Defining predictive factors for reproductive output in captive common marmosets (Callithrix jacchus)

Common marmosets (Callithrix jacchus) demonstrate variations in reproductive output, not only in terms of total reproductive output during a lifetime but also in litter size per parturition. The present study explores factors, such as parents’ litter size, parturition number, maternal body weight at conception and maternal age, which may account for this variation. A retrospective analysis of clinical records of a captive breeding colony was conducted over a 9‐year period yielding reproductive summaries of 26 dams and 22 sires producing a total of 115 litters. Dams born from litters of  ≤2 (N = 20) more often produced litters of ≤2, whereas dams born from litters of >2 (N = 6) more often produced litters of >2 (p < 0.05). The dams’ maternal body weight at the time of conception had also a significant effect on subsequent litter size. In addition, the chance of triplets was higher after the second parturition. Maternal age, interbirth interval, and season of birth had no effect on litter size. Factors relating to the sire had a negligible effect on the size of the litter. Multivariate statistical modeling revealed that the dams’ original litter size, maternal bodyweight at conception and parturition number are determining factors for the number of babies per litter. This study identified factors determining marmoset litter size, some of which (maternal litter size) are novel to this study and were not reported previously. Further exploration of the potential role of maternal litter size as a determinant of the litter sizes produced by marmoset breeders is warranted.

12:12-h light:dark cycle (lights on, 0700-1900). Lighting in the indoor compartment is provided using full spectrum fluorescent bulbs placed close to the cages in addition to natural light through windows.
The room ventilation rate is around eight air changes per hour. The daily diet consists of commercial primate pellets for New World Monkeys (Sniff, Soest, Germany) offered ad libitum and supplemented with limited amounts of fresh fruit, vegetables, Arabic gum, and homemade porridge.
Tap water is provided ad libitum by way of automatic watering nipples.
As part of routine husbandry, weights of all marmosets are obtained, on average, once weekly. Weights are taken by placing a scale in the animal's home cage, using PRT-training-i.e., marmosets did not have to be handled for weighing. Weights at likely conception dates, approximately 144 days prior to date of birth of the produced litter (Hearn, 1986), were used in the analysis as well as the body weight measurements recorded closest to the date of parturition. All animal procedures, husbandry, and housing were conducted according to BPRC Animal Welfare Body requirements. The research in this study adhered to the American Society of Primatologists' Principles for Ethical Treatment of Non-Human Primates.
The data used for the current study were obtained between August 2009 and February 2018. The data set consisted a total of 149 births in 32 dams and 25 sires for body weight and inter parturition interval analysis (data set 1). Reproductive output per litter size per parturition was defined as number of infants produced at full term gestation, death and alive. For 34 births in 6 dams, dam's litter size was unknown, leaving a total of 115 births in 26 dams and 22 sires for the analysis of factors influencing litter size (data set 2). For 10 births involving three sires, sire's litter size was not known, therefore analyses involving sire's litter size were performed on a dataset with 105 parturitions in 25 dams (data set 3). Dataset 3 was also used to re-run models, not involving sire litter size of origin, to allow comparison of AIC values as this requires the same dataset. Age range of the marmoset dams and sires in data set 2 was 2.2-7.9 years and 2.4-7.9 years, respectively.  The increase in the dams' bodyweight just before parturition compared to the body weight at conception was related to the produced litter size. The mean body weight increase of the dam (95% CI) with singletons, twins, or ≥triplets was 83 gr (61 to 105), 88 gr (81 to 95), and 116 gr (109 to 122), respectively.

