Pregnancy renders anatomical changes in hypothalamic substructures of the human brain that relate to aspects of maternal behavior

Animal studies have shown that pregnancy is associated with neural adaptations that promote maternal care. The hypothalamus represents a central structure of the mammalian maternal brain and hormonal priming of specific hypothalamic nuclei plays a key role in the induction and expression of maternal behavior. In humans, we have previously demonstrated that becoming a mother involves changes in grey matter anatomy, primarily in association areas of the cerebral cortex. In the current study, we investigated whether pregnancy renders anatomical changes in the hypothalamus. Using an advanced delineation technique, five hypothalamic substructures were defined in longitudinal MRI scans of 107 women extracted from two prospective pre-conception cohort studies, including 50 women who were scanned before and after pregnancy and 57 nulliparous control women scanned at a similar time interval. We showed that becoming a mother is associated with volume reductions in the anterior-superior, superior tuberal and posterior hypothalamus. In addition, these structural changes related to hormonal levels during pregnancy and specific aspects of self-reported maternal behavior in late pregnancy, including maternal-fetal attachment and nesting behavior. These findings show that pregnancy leads to changes in hypothalamic anatomy and suggest that these contribute to the development of maternal behavior in humans, supporting the conservation of key aspects of maternal brain circuitry and their role in maternal behavior across species.


Introduction
Pregnancy represents one of life's major transition periods and is associated with unparalleled endocrine surges.This cascade of hormonal fluctuations orchestrates radical physiological and physical adaptations in the female body.Animal studies have consistently shown that reproduction is also associated with dramatic changes in the brain (Brunton and Russell, 2008;Kohl and Dulac, 2018;Leuner and Sabihi, 2016;Numan, 2020;Pawluski et al., 2022).These neural adaptations contribute to the emergence of maternal behaviors, a set of species-specific behaviors promoting the development and protection of the highly dependent offspring (Brunton and Russell, 2008;Kohl and Dulac, 2018;Pawluski et al., 2022).In rodents, these behaviors include nesting behavior, pup retrieval, nursing, maternal aggression and grooming/licking of the offspring (Kohl and Dulac, 2018;Numan and Insel, 2003).
One of the core brain structures of the mammalian maternal brain is the hypothalamus.Several nuclei located in this brain structure, such as the medial preoptic area (mPOA) and paraventricular nucleus (PVN), are known to undergo reproduction-related morphological and functional adaptations that trigger the onset and expression of maternal behaviors in rodents.For instance, hormonal priming of the mPOA during late pregnancy and parturition in rats plays a central role in the switch from pup avoidance to pup approach and the development of maternal caregiving (Brunton and Russell, 2008;Kohl and Dulac, 2018;Pawluski et al., 2022).Accordingly, inhibition of the mPOA during pregnancy in rats was found to reduce basic maternal behavior, including the quality of nestbuilding and pup retrieval (Gómora-Arrati et al., 2016;Kohl et al., 2018;Numan and Stolzenberg, 2009).Other hypothalamic structures such as the PVN and supraoptic nucleus (SON) also play a critical role in the initiation and maintenance of maternal behavior (Bayerl et al., 2016;Francis et al., 2000;Hillerer et al., 2014;Insel et al., 1989).
Although extensively studied in other animals, in humans, relatively little is known about the impact of pregnancy on the brain.We have previously discovered that pregnancy renders long-lasting changes in the grey matter anatomy of the human brain (Hoekzema et al., 2022(Hoekzema et al., , 2020(Hoekzema et al., , 2017)).These changes were primarily located in higher-order association areas of the cerebral cortex.These areas have greatly expanded in the human lineage, which could point to species-specific rather than evolutionarily conserved effects.To examine whether becoming a mother in humans is associated with changes in core structures of the mammalian maternal brain such as the hypothalamus, it is critical to apply methodological approaches that allow investigating these small and functionally diverse structures in the human brain.
To examine hypothalamic anatomy, a highly specialized approach was developed based on a combination of neuroimaging and histological measures that allows investigating the hypothalamus in detail using magnetic resonance imaging (MRI), involving individually delineating five substructures of the hypothalamus (Makris et al., 2013).Albeit time-consuming and labor-intensive, this histologically-tested approach represents a reliable method for investigating the anatomy of hypothalamic substructures in humans in vivo.
In this study, we applied this approach to the longitudinal datasets of 107 women extracted from two prospective pre-conception cohort studies (Hoekzema et al., 2022(Hoekzema et al., , 2017)).Of these women, 50 became first-time mothers during the course of the project and 57 women remained nulliparous.For all 214 brain scans, the hypothalamus and its substructures were individually delineated on the baseline and follow-up anatomical brain scans.This approach allowed us to investigate for the first time whether becoming a mother alters the anatomy of the human hypothalamus.In addition, considering the primary role of hypothalamic substructures in the expression of maternal behaviors, we tested whether anatomical changes in the hypothalamus across pregnancy were associated with prenatal and postnatal aspects of maternal behavior.

