The Association Between Maternal Psychological Health and Human Milk Oligosaccharide Composition

Abstract

Background and Objective:

Human milk oligosaccharides (HMOs) are carbohydrates abundant in human breast milk. Their composition varies widely among women, and prior research has identified numerous factors contributing to this variation. However, the relationship between maternal psychological health and HMO levels is currently unknown. Thus, our objective was to identify whether maternal stress, anxiety, or depressive symptoms are associated with HMOs.

Methods:

Data originated from 926 lactating individuals from the UC San Diego Human Milk Biorepository. Nineteen prevalent HMOs were assayed using high-performance liquid chromatography. Participants self-reported measures of the Edinburgh Postnatal Depression Scale (n = 495), State-Trait Anxiety Inventory S-Scale (n = 486), and/or Perceived Stress Scale (n = 493) within 60 days of their milk collection; their results were categorized using standard screening cutoffs. HMOs were assessed individually and grouped by principal component analysis (PCA), and associations with maternal psychological symptoms were analyzed using multivariable linear regression adjusted for covariates.

Results:

After Bonferroni correction (p < 0.002), the following HMOs significantly varied with maternal psychological distress in multivariate analysis: lacto-N-fucopentaose III (LNFP III) and lacto-N-hexaose (LNH) among Secretors with depressive symptoms and difucosyllactose (DFLac), LNFP III, and disialyl-LNH (DSLNH) among Secretors with stress. In PCA, depressive symptoms and stress were associated with one principal component among Secretors. No HMOs varied with anxiety symptoms.

Conclusions:

Several HMOs varied with maternal depressive symptoms and stress, suggesting a relationship between maternal psychological health and breast milk composition. Additional studies are needed to determine the impact of this variation on infant health.

Introduction

Human milk oligosaccharides (HMOs) are abundant glycans that comprise the third largest solid component of breast milk.^1,2 Although there are more than 200 types of HMOs, more than 90% of mature milk’s total HMO content consists of only 19 structures: 2′-fucosyllactose (2′FL), 3-fucosyllactose (3FL), 3′-sialyllactose (3′SL), 6′-sialyllactose (6′SL), difucosyllactose (DFLac), lacto-N-tetraose (LNT), lacto-N-neotetraose (LNnT), lacto-N-fucopentaose I, II, and III (LNFP I/II/III), sialyl-LNT (LST) b, LSTc, difucosyl-LNT (DFLNT), disialyl-LNT (DSLNT), lacto-N-hexaose (LNH), fucosyl-LNH (FLNH), difucosyl-LNH (DFLNH), fucosyl-disialyl-LNH (FDSLNH), and disialyl-LNH (DSLNH).³

Prior research has identified numerous ways by which HMOs promote infant health. Some examples include functioning as prebiotics to develop a healthy gut microbiome, exhibiting immunoprotective effects, such as by hindering pathogen binding to host tissue, and altering gene expression.^1,4–8 HMOs have also been associated with healthy neurodevelopment and increased executive functions in toddlers.⁹

Many groups have demonstrated variation in HMO composition among women. One contributor to the variation is the activity of the galactoside 2-alpha-L-fucosyltransferase 2 (FUT2) enzyme, which synthesizes α1,2-fucosylated HMOs, such as 2′FL and LNFP I.^1,2,10 This enzyme is encoded by the Secretor gene, Se; thus, women with this gene are coined Secretors, while non-Secretors have lower to absent FUT2 activity levels and consequently diminished α1,2-fucosylated HMOs. Also often included in this designation is the presence or absence of the Le gene, which encodes the enzyme galactosidase 3/4-L-fucosyltransferase (FUT3) and synthesizes α1,4-fucosylated structures, such as LNT and LNFP II. Thus, four phenotypes are generated based on the activities of FUT2 and FUT3: Se+Le+, Se+Le−, Se-Le+, and Se−Le−.

In addition to SeLe genotype, previous research has examined many factors that may contribute to HMO variability, including lactation stage, parity, maternal age, diet, weight, body mass index (BMI), mode of delivery, and geographic region.^10–19 Additional factors, such as seasonality, gestational diabetes mellitus, maternal allergy, infant sex and size, and socioeconomic status, have also been studied.^{11,12,20–23}

Prior studies have also identified factors affecting other components of breast milk, including a positive relationship between maternal symptoms of depression and anxiety and levels of sodium and transforming growth factor beta (TGF-β).^24–26 Mood-mediated effects on mammary gland permeability and immune system function are hypothesized to contribute to these associations. Given HMOs’ excretion through the mammary gland and their role in the immune system, it is plausible that maternal psychological distress may also impact HMO composition. However, to our knowledge, no groups have examined this potential relationship. Therefore, this study’s primary aim was to determine whether maternal symptoms of depression, anxiety, or stress were significantly associated with HMO composition in lactating women.

Materials and Methods

Subjects and sample collection

Participants consented to provide milk samples to the UC San Diego Human Milk Biorepository (HMB) between 2014 and 2022. Eligible volunteers were women 18 years and older residing in the United States or Canada. Milk samples were collected according to HMB protocols.²⁷ Participants provided written informed consent for the use of their milk for research and completed an interview about their sociodemographics, maternal and infant health history, lifestyle, and exposures to medications and recreational substances. Medication history was captured day-by-day for the 14 days prior to sample collection by specific medication, dose, and indication. Participants were asked to collect at least 50 mL of milk up to 24 hours before their scheduled interview. Samples were collected off-site in participants’ homes or at the HMB research center. Home collections were refrigerated at 0–4°C for up to 24 hours and shipped overnight on ice to the HMB. All milk was aliquoted and stored at the HMB at −80°C.²⁸ The study was approved by the institutional review board at UC San Diego under project #130658.

Assessment of maternal psychological state

After submitting milk samples and completing an interview, participants were emailed questionnaires, which they completed online to capture their psychological state. If participants had ≥1 missing response to the Edinburgh Postnatal Depression Scale (EPDS), State-Trait Anxiety Inventory S-Scale (STAI-S), or Perceived Stress Scale (PSS), their responses were not scored for that survey. Data were only included in the analysis if surveys were completed within 60 days of milk collection.

