In this prospective cohort pilot study, subjects were recruited from Tufts Medical Center in Boston, Massachusetts between May 2019 and September 2021. Eligible infants were born at ≥ 35 weeks’ gestation. Infants with a birth weight < 10^th percentile (28), or with a history of intrauterine growth restriction, maternal diagnosis of preeclampsia or type I DM, or major congenital or chromosomal anomalies were excluded to avoid infants at risk for fetal growth restriction or placental insufficiency that may have skewed the data. Infants with fetal growth restriction may have a different physiologic response due to the paucity of nutrition in utero. Infants with in utero opioid exposure were also excluded based on prior research in this population showing alterations in hunger signaling and hypothalamic gene expression (29). We studied two cohorts of infants—IDMs and infants born to mothers with normal glycemic tolerance during pregnancy. Infants in the diabetic cohort were defined as infants born to mothers with GDM during the current pregnancy or a history of type II DM. Infants in the normoglycemic cohort were born to mothers with normal glucose tolerance testing during pregnancy.

Study eligibility was assessed upon maternal admission to the labor and delivery unit or while the infant was admitted to the newborn nursery or neonatal intensive care unit. All study data, measurements and samples were collected during the infant’s hospitalization. Maternal pregnancy information and demographic characteristics, as well as infant growth parameters, glucose levels when acquired, medications, and feeding data were extracted from the medical record. Infants were characterized as having feeding issues if there was documentation in the medical record about feeding concerns, such as prolonged hospitalization for suboptimal oral intake in the newborn nursery or requiring nasogastric tube for feeding due to poor oral intake. Since infant sex may influence salivary gene expression of hypothalamic and hunger signaling genes (29), efforts were made to recruit an equal number of male and female infants in each cohort. This study was reviewed and approved by the IRB at Tufts Medical Center. Written informed consent was obtained from a parent or guardian prior to infant participation in this study.

Infant length was measured to the nearest 0.1 cm using an infant length board. Two measurements were taken and if the difference between the two measurements was greater than or equal to 0.5 cm, a third measurement was obtained. The infant’s length was recorded as the average of all measurements. Skin fold thickness at the left hip was measured using calibrated skinfold calipers (Harpenden) and was measured to the nearest 0.1 mm. Two measurements were taken and if the difference between the two measurements was greater than or equal to 0.5 mm, a third measurement was obtained. Infant weight, length, and skinfold measurements were inputted into the Catalano formula to calculate fat mass in kg and percent fat mass (30). Percentiles were calculated for infant fat mass and percent fat mass taking into account infant sex and gestational age at birth using body composition curves published by Norris, et al., 2019 (31). Infant birth weight percentiles also took into account infant sex and gestational age using the Olsen, 2010 growth curves (28, 32). Infant weight was measured using a bedside scale and extracted from the medical record.

Gentle suctioning of the oropharynx was performed to collect two infant salivary samples anytime during the birth hospitalization. To minimize breast milk and associated maternal RNA contamination, saliva was collected prior to feeding when possible. The samples were placed in an RNA stabilization solution and were stored until salivary RNA extraction was performed using RNeasy Micro Kit (Qiagen, Hilden, Germany) per manufacturer’s instructions. On column DNase treatment was performed using RNase-free DNase I Set (Qiagen, Hilden, Germany) to minimize genomic contamination. Samples were then converted to complementary DNA (cDNA) using SuperScript VILO cDNA Synthesis Kit (Invitrogen, Carlsbad, California) and underwent targeted preamplification using TaqMan PreAmp Master Mix Kit (Applied Biosystems, Foster City, California). Detailed methods for salivary sample processing are outlined in the methods section of Yen, et al., 2019 (29).

