Prenatal tobacco exposure and adult offspring brain health: the mediating role of leukocyte telomere length in a large population-based cohort Authors: Zirui Wang, Xingyu Zhang, Xianglu Meng, Jiamei Xie, Qiang Xu, Jiayuan Xu, Junping Wang Published: 2026 ABSTRACT AbstractMaternal smoking during pregnancy (MSDP) has been linked to adverse effects on offspring brain health. However, the underlying mechanisms remain unclear. Given that offspring exposed to MSDP exhibit shortened leukocyte telomere length (LTL)—a potential biomarker of accelerated biological ageing—we hypothesized that LTL may mediate the association between MSDP and long-term brain health consequences in adult offspring. Using data from the UK Biobank cohort, we performed association analyses to assess the effects of MSDP (almost 100 000 patients) on LTL, 363 imaging derived phenotypes (IDPs) from structural MRI, 11 kinds of cognitive performances, and 4 mental health outcomes in adult offspring. Mediation analysis was further conducted to evaluate whether LTL mediates the relationship between MSDP and these brain health outcomes. Offspring exposed to MSDP exhibited significantly shorter LTL, greater atrophy in 30 IDPs, poorer performance on 3 cognitive tests, and higher scores on 4 mental health questionnaires compared to unexposed offspring. Mediation analysis revealed that shortened LTL partially mediated the association between MSDP and atrophy in 11 IDPs—particularly in the hippocampus and its subregions—as well as reduced fluid intelligence performance and increased depressive symptoms. In conclusion, this study demonstrates that MSDP may contribute to long-term adverse brain health outcomes in adult offspring. Crucially, we identify accelerated telomere shortening as a partial mediator reflecting cumulative aging that partially explains the link between MSDP and specific adverse outcomes in adult offspring, particularly hippocampal atrophy, declines in fluid intelligence, and increased depressive symptoms. These results underscore the importance of reducing prenatal tobacco exposure to improve population-level brain health. Subject terms: Hippocampus, Clinical genetics Introduction Extensive research has demonstrated that maternal smoking during pregnancy (MSDP) significantly impacts offspring neurodevelopment, with nicotine and other toxins readily crossing the placental barrier [1]. These substances critically disrupt fundamental processes of central nervous system development including neurogenesis, neuronal migration, differentiation, synaptogenesis, and myelination [2]. Neuroimaging studies over the past two decades have consistently identified structural brain alterations in offspring associated with MSDP [3–7]. Preterm infants exposed to MSDP exhibit reduced volumes in the frontal lobe and cerebellum (N = 232) [8]. Recent large-scale research involving children aged 9–11 years (N = 2 704) revealed global reductions in brain volume and surface area, along with decreased cortical gyrification [9]. Furthermore, adolescents with MSDP exposure demonstrate significantly smaller amygdala volume compared to unexposed peers (N = 378) [10]. Furthermore, MSDP has been robustly associated with diverse cognitive deficits across developmental stages in offspring. Children exposed to MSDP show poorer performance in arithmetic and spelling tasks (N = 1 186) [11], while adolescents exhibit marked deficits in auditory and visual attention (N = 181) [12]. Longitudinal studies document persistent cognitive effects, with lower intelligence quotient (IQ) scores observed in childhood (N = 798) [13], adolescence, and young adulthood (N = 3 611 and 4 050) [14]. These findings collectively demonstrate MSDP’s detrimental impact on attention, working memory, and academic abilities [14–18]. The psychiatric implications of MSDP in offspring are equally concerning. For instance, children exposed to MSDP exhibit a 45% increased risk of generalized anxiety disorder and a 75% higher likelihood of developing depression (N = 3 781) [19]. Additionally, MSDP has been identified as a significant risk factor for attention deficit hyperactivity disorder (ADHD) (N = 4 517 and 2 861) [20, 21] and autism spectrum disorder (ASD) in children (N = 8 648) [22]. Longitudinal research further revealed that MSDP elevates the risk of personality disorders and any Axis I psychiatric disorders-including mood, anxiety, and psychotic disorders-by the time offspring reach 30 yo (N = 3 633) [23]. While existing studies have established significant associations between MSDP and adverse neurodevelopmental outcomes in infants, children, and adolescents, critical knowledge gaps remain regarding its long-term effects in adulthood. Furthermore, the generalizability of previous findings has been constrained by relatively small sample sizes. These limitations highlight the need for large-scale investigations examining the enduring consequences of MSDP on brain health in adult populations. Equally important is elucidating the underlying biological mechanisms driving these associations, as such insights could yield valuable biomarkers for early detection and intervention strategies targeting MSDP-related neurological impairments. Addressing these research gaps would significantly advance our understanding of the lifelong neurodevelopmental consequences of prenatal smoke exposure in offspring. Growing evidence suggests that telomere biology may represent a key mechanism underlying the long-term neurodevelopmental consequences of