Depression

Depressive symptoms were assessed using the EPDS. This scale is useful in identifying depression symptoms in women during pregnancy and up to one year postpartum.²⁹ It consists of 10 screening questions, and responses range from 0 to 3. Participants’ responses were summed to determine overall risk for depression. While this is not diagnostic of depression, scores ≥12 are highly suggestive of symptoms of depression.³⁰ Thus, we defined high risk for depression as a score ≥12.

Anxiety

STAI was used to measure symptoms of anxiety. This scale has two components: the State Anxiety Scale (S-Anxiety) measures current anxiety symptoms while the Trait Anxiety Scale (T-Anxiety) measures stable anxiety predispositions.²⁹ Each scale consists of 20 questions, with responses ranging from 1 to 4. Participants’ responses were summed to determine their overall risk for anxiety. Because our focus was on women currently experiencing symptoms of anxiety, high risk for anxiety was defined as a score of ≥45 on the STAI S-Scale (STAI-S) only.³¹

Stress

PSS is a 10-question survey that assesses the frequency of perceived stress over the last month.²⁹ Responses to these questions ranged from 0 to 4, and these values were summed to determine an overall risk for stress. A PSS score ≥20 was defined as being high risk for stress.³⁰

Covariates

Maternal factors of interest included maternal age (in years; continuous) and BMI (in kilogram per square meter; continuous) at the time of sample collection, race (White, Black, Asian, American Indian, Native Hawaiian/Pacific Islander), exclusive breastfeeding status (yes/no), and current use of antidepressants and/or anxiolytics (yes/no). Infant factors of interest included infant age at the time of sample collection (in years; continuous) and sex (male/female).

HMO quantification

As previously described, HMOs were quantified using high-performance liquid chromatography (HPLC).^3,28 Briefly, the oligosaccharide raffinose was added as an internal standard prior to sample processing. After HMOs were extracted by high-throughput solid-phase extraction, they were labeled with 2-aminobenzamide and subsequently separated by HPLC with fluorescent detection (HPLC-FL). Peak annotation was based on standard retention times. The absolute concentrations of the following 19 abundant HMOs were calculated based on standard response curves and corrected for internal standard recovery: 2′FL, 3FL, 3′SL, 6’SL, DFLac, DFLNH, DFLNT, DSLNH, DSLNT, FDSLNH, FLNH, LNFP I, LNFP II, LNFP III, LNH, LNT, LNnT, LSTb, and LSTc. Secretor status was defined by the quantity of 2′FL or LNFP I, with non-Secretors having near-absent levels of both. For each participant, the concentrations of all HMOs were summed together to estimate total HMO concentration and the relative abundance of each HMO (Supplementary Fig. S1).

Statistical analysis

Statistical analysis was performed using SAS version 9.4. To evaluate for associations between symptoms of depression, anxiety, or stress and percent concentration of each HMO, univariate analysis was performed with a two-sided p-value significance cutoff of <0.002 (Bonferroni corrected for multiple comparisons). Those that remained significant were further analyzed using linear regression models adjusted for infant sex, exclusive breastfeeding, infant collection age, maternal age, race, depression and anxiety medications, and maternal BMI. Separate models were performed for each significant HMO with depressive, anxiety and stress symptoms. Risk ratios and 95% confidence intervals (CI) were reported. In a secondary analysis, principal component analysis (PCA) was performed as a data reduction technique stratified by Secretor status. The optimal number of components was based on eigenvalues >1. Multivariable linear models were repeated with each PC, and all models were stratified by Secretor status and adjusted for the aforementioned covariates.

Results

Basic information on maternal and infant participants

Of the 926 HMO profiles assayed from the HMB, 562 had complete responses to the EPDS, STAI-S, and/or PSS surveys (Fig. 1). Of the 562 samples, 6 were excluded due to survey completion >60 days after milk sample collection, and 52 were excluded because they were serial samples from the same participant (in these cases, the first enrolled sample was analyzed). The final cohort consisted of 504 women with complete HMO assays and at least one complete mood survey; of these, 495 had EPDS measures, 486 had STAI-S measures, and 493 had PSS measures.

FIG. 1.

Flowchart of the study population. EPDS, Edinburgh Postnatal Depression Scale; HMB, Human Milk Biorepository; HMO, human milk oligosaccharide; PSS, Perceived Stress Scale; STAI-S, State-Trait Anxiety Inventory S-Scale.

Characteristics of the maternal and infant participants are shown in Table 1. From the sample, 11.5% had elevated depressive symptoms, 14.4% had elevated anxiety symptoms, and 25.2% had elevated stress. There was significant overlap (n = 35) between those with elevated depressive and anxiety symptoms; 61% of those with depressive symptoms above the cutoff also had anxiety symptoms above the cutoff, and 50% of those with symptoms of anxiety also had symptoms of depression above the cutoff. Nearly three-quarters of the women were Secretors, and most mothers identified as White (91.7%). The average infant age was 6 months at the time of milk collection.

Table 1.

