Cord Blood-Derived Exosomal CNTN2 and BDNF: Potential Molecular Markers for Brain Health of Neonates at Risk for Iron Deficiency

Maternal iron deficiency anemia, obesity, and diabetes are prevalent during pregnancy. All are associated with neonatal brain iron deficiency (ID) and neurodevelopmental impairment. Exosomes are extracellular vesicles involved in cell–cell communication. Contactin-2 (CNTN2), a neural-specific glycoprotein, and brain-derived neurotrophic factor (BDNF) are important in neurodevelopment and found in exosomes. We hypothesized that exosomal CNTN2 and BDNF identify infants at risk for brain ID. Umbilical cord blood samples were measured for iron status. Maternal anemia, diabetes, and body mass index (BMI) were recorded. Cord blood exosomes were isolated and validated for the exosomal marker CD81 and the neural-specific exosomal marker CNTN2. Exosomal CNTN2 and BDNF levels were quantified by ELISA. Analysis of CNTN2 and BDNF levels as predictors of cord blood iron indices showed a direct correlation between CNTN2 and ferritin in all neonates (n = 79, β = 1.75, p = 0.02). In contrast, BDNF levels inversely correlated with ferritin (β = −1.20, p = 0.03), with stronger association in female neonates (n = 37, β = −1.35, p = 0.06), although there is no evidence of a sex-specific effect. Analysis of maternal risk factors for neonatal brain ID as predictors of exosomal CNTN2 and BDNF levels showed sex-specific relationships between infants of diabetic mothers (IDMs) and CNTN2 levels (Interaction p = 0.0005). While male IDMs exhibited a negative correlation (n = 42, β = −0.69, p = 0.02), female IDMs showed a positive correlation (n = 37, β = 0.92, p = 0.01) with CNTN2. A negative correlation between BNDF and maternal BMI was found with stronger association in female neonates (per 10 units BMI, β = −0.60, p = 0.04). These findings suggest CNTN2 and BNDF are respective molecular markers for male and female neonates at risk for brain ID. This study supports the potential of exosomal markers to assess neonatal brain status in at-risk infants.


Introduction
Fundamentally, iron is a critical micronutrient for tissue oxygenation, cellular metabolism, energy generation, and the metabolism of toxins [1]. Iron and iron-containing enzymes in the brain are involved in neuronal energy metabolism, myelination, and neurotransmission [2]. Iron deficiency (ID) occurs when the body lacks sufficient iron to supply these needs. When left untreated, ID can progress to iron deficiency anemia (IDA). Anemia impacts more people globally than any other health impact early life iron status, particularly MOD. Given the evidence of impaired neural development in early-life ID and impaired cognitive function of IDMs [2,9,13,18], we hypothesized that neonates exposed to these risk factors would have lower levels of circulating exosomal CNTN2 and BDNF.

Study Groups
Infants delivered at UnityPoint Meriter Hospital, a tertiary care teaching hospital with 3900 delivers per year, were recruited between 2008 and 2011. The study inclusion criteria included healthy newborns born to English and Spanish-speaking women between 18 and 40 years of age, who were delivered ≥35 weeks of gestation. The newborns were without infection or other complication. Informed consent was obtained for all subjects in the study, which was approved by the Institutional Review Boards from the University of Wisconsin-Madison and Meriter Hospital (Meriter IRB #: 2011-005). Electronic medical records were screened for known risk factors for infantile IDA. This study cohort contains new data generated from archived specimens collected from a larger previously published birth cohort [13].

Population Characteristics
Population characteristics were previously described [13]. Briefly, maternal age and diabetes mellitus (both pre-existing and gestational) at delivery were assessed. Maternal obesity was evaluated using morphometric measures and pre-pregnancy and delivery body mass index (BMI kg/m 2 ) was determined.

Cord Blood Sample Collection and Lab Tests
EDTA-anticoagulated umbilical cord blood samples were obtained at both vaginal and Caesarean deliveries, stored at 4 • C, and processed within 8 days. Plasma was aliquoted and stored at −80 • C until assayed. Plasma ferritin was quantified with a commercial human enzyme-linked immunosorbent assay (ELISA) kit (Bio-Quant). Zinc protoporphyrin/heme (ZnPP/H) was measured in whole blood using a Front-Face Hematofluorometer (Aviv Biomedical) [10]. ZnPP/H rises as zinc substitutes for iron when iron is deficient in the protoporphyrin ring, measuring red blood cell ID [13]. Hematocrit was measured in whole blood by PoCH-100i hematology analyzer (Sysmex) [13].