Design and participants
The data for the current study were collected using a prospective cohort design.Data were extracted from two different pre-conception cohort studies, performed in Leiden (the Netherlands) and Barcelona (Spain).The research studies were evaluated and approved by the relevant ethics committees (the Ethics Review Board of the Leiden University Medical Center and the Comitè Ètic d'Investigació Clinica de l'Institut Municipal d'Assistència Sanitària respectively).In these prospective cohort studies, nulliparous women were examined and followed during the following period.Women with and without the intention to become pregnant in the near future took part in the initial session.At the start of the study, women with the intention to become pregnant were included in the PRG group, and women without this intention in the CTR group.The final group allocation depended on the transition from nulliparity to primiparity during the course of this study.In the Leiden study, seven women who wanted to get pregnant did not get pregnant during the study, so switched from the PRG to the CTR group.Two women originally in the CTR group got pregnant and switched to the PRG group.In the Barcelona study, two women switched from the PRG group to the CTR group, and two women switched the other way around.In both studies, anatomical brain scans were acquired before conception (the Pre session) and in the early postpartum period (the Post session) (78.02 ± 47.76 days after giving birth).The nulliparous control women were investigated at similar time intervals.In the Leiden sample, most women also took part in an experimental session in the 36th week of pregnancy.During this session, various questionnaires were administered that allowed us to investigate different aspects of prenatal maternal behavior.These questionnaires included the Maternal Antenatal Attachment Scale (MAAS, (Condon, 1993), n=22), the Prenatal Attachment Inventory (PAI, (Muller, 1993), n=22) and the Nesting Behaviour Questionnaire (Anderson and Rutherford, 2013), n=22).To measure postnatal maternal behavior, the Maternal Postnatal Attachment Scale (MPAS, (Condon and Corkindale, 1998), applied in both the Leiden and Barcelona sample, n=47) and the Postpartum Bonding Questionnaire (PBQ, (Brockington et al., 2006), applied only in the Leiden sample, n=24) were administered during the Post session.Postnatal depression was assessed by the Edinburgh Postnatal Depression Scale (EPDS, (Cox et al., 1987) in the PRG group only in both samples (Leiden: n=23, Barcelona: n=25).For more details on the approach and samples, please see (Hoekzema et al., 2022(Hoekzema et al., , 2017)).
For this project, longitudinal data were analyzed from 107 participants with complete datasets, extracted from the two described Note.Demographic information of the sample and between-group differences.PRG = nulliparous women who became pregnant between the sessions, CTR = nulliparous women who did not become pregnant between the sessions.M = mean, SD = standard deviation.
prospective cohort studies (Table 1).While the Leiden study currently involves a larger sample (n=40 PRG and n=40 CTR), not all women were included in the sample of the present study (PRG (n=25) and CTR (n=37) since, due to the time-consuming nature of the delineation approach, hypothalamic processing and parcellations had already commenced prior to completing the data collection of the entire sample.Demographic information for the sample is provided in  S2).Therefore, we included site (Barcelona or Leiden) and age at the PRE session as a covariate in the analyses.

MRI imaging
For both samples, MRI acquisitions were performed on 3-Tesla Philips MRI scanners.For the Leiden sample, high-resolution anatomical MRI brain scans were acquired using a T1-weighted gradient echo pulse sequence (TR = 9.8 ms, TE = 4.6 ms, 178 slices, FOV = 224 mm, flip angle = 8 • , voxel size = 0.875 × 0.875 × 1.20 mm 3 ).For the Barcelona sample, high-resolution anatomical MRI brain scans were acquired using a T1-weighted gradient echo pulse sequence (TR = 8.2 ms, TE = 3.7 ms, 180 slices, FOV = 240 mm, flip angle = 8 • , voxel size = 0.938 × 0.938 × 1.00 mm 3 ).Due to an unexpected technical problem, the radiofrequency (RF) head coil was replaced for some time with another head coil in the Barcelona sample, and 28 scans in total were acquired using the latter coil.There were no significant differences between the groups in the number of scans acquired with this head coil (X 2 = 4.21, p = 0.240).Nevertheless, in order to control for these differences in head coils, we included in all our models a covariate with three levels (1=Barcelona coil one, 2 = Barcelona coil two, 3 = Leiden coil).All MRI scans were visually checked for quality before data processing, but no images had to be excluded.