MSDP in offspring. Telomeres, nucleoprotein complexes that cap eukaryotic chromosomes, play a crucial role in maintaining genomic stability[24]. These protective structures progressively shorten with each cellular division and are widely recognized as biomarkers of cellular aging [24]. Importantly, research has consistently demonstrated that MSDP is associated with accelerated telomere shortening during early development. Longitudinal studies have documented progressive leukocyte telomere length (LTL) reduction in infants with MSDP at 4, 12, and 18 months of age [25], while population-based analyses in the UK Biobank (UKB) identified MSDP as the most significant modifiable early-life factor associated with shorter LTL in offspring [26]. The clinical relevance of these findings is underscored by emerging evidence linking shortened LTL to multiple neuropsychiatric outcomes [27]. A study of 435 046 UKB participants of European ancestry found that shorter LTL was associated with adverse neuroimaging phenotypes, such as reduced hippocampal volume, increased white matter hyperintensity burden and disrupted fornix microstructure (evidenced by lower fractional anisotropy and higher mean diffusivity), as well as poorer cognitive performance [28, 29]. Similarly, a cross-sectional study of 1 960 Texas residents demonstrated positive associations between LTL and both global (total, gray, and white matter) and regional (hippocampus, amygdala, and inferior temporal gyrus) brain volumes [30]. Consistent with these observations, a study of 976 community-dwelling Chinese men aged ≥ 65 years in Hong Kong revealed that shorter LTL correlated with poorer performance on delayed recall and category verbal fluency tasks [31]. Furthermore, findings from a 10-year longitudinal community study indicated that individuals with persistently elevated depressive symptoms exhibited accelerated LTL attrition over time [32]. Numerous studies have demonstrated that MSDP is associated with both LTL shortening and impaired brain health in offspring. Furthermore, evidence suggests that shortened LTL may itself contribute to impaired brain health independently [24, 33]. This observation led us to hypothesize that LTL shortening might serve as a mediating biomarker in the pathway through which MSDP influences offspring brain health. To investigate this potential mechanism, we leveraged data from the UKB, the largest population-based prospective cohort study to date, comprising over 500 000 community-dwelling participants aged 37–73 years who provided comprehensive genetic, lifestyle, and health information. This extensive dataset provides an unparalleled resource for examining the long-term effects of MSDP on adult offspring brain health, which we operationalized through multiple measures including brain structure, cognitive performance, and mental health phenotypes. Our analytical approach proceeded in several stages: First, we investigated the associations between MSDP and offspring brain health, while controlling for potential confounding variables. To ensure robust findings, we further implemented two types of sensitivity analyses restricting the sample to non-smokers or individuals free of neuropsychiatric disorders to examine whether the observed associations remained significant in these subgroups. Next, we applied propensity score matching (PSM) to address potential selection bias and balance characteristics between exposed and unexposed groups. PSM is a statistical technique designed to reduce selection bias in observational studies. It calculates the probability (the propensity score) of each subject receiving a specific intervention based on their observed characteristics. Subjects who received the intervention are then matched with untreated subjects with similar propensity scores. This creates comparable groups, mimicking randomized controlled trial conditions and enabling more reliable estimation of the intervention’s effect [34]. Mediation analyses were conducted on the associations that persisted after sensitivity analysis and PSM to assess whether shortened LTL mediated the relationship between MSDP and offspring brain health outcomes (Fig. 1A). Study overview and the relationships between MSDP and brain health in adult offspring. A Study overview: LTL shortening serves as a mediating biomarker in the pathway through which MSDP influences adult offspring brain health outcomes. B Results of PSM analyses of MSDP and cognitive performances, mental health phenotypes and LTL in adult offspring. The x-axis shows each phenotype, and the y-axis shows β value representing the association between MSDP and each phenotype. C Results of PSM analyses of MSDP and IDPs. The x-axis shows each IDP and the y-axis shows β value representing the association between MSDP and each IDP. Gray points represent non-significant results. CT cortical thickness, EPQ-R Eysenck Personality Questionnaire-Revised, GAD Generalized Anxiety Disorder, IDPs imaging derived phenotypes, LTL leukocyte telomere length, MSDP maternal smoking during pregnancy, SA surface area, PHQ Patient Health Questionnaire, PSM propensity score matching, VL volume. Materials and methods Participants We conducted this study using data from the UKB (https://www.ukbiobank.ac.uk/). Baseline assessments were performed at 22 dedicated assessment centers throughout England, Scotland, and Wales during 2006–2010 [35]. The UKB study received ethical approval from the North West Multi-center Research Ethics Committee, and all