Exosome Isolation
Cord blood exosomes were isolated using a previously described protocol [40]. In short, precipitation solution (50% PEG8000, 0.5M NaCl in DPBS) was added to 50 µL of lysed whole blood at a 1:6 volume ratio for a final PEG8000 concentration of 8.3%. The whole blood and precipitation solution combination was mixed by repeated inversion overnight at 4 • C. The samples were then centrifuged at 2000 g for 30 minutes at 4 • C to pellet exosomes. Supernatant was removed and pellets were re-suspended in lysis buffer (150 mM NaCl, 50 mM Tris pH 7.6, 1% Igepal CA-630) for BDNF or Reagent Diluent (DuoSet ELISA Ancillary Reagent Kit DY008) for CNTN2. Re-suspended pellets were stored at −80 • C.

Dot Blot and Western Blot Validation of Exosomes
Exosomal enrichment was validated by dot blot and Western blot (WB) for the exosomal marker CD81 and the neural-specific exosomal marker CNTN2 using a previously described protocol [37]. In brief, for the dot blot experiment, serial dilutions in PBS of re-suspended exosomal pellet, supernatant, and input whole blood samples were blotted onto a nitrocellulose membrane, blocked with blocking buffer for fluorescent Western blotting, and incubated with anti-CD81 (1:10,000×, mouse monoclonal, RnD Systems) and anti-CNTN2 (1:10,000×, mouse monoclonal, RnD Systems) antibodies for 1 hour at room temperature. Following PBS + 0.1% Tween-20 washes, blots were incubated with Alexa700-anti mouse IgG (Rockland). WB images were captured using Odyssey Infrared Imaging System (LI-COR Biosciences). For the Western blot, 10 µg of protein quantified by Bradford assay (Sigma) from re-suspended exosomal pellets, supernatants, and input whole blood samples was separated using a gradient 4%-20% SDS-PAGE gel (Invitrogen). Proteins were blotted onto a nitrocellulose membrane, blocked with blocking buffer for fluorescent Western blotting, and incubated with anti-CNTN2 (1:1500×, mouse monoclonal, RnD Systems) overnight at 4 • C. Following PBS ± 0.1% Tween-20 washes, blots were incubated with Alexa700-anti mouse IgG (Rockland). WB images were captured using Odyssey Infrared Imaging System.

ELISA Quantification of CNTN2 and BDNF
Cord blood exosomal CNTN2 and BDNF levels were quantified by ELISA. CNTN2 levels were determined using DuoSet Human CNTN2/TAG1 ELISA Kit (DY1714-05) and DuoSet ELISA Ancillary Reagent Kit (DY008) using the protocol provided (R & D Systems). BDNF levels were determined using Quantikine Total BDNF ELISA Kit (DBNT00) using the protocol provided (R & D Systems). CNTN2 and BDNF protein concentrations were determined using standard curves (CNTN2 R 2 = 0.96−0.99, BDNF R 2 = 0.99) created using the optical density of known concentrations of a standard solution per the manufacturer's protocols, and a best fit line was generated using linear regression analysis. The flowchart of experimental study and the validation of exosomal isolation are shown in Figure 1. The levels of exosomal CNTN2 and BDNF in 50 µL of cord blood samples were calculated using the following formula: Infrared Imaging System (LI-COR Biosciences). For the Western blot, 10 µg of protein quantified by Bradford assay (Sigma) from re-suspended exosomal pellets, supernatants, and input whole blood samples was separated using a gradient 4%-20% SDS-PAGE gel (Invitrogen). Proteins were blotted onto a nitrocellulose membrane, blocked with blocking buffer for fluorescent Western blotting, and incubated with anti-CNTN2 (1:1500×, mouse monoclonal, RnD Systems) overnight at 4 °C. Following PBS ± 0.1% Tween-20 washes, blots were incubated with Alexa700-anti mouse IgG (Rockland). WB images were captured using Odyssey Infrared Imaging System.

