Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Dec 1.
Published in final edited form as: Environ Res. 2020 Aug 22;191:110113. doi: 10.1016/j.envres.2020.110113

Association between prenatal immune phenotyping and cord blood leukocyte telomere length in the PRISM pregnancy cohort

E Colicino 1, W Cowell 2, A Bozack 2, Pedretti N Foppa 1, A Joshi 1, MM Niedzwiecki 1, V Bollati 3, C Berin 2, RO Wright 1,2, RJ Wright 1,2,4
PMCID: PMC7883408  NIHMSID: NIHMS1626368  PMID: 32841635

Abstract

Background:

Environmental exposures including air pollutants, toxic metals, and psychosocial stress have been associated with shorter telomere length (TL) in newborns. These exposures have in turn been linked to an enhanced inflammatory immune response. Increased inflammation during pregnancy may be a central biological pathway linking environmental factors with reduced TL at birth. Approaches that more comprehensively characterize the prenatal inflammatory milieu rather than targeting specific individual cytokines in relation to newborn TL may better elucidate inflammatory mechanisms.

Methods:

Analyses included 129 mother-child dyads enrolled in the PRogramming of Intergenerational Stress Mechanisms (PRISM) pregnancy cohort. We measured 92 inflammation related proteins during pregnancy in maternal serum using the Olink protein array and quantified cord blood relative leukocyte TL (rLTL) via qPCR. We leveraged a tree-based machine learning algorithm to select the most important inflammatory related proteins jointly associated with rLTL. We then evaluated the combined association between the selected proteins with rLTL using Bayesian Weighted Quantile Sum (BWQS) Regression. Analyses were adjusted for gestational week of serum collection, maternal race/ethnicity, age, and education, and fetal sex. We evaluated major biological function of the identified proteins by using the UniProtKB, a centralized repository of curated functional information.

Results:

Three proteins were negatively and linearly associated with rLTL (CASP8 β: −0.22 p=0.008, BNGF β: −0.43 p=0.033, TRANCE β:−0.38 p=0.004). Results from BWQS regression showed a significant overall decrease in rLTL (β: −0.26 95%CrI: −0.43, −0.07) per quartile increase of the mixture, with CASP8 contributing the greatest weight (CASP8 50%; BNGF 27%, and TRANCE 23%). The identified proteins were involved in the regulation of apoptotic processes and cell proliferation.

Conclusions:

This proteomics approach identifies novel maternal prenatal inflammatory protein biomarkers associated with shortened rLTL in newborns.

Keywords: immune, Olink, telomere length, inflammation, birth, exposomics

Introduction

Telomeres and their associated proteins play a key role in maintaining the integrity and stability of both the genome and cells, for example by protecting against the loss of protein-coding DNA sequences and preventing inappropriate chromosome fusion during cell division among other roles (Blackburn, 2001; Blasco, 2005; Bosquet Enlow et al., 2019). Telomere length (TL) reflects the replicative history of cells and thus can be considered a mitotic clock and biomarker for aging and cellular function (Aviv, 2008). Age-adjusted shorter TL has been repeatedly linked with premature all-cause mortality and age-related disease risk in studies of adults (Epel et al., 2008; Kimura et al., 2008; RM et al., 2003). In addition, evidence supports that newborn TL is a key driver of inter-individual variability in TL across the lifespan (Aviv, 2008). Although TL is highly heritable, genetic variants explain only a small proportion of variability in TL at birth (Codd et al., 2010; Levy et al., 2010; Prescott et al., 2011), thus highlighting the potential importance of other prenatal factors in the setting of newborn TL.

Prenatal environmental exposures including air pollutants, toxic metals, and psychosocial stress have been associated with shortened TL in newborns (Bosquet Enlow et al., 2018; Marchetto et al., 2016; Rosa et al., 2019; Wai et al., 2018; Zhang et al., 2019). A potential mechanism of telomere shortening in response to these exposures includes an enhanced maternal inflammatory immune response. Innate immune cells possess receptors that signal the activation and production of an array of biologically active proteins in response to environmental stimuli (Newton and Dixit, 2012). Among those proteins, several cytokines, chemokines, and pro-inflammatory transcription factors have been linked to TL shortening in adults; however, the majority of this research has been conducted in non-pregnant women and has been restricted to a small number of targeted inflammatory markers, such as TNF-α and IL-10 (Lazarides et al., 2019; Wong et al., 2014).

Recent achievements in proteomic technologies have improved the detection of diverse inflammatory markers, but high-dimension analyses between these markers and newborn TL have not been conducted. In this context, a particular interesting advance is the recently developed proximity extension assay (PEA), commercialized by Olink (Assarsson et al., 2014). PEA is a high-throughput proteomics approach that allows interrogation of 92 markers associated with inflammation (Assarsson et al., 2014). Advantages of this approach include the high sensitivity and specificity of the assay, low sample volume requirements, and its suitability for hypothesis‐free biomarker discovery. Characterizing the prenatal inflammatory milieu through a broader range of cytokine and chemokines, rather than focusing on individual cytokine levels in relation to newborn TL, may more comprehensively elucidate underlying associations.

