The Utility of Human Milk Oligosaccharides against Group B Streptococcus Infections of Reproductive Tissues and Cognate Adverse Pregnancy Outcomes

Preterm birth affects nearly 10% of all pregnancies in the United States, with 40% of those due, in part, to infections. Streptococcus agalactiae (Group B Streptococcus, GBS) is one of the most common perinatal pathogens responsible for these infections. Current therapeutic techniques aimed to ameliorate invasive GBS infections are less than desirable and can result in complications in both the neonate and the mother. To this end, the need for novel therapeutic options is urgent. Human milk oligosaccharides (HMOs), an integral component of human breast milk, have been previously shown to possess antiadhesive and antimicrobial properties. To interrogate these characteristics, we examined HMO-mediated outcomes in both in vivo and ex vivo models of GBS infection utilizing a murine model of ascending GBS infection, an EpiVaginal human organoid tissue model, and ex vivo human gestational membranes. Supplementation of HMOs resulted in diminished adverse pregnancy outcomes, decreased GBS adherence to gestational tissues, decreased colonization within the reproductive tract, and reduced proinflammatory immune responses to GBS infection. Taken together, these results highlight the potential of HMOs as promising therapeutic interventions in perinatal health.


Safety Statement
No unexpected or unusually high safety hazards were encountered in this study.

Bacterial Strains and Culture Conditions
The bacterial strains used in this study are shown in Table S1.All bacterial strains were grown on tryptic soy agar plates supplemented with 5% sheep blood (blood agar) plates at 37˚C in ambient air overnight.Bacteria were sub-cultured from blood agar plates into Todd-Hewitt broth (THB) and incubated at 37°C in ambient air overnight.The following day, bacterial density was measured spectrophotometrically at an optical density of 600 nm (OD 600 ), and bacterial numbers were determined with a coefficient of 1 OD 600 = 10 9 CFU/mL.

HMO isolation
Human milk was obtained from healthy, lactating women between 3 days and 3 months postpartum and stored between −80 and -20°C.Deidentified milk was provided by Dr. Jörn-Hendrik Weitkamp from the Vanderbilt Department of Pediatrics, under a collection protocol approved by the Vanderbilt University institutional review board (IRB #100897), or from Medolac.
Milk samples were thawed and then centrifuged for 45 min.Following centrifugation, the resultant top lipid layer was removed.The proteins were then removed by diluting the remaining sample with roughly 1:1 (vol/vol) 180 or 200 proof ethanol, chilling the sample briefly, and centrifuging for 45 min, followed by removal of the resulting HMO-containing supernatant.Following concentration of the supernatant in vacuo, the HMO-containing extract was dissolved in 0.2 M phosphate buffer (pH 6.5) and heated to 37°C.β-Galactosidase from Kluyveromyces lactis was added, and the reaction mixture was stirred until lactose hydrolysis was complete.The reaction mixture was diluted with roughly 1:0.5 (vol/vol) 180 or 200 proof ethanol, chilled briefly, and then centrifuged for 30 min.The supernatant was removed and concentrated in vacuo, and the remaining salts, glucose, and galactose were separated from the oligosaccharides using size exclusion chromatography with P-2 gel (H2O eluent).The oligosaccharides were then dried by lyophilization.Correspondingly, HMO isolates from donors were combined and solubilized in water to reach a final concentration of 102.6 mg/ml.

MS and MS/MS Analysis of HMO Samples
Dried HMO samples were prepared and processed for evaluation by reconstitution in water to approximately 1 mg/mL.These solutions were deposited on a matrix-assisted laser desorption/ionization (MALDI) target plate as follows: 1 μL of HMO was spotted followed by 0.2 μL of 10 mM NaCl and 1 μL of DHB matrix (60 mg/mL in 50% methanol).The spots were allowed to air-dry and then were analyzed in positive ion mode on a 9.4T Fourier transform ion cyclotron resonance (FT-ICR) mass spectrometer (MS) (Bruker Solarix).Mass spectra were acquired in positive ion mode from m/z 300 to 2500.Sodium ion adducts of HMOs were detected with a mass accuracy of >2 ppm.MS/MS analysis was performed for selected ions with a linear ion trap mass spectrometer equipped with a MALDI source (LTQ XL, Thermo Scientific).Selected sodium adduct ions of interest were isolated with a 1 amu window and fragmented via CID using a collision energy of 35 eV (Supplemental Figure 1-3). 1