Data processing and segmentation of the hypothalamus
First, Freesurfer (https://surfer.nmr.mgh.harvard.edu/)was used to process the anatomical MRI brain images, using the recon-all function for automatic segmentation.We used the automatic segmentation of the basal forebrain structure as an indication for hypothalamus location and potential borders in each individual brain.Subsequently, the hypothalamic subregions were delineated entirely manually, using the Freeview tool (version 5.3.0,https://surfer.nmr.mgh.harvard.edu/).Data were anonymized to ensure rater blindness to the individuals' group allocation and session.
The hypothalamus was segmented using the volumetric parcellation method of Makris et al. (Makris et al., 2013), into five subregions: the anterior-superior (a-sHypo), anterior-inferior (a-iHypo), superior tuberal (supTub), inferior tuberal (infTub) and posterior hypothalamus (postHyp).One of the co-authors (S.M. Burke) was trained by the neuroanatomist who created this approach (Prof.Makris), who subsequently trained three raters to perform and check hypothalamic delineations.After an extensive training period of several months, the inter-rater reliability (IRR) between these raters was calculated for 10 subjects using Intra-Class Correlations (ICCs).This rendered excellent ICCs for the hypothalamic subregions (see Supplementary Table S1).After completing this training period, delineations of the hypothalamus were performed by two raters who subdivided the 214 anatomical brain scans.
In short, using this method, the hypothalamus was divided into the five measurable subregions based on visible anatomical landmarks (Fig. 1).The anterior hypothalamic parcellation units extend from the highest point of the anterior commissure to the anterior-most tip of the infundibulum stalk.The tuberal parcellation units extend from the anterior-most section containing the infundibulum stalk, to the coronal section just anterior to the mammillary bodies.The section where the mammillary bodies are visible were segmented as posterior hypothalamus.Further, the border between the inferior and superior regions was set at the superior-most level of the floor of the substantia innominata or the anterior and lateral perforated substance.
The superior border of the hypothalamus was then established by drawing the hypothalamic fissure in (both left and right) sagittal slices, cupping the thalamus.The rest of the segmentation was done on coronal slices.The anterior border was determined as the first slice on which the anterior commissure (AC) was clearly visible and continuous.The posterior border was the slice with the most posterior part of the mammillary bodies.Next, the hypothalamus was divided into anterior, tuberal and posterior parts.The anterior and tuberal hypothalamus were further divided into superior and inferior parts; the border was defined by drawing a horizontal line from the floor of the basal forebrain through the hypothalamus.The anterior hypothalamus embodied all slices where the AC was visible; the inferior part of the AC being the superior border.The lateral borders of the anterior section were determined by drawing a vertical line just lateral to the optic tracts/optic chiasm (both were excluded from the segmentation).The tuberal section began with the slice where the AC is no longer visible, included the slices on which the fornix was visible, and ended when the mammillary bodies came into view.The fornix was taken as the superior border where visible, and the fissure in the remaining ones.Lastly, the posterior section included all slices where the mammillary bodies are visible.The fissure was taken as the superior border in all the posterior slices.The lateral borders of the tuberal and posterior sections were established using the Freeview contour tool.The inferior border of the hypothalamus was formed by manually correcting the automatic segmentation by Freesurfer, where the cerebrospinal fluid was omitted.Lastly, the third ventricle was also segmented using Freesurfer's automatic segmentation and adapting it as necessary, which serves as a medial border in most slices.After completion of all manual delineations, volumes of all delineated substructures were extracted to be used for further analyses.

Hormone sampling
Estradiol (pg/ml) and Progesterone (pg/ml) identification.Every 4 weeks throughout their pregnancy and every 2 weeks near the end of pregnancy, women collected their first morning urine (pregnancy weeks 8, 12, 16, 20, 24, 28, 32, 36, 38, 40).Urine samples were saved in 10 ml polypropylene tubes and stored in a − 20 freezer.Hormones were sampled at the Technical University of Dresden, using a high-throughput liquid chromatography-tandem mass spectrometry (LC-MS/MS) assay with atmospheric pressure chemical ionization coupled with online solid phase extraction (Gao et al., 2015).The intra-and inter-assay Coefficients of Variability for all steroid hormones measured were less than 12%.
To correct for urine concentration, levels of creatinine (mg/dl) were determined with a kinetic colorimetric test (Jaffé method) which is retraceable to the isotope-dilution mass spectrometry (IDMS) reference method.Intra-and inter-assay Coefficients of Variability were below 3%.All measures were corrected for urine concentration in the sample by adjusting for creatinine levels.
Oxytocin identification.Oxytocin levels (pg/ml) were extracted from the same urine samples using radioimmunoassays in an RIAgnosis lab, for week 12 (n=20), week 28 (n=21) and week 36 (n=19).There were four missing samples for week 12.The values of the oxytocin were divided by the creatine level measured at that same timepoint (mg/dl) to control for variation in urine concentration.Intra-and inter-assay Coefficients of Variability were below 3%.We removed outliers more than two times the standard deviation from the mean (in week 12 and 36 one outlier and in week 28 two outliers).