Here, we leveraged the ethnically diverse PRogramming of Intergenerational Stress Mechanisms (PRISM) pregnancy cohort to evaluate the relationship between maternal inflammation-related immune system status during pregnancy and newborn TL. We hypothesized that higher levels of maternal inflammatory markers would be associated with shorter newborn leukocyte TL, measured in cord blood. We first selected the most important inflammatory-related proteins jointly associated with TL via a tree-based machine learning algorithm. We then assessed the combined association between the selected proteins and TL using Bayesian Weighted Quantile Sum (BWQS) regression, a novel mixtures approach that captures the association of the overall inflammatory response with newborn TL and identifies the individual contribution of each protein to the overall response. We also provided results from several sensitivity analyses.

Methods

Study Population

At the time of immune profiling, 843 women were enrolled in the ongoing PRogramming of Intergenerational Stress Mechanisms (PRISM) study, an urban ethnically diverse pregnancy cohort designed to evaluate the role of early life environmental exposures on children’s development across the early life course. Beginning in 2011, pregnant women were recruited from Boston and New York City hospitals and affiliated prenatal clinics. Eligibility criteria included English or Spanish-speaking, 18 years or older at enrollment, and singleton pregnancy; exclusion criteria included HIV+ status or self-reported drinking ≥7 alcoholic drinks per week before pregnancy recognition or any alcohol after pregnancy recognition. Maternal serum samples were collected during the 2nd or 3rd trimester of pregnancy and cord blood was collected at birth. In New York City the study protocol was overseen by the Institutional Review Board (IRB) at the Icahn School of Medicine at Mount Sinai (ISMMS), while in Boston the study protocol was overseen by the IRB at the Brigham and Women’s Hospital (BWH). Both study protocols were approved by the human studies’ committees at the ISMMS and the BWH; the Beth Israel Deaconess Medical Center relied on BWH for review and oversight of the protocol. All participants provided written informed consent in their primary language.

Leukocyte Cord Blood Analyses

Newborn relative leukocyte telomere length (rLTL) was assessed in 155 biobanked cord blood samples. Briefly, cord blood was collected at delivery in EDTA tubes and the buffy coat fraction was isolated and stored at −80°C. DNA extraction was performed using a Promega Wizard DNA extraction system (Madison, WI, USA) (Bosquet Enlow et al., 2019). Quantitative real-time polymerase chain reaction (PCR) was used to measure the ratio of telomeric repeat copy number (T) to a nuclear single copy gene (S, human beta-globulin gene) copy number (T/S ratio) in each sample. To control for inter-assay variability, a pooled DNA sample was included in each run. For each plate, we divided the T/S ratio of each pooled DNA sample by the average T/S ratio for all pooled samples across all plates to obtain a normalizing factor. We then divided samples on a given plate by the plate-specific normalizing factor to adjust for potential batch effects. Additional details of the PCR assay have been previously published (see Bosquet Enlow et al., 2019).

Pregnancy immune phenotyping

We collected maternal blood by venipuncture during pregnancy (mean ± SD: 29.5 ± 5.0 weeks, range: 11–40 weeks). After collection, blood was separated into isolated serum aliquots, snap frozen in liquid nitrogen, and stored them at −80°C. We conducted proteomics analyses at the Human Immune Monitoring Center at ISMMS, a center certified by Olink Proteomics to perform PEA workflow. PEA requires minimal quantity (<1 μL) of biospecimen and enables high-multiplex capabilities with minimal antibody cross-reactivity (Olink Bioscience, Sweden) (Assarsson et al., 2014). Briefly, pairs of complementary oligonucleotide-labeled antibody probes bind to a shared target protein, bringing them into proximity and enabling extension by DNA polymerase. The resulting DNA amplicons are quantified by polymerase chain reaction (PCR) using a microfluidic real-time platform. We present final protein data as Normalized Protein eXpression (NPX) levels, which are log2-transformed values relative to extension and inter-plate control samples. Quality control procedures included the exclusion of eight samples that deviated by more than 0.3 units from the median value of four internal controls. One protein (BDNF) was not reported due to technical issues identified by the manufacturer; of the remaining 91 biomarkers, 26 were detected in fewer than 60% of participants and were excluded from analyses. For the remaining proteins, we replaced values below the limit of detection (LOD) with the LOD divided by the square root of 2. Gestational week of serum collection was recorded and included in all analyses.

Covariates

Maternal age, self-reported ethnicity, and education level were determined by questionnaires completed during pregnancy and fetal sex was reported by the mother after delivery.

Statistical methods

In the main analyses, we included mother-child pairs who had complete information on cord blood telomere length, maternal ethnicity, maternal age, maternal education, and week of serum collection, resulting in a final analytic sample of 129 mother-child pairs. We first examined the correlations between proteins using a heat map and we visually inspected their distribution using boxplots.