participants provided written informed consent. Our analysis focused on data pertaining to MSDP, LTL, brain structure, cognitive function tests, and mental health questionnaires. Demographic data We extracted demographic variables, including sex, age, self-reported ethnicity, smoking status, alcohol intake frequency, body mass index (BMI), education level, household income, and Townsend deprivation index (TDI). After excluding participants with incomplete demographic data, our analytical sample comprised 404 925 individuals (Please see Table 1 for the detailed demographic characteristics). Table 1Demographic characteristics of the study population.MSDPWithout MSDPΖ / χ2P valueNumbers of participants104,658247,294Sex (M / F)50,597/54,061115,322/131,97286.451.44E-20Age (Years) [M (Q1, Q3)]56 (50, 62)57 (49, 63)−18.531.10E-76Ethnicity1012.486.66E-214 White British97,793 (93.43%)222,807 (90.10%) Other6865 (6.56%)24,487 (9.90%)BMI (kg/m2) [M (Q1, Q3)]27.20 (24.50, 30.40)26.50 (23.90, 29.50)−44.360.00E + 00Smoking Status−19.001.79E-80 Never56,169 (53.67%)139,324 (56.34%) Previous35,933 (34.34%)84,912 (34.34%) Current12,556 (12.00%)23,058 (9.32%)Alcohol intake frequency−13.701.04E-42 Daily or almost daily23,122 (22.09%)51,354 (20.77%) Three or four times a week25,241 (24.12%)59,009 (23.86%) Once or twice a week27,232 (26.02%)63,296 (25.60%) One to three times a month11,727 (11.21%)27,342 (11.06%) Special occasions only10,772 (10.29%)27,259 (11.02%) Never6564 (6.27%)19,034 (7.70%)Education−12.697.08E-37 College or University degree31,567 (30.16%)95,352 (38.56%) A levels/AS levels or equivalent11,755 (11.23%)29,638 (11.99%) O levels/GCSEs or equivalent23,064 (22.04%)51,115 (20.67%) CSEs or equivalent6982 (6.67%)11,661 (4.72%) NVQ or HND or HNC or equivalent8030 (7.67%)14,754 (5.97%) Other professional qualifications eg: nursing, teaching5223 (4.99%)12,579 (5.09%) None of the above18,037 (17.23%)32,195 (13.02%)Household income (in GBP per year)−6.421.34E-10 Less than 18,00023,302 (22.26%)53,209 (21.52%) 18,000–30,99926,084 (24.92%)62,587 (25.31%) 31,000–51,99928,290 (27.03%)64,488 (26.08%) 52,000–100,00021,473 (20.52%)52,434 (21.20%) Greater than 100,0005509 (5.26%)14,576 (5.89%)Townsend [M (Q1, Q3)]−2.03 (−3.57, 0.70)−2.27 (−3.72, 0.24)−22.896.20E-116All values in the table are presented with two decimal places.Descriptive statistics are presented as median (interquartile range, IQR) for continuous variables with non-normal distributions (age, BMI, Townsend deprivation index), as frequency (percentage) for categorical variables (sex, ethnicity) and ordinal variables (smoking status, alcohol intake frequency, education, household income). Group differences were assessed using the Mann-Whitney U test for continuous and ordinal variables, and the Pearson chi-square test for categorical variables.MSDP maternal smoking during pregnancy, BMI body mass index, M median, Q1 first quartile, Q3 third quartile, GBP Great British Pound. Demographic characteristics of the study population. All values in the table are presented with two decimal places. Descriptive statistics are presented as median (interquartile range, IQR) for continuous variables with non-normal distributions (age, BMI, Townsend deprivation index), as frequency (percentage) for categorical variables (sex, ethnicity) and ordinal variables (smoking status, alcohol intake frequency, education, household income). Group differences were assessed using the Mann-Whitney U test for continuous and ordinal variables, and the Pearson chi-square test for categorical variables. MSDP maternal smoking during pregnancy, BMI body mass index, M median, Q1 first quartile, Q3 third quartile, GBP Great British Pound. MSDP MSDP exposure was assessed via a self-reported touchscreen question in the Early Life Factors category: “Did your mother smoke regularly around the time when you were born?” Possible responses included “Yes,” “No,” “Do not know,” and “Prefer not to answer.” We classified “Do not know” and “Prefer not to answer” as missing values and excluded them from analysis. After further removing participants with missing MSDP or demographic data, 351 952 individuals remained (104 658 exposed to MSDP; 247 294 unexposed). Imaging derived phenotypes (IDPs) IDPs were obtained from participants who underwent structural MRI (sMRI) scanning using Siemens Skyra 3 T scanners with 32-channel head coils across four UKB assessment centers. The UKB team processed the raw T1-weighted images through their standardized quality-control and image-processing pipeline [36] (https://www.fmrib.ox.ac.uk/ukbiobank/), yielding 363 IDPs comprising 7 global measures (normalized total brain, gray matter, and white matter volume (VL), total surface area (SA) and mean cortical thickness (CT) of bilateral hemispheres) and 356 local measures (224 VL measures including 66 cortical, 14 subcortical, and 28 cerebellar regions from the Desikan-Killiany, Harvard-Oxford, and Diedrichsen atlases respectively, plus 20 amygdala, 44 hippocampal, and 52 thalamic subregions segmented using FreeSurfer v7.0; along with 66 SA and 66 CT measures from the Desikan-Killiany atlas). After excluding outliers (values beyond median ± 6 median absolute deviations (MAD)) and participants with missing MSDP or demographic data, final sample sizes for IDPs ranged from 31 017–31 639. Cognitive function assessments We analyzed nine cognitive tests administered via touchscreen questionnaires, assessing: numeric memory, fluid intelligence, trail making, matrix pattern