Selected Characteristics of Maternal and Infant Participants

Characteristics	n (%)
	All	Depression^a		Anxiety^b		Stress^c
	n = 504	Yes (n = 57)	No (n = 438)	Yes (n = 70)	No (n = 416)	Yes (n = 124)	No (n = 369)
Secretor	375 (74.4)	41 (71.9)	327 (74.7)	53 (75.7)	309 (74.3)	94 (75.8)	274 (74.3)
Maternal race^d
White	462 (91.7)	52 (91.2)	403 (92.0)	58 (82.9)	388 (93.3)	109 (87.9)	343 (93.0)
Black	8 (0.02)	1 (0.02)	6 (0.01)	3 (0.04)	5 (0.01)	4 (0.03)	4 (0.01)
Asian	19 (0.04)	1 (0.02)	18 (0.04)	3 (0.04)	14 (0.03)	5 (0.04)	14 (0.04)
American Indian	12 (0.02)	1 (0.02)	11 (0.03)	4 (0.06)	8 (0.02)	4 (0.03)	8 (0.02)
Native Hawaiian/Pacific Islander	2 (0.00)	2 (0.04)	0 (0.00)	2 (0.03)	0 (0.00)	2 (0.02)	0 (0.00)
Taking medications for anxiety or depression	166 (32.9)	34 (59.6)	128 (29.2)	41 (58.6)	117 (28.1)	68 (54.8)	95 (25.7)
Exclusive breastfeeding at the time of the sample	418 (82.9)	45 (78.9)	365 (83.3)	57 (81.4)	347 (83.4)	103 (83.1)	306 (82.9)
Infant sex male	269 (53.4)	34 (59.6)	231 (52.7)	38 (54.3)	221 (53.1)	69 (55.6)	197 (53.4)
Infant age (years) (mean, SD)	0.6 (0.5)	0.7 (0.4)	0.6 (0.4)	0.7 (0.7)	0.6 (0.5)	0.7 (0.5)	0.6 (0.5)
Maternal BMI^e (mean, SD)	26.2 (5.9)	28.0 (8.4)	25.9 (5.4)	27.4 (7.6)	26.0 (5.4)	27.8 (7.3)	25.8 (5.3)
Maternal age (mean, SD)	33.2 (4.3)	32.8 (4.0)	33.2 (4.3)	33.7 (4.1)	33.2 (4.3)	33.4 (4.1)	33.2 (4.3)

Depression defined by Edinburgh Postnatal Depression Scale (EPDS) ≥12. Complete EPDS responses missing for 10 participants.

Anxiety defined as by state State-Trait Anxiety Index S-Scale (STAI-S) score ≥45. Complete STAI-S responses missing for 18 participants.

Stress defined by Perceived Stress Scale (PSS) ≥20. Complete PSS responses missing for 11 participants.

Race missing for one participant.

BMI = bodyweight in kilograms divided by height in meters squared.

BMI, body mass index; SD, standard deviation.

Maternal sociodemographic and health characteristics varied with symptoms of depression, anxiety, and stress. Women with symptoms of depression, anxiety, and stress were more likely to be taking medications for depression or anxiety and more likely to have a higher BMI. Women with symptoms of depression were less likely to be exclusively breastfeeding their infant at the time of the sample collection.

Looking at overall HMO profiles, the range of total HMO concentration was from 4,316 to 23,960 nmol/mL with an average of 13,343.4 nmol/mL (standard deviation [SD] 3651.4). As expected, non-Secretors had lower than average total HMO levels (mean 9,869.8 nmol/mL, SD 2396.4) in comparison with Secretors (mean 14,563.3 nmol/mL, SD 3184.5). On average, secretors also had more diverse HMO profiles than non-Secretors (4.75 ± 1.43 versus 4.12 ± 1.49) (data not shown).

Clustering of HMOs

The concentrations of HMOs (nmol/mL) were assessed for clustering using Pearson correlation coefficients. This analysis was performed separately for Secretors and non-Secretors due to the groups’ differing activities of FUT2. In Secretors, among FUT2-dependent HMOs (2′FL, LNFP I, and DFLac), 2′FL and LNFP I had the only significant positive correlation (ρ = +0.52), while DFLac was only weakly correlated with 2′FL (ρ = +0.20) and LNFP I (ρ = +0.03) (Supplementary Fig. S2). For the other HMOs, many structures were positively clustered with ρ ≥ +0.50, and significant negative correlations existed between 3FL and LNT (ρ = −0.63) and LNFP III (ρ = −0.55).

Similar to Secretors, many HMOs were positively correlated in non-Secretors (Supplementary Fig. S2). 3FL was also negatively correlated with LNT (ρ = −0.71) and LNFP III (ρ = −0.74) in non-Secretors. Non-Secretors demonstrated additional negative correlations between HMOs, including 3FL and LNnT (ρ = −0.61), 3FL and LNFP II (ρ = −0.57), 3FL and DSLNT (ρ = −0.62), and LNT and DFLNH (ρ = −0.55), which were not seen in Secretors.

Depressive symptoms

In Secretors, 8 HMOs (3FL, LNT, LNFP III, LSTc, LNH, DSLNT, FDSLNH, and DSLNH) were significantly different between those with depressive symptoms and those without (p < 0.05) (Table 2). After applying Bonferroni correction, three remained significant (LNFP III, LSTc, and LNH). Multivariable linear regression was performed for these HMOs. LNFP III (adjusted risk ratio [aRR] −0.1, 95% CI −0.1 to −0.1) and LNH (aRR −0.1, 95% CI −0.2 to −0.0) were negatively associated with maternal depressive symptoms in Secretors.

Table 2.

The Impact of Depression on Human Milk Oligosaccharide Profiles, Stratified by Secretor Status

	Secretor (n = 364)				Non-Secretor (n = 131)
% HMO	Depression^a (n = 40)	No depression^b (n = 324)	p-Value	aRR 95% CI^c	Depression^a (n = 17)	No depression^b (n = 114)	p-Value	aRR 95% CI^c
2′FL	41.2	38.2	0.15		1.1	0.7	0.56
3FL	15.7	10.6	0.02		41.4	27.6	0.07
LNnT	4.5	4.5	0.91		1.8	3.8	0.008
3′SL	4.0	5.1	0.07		3.4	4.6	0.10
DFLac	2.6	3.5	0.06		0.8	1.5	0.004
6′SL	3.4	3.5	0.82		7.2	5.2	0.13
LNT	3.4	4.9	0.04		6.6	9.7	0.26
LNFP I	4.7	5.4	0.28		1.7	1.9	0.34
LNFP II	7.7	8.6	0.11		21.9	23.8	0.52
LNFP III	0.2	0.4	0.001 ^d	−0.1 −0.1 to −0.1^e	0.5	0.6	0.33
LSTb	0.6	0.7	0.11		1.4	1.5	0.57
LSTc	0.3	0.5	0.0008 ^d	−0.1 −0.3 to 0.0	0.2	0.4	<0.001 ^d	−0.2 −0.4 to 0.0
DFLNT	7.5	8.6	0.13		6.1	6.0	0.93
LNH	0.3	0.5	<0.0001 ^d	−0.1 −0.2 to −0.0^e	0.7	0.8	0.61
DSLNT	1.3	1.9	0.02		2.3	3.0	0.33
FLNH	0.5	0.6	0.09		0.5	1.3	0.0004 ^d	−0.7 −1.3 to 0.0
DFLNH	0.4	0.4	0.60		0.8	0.6	0.31
FDSLNH	1.2	1.5	0.03		3.1	4.2	0.12
DSLNH	0.3	0.5	0.003		0.4	0.8	0.003

