Individual and joint association of maternal and cord blood mitochondrial DNA heteroplasmy and copy number with preterm birth

Abstract

Background:

The biological mechanisms underlying preterm birth (PTB) are not fully understood. This study examined the individual and joint associations of mitochondrial DNA (mtDNA) heteroplasmy and copy number (mtDNA-CN) in maternal and cord blood with PTB.

Methods:

This study analyzed 998 mother-infant dyads (18% PTB). MtDNA heteroplasmy and mtDNA-CN were measured by a targeted DNA sequencing approach using umbilical cord blood collected at birth and maternal blood collected 24–72 hours postpartum. Multivariable logistic regressions were used to evaluate the associations with PTB.

Results:

Maternal predicted pathogenic heteroplasmy was associated with a twofold increased risk of PTB (adjusted OR (95% CI) = 2.08 (1.07, 4.05)), whereas no significant associations were observed for maternal overall or functional heteroplasmy. No significant associations were observed for cord blood heteroplasmy measures. A U-shaped association between maternal mtDNA-CN and PTB was observed. Higher cord mtDNA-CN were associated with increased PTB odds (1.36 (1.14, 1.62)). Notably, a significant interaction between maternal and cord mtDNA-CN was found for PTB risk.

Conclusion:

Maternal, but not fetal, mtDNA heteroplasmy was associated with increased PTB risk. Both maternal and cord blood mtDNA-CN demonstrated individual and interactive associations with PTB, suggesting that maternal and fetal mitochondrial genomic variations may jointly influence PTB risk.

Introduction

Preterm birth (PTB), defined as shorter than 37 weeks of gestation, is one of the leading causes of mortality in children younger than 5 years.¹ Moreover, survivors of PTB are at higher risk of neurodevelopmental, cardiometabolic, and respiratory disorders in later life compared to their term peers.^{2, 3} In 2020, there were 13.4 million newborns with preterm births worldwide.⁴ In the United States, 10% of all births were PTB.⁵ Notably, racially and ethnically minoritized women have a higher risk of PTB compared to White women.^{5, 6} Although numerous risk factors for PTB have been identified, the underlying biological mechanisms remain incompletely understood, particularly for spontaneous PTB.^{7, 8} Mitochondria have gained increasing attention given their central role in regulating energy production, oxidative stress, and immune signaling, biological pathways increasingly implicated in the pathophysiology of PTB.^9–11 Accordingly, attention has recently turned to mitochondrial DNA (mtDNA), although prior genetic studies have mostly focused on nuclear DNA.^{12, 13}

MtDNA encodes protein subunits essential for OXPHOS in the energy production and RNAs required for mitochondrial protein synthesis, thus playing a vital role in maintaining mitochondrial biogenesis and function. MtDNA copy number (mtDNA-CN) and heteroplasmy are commonly used measures of mtDNA abundance and sequence variation that can be quantified in tissues. Blood samples are frequently used in epidemiologic studies due to their feasibility. MtDNA-CN represents the number of mitochondrial genomes per nucleated cell and has been widely used as a proxy for mitochondrial function, though its biological significance remains debated.^{14, 15} Both elevated and reduced blood mtDNA levels have been linked to various health outcomes, especially in the aged population.^{9, 15} Heteroplasmy refers to the presence of a mixture of wild-type and mutant mtDNA within an individual.^{16, 17} Heteroplasmy burden has been associated with mitochondrial disease and dysfunction, and increasingly with a broader range of human diseases.^18–20 Recent large population studies have shown that blood mtDNA-CN declines with age, whereas mtDNA heteroplasmic variants tend to accumulate over the human lifespan.¹⁷ In addition, mtDNA measures vary across tissues and may change in response to physiologic stress, underscoring the dynamic nature of these biomarkers.^{16, 21}

Several recent studies also suggest the potential relevance of mtDNA variations in the etiology of PTB. Women with primary mitochondrial disorders caused by mtDNA mutations have been found to have higher frequencies of PTB.^{22, 23} PTB infants have also been reported to carry known pathogenic mtDNA variants.²⁴ These observations motivate further study of mtDNA heteroplasmy and mtDNA-CN in relation to PTB. To date, no studies have examined whether mtDNA heteroplasmy in mothers or newborns is associated with PTB, although one study reported that higher placental mtDNA heteroplasmy was associated with shorter gestational age, with a stronger magnitude of the association observed among females.²⁵ With respect to mtDNA-CN, the existing evidence on how maternal mtDNA-CN associates with PTB is still inconsistent.^{26, 27} A few studies reported that a higher level of cord blood mtDNA-CN was correlated with an increased risk of PTB.^{26, 28}