ELISA Quantification of CNTN2 and BDNF
Cord blood exosomal CNTN2 and BDNF levels were quantified by ELISA. CNTN2 levels were determined using DuoSet Human CNTN2/TAG1 ELISA Kit (DY1714-05) and DuoSet ELISA Ancillary Reagent Kit (DY008) using the protocol provided (R & D Systems). BDNF levels were determined using Quantikine Total BDNF ELISA Kit (DBNT00) using the protocol provided (R & D Systems). CNTN2 and BDNF protein concentrations were determined using standard curves (CNTN2 R 2 = 0.96−0.99, BDNF R 2 = 0.99) created using the optical density of known concentrations of a standard solution per the manufacturer's protocols, and a best fit line was generated using linear regression analysis. The flowchart of experimental study and the validation of exosomal isolation are shown in Figure 1. The levels of exosomal CNTN2 and BDNF in 50 µL of cord blood samples were calculated using the following formula:

Statistical Analysis
Due to the limited volume of available cord blood samples, there were a number of neonates who were missing values for either CNTN2 or BDNF: 38 have both, 20 and 21 are missing CNTN2 and BDNF, respectively. There was some evidence of differences in other characteristics according to whether CNTN2 or BDNF were missing. While those missing CNTN2 tended to have lower BMI and risk levels, those missing BDNF tended to have lower HCT. To account for these differences and to compare relationships with BDNF and CNTN2 among the same group of patients, a multiple imputation strategy was employed. Using the Multiple Imputation using Chained Equations procedure [41], 10 data sets were created with missing values imputed stochastically using the observed patterns among the non-missing data. Linear regression models were fit in each data set and the results were pooled using the Rubin method to account for imputation uncertainty. The pooled results are shown in Tables 2 and 3

Statistical Analysis
Due to the limited volume of available cord blood samples, there were a number of neonates who were missing values for either CNTN2 or BDNF: 38 have both, 20 and 21 are missing CNTN2 and BDNF, respectively. There was some evidence of differences in other characteristics according to whether CNTN2 or BDNF were missing. While those missing CNTN2 tended to have lower BMI and risk levels, those missing BDNF tended to have lower HCT. To account for these differences and to compare relationships with BDNF and CNTN2 among the same group of patients, a multiple imputation strategy was employed. Using the Multiple Imputation using Chained Equations procedure [41], 10 data sets were created with missing values imputed stochastically using the observed patterns among the non-missing data. Linear regression models were fit in each data set and the results were  Tables 2 and 3. CNTN2 and BDNF response variables were standardized using the empirical normal quantile transform to satisfy distributional assumptions. Predictor-by-sex interaction terms were included to test for sex-specific relationships. Analyses were conducted using R version 3.5.2 (R Foundation for Statistical Computing, Vienna, Austria) [42].

