Estrogen Signaling During Abrupt Involution and Long-Term Metabolic Signature Similar to Estrogen Receptor-Negative Breast Cancer

TL;DR

Breast cancer risk may be linked to abrupt involution, which alters metabolism and estrogen signaling in ways similar to estrogen receptor-negative breast cancer.

Contribution

The study reveals long-term metabolic changes in abrupt involution resembling estrogen receptor-negative breast cancer.

Findings

Day 28 abrupt involution glands showed increased estrogen signaling and glucose metabolism compared to gradual involution.

Day 120 abrupt involution glands exhibited metabolic features similar to estrogen receptor-negative breast cancer.

Tamoxifen treatment in abrupt involution mice reversed some metabolic differences toward those seen in gradual involution.

Abstract

Epidemiological data link a lack of breastfeeding with an increased risk of breast cancer. Breast tissue remodels after pregnancy through involution. Long-term breastfeeding results in gradual involution (GI), and a lack of breastfeeding leads to abrupt involution (AI). AI causes increased mammary gland estrogen signaling, causing adipocyte redifferentiation through neutrophil infiltration. Adipocyte differences and metabolic implications of involution have not been explored between AI and GI. As breast cancer is characterized as highly metabolic, we explored how adipocyte differences and metabolism during involution may support breast cancer risk. FVB/n was randomized to AI/GI and standardized to 6 pups on day 0/birth. AI mice had pups removed on day 7. GI mice had 3 pups removed on days 28 and 31. Mammary glands were harvested at 28, 56, and 120 days. A subset of AI mice were given tamoxifen for 21 days. Day 28 AI glands had upregulation of estrogen signaling, neutrophil degranulation, and glucose metabolism and downregulation of adipogenesis and glycolysis compared to Day 56 GI. Day 120 AI glands had downregulation of oxidative phosphorylation and upregulation of mitochondrial dysfunction similar to estrogen receptor-negative (ER−) pregnancy-associated breast cancer (PABC). AI with tamoxifen resulted in a similar metabolic phenotype to GI. Early metabolic phenotypes in AI and GI glands may be related to estrogen signaling. AI long-term transcriptional metabolic effects were similar to breast cancer.

Figures

Full text

1. Introduction

Breast cancer is the most prevalent cancer among women in the United States, with 1 in 8 women being diagnosed [1]. Epidemiological studies have linked a lack of breastfeeding with an increased risk of breast cancer, specifically triple negative breast cancer (TNBC) [2,3,4,5,6]. A large collaborative analysis from multiple epidemiological studies comprising over 50,000 breast cancer cases and controls concluded that for 12 months of breastfeeding, the breast cancer risk was reduced by 4.3% [5]. Several studies show a 25–50% lower risk of TNBC in women who breastfed four to six months compared to parous women who never breastfed [4].

The mammary gland has a remarkable ability to rapidly proliferate and shift to lactation during pregnancy [7,8,9,10]. After the completion of breastfeeding, the mammary gland remodels to a near pre-pregnancy state through involution [7]. Long-term breastfeeding and gradual weaning leads to a gradual remodeling process: gradual involution (GI) [10]. Abrupt discontinuation of breastfeeding or lack of initiation after birth leads to abrupt involution (AI) [10]. While the epidemiological link between the lack of breastfeeding and breast cancer is clear [2,3,4,5], the mechanisms remain poorly understood. Research examining lack of breastfeeding following full-term pregnancy has suggested involution as the critical window involved in the risk of breast cancer [8,9,10].

Mouse models of AI resulted in a pro-tumorigenic environment in the mammary gland [8,9,10]. Studies show the short-term effects of AI include greater cell death and turnover, heightened immune response, and tissue remodeling, likened to wound healing [7,8,9,10,11]. AI leads to increased and sustained inflammation over prolonged periods, increased collagen deposition, and estrogenic signaling [10]. The mammary gland of mice that underwent AI displayed long-term increased cell proliferation, hyperplasia, and squamous metaplasia [10]. These studies conclude that AI leads to pre-cancerous changes within the mammary gland [7,8,9,10,11].