PTB is increasingly recognized as a complex condition involving dynamic interactions across maternal, placental, and fetal compartments,²⁹ suggesting that biomarkers measured in a single compartment may provide only a partial view of the biological processes relevant to PTB. In line with this, some studies have reported that the same biomarker can show different associations with PTB in maternal blood and cord blood, highlighting the importance of assessing both compartments.^30–32 However, few studies have simultaneously examined mtDNA variation measures in both mothers and newborns. Although mtDNA is maternally inherited, the mtDNA measures in mothers and newborns are not expected to be identical. Heteroplasmy patterns may shift during transmission because of the mitochondrial bottleneck and purifying selection, and their relative levels may also be influenced by nuclear genetic variation.¹⁷ As for mtDNA copy number, our previous work found that mtDNA-CN in maternal peripheral blood and umbilical cord blood was only modestly correlated (r=0.26),¹⁸ suggesting that this measure may vary across maternal and newborn biological contexts.

To address these gaps, we evaluated whether mtDNA heteroplasmy and copy number measured in maternal peripheral blood and umbilical cord blood were associated with PTB. We first examined associations for maternal and cord blood mtDNA measures separately. We then evaluated maternal and cord blood mtDNA measures together to assess their joint relationship with PTB. Associations with spontaneous PTB were additionally examined as a secondary outcome.

Methods

Study Population

The study sample was a subset of the Boston Birth Cohort (BBC), a cohort initiated in 1998 to investigate the causes and long-term health consequences of PTB and its related conditions, including low birth weight and small for gestational age. The study design and procedures of BBC were described in detail elsewhere.³³ The cohort recruited mothers postpartum at the Boston Medical Center (BMC) through rolling enrollment. Women who delivered a singleton live newborn were eligible for enrollment. Cases were defined as mothers who delivered a preterm (<37 weeks of gestation) and/or low birth weight (<2500 grams) infants. Controls were enrolled at approximately 2:1 control/case ratio and were defined as mothers who delivered a term infant without low birth weight. Pregnancies involving vitro fertilization, multiple gestations, chromosomal abnormalities, or major birth defects were excluded. Standardized questionnaire interviews were conducted with mothers 24–72 hours postpartum after obtaining written informed consent to collect demographic and lifestyle data. Clinical information on mother-infant pairs was extracted from electronic medical records. Maternal peripheral blood and umbilical cord blood samples were collected for biospecimen analysis. Among 8631 mother-infant pairs enrolled in BBC, mtDNA sequencing was performed in maternal and cord blood samples from 1067 pairs, and 998 pairs were included in the final analysis after quality controls (Figure S1). We compared the baseline characteristics of mother-infant pairs included in the current study to those excluded (Table S1). All study protocols of the BBC were approved by the Institutional Review Boards of BMC and Johns Hopkins University.

MtDNA Measurement

Sequencing and Variant Calling.

Genomic DNA was extracted from cord umbilical blood samples collected at delivery and maternal peripheral blood samples collected within 24–72 hours after delivery. A targeted sequencing method, STAMP (sequencing by targeted amplification of multiplex probes), was used to simultaneously measure mtDNA heteroplasmy and copy number with 46 pairs of oligonucleotide probes targeting mtDNA and five pairs of probes targeting nuclear DNA.³⁴ STAMP sequencing libraries were sequenced on the Illumina HiSeq 2500 with 2X250-bp reads. Paired-end reads were aligned to the human genome (GRCh38 assembly), and a modified mtDNA sequence based on the Revised Cambridge Reference Sequence (rCRS) using the “stamp toolkit”. Consensus reads were generated by merging base information from paired-end reads sharing the same barcode using a Bayesian approach. Across 998 mother-newborn pairs included in our analysis, the average per-sample sample median depth was 3786x, and the average per-sample mean depth was 4867x, indicating high sequencing depth. Variant calling was performed on consensus reads with standard quality filters applied, including base alignment quality (BAQ) ≥ 30 and mapping quality (MAPQ) ≥ 20, while nuclear mitochondrial segments were excluded and variants with excessive mismatches were removed. Variant functional annotation was subsequently conducted using the ANNOVAR pipeline.

MtDNA Heteroplasmy.