| Statistical analysis
Inter birth intervals (IBI) did not differ for the produced litter size.
Over 90% of the litter sizes analyzed were multiples (Table 1).
Most dams were born as part of twin or triplet litters (Table 2). Dams born from litters of ≤2, produced litters of ≤2 in 70 % of cases, whereas dams born from litter size >2 produced litters of >2 in 69% of cases ( Figure 1). Dams born as singleton or twin litters are 2.25 times (1.28-3.94) more likely to produce litters ≤2 (p < 0.05), whereas dams born as triplet litters are 2.28 times (1.52-3.41) more likely to produce litters >2 (p < 0.05). Dams with body weights below the median at conception produced litters ≤2 in 75% of cases, whereas dams with body weights ≥ the median produced litters >2 in 57% of cases Dams with body weights below the median are 1.78 times (1.26-2.59) more likely to produce litters ≤2 (p < 0.05), whereas dams with body weights ≥ the median are 2.33 times (1.42-3.83) more likely to produce litters >2 (p < 0.05) ( Figure 2).
Most sires were born as part of twin and triplet litters (Table 3).
Sires born from litters of ≤2 generally produce litters of ≤2. However, this finding was not statistically significant. An interaction between dam and sire litter size of origin was determined but revealed no significant effect.
Logistic regression analysis, accounting for pseudoreplication, was performed with explanatory variables added in a stepwise manner. The best fitting model based on the lowest Aikake Information Criterion revealed that the dams' litter size of origin, maternal bodyweight and parturition number predicted produced litter size (Table 4). Sires' litter size of origin showed a small non-significant effect, nevertheless, same direction as dams. IBI, age and day of the year when parturition took place did not hold predictive value for produced litter sizes.
To translate from logistic regression estimates to probabilities, a triplet "risk" table (Table 5)

| DISCUSSION
The present study shows that the dams' litter of origin size has predictive value in relation to litter size. Dams born from a twin litter are more likely to give birth to twin litters and dams born from triplet litters are more likely to produce triplets. Therefore, it is tempting to FIGURE 1 Overview of the dams' litter size of origin versus produced litter size (n.a. = not available). Most dams were born as part of twin or triplet litters. Dams born from litters of ≤2, produced litters of ≤2 in 70% of cases, whereas dams born from litter size >2 produced litters of >2 in 69% of cases Overview of the produced litters. Presented in absolute numbers but also in frequency of occurrence. The data set consisted a total of 149 births in 32 dams and 25 sires for body weight and inter parturition interval analysis (data set 1). For 34 births in 6 dams, dam's litter size was unknown, leaving a total of 115 births in 26 dams and 22 sires for the analysis of factors influencing litter size (data set 2). Data show that over 90% of the litter sizes analyzed were multiples. It has been suggested that marmosets may be able to adjust litter size in pregnancy in response to proximate environmental factors as triplet females were not more likely than twin females to produce triplet litters (Jaquish, Cheverud, et al., 1996;Rutherford et al., 2014;. However, the data of this long-term follow-up study show that the dams litter size of origin had predictive value in produced litter size. It has been suggested that stressors, such as noise disturbance, changes in food availability or maternal body condition, or changes in social setting, may be associated with loss of litter (Tardif & Jaquish, 1997). These losses might be way of controlling reproductive investment and ensure that the number of infants produced reflects the dam's possibility to provide optimal feed and care for them. Rutherford et al. (2014) showed that although there are no differences in the number of offspring between triplet females and twin females, the loss of offspring during pregnancy is significantly higher in triplet females. This manuscript reports a finding that is the opposite of that reported on a different marmoset population (Rutherford et al., 2014), the earlier study reporting that females born into triplet litters were not more likely than twin females to produce triplets. It should be taken into account that the study by Rutherford et  Longevity and total reproductive output of the dams were not determined, but it is conceivable that a high reproductive output is at the cost of longevity (Kirkwood, 1977;Westendorp & Kirkwood, 1998). Indirect factors possibly influencing reproductive variation such as group size (i.e., number of potential helpers), and group composition were also excluded from analysis. Where possible, future analysis should include all these parameters.
The influence of the maternal body weight at conception as predictor of litter size may be related to energy availability as marmosets opportunistically adjust their reproductive output in Overview of the dams litter size of origin with absolute number of parturitions produced and the frequency.