Epi Vaginal Coculture
EpiVaginal™ (VEC-100™) tissues were purchased from MatTek Life Sciences.These tissues are cultured from normal, primary human-derived vaginal epithelial cells, and are similar to studying in vivo tissue.The tissues were processed into 8-mm diameter sections, and sections were placed in Nunc™ single well tissue culture plate inserts containing DMEM medium, F12 medium, phenol-red cell culture medium, supplemented with 5 µg/ml gentamicin (10% of normal gentamicin level) and 0.25 µg/ml amphotericin B. Sections were thawed and incubated overnight at 37°C in ambient air containing 5% CO 2 ; washed 3 times with prewarmed, sterile phosphate buffered saline (pH 7.4); and placed in VEC-100-MM™ (MatTek Life Sciences) maintenance medium.Bacterial cells were added to the top face of the tissue at a multiplicity of infection of 1 × 10 6 cells per 8-mm diameter membrane, using a predetermined coefficient of bacterial density of 1 OD 600 = 10 9 CFU/mL.HMOs were added to achieve a final carbohydrate concentration of ca. 5 mg/mL.Concomitantly, uninfected vaginal tissue samples were also maintained.Cocultures were incubated at 37°C in ambient air containing 5% CO 2 for 24 hours and cells were fixed with 2.0% paraformaldehyde and 2.5% glutaraldehyde in 0.05 M sodium cacodylate buffer (Electron Microscopy Sciences, Hatfield, PA) for at least 12 hours prior to processing for microscopy.

Gestational Membrane Coculture
De-identified gestational membrane tissue samples were procured from term, nonlaboring Caesarean section-delivery live births at Vanderbilt University Medical Center with approval from the Vanderbilt University Medical Center Institutional Review Board (VUMC IRB #181998).12-mm gestational membranes biopsy punches were isolated and cultured in RPMI 1640 medium (ThermoFisher, Waltham, MA) with 10% charcoal stripped fetal bovine serum (ThermoFisher) and 1% antibiotic/antimycotic solution (ThermoFisher) overnight at 37°C in room air supplemented with 5% carbon dioxide.The membranes were washed 3 times, infected with 10 6 CFU/mL of GBS in RMPI 1640 medium without antibiotics in the absence of HMO treatment or supplemented with HMOs at a concentration of 5 mg/mL.A predetermined coefficient of bacterial density of 1 OD 600 = 10 9 CFU/mL.Cocultured tissues were incubated at 37°C in air supplemented with 5% carbon dioxide overnight and cells were fixed with 2.0% paraformaldehyde and 2.5% glutaraldehyde in 0.05 M sodium cacodylate buffer (Electron Microscopy Sciences, Hatfield, PA) for at least 12 h prior to processing for microscopy.

Quantifying Bacterial Adherence in Gestational Membranes
To determine bacterial adherence in gestational membranes, quantitative culture methods were employed.Following overnight incubation at 37°C in air supplemented with 5% carbon dioxide, gestational membranes were washed three times and placed in mL of sterile THB.
Tissues were homogenized and subjected to serial dilution and plating onto blood agar to enumerate bacteria (CFU/mL) in host tissue.

High-Resolution Field-Emission Gun Scanning Electron Microscopy (FEG-SEM) Analyses
3][4] Briefly, bacterial cells were cultured in biofilms adhering to gestational membranes overnight in the culture conditions described above.HMOs were dissolved in DI water to achieve a concentration of 102.6 mg/mL and filtered through a 0.2 μm syringe filter.HMOs were added to achieve a final carbohydrate concentration of ca. 5 mg/mL.The following day, bacterial cells were fixed in a solution of 2.5% glutaraldehyde, 2.0% paraformaldehyde, and 0.05 M sodium cacodylate buffer pH 7.4.Samples were dehydrated with sequential washes of increasing concentrations of ethanol before being subjected to critical point drying, mounting on aluminum stubs, and sputter coating with 20 nm of gold-palladium.Samples were viewed using an FEI Quanta 250 field-emission gun scanning electron microscope at 5 kEV with a spot size of 2.5.

Mouse Model of Ascending Vaginal GBS infection During Pregnancy
GBS infection of pregnant mice and subsequent analyses were performed as previously described. 5,6 riefly, C57BL6/J mice were purchased from Jackson laboratories and mated in harem breeding strategies (1 male to 3-4 females) overnight.The following day, pregnancy was confirmed by the presence of a vaginal mucus plug establishing the embryonic date (E0.5).On embryonic day 12.5 (E12.5)pregnant dams were anesthetized via inhalation of isoflurane and vaginally injected or dosed orally with an HMO cocktail at a concentration of ca. 5 mg/mL.On embryonic day 13.5 (E13.5)pregnant dams were anesthetized via inhalation of isoflurane and vaginally infected by pipetting 5x10 2 to 10 3 colony forming units (CFU) in 0.05 mL of THB plus 10% gelatin directly into the vagina.It is important to note that the release of the liquid might create a pressure effect where some of the inocula could be propelled into the cervix.Uninfected controls were also maintained.On embryonic day 15.5 (E15.5)animals were euthanized by carbon dioxide asphyxiation and necropsy was performed to harvest reproductive tissues including vagina, uterus, placenta, decidua, fetal membranes, and fetus.