completion, symbol digit substitution, tower rearranging, paired associate learning, prospective memory and pairs matching. Outliers-defined as values outside (median ± 6 MAD)-were excluded. Final sample sizes for cognitive tests ranged from 24 541–351 460, depending on data availability. Detailed descriptions of 11 cognitive performances derived from these 9 cognitive tests are available on the UKB website (https://biobank.ndph.ox.ac.uk/showcase/label.cgi?id=100026). For all cognitive tests, higher scores indicate better performance. Mental health phenotypes We analyzed four key mental health domains: neuroticism, depression, anxiety and life satisfaction. Sample sizes for these assessments ranged from 86 291–291 961 participants based on data availability. Neuroticism assessment Neuroticism, representing the tendency to experience negative emotions, was evaluated using the 12-item neuroticism scale from the Eysenck Personality Questionnaire-Revised (EPQ-R) [37]. Participants responded to each item with “yes”, “no”, “do not know”, or “prefer not to answer”. The total score (range: 0–12) was calculated by summing affirmative responses, with higher scores indicating greater neuroticism severity. “Do not know” and “prefer not to answer” responses were treated as missing data. Depression symptoms Depressive symptoms were measured using the 9-item Patient Health Questionnaire (PHQ-9) [38]. Each item assessed symptom frequency over the preceding two weeks using a 4-point scale: “not at all,” “several days,” “more than half the days,” and “nearly every day,” with an additional “prefer not to answer” option. Total scores ranged from 9–36, with higher scores reflecting more severe depressive symptoms. Anxiety symptoms Anxiety was evaluated through the 7-item Generalized Anxiety Disorder scale (GAD-7) [39]. Similar to the PHQ-9, responses were recorded on a 4-point frequency scale (“not at all” to “nearly every day”) plus “prefer not to answer.” The total score (range: 7–28) represented anxiety severity, with higher scores indicating greater symptom burden. Life satisfaction Life satisfaction was assessed across five domains (family relationships, financial situation, friendships, health, and work/job) using a 6-point Likert scale ranging from “extremely happy” to “extremely unhappy.” The composite score (range: 5–30) was derived by summing responses across domains, with higher scores indicating lower overall life satisfaction. For all mental health phenotypes (EPQ-R neuroticism score, PHQ-9 depression score, GAD-7 anxiety score, and life satisfaction composite score), higher scores indicate worse outcomes (i.e., greater neuroticism, more severe depressive symptoms, more severe anxiety symptoms, and lower life satisfaction, respectively). Measurement of LTL LTL was quantified using a validated quantitative polymerase chain reaction (qPCR) assay that measures the relative ratio of telomere repeat copy number to single-copy gene copy number (T/S ratio) [40]. This method underwent rigorous quality control procedures, with all measurements adjusted for potential technical confounding factors [40]. To ensure normal distribution of the data, raw LTL values were log-transformed and subsequently standardized to Z-score by the UKB team, enabling cross-dataset comparability. For all subsequent analyses in our study, we utilized these log-transformed and Z-score standardized LTL values. After excluding participants with missing MSDP information or incomplete demographic data, the final analytical sample comprised 330 731 individuals. Association analyses We conducted linear regression analyses to examine the associations between MSDP and four categories of outcomes: IDPs, cognitive performance, mental health phenotypes and LTL. All outcome variables were standardized using Z-score transformation. The linear regression models were employed adjusting for age, sex, ethnicity, lifestyle factors (smoking status, alcohol intake frequency, and BMI), socioeconomic status (education, income and TDI) and the first 10 genetic principal components. For IDPs, assessment center was included as an additional covariate, while complete blood count was adjusted for in LTL analyses. Local IDP measures (VL, SA and CT) were further adjusted for total intracranial volume, total hemispheric surface area, and global mean thickness, respectively. To assess robustness, we performed two sets of sensitivity analyses using the same covariates as in the above correlation analysis: (1) restricting to never-smokers to eliminate confounding by personal smoking, and (2) excluding participants with neuropsychiatric disorders (ICD-10 chapters V and VI). To further address potential selection bias, we implemented 1:1 nearest-neighbor PSM without replacement, balancing participants on demographic (age, sex and ethnicity), lifestyle (smoking status, alcohol intake frequency, BMI) and socioeconomic (education, income and TDI) factors. Sequentially, the results from the primary association analyses underwent two sets of sensitivity analyses; only those that remained significant were then advanced to PSM and finally to the following mediation analyses. Mediation analyses To investigate the potential mediating role of LTL in the relationships between MSDP and brain health phenotypes in offspring, we performed mediation analyses based on prior association results. The mediation model was constructed with MSDP as the exposure (X), LTL as the mediator (M), and each brain health measure as the outcome (Y), while