FIGURE 2
Dams' body weight at conception versus produced litter size. Dams with body weights below the median at conception produced litters ≤2 in 75% of cases, whereas dams with body weights ≥the median produced litters >2 in 57% of cases response to changes in diet or energy resources (Jaquish, Tardif, Toal, & Carson, 1996;Tardif & Jaquish, 1997;. This confirms the link between higher maternal body weight (energy availability) and higher reproductive output (Ash & Buchanan-Smith, 2014;Box & Hubrecht, 1987;Luke & Keith, 1992;Rutherford et al., 2014;Tardif & Bales, 2004;Tardif & Jaquish, 1997;Tardif et al., 2003;Tardif, Ziegler, Power, & Layne, 2005). Most likely, maternal body condition rather than body weight will be a better  Multivariate analyses were performed by mixed logistic regression models, where a litter size >2 was coded 1 and litters ≤2 as 0. Dam was included as a random variable, thereby accounting for pseudoreplication. Several explanatory variables (dams' litter size of origin, dams' body weight, >2rd litter for dam, dam age, sires' litter size of origin) were sequentially added as fixed variables. Dam weight was entered as deca-gram (dag) thus the parameter estimate indicates the change in log Odds for a 10 g weight change.Collinearity was addressed by calculation of the Variance Inflation Factor (VIF) for each of the explanatory variables in models involving more than one explanatory variable. Aikake Information Criterion (AIC) = 119.5282. | 5 of 7 (Box & Hubrecht, 1987). This result might have been caused by an abundance of high quality food instead of aging. However, Rothe et al. (1992) describes that age or parity does not affect produced litter size while others even described declining litter size with age . In the present study, no age effect on litter size was observed.
The influence of the parturition number on litter size has not been reported before; we observed that triplets are less likely to be produced in the first two litters, but the chance of triplets is increased thereafter.
The absence of seasonal influence on the litter size is in agreement with earlier studies (Box & Hubrecht, 1987;Hearn, 1986). The suggestion of a potential seasonal influence in some wild populations is probably related to the availability of food (Sousa et al., 1999), which is not an issue in captivity.
In general twin and singleton litters do not require food supplementation. However, in captivity, twin and triplet litters are common, which is confirmed in this study. Larger litters have considerably greater perinatal mortality than twin litters, ranging from 30% of infants from triplet litters to 65% from quintuplets (Ash & Buchanan-Smith, 2014;Jaquish, Gage, & Tardif, 1991). Poor coordination of cooperative parental behavior has been suggested as a cause of this loss (Tardif, Layne, Cancino, & Smucny, 2002), but also infant condition might play a role .
It is clear that, in this species, infants must be able to cling and locomote well, from birth, if they are to survive (Rothe, 1974). Another cause could be due to lack of milk of the dam to feed three infants.
Irrespective the cause, one of the three infants of a triplet often loses weight within 2-3 days and dies within a week of birth (Hearn & Burden, 1979;Hearn et al., 1975). Triplets can also be cross fostered if an appropriate dam is available. It must be realized that dams seem to dry off quickly, if the foster dam is more than 3 days out from the last time she nursed, she probably doesn't have any milk left. To improve survival of triplets, infants from triplet litters can be rotationally handreared. In general, this human intervention is necessary to get them through the first 2-3 weeks. Despite the above described human intervention, triplets are still associated with higher infant mortality (Ash & Buchanan-Smith, 2014;Hearn & Burden, 1979). Hand-rearing techniques have also serious adverse effects, as normal maternal and family relations are disturbed. The early experiences of marmosets are critical in influencing their development and ability to cope with later events, changes or other stressors (Dettling, Feldon, & Pryce, 2002;Dettling, Schnell, Maier, Feldon, & Pryce, 2007;Pryce, Dettling, Spengler, Schnell, & Feldon, 2004;Pryce et al., 2005). As alternative for the cross fostering and hand-rearing, one of the infants of a triplet can be euthanized at day 1 to bring total litter size down to two. Besides ethical dilemmas this also has an emotional impact. The human-animal bond positively impacts quality of life for research animals, but staff caring for the animals often experience euthanasia-related stress symptoms comparable to those encountered in veterinary clinics and animal shelters. Constant exposure to or participation in euthanasia procedures can cause a psychological state characterized by a strong sense of work dissatisfaction, alienation, or careless and callous handling of animals. The results of this study can potentially be used as an additional measure to the various breeding policies that are currently applied to help preventing production of triplet litters as much as possible.
By defining the factors that predict the produced litter size per parturition, practical aspects of managing marmoset breeding colonies can be enhanced. Maintaining a colony of breeders, with longer healthy life spans and an increased incidence of twin litters could have far reaching implications to improve the quality of life for marmosets in breeding facilities.