Quantitative real-time polymerase chain reaction (qPCR) on cDNA was performed to relatively quantify gene expression of NPY2R, PRKAA2, leptin (LEP), ghrelin (GHRL), POMC and adiponectin (ADIPOQ) as previously described by Yen, et al., 2019 (29). Samples were run on Custom TaqMan Array qPCR plates (Applied Biosystems, Foster City, California). Each plate was run with a control human mRNA sample provided by Applied Biosystems. Reference genes (YWHAZ, GAPDH, and HPRT1) demonstrated by Khanna et al. to have stable expression across sex and gestational age, were quantified for each sample (33). Genes of interest (NPY2R, PRKAA2, LEP, GHRL, POMC and ADIPOQ) were run in duplicate for each patient sample. Amplification of all 3 reference genes as an assessment of RNA integrity in each sample was a prerequisite for inclusion of samples in the final analysis. If any reference gene failed to amplify, the secondary sample was analyzed as described above. A negative water control was run once to ensure there were no unintended amplification products (primer dimers).

Data analysis was conducted using R version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). Maternal and infant demographics were summarized using means and standard deviations for parametric continuous data and medians and interquartile ranges for non-parametric continuous data. Salivary gene expression was treated as a binary variable. If cycle threshold (Ct) for amplification was met by 40 cycles, the gene was noted to be present in the saliva sample. If the Ct was not met by 40 cycles, the gene did not amplify and was deemed to be absent in the saliva sample. Salivary gene expression and body composition were summarized by cohort and feeding outcome. Salivary gene expression was compared between our 2 study cohorts using Fisher’s exact test. Body composition measurements were compared between infants born to diabetic and normoglycemic mothers using the Wilcoxon rank sum test. Univariate logistic regression analyses using binary hunger signaling gene expression of PRKAA2 and NPY2R as the outcome variable were run to explore associations between infant body composition or expression of energy homeostasis regulators (POMC and GHRL) and hunger signaling genes independent of study cohort. Given the nature of a pilot study with inherent limitations of sample size, regression analyses were done to provide estimates and variability.

A total of 45 infants were recruited, with 22 infants in the normoglycemia group and 23 in the diabetic group. Mothers diagnosed with diabetes were older and had higher pre-pregnancy BMIs as compared to mothers with normoglycemia during pregnancy. IDMs had lower gestational ages at birth and higher birth weight for age percentiles (28) as compared to the normoglycemia cohort. There were no infants in the normoglycemia cohort with feeding issues and only 3 infants (13%) in the diabetic cohort with feeding issues noted in the medical record ( Table 1 ).

Salivary samples were collected once from all 45 infants and collection time ranged from day of life 0 to 7. Three samples from the diabetic cohort did not amplify one of the three reference genes, and so the secondary sample was processed, which passed quality control. LEP was only able to be quantified in three samples in the normoglycemia cohort and one sample in the diabetic cohort. ADIPOQ was only detectable in one sample in the normoglycemia cohort. PRKAA2, GHRL, and NPY2R were less likely to be detected in the diabetic cohort, whereas POMC was more likely to be detected in the diabetic cohort ( Table 2 ). There was no gene expression detected for the negative water control. The positive control had minimal plate-to-plate variability with respect to the geomean of the three reference genes (mean 23.7± 0.4).

Body composition measurements were collected once from all 45 infants and collection time ranged from day of life 0 to 7. Infants born to mothers with diabetes had a slightly higher absolute fat mass as well as percentage of fat mass as compared to infants born to mothers who were normoglycemic, despite having a lower gestational age at birth. When looking at body composition percentiles by gestational age and sex using body composition growth curves (31), infants in the diabetic cohort had a greater fat mass percentile compared to infants born to mothers who were normoglycemic during their pregnancy ( Table 2 ).

Three infants in the diabetic cohort had poor feeding documented in their medical record, while none in the control group had documented feeding issues. These infants were eventually able to eat by mouth and were discharged home. No additional work up was required for poor feeding. Generally, infants with documented feeding issues were born at a lower gestational age, with a median age of 37.6 weeks as compared to infants without feeding issues, with a median age of 39.1 weeks. Poor feeders had a greater birth weight of 3.57 kg and birth percentile of 72, as compared to feeders who had a median birth weight of 3.38 kg and birth percentile of 50.5. The poor feeders also had a greater fat mass of 0.45 kg and percentage of fat of 14.2, as compared to infants without feeding issues which was 0.33 kg and 10.3 respectively. Feeders were more likely to express GHRL, NPY2R and POMC and less likely to express PRKAA2 than non-feeders. Greater infant adiposity was associated with lower odds of expression of NPY2R. Expression of GHRL was associated with increased odds of expression of PRKAA2 ( Table 3 ).