PPROM, Preterm Birth, and Survival Analyses
GBS infection of pregnant mice and subsequent analyses were performed as previously described. 5,6 riefly, C57BL6/J mice were purchased from Jackson laboratories and mated in harem breeding strategies (1 male to 3-4 females) overnight.The following day, pregnancy was confirmed by the presence of a vaginal mucus plug establishing the embryonic date (E0.5).On embryonic day 12.5 (E12.5)pregnant dams were anesthetized via inhalation of isoflurane and vaginally injected with an HMO cocktail at a concentration of ca. 5 mg/mL.On embryonic day 13.5 (E13.5)pregnant dams were anesthetized via inhalation of isoflurane and vaginally infected (directly into the cervix) with 5x10 3 to 1x10 4 colony forming units (CFU) in 0.05 mL of THB plus 10% gelatin.Uninfected, and infected (untreated) controls were also maintained.Animals were monitored daily for PPROM, preterm birth, and maternal survival.On embryonic day 21.5 (E21.5)animals were euthanized by carbon dioxide asphyxiation.

Quantifying Bacterial Burden in Host Tissues
To determine bacterial burden in reproductive tissues quantitative culture methods were employed as previously described. 5Briefly, reproductive tissues were weighed and placed in sterile THB (Todd-Hewitt broth which is optimized for growth of streptococcal species).Tissues were homogenized and subjected to serial dilution and plating onto blood agar to enumerate bacteria (CFU/mg) in host tissue.

Histopathological Analyses
Reproductive tissues were subjected to a primary fixation in 4% formalin (neutral buffered) overnight.The following day, tissues were embedded in paraffin and sectioned into 5 m thick sections for staining and microscopical analyses.Sections were stained with hematoxylin and eosin for histopathological examination and imaged with an OMAX M83ES compound light microscope with ToupView software package.

Immunohistochemical Analyses
Tissues were fixed in 4% neutral buffered formaldehyde overnight before being embedded into paraffin blocks.Samples were cut into 5-μm sections, and multiple sections were placed on each slide for analysis.Samples were deparaffinized with xylene, and heat-induced antigen retrieval was performed on the Bond Max automated IHC stainer (Leica Biosystems) using Epitope Retrieval 2 solution for 5 to 20 min.Slides were incubated with a rabbit polyclonal anti-GBS antibody (ab78846; Abcam) for 1 h.The Bond Polymer Refine detection system (Leica Biosystems) was used for visualization.Slides were counter stained with eosin, dehydrated and cleared, and coverslips were added before light microscopy analysis was performed.

Cytokine Analyses
Mouse reproductive tissues, maternal sera, and amniotic fluid were analyzed by multiplex cytokine assays.Mouse tissues were placed in 1 mL of sterile PBS or THB+ 10 mg/mL penicillin and homogenized and passed through a 0.22 µm filter.Samples were frozen at -80˚C or on dry ice until analyses were performed.Samples were analyzed by Eve Technologies via multiplex cytokine array (Eve Technologies, Alberta, Canada) as previously described 77 .We have previously validated host targets for specific cytokines (IL-1β, IL-6, KC, and TNF-α) by sandwich ELISA (AbCam). 7,8