The heteroplasmic variants were defined as those with a variant allele fraction (VAF) ≥ 1.5%, calculated as the number of reads of the mutated allele divided by the total number of reads. The threshold was applied to minimize the influence of spurious variants arising from nuclear mitochondrial segments, and detailed quality control filters were described before.¹⁸ In addition to total heteroplasmy counts, we further defined functional and predicted pathogenic heteroplasmy, due to their functional relevance in mitochondrial energy production and protein synthesis. The presence of functional heteroplasmy was defined as having any heteroplasmic variant at nonsynonymous sites of OXPHOS genes and all sites of tRNA and rRNA genes. The presence of predicted pathogenic heteroplasmy refers to having any predicted pathogenic heteroplasmic variants specifically within the above functional genes, identified based on known disease association, bioinformatic pathogenicity predictions, and general population variant frequencies (detailed criteria described elsewhere¹⁸).

MtDNA-CN and Haplogroup.

MtDNA-CN was calculated from STAMP data using a residual method. The crude value of mtDNA-CN was calculated as the residual from the following linear model:

log2(number of mtDNA consensus reads) ~ log2(number of nuclear DNA consensus reads) + sample plate. The sample plate was added to control for batch effect. Separate mtDNA-CN z scores were generated for mothers and infants by standardizing values (subtracting the mean and dividing by the standard deviation) to facilitate comparison and interpretation. HaploGrep2 (v2.1.1) was used to derive the mtDNA haplogroup information based on each sample’s main mtDNA sequence.³⁵

Outcome and Covariates

The primary outcome was PTB, defined previously as delivery before 37 weeks of gestation. Infants born at term (≥37 weeks) served as the control group for analysis, regardless of birth weight. A secondary outcome, spontaneous PTB (sPTB), was further defined as a birth that occurred secondary to documented preterm labor (cervical effacement and dilation at <37 weeks), or premature rupture of membranes at <37 weeks without uterine contractions, or both. Maternal and newborn characteristics were collected via standardized postpartum interviews, including maternal age at delivery, self-reported race/ethnicity, maternal education, maternal smoking before or during pregnancy, parity, and from medical record abstraction, such as child sex, delivery type, and maternal diabetes (gestational or pre-gestational). Pre-pregnancy BMI was calculated using self-reported height and weight, with obesity defined as BMI ≥ 30 kg/m². The variable maternal metabolic conditions was categorized as neither condition, obesity only, diabetes only (including both pre-gestational and gestational diabetes), or both obesity and diabetes. Single imputation was used for missing values for variables with a low missing rate (<5%). Maternal education, smoking history, and delivery type were imputed with their most frequent categories (‘high school or less’, ‘never’, and ‘vaginal delivery’, respectively). Missing BMI values were imputed using the mean BMI within each race/ethnicity group before defining the maternal metabolic conditions variable.

Statistical Analysis

All statistical analyses were conducted using STATA 18 (StataCorp LLC, College Station, TX). Basic characteristics of mother-infant pairs were compared between PTB and term groups, using t-tests for continuous variable (maternal age at delivery) and chi-square tests for categorical variables. Logistic regression models were used to estimate associations between maternal and cord blood mtDNA measures and PTB. Covariates were selected a priori based on existing literature and included ancestry group (defined by mtDNA haplogroup), maternal age at delivery, maternal education, maternal smoking before or during pregnancy, maternal metabolic conditions, parity, delivery type, and child sex.^{36, 37}

Associations of maternal and cord blood heteroplasmy measures with PTB were evaluated for overall heteroplasmic sites, functional heteroplasmy, and predicted pathogenic heteroplasmy. In addition to the covariates listed above, models for maternal heteroplasmy further adjusted for maternal blood mtDNA-CN z scores, and models for cord blood heteroplasmy further adjusted for cord blood mtDNA-CN z scores. To evaluate the joint effects of maternal and cord heteroplasmy, separate four-level categorical variables were constructed for overall heteroplasmic sites, functional heteroplasmy, and predicted pathogenic heteroplasmy, with categories defined as neither maternal nor cord heteroplasmy, maternal heteroplasmy only, cord heteroplasmy only, and both maternal and cord heteroplasmy. The group with neither maternal nor cord heteroplasmy served as the reference.