During AI, estradiol binds to estrogen receptor alpha (ERα) causing an increase in neutrophils and extracellular remodeling, which cause fibroblast-like stem cells to redifferentiate to adipocytes during the second phase of involution [12]. Adipocytes in AI glands expand after differentiation to repopulate the mammary gland, due to the hypertrophy of excess milk lipids released from collapsed alveolar structures [9]. Adipocyte redifferentiation and population in GI glands have not been explored. Research has shown that rapid adipocyte repopulation and hypertrophy cause hypoxia and metabolic reprogramming [13]. Breastfeeding and breast cancer are two highly metabolic processes. However, the changes in metabolism related to AI and GI within the mammary gland have not been studied. We hypothesized that the metabolic alterations induced by adipocyte repopulation of AI versus GI may drive resultant phenotypic differences in hyperplasia and inflammation. The objective of this study was to evaluate the short-term and long-term impact of AI on overall mammary gland metabolism using transcriptomic and functional analyses.

Our study shows similar AI adipocyte repopulation to that described in previous studies, but the completion of involution and the type of involution cause differences in adipocyte repopulation. Accompanied with adipocyte repopulation, AI and GI processes lead to differences in transcriptional metabolic outcomes, with long-term AI mammary glands mimicking the metabolic phenotype of estrogen receptor-negative (ER−) pregnancy-associated breast cancer (PABC).

2. Results

2.1. Day 28 AI vs. GI Comparison of Mammary Glands

Previous findings revealed that AI and GI mammary glands were histologically similar on day 28 postpartum, despite GI mice still being housed with pups. To determine if the mammary glands were still producing milk, we examined the gene expressions of two major milk proteins overtime in both AI and GI glands. Casein beta 2 (CSN2) decreased in the AI glands the day after offspring removal (day 8.5) and continued to decrease over time. The GI glands had an increase in CSN2 expression on day 12 postpartum that began to gradually decrease on day 17 postpartum (Figure 1b). Comparison of CSN2 on day 28 showed that the GI glands had a 10.5 fold increase in expression compared to the AI glands (Figure 1c). Similar effects of AI and GI are shown with whey acidic protein (WAP). The overall gene expression of WAP on day 28 was lower in the AI and GI glands compared to CSN2 (Figure 1d). However, the GI glands had a 47.5 fold higher expression of WAP than the AI glands (Figure 1e). Based on the presence of milk protein genes and the literature evidence of offspring breastfeeding past normal weaning time frames, we decided that comparing glands on day 28 was not an accurate representation of the metabolic changes induced by differences in involution. We examined the metabolic differences between the AI and GI glands based on the relative time since complete weaning of offspring: day 28 postpartum for AI glands and day 56 postpartum for GI glands.

2.2. Day 28 AI vs. Day 56 GI Mammary Glands Alterations in Adipocytes and Energy Metabolism

AI mammary glands upon unadjusted Ingenuity Pathway Analysis (Figure 2a) had a significant upregulation (p < 0.05) of numerous metabolic pathways compared to GI glands including mitochondria complex I and IV, mitochondria protein import, mitochondria biogenesis, mitochondria RNA degradation, mitochondria protein degradation, cristae formation, respiratory electron transport, oxidative phosphorylation, glucose metabolism, insulin receptor signaling, insulin secretion signaling, GLUT4 translocation to the cellular membrane, transcriptional regulation of white adipocyte differentiation, glycerophospholipid biosynthesis, and RNA/DNA biosynthesis. Adipogenesis, glycolysis, and complex III pathways were significantly downregulated in AI glands. Additionally, AI glands had a significant enrichment in cellular stress pathways such as cellular response to hypoxia, cellular response to heat stress, cellular response to mitochondrial stress, NRF2-mediated oxidative stress response, and detoxification of reactive oxygen species.

While transcriptomic analyses indicated an increase in oxidative phosphorylation and mitochondrial complexes, there was no significant difference (Figure 2b) in complex I (p = 0.4388), II (p = 0.2648), IV (p = 0.0724), or V (p = 0.2633) between groups. GI had a significantly higher concentration of Complex III than AI (p = 0.0345). A seahorse assay was used to examine the functional differences. Oxygen consumption rates were undetectable in AI glands unlike GI glands, despite identical conditions. AI had significantly higher levels of mitochondrial oxidative stress (p = 0.0002; Figure 2c) measured via flow cytometry, but there was no significant difference in whole cell oxidative stress measured by H2CDFDA (p = 0.1070; Figure 2d).

