately first-borns was not significant [analysis not reported here].) Lammi et al. [16] hypothesized that this effect may have been a reflection of low birth weight; as this measure was unavailable to us, we cannot address this explanation. We do know, however, that young mothers have a milk composition that differs from older mothers: Hinde & Capitanio [31] focused on available milk energy (the product of milk gross energy and milk yield), and found that, “for each additional parity, mothers produced 0.4 and 0.9 additional kcal of milk at each time point [time points were 1 month and 3.5 months of infant age, respectively] (p. 525).” We don’t know if there is a significant cut-point in milk composition at the parity found when mothers reach 5.5 years of age, however, nor are we aware of any scientific literature relating naturally occurring variation in components of mother’s milk to risk for T2D (although there is some evidence that breast-fed infants have a reduced risk of T2D compared to bottle-fed infants [32]). This is an exciting area for future study.

Infant weight.

Weights obtained during the BBA program (between 3-4 months of age) were unrelated to risk for T2D. Other studies focused specifically on weights at birth, however, which were not available to us for all animals. We did have birth weights on animals assigned to our nursery, however, and a random selection of n=21 such animals showed that infant weight and birth weight are significantly correlated, even when accounting for variation in the age of testing in the BBA program: partial r = 0.799, p < .001 (data not shown). These data suggest infant weight may indeed be a suitable proxy for birth weight. While we found no evidence that infant weight predicted risk for T2D diagnosis, we believe that it remains possible that unusually high or low birth weights among captive primates may put animals at risk for later development of T2D, as has been found in humans [18]. Our small sample of diabetic animals may have insufficiently sampled the range of infant weights with enough power to detect such a relationship.

Depression (proxy).

Activity recorded in the BBA program was a significant predictor of risk for T2D, and Emotionality was a near-significant secondary predictor. In 2014, Hennessy et al. [21] reported that Activity and Emotionality, as determined in the BBA program at 3-4 months of infant age, were related to a tendency for adult male rhesus monkeys to display the hunched posture that characterizes depression, when the animals were removed from their field corrals and held in individual cages indoors. Previously [27], we considered Activity and Emotionality in this assessment as indicators of responsiveness (to the relocation and separation) and adaptation (where the more “adaptive” pattern involved lower Activity on Day 1 and increased Activity on Day 2). Because we found no interaction with Day in our analyses, we propose that the pattern of high Activity and Emotionality on both days is a somewhat maladaptive response in the BBA testing situation [27]. And of the three measures that showed either significant (maternal age, Activity) or near-significant (Emotionality) bivariate impacts on T2D, only Activity survived the omnibus logistic regression, making a behavioral measure, obtained years prior to the diagnosis, the principal predictor of risk for T2D. If we can consider our measures of Activity and Emotionality as proxy measures for a tendency to display depressive behavior later in life in response to social manipulations, then our results are consistent with those reported in the human literature [e.g., 19].

Limitations.

The most obvious limitation of our study is the small number of cases of T2D in our sample. Fortunately for the animals themselves, and because of the expense of treatment, spontaneously occurring cases of T2D in our (and probably others’) primate colony are relatively rare. This makes accruing a reasonable sample size a challenge. And we note that the way in which T2D cases are identified (described above) likely under-identifies cases in our outdoor colony – less detailed surveillance for illness might be a trade-off to a housing strategy that emphasizes psychological well-being and mimics group composition in the wild. While we urge caution in interpreting and utilizing our results until they can be replicated by others, we believe our results are suggestive of new directions to pursue in understanding the development of this condition. A related limitation of our relatively small sample size is that we simply could not test, for example, the hypothesis that advanced maternal age might also be a risk factor for T2D in offspring. This and other questions must remain open until a larger sample is found.

Implications.

The goal of this analysis was two-fold. First, we were interested in whether the risk factors that we considered might provide opportunities for use of this animal model to study mechanisms and pathogenesis of T2D. Our results indicate qualified support for the idea that maternal age might be worthy of continued study (and we propose one mechanism whereby this risk factor might be influential, namely the quality of mother’s milk), and significant support that behavioral measures in infancy relating to later depression in adulthood might be deserving of further study. There was insufficient support for the role of CMV status, infant weight, and season of birth. The second goal was to identify potential additional risk factors (beyond age and weight) that might assist colony managers to identify at-risk individuals and possibly to intervene to forestall development of T2D. Maternal age and behavioral responses to separation/relocation in infancy are the two measures with the greatest potential. While colony managers may not have systematic, quantitative behavioral data on their animals (such as we have collected with the BBA program), one can easily identify animals that respond to cage relocations with higher levels of activity and emotionality, or even with frank depressive behavior (e.g., hunched posture). Moreover, it’s clear that in studies of diet-induced diabetes [33], only a subset of treated animals develops diabetes. Perhaps selecting animals with younger mothers and with a behavioral tendency toward stress-related depressive responses might increase the number of animals that develop the desired phenotype. Using this information to select animals could also be valuable in exploring the role of other viral triggers to T2D, as has been suggested for humans [7].

Data Availability:

The data that support the findings of this study are available from the corresponding author upon reasonable request.