Associations of maternal and cord blood mtDNA-CN z scores with PTB were first explored using restricted cubic spline (RCS) models with three knots placed at the 10th, 50th, and 90th percentiles to assess potential non-linearity. Statistical deviation from linearity was evaluated using the P value for the second spline term. Based on patterns observed in the RCS plots, mtDNA-CN was then modeled either as a linear term or as a linear spline with a knot at z score = 0. The linear spline specification allowed different slopes above and below z score = 0 and provided more interpretable estimates of the associations. To evaluate the joint effects of maternal and cord mtDNA-CN, both variables were dichotomized at z score = 0. Interaction terms between continuous maternal mtDNA-CN and binary cord mtDNA-CN, and between continuous cord mtDNA-CN and binary maternal mtDNA-CN, were added to the models. For models using linear spline terms, interaction terms were constructed for each spline component. Statistical interaction was assessed using likelihood ratio tests.

Parallel analyses were performed for spontaneous PTB. Effect modification by child sex and ancestry group defined by mtDNA haplogroup was assessed by including interaction terms in the corresponding models, with statistical significance evaluated using likelihood ratio tests. To assess robustness, we conducted two sensitivity analyses. First, given that blood cell composition is an important source of variation in blood mtDNA measurements,³⁸ primary analyses were repeated among participants with available blood cell composition data (n = 388 for maternal blood cell composition; n = 646 for cord blood cell composition), with additional adjustment for maternal or cord blood cell type composition. Second, primary analyses were repeated among mother–infant pairs with complete covariate information to assess potential bias arising from missing data imputation.

Results

Among 998 mother-infant dyads, 182 (18.2%) infants were born preterm. Table 1 compares maternal and child characteristics between preterm and term births. Mothers who delivered preterm were slightly older on average, more likely to have a smoking history, have metabolic conditions, and experience C-section delivery. In the total sample, 447 mothers and 491 newborns have at least one mtDNA heteroplasmy. Specifically, functional heteroplasmy was observed in 244 mothers and 264 newborns, and predicted pathogenic heteroplasmy was observed in 46 mothers and 71 newborns.

Table 1.

Maternal and child characteristics by preterm birth

Characteristic	Preterm Birth	Term Birth	P
N	182	816
Maternal age at delivery, mean (SD)	29.2 (6.1)	28.1 (6.6)	0.034
Maternal self-reported race/ethnicity			0.29
Non-Hispanic Black	116 (63.7)	474 (58.1)
Non-Hispanic White	9 (5.0)	41 (5.0)
Hispanic	42 (23.1)	194 (23.8)
Othersa	15 (8.2)	107 (13.1)
Maternal education			0.72
High school or less	121 (66.9)	531 (65.5)
Some college or above	60 (33.1)	280 (34.5)
Missing	1	5
Parity			0.96
Nulliparous	76 (41.8)	339 (41.5)
Multiparous	106 (58.2)	477 (58.5)
Maternal smoking before or during pregnancy			0.047
Never	137 (76.5)	668 (82.9)
Have ever	42 (23.5)	138 (17.1)
Missing	3	10
Maternal metabolic conditions			<0.001
Neither	106 (60.9)	539 (70.0)
Obesity only	26 (14.9)	144 (18.7)
Diabetes only	24 (13.8)	52 (6.8)
Both	18 (10.4)	35 (4.5)
Missing	8	46
Delivery type			0.021
C-section	72 (39.8)	251 (30.9)
Vaginal	109 (60.2)	562 (69.1)
Missing	1	3
Child sex			0.024
Female	68 (37.4)	380 (46.6)
Male	114 (62.6)	436 (53.4)
Ancestry group defined by mtDNA haplogroup			0.94
African	128 (70.3)	564 (69.1)
Europe	33 (18.1)	157 (19.2)
Asian	21 (11.6)	95 (11.7)

Continuous variables are presented as mean (SD) and categorical variables as No. (%). Percentages were calculated using non-missing data for each variable. P values were calculated using the t test for continuous variables and the χ² test for categorical variables, as appropriate.

^aThe “others” group includes those who self-identified as Asian, Cape Verdean, Pacific Islander, more than one race, and other races.

MtDNA Heteroplasmy and PTB

Mothers with two or more overall heteroplasmic sites and mothers with functional heteroplasmy had a slightly higher likelihood of PTB, yet the associations were not statistically significant (Table 2). Maternal predicted pathogenic heteroplasmy was associated with increased odds of PTB (adjusted OR = 2.08, 95% CI: 1.07, 4.05, P = .031). There was no association between the cord blood heteroplasmy measures and PTB. In exploratory joint analyses, compared with the group with neither maternal nor cord predicted pathogenic heteroplasmy, the group with maternal predicted pathogenic heteroplasmy only was associated with increased odds of PTB (OR = 2.99, 95% CI: 1.36–6.57, P = .007, Table S2), consistent with the primary finding for maternal heteroplasmy.