This pilot study was able to demonstrate the feasibility of detecting PRKAA2, NPY2R, GHRL, and POMC in saliva of our study population, but we were unable to reliably detect LEP and ADIPOQ. These results do not necessarily mean that leptin and adiponectin are not expressed in infant saliva, as there may be more sensitive assays that could detect their gene expression. It is important to consider whether one should target gene expression for the hormone of interest or the receptor. While we were unable to detect LEP in our study, prior studies have detected LEPR expression in neonatal saliva (29). Receptor expression may be better quantified in saliva. We did not measure serum levels of target hormones for comparative analysis.

Neonatal saliva is a non-invasive biofluid that contains a wealth of information which can be collected serially with minimal impact on the infant (25). Salivary diagnostics may be helpful in not only identifying infants at risk for poor feeding, but more importantly, elucidating disruption in regulatory pathways responsible for outcomes. The salivary transcriptome also allows for non-invasive serial monitoring of changes in energy regulators and hunger signaling over time.

PRKAA2, GHRL, and NPY2R were less likely to be detected in the diabetic cohort, whereas POMC was more likely to be detected in the diabetic cohort. Decreased expression of PRKAA2 and GHRL and increased expression of POMC is consistent with biological mechanisms relating these hormones to adiposity, decreased oral intake and poor feeding outcomes. Post-hoc power analysis yielded a power of 3 to 12 percent, given the small sample size.

Infants in the diabetic cohort had greater overall fat mass and percent fat mass despite being born at a younger gestational age. It has been shown that body fat mass and body fat percentage increases as gestational age increases and females have greater adiposity than males (31). To account for changes in body composition with gestational age and sex, Norris et al (31) compiled body composition data from four studies in the United States and Ireland to create body composition curves stratified by percentile. Our cohort generally had lower fat mass, but a greater percentage of fat mass to total body weight, as compared to the cohort used to derive the body composition curves. This may be due to differences in modalities used to measure body composition, as we used anthropometric measurements and skin fold thickness rather than air displacement plethysmography. Future studies may consider adjusting for infant sex and gestational age when interpreting body composition measurements or using body composition curves, although care should be taken in considering how body composition data are obtained.

Our study utilized non-invasive methods to investigate the relationship between body composition and regulation of appetite through adipoinsular and hypothalamic hormones in infants of diabetic mothers. By demonstrating the feasibility of salivary gene expression analysis of these pathways, future investigators can study how these pathways change over time. This method may be preferred over cord blood which is also non-invasive but only provides information on the in utero environment, as well as serial blood draws which are challenging to obtain and inflict harm. Since neonatal normative serum gene expression levels for our target hormones do not exist, we need not perform a comparative analysis between serum and saliva to determine a gold standard biofluid for measurements. Rather, we aimed to determine if levels in saliva would be independently informative of feeding outcomes. While it is possible that salivary gene expression is impacted by the oral microbiome and metabolome, it was beyond the scope of this study to determine what if any impact these would have on gene expression (34).

One limitation of our study was that maternal glucose control and treatment with medications during pregnancy in the diabetic mothers was not quantified but may be an important factor related to infant adiposity, energy homeostasis, and hunger signaling. It may also be informative to separate infants born to mothers with pregestational versus gestational diabetes, as the duration of exposure to an elevated glucose concentration varies between the groups. However, we were unable to do so given our sample size. Moreover, placental nutrient metabolism and delivery impact (35) on the newborn is outside the scope of the current project. An additional limitation of this study is that only three infants were noted to have feeding difficulties, which is far fewer than the 37% of IDMs cited in prior literature (36). It will be important for future studies to specifically target IDMs who struggle to feed, as these infants may be more likely to have dysregulation of their hypothalamic signaling and/or energy expenditure pathways directly impacting feeding success.