adjusting for the same set of confounders used in the association analyses. The total effect (c) represented the unadjusted association between MSDP and each brain health phenotype, whereas the direct effect (c′) reflected the MSDP-brain health relationship after adjusting for LTL. The indirect (mediating) effect, calculated as the product of two regression coefficients (a*b; where a denotes the effect of MSDP on LTL and b represents the effect of LTL on brain health measures adjusted for MSDP), quantified the proportion of the total effect mediated through LTL. Statistical analyses All the association analyses and mediation analysis were corrected for multiple testing using the Bonferroni method. For the primary association analyses, the significance threshold was set at P = 1.32 × 10⁻⁴ (0.05/379), corrected for 379 total tests (363 IDPs, 11 cognitive, 4 mental health measures and 1 LTL). For the two sets of sensitivity analyses, the significance threshold was set at P = 3.29 × 10⁻⁴ (0.05/76/2), corrected for 76 total tests (67 IDPs, 4 cognitive, 4 mental health measures and 1 LTL). For the PSM analyses, the significance threshold was set at P = 9.62 × 10⁻⁴ (0.05/52), corrected for 52 total tests (44 IDPs, 3 cognitive, 4 mental health measures and 1 LTL). In all mediation analyses, confidence intervals (CIs) were calculated using a bootstrap method with 1000 resampled iterations to estimate the 95% CIs for all mediation effects. The significance of any mediation effect was determined if the 95% CIs did not contain zero and the associated P value was less than 1.16 × 10⁻³ (0.05/43), corrected for 43 total tests (36 IDPs, 3 cognitive, and 4 mental health measures). All statistical analyses were performed using R software (version 4.2.2; R Foundation for Statistical Computing, Vienna, Austria). The PSM and mediation models were implemented using the “MatchIt” and “mediation” packages, respectively. Results MSDP associated with widespread alterations in IDPs in offspring The primary analyses revealed significant associations between MSDP and 67 IDPs (33 VL, 23 SA, and 11 CT-related measures). Sensitivity analyses in non-smoking (N = 19 663–20 032) and neuropsychiatric disorder-free subgroups (N = 25 921–26 424) confirmed 45 and 57 significant IDPs, respectively, with 44 overlapping findings (21 VL, 16 SA, 7 CT-related measures). Thirty-six IDPs remained significant in PSM analyses (N = 18 590–18 956). Notably, offspring with MSDP exhibited widespread reductions in 33 IDPs, including all 21 VL measures, 10 SA measures, and 2 CT measures, while only 3 CT measures showed positive associations. The most pronounced effects were observed in VL and SA measures, with MSDP-exposed individuals showing significantly smaller VL and SA in multiple regions including bilateral inferior parietal gyrus (right VL: β = −0.15, P = 2.85 × 10⁻³⁵, N = 18 954; left VL: β = −0.15, P = 2.41 × 10⁻³⁴, N = 18 954; right SA: β = −0.15, P = 7.83 × 10⁻³⁴, N = 18 954; left SA: β = −0.15, P = 1.94 × 10⁻³², N = 18 956), middle temporal gyrus (right VL: β = −0.13, P = 1.08 × 10⁻²⁷, N = 18 954; left VL: β = −0.13, P = 7.42 × 10⁻²⁷, N = 18 954), and pericalcarine cortex (right SA: β = −0.14, P = 3.17 × 10⁻³², N = 18 954; left SA: β = −0.14, P = 3.18 × 10⁻³², N = 18 956). Complete association results for all 363 examined IDPs are detailed in Supplementary Table 1–3 and Fig. 1C. MSDP linked to impaired cognitive performance in offspring The primary analyses demonstrated that MSDP was significantly associated with poorer performance in four cognitive tests in offspring, three of which were further supported by sensitivity analyses in both non-smoking (N = 15 698-65 566) and neuropsychiatric disorder-free subgroups (N = 20 871-88 463). PSM analyses confirmed three results with consistent significance: fluid intelligence (β = −0.08, P = 2.61 × 10⁻²⁷, N = 67 938), matrix pattern completion (β = −0.11, P = 3.09 × 10⁻¹¹, N = 14 830), and paired associate learning tests (β = −0.09, P = 9.24 × 10⁻⁸, N = 14 974), indicating that offspring exposed to MSDP exhibited significantly poorer cognitive performance compared to unexposed individuals. Complete association results between MSDP and all 11 cognitive measures are presented in Fig. 1B and Supplementary Tables 4–6. MSDP linked to adverse mental health outcomes in offspring The primary analyses revealed significant associations between MSDP and poorer mental health outcomes in four questionnaires in offspring, which were robust in sensitivity analyses restricted to non-smoking (N = 49 930–162 137) and neuropsychiatric disorder-free subgroups (N = 67 220–221 963), with all four measures maintaining significance. PSM analyses further confirmed these results: EPQ-R (β = 0.20, P = 6.78 × 10⁻⁴⁹, N = 176 468), PHQ-9 (β = 0.11, P = 1.27 × 10⁻¹¹, N = 67 782), GAD-7 (β = 0.06, P = 1.03 × 10⁻⁴, N = 68 198), and life satisfaction (β = 0.06, P = 2.56 × 10⁻¹¹, N = 49 980). Notably, offspring exposed to MSDP exhibited higher scores on the EPQ-R (indicating greater emotional instability), PHQ-9 (suggesting depressive symptoms), GAD-7 (reflecting anxiety symptoms), and Life Satisfaction Questionnaire (suggesting life dissatisfaction) compared to unexposed individuals. These results collectively indicate that MSDP is associated with worse mental health outcomes in offspring. Complete association results between MSDP and all four mental health phenotypes are presented in Fig. 1B and Supplementary Table 7–9. MSDP associated with shorter LTL in offspring The