To validate the IPA and examine substrate utilization, targeted metabolomics was performed between AI and GI mammary glands (Figure 2e. There was no significant difference for glucose-6-phosphate (p = 0.3799) between groups. However, AI had significantly more L-lactic acid (p = 0.0011) and significantly less pyruvic acid (p < 0.0001). AI had significantly more citric acid (p < 0.0001) and palmitic acid (p = 0.0021), while GI had significantly more succinic acid (p = 0.0009) and L-glutamic acid (p = 0.0004).

To examine the lipid metabolic pathways, adipocyte population of AI and GI glands were analyzed (Figure 3). The AI mammary glands had a significantly larger area of adipose tissue than GI glands (p = 0.0261). The AI and GI glands did not have a significantly different number of adipocytes (p = 0.4226); however, the AI mammary glands had a trend toward a larger average size of adipocytes than GI glands (p = 0.1234). Adipocytes were categorized by size for each group. The GI glands had more adipocytes <1000 microns, and AI had more adipocytes larger than 1000 microns. Exploratory untargeted lipidomics showed differences, with GI glands (Figure 3f,g) having a higher amount of cholic acid glucuronide (nominal p = 0.0297), which has shown to be protective against breast cancer proliferative effects, and sphingomyelin (40:1; nominal p = 0.0433). The AI glands had higher concentrations of four lipid metabolites (Figure 3h–k): diacylglycerol (PGJ2/18:0; p = 0.0238), sphingomyelin (18:2/18:1; p = 0.0418), oxidized prostaglandin (O-28:1; p = 0.0332), and triglyceride (12:0/16:0/18:0; p = 0.0007). Figure 3f–h show nominal p-values.

Overall, these findings show that AI glands have an upregulation of transcriptional oxidative phosphorylation and an increase in relative amounts lactic acid, which may be related to mitochondrial and oxidative stress. Upregulation of these pathways and metabolites may contribute to an increase in adipocyte size and support RNA/DNA synthesis for cell proliferation. The data support that GI mammary glands may use multiple pathways for energy production (glycolysis, TCA cycle) and smaller adipocyte accumulation.

2.3. AI Long-Term Metabolic Alterations

Given substantial differences in mammary gland metabolism and mitochondrial oxidative stress, the AI and GI glands were evaluated at day 120 (long-term effect). The Unadjusted Ingenuity Pathway Analysis of day 120 transcriptomic data revealed several metabolic pathways impacted by AI compared to GI (Figure 4a). The AI glands showed significant downregulation of metabolic pathways (p < 0.01 for all) including oxidative phosphorylation, mitochondrial translation, glycolysis, gluconeogenesis, mitochondrial protein import, serine and glycine biosynthesis, cristae formation, and mitochondrial biogenesis. There was upregulation of mitochondrial dysfunction pathway. There were no significant differences in the mitochondrial complexes (p > 0.05) (Figure 4b).

While transcriptomic analyses indicated the downregulation of multiple mitochondria and oxidative phosphorylation pathways, functional analysis showed the AI glands had significantly higher basal respiration (p = 0.0011) and maximal respiration (p = 0.0027) compared to the GI glands (Figure 4c–e). There was no significant difference between the two groups for mitochondrial respiration (p = 0.2190). Analysis of H2DCFDA and MitoSox revealed no significant differences between the groups, although the AI glands had a trend for higher MitoSox levels than GI (Figure 4f,g). Exploratory lipidomics analysis showed the relative abundance of oxidized phosphatidic acid (30:3-O2) was higher in the AI glands than in GI (nominal p = 0.0476; Figure 4h).

Collectively, day 120 AI metabolic profiling demonstrated that transcriptional profiles were largely opposite of day 28, though the functional changes did not corroborate these findings. This may indicate that transcriptional changes precede the functional outputs.

2.4. Metabolic Pathways in Human ER-Negative Pregnancy-Associated Breast Cancer Similar to Day 120 AI Mammary Glands

To translate our findings and understand the metabolic differences in breast tumors following pregnancy in humans, a dataset examining the gene expression of tumors categorized as PABC was analyzed. Epithelial and stroma tumor gene expressions were compared and analyzed by Ingenuity Pathway Analysis software (Version 153384343) to identify differentially enriched metabolic pathways (Figure 5a). Unadjusted analysis of PABC ER− tumors displayed a significant enrichment (all p < 0.05) in several metabolic pathways such as mitochondrial dysfunction, triglyceride metabolism, mitochondrial biogenesis, and cholesterol biosynthesis. PABC ER− tumors had significantly reduced pyruvate metabolism, fatty acyl-CoA biosynthesis, signaling by insulin receptor, mitochondrial translation, and integration of energy metabolism. There were significant differences between PABC ER− and ER+ tumors with respect to insulin secretion signaling, amyloid processing, serine biosynthesis, and mitochondrial division signaling. Several pathways altered in PABC ER− tumors overlap with pathways in the day 120 AI glands. These results were compared to pathways impacted by AI on day 120 (Figure 5b). Several metabolic pathways were found down regulated in both PABC ER− tumors and AI MG including mitochondrial translation, oxidative phosphorylation, mitochondrial protein import, and cristae formation. Mitochondrial dysfunction was upregulated in both PABC ER− and AI MG.

