Estrogen exacerbates mammary involution through neutrophil dependent and independent mechanism

There is strong evidence that the pro-inflammatory microenvironment during post-partum mammary involution promotes parity-associated breast cancer. Estrogen exposure during mammary involution drives tumour growth through the activity of neutrophils. However, how estrogen and neutrophils influence mammary involution are unknown. Combined analysis of transcriptomic, protein, and immunohistochemical data in Balb/c mice with and without neutrophil depletion showed that estrogen promotes involution by exacerbating inflammation, cell death and adipocytes repopulation through neutrophil-dependent and neutrophil-independent mechanisms. Remarkably, 88% of estrogen-regulated genes in mammary tissue were mediated through neutrophils, which were recruited through estrogen-induced CXCL2-CXCR2 signalling. While neutrophils mediate estrogen-induced inflammation and adipocytes repopulation, estrogen-induced mammary cell death was mediated by neutrophils-independent upsurges of cathepsins and their lysosomal leakages that are critical for lysosome-mediated cell death. Notably, these multifaceted effects of estrogen are unique to the phase of mammary involution. These findings are important for the development of intervention strategies for parity-associated breast cancer.


Figure 2. Estrogen regulates a multitude of neutrophil-dependent and -independent biological processes in involuting mammary gland.
Mice at INV D1 were treated with anti-Ly6G antibody (Ly6G) or isotype control (IgG). 24h later, they were treated with vehicle control (Ctrl) or E2B for 24h (Ctrl+IgG n=3, Ctrl+Ly6G n=3, E2B+IgG n=3, E2B+Ly6G n=3). RNA-Seq data were processed and analysed with DESeq2 followed by GO over-representation analysis. A, Volcano plot for the differentially expressed E2B regulated genes in mammary gland (MG) from IgG-and Ly6G-treated animals. B, Venn diagram for the differentially expressed genes identified from the DESeq2 analysis of the RNA-Seq data. C, Top 20 Gene Ontology (GO) terms for E2B regulated genes in MG without neutrophil depletion. D, Top 20 GO terms for the E2B regulated genes in MG after neutrophil depletion. E, Top 20 GO terms for E2B regulated genes are lost as a result of neutrophil depletion.
infiltration 170 Based on the GO over-representation analysis, 63 E2B-regulated genes with fold 171 change ≥ 3 were associated with leukocyte migration and inflammation (Fig. 3A). All 63 genes 172 were no longer E2B-regulated after neutrophil depletion (in Ly6G group). This suggests that 173 the regulation of inflammation by estrogen is primarily exerted through neutrophils. Consistent 174 with the previous report [4], S100 calcium-binding protein A8 (S100a8) and A9 (S100a9) were 175 all induced by E2B in isolated mammary neutrophils using magnetic beads (Dynabeads®) 176 coupled to anti-Ly6G antibody (Fig. 3B, S100a8, p=0.0079; S100a9, p=0.0035). C-X-C motif 177 chemokine receptor 2 (Cxcr2), the receptor for C-X-C motif chemokine ligand 1 and 2 (Cxcl1 178 and Cxcl2) which was also previously demonstrated to be induced by estrogen in mammary 179 neutrophils [4], was also significantly up-regulated (Fig. 3B, p=0.0139). Aconitate in vivo and in vitro [22]. Thus, Mirt2 regulation by E2B in neutrophils indicates its involvement 190 in the moderation of the inflammatory response during mammary involution. 191 Next, the study investigated the mechanism of estrogen-induced neutrophils infiltration.
192 S100A8, S100A9, CXCL1, and CXCL2 are known neutrophil chemoattractants that promote neutrophil migration during inflammation [23][24][25]. S100A8 and S100A9 are small calcium-194 binding proteins that activate calcium-dependent signalling through receptor for advanced 195 glycation endproducts (RAGE) or toll-like receptor 4 (TLR4). Paquinimod (PAQ) is a 196 derivative of quinoline-3-carboxamide that has been shown to inhibit S100A9 activity through 197 binding with S100A9 [26,27]. The binding inhibits S100A9 dimerization or the formation of 198 heterodimer with S100A8, thereby preventing the activation of RAGE or TLR4.  A, Heatmap representation of estrogen-regulated genes associated to leukocyte migration and inflammation in neutrophils (≥ 3 and ≤ -3-fold); Experiment is conducted according to the description in Fig. 2. B, qPCR analysis of estrogen-regulated expression of Acod1, Mirt2, Trem1, Trem3, S100a9, S100a8, Cxcr2, and Il1b relative to Gapdh in isolated mammary neutrophils following treatment with or without E2B for 24h (Ctrl n=5, E2B n=5 repopulation is a hallmark of post-lactational mammary involution. Fig. 1 shows that estrogen  Since adipogenesis genes induced by estrogen appears to be mediated through the  depletion had no effect on E2B-induced increase of active form of CTSB (Fig. 5C).