Tree-based machine learning for immuno-protein selection

We next leveraged the Boruta algorithm, which is a tree-based machine learning approach, to select the most important immune-related proteins associated with rLTL when all proteins are jointly considered in the model (Kursa and Rudnicki, 2010). Briefly, this approach ranks the independent variables (i.e. proteins) based on their importance in relation to the outcome (i.e. rLTL) using iterative sets of decision trees. This algorithm is an extension of the random forest method and runs a forest of (NF=500) decisional trees multiple times (NT=400), including shadows (permutations) of the independent variables, to reduce randomness of the prediction and correlation among variables. Similar to the random forest method, this approach considers an average absolute error to evaluate the accuracy of the model. The algorithm is also robust against outliers and skewed distributions. To determine model fit of the data we implemented a 10-fold cross-validation step. After generating the decisional trees, we used the mean probability across trees to determine the importance of each protein in relation to rLTL. We selected those proteins with a mean probability exceeding a 70% threshold as non-randomly related to rLTL (Kursa and Rudnicki, 2010). We selected a threshold above the (50%) randomness level due to the relatively limited sample size and because all proteins with a mean probability greater than 70% showed a nominally significant association with rLTL. We then regressed the selected protein on rLTL in separate linear models adjusting for maternal age (continuous in years), maternal race/ethnicity (Black vs. White vs. Hispanic vs. other), maternal education level (less than high school degree vs. high school or more), fetal sex (girl vs. boy) and week of serum collection (continuous in weeks).

Bayesian Weighted Quantile Sum regression for cumulative inference

Finally, we used Bayesian Weighted Quantile Sum (BWQS) regression to assess the overall association between the selected proteins, treated as a mixture, and rLTL. This novel mixture method is a supervised quantile-based approach that combines multiple independent variables additively into a weighted index, with weights capturing the contribution of each variable to the mixture (Colicino et al., 2019, Gennings et al., 2013). The BWQS regression provides simplicity and stability of inference, easy interpretability, and, due to quartile transformations of the independent variables, results that are insensitive to exposures outliers and skewed distribution (Colicino et al., 2019). Proteins selected in the Boruta algorithm were included in the mixture; we assessed their overall association with the newborn rLTL and identified the individual contribution of each component. We adjusted the analysis for the same set of covariates included in the linear regressions.

Sensitivity analysis

Protein-by-protein association.

We assessed the linear association between individual proteins and the rLTL with linear regression models. To reduce the right-skewed distribution of all proteins and to fulfill normality assumption requirements of this set of models, we centered and scaled all protein levels. All models were adjusted for the same set of covariates used in the main analyses. We corrected our results for multiple comparisons with the False Discovery Rate (FDR) and summarized individual associations with a Volcano plot. To offer a parallel comparison to this approach, we provided results for the tree-based method and BWQS regression after the transformation of protein levels.

Cell proportion adjustment.

Leukocyte samples are composed by different cell types and blood counts, which depend on the individual difference. To limit the effect that differences of cord blood cell proportions could have on rLTL detection, we included cell proportion estimates in the covariate adjustment of the main BWQS regression. We estimated cord blood cell proportions—granulocytes, monocytes, basophils, natural killer cells, CD4 and CD8 T cells, and nucleated red blood cells—leveraging the validated approach in the EWAStool R-package (Heiss, et al. 2019), epigenetic marks from cord blood samples (Brunst, et al 2018) and Balkuski reference data (Bakulski, et al 2016). The epigenetic marks were measured on cord blood from a subset of the initial set of PRISM participants (n=116) by using the Illumina Infinium HumanMethylation450 BeadChip array technology, as previously described (Brunst, et al 2018).

Inverse Probability Weighting (IPW) Correction.

We compared the main characteristics of PRISM participants based on their inclusion/exclusion in this analysis. To reduce any potential selection bias, we incorporated weights in the main analysis. We estimated weights by using the inverse probability weighting approach, which was based on a logistic regression predicting the probability of being included in the analysis from the main characteristics.

Additional covariate adjustment.

We also adjusted our analysis for additional covariates (gestational age at delivery, mother marital status, and birth weight). We compared results between main and this sensitivity analysis by using the Watanabe-Akaike Information Criteria (WAIC), which allows comparisons between Bayesian models on the same participants.

Continuous NPX levels.

Due to the normal distribution of the proteins, and their individual linear association with the rLTL, we performed the main analysis with continuous normalized protein levels.

Sex-stratified analysis.

Previous literature has shown a differential immune response based on sex, therefore we explored whether the mixture-rLTL association differed by fetal sex in a stratified analysis.

Functional enrichment annotation

We annotated the function of each identified protein that was significantly related to rLTL using UniProtKB, a centralized repository of curated functional information, including protein description, taxonomy, and Gene-Ontology pathways (Poux et al., 2017).