Results
A total of 42 male and 37 female neonates were included in the study. No significant sex difference was found in terms of maternal age, gestational age at delivery, birth weight, or birth weight for gestational age z-score, as well as between neonatal indices, including ferritin, hematocrit, and zinc protoporphyrin ( Table 1). The composition of studied neonates included approximately 29%-30% from diabetic mothers, 55%-60% from obese mothers, and 24-27% from anemic mothers ( Table 1). Analysis of CNTN2 and BDNF levels as predictors of Fe-related outcomes among all, female, and male neonates ( Table 2) showed a positive relationship between CNTN2 and ferritin level for all neonates (β = 1.75, p = 0.02). Based on this analysis, it is estimated that among all neonates, a one unit increase in CNTN2 (an increase of 20 pg/mL) is associated with a mean increase in cord blood ferritin of 1.75 ng/mL (95% CI: 0.28−3.21, p = 0.02). Though there was no evidence that the slopes for males and females differ (predictor-by-sex interaction p = 0.43)-male neonates tended to have a stronger association (β = 2.78, p = 0.07) compared to female (β = 1.45, p = 0.11) neonates. Conversely, a negative relationship was found between BDNF and ferritin level for all neonates (β = −1.20, p = 0.03) with a stronger association with female (β =·1.35, p = 0.06) than male (β = −1.07, p = 0.18) neonates, which means a one unit increase in BDNF (20 pg/mL) is associated with a mean decrease of 1.35 ng/mL ferritin.
Analysis of neonatal risk factors for brain ID (baby weight Z-score, maternal BMI, maternal anemia, and maternal diabetes) as predictors of exosomal CNTN2 and BDNF levels were performed (Table 3). Here, CNTN2 and BDNF were transformed to standard normal variables (Z-score) using the empirical normal quantile transform to account for their skewed distribution. The results showed anemia (vs. no anemia) tended to be associated with an increase in BDNF Z-score of 0.45 (95% CI: 0.17-1.07, p = 0.14); this relationship appears stronger in females (β = 0.65, p = 0.08) than males (β = 0.24, p = 0.59), although there is no evidence of sex difference (p = 0.45). These changes mean that neonates from anemic mothers (vs. non-anemic mothers) are associated with an average increase of 22% exosomal BDNF level. There was a strong sex-specific relationship between maternal diabetes and exosomal CNTN2 (predictor-by-sex interaction p = 0.0005). While females showed a positive correlation (β = 0.92, p = 0.01), males showed a negative correlation (β = −0.69, p = 0.02). When stratified by sex, baseline CNTN2 levels were not different between male and female neonates; however, they showed a sex-differential response when exposed to maternal diabetes ( Figure 2). CNTN2 and BDNF levels were analyzed as predictors of iron-related markers in all, male, and female neonates. A significant positive relationship was found between CNTN2 and ferritin level for all neonates, with a stronger association in males. A significant negative relationship was found between BDNF and ferritin level for all neonates, with a stronger association in females. Asterisk indicates a significant result (* p < 0.05). Neonatal risk factors were analyzed as predictors of CNTN2 and BDNF levels. Maternal diabetes showed a positive association with CNTN2 in females and a negative association in males. Asterisk indicates a significant result (* p < 0.05, ** p ≤ 0.01; *** p < 0.001). Table 2. Contactin-2 (CNTN2) and brain-derived neurotrophic factor (BDNF) as predictors of ironrelated markers. CNTN2 and BDNF levels were analyzed as predictors of iron-related markers in all, male, and female neonates. A significant positive relationship was found between CNTN2 and ferritin level for all neonates, with a stronger association in males. A significant negative relationship was found between BDNF and ferritin level for all neonates, with a stronger association in females. Asterisk indicates a significant result (* p < 0.05).

Discussion
IDA is a prevalent consequence of micronutrient deficiency, affecting approximately 19% of pregnant women and 18% of preschool-age children globally [4]. Identification and treatment of ID