Table 2.

Association of maternal and cord blood mtDNA heteroplasmy with preterm birth

Maternal Blood MtDNA Heteroplasmya	No. of Preterm n=182	No. of Term n=816	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Overall Heteroplasmy Count
0	93	458	1[Ref]	1[Ref]
1	59	255	1.14 (0.79, 1.63)	1.15 (0.80, 1.66)
2 or more	30	103	1.43 (0.90, 2.28)	1.38 (0.86, 2.22)
Presence of Functional Heteroplasmy
No (0)	130	624	1[Ref]	1[Ref]
Yes (≥1)	52	192	1.30 (0.91, 1.86)	1.29 (0.89, 1.86)
Presence of Predicted Pathogenic Heteroplasmy
No (0)	168	784	1[Ref]	1[Ref]
Yes (≥1)	14	32	2.04 (1.07, 3.91)*	2.08 (1.07, 4.05)*
Cord Blood MtDNA Heteroplasmyb	No. of Preterm n=182	No. of Term n=816	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Overall Heteroplasmy Count
0	90	417	1[Ref]	1[Ref]
1	62	270	1.06 (0.74, 1.52)	1.04 (0.72, 1.51)
2 or more	30	129	1.08 (0.68, 1.70)	0.98 (0.61, 1.57)
Presence of Functional Heteroplasmy
No (0)	131	603	1[Ref]	1[Ref]
Yes (≥1)	51	213	1.10 (0.77, 1.58)	1.08 (0.74, 1.56)
Presence of Predicted Pathogenic Heteroplasmy
No (0)	170	757	1[Ref]	1[Ref]
Yes (≥1)	12	59	0.91 (0.48, 1.72)	0.82 (0.43, 1.60)

ORs and 95% CIs for preterm birth in relation to mtDNA heteroplasmy measures were estimated using logistic regression models. Counts indicate the numbers of preterm and term births in each exposure group.

^aMaternal blood models were adjusted for ancestry groups defined by mtDNA haplogroups, maternal age at delivery, maternal education, maternal metabolic conditions, maternal smoking, parity, delivery type, child sex, and maternal blood mtDNA-CN z score.

^bCord blood models were adjusted for ancestry groups defined by mtDNA haplogroups, maternal age at delivery, maternal education, maternal metabolic conditions, maternal smoking, parity, delivery type, child sex, and cord blood mtDNA-CN z score.

^*P < .05.

Abbreviations: CI, confidence interval; mtDNA, mitochondrial DNA; mtDNA-CN, mitochondrial DNA copy number; OR, odds ratio.

Results for SPTB were consistent with the primary analyses (Table S3). No statistically significant effect modification by child sex or ancestry group was observed, although some variation in effect estimates across strata was noted (Tables S4 & S5). In analyses restricted to participants with available blood cell composition data, additional adjustment for cell composition did not substantially alter the heteroplasmy associations (Table S6). In analyses restricted to participants with complete covariate information, estimates remained directionally consistent (Table S7).

MtDNA-CN and PTB

RCS modeling indicated a U-shaped relationship between maternal mtDNA-CN z scores and the likelihood of PTB (P for non-linearity < .001), with a changing point near z score = 0 (Figure 1-A). Based on the adjusted spline logistic regression model with a spline at z score = 0 (Table 3), one SD increase in z scores below z score = 0 was associated with a 35% reduction in the odds of PTB (adjusted OR = 0.65, 95% CI: 0.50, 0.84, P = .001); however, one SD increase in z scores above z score =0 corresponded to a 1.73-fold increase in the odds of PTB (adjusted OR = 1.69, 95% CI: 1.25, 2.30, P =.001). When modeled as a linear term, maternal mtDNA-CN z scores were not statistically associated with PTB. There was a nearly linear association between cord mtDNA-CN z scores and PTB revealed by RCS analysis (P for non-linearity = .148) (Figure 1-B). When modeled as a linear term in the logistic regression, per one SD increase in cord blood mtDNA-CN z scores was associated with 36% increased odds of PTB (adjusted OR = 1.36, 95% CI: 1.14, 1.62, P = .001). Adjusted logistic regression with a linear spline (knot at z = 0) showed that cord mtDNA-CN was statistically significantly associated with increased odds of PTB when z scores ≥ 0 (adjusted OR = 1.59, 95% CI: 1.18, 2.15, P = .002), while the association was weaker and not statistically significant when z scores < 0 (adjusted OR =1.13, 95% CI: 0.83, 1.55, P = .434) (Table 3).