2.5. Blocking of Estrogen Receptor with Tamoxifen Study

The literature on involution has shown that the estrogen return during AI leads to neutrophil infiltration and extracellular matrix remodeling that directly and indirectly promotes adipocyte redifferentiation, as previously discussed. In addition to the metabolic pathways impacted by AI shown in Figure 2a, Ingenuity Pathway Analysis of day 28 AI glands (Figure 6a) showed the upregulation of estrogen signaling, neutrophil degranulation, and several RNA/DNA synthesis pathways compared to the GI glands. Comparison of day 28 AI glands and day 56 GI glands showed that the AI glands had significantly higher ERα positivity via IHC in epithelial populations and stromal compartments than day 56 GI glands (p < 0.0001 for both) (Figure 6b). As previously mentioned, the AI glands had a larger adipocyte area and trend for larger adipocytes than GI glands. Based on the literature showing estrogen’s impact on adipocyte repopulation and upregulation of estrogen mediated signaling and neutrophil degranulation, we hypothesized that the return of estrogen during AI contributes to the metabolic changes that could support altered RNA metabolism and cell proliferation.

To test this hypothesis, a subset of FVB/n mice underwent breeding and AI as previously described. On day 8 postpartum, sustained release tamoxifen citrate (5 mg) or a placebo pellet was placed in the subscapular region, and mice were harvested on day 28 postpartum. Unadjusted Ingenuity Pathway Analysis showed that the tamoxifen- (Figure 7a) treated mice had significant (p < 0.05) downregulation of neutrophil degranulation pathways and upregulation of adipogenesis. Additionally, the tamoxifen-treated mice had upregulation (p < 0.05) of oxidative phosphorylation, respiratory electron transport, complex I biogenesis, glucose metabolism, glycolysis, and insulin receptor signaling and downregulation of glycerophospholipid biosynthesis and mitochondrial dysfunction. Western blot of mitochondrial complexes (Figure 7b) showed tamoxifen-treated glands had significantly more complex I (p = 0.0254) than placebo-treated glands. While not significant, there was a trend for tamoxifen-treated glands to have higher concentrations of complex IV (p = 0.1780) and V (p = 0.1022). There were no significant differences between groups for the number of adipocytes, total adipocyte area, or average adipocyte size (Figure 7c–f).

Based on the above findings, it is hypothesized that estrogen in AI may contribute to alterations in adipocyte redifferentiation and not the size of adipocytes. Adipocyte redifferentiation could alter the metabolic environment of day 28 AI glands, which contributes to a long-term metabolic prolife similar to PABC ER− (Figure 7g).

3. Discussion

To our knowledge, this is the first study to examine metabolic changes within the mammary gland of mice undergoing AI and GI. Expression of milk protein genes and the presence of milk lipids during early analyses showcased that while AI and GI glands were histologically similar at day 28 postpartum [10], the glands were not metabolically comparable. Extended weaning in mice has shown that the offspring will continue to breastfeed from the mother, even though majority of their nutrition comes from solid food [14,15]. Suckling plays a major role in hormones that control milk protein synthesis [16]. Comparison of day 28 AI and day 56 GI glands (similar relative time following complete weaning) showed AI led to an increase in estrogen signaling, transcriptional oxidative phosphorylation pathway and lactic acid, increased mitochondrial stress, and an increase in the metabolic pathways linked to RNA/DNA synthesis. Tamoxifen treatment in AI mice resulted in transcriptional metabolic outcomes similar to GI glands, implicating a critical role of estrogen in metabolic phenotypes observed in AI glands. Long-term transcriptomic analysis revealed the reversal of metabolic pathways shown in AI on day 28 postpartum. Several transcriptional pathways altered in day 120 AI glands were similar to PABC ER− tumors [17], suggesting a link of AI to TNBC through metabolism.