301
E2B also significantly induced the expression of tumour necrosis factor (Tnf) (Fig. 5A,   302 IgG, p=0.0161; Ly6G, p=0.0343, and Suppl. Fig. 6), which is a known upstream activator of STAT3. However, E2B did not affect the levels of phosphorylated and total STAT3 (Fig. 5B). 304 This suggests that the up-regulation of Ctsb expression by E2B is a direct event independent 305 of STAT3 activation. Furthermore, E2B also up-regulated the expression of BH3 interacting 306 domain death agonist (Bid), independent of neutrophils ( Fig. 5A, IgG, p=0.0119; Ly6G, 307 p=0.0248, and Suppl. Fig. 6). Bid is a well-established pro-apoptotic marker, and its 308 overexpression has been reported to promote cell death [40,41].   Fig. 6) (Ctrl+IgG n=3, Ctrl+Ly6G n=3, E2B+IgG n=3, E2B+Ly6G n=3). Mice on INV D1 were treated with Ctrl or E2B for 48h before MG were collected for analysis; B, Western blots of cathepsin B, D and L proteins (sc, single-chain; dc, heavy chain of the double-chain form) in mammary tissue of 48h treatment (Ctrl n=3, E2B n=4). C, Western blotting analysis shows that depletion of neutrophils did not affect estrogen-induced increase of single-chain (sc) and double-chain (dc) forms of CTSB (Ctrl+IgG n=3, E2B+IgG n=4, Ctrl+Ly6G n=3, E2B+Ly6G n=3). D, Effect of E2B on protein levels of lysosomal and cytosolic CTSB and CTSL proteins after subcellular fractionation. LAMP2 is used as a lysosomal marker (s.e, short exposure; l.e., long exposure) (Ctrl n=3, E2B n=3). Data are presented as Mean ± SEM.   cell death-related genes. E2B significantly down-regulated S100a9 (p=0.0095), Cxcl2 364 (p=0.0382), and Cxcr2 (p=0.0236) while had no effect on the expression of S100a8, Clec4d, 365 Il1b, and Trem3 (Fig. 7C). The expression of Clec4e, Mirt2, and Trem1 were also analysed but 366 had no amplification in the qPCR reactions. These observations contrast with that in the 367 involuting mammary gland where the expression of all these pro-inflammatory genes were up-regulated with E2B treatment. As for the cell death-related genes, E2B up-regulated Ctsb 369 (p=0.0166), consistent with the understanding that Ctsb is an ER target gene [44]. However, 370 E2B had no effect on the expression of Bid while Tnf displays no amplification in the qPCR 371 reaction due to the low level of expression (Fig. 7D). Hence, unlike its effect during mammary 372 involution, E2B exerts an anti-inflammatory effect on the mammary gland of nulliparous mice.   395 In sharp contrast to the effect on neutrophils during mammary involution, estrogen 396 reduced mammary neutrophil infiltration in age-matched nulliparous mice. Estrogen-regulated 397 expression of cytokines such as S100a9, Cxcl2, and Cxcr2 in nulliparous mice also occurs in 398 the opposite direction as that in mice undergoing mammary involution. The observation in nulliparous mice is consistent with reports that estrogen inhibits inflammation in obesity- 400 induced mammary inflammation [45], and in Staphylococcus aureus infected bovine mammary 401 epithelial cells [46]. To our knowledge, this is the first evidence of the plasticity of estrogen 402 action on neutrophils that is shaped by the tissue microenvironment in an in vivo model. This  Neutrophils are known to express ERα, ERβ, and GPER30 [49]. ERα and ERβ are 410 members of the nuclear receptor superfamily of transcription factors, whereas GPER30 is a G 411 protein-coupled membrane receptor. We speculate that ERα plays a major part in regulating the 412 gene expression in neutrophils during mammary involution based on the following evidence.