Results

Study Population

Women were racially/ethnically diverse (42% White, 34% African American, 12% Hispanic) with 15% reporting less than a high school education. Women were, on average, 31 years old (standard deviation (SD): 5.5) at time of delivery and had an average pregnancy length of 39 weeks (SD: 1.46) (Table 1). Cord blood rLTL was approximately normally distributed with a mean of 2.32 (SD: 0.8) (Table 1).

Table 1:

PRogramming of Intergenerational Stress Mechanisms (PRISM) participant characteristics (N = 129)

Variables N (%)
Maternal age (years), mean ±SD 30.98 ± 5.48
Serum collection during gestation (weeks), mean ± SD 28.76 ± 4.37
Race/Ethnicity
     White 54 (42%)
     Black 43 (34%)
     Hispanic 16 (12%)
     Other 16 (12%)
Maternal education
     <High school 19 (15%)
     Some college/4 year’s degree 62 (48%)
     Graduate degree 48 (37%)
Child sex (boy) 69 (53%)
Cord blood rLTL, mean ± SD 2.32 ± 0.75
Open in a new tab

SD= Standard Deviation; rLTL = relative leukocyte TL

Immuno-protein selection

We detected 65 proteins (71%) with levels above the limit of detection in at least 60% of samples. The correlation heat map showed low to moderate Pearson correlations between levels of all proteins, with most of the proteins being positively associated with each other (Figure S1). Most markers were log-normally distributed with long right tails and high inter-individual variability (Figure 1B). We report the probability of an association between each protein and rLTL in Figure 1A. This approach selected three protein markers that were jointly associated with the outcome: Beta-nerve growth factor (BNGF [Uniprot P01138], probability mean: 0.70, SD: 0.08), caspase-8 (CASP8 [Uniprot Q14790], probability mean: 0.76, SD: 0.20), and TNF-related activation-induced cytokine (TRANCE [Uniprot O14788], probability mean: 0.75, SD: 0.15) (Table S1). Pearson correlations between the NPX levels of BNGF and CASP8 and of BNGF and TRANCE were positive (r=0.15 p=0.08; r=0.21 p=0.02, respectively), while the correlation between CASP8 and TRANCE NPX levels was null (r=0.01 p=0.94). Levels of each protein had a significant, negative linear association with rLTL (BNGF: β = −0.43, p=0.033, CASP8: β= −0.22, p=0.008, and TRANCE: β= −0.38, p=0.004) (Figure 2).

Figure 1. Protein selection.

Figure 1.

Open in a new tab

A) Boruta algorithm results. Mean probability of protein levels to be associated with the cord blood relative leukocyte TL (rLTL). The most important proteins for the rLTL were selected based on a 70% threshold (yellow color) and would likely not be selected by chance. B) Distributions of Normalized Protein eXpression (NPX) levels of each protein.

Figure 2.

Figure 2.

Open in a new tab

Linear association between Normalized Protein eXpression (NPX) levels of the proteins, selected by the Boruta algorithm, (x-axis in each plot) and cord blood relative leukocyte TL (rLTL) (y-axis in each plot). Blue lines and grey shading indicate the estimated linear associations and 95% confidence intervals, respectively. P-values for the association are reported on the top right corner.

Bayesian Weighted Quantile Sum regression

Results of BWQS showed a significant overall decrease in rLTL (β=−0.26, 95% credible interval (CrI): −0.43, −0.07) per quartile increase of the BWQS index, representing the combined mixture of the three proteins (Figure 3A, Table S3). CASP8 contributed the greatest weight to the mixture (CASP8 50%; BNGF 27%, and TRANCE 23%) (Figure 3B, Table S3).

Figure 3. Bayesian Weighted Quantile Sum (BWQS) results.

Figure 3.

Open in a new tab

A) Association between the Bayesian Weighted Quantile Sum (BWQS) index, representing the mixture of the proteins on the x-axis, and cord blood cord blood relative leukocyte TL (rLTL) on the y-axis. Blue line and grey shading area represent the estimated association and the 95% credible intervals, respectively. Each dot represents an individual mother-child pair. B) Bar plot showing the contribution (relative weight) of each protein to the mixture. Weights sum up to one. The analysis was adjusted for maternal age (continuous in years), maternal race/ethnicity (White, Black, Hispanic, other), maternal education level (< or ≥ high school degree), fetal sex (boy, girl) and week of serum collection (continuous in weeks).

Sensitivity analysis

Protein-by-protein association.

No protein levels were individually and linearly associated with cord blood rLTL at 5% FDR (Figure S2, Table S2). Results on centered and scaled protein levels from the Boruta algorithm and BWQS regression were similar to the main analysis, due to the methods’ characteristic of being robust against outliers and skewed distributions (Figure S3, Table S3). The WAIC comparison between BWQS regressions, before and after centering and scaling protein levels, showed better fit to the data when protein levels were not transformed (Table S3).

Cell proportion adjustment.