Figure 1. Association of maternal and cord blood mtDNA-CN z scores with preterm birth.

Panel A and B show maternal and cord blood mtDNA-CN z scores in relation to the odds ratio of preterm birth using estimates from logistic restricted cubic spline models with 3 knots at 10th, 50th and 90th percentiles. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated, with a z score of 0 used as the reference point. The models were adjusted for ancestry groups defined by mtDNA haplogroups, maternal age at delivery, maternal education, maternal metabolic conditions, maternal smoking, parity, delivery type, and child sex.

Table 3.

Association of maternal and cord blood mtDNA copy number z scores with preterm birth

Maternal MtDNA-CN z Scores	Overall		Stratified by Cord MtDNA-CN
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)	Low (n=462) Adjusted OR (95% CI)	High (n=536) Adjusted OR (95% CI)	P for interaction
Continuous z	1.00 (0.85, 1.18)	1.00 (0.85, 1.17)	0.66 (0.51, 0.85)**	1.23 (0.99, 1.54)	<0.001
z < 0 (spline)	0.66 (0.51, 0.85)**	0.65 (0.50, 0.84)**	0.65 (0.46, 0.93)*	0.62 (0.41, 0.93)*	0.010
z ≥ 0 (spline)	1.67 (1.25, 2.25)***	1.69 (1.25, 2.30)***	0.67 (0.31, 1.45)	2.04 (1.44, 2.89)***
Cord MtDNA-CN z Scores	Overall		Stratified by Maternal MtDNA-CN
	Unadjusted OR (95% CI)	Adjusted OR (95% CI)	Low (n=491) Adjusted OR (95% CI)	High (n=507) Adjusted OR (95% CI)	P for interaction
Continuous z	1.39 (1.17, 1.65)***	1.36 (1.14, 1.62)***	1.22 (0.97, 1.52)	1.74 (1.32, 2.28)***	0.048
z < 0 (spline)	1.13 (0.83, 1.54)	1.13 (0.83, 1.55)	0.98 (0.70, 1.37)	2.45 (1.05, 5.72)*	0.077
z ≥ 0 (spline)	1.66 (1.24, 2.21)***	1.59 (1.18, 2.15)**	1.66 (1.07, 2.56)*	1.52 (1.02, 2.27)*

ORs and 95% CIs for associations of mtDNA-CN with preterm birth were estimated using logistic regression models. mtDNA-CN z score was modeled either as a continuous variable or as a linear spline with a knot at z = 0. All adjusted models were adjusted for ancestry groups defined by mtDNA haplogroups, maternal age at delivery, maternal education, maternal metabolic conditions, maternal smoking, parity, delivery type, and child sex.

For maternal mtDNA-CN, effect modification by cord mtDNA-CN was assessed by adding a product term between maternal mtDNA-CN (continuous or spline terms) and binary cord mtDNA-CN (low (z scores < 0) vs high (z scores ≥ 0)) to the model. For cord mtDNA-CN, effect modification by maternal mtDNA-CN was assessed by adding a product term between cord mtDNA-CN (continuous or spline terms) and binary maternal mtDNA-CN (low (z scores < 0) vs high (z scores ≥ 0)) to the model. Stratum-specific estimates from these models are presented in the table. P values for interaction were obtained from likelihood ratio tests comparing models with and without the product term.

^*P < .05.

^**P < .01.

^***P < .001.

Abbreviations: CI, confidence interval; mtDNA-CN, mitochondrial DNA copy number; OR, odds ratio.

We further explored whether maternal and cord mtDNA-CN modified each other’s associations with PTB. Among mother-infant dyads with high cord mtDNA-CN (z scores ≥ 0), maternal mtDNA-CN z scores were statistically significantly associated with PTB likelihood in a U-shaped pattern (Table 3 and Figure S2-A). However, among mother-infant dyads with low cord mtDNA-CN (z scores <0), higher maternal mtDNA-CN z scores were monotonically associated with decreased odds of PTB across both sides of the knot at z scores = 0. The direction of association between cord mtDNA-CN and PTB remained positive across strata of maternal mtDNA-CN, although the association was not statistically significant when maternal mtDNA-CN was low (Table 3 and Figure S2-B). The p values from likelihood ratio tests further suggested a statistically significant interaction between maternal and cord mtDNA-CN z scores.