Bold denotes significant p < 0.05.

Model adjusted for infant sex, exclusive breastfeeding, infant collection age, maternal age, race, depression and anxiety medications, and maternal body mass index.

Depression defined by Edinburgh Postnatal Depression Scale (EPDS) ≥12.

No depression defined by EPDS <12.

Confidence intervals rounded to nearest one decimal point.

Indicates significant p-value (<0.002) after Bonferroni correction.

Statistical significance based off actual numbers; rounded numbers displayed.

2′FL, 2′-fucosyllactose; 3FL, 3-fucosyllactose; 3′SL, 3′-sialyllactose; 6′SL, 6′-sialyllactose; aRR, adjusted risk ratio; CI, confidence interval; DFLac, difucosyllactose; DFLNH, difucosyl-LNH; DFLNT, difucosyl-LNT; DSLNH, disialyl-LNH; DSLNT, disialyl-LNT; FDSLNH, fucosyl-disialyl-LNH; FLNH, fucosyl-LNH; HMO, human milk oligosaccharide; LNFP, lacto-N-fucopentaose; LNH, lacto-N-hexaose; LNT, lacto-N-tetraose; LNnT, lacto-N-neotetraose; LST, sialyl-LNT.

In non-Secretors, 5 HMOs (LNnT, DFLac, LSTc, FLNH, and DSLNH) varied between those with depressive symptoms and those without (p < 0.05). LSTc and FLNH remained significant after applying Bonferroni correction; however, none were significant in multivariate analysis.

Anxiety symptoms

In Secretors, DFLac, LNFP III, and DSLNT varied between those with anxiety symptoms and those without in univariate analysis (p < 0.05) (Table 3). None remained significant after applying Bonferroni correction. In non-Secretors, LSTc, FLNH, and DSLNH initially varied in univariate analysis (p < 0.05) but not after adjusting for multiple comparisons. Thus, maternal anxiety symptoms were not associated with the concentrations of any of our 19 studied HMOs.

Table 3.

The Impact of Anxiety on Human Milk Oligosaccharide Profiles, Stratified by Secretor Status

	Secretor (n = 358)			Non-Secretor (n = 128)
% HMO	Anxiety^a (n = 52)	No anxiety^b (n = 306)	p-Value	Anxiety^a (n = 18)	No anxiety^b (n = 110)	p-Value
2′FL	39.7	38.3	0.47	1.1	0.7	0.54
3FL	13.6	10.6	0.12	37.1	28.7	0.21
LNnT	4.6	4.4	0.64	2.4	3.7	0.17
3′SL	4.7	5.0	0.65	4.4	4.4	0.99
DFLac	2.6	3.6	0.04	1.1	1.5	0.11
6′SL	3.7	3.5	0.56	5.6	6.8	0.51
LNT	4.3	4.8	0.36	7.4	9.3	0.46
LNFP I	5.1	5.3	0.63	1.8	1.8	0.99
LNFP II	8.2	8.7	0.42	22.4	23.9	0.60
LNFP III	0.3	0.4	0.003	0.6	0.6	0.65
LSTb	0.6	0.7	0.35	1.4	1.4	0.82
LSTc	0.4	0.5	0.38	0.3	0.4	0.008
DFLNT	7.6	8.7	0.09	5.4	6.2	0.57
LNH	0.4	0.5	0.27	0.8	0.7	0.75
DSLNT	1.4	1.9	0.01	2.4	3.0	0.45
FLNH	0.6	0.6	0.64	0.6	1.2	0.008
DFLNH	0.4	0.4	0.77	0.6	0.8	0.41
FSDLNH	1.3	1.5	0.32	3.2	4.2	0.13
DSLNH	0.4	0.5	0.38	0.5	0.8	0.01

Bold denotes significant p < 0.05.

Anxiety defined by State-Trait Anxiety Index S-Scale (STAI-S) ≥45.

No anxiety defined by STAI-S score <45.

2′FL, 2′-fucosyllactose; 3FL, 3-fucosyllactose; 3′SL, 3′-sialyllactose; 6′SL, 6′-sialyllactose; DFLac, difucosyllactose; DFLNH, difucosyl-LNH; DFLNT, difucosyl-LNT; DSLNH, disialyl-LNH; DSLNT, disialyl-LNT; FDSLNH, fucosyl-disialyl-LNH; FLNH, fucosyl-LNH; HMO, human milk oligosaccharide; LNFP, lacto-N-fucopentaose; LNH, lacto-N-hexaose; LNT, lacto-N-tetraose; LNnT, lacto-N-neotetraose; LST, sialyl-LNT.

Stress

In Secretors, the concentrations of the following 8 HMOs significantly varied in women with stress versus those without (p < 0.05): 3FL, LNnT, DFLac, LNFP III, LSTc, LNH, DSLNT, and DSLNH (Table 4). DFLac, LNFP III, and DSLNH remained significant after applying Bonferroni correction. In multivariable linear regression, DFLac (aRR −1.0, 95% CI −1.5 to −0.2), LNFP III (aRR −0.1, 95% CI −0.1 to −0.0), and DSLNH (aRR −0.1, 95% CI −0.2 to −0.0) were negatively associated with maternal stress.