During involution, estradiol binds to estrogen receptor alpha (ERα) causing an increase in neutrophils and extracellular remodeling, which directly cause adipocyte re-population during the second phase of involution (~day 14 postpartum) [12]. Adipocytes in AI glands expand after redifferentiation due to hypertrophy of milk lipids [9]. Interestingly, our results on day 28 AI glands show similar effects with transcriptional upregulation of estrogen receptor signaling, neutrophil pathways, and larger adipocytes/adipocyte area. AI glands have an increase in glucose metabolism but downregulation of glycolysis. This was supported by targeted metabolomics, as the AI glands had higher amounts of lactic acid and lower amounts of pyruvic acid. The combination of targeted metabolite data with transcriptional upregulation of RNA and DNA synthesis and mitochondrial biogenesis pathways suggests the divergence of metabolic pathways away from energy production to cell growth [18]. These data support the previously shown increase in cell proliferation caused by AI within the mammary gland [10].

Transcriptional upregulation of the GLUT4 pathway gives insight into the cell population responsible for the metabolic phenotype: adipocytes. While confirmation studies need to be completed to validate this hypothesis, rapid adipocyte repopulation and hypertrophy have been shown to increase adipocyte glucose uptake via GLUT4, mitochondrial biogenesis, oxidative phosphorylation, protection against oxidative stress (NRF2 pathways), and promotion of cell proliferation, similar to previously shown outcomes [19,20,21]. Future studies will need to focus on isolating adipocytes from involuting glands and comparing metabolic features to whole gland metabolism.

The GI glands had a downregulation of estrogen signaling, upregulation of adipogenesis, and a subsequent different metabolic phenotype compared to AI. The results of anti-estrogen therapy, tamoxifen, in AI glands mimicked several transcriptional findings in GI glands including the upregulation of glycolysis and adipogenesis and the downregulation of neutrophil pathways. The transcriptional changes due to tamoxifen may be related to the bypass of ERα and activation of g protein-coupled estrogen receptor (GPER/GPR30) [22]. Numerous studies have indicated that while tamoxifen is an ERα antagonist, it is also a GPER agonist [22]. Activation of GPER by estrogen has shown to regulate tissue metabolism specifically by increasing oxidative phosphorylation, promoting aerobic glycolysis, and reducing adipocyte hypertrophy [23,24,25].

However, adipocytes in the mammary gland from AI tamoxifen-treated mice were similar in size to AI placebo-treated mice. Two causes for this discrepancy are plausible. Previous studies indicate the milk lipids (MCFA) are taken up into redifferentiated adipocytes via passive diffusion [9]. In AI mice, regardless of treatment, milk lipids would still be present at the time of forced weaning and need to be relocated from the collapsing alveolar structures [9]. Thus, adipocytes would fill with milk lipids regardless of tamoxifen treatment.

The long-term effects of AI reveal a reversal in transcriptional changes originally shown on day 28 AI glands. There is downregulation of energy pathways and upregulation of mitochondrial dysfunction. PABC ER− tumors showed similar affected metabolic pathways to AI glands. Both AI glands and PABC ER− tumors displayed transcriptional upregulation of mitochondrial dysfunction and downregulation of oxidative phosphorylation. PABC tumors had higher amounts of cell death and extracellular matrix remodeling similar to the previous findings of AI [10,11]. These results provide potential links between AI and ER− breast cancer, which have been linked in epidemiological studies [2,3,4,5]. As work on AI and receptor status breast cancer is being explored [17], our current study adds insight into the role of GI.

This study has acknowledged strengths and limitations. This research focused on metabolic changes over time related to AI and GI, but earlier timepoints need to be evaluated to understand the short-term impact that causes downstream long-term effects. Potential mechanisms of estrogen signaling on adipocyte repopulation provided critical information regarding estrogen’s role in mammary gland biology, metabolic outcomes, and subsequent long-term impact of its effects. Further studies need to be conducted to examine the differences in estrogen signaling between AI and GI glands and how the adipocyte repopulation process leads to metabolic differences.