Discussion
IDA is a prevalent consequence of micronutrient deficiency, affecting approximately 19% of pregnant women and 18% of preschool-age children globally [4]. Identification and treatment of ID at 9-12 months of life is a late intervention that does not wholly eliminate deficits in behavioral and neurodevelopmental outcomes in affected children [2,10]. One potential explanation of this finding is the recognition that poor fetal/neonatal iron endowment, congenital ID, occurs in neonates with certain risk factors [13][14][15][16][17] and that iron is preferentially shuttled to the periphery for erythropoiesis in states of deficiency leaving less iron available for brain metabolism before ID is traditionally detected in the blood [15,16]. Thus, it is imperative to develop new clinical tools that can non-invasively index brain iron status, thereby allowing earlier therapeutic iron repletion. This is the first study that examined the relationship between neurological markers CNTN2 and BDNF in circulating exosomes with biological predictors for poor neonatal iron endowment, including maternal anemia, obesity, and diabetes. This is a hypothesis-generating study to suggest further research into molecular markers that could be used to appraise brain iron status at birth or early infancy. In a broader sense, this study adds to the growing evidence supporting the feasibility and utility of markers found within exosomes as indicators of parental cell function and dysfunction. Exosomal CNTN2 and BDNF levels in umbilical cord blood were assessed using ELISA and were compared across groups stratified by maternal and neonatal BMI, maternal iron status, and maternal diabetic status.
The direct correlation between neural-specific exosomal CNTN2 and cord blood ferritin suggests a putative molecular marker for assessing neural impairment associated with brain ID in neonates, particularly in males. The finding of negative relationship between IDM males and exosomal CNTN2 as a variable further underscores the relationship between CNTN2 and brain iron status. Previous studies show that IDMs are at risk for brain ID accompanied by impaired cognitive function [14,18]. Our data suggest that CNTN2 levels may reflect impairment of brain development and indirectly brain iron stores, which are likely deficient in infants of maternal IDA or chronic hypoxic states like IDMs [15]. CNTN2 could be used as a molecular marker newborn screen to identify neonates at risk for brain ID. Although more studies are needed, it is important to identify these infants early to allow timely interventions that are likely more effective in preventing long-term abnormal neurodevelopment.
To date, no mechanism is known for the interaction of CNTN2 with nutritional status. However, CNTN2 is cleaved by β-site APP-cleaving enzyme 1 (BACE1), which generates soluble CNTN2 and decreases surface levels of CNTN2 [30]. BACE1 also cleaves amyloid-β precursor protein (β-APP), whose levels correlate with iron levels [43]. In line with this concept, the function of α-secretase, another enzyme that cleaves β-APP, is modulated by furin, an iron-dependent enzyme [43]. Moreover, early-life ID alters β-APP signaling and metabolism [44]. Taken together, it is likely that CNTN2 is cleaved by enzymes that are directly or indirectly modulated by iron, which could underlie the connection between brain iron status and CNTN2 levels. This proposed mechanism could be further investigated in future studies.
Conversely, although less robust, the negative relationship between maternal BMI and exosomal BDNF (as a variable) suggests a possible molecular marker for indexing neural impairment associated with brain ID specifically in female neonates. These data imply that BDNF may also correlate with brain-specific iron stores. This is plausible as preclinical studies have shown that brain BDNF levels are reduced by fetal/neonatal ID and that ID induces chromatin remodeling of the BDNF gene [36,37]. In support, clinical studies showed lower cord plasma BDNF with maternal ID [45]. Collectively, these findings suggest that BDNF could be used as a metric to identify female neonates at risk for the neural consequences of brain ID.
Our study found novel sex differences in the effects of iron status on CNTN2 and BDNF levels, with CNTN2 correlating with risk of ID more strongly in male neonates and BDNF correlating more strongly in female neonates. The mechanism underlying these gender differences remains to be elucidated. Male infants are known to be more susceptible to IDA in early infancy [13,46,47]. Similar to other early-life injurious events, male neonates show more severe effects than female neonates with comparable injuries [48,49]. Differential inflammatory responses, particularly cytokine expression and microglia activation, have been proposed to underlie such sex-specific effects [50]. Iron metabolism is known to be affected by inflammation [51] and sex-specific differences in inflammatory pathways may mediate the differences seen in iron-related molecular marker candidates. Other possible mechanisms include neuroprotective effects of estrogen [52] and differential activation of apoptotic pathways [48], all of which might interact to produce the sex-specific responses seen in this study. There were limitations of this study. A significant limitation is a lack of brain iron status that could be used to validate these findings. Future studies could establish this correlation using cerebrospinal fluids and brain samples from primate and non-primate models of fetal/neonatal IDA. Some metrics in this study were inconsistent across different groups. For example, the inverse relationships of cord ferritin and maternal anemia with exosomal BDNF in female neonates suggest additional roles for circulating BDNF. It is possible that the BDNF responses in female neonates may have neuroprotective activity. Finally, the limited sample size in this study could not fully investigate the variability of fetal influences on these exosomal markers. Thus, additional studies are needed to establish the basis behind these differential responses.

Conclusions
This study found that exosomal CNTN2 is an attractive candidate for further determination as a molecular marker for neonates at risk for brain ID, particularly for male neonates. Conversely, BDNF is likely a better marker for female infants at risk for brain ID. The neurological and behavioral effects of early-life IDA are significant and long lasting. Moreover, IDA is often detected after the brain may have already been iron deficient for some time. This study provides possible tools for non-invasive assessment of brain health related to iron status and provide an early, evidence-based interventional approach to prevent progression to IDA. Employing earlier evidence-based tools could then synergize with needed global nutritional interventions aimed to prevent IDA in early childhood. An important future step is to evaluate behavioral outcomes in the infants from whom umbilical cord samples were collected to determine whether the candidate molecular markers found in this study correlate with neurodevelopmental deficits associated with ID. Additional future directions could include the isolation of CNTN2-specific exosomes to identify associated markers, allowing the elucidation of the pathways underlying iron homeostasis and circulating exosomal CNTN2.