Table 4.

The Impact of Stress on Human Milk Oligosaccharide Profiles, Stratified by Secretor Status

	Secretor (n = 364)				Non-Secretor (n = 129)
% HMO	Stress^a (n = 93)	No stress^b (n = 271)	p-Value	aRR 95% CI^c	Stress^a (n = 31)	No stress ^b (n = 98)	p-Value	aRR 95% CI^c
2’FL	38.4	38.5	0.91		1.0	0.7	0.41
3FL	14.6	9.8	0.005		40.4	26.5	0.01
LNnT	5.4	4.2	0.03		2.1	3.9	0.02
3′SL	4.6	5.1	0.32		4.1	4.6	0.51
DFLac	2.6	3.7	0.0009 ^d	−1.0 −1.5 to −0.2	1.8	1.3	0.19
6′SL	3.4	3.5	0.84		7.4	6.7	0.57
LNT	4.2	4.9	0.19		5.8	9.9	0.06
LNFP I	4.8	5.4	0.13		1.8	1.9	0.25
LNFP II	8.1	8.8	0.22		19.9	24.7	0.04
LNFP III	0.3	0.4	0.0002 ^d	−0.1 −0.1 to −0.0^e	0.4	0.7	0.03
LSTb	0.6	0.7	0.13		1.2	1.5	0.01
LSTc	0.4	0.5	0.02		0.4	0.4	0.72
DFLNT	7.9	8.8	0.11		4.4	6.6	0.01
LNH	0.4	0.5	0.008		0.7	0.8	0.56
DSLNT	1.6	2.0	0.03		2.2	3.1	0.14
FLNH	0.5	0.6	0.13		1.1	1.1	0.90
DFLNH	0.4	0.4	0.26		0.9	0.6	0.008
FDSLNH	1.3	1.5	0.04		4.3	3.2	0.05
DSLNH	0.3	0.5	0.001 ^d	−0.1 −0.2 to −0.0^e	0.6	0.8	0.31

Bold denotes significant p < 0.05.

Stress defined by Perceived Stress Scale (PSS) ≥20.

No stress defined by PSS <20.

Confidence intervals rounded to nearest one decimal point.

Indicates significant p-value (<0.002) after Bonferroni correction.

Statistical significance based off actual numbers; rounded numbers displayed.

Model adjusted for infant sex, exclusive breastfeeding, infant collection age, maternal age, race, depression and anxiety medications, and maternal body mass index.

In non-Secretors, 3FL, LNnT, LNFP II, LNFP III, LSTb, DFLNT, DFLNH, and FDSLNH initially varied in univariate analysis (p < 0.05). However, none remained significant after applying Bonferroni correction, suggesting that stress is not associated with the concentrations of these HMOs in non-Secretors.

Principal component analysis

To further analyze the relationship between maternal psychological state and HMO composition, a PCA was performed on all 19 HMOs. Similar to the analysis above, PCA was performed independently for Secretors and non-Secretors.

In Secretors, the top six PCs had an eigenvalue >1, explaining ∼78% of the total variance in the data (Supplementary Table S1). Non-Secretors had five PCs with an eigenvalue >1, which similarly explains ∼78% of the total variance in the data.

Within each group, additional analysis on PCs was performed to identify which HMOs had the largest impact on their profiles by ranking the magnitude of the loading scores of the HMOs in the profile (Supplementary Table S2). In Secretors, LNFP III, LNT, DSLNT, LNFP II, DFLNT, and 3FL contributed to PC1, while DSLNH, LSTc, FLNH, LNH, DFLNH, and 6′SL contributed to PC2. 2′FL, LNFP I, FDSLNH, LNFP II, LNH, and 3FL contributed to PC3, LSTb, LNnT, 3′SL, DFLac, LNFP I, and LNT contributed to PC4, and 6′SL, LSTb, 2′FL, DSLNT, DFLac, and LNH contributed to PC5. PC6 had the largest contributions from LNnT, LSTc, DFLNT, FDSLNH, FLNH, and 3′SL in Secretors.

In non-Secretors, LNFP III, 3FL, LNT, LNnT, DSLNT, and LNFP II contributed most to PC1, while FLNH, LSTc, DFLac, DSLNH, 6′SL, and LNH contributed to PC2. LNFP I, LSTb, 3′SL, 6′SL, FDSLNH, and DSLNH contributed to PC3, DFLNT, LNFP II, LNnT, LNT, FDSLNH, and DFLNH contributed to PC4, and 2′FL, DSLNT, 3′SL, LNFP III, DSLNH, and LNFP I contributed to PC5.

For both Secretors and non-Secretors, PCs were evaluated with respect to maternal depressive, anxiety, and stress symptoms (Table 5). In Secretors, PC2 was significantly associated with maternal symptoms of depression (aRR −0.6, 95% CI −1.1 to −0.1) and stress (aRR −0.3, 95% CI −0.7 to −0.0). No PCs were significantly associated with any of the psychological states in non-Secretors.

Table 5.

Principal Component Analysis in Secretors and non-Secretors

Principal component	Secretors
Principal component	Depression^a	Anxiety^b aRR (95% CI)	Stress^c
PC1	−0.1 (−0.7, 0.6)	−0.1 (−0.7, 0.4)	−0.4 (−0.9, 0.0)
PC2	−0.6 (−1.1, −0.1)	−0.3 (−0.8, 0.1)	−0.3 (−0.7, −0.0) ^d
PC3	−0.3 (−0.8, 0.2)	−0.2 (−0.6, 0.3)	0.0 (−0.3, 0.4)
PC4	0.1 (−0.4, 0.5)	0.1 (−0.2, 0.5)	0.3 (−0.0, 0.6)
PC5	−0.2 (−0.6 0.2)	−0.2 (−0.5, 0.2)	−0.2 (−0.5, 0.1)
PC6	−0.2 (−0.6, 0.1)	0.0 (−0.3, 0.2)	0.1 (−0.2, 0.3)
	Non-Secretors
PC1	−0.2 (−1.3, 1.0)	0.8 (−0.4, 1.9)	0.1 (−0.8, 1.1)
PC2	−0.8 (−1.6, 0.1)	−0.4 (−1.3, 0.4)	−0.1 (−0.8, 0.7)
PC3	−0.0 (−0.8, 0.7)	0.3 (−0.3, 0.9)	0.1 (−0.5, 0.8)
PC4	0.3 (−0.3, 0.9)	0.2 (−0.4, 0.8)	0.2 (−0.3, 0.7)
PC5	0.5 (−0.0, 1.0)	0.4 (−0.1, 1.0)	−0.1 (−0.5, 0.4)

Bold denotes significant confidence interval.