primary association analysis demonstrated a consistent negative association between MSDP and offspring LTL which remained robust in sensitivity analyses restricted to non-smoking (N = 183 536) and neuropsychiatric disorder-free subgroups (N = 249 689). PSM analysis further confirmed this association (β = −0.05, P = 1.82 × 10⁻²⁶, N = 197 080), indicating that offspring exposed to MSDP had significantly shorter LTL compared to unexposed individuals. These results suggest that prenatal tobacco exposure may accelerate cellular aging in offspring, as reflected by telomere shortening. Complete association results are presented in Fig. 1B and Supplementary Table 10–12. LTL partially mediates the association between msdp and structural brain alterations in offspring In light of the significant associations of MSDP with LTL and 36 IDPs, we further investigated whether shortened LTL significantly mediated the associations between MSDP and changes in IDPs. Our mediation analysis revealed significant indirect effects of LTL in the relationship between MSDP and structural brain alterations. Among 36 IDPs associated with MSDP, shortened LTL significantly mediated the association with atrophy in 11 IDPs (9 VL-related and 2 SA -related measures), accounting for 0.96–3.32% of the total effects. The mediated effects were particularly prominent in the VLs of hippocampus and its subregions—key structures for memory formation and emotional regulation, including right CA1-head (indirect effect β = −1.23 × 10⁻³, 3.32% of total effect), right hippocampal-head (indirect effect β = −1.12 × 10⁻³, 3.16% of total effect), right whole hippocampus (indirect effect β = −1.03 × 10⁻³, 2.57% of total effect), bilateral subiculum-head (right: indirect effect β = −1.31 × 10⁻³, 2.48% of total effect; left: indirect effect β = −1.11 × 10⁻³, 2.32% of total effect), right hippocampal-body (indirect effect β = −9.70 × 10⁻⁴, 2.42% of total effect) and right CA1-body (indirect effect β = −9.64 × 10⁻⁴, 1.46% of total effect). Additional mediated effects were observed in the VLs of right anteroventral thalamic nucleus—a component of the limbic system involved in emotional and memory processing (indirect effect β = −1.42 × 10⁻³, 2.85% of total effect), right bankssts (indirect effect β = −1.10 × 10⁻³, 1.54% of total effect) and the total SAs of bilateral hemisphere (left: indirect effect β = −8.07 × 10⁻⁴, 1.07% of total effect; right: indirect effect β = −7.58 × 10⁻⁴, 0.96% of total effect).These findings suggest that MSDP contributes to localized and global brain structural atrophy, particularly in memory-related regions, partially through its impact on cellular ageing processes as reflected by telomere shortening. Complete mediation results for all 36 IDPs are presented in Fig. 2A–C, with detailed findings in Supplementary Table 13 and Supplementary Fig. 1. The results of mediating role of LTL in the relationships between MSDP and brain health outcomes in adult offspring. A Mediation results of MSDP, LTL, and brain health outcomes. The x-axis shows each brain health phenotype and the y-axis shows the proportion of the total effect mediated through LTL. Gray points represent non-significant results. B The forest plot shows the significant results of the mediation analyses. The x-axis shows indirect (mediating) effect and the y-axis shows significant brain health phenotypes. C, D, E The detailed results of mediation analyses when LTL as the mediator and the volume of right CA1-head, the performance of fluid intelligence test, and the score of PHQ-9 as outcomes respectively. CT cortical thickness, IDPs imaging derived phenotypes, LTL leukocyte telomere length, MSDP maternal smoking during pregnancy, SA surface area, PHQ Patient Health Questionnaire, VL volume. Shortened LTL partially mediates the association between MSDP and poor cognitive performances in offspring Building upon the established associations between MSDP, shortened LTL, and impaired cognitive performances, we conducted mediation analyses to examine potential mechanistic pathways. Our findings revealed that shortened LTL significantly mediated 1.12% of the total association between MSDP and poor performance on the fluid intelligence (indirect effect β = −8.42 × 10⁻⁴), as detailed in Supplementary Table 14 and illustrated in Figs. 2A, B, D. This statistically significant, albeit modest, mediation effect indicates that MSDP might partially contribute to impaired fluid intelligence in offspring through accelerated cellular ageing, as measured by telomere shortening. Notably, no significant mediation effects were observed for the paired associate learning or matrix pattern completion test, indicating domain-specific pathways in the relationship between prenatal tobacco exposure and cognitive outcomes. LTL partially mediates the association between msdp and mental health phenotypes in offspring Given the established associations between MSDP, LTL shortening, and adverse mental health outcomes, we investigated LTL as a potential mediator in these relationships. Mediation analyses revealed a small but statistically significant indirect effect of shortened LTL on the relationship between MSDP and increased depression symptoms (indirect effect β = −3.58 × 10⁻⁴, 0.39% of total effect; Supplementary Table 15, Fig. 2A, B, E). This finding suggests that MSDP may contribute to increased depressive symptoms in offspring partially through accelerated cellular ageing, as indicated by telomere shortening. No significant mediation effects were