Results from the BWQS regression with the additional adjustment for cord blood cell proportions were similar to the main analysis in terms of direction, magnitude, and significance, with an overall association between the mixture and rLTL of β= −0.23 (95%CrI: −0.42, −0.04), and a contribution of each marker to the mixture of CASP8=47%, TRANCE=27%, and BNGF=26% (Table S4).

Inverse Probability Weighting Correction.

Mothers included in the main analysis compared to those excluded were older (31±5 vs 29±6, p<0.001), mostly white (42% vs 18%, p<0.001) with at least high school education (48%+37% vs 47%+14%, p<0.001), had a living partner (81% vs 70%, p=0.003) and reported a longer gestational period (39.1 ±1.5 vs 38.7±2.2, p=0.003) (Table S5). We reduced any potential selection bias with an IPW approach and results were consistent with those of the main analysis, showing an overall mixture-rLTL association of β= −0. 31 (95%CrI: −0.42, −0.04) and individual contribution to the mixture of CASP8=52%, TRANCE=27%, BNGF=21% (Table S6). IPW had a median of 1.12 and an inter-quartile range of 0.36.

Additional covariate adjustment.

Results including additional covariates were similar to the main analysis (Table S7). The WAIC comparison between the main and sensitivity models identified that the analysis with the reduced set of covariates was more appropriate for fitting the data (Table S7).

Continuous NPX levels.

The overall results modeling continuous levels of proteins were similar to the main analysis; the relative contribution of each protein to the overall mixture was similar to the initial analysis (Table S8).

Sex-stratified analysis.

The direction of the mixture-rLTL association was similar in both boy and girl pregnancies, as was the contribution of each protein to the mixture. However, the association between the protein mixture and rLTL was stronger among girl pregnancies compared to boy pregnancies and was only significant among girls (Table S9).

Functional enrichment annotation

Table S10 presents the major biological function of the three proteins in humans. Both CASP8 and BNGF are involved in the regulation of apoptotic processes. BNGF also plays a role in cell population proliferation and in regulation of neuronal activity, including neuron differentiation and neuronal apoptotic processes. TRANCE, a protein from the TNFSF11 gene, is involved in cell differentiation, including osteoclast differentiation, and in the regulation of human immune responses.

Discussion

We examined the maternal inflammatory milieu during pregnancy in relation to cord blood leukocyte TL. We found levels of three inflammatory markers (BNGF, CASP8, TRANCE) were negatively, individually and jointly associated with shorter cord blood rLTL. Results from BWQS regression showed a significant overall decrease in cord blood rLTL per quartile increase of the protein mixture, with CASP8 contributing the greatest weight.

The association between enhanced inflammation and shortened TL is well substantiated in adults. For example, increased production of cytokines, chemokines and inflammation-responsive C-reactive protein have been consistently and negatively correlated with TL (Brown et al., 2018; Masi et al., 2012). Research has also linked a maternal pro-inflammatory state, operationalized as the ratio of tumor necrosis factor-alpha (a pro-inflammatory cytokine) to interleukin-10 (an anti-inflammatory cytokine) with shorter newborn rLTL (Lazarides et al., 2019). Less research has examined TL in relation to a broad spectrum of inflammation-related immune proteins measured during gestational period, as we did here. Our approach allowed us to identify associations with three proteins (CASP8, BNGF, TRANCE) that have typically not been included in high throughput immunoassays.

CASP8 is among the most upstream proteases involved in the cascade responsible for inducing programmed cell death. Critically short telomere length also serves to signal cellular apoptosis (Tumpel and Rudolph, 2012), providing a putative biological function shared by telomeres and CASP8. Additionally, while little research has examined the direct relationship between CASP8 and TL, upregulation of telomerase, an enzyme active during gestation that is capable of elongating telomeres in stem and other replicative cells, has been shown to reduce activation of CASP8 and other caspases (Bermudez et al., 2006), further supporting a potential underlying pathway linking these cellular components. BNGF plays key roles in neuronal proliferation and prenatal development of the sympathetic and sensory nervous systems (Carvalho et al., 2011; Einarsdottir et al., 2004). While research directly examining associations between TL and BNGF is limited, recent evidence supports that neurotrophic factors more generally may be involved in the regulation of telomerase (Niu and Yip, 2011), including during fetal development. For example, brain derived neurotrophic factor (BDNF) has been shown to promote the survival of embryonic neurons via upregulation of telomerase activity (Fu and Chen, 2002). TRANCE is involved in the differentiation of cells, including osteoclasts. It is also known to augment the ability of dendritic cells to stimulate and regulate naive T-cell proliferation and is thought to play a role in the regulation of T-cell-dependent immune responses (Wong et al., 1997).