4. Materials and Methods

4.1. Animals

All experiments were conducted according to National Institute of Health Guide for the Care and Use of Laboratory Animals. All protocols were approved by the Ohio State Institutional Animal Care and Use Committee (IACUC), Approval Code: 2009A0114R5, Approval Date: 12 October 2021. FVB/n female mice from Jackson Laboratory (Bar Harbor, ME) were paired for breeding at 8 weeks old. Female mice were separated prior to giving birth and housed individually. On day 0 (birth), mice were standardized to 6 pups per dam. On day 7 postpartum, mice were randomized to AI or GI. For AI, pups were removed from the dam on day 7. For GI, 3 pups were removed from the dam on day 28, and the remaining 3 pups were removed on day 31. Our model utilizes an extended weaning procedure for GI mice (day 28–31) in an effort to mimic long-term breastfeeding in accordance with the World Health Organizations (WHO) recommendations of breastfeeding for two years [14]. Research supports that rodent offspring will continue to periodically nurse when housed with the mother, even though offspring predominantly consume solid food [14,15]. Mice were harvested around 28-, 56-, or 120-days postpartum (Figure 1a). Supplementary File S1 provides detailed information on harvest and postpartum days. Mice were harvested in estrous, metestrous, or pro-estrous stage, avoiding di-estrous. To stage the estrous cycle, the mouse vagina was flushed with 20 µL of 1X sterile phosphate buffered saline (PBS). The flushed contents were collected and placed on a slide. The slide was fixed and stained with the three step Quick Dip solution (#J0322A1, #J0322A2, #J0322A3; Jorgenson Laboratories LLC, Loveland, CO, USA). Slides were viewed under a microscope to confirm the estrous stage.

4.2. Treatment with Tamoxifen

A subset of FVB/n mice underwent breeding and AI. On day 8 postpartum, AI mice had either sustained release tamoxifen citrate (5 mg) or a placebo pellet placed in the subscapular region for 21 days. Mice were harvested on day 28.

4.3. RNA Extraction

The mammary gland (50 mg piece) was cut and homogenized in Trizol using Precellys Lysing Kit (Bertin Technologies, Montigny-le-lBretonneux, France) and Precellys Evolution Homogenizer. Total RNA was isolated following Trizol-RNA isolation protocol. RNA was subjected to clean-up and concentration (Norgen Kit #23600; Biotek Corp., Thorold, Ontario, Canada) per protocol. The RNA concentration and purity were determined using Nanodrop (ThermoFisher, Waltham, MA, USA).

4.4. Gene Expression Analysis

RNA extracted from mammary glands on days 28, 56, and 120 were analyzed via Affymetrix Clairom D Mouse Assay (Applied Biosystem; Waltham, MA, USA) by Genomic Shared Resource at Ohio State University Comprehensive Cancer Center. The quality and purity were validated through tapestation analysis (n = 3/group). The data were quality checked, and the gene expression was analyzed using Transcription Analysis Console software (Version 4.0; ThermoFisher).

4.5. Ingenuity Pathway Analysis

Ingenuity Pathway Analysis software (Qiagen, Hilden, Germany) was used to compare the gene expression data between AI and GI for impacted pathways. Pathways were selected based on the previous pathways mentioned in the literature, metabolism, and energy production. Pathways having a p-value < 0.05 were reported. The BH/FDR results are provided in Supplementary File S2.

4.6. qPCR

Total RNA extracted from mammary glands on days 28, 56, and 120 was utilized to synthesize cDNA using a High Capacity cDNA Reverse Transcription kit (Applied Biosystems). qPCR was run (n = 4–5 per group) using SYBR green (Biorad; Hercules, CA, USA). Primers were purchased through Millipore-Sigma (Burlington, MA, USA). The primer sequences are provided in Supplementary File S3. Primer fidelity was confirmed using agarose gel electrophoresis.

4.7. Western Blot

Mammary gland samples (n = 4–5; 50 mg) were added to RIPA buffer with 0.1 M PMSF, 1% v/v phosphatase inhibitor cocktail 2 (P-5726; Sigma-Aldrich, St. Louis, MO, USA), phosphatase inhibitor cocktail 3 (P0044-5; Sigma-Aldrich), and protease inhibitor (P8340; Sigma-Aldrich). Tissues were homogenized using Precellys lysing, allowed to sit on ice for 30 min, and centrifuged at 16,100× g for 10 min at 4 degrees Celsius. Protein was estimated by a Pierce BCA protein assay kit #23225 (ThermoFisher). The mammary gland lysates were separated by SDS-PAGE buffer and transferred to PVDF membranes. Protein was immunoblotted with primary anti-bodies: Ox-phos Rodent WB Cocktail (45-8099; Invitrogen; Waltham, MA). Blots were incubated with corresponding secondary antibodies IRDye 680 RD donkey anti-mouse IgG or IRDye 800 CW goat anti-rabbit IgG and developed Odyssey CLX (Licor; Lincoln, NE, USA). Images were quantified via Image Studio Version 5.3 (Licor). Full blots are available in Supplementary File S4.