Statistical analyses
All statistical analyses were done in RStudio (Rstudio, 2022).In order to account for multiple testing of intercorrelated variables, we used a correlation-adjusted Bonferroni correction threshold (Sankoh et al., 1997), http://www.quantitativeskills.com/sisa/calculations/b onfer.htm).For the volumetric analyses, we corrected for the five subregions and the total hypothalamic volume (six tests).For the correlation analyses, we corrected for the number of measures of maternal behavior per subregion (five total scores of questionnaires).Since the procedure to define the correlation-adjusted Bonferroni threshold asks for the averaged correlation between all variables, performing the adjustment over all subregions together would result in a correction for the correlations of interest (the correlations between maternal behavior and hypothalamic subregional volumes).Additionally, different subregions might be involved in specific aspects of maternal behavior.Therefore, we performed the correlation-adjusted Bonferroni correction per subregion separately.When total scores revealed significant associations, we accounted within the respective questionnaire for the number of subscales (specified below per questionnaire).
Volumetric change across pregnancy (general linear models (GLM)).Volumetric change was calculated by substracting the Post from the Pre values for each of the hypothalamic subregions as well as the total hypothalamic volume.To investigate whether pregnancy was associated with volumetric changes, we applied general linear models for each of the hypothalamic subregions (anterior-superior, anterior-inferior, superior tuberal, inferior tuberal, posterior hypothalamus) and total hypothalamus.In each model, we used volumetric change (Post session volume minus Pre session volume) per region as dependent variable and group (PRG or CTR) as between-factor.Additionally, in each model, age (at Pre session), RF coil (coded with three levels), study site (Barcelona or Leiden) and time interval between the Pre and Post session were included as covariates.The results were corrected for multiple comparisons using a correlation-adjusted Bonferroni correction (number of comparisons = six (five hypothalamic subregion and total hypothalamus), average inter-correlation coefficient = 0.20.Therefore, the significance threshold was set at p < 0.01.Subsequently, we performed one sample t-tests within each group to specify the direction of effects.
Additional variables and outliers.To test whether the observed changes were driven by changes in total gray matter (GM) volume, we also repeated the main models including total GM volume change (GM post minus pre session) as a covariate.The results were corrected for multiple comparisons using a correlation-adjusted Bonferroni correction as described above, resulting in a significance threshold of p < 0.01.To exclude the possibility that the results were biased by outliers, we repeated the linear model calculations without outliers, defined as ± two standard deviations per group and time point.This led to the exclusion of eight subjects for the anterior-inferior hypothalamus (four PRG, four CTR), seven subjects for the anterior-superior hypothalamus (three PRG, four CTR), six subjects for the inferior tuberal hypothalamus (four PRG, three CTR), five subjects for the superior tuberal hypothalamus (three PRG, two CTR), nine subjects for the posterior hypothalamus (four PRG, five CTR) and six subjects for the total hypothalamus (two PRG, four CTR).
Hypothalamic volume at PRE and POST session (mixed models).In order to verify that the results stay the same when using a repeated measures approach by modeling the time factor as a variable of interest, we calculated mixed models, by including the two time points (pre and post session), for each region and total hypothalamus, and including all above described covariates.The results were corrected for multiple comparisons using a correlation-adjusted Bonferroni correction (number of comparisons = 12 (five hypothalamic subregion and total hypothalamus), average inter-correlation coefficient = 0.25).Therefore, the significance threshold was set at p < 0.008.
Association of volumetric change with maternal behavior.To examine whether changes in hypothalamic anatomy were associated with aspects of maternal behavior, we performed Pearson correlation analyses only involving the hypothalamic subregions showing significant group differences.In case variables were not normally distributed, we performed Spearman correlation analyses, and we reported correlation coefficients as r (Pearson correlation) or rho (Spearman correlation).We determined correlations between hypothalamic volumetric change and prenatal maternal behavior (MAAS and PAI), nesting behavior (Nesting Behavior Questionnaire) and postnatal maternal behavior (MPAS and PBQ).First, correlation analyses were performed with the total scores of these questionnaires.The results were corrected for multiple comparisons using a correlation-adjusted Bonferroni correction (number of comparisons = five (total scores of questionnaires), average inter-correlation coefficient = 0.06).Therefore, the significance threshold was set at p < 0.01.Furthermore, when the total score for a scale rendered a significant result or were significant at nominal level (p < 0.05) for an association with hypothalamic changes across pregnancy, the subscales of that questionnaire with that specific subregion were also examined.Again, these analyses were corrected for multiple testing (for the number of tested subscales) with a correlation-adjusted Bonferroni approach.Therefore, the MAAS significance threshold was set at p < 0.038 (two subscales, average inter-correlation coefficient = 0.59) the PAI significance threshold was set at p < 0.029 (three subscales, average inter-correlation coefficient = 0.50), nesting behavior significance threshold was set at p < 0.017 (four subscales, average inter-correlation coefficient = 0.24),the MPAS significance threshold was set at p < 0.026 (three subscales, average inter-correlation coefficient = 0.40), and the PBQ significance threshold was set at p < 0.024 (four subscales, average inter-correlation coefficient = 0.46).Finally, to check for the influence of mood-associated factors, we ran correlation analyses between volumetric change measures and the EPDS scores in the PRG group.
Association of volumetric change with hormone levels during pregnancy.Normality distribution of the data was examined using Shapiro-Wilk tests.Correlation analyses were performed with the averaged hormone levels for each pregnancy trimester separately for estradiol, and progesterone as well as oxytocin levels at week 12, week 28 and week 36.In order to control for multiple comparisons, we used a correlationadjusted Bonferroni correction (number of comparisons = nine hormone measures, average inter-correlation coefficient = 0.09).Therefore, the significance threshold was set at p < 0.006.
Additional variables and outliers.To examine whether these group differences were also evident when correcting for changes in total grey matter, we repeated these models including total grey matter volume change as an additional covariate.These analyses rendered similar results (see Table 2).Furthermore, removing outliers from the main analyses did not affect the results (see Supplementary Table S3).Therefore, we chose to not exclude these outliers for follow-up analyses.
Hypothalamic volume at PRE and POST session (mixed models).The results of the models considering the hypothalamic volumes per time point were similar to the results of the linear models using volume change (see Supplementary Table S4), with significant changes in anterior-superior, superior tuberal and posterior hypothalamus.Additionally, we did not find any differences at the pre session between the groups in any of the hypothalamic subregions.Therefore, these results support that differences between groups are showing a volumetric decrease in specific hypothalamic subregions only in the PRG group (see Supplementary Table S5).