Models adjusted for infant sex, exclusive breastfeeding, infant collection age, maternal age, race, depression and anxiety medications, and maternal body mass index.

Depression defined as Edinburgh Postnatal Depression Scale (EPDS) ≥12.

Anxiety defined by State-Trait Anxiety Index S-Scale (STAI-S) ≥45.

Stress defined as Perceived Stress Scale (PSS) ≥20.

Statistical significance based off actual numbers; rounded numbers displayed.

aRR, adjusted risk ratio; CI, confidence interval, PC, principal component.

Discussion

Our results demonstrated that HMOs vary with maternal psychological symptom level above standard cutoff scores. We identified associations between HMOs and symptoms of maternal depression and stress, both at the individual HMO level and among one cluster of HMOs. All significant associations were negative, indicating that depressive symptoms and stress may have a negative effect on HMO levels.

To our knowledge, this is the first study to evaluate the relationship between maternal psychological health and HMO composition. Previous research has found significant associations between maternal psychological distress and other components of breast milk, including a positive relationship between maternal depressive and anxiety symptoms and the levels of sodium and the sodium-to-potassium ratio in breast milk.²⁴ In this study, researchers hypothesized that women with symptoms of anxiety or depression may breastfeed less, resulting in incomplete duct emptying and increased mammary gland permeability. Another study supported this proposed mechanism, demonstrating an association between postpartum depression and increased mammary gland permeability only when depression hindered exclusive breastfeeding.²⁵ In our study, HMOs varied with depressive symptoms and stress even after controlling for exclusive breastfeeding, suggesting that this mechanism is less likely to account for our findings.

Prior research has also demonstrated increased levels of TGF-β in women with symptoms of depression and anxiety.²⁶ Researchers hypothesized that serotonin receptor-mediated perturbations of the immune system may be one mechanism underlying this association. Given HMOs’ significant role in the immune system, it is plausible that this mechanism may contribute to HMO variation in women with symptoms of depression and stress. Further studies are needed to test this hypothesis, as the scope of this study was unable to determine causal relationships.

In our study, 74% of the women were Secretors, which is comparable with the global Secretor prevalence of 80%.¹⁶ As expected, HMOs varied by Secretor status, and Secretors had a higher total concentration of HMOs than non-Secretors, on average. However, we also observed a high degree of variation within Secretor groups; the total HMO content varied 4.8-fold in non-Secretors and 4.5-fold in Secretors. These findings are consistent with other studies, which similarly demonstrate high variability of HMO profiles within each group.¹⁰ Although we would expect Secretors to have elevated levels of α1,2-fucosylated HMOs due to their presence of FUT2, they varied in other HMO structures as well. This suggests that the complexity of HMO profiles is multifactorial, perhaps influenced by other enzymes, genetic factors, and epigenetics.^1,32 Of note, our study did not consider Le status in addition to Secretor status; thus, we stratified based on two groups instead of four. However, HMOs that are not α1,4-fucosylated, such as LNH, also varied between the groups, reiterating the complexity of HMO composition.

We applied multiple statistical approaches to identify associations between maternal psychological state and HMO composition. We controlled for most of the factors known to impact HMO profiles as reported by prior studies. In our analysis, although many HMOs initially varied with maternal depressive, anxiety, and stress symptoms, only a few remained after adjusting for multiple comparisons. Overall, the HMOs that significantly varied with maternal psychological state were LNFP III and LNH for Secretors with depressive symptoms, and DFLac, LNFP III, and DSLNH for Secretors with stress. Given that DFLac is the only HMO of the listed expected to significantly vary between Secretors and non-Secretors due to its dependence on FUT2 for its synthesis, it is unclear why the other structures also vary with Secretor status, making this an important area of interest for future research.

For all significant associations, there was a negative relationship between maternal psychological distress and the percent concentration of the HMO. This suggests that maternal depressive symptoms and stress may negatively affect the levels of some HMOs, which could have important implications for infant health. Previous findings have identified that DSLNT may protect premature infants against necrotizing enterocolitis, LNFP II may reduce gastroenteritis and respiratory infections in infants exposed to human immunodeficiency virus (HIV), and 2′FL and LNH may promote healthy cognitive development.^9,33–35 Of these examples, LNH significantly varied with depressive symptoms in Secretors. Future studies should analyze whether associations between maternal psychological state and infant outcomes are potentially mediated by changes in milk composition.

Interestingly, HMOs did not significantly vary with anxiety symptoms at the individual level or in PCA. By utilizing STAI-S ≥45 as our cutoff for anxiety, we may have missed participants with clinically meaningful symptoms of anxiety. Although this cutoff has been utilized in prior studies,²⁹ other groups have used alternative thresholds, including STAI ≥40.³⁶ In addition, the STAI is not the only validated tool to screen for anxiety. Another such scale is the Brief Measure of Worry Severity survey, which is preferred by some groups for measuring postnatal anxiety.³⁷ As such, future research should evaluate the impact of anxiety on HMO profiles by using these alternative measures.

Of note, controlling for maternal use of antidepressants and/or anxiolytics in our analysis altered our findings. In models without inclusion of this covariate, we identified additional significant associations (p < 0.002) in comparison with the findings described above. However, after determining that the use of these medications is significantly associated with HMOs, we adjusted for it in the final models, which nullified many of the associations between the psychiatric scales and HMOs. Thus, we conclude that treatments for depression and/or anxiety seem to be associated with HMOs, which remains an important area for further investigation.