4.8. Immunohistochemistry (IHC)

Mammary glands (n = 3–5/group) were fixed in 10% neutral-buffered formalin for 72 h and paraffin embedded. Ten micron sections were cut and fixed on glass slides. Sections were stained utilizing anti-F4/80 (1:500 Invitrogen MF48000) and anti-ERα (1:2000, Abcam ab32063). For adipocytes measurements, F4/80 slides were imaged using an EVOS 5700 microscope (Thermo Fisher Scientific, Waltham, MA, USA), and adipocytes were quantified using ImageJ (NIH). ERα slides were analyzed using Vectra Microscope (Akoya Biosciences, Marlborough, MA, USA) and InForm Software version 1 (Akoya Biosciences).

4.9. Targeted Metabolomics

Using Progenesis QI, pooled sample runs were selected for feature alignment with internal standard signals of heavy labeled ^13^C_3_-lactic acid (Avanti Polar Lipids) used to normalize between samples. Inguinal mammary glands (50 mg) were processed for metabolomics analyses using Liquid Chromatography-column isolation (n = 3/group). Tissue samples were weighed, and extraction solution was added (MeOH:Chloroform 1:1) at 200 mg/mL. Extractions were performed using a C18 column (non-polar e.g., lipids) and a Hilic column (polar e.g., glycolytic metabolites) to capture non-polar and polar metabolites. Metabolomics relative intensity measurements were performed using QTOF. All analyses were performed using XCMS online tools. For targeted metabolomics, labelled internal standards for lactic acid, succinic acid, citric acid and palmitic acid (Cambridge Isotope Laboratories, Inc., Tewksbury, MA, USA) were used to quantify these metabolites in the AI and GI mammary glands. The false discovery rate (FDR) was calculated using the Benjamini–Hochberg method, and corrected p-values are presented. All samples were aligned with a score of 90% or above, and upon feature detection, t-test p-values between the groups were calculated with a cutoff of 0.05.

4.10. Sample Extraction for Untargeted Lipidomics

The extraction protocol was adapted from a previously described method [26]. Lipids were extracted from the mammary gland with n = 5 per group. The mammary gland (25 mg) was mixed with Optima methanol (250 µL), type I deionized water (125 µL) and zirconia beads (0.5 mm). Samples were homogenized using a Mini-bead beater-16 (Biospec Products, Tulsa, OK, USA) for 30 s and then vortexed for 1 min. Afterwards, 1000 µL of HPLC-grade methyl-tert butyl ether (MTBE) was added, the samples were vortexed for 10 min, and then, they were centrifuged for 5 min (4 °C, 10,000 rpm) using a Microfuge 22R Centrifuge (Beckman Coulter, Brea, CA, USA). The supernatant was transferred to a fresh tube. The lipid extraction was repeated once with the MTBE. The MTBE fractions were pooled, dried under argon gas, and stored at −80 °C.

4.11. UHPLC-MS Untargeted Lipidomics

Extracts were reconstituted in 150 µL of acetonitrile/isopropanol (7:3, v/v), followed by 1 min of vortexing and 5 min of centrifugation (4 °C, 14,000 rpm) using a Microfuge 22R Centrifuge (Beckman Coulter). Extract components were separated with a C8 column (Acquity Plus BEH, Waters, Milford, MA, 100 mm × 2.1 mm, 1.7 µm particle size) on an Agilent 1290 UHPLC coupled to an Agilent 6545 quadrupole time-of-flight mass spectrometer (Agilent Technologies, Santa Clara, CA). Samples were ionized using an ESI probe operated in positive mode, followed by negative mode. Chromatographic separation conditions and source ionization parameters were shared previously [26]. The injection volume was 5 µL. Quality control samples (QC) were analyzed every 6th injection to correct for instrument performance. Blanks were analyzed before running the samples and analyzed every 25th injection to remove persistent contaminant features from the data. Iterative MS-MS analysis on one QC was performed to capture MS2 data for assistance with metabolite identification.