Associations of changes in hypothalamic subregions with maternal behavior
To investigate whether these changes in hypothalamic anatomy were related to maternal behavior, we tested for associations with prenatal (maternal-fetal attachment (PAI and MAAS) and nesting behavior) and postnatal (mother-infant bonding (MPAS and PBQ)) aspects of maternal behavior.All correlation results are summarized in Supplementary Table S6.

Table 2
General linear models without and with grey matter volume change as additional covariate.

Prenatal maternal behavior
The overall score of the PAI showed a significant negative association with the volume change in the superior tuberal hypothalamus (r = − 0.68, p = 0.0005; Fig. 3A) and the total hypothalamus (r = − 0.64, p = 0.001), indicating that stronger volume reductions within this structure related to higher maternal-fetal attachment.This association was mostly driven by the differentiation subscale of the PAI (superior tuberal: r = − 0.67, p = 0.001; total: r = − 0.59, p = 0.004, Fig. 3A), which quantifies the degree to which the mother differentiates herself from the fetus and regards her fetus as an individual.In addition, the anticipation subscale of the PAI, which quantifies the ability to fantasize about the baby, was significantly associated with the volumetric change in the superior tuberal hypothalamus (r = − 0.52, p = 0.013, Fig. 3A) and the total hypothalamus (r = − 0.57, p = 0.006, Fig. 3A).Additionally, the total MAAS score was associated at nominal significance level with volume change in the superior tuberal total hypothalamus (r = − 0.45 p = 0.037, Supplementary Figure S1).This association was driven by the quality of attachment subscale of the MAAS (rho = − 0.51, p = 0.016, Supplementary Figure S1), quantifying the quality of attachment between fetus and mother.

Nesting behavior
When testing for associations between hypothalamic volumetric change and nesting behavior, the total score of the nesting behavior questionnaire was nominally associated with the volume change in the anterior-superior hypothalamus (r = − 0.50, p = 0.019; Supplementary Figure S1) and superior tuberal hypothalamus (r = − 0.45, p = 0.034; Supplementary Figure S1).Subscale analyses demonstrated that the volumetric change in the superior tuberal hypothalamus was negatively associated with the space preparation: cleaning subscale (rho = − 0.55, p = 0.008, Supplementary Figure S1), revealing a stronger urge for cleaning the space as preparation for the arrival of the baby during the third trimester of pregnancy when undergoing stronger volume reductions within this substructure.No association of anterior-superior hypothalamus volume change and any of the subscales of the nesting behavior questionnaire survived the correlation-adjusted Bonferroni significance level.

Postnatal maternal behavior
The total score of PBQ showed a significant positive association with the total hypothalamic volume change (r = 0.52, p = 0.011, Fig. 3B).Additionally, the PBQ total score was positively associated with the superior tuberal hypothalamus (r = 0.48, p = 0.020, Supplementary Figure S1) at nominal level.These associations indicate that stronger volume reductions are related to better postpartum bonding as the subscale impaired bonding was driving the overall association (total: r = 0.49, p = 0.019, Fig. 3B; superior tuberal: r = 0.59, p = 0.003, Supplementary Figure S1).MPAS showed a negative association with volume change in the posterior (rho = − 0.31, p = 0.036, Supplementary Figure S1) and the total hypothalamus (rho = − 0.32, p = 0.028) at nominal level, indicating that stronger volume reductions are related to higher postnatal attachment.The subscale assessing the quality of maternal attachment to the newborn was driving the effect (posterior: rho = − 0.37, p = 0.011; total: rho = − 0.39, p = 0.007, Supplementary   Figure S1), suggesting that maternal attachment is stronger with higher volume reductions in the posterior and whole hypothalamus.