There were several limitations to our study. One constraint was our study’s overrepresentation of women identifying as White, limiting the ability to generalize our findings to more diverse populations. Another limitation was the potential for misclassification of the exposure due to allowing up to 60 days between the participants’ milk collection and completion of psychiatric scales. Notably, however, the median time between milk donation and completion of the questionnaires was one day. Finally, because our data relied on self-reported questionnaires, our study was limited by reporting bias.

Conclusion

To our knowledge, this is the first study evaluating the association between maternal psychological distress and HMOs. Our results show variations in the concentrations of several HMOs with symptoms of maternal depression and stress. These psychological symptoms also varied with one cluster of HMOs in PCA. Together, these findings indicate a relationship between maternal psychological health and breast milk composition. Although we found no association between maternal anxiety and HMOs, additional studies using alternative scales and cutoffs for anxiety are needed to replicate this finding. Future research should also determine whether our observed variations in HMOs affect infant health.

Footnotes

Authors’ Contributions

All authors conceived of the presented idea. G.B. performed the analysis, and H.R. wrote the article with the supervision of G.B. and K.B. C.C. reviewed the content, and all authors discussed the results and contributed to the final article.

Data Availability

The data that support the findings of this study can be requested from the UC San Diego Human Milk Biorepository. The code for statistical analysis is available from the authors upon request.

Ethical Approval

This study was approved by the University of California, San Diego Institutional Review Board (Project #130658) on December 6, 2023. All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Declaration of Helsinki of 1975, as revised in 2000.

Author Disclosure Statement

C.C. receives research funding from the following industry sponsors and a foundation: Amgen, AstraZeneca; GlaxoSmithKline; Janssen Pharmaceuticals; Pfizer, Inc.; Regeneron; Hoffman La-Roche-Genentech; Genzyme Sanofi-Aventis; Takeda Pharmaceutical Company Limited; Sanofi; UCB Pharma, USA; Leo Pharma, Sun Pharma Global FZE; Gilead; Novartis; and the Gerber Foundation. None of the other authors have any disclosures.

Funding Information

The UC San Diego HMB is funded by the UC San Diego Center for Better Beginnings, the National Institutes of Health, grant UL1TR001442 (PI Firestein), and the Gerber Foundation, project 4998 and project 6488 (PI Chambers), and National Institutes of Health, National Institute of Child Health and Development, grant P50HD106463 (PI Tremoulet).

Supplementary Material

References

Bode

. Human milk oligosaccharides: Every baby needs a sugar mama. Glycobiology, 2012; 22(9):1147–1162; doi: 10.1093/glycob/cws074

Moubareck

. Human milk microbiota and oligosaccharides: A glimpse into benefits, diversity, and correlations. Nutrients, 2021; 13(4):1123; doi: 10.3390/nu13041123

Berger

, Hampson

, Schmidt

, et al. Stability of human-milk oligosaccharide concentrations over 1 week of lactation and over 6 hours following a standard meal. J Nutr, 2022; 152(12):2727–2733; doi: 10.1093/jn/nxac214

Z-T

, Nanthakumar

, Newburg

. The human milk oligosaccharide 2′-fucosyllactose quenches campylobacter jejuni–induced inflammation in human epithelial cells HEp-2 and HT-29 and in mouse intestinal mucosa. J Nutr, 2016; 146(10):1980–1990; doi: 10.3945/jn.116.230706

Weichert

, Jennewein

, Hüfner

, et al. Bioengineered 2′-fucosyllactose and 3-fucosyllactose inhibit the adhesion of Pseudomonas aeruginosa and enteric pathogens to human intestinal and respiratory cell lines. Nutr Res, 2013; 33(10):831–838; doi: 10.1016/j.nutres.2013.07.009

Kuntz

, Rudloff

, Kunz

. Oligosaccharides from human milk influence growth-related characteristics of intestinally transformed and non-transformed intestinal cells. Br J Nutr, 2008; 99(3):462–471; doi: 10.1017/S0007114507824068

Kuntz

, Kunz

, Rudloff

. Oligosaccharides from human milk induce growth arrest via G2/M by influencing growth-related cell cycle genes in intestinal epithelial cells. Br J Nutr, 2009; 101(9):1306–1315; doi: 10.1017/S0007114508079622

Charbonneau

, O’Donnell

, Blanton

, et al. Sialylated milk oligosaccharides promote microbiota-dependent growth in models of infant undernutrition. Cell, 2016; 164(5):859–871; doi: 10.1016/j.cell.2016.01.024

Willemsen

, Beijers

, Gu

, et al. Fucosylated human milk oligosaccharides during the first 12 postnatal weeks are associated with better executive functions in toddlers. Nutrients, 2023; 15(6):1463; doi: 10.3390/nu15061463

10.

Azad

, Robertson

, Atakora

, et al. Human milk oligosaccharide concentrations are associated with multiple fixed and modifiable maternal characteristics, environmental factors, and feeding practices. J Nutr, 2018; 148(11):1733–1742; doi: 10.1093/jn/nxy175

11.

Dou

, Luo

, Xing

, et al. Human milk oligosaccharides variation in gestational diabetes mellitus mothers. Nutrients, 2023; 15(6):1441; doi: 10.3390/nu15061441

12.

Wang

, Zhao

, et al. Neutral human milk oligosaccharides are associated with multiple fixed and modifiable maternal and infant characteristics. Nutrients, 2020; 12(3):826; doi: 10.3390/nu12030826

13.

Ferreira

, Alves

, Figueiredo

, et al. Human milk oligosaccharide profile variation throughout postpartum in healthy women in a Brazilian cohort. Nutrients, 2020; 12(3):790; doi: 10.3390/nu12030790

14.