In addition to evidence of direct associations between TL and immune system proteins that we and others have observed, several studies have linked pro-inflammatory chemical and nonchemical stressors with shorter TL at birth. For example, air pollution (Lee et al. 2020), metals (Herlin et al., 2019), and psychosocial stress (Bosquet Enlow et al., 2019) have been independently associated with shorter cord blood TL. In turn, environmental and psychosocial exposures are increasingly linked to altered immune system status, including with the proteins we found to be associated with shorter TL. For example, BNGF levels have been shown to increase with chronic and acute stress (Alleva et al., 1996; Smith, 1996) and toxic metal exposures have been shown to induce apoptosis via the CASP8 pathway (Eichler et al., 2006). Future targeted research investigating how these proteins relate to external environmental exposures to affect telomere length will be needed to fully understand the underlying biological mechanisms at play.

Shorter age-adjusted TL has been associated with several chronic, inflammation-related diseases, including obesity in childhood, hypertension, cognitive decline, Alzheimer’s and depression in later in life (Clemente et al., 2019, Colicino et al., 2017; Ridout et al., 2016; Rizvi et al., 2014). Several complications and conditions related to pregnancy, including recurrent miscarriage, pre-eclampsia, fetal growth restriction and low birthweight, have also been linked to shorter TL at birth (Davy et al., 2009; Hanna et al., 2009, Lee et al., 2017). Similar to many chronic conditions, these disorders are characterized by pro-inflammation, including within the intra-uterine environment (Cornelius, 2018; Kwak-Kim et al., 2009). While relationships between telomere length and indicators of immune system status and telomere dynamics within the context of these pregnancy-related disorders remain to be determined, it is plausible that disease-related changes to the maternal immune system plays a role in the prenatal programming of newborn TL, with major health implications at birth and later in life. Indeed, accumulating research suggests that variability in rLTL at birth and during early life drives inter-individual differences across the life course (Asghar 2015, Bateson 2015).

Maternal adaptations to pregnancy have been shown to vary by fetal sex (Al-Qaraghouli 2017), as has the prevalence of several maternal pregnancy-related disorders and perinatal outcomes (Di Renzo 2007, Stark 2006). Results of our exploratory sex-stratified analysis, although limited by a small sample size, were consistent with previous literature showing sex-differential fetal and placental influences on the maternal immune milieu (Sood 2006, Ghidini 2005). While the mechanistic underpinnings of these differences remain incompletely understood, they could relate to differences in endocrine and immune active placental signaling (Ghidini 2005) or maternal immune system responsivity to paternally-derived fetal antigens (Petroff 2011).

We note several strengths of our study. We assessed a panel of inflammatory markers using recently developed high-throughput technology, which is not susceptible to specificity issues typical of multiplexed immunoassays. We were able to adjust for important covariates including maternal age, race/ethnicity, and socioeconomic status. Although the racial/ethnic diversity of the sample is a strength of this study, it may reduce the generalizability of our findings to more racially homogenous populations. Future studies with larger sample size and with more homogenous populations may shed light on the role of inflammation in fetal development and its consequences for early life. Our sample having both the inflammatory panel and newborn TL included a relatively small number of mother-child pairs (N=129), and this may have contributed to our model’s selection of only three inflammation-related markers; however, the Boruta algorithm showed robust results across seeds and random forests.

Conclusions

We identified three inflammatory markers, measured during pregnancy, that were individually and jointly associated with shorter leukocyte TL at birth. Our results provide further evidence that increased inflammation during pregnancy may be a biological pathway linking prenatal environmental factors with reduced TL at birth.

Supplementary Material

1
2

Highlights:

  1. Maternal prenatal inflammatory protein biomarkers are associated with shortened cord blood leukocyte telomere length in newborns

  2. Three inflammatory markers (BNGF, CASP8, TRANCE) were individually and jointly associated with shorter cord blood rLTL.

  3. The joint association was evaluated via the Bayesian Weighted Quantile Sum Regression.

  4. The identified inflammatory markers were involved in the regulation of apoptotic processes and cell proliferation.

  5. Increased inflammation during pregnancy may be a biological pathway linking prenatal environmental factors with reduced TL at birth.

Acknowledgments

Funding: This work was supported by the National Institutes of Health [grant numbers: R01 HL095606, R01 HL114396, P30 ES023515, and UG3 OD023337]. During the preparation of this manuscript, WJC was supported by T32 HD049311.