Postnatal depression
Finally, to check for the influence of mood-associated factors related to pregnancy, we ran correlation analyses between volumetric change measures and the EPDS scores in the PRG group.EPDS scores were not significantly associated with volume change in any of the hypothalamic subregions nor total hypothalamus (p > 0.137).

Association of volumetric change with hormone levels during pregnancy
To investigate whether the observed changes in hypothalamic volumes were related to hormone levels during pregnancy, we tested for associations with averaged estradiol and progesterone levels per pregnancy trimester and oxytocin levels measured in week 12, week 28 and week 36.
Averaged estradiol levels during the second trimester showed a significant negative association with volume changes in the posterior hypothalamus (rho = − 0.56, p = 0.006, Fig. 3C) and the total hypothalamus (rho = − 0.57, p = 0.006, Fig. 3C).These results indicate that an increased volume reduction in those regions was associated with higher estradiol levels during the second trimester.At nominal level, volume change in the total hypothalamus was negatively associated with estradiol levels during the first trimester (rho = − 0.47, p = 0.036, Supplementary Figure S2).
Further at nominal significance level, change in posterior hypothalamic volumes was positively associated with oxytocin values in week 12 (r = 0.45, p = 0.046, Supplementary Figure S2), indicating that lower oxytocin levels at week 12 are related to stronger volume reductions in this region.Finally, total hypothalamus volume change was positively associated with progesterone levels at second trimester (rho = 0.43, p = 0.043, Supplementary Figure S2), pointing to lower progesterone levels in this trimester in relation to increased volume reductions in the hypothalamus.
For an overview over correlations between all variables of interest a full-zero correlation figure is in the Supplementary Figure S3.