The mtDNA- CN findings for SPTB aligned with those from the primary analysis on overall PTB (Table S8). The findings were also consistent across child sex and ancestry subgroups, with no significant evidence of effect modification (Tables S9 & S10). The association between maternal mtDNA-CN and PTB remained stable after additional adjustment for maternal blood cell composition (Table S11). In contrast, adjustment of cord blood cell composition substantially attenuated the association between cord mtDNA-CN and PTB. To better understand this attenuation, we examined the relationship of cord blood cell composition with gestational age and cord blood mtDNA-CN z scores and found only mild to modest correlations (Figure S3). We also compared the basic characteristics across the overall sample and the above two subsamples and did not observe substantial differences (Table S12). Analyses limited to participants with complete covariate information produced results closely aligned with those from the primary analyses (Table S13).

Discussion

This study investigated maternal and cord blood mtDNA heteroplasmy and mtDNA-CN in relation to PTB, providing novel insights into how mtDNA variations may contribute to PTB. Predicted pathogenic heteroplasmy in maternal blood, but not cord blood, was associated with higher odds of PTB. Maternal mtDNA-CN showed a U-shaped relationship with PTB, and cord mtDNA-CN was positively correlated with PTB. By systematically characterizing the independent and joint associations of these mtDNA variation measures from both maternal and fetal compartments with PTB, our study emphasized the need for further research into the role of mtDNA variations in PTB underlying mechanisms and clinical implications.

We found that maternal predicted pathogenic heteroplasmy was significantly associated with increased odds of PTB, which was consistent with previous small-scale studies that mothers with specific mtDNA variants have a higher proportion of PTB.^{22, 23, 39} In contrast, cord heteroplasmy was not significantly associated with PTB, suggesting that the mtDNA mutational load of fetuses may have a limited immediate impact at birth. While few studies have examined heteroplasmy in maternal or cord blood in relation to PTB, one study reported an association between increased placenta heteroplasmic loads and shorter gestational age.²⁵ While the prior study reported a stronger magnitude of associations in females compared to males, we did not observe substantial sex differences in the current study.

We found that both lower and upper extremes of maternal blood mtDNA-CN z scores were associated with PTB. This U-shaped relationship pattern may help explain inconsistencies in prior research. Wang et al. 2024 reported a negative association between lower pre-conception mtDNA-CN and PTB risk, while Fukunaga et al. 2023 found no linear association.^{26, 27} These discrepancies might also be partially explained by the differences in the timing of mtDNA-CN measurement or study population characteristics. Biologically, both high and low mtDNA-CN may suggest mitochondrial dysfunction. Elevated mtDNA-CN may reflect compensatory upregulation in response to impaired mitochondrial bioenergetics,^{14, 40} but excessive mtDNA-CN may promote mutation accumulation or oxidative stress, exacerbating dysfunction over time.⁴¹ Conversely, low mtDNA-CN may suggest a failure to generate additional mtDNA to maintain metabolic balance, signaling more severe mitochondrial dysfunction. This aligns with prior studies linking low mtDNA-CN to cardiometabolic and neurological diseases in aged populations.¹⁵

We also observed a positive association between cord blood mtDNA-CN level and PTB likelihood, which reinforces findings in previous studies through a larger, racially diverse cohort. Notably, adjusting for white blood cell counts substantially attenuated the association between cord blood mtDNA-CN and PTB, suggesting that variation in cellular composition may account for a substantial portion of the observed relationship. This warrants cautious interpretation of the cord blood findings. Beyond cell composition, MtDNA-CN level in cord blood may reflect a fetal stress response to adverse intrauterine conditions, which is supported by previous studies linking prenatal exposures to environmental toxicants (e.g., tobacco smoking, metal exposures)^{28, 42, 43} with increased cord blood mtDNA-CN. Higher cord mtDNA-CN may reflect a compensatory mitochondrial response to acute prenatal stressors. Alternatively, it may reflect increased oxidative stress, leading to mitochondrial damage and compensatory upregulation of mtDNA replication.⁴⁴ Increased oxidative stress markers, such as malondialdehyde (MDA), have been detected in the cord blood of PTB infants.⁴⁵ Furthermore, elevated mtDNA-CN may also indicate immune system alteration, as some researchers considered blood mtDNA-CN as a biomarker of immune regulation.^{44, 46}

Moreover, our study examined the joint association between maternal and fetal mtDNA-CN and PTB. When cord mtDNA-CN was low, maternal mtDNA-CN showed a negative, monotonic association with PTB, possibly reinforcing the role of chronic maternal mitochondrial dysfunction in maintaining pregnancy. In contrast, when cord mtDNA-CN was high, maternal mtDNA-CN retained a U-shaped relationship with PTB. The co-occurrence of maternal and cord high mtDNA-CN may reflect a shared mitochondrial response to adverse intrauterine conditions. Further research should investigate how maternal health conditions and environmental exposures influence maternal-fetal mtDNA-CN correlations and their implications for PTB and other birth outcomes.