413
First, the relative Esr1 (ERα) expression in mammary neutrophils during involution is 414 approximately 40 times more than that in Esr2 (ERβ) (Suppl. Fig. 8). Second, ERα has been 415 reported to mediate estrogen-induced neutrophil migration in the uterus through ERα 416 phosphorylation at serine 216 [50]. Third, ERα was also reported to mediate the effect of 417 estrogen on myeloid-derived suppressor cells, which are mostly granulocytic cells, in 418 stimulating tumour development in mice model [51]. The cellular factors and the signalling 419 pathway that elicit the epigenetic changes in ERα-cistrome in neutrophils during mammary 420 involution is an interesting area for future study. 422 Expectedly, hundreds of estrogen-regulated genes through neutrophils are linked to 423 immune functions such as inflammatory response, chemotaxis, leukocytes adhesion and 424 migration. Using specific CXCR2 antagonist SB225002, this study identified CXCL2-CXCR2 425 signalling as a major pathway for estrogen to induce neutrophil infiltration. This is consistent 426 with the up-regulation of Cxcl2 [4], and Cxcr2 by estrogen in neutrophils (Fig. 3B). CXCR2 427 has been reported to be important for neutrophil infiltration in several mouse models. In an 428 acute lung injury model, Cxcr2 gene deletion abolished hyperoxia-induced neutrophil 429 accumulation in the lungs [52]. In studies of reperfusion injury, inhibition of CXCR2 with 430 repertaxin or anti-CXCR2 antibodies led to the reduction of neutrophils accumulation [53][54][55].

431
The CXCR2 antagonist SB225002 has also been reported to reduce neutrophil recruitment and 432 pro-inflammatory factor expression in LPS-induced acute lung injury [56]. Although the source 433 of CXCL1 and CXCL2 in those studies were not clear, our data clearly indicate that estrogen-  (Fig. 3A). TREM1 was first identified to be selectively expressed on neutrophils 438 and monocytes [57]. TREM3 is highly homologous to TREM1 and is believed to have 439 overlapping function. TREM1/3-deficient mice displayed impaired neutrophil trans-epithelial 440 infiltration into the lung when challenged with P. aeruginosa [58]. TREM1 was also known to 441 amplify inflammation, as TREM1 overactivation with agonistic antibodies following LPS 442 treatment led to the up-regulation of cytokines such as TNFα, MCP-1, and IL8 [57]. It is 443 plausible that Trem1/3 up-regulation is involved in estrogen-induced neutrophil infiltration. and the commitment of white fat cell lineage [60]. The present study provides the first evidence  The study provides the first evidence that estrogen accelerates cell death when there is 462 ongoing LM-PCD. This is mediated by increased expression of Ctsb, and the cytosolic protein 463 levels of active (cleaved) CTSB, CTSD, and CTSL (Fig. 5B). We propose the following model   Estrogen treatment also induces expression of Bid and Tnf gene which are reported to be involved in the induction of LM-PCD. The apoptotic protein BID is cleaved into the active tBID by the activated cytosolic CTSs. TNFα is known to induce LM-PCD via the ZnT2-mediated zinc accumulation in lysosomes, leading to PCD. The study also finds that estrogen stimulates neutrophil infiltration into the involuting mammary gland via the CXCL2/CXCR2 pathway. Meanwhile, estrogen promotes the expression of numerous proinflammatory genes such as Trem1, Trem3, Il1b, S100a8, S100a9 in neutrophils that heighten mammary inflammation. Furthermore, increased neutrophil infiltration can also recruit macrophage into the involuting gland contributing to the observed estrogen induced adipocyte repopulation. Estrogen-induced expression of genes coding for extracellular matrix remodelling enzymes such as Mmp19, Mmp3, Mmp8, Ptx3, Col8a2, Has2 further facilitate the adipocyte repopulation during mammary involution.
All animal experiments were performed in accordance with the protocol approved by Putative S100A9 inhibitor Paquinimod (PAQ) or ABR-215757 was synthesized (Suppl. 540 Fig. 9) by following the reported protocol [28]. The stock solution was prepared by dissolving 541 PAQ in DMSO to a concentration of 50mg/ml. Mice were treated with a daily dosage of 542 20mg/kg body weight via intraperitoneal injection of PAQ diluted with 1xPBS to a working 543 concentration of 5mg/ml. Control mice were treated with DMSO in 1xPBS.   Illumina adapter sequence was removed using Trimmgalore. Processed sequences were 561 subsequently mapped to the Mus musculus BALB/cJ reference genome (obtained from Ensembl) and counted using the stringtie and featurecount program. Gene annotation files were 563 also obtained from Ensembl.