4.12. Lipidomics Data Processing

Raw UHPLC-MS data were processed using Agilent Profinder (version 10.0) to extract and align molecular features and group by metabolite, using the following parameters: noise ≥ 10,000 counts, retention time tolerance of 0.1 min, and a 10 ppm cutoff set for binning and alignment. Post-processing filtration was applied to eliminate metabolites with intensity < 30,000 counts and to eliminate metabolites present in the process blanks. The Lipid Annotator software was used to process the iterative MS/MS files to generate a database, in order to annotate the previously extracted compounds using Agilent Mass Profiler Professional software. The Benjamini–Hochberg method was used to correct FDR, with p = 0.00077 as significant. No lipid metabolites were significantly different after FDR. Lipid metabolite nominal p-values were presented.

4.13. Mitochondria and Whole Cell Reactive Oxygen Species

Mammary glands without lymph nodes from days 28, 56, and 120 postpartum (n = 4–5/group) were digested with Type 1 Collagenase (Worthington Biochemical, Lakewood, NJ) and hyaluronidase (Millipore-Sigma) in Hank’s Balanced Salt Solution (HBSS) (Gibco, Evansville, IN, USA) with 2% FBS. Cells were counted via Luna II (Logos Biosystems, Annandale, VA, USA). The redox-sensitive fluorochrome 5-(and 6)-chloromethyl-2′,7′-dichlorodihydroflurescein diacetate dye (CM-H_2_DCFDA) (Invitrogen) was used to measure the intracellular reactive oxygen species. Mouse MG (n = 5/group) cells were treated with Mitosox-Red, and the intracellular ROS concentration was evaluated at 6 and 24 h. In total, 5 × 10^4^ treated/untreated cells were loaded with 2 μM CM-H_2_DCFDA for 20 min at 37 °C. Before analysis, the cells were removed from the loading buffer and incubated in growth medium for 30 min at 37 °C. Data acquisition and analysis were performed using a Cytek Northern Light (Becton Dickinson, BD, Franklin Lakes, NJ, USA). At least 100,000 events were detected for each sample to guarantee statistical significance. The data were normalized to the number of live cells. The data were analyzed by FlowJo version 10.8 (Becton Dickinson). Gating strategy and viability were determined via positive cells, singlet/FSC-A vs. SSC-A/Live/H2DCFDA and MitoSOX. These data are provided in Supplementary File S5.

4.14. Seahorse Analysis

Mammary glands without lymph nodes from days 28, 56, and 120 postpartum (n = 3–6/group) were digested with Type 1 Collagenase and hyaluronidase in HBSS with 2% FBS. Cells were counted via Luna II. A total of 50,000 cells from each sample were plated in triplicate on Seahorse cell culture plates and incubated at 37 °C for four hours. Cells were analyzed by Seahorse Bioanalyzer XFe24 (Agilent Technologies) using an XF Cell Mito Stress Test Kit (Agilent Technologies #103015-100). Protein was extracted and quantified using the Pierce BCA assay, as previously described. The results were normalized to the protein quantity.

4.15. Ingenuity Pathway Analysis of PABC Samples

The “Molecular Signature of Pregnancy-Associated Breast Cancer (PABC)” dataset was identified on National Center for Biotechnology Information Gene Expression Omnibus (NCBI GEO #GSE31192). The gene expression profile of laser-captured tumor epithelial cells and stroma were made publicly available by Harvell et al. [17]. Affymetrix Human Genome U133 Plus 2.0 Array data were downloaded and subjected to Ingenuity Pathway Analysis, as with the mouse experiments (above). PABC estrogen receptor positive (ER+) and PABC ER− tumor gene expression was evaluated for differential metabolic pathways via Ingenuity Pathway Analysis.

4.16. Statistical Analysis

Data analyses were completed using statistical software SAS version 9.4 (SAS Institute Cary, North Carolina) and Prism (GraphPad, San Diego, CA, USA). The data were analyzed at individual timepoints between AI and GI. Unless specified above, other comparisons were made using a two-sample t-test. The significance level for all analyses was set a priori at p < 0.05, either for single comparisons or after Holm’s adjustment for multiple comparisons.

5. Conclusions

This research study shows tissue-specific metabolic effects of AI over time and the potential role of estrogen signaling. Additionally, AI leads to long-term mitochondrial dysfunction. These alterations closely mirror metabolic signatures observed in PABC ER− tumors. Together, these findings support a potential link between the lack of breastfeeding and increased breast cancer risk.