Discussion
We investigated whether pregnancy induces volumetric changes in hypothalamic subregions by applying an advanced delineation approach to longitudinal brain MRI scans of 107 women.Our results showed that women who became mothers during the course of the study underwent a decrease in volume from the Pre to the Post session compared to control women, who remained nulliparous, in the anterior-superior, the superior tuberal, and the posterior hypothalamus as well as the whole hypothalamus.These results stayed the same in a mixed model approach, and they were not affected by GM volume change nor outliers, which further strengthens our findings.Additionally, we showed that these changes in hypothalamic anatomy were related to specific maternal behaviors.Prenatal maternal-fetal attachment, as measured by the PAI, was negatively associated with the volume change in the superior tuberal hypothalamus.More specifically, the volumetric change in the superior tuberal hypothalamus was negatively related to the Differentiation and Anticipation subscales of the PAI, indicating that stronger volume decreases were associated with more imagination and fantasy about the baby and a stronger differentiation of the fetus as an individual.Furthermore, the volumetric decrease in the anterior-superior and superior tuberal hypothalamus was associated with the degree of nesting behavior during the third trimester of pregnancy, specifically with the space preparation and cleaning subscale.These results indicate that volumetric losses in these hypothalamic subregions are associated with increased maternal-fetal attachment and an increased exhibition of nesting behaviors during pregnancy.Further, volume decreases in the superior tuberal and total hypothalamus were associated with increased postnatal bonding of the mother to the baby.Finally, maternal postnatal attachment, assessed with the MPAS, was negatively associated with the volume change in the posterior and total hypothalamus, indicating that stronger volume decreases were related to increased quality of attachment.Additionally, we were able to show that increased volumetric change during pregnancy in the posterior and total hypothalamus was related to higher estradiol levels during the second trimester of pregnancy.Overall, our data show for the first time that becoming a mother renders changes in substructures of the hypothalamus and provide evidence for an association with pregnancy hormones and aspects of maternal behavior.These findings suggest that, in accordance with findings in other mammals, the hypothalamus represents a core structure of the maternal brain in humans and changes in its substructures may play a role in the induction of maternal behavior.
The hypothalamus is one of the oldest brain regions and its anatomy is highly conserved throughout vertebrate species (Lemaire et al., 2021;Saper and Lowell, 2014;Xie and Dorsky, 2017).It is a functionally diverse region consisting of multiple nuclei, which are involved in the coordination of endocrine, emotional and somatic processes in order to sustain a healthy physiological equilibrium (Makris et al., 2013).Specific hypothalamic subnuclei, including the mPOA and PVN, have consistently been associated with reproduction and maternal behavior.We know, mostly from rodent studies, that these hypothalamic nuclei are remarkably plastic during pregnancy and the postpartum period and undergo various neurochemical, neuroendocrine, morphological, gene expression, activational and functional changes (Leuner and Sabihi, 2016).For example, adult neurogenesis driven by pregnancy, parenting and pup exposure has been shown in hypothalamic nuclei (Leuner and Sabihi, 2016).In addition, hormonal priming of the mPOA during late pregnancy upregulates hormone receptor expression and receptor expression and oxytocin release are increased in the PVN during the peripartum period (Hillerer et al., 2014).Interestingly, a neuroimaging study in mice has shown grey matter volumetric increases in several hypothalamic subregions, including the mPOA, during late pregnancy and the early postpartum period, which reduced to pre-pregnancy values again after weaning (Barrière et al., 2021).Furthermore, various research findings have shown that reproduction-related changes in the hypothalamus are essential for initiating maternal behavior.For instance, hormonal priming of the mPOA results in pup approach behavior, volumetric changes in various hypothalamic substructures relate to the performance of maternal behavior (Barrière et al., 2021) and dysfunction of the mPOA leads to reduced or disrupted maternal behavior (Gómora-Arrati et al., 2016;Kohl et al., 2018;Numan and Stolzenberg, 2009).
Overall, our results are in line with these findings, since we also observe structural changes in specific subregions of the hypothalamus, namely the anterior-superior, the superior tuberal, and the posterior hypothalamus.These subregions, as defined by (Makris et al., 2013), encompass some of the most essential parts of the hypothalamus involved in rodent maternal behaviors such as the mPOA (anterior-superior hypothalamus), and the PVN (anterior-superior and superior tuberal hypothalamus).We also showed that higher levels of estradiol during the second trimester were related to larger decreases in the posterior hypothalamic volume.Similarly, estradiol in the second and third trimester have previously been shown to relate to volumetric decreases in different areas of the brain during pregnancy (Hoekzema et al., 2022).These results suggest that pregnancy hormones, and mainly estradiol, are supporting brain changes that happen during pregnancy.
However, it should be noted that, despite the unprecedented detail offered by the applied parcellation approach, this method does not approach the degree of information produced by in vitro animal studies and individual nuclei within the hypothalamus cannot be investigated in further detail.Therefore, at this moment, no conclusions can be drawn with respect to the specific role of hypothalamic nuclei contained within these substructures.In addition, while our findings support that changes K. Spalek et al. in hypothalamic substructures may be conserved across species, these effects do not necessarily represent similar processes.In fact, while we observed volume reductions in hypothalamic substructures, volume increases were observed in mice in various hypothalamic nuclei.Unfortunately, studies involving MRI do not allow investigating the neural processes underlying these macroscopic changes in detail.Future studies acquiring post-mortem measurements in deceased pregnant women might shed more light on the way specific hypothalamic nuclei are affected by pregnancy in humans.
In agreement with our findings, some early studies in humans have already suggested that the hypothalamus represents an important part of the maternal brain in humans (for an overview see (Kim, 2016)).Hypothalamic and medial prefrontal activity has been found to increase in mothers in reaction to hearing their own baby during the early postpartum period (Kim et al., 2011;Swain et al., 2011).In addition, using a whole-brain structural MRI approach, an increase in grey matter volume across the postpartum volume was found in the hypothalamus and other brain regions central for maternal motivation and reward processing (Kim et al., 2010).Accordingly, our data also point to volumetric changes in the hypothalamus within the peripartum period, although we found volume reductions when comparing pre-pregnancy and post-pregnancy scans within specific substructures of the hypothalamus while Kim et al. (2010) observed volume increases in the hypothalamus across the postpartum period.Together, these findings might indicate that hypothalamic substructures undergo volume reductions from preto post-pregnancy, which partially reverse during the postpartum period.Future research capturing the maternal brain at different stages in this transition might help to further elucidate the timing of these changes in hypothalamic volumes.Taken together, in accordance with findings from animal studies, research findings point to a pronounced plasticity of the hypothalamus in the peripartum period in humans.
In addition, similar to what has been found in rodent studies, we also showed that the volumetric changes in hypothalamic subregions were associated with specific aspects of maternal behavior, including nesting behavior and maternal-fetal attachment.Interestingly, changing activity in the mPOA in rats has been found to influence different types of maternal behavior, including the quality of nest building and the amount of pup retrieval (Gómora-Arrati et al., 2016).Additionally, acute deactivation of the PVN decreased the occurrence of maternal nursing behavior (Bayerl et al., 2016).Although our MRI approach does not allow the investigation of these specific hypothalamic nuclei, it is interesting to note that we showed a correlation between nesting behavior and the anterior-superior and superior tuberal hypothalamus as well as between maternal-fetal attachment and the superior tuberal hypothalamus, which include the mPOA and PVN respectively (Makris et al., 2013).Furthermore, we observed a correlation between the superior tuberal, posterior and total hypothalamus with postnatal bonding and attachment.The observed correlations between hypothalamic volume changes and maternal behavior suggest that the hypothalamus and its substructures may also play a role in stimulating maternal behavior in humans.
Interestingly, associations were observed with mainly prenatal but also some postnatal aspects of maternal behavior, suggesting that these hypothalamic alterations may primarily be involved in the early induction of maternal behaviors during pregnancy rather than the maintenance or postnatal expression of such behaviors.Previous findings in humans have shown evidence for the prenatal onset of aspects of maternal behavior such as nesting behavior (Anderson and Rutherford, 2013) and changes in maternal responsiveness, including an increase in feelings of nurturance and attachment (Bridges, 2015;Fleming et al., 1997;Saltzman and Maestripieri, 2011).Prenatal maternal behaviors such as nesting behaviors and changes in prenatal maternal responsiveness are also known to occur in numerous other animal species, and have been linked to pregnancy hormones (Bridges, 2015;Saltzman and Maestripieri, 2011).For instance, the administration of pregnancy-mimicking hormone regimens in rodents and primates has been shown to increase the frequency of bar pressing for stimuli related to the young and to induce a preference for young animals over food (Champagne et al., 2004;Li and Fleming, 2003;Pryce et al., 1993).In addition, changes in hormonal receptor expression in hypothalamic substructures such as the mPOA are known to play an important role in the onset and early expression of maternal behavior in rodents (Kohl and Dulac, 2018).Our findings provide the first indication that evolution of maternal behaviors in humans, partially forming already prenatally, is related to volumetric changes in hypothalamic substructures during pregnancy.Although our results do not indicate when exactly these brain changes take place, the correlation with second trimester estradiol suggest that the induction of pregnancy-related plasticity in the brain already starts during pregnancy.This might enable the start of prenatal maternal behavior during pregnancy and might prepare women for postnatal maternal care.
Finally, there are some limitations to be considered.We cannot make a comparison between the accuracy of the manual delineation method applied in this manuscript and the recently introduced automated method by Billot et al., (2020), as implemented in Freesurfer, since this method was not yet available at the time of our manual delineations.Although the manual delineation as we performed in this study could be considered as a golden standard, it is very time consuming.Billot et al. (2020) showed comparable results between the much faster automatic segmentation and their manual delineation.A comparison between the two methods was out of the scope of this article, but would be of interest to investigate the reproducibility of our results.Second, we believe that our results point to region-specific effects, given that we showed group differences in three subregions of the hypothalamus (anterior-superior, superior tuberal and posterior) and not in all of the subregions.This is supported by findings in non-human animals showing that specific subregions (respectively subnuclei) are involved in pregnancy-related and maternal behaviour.However, we cannot rule out that the subregion specificity is driven by other factors, such as measurement error.Nevertheless, we can exclude that the non-significant subregions, which are the smallest ones, had an increased measurement variation (standard deviation/mean).Third, biophysical factors such as BMI, gestational weight gain and stressors could potentially have influenced our results.Since these data were either not available or only available for one of the two samples, we were not able to account for those factors.Nevertheless, in a previous publication (Hoekzema et al., 2022), using one of the two samples, we were able to show that stress did not represent a major factor contributing to changes in neural architecture and neural network organization related to pregnancy.Finally, it should be pointed out that the behavioral measures used for assessing maternal behavior were available only for a subsample of subjects (and in several cases only for one of the two sites).Therefore, these associations need to be interpreted with caution.Additionally, there were differences between the two sites in several variables of interest.Although we corrected for research site as a covariate, these differences may have influenced the results.
To conclude, our data show that pregnancy leads to anatomical changes in subregions of the hypothalamus, including substructures containing core regions of the maternal brain circuit as identified in animal studies.In addition, associations were observed between these anatomical changes and specific aspects of maternal behavior such as nesting behaviors and maternal-fetal attachment.These findings suggest that the hypothalamus represents a core structure of the human maternal brain that may contribute to the development of maternal behavior in humans, supporting the conservation of important aspects of the maternal brain circuitry and its role in maternal behavior across species.