RNA Extraction and Transcriptomic Analyses
Transcriptomic analyses were performed as previously described. 8Briefly, bacteria were cultured in THB broth alone or supplemented with 2.5 mg/mL of HMOs (a sub-inhibitory concentration) at 37˚C for 4 hours.15 mL cultures were centrifuged at 8,000 x g and the supernatant was removed from the cell pellet.Cells were resuspended in 1 mL of RNAprotect Bacteria reagent (Qiagen) and stored at -80˚C until RNA extraction protocol was performed.RNA was extracted using the RNeasy Mini Kit (Qiagen) with enzymatic lysis, proteinase K treatment, and mechanical disruption.DNA was removed using the Turbo DNA-free kit (Ambion) and ribosomal RNA was removed using the Ribo-Zero rRNA removal kit for Gram-positive bacteria (Epicentre).Libraries were prepared using TruSeq Stranded mRNA library preparation kit (Illumina) and RNAseq was performed by Hudson Alpha using a HiSeq X Ten.Base calling was performed with the Illumina Real Time Analysis (RTA) v1.18.64 and the resulting data was converted to FastQ files with Illumina Bcl2fastq, v1.8.4.Adapter and quality trimming were carried out with Trim Galore! v.0.6.4;while, ribosomal RNA was removed with SortMeRNA v.2.1. 9The accession information for the reference GBS genome, utilized for aligning the RNA-Seq reads, as well as the locus tag key employed for facilitating cross-identification of annotated transcripts, were obtained from the genome of Streptococcus agalactiae strain NEM316, assembly ASM19605v1 (GeneBank assembly GCA_000196055.1).Transcripts were mapped to the published genome of Streptococcus agalactiae strain NEM316 using STAR v.2.6.1d and quantified with Salmon v.1.2.1. 10Differential expression analysis was performed using DESeq2 comparing samples grown in medium supplemented with HMOs to samples grown in medium alone.Genes with at least a 2-fold change in expression and P<0.05 as determined by one-way ANOVA were considered to have significant differences in expression.

STRING Analyses 11
To identify putative protein-protein interactions and networks of changes in gene expression, differentially expressed genes were uploaded to the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (https://string-db.org/)and significant protein-protein interactions were determined at the criterion of confidence (combined score) > 0.4.

Statistical Analyses
Statistical analysis of parametric data with more than two groups was performed using one-way ANOVA with either Tukey's or Dunnet's post hoc correction for multiple comparisons; all reported P values were adjusted to account for multiple comparisons.For parametric data with two groups, a Student's t test was used.P values of ≤0.05 were considered significant.Nonparametric data (such as log-transformed CFU data) were analyzed by Mann-Whitney U or Kruskal-Wallis tests.Effect size (Cohen's D) was calculated from observed differences between the uninfected and infected groups to determine the smallest group size that would enable detection of a significant effect at a minimum desired power of 80%.The power statistics were calculated based on the hypothesis focused on proinflammatory cytokines where an effect was expected; namely, IL-6, IFN-g, and IL-1B.To justify the minimum group size, we focused on the smallest possible effect size among the comparisons of interest, which was IFN-g.Using the 'pwr' package, a minimum effect size of 1.53 was necessary for power of 80% with a group size of n=3.
Cohen's D for the comparison of IFNg between groups was 1.67; thus, a minimum group size of 3 individuals is sufficiently powered to detect differences.All data analyzed in this work were derived from at least three biological replicates (representing different placental samples).

Ethics Statement
This study was carried out in accordance with the recommendations of the Vanderbilt University Medical Center Institutional Review Board.This protocol was approved by the Institutional Review Board (IRB #181998 and #00005756).All animal experiments were performed in accordance with the Animal Welfare Act, U.S. federal law, and NIH guidelines.All experiments were carried out under a protocol approved by Vanderbilt University Institutional Animal Care and Use Committee (IACUC: M/14/034 and M/17/012), a body that has been accredited by the Association of Assessment and Accreditation of Laboratory Animal Care Act (AAALAC).

Figure S3 .
Figure S3.Human milk oligosaccharides reduce bacterial adherence to gestational membranes.Bacterial adherence was evaluated by quantitative culture (Log CFU/mL) in response to HMO supplementation.Data displayed represent ± SEM of three independent experiments, each with two technical replicates.In the presence of 5 mg/mL of HMOs, four strains of GBS exhibited significantly diminished adherence as determined by Student's ttest **P < 0.01, ***P < 0.001, and ****P < 0.0001).

Figure S9 .S23Figure S10 .
Figure S9.Heatmap of 474 significant transcripts identified with DESeq2.The level of expression is shown as the logarithm of TPMs.Medium alone control (untreated, blue) versus HMO-treated samples (HMO, red) are shown.Columns indicate three independent biological replicates utilized for analysis.

Table S1 .
Group B Streptococcus strains evaluated in the study.

Table S2 .
Generic structural descriptions of the molecular ions observed during analysis of HMOs in human milk samples.The exact mass is the mass calculated from a molecular formula using known masses of the most abundant isotopes of each element in the structure.[M+Na+] is the calculated mass of the sodium adduct that's expected after ionization.Measured is the mass value observed by the instrument.Ppm is a measurement of the error that exists between the expected mass and the observed mass.A negative ppm represents a measured mass that's greater than what was expected.A positive ppm represents a measured mass that's lower than what was expected.

Table S4 .
STRING analyses reveals predicted protein-protein-interactions implicating functional enrichment of specific metabolic pathways.