Several limitations should be considered. First, maternal and cord blood mtDNA measures were assessed at a single time point at delivery, limiting our ability to establish a temporal relationship between mtDNA changes and PTB. As a result, the observed associations should not be interpreted as evidence of a causal relationship. It is therefore difficult to determine whether differences in mtDNA measures, particularly mtDNA-CN levels, preceded the onset of PTB or instead reflect acute physiological stress related to labor and delivery. Future prospective studies with biospecimen collection earlier in pregnancy are needed to clarify the potential causal relevance of mtDNA measures for PTB. Second, although we adjusted multiple white blood cell counts in maternal and cord blood, the lack of blood platelet data may introduce measurement errors in mtDNA-CN estimates. Previous research has shown that the relative proportion of platelets was positively correlated with blood mtDNA-CN.³⁸ Third, although our study includes the largest number of PTB newborns compared with previous studies, statistical power may still be limited because functional or predicted pathogenic heteroplasmy was observed in only a small proportion of participants. This concern is particularly relevant when evaluating the joint effect of maternal and cord blood mtDNA measures. In addition, although these analyses were motivated by a priori hypotheses, we conducted multiple related analyses across maternal and cord blood mtDNA measures. The possibility of false positive findings due to multiple comparisons cannot be completely ruled out, and the results should therefore be interpreted cautiously and confirmed in independent studies. Finally, although we carefully adjusted for multiple covariates based on prior literature, residual confounding cannot be completely ruled out, given the incomplete understanding of genetic, environmental, and metabolic influences on mtDNA measures.

To our knowledge, this is the first study to comprehensively examine mtDNA heteroplasmy and mtDNA-CN in relation to PTB, addressing a critical gap in the underlying mechanism of PTB from mtDNA variations and mitochondrial dysfunction perspective. Importantly, our study provides an innovative framework for future research in considering the joint association of mtDNA variations from both maternal and fetal compartments with perinatal health outcomes. Additionally, the U-shaped relationship between maternal mtDNA-CN and PTB observed in our study challenges prior assumptions of linearity in mtDNA-CN research. Future research is expected to consider potential health consequences of both lower and upper extremes of mtDNA-CN level. These findings underscore the potential of mtDNA variations as biomarkers for early risk assessment in PTB. Also, given an increasing interest in mitochondrial-focused therapies,⁴⁷ our findings may help inform future translational medicine research on developing PTB prevention and intervention strategies targeting mitochondrial health.

In conclusion, mtDNA heteroplasmy burden and increased mtDNA-CN level in maternal blood were associated with PTB. Cord blood mtDNA-CN was also positively associated with PTB; however, this association was attenuated after adjustment for white blood cell counts, indicating that cellular composition may influence this relationship. Interestingly, we observed distinct patterns in the joint association of maternal and cord blood mtDNA-CN with PTB, suggesting potential interplay between maternal and fetal mitochondrial variation. These findings highlight the potential involvement of mitochondrial variation in PTB and warrant further investigation in prospective studies.

Impact:

• We found that maternal predicted pathogenic mtDNA heteroplasmy was associated with increased PTB risk; maternal mtDNA-CN showed a U-shaped association with PTB; and higher cord blood mtDNA-CN was positively associated with PTB.

• This study provides the first evidence from a large birth cohort simultaneously assessing maternal and fetal mtDNA heteroplasmy and mtDNA-CN in relation to PTB, adding a mitochondrial genetics perspective to PTB research.

• This study highlights mitochondrial genomic variation as a novel pathway in PTB pathogenesis and suggests the potential utility of mtDNA measures as biomarkers for early risk assessment and as a basis for future mitochondria-targeted interventions.

Data Availability Statement:

The data used in this study are not publicly available due to participant confidentiality and limitations of the informed consent. However, the analytical methods and computer code used for data processing and analysis are available upon reasonable request from qualified researchers.