Fig. 1 .
Fig. 1.Overview figure of the five hypothalamic subregions in one of the participants in the sagittal (left top), axial (middle and right top) and coronal (bottom) view.Green = anterior-superior hypothalamus, blue = anterior-inferior hypothalamus, orange = superior tuberal hypothalamus, yellow = inferior tuberal hypothalamus and red = posterior hypothalamus.

Fig. 2 .
Fig. 2. Volumes of hypothalamic subregions that significantly changed differently in pregnant and control groups.Individual measures of hypothalamic volume at both sessions are shown in mm 3 , the mean and standard deviation per group are separately plotted for the two timepoints (Pre and Post) and the two groups (PRG = pregnant (n=50)(red), CTR = control (n=57)(blue)).The Pre timepoint was before conception, the Post time point during the early postpartum period.a-sHypo = anterior-superior hypothalamus; supTub = superior tuberal hypothalamus; postHyp = posterior hypothalamus.** p < 0.01 (Statistically significant), as determined with a correlation-adjusted Bonferroni approach.

Table 3
Post-hoc differences in hypothalamic volume change between pregnant and control participants.
*Note.The mean and standard deviation (SD) of volumetric change in mm 3 per hypothalamic subregion, and total hypothalamic volume are reported per group (PRG n=50, CTR n=57).Results of the t-tests are shown, including the t-value (t) and p-value (p).* p < 0.05 statistically significant effects.