Sprenger

, Lee

, De Castro

, et al. Longitudinal change of selected human milk oligosaccharides and association to infants’ growth, an observatory, single center, longitudinal cohort study. PLoS One, 2017; 12(2):e0171814; doi: 10.1371/journal.pone.0171814

15.

Vinjamuri

, Davis

JCC

, Totten

, et al. Human milk oligosaccharide compositions illustrate global variations in early nutrition. J Nutr, 2022; 152(5):1239–1253; doi: 10.1093/jn/nxac027

16.

Soyyılmaz

, Mikš

, Röhrig

, et al. The mean of milk: A review of human milk oligosaccharide concentrations throughout lactation. Nutrients, 2021; 13(8):2737; doi: 10.3390/nu13082737

17.

Samuel

, Binia

, De Castro

, et al. Impact of maternal characteristics on human milk oligosaccharide composition over the first 4 months of lactation in a cohort of healthy European mothers. Sci Rep, 2019; 9(1):11767; doi: 10.1038/s41598-019-48337-4

18.

, Mao

, Liu

, et al. Vitamins, vegetables and metal elements are positively associated with breast milk oligosaccharide composition among mothers in Tianjin, China. Nutrients, 2022; 14(19):4131; doi: 10.3390/nu14194131

19.

Biddulph

, Holmes

, Tran

, et al. Associations between maternal nutrition and the concentrations of human milk oligosaccharides in a cohort of healthy Australian lactating women. Nutrients, 2023; 15(9):2093; doi: 10.3390/nu15092093

20.

Pell

, Ohuma

, Yonemitsu

, et al. The human-milk oligosaccharide profile of lactating women in Dhaka, Bangladesh. Curr Dev Nutr, 2021; 5(12):nzab137; doi: 10.1093/cdn/nzab137

21.

Sjögren

, Duchén

, Lindh

, et al. Neutral oligosaccharides in colostrum in relation to maternal allergy and allergy development in children up to 18 months of age. Pediatr Allergy Immunol, 2007; 18(1):20–26; doi: 10.1111/j.1399-3038.2006.00486.x

22.

Wang

, Liu

, Li

, et al. Pregnancy-related diseases and delivery mode can affect the content of human milk oligosaccharides: A preliminary study. J Agric Food Chem, 2022; 70(16):5207–5217; doi: 10.1021/acs.jafc.2c00147

23.

M. Tonon

, De Morais

MFV

, Abrão

, et al. Maternal and infant factors associated with human milk oligosaccharides concentrations according to secretor and Lewis phenotypes. Nutrients, 2019; 11(6):1358; doi: 10.3390/nu11061358

24.

Serim Demirgoren

, Ozbek

, Ormen

, et al. Do mothers with high sodium levels in their breast milk have high depression and anxiety scores? J Int Med Res, 2017; 45(2):843–848; doi: 10.1177/0300060517700013

25.

Flores-Quijano

, Córdova

, Contreras-Ramírez

, et al. Risk for postpartum depression, breastfeeding practices, and mammary gland permeability. J Hum Lact, 2008; 24(1):50–57; doi: 10.1177/0890334407310587

26.

Shariat

, Abedinia

, Rezaei

, et al. Increase concentration of transforming growth factor beta (TGF-β) in breast milk of mothers with psychological disorders. Acta Med Iran, 2017; 55(7):429–436.

27.

Bandoli

, Bertrand

, Saoor

, et al. The design and mechanics of an accessible human milk research biorepository. Breastfeed Med, 2020; 15(3):155–162; doi: 10.1089/bfm.2019.0277

28.

Renwick

, Rahimi

, Sejane

, et al. Consistency and variability of the human milk oligosaccharide profile in repeat pregnancies. Nutrients, 2024; 16(5):643; doi: 10.3390/nu16050643

29.

Obrochta

, Chambers

, Bandoli

. Psychological distress in pregnancy and postpartum. Women Birth, 2020; 33(6):583–591; doi: 10.1016/j.wombi.2020.01.009

30.

Navaratne

, Foo

, Kumar

. Impact of a high Edinburgh Postnatal Depression Scale score on obstetric and perinatal outcomes. Sci Rep, 2016; 6(1):33544; doi: 10.1038/srep33544

31.

Crouch

, Chambers

, Bertrand

, et al. The association between adverse and positive childhood experiences and marijuana use during lactation. Breastfeed Med, 2023; 18(3):212–218; doi: 10.1089/bfm.2022.0243

32.

McGuire

, Meehan

, McGuire

, et al. What’s normal? Oligosaccharide concentrations and profiles in milk produced by healthy women vary geographically. Am J Clin Nutr, 2017; 105(5):1086–1100; doi: 10.3945/ajcn.116.139980

33.

Masi

, Embleton

, Lamb

, et al. Human milk oligosaccharide DSLNT and gut microbiome in preterm infants predicts necrotising enterocolitis. Gut, 2021; 70(12):2273–2282; doi: 10.1136/gutjnl-2020-322771

34.

Sudarma

, Hegar

, Hidayat

, et al. Human milk oligosaccharides as a missing piece in combating nutritional issues during exclusive breastfeeding. Pediatr Gastroenterol Hepatol Nutr, 2021; 24(6):501–509; doi: 10.5223/pghn.2021.24.6.501

35.

Berger

, Plows

, Jones

, et al. Human milk oligosaccharide 2’-fucosyllactose links feedings at 1 month to cognitive development at 24 months in infants of normal and overweight mothers. PLoS One, 2020; 15(2):e0228323; doi: 10.1371/journal.pone.0228323

36.

Facco

, Stellini

, Bacci

, et al. Validation of visual analogue scale for anxiety (VAS-A) in preanesthesia evaluation. Minerva Anestesiol, 2013; 79(12):1389–1395.

37.

Austin

M-P

, Tully

, Parker

. Examining the relationship between antenatal anxiety and postnatal depression. J Affect Disord, 2007; 101(1–3):169–174; doi: 10.1016/j.jad.2006.11.015

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

2.09 MB

0.18 MB

0.10 MB

0.08 MB