observed for neuroticism, anxious symptoms or life satisfaction, indicating specificity in the pathway from prenatal tobacco exposure to increased depressive symptoms via LTL. Discussion Leveraging the extensive UKB cohort, this large-scale study demonstrates that MSDP is associated with measurable and enduring neurobiological consequences in adult offspring. Specifically, adult offspring exposed to MSDP exhibit: (i) shorter LTL, (ii) widespread structural brain alterations, particularly reduced gray matter VL and SA, (iii) poorer performance on fluid intelligence, matrix pattern completion and paired associate learning tests, and (iv) elevated scores on indices of neuroticism, depressive symptoms, generalized anxiety and diminished life satisfaction. Crucially, the mediation analyses reveal that accelerated telomere attrition partially mediates the relationships between MSDP and structural brain atrophy in hippocampal and cortical regions, reduced fluid intelligence and increased depressive symptomatology. These data provide the first population-scale evidence that telomere biology may serve as an early-life biomarker and mechanistic pathway linking prenatal tobacco exposure to long-term brain health in offspring. The consistency of MSDP-related deficits across neuroanatomical, cognitive and psychiatric domains aligns with a developmental-programming framework in which gestational insults are biologically embedded and propagate across the lifespan. MSDP-associated VL reductions in both the cerebral cortex and basal ganglia, together with SA reductions in the cerebral cortex, which are accompanied by poorer performance on key cognitive tasks and increased symptoms related to emotional instability, depression, and anxiety. These MSDP-related brain impairments recapitulate patterns observed in smaller pediatric and adolescent samples [20–22, 41] and extend them into middle and older adulthood. The convergence of these structural, cognitive, and emotional correlates suggests that MSDP may be linked to a broad neurodevelopmental detrimental impact in offspring. Most importantly, this study suggests that shortened LTL serves as a significant, albeit partial, mediator in the association between MSDP and brain structural, cognitive, and mental health outcomes in offspring, which is consistent with extensive evidence that prenatal tobacco smoke exposure induces oxidative stress, inflammation, and mitochondrial dysfunction—the key drivers of telomere attrition [42–44]. The most pronounced mediation effects were observed in brain structure, where shortened LTL significantly mediated the association between MSDP and the loss of VLs and SAs in 11 brain regions. Notably, the strongest mediations (explaining 1.46–3.32% of total effects) were concentrated in the hippocampus and its subregions—critical structures for memory formation and emotional regulation—whereas no such mediating effects were observed for CT measures. The hippocampus exhibits a unique vulnerability to aging and stress, attributable in part to its integral role in regulating the hypothalamic-pituitary-adrenal (HPA) axis [45]. This characteristic may render hippocampal volume particularly susceptible to the effects of accelerated cellular aging, as indexed by LTL shortening. In contrast, CT is a complex trait reflecting neuronal size, neuropil, and glial cell density, which may have different developmental origins and be less directly tied to the cumulative cellular replicative history captured by LTL [46]. This anatomical dissociation underscores the value of examining multiple brain parameters and suggests that MSDP may impact the brain through distinct mechanisms, with telomere-mediated pathways playing a more prominent role in subcortical structures like the hippocampus. Similarly, in cognitive and mental health domains, we observed domain-specific mediation effects. Shortened LTL significantly mediated 1.12% of the association between MSDP and fluid intelligence deficits, and 0.39% of the association with depressive symptoms. The selective mediation of fluid intelligence (but not other cognitive domains) and depressive symptoms (but not anxiety or neuroticism) suggests pathway specificity. Fluid intelligence relies heavily on prefrontal-hippocampal networks [47]- precisely the regions showing significant structural mediation. Similarly, the hippocampus regulates hypothalamic-pituitary-adrenal axis function [48], and its MSDP-associated atrophy via telomere shortening may preferentially increase depression vulnerability. The modest mediation proportions observed in our study (0.39–3.32%) suggest that LTL shortening is likely one of multiple interconnected pathways linking MSDP to brain health outcomes in offspring. Rather than viewing LTL as an isolated primary driver, we propose that it may be more accurately conceptualized as a proxy marker of cumulative biological aging, reflecting the integrated impact of upstream processes such as MSDP-induced oxidative stress, systemic inflammation, and mitochondrial dysfunction [42–44]. From this perspective, the mediation effect of LTL partially captures the contribution of these broader accelerated aging mechanisms (including epigenetic modifications or neuroinflammation) to the observed neuroanatomical, cognitive, and mental health outcomes. The robustness of our findings is supported by several key methodological strengths including the large population-based cohort, harmonized neuroimaging, cognitive and mental health protocols, rigorous