Footnotes

Declaration of interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Alleva E, et al. , 1996. NGF regulatory role in stress and coping of rodents and humans. Pharmacol Biochem Behav. 54, 65–72. [DOI] [PubMed] [Google Scholar]
  2. Al-Qaraghouli M, and Fang YMV, 2017. Effect of Fetal Sex on Maternal and Obstetric Outcomes. Front Pediatr.; 5:144. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Asghar M, et al. Chronic infection. Hidden costs of infection: chronic malaria accelerates telomere degradation and senescence in wild birds. Science. 2015;347(6220):436–438. [DOI] [PubMed] [Google Scholar]
  4. Assarsson E, et al. , 2014. Homogenous 96-plex PEA immunoassay exhibiting high sensitivity, specificity, and excellent scalability. PLoS One. 9, e95192. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Aviv A, 2008. The epidemiology of human telomeres: faults and promises. J Gerontol A Biol Sci Med Sci. 63, 979–83. [DOI] [PubMed] [Google Scholar]
  6. Bakulski KM, Feinberg JI, Andrews SV, Yang J, Brown S, L McKenney S, et al. DNA methylation of cord blood cell types: applications for mixed cell birth studies. Epigenetics. 2016;11:354–362. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bateson M. Cumulative stress in research animals: Telomere attrition as a biomarker in a welfare context? Bioessays. 2016;38 (2):201–212. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bermudez Y, et al. , 2006. Telomerase confers resistance to caspase-mediated apoptosis. Clin Interv Aging. 1, 155–167. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Blackburn EH, 2001. Switching and signaling at the telomere. Cell. 106, 661–73. [DOI] [PubMed] [Google Scholar]
  10. Blasco MA, 2005. Telomeres and human disease: ageing, cancer and beyond. Nat Rev Genet. 6, 611–22. [DOI] [PubMed] [Google Scholar]
  11. Bosquet Enlow M, et al. , 2018. Sex differences in effects of maternal risk and protective factors in childhood and pregnancy on newborn telomere length. Psychoneuroendocrinology. 95, 74–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Bosquet Enlow M, et al. , 2019. Maternal cortisol output in pregnancy and newborn telomere length: Evidence for sex-specific effects. Psychoneuroendocrinology. 102, 225–235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Brown LL, et al. , 2018. Does Telomere Length Indicate Biological, Physical, and Cognitive Health Among Older Adults? Evidence from the Health and Retirement Study. J Gerontol A Biol Sci Med Sci. 73, 1626–1632. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Brunst KJ, Tignor N, Just A, et al. Cumulative lifetime maternal stress and epigenome-wide placental DNA methylation in the PRISM cohort. Epigenetics. 2018;13(6):665–681. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Carvalho OP, et al. , 2011. A novel NGF mutation clarifies the molecular mechanism and extends the phenotypic spectrum of the HSAN5 neuropathy. J Med Genet. 48, 131–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Clemente DBP, Maitre L, Bustamante M. et al. Obesity is associated with shorter telomeres in 8 year-old children. Sci Rep 9, 18739 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Codd V, et al. , 2010. Common variants near TERC are associated with mean telomere length. Nat Genet. 42, 197–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Colicino E, Wilson A, Frisardi MC, et al. 2017. Telomere Length, Long-Term Black Carbon Exposure, and Cognitive Function in a Cohort of Older Men: The VA Normative Aging Study Environ Health Perspect. 125(1):76–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Colicino E, et al. , 2020. Per- and poly-fluoroalkyl substances and bone mineral density: results from the Bayesian weighted quantile sum regression. Envirnmental Epidemiology 4(3):e092 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Cornelius DC, 2018. Preeclampsia: From Inflammation to Immunoregulation. Clin Med Insights Blood Disord. 11, 1179545X17752325. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Davy P, et al. , 2009. Fetal growth restriction is associated with accelerated telomere shortening and increased expression of cell senescence markers in the placenta. Placenta. 30, 539–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Di Renzo GC, et al. 2007. Does fetal sex affect pregnancy outcome?. Gend Med.; 4(1):19–30. [DOI] [PubMed] [Google Scholar]
  23. Eichler T, et al. , 2006. Single and combination toxic metal exposures induce apoptosis in cultured murine podocytes exclusively via the extrinsic caspase 8 pathway. Toxicol Sci. 90, 392–9. [DOI] [PubMed] [Google Scholar]
  24. Einarsdottir E, et al. , 2004. A mutation in the nerve growth factor beta gene (NGFB) causes loss of pain perception. Hum Mol Genet. 13, 799–805. [DOI] [PubMed] [Google Scholar]
  25. Epel ES, et al. , 2008. The rate of leukocyte telomere shortening predicts mortality from cardiovascular disease in elderly men. Aging. 1, 81–88. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Fu C, Chen Z, 2002. Telomerase activity in myelodysplastic syndrome. Chin Med J (Engl). 115, 1475–8. [PubMed] [Google Scholar]
  27. Gennings C, et al. , 2013. A Cohort study evaluation of maternal PCB exposure related to time to pregnancy in daughters. Environ Health. 12, 66. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Ghidini A. and Salafia CM 2005. Gender differences of placental dysfunction in severe prematurity. BJOG: An International Journal of Obstetrics & Gynaecology, 112: 140–144. [DOI] [PubMed] [Google Scholar]