564
Processed RNA-Seq data were analysed for differential gene expression using the 565 DESeq2 package with contrast method [21]. Statistically significant differential gene 566 expression was determined by Benjamini-Hochberg adjusted p-value (padj). Volcano plot 567 based on the results of DESeq2 analysis was generated using the plotly package [76]. Venn 568 diagram was plotted using the online software Venny [77]. Pathway analysis of the 569 differentially expressed (DE) gene (padj<0.05) was then conducted using the clusterProfiler 570 package [78] where gene ontology (GO) over-representation analysis was performed. The 571 enriched GO terms obtained from the GO over-representation analysis were removed of 572 redundancy using the 'simplify' function which removes highly similar enriched GO terms and 573 keeps only one representative term. The DESeq2, plotly, and clusterProfiler package was run 574 in R using RStudio [79,80].  Pefabloc, at pH7.5) and centrifuged at 750g for 10min at 4°C to remove cell nuclei and debris.

628
The supernatant was then spun at 10,000g for 15min at 4°C to pellet organelles. The pellet was 629 washed and re-suspended in subcellular fractionation buffer as lysosomal fraction. Organelles 630 were disrupted by three cycles of freezing and thawing. To collect the cytosolic fraction, the 631 supernatant collected after pelleting organelles was spun at 100,000g for 1h at 4°C to remove 632 microsomes. Protein concentration was determined by the Bradford protein assay (Bio-rad).  Antibodies used for western blot analysis are listed in Suppl. Table 2. (HyClone) and kept at 37 o C in a humidified 5% carbon dioxide and 95% air atmosphere.

672
Graphs were plotted using the mean value with the standard error of the mean (SEM).

673
When comparing 2 groups, statistical significance was determined using a two-tailed unpaired 674 student's t-test. When comparing between more than 2 groups, one-way ANOVA followed by 675 post-hoc turkey test was performed. All statistical analysis was performed using the GraphPad Mice at INV D1 was treated with anti-Ly6G antibody (Ly6G) or isotype control (IgG). 24h later, they were treated with vehicle control (Ctrl) or E2B for 24h. Flow cytometry analysis were performed on total blood cells and from digested mammary gland (MG) tissue after red blood cells lysis. A, Ly6G treatment significantly reduces circulating blood neutrophils by more than 90% while not affecting blood monocytes; Percentage of blood neutrophils (CD45+ CD11b+ Gr1 hi ) and monocytes (CD45+ CD11b+ Ly6C hi ) out of live CD45+ population. B, E2B treatment in mice given IgG increased the percentage of mammary neutrophils by 6 folds and this effect was abolished by neutrophil depletion with Ly6G. Mammary macrophages was also increased significantly by E2B treatment and was attenuated with Ly6G but to a non-statistically significant level; Ly6C hi ) out of live CD45+ cell population. B, Representative flow cytometry dot plot for the percentage of neutrophils in the MG. C, Treatment with E2B and PAQ increases the expression of some inflammatory genes as compared to E2B-treated; Gene expression of inflammatory genes Cxcl1, Cxcl2, S100a8, S100a9, Clec4d, and Clec4e relative to 36b4 by qPCR analysis. E2B n=6, E2B+PAQ n=5. Data represented as mean ± SEM. Scale bars: 50µm.