quality control, comprehensive adjustment for confounders (sociodemographic, lifestyle, clinical factors) and a suite of sensitivity analyses (PSM, restriction to never-smokers and neuropsychiatric-naïve subgroups). Several limitations merit consideration. First, the retrospective ascertainment of MSDP introduces potential recall bias, which could operate in two directions: (1) differential misclassification—offspring with poorer brain health may over-report exposure, inflating associations; or (2) non-differential misclassification—random recall errors would bias results toward the null, attenuating true effects. In mediation analysis, such misclassification can bias both total and indirect effect estimates, typically reducing the mediated proportion. Moreover, our analysis lacked information on crucial dimensions of exposure such as smoking intensity (e.g., number of cigarettes per day), duration, and timing during pregnancy (e.g., first trimester vs. throughout). This limitation prevents us from exploring potential dose-response relationships and may lead to an underestimation of the true association between MSDP and offspring brain health outcomes. Third, the modest mediation proportions suggest other pathways (e.g., epigenetic modifications, neuroinflammation and et al.) likely contribute significantly to MSDP’s neurodevelopmental consequences [49–51]. Fourth, while our mediation models were theoretically grounded in a temporal framework where MSDP precedes LTL attrition and subsequent brain changes, the cross-sectional nature of the LTL, neuroimaging, and behavioral assessments in adulthood precludes definitive conclusions about causality. Therefore, our mediation findings should be interpreted as statistical mediation, indicating that LTL statistically accounts for a portion of the MSDP-outcome association, rather than as definitive proof of a causal pathway. Future longitudinal studies with repeated measures from birth onwards are essential to validate the temporal dynamics of this proposed mechanism. Fifth, LTL was measured in leukocytes and may not fully reflect telomere dynamics in neural tissue; however, peripheral and central telomere lengths are moderately correlated and jointly influenced by systemic oxidative stress [52–54]. Sixth, UKB participants are healthier and more socio-economically advantaged than the general UK population, potentially underestimating true effect sizes. Seventh, our findings should also be interpreted with caution given that we lacked data on several prenatal programming factors (e.g., maternal stress, nutrition, and secondhand smoke exposure) and maternal age that could potentially confound the observed associations. Finally, the generalizability of these findings, derived from a predominantly White British cohort, needs to be confirmed in more diverse populations. In summary, MSDP is a modifiable risk factor with lifelong detrimental consequences for offspring brain health, imposing a significant public health burden. The population-level impact is substantial given smoking prevalence during pregnancy (∼10% in high-income countries) [55, 56]. This underscores the critical importance and cost-effectiveness of smoking cessation programs for pregnant women as a primary neuroprotective public health strategy. The identification of LTL shortening as a partial mediator offers potential avenues for early-risk stratification and targeted interventions. Shortened LTL can serve as a biomarker to identify high-risk individuals for strategies such as cognitive training, stress reduction, or investigating telomere-protective compounds. Ultimately, these findings illuminate the biological pathway from prenatal insult to adult brain phenotypes, highlighting the imperative to mitigate the lifelong neurological legacy of prenatal tobacco exposure. Supplementary information Supplementary Materials Supplemental Tables Supplementary Figure legends Acknowledgements The authors wish to express their gratitude to all the investigators and participants who contributed the data to the UKB. This work was supported by the following grants: the Tianjin Health Research Project (Grant No. TJWJ2025ZD001) and Tianjin Science and Technology Plan Project (25ZXWZSY00070) to J.W; the National Natural Science Foundation of China (Grant No. 82371924), National Key Project of “Inter-governmental International Scientific and Technological Innovation Cooperation” (Grant No. 2023YFE0199700), the Tianjin Young Talents in Science and Technology (Grant No. QN20230336), the Non-profit Central Research Institute Fund of Chinese Academy of Medical Sciences (Grant No. 2024-JKCS-18), the Tianjin Science and Technology Commission Major Special Project in Public Health Science and Technology (Grant No. 24ZXGQSY00050), and the Tianjin Medical University “Clinical Talent Training 123 Climbing Plan” to J.X; the New Star of Excellence Project of Tianjin Medical University General Hospital to Q.X (Grant No. 209060403205); and the Tianjin Key Medical Discipline (Specialty) Construction Project (Grant No. TJYXZDXK-3-008C). Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. These authors contributed equally: Zirui Wang, Xingyu Zhang, Xianglu Meng, Jiamei Xie. These authors jointly supervised this work: Qiang Xu, Jiayuan Xu, Junping Wang. Contributor Information Qiang Xu, Email: xuqiang@tmu.edu.cn. Jiayuan Xu, Email: xujiayuan@tmu.edu.cn. Junping Wang, Email: wangjunping_tj@163.com. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04170-8. References