  29. Hanna CW, et al. , 2009. Telomere length and reproductive aging. Hum Reprod. 24, 1206–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Heiss JA, Just AC. Improved filtering of DNA methylation microarray data by detection p values and its impact on downstream analyses. Clin Epigenetics. 2019;11(1):15. Published 2019 Jan 24. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Herlin M, et al. , 2019. Exploring telomere length in mother-newborn pairs in relation to exposure to multiple toxic metals and potential modifying effects by nutritional factors. BMC Med. 17, 77. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Kimura M, et al. , 2008. Telomere length and mortality: a study of leukocytes in elderly Danish twins. Am J Epidemiol. 167, 799–806. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Kursa MB, Rudnicki WR, 2010. Feature Selection with the Boruta Package. Journal of Statistical Software. 36, 1–13. [Google Scholar]
  34. Kwak-Kim J, et al. , 2009. Recurrent pregnancy loss: A disease of inflammation and coagulation. J Obstet Gynaecol Res. 35, 609–622. [DOI] [PubMed] [Google Scholar]
  35. Lazarides C, et al. , 2019. Maternal pro-inflammatory state during pregnancy and newborn leukocyte telomere length: A prospective investigation. Brain Behav Immun. 80, 419426. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Lee S, Hande P, Yeo GS et al. Correlation of cord blood telomere length with birth weight. BMC Res Notes 10, 469 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Levy D, et al. , 2010. Genome-wide association identifies OBFC1 as a locus involved in human leukocyte telomere biology. Proc Natl Acad Sci U S A. 107, 9293–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Marchetto NM, et al. , 2016. Prenatal stress and newborn telomere length. Am J Obstet Gynecol. 215, 94.e1–8. [DOI] [PubMed] [Google Scholar]
  39. Masi S, et al. , 2012. Inflammation and not cardiovascular risk factors is associated with short leukocyte telomere length in 13- to 16-year-old adolescents. Arterioscler Thromb Vasc Biol. 32, 2029–2034 [DOI] [PubMed] [Google Scholar]
  40. Newton K, Dixit VM, 2012. Signaling in innate immunity and inflammation. Cold Spring Harb Perspect Biol. 4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Niu C, Yip HK, 2011. Neuroprotective signaling mechanisms of telomerase are regulated by brain-derived neurotrophic factor in rat spinal cord motor neurons. J Neuropathol Exp Neurol. 70, 634–52. [DOI] [PubMed] [Google Scholar]
  42. Poux S, et al. , 2017. On expert curation and scalability: UniProtKB/Swiss-Prot as a case study. Bioinformatics. 33, 3454–3460. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Petroff MG. 2011. Review: Fetal antigens--identity, origins, and influences on the maternal immune system. Placenta.; 32 Suppl 2(Suppl 2):S176–S181. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Prescott J, et al. , 2011. Genome-wide association study of relative telomere length. PLoS One. 6, e19635. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Ridout KK, et al. , 2016. Depression and telomere length: A meta-analysis. J Affect Disord. 191, 237–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Rizvi S, et al. , 2014. Telomere length variations in aging and age-related diseases. Curr Aging Sci. 7, 161–7. [DOI] [PubMed] [Google Scholar]
  47. RM C, et al. , 2003. Association between telomere length in blood and mortality in people aged 60 years or older. Lancet. 361, 393–395. [DOI] [PubMed] [Google Scholar]
  48. Rosa MJ, et al. , 2019. Association between prenatal particulate air pollution exposure and telomere length in cord blood: Effect modification by fetal sex. Environ Res. 172, 495–501. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Smith MA, 1996. Hippocampal vulnerability to stress and aging: possible role of neurotrophic factors. Behav Brain Res. 78, 25–36. [DOI] [PubMed] [Google Scholar]
  50. Sood R, et al. ,2006. Gene expression patterns in human placenta. PNAS, 103 (14) 5478–5483; [DOI] [PMC free article] [PubMed] [Google Scholar]
  51. Stark MJ, et al. , 2006. Alterations in the maternal peripheral microvascular response in pregnancies complicated by preeclampsia and the impact of fetal sex. J Soc Gynecol Investig.;13(8):573–578. [DOI] [PubMed] [Google Scholar]
  52. Tumpel S, Rudolph KL, 2012. The role of telomere shortening in somatic stem cells and tissue aging: lessons from telomerase model systems. Ann N Y Acad Sci. 1266, 28–39. [DOI] [PubMed] [Google Scholar]
  53. Wai KM, et al. , 2018. Impact of prenatal heavy metal exposure on newborn leucocyte telomere length: A birth-cohort study. Environ Pollut. 243, 1414–1421. [DOI] [PubMed] [Google Scholar]
  54. Wong BR, et al. , 1997. TRANCE is a novel ligand of the tumor necrosis factor receptor family that activates c-Jun N-terminal kinase in T cells. J Biol Chem. 272, 25190–4. [DOI] [PubMed] [Google Scholar]
  55. Wong JYY, et al. , 2014. The Relationship between Inflammatory Biomarkers and Telomere Length in an Occupational Prospective Cohort Study. PLoS ONE. 9, e87348. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Zhang L, et al. , 2019. Prenatal cadmium exposure is associated with shorter leukocyte telomere length in Chinese newborns. BMC Med. 17, 27. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1626368-supplement-1.docx (1.1MB, docx)
2
NIHMS1626368-supplement-2.xlsx (13KB, xlsx)

RESOURCES