Escherichia coli strain P4-NR O15:H21H54, phylogroup B1 was isolated from the milk of an experimentally infected dairy cow (Shpigel et al., 1997; Leimbach et al., 2017; Mintz et al., 2019) and prepared for challenge in lactating mice as previously reported by us (Salamon et al., 2020). Briefly, bacteria from frozen stock were grown on Luria Bertani (LB) agar plates at 37°C for 24 h after which one isolated bacteria colony was transferred to 10 mL of LB broth for overnight growth at 37°C with shaking at 110 rpm. Pre-cultured bacteria were diluted 1:100 with fresh LB broth and grew to log phase for 3 h at 37°C with shaking at 110 rpm. Next, culture was centrifuged for 10 min at 4,000 × rpm and bacteria were washed twice with sterile PBS. Bacteria were resuspended in sterile PBS and inoculum was divided to 50 μL doses in 0.3 mL insulin syringes, approximately ~10⁴ CFUs and kept on ice until the IMM challenge. Challenge was quantified and confirmed by plating the inoculum as serial 10-fold dilutions on LB agar plates.

Streptococcus uberis strain PS10 and Staphylococcus epidermidis strain PS20 were isolated by the National Service for Udder Health and Milk Quality (NSUHMQ) laboratory from quarter milk samples originating from Israeli dairy cows affected by clinical mastitis. Bacteriological examination was performed according to National Mastitis Council guidelines (Shwimmer et al., 2007). Strains identification was further validated using 16S rRNA PCR analysis and sequencing as previously described (Wald et al., 2017). Strains were cryopreserved in 20% glycerol at −80°C until further processing. For IMM challenge, Streptococcus uberis and Staphylococcus epidermidis bacteria were grown on Heart Infusion Broth (238,400, HI, BD DIFCO) agar plates at 37°C for 24 h after which one isolated bacteria colony was transferred to 10 mL of HI broth for overnight growth at 37°C with shaking at 110 rpm. Pre-cultured bacteria were diluted 1:100 with fresh HI broth and grew to log phase for 3 h at 37°C with shaking at 110 rpm. Next, culture was centrifuged for 10 min at 4,000 × rpm and bacteria were washed twice with sterile PBS. Bacteria were resuspended in sterile PBS and inoculum was divided to 50 μL doses in 0.3 mL insulin syringes, approximately ~10⁴ CFUs and kept on ice until the IMM challenge. Challenge was quantified and confirmed by plating the inoculum as serial 10-fold dilutions on HI agar plates.

Mycoplasma bovis strain PG45 (ATCC 25523 / NCTC 10131) (Hale et al., 1962; Wise et al., 2011) was used in this study and prepared for challenge in lactating mice as previously reported by us (Schneider et al., 2022). Bacteria were propagated at 37^oC in FF modified broth medium (Friis, 1975) supplemented with 0.5% (w/v) sodium pyruvate and 0.005% (w/v) phenol red. Stock cultures were grown to the titers of 10⁸–10⁹ CFUs/mL, aliquoted, and maintained at −80^oC. For each stock, the number of CFUs per mL was determined by performing serial 10-fold dilutions in FF broth and by plating each dilution on agar in triplicates (Rodwell and Whitcomb, 1983); the agar plates were grown at 37°C, under an atmosphere of 5% CO2/95% air. For challenge, mycoplasma cultures were grown until mid-log phase and then harvested by centrifugation at 10,000 × g for 10 min at 4°C. The bacterial pellets were washed twice with the same volume of sterile PBS × 1 (Gibco1by Life Technologies) and centrifugated as described above. The bacterial cultures were resuspended in 500 μL of PBS to a final concentration of ~2 × 10¹⁰ CFUs/mL. The titers of M. bovis were calculated by plating serial 10-fold dilutions on FF agar.

Twelve to fourteen week-old female BALB/c mice were used in this project (Envigo, Israel). Intramammary challenge (IMM) was performed 8 days post-partum. Mice pups were removed before the challenge. Mice were first anesthetized (1.5–2.5% isoflurane in O₂) and then challenged with 10⁴ CFUs of E. coli P4-NR, Strep. uberis, and S. epidermidis or with 10⁹ CFUs of M. bovis PG45. IMM infusion was performed through the teat canal in both L4 and R4 abdominal mammary glands (the fourth pair found from head to tail). Three mice/six glands were used in every experiment. Injections were conducted under a binocular using 0.3 mL insulin syringes with a 33-gauge blunt needle. Mice were sacrificed 24 h (E. coli, Strep uberis, and Staph. epidermidis) or 48 h (M. bovis) post challenge, and mammary tissues were harvested for histology, total RNA extraction and total bacterial counts. Glands harvested from normal non-challenged mice were used as controls.

Mammary tissues were trisected for histology, total RNA extraction, and total bacterial counts as previously described (Mintz et al., 2019; Salamon et al., 2020). Harvested mammary tissues were weighed and homogenized in ice-cold PBS immediately after their removal and homogenates were plated as serial 10-fold dilutions on agar plates with and without antibiotics and bacterial colonies were counted following incubation at 37°C for 24 h to determine the output CFUs per 1 g of tissue. Specifically, E. coli was incubated on LB agar plates with and without chloramphenicol (25 μg/mL) for 24 h, Streptococcus uberis and Staphylococcus epidermidis were incubated on HI agar plates for 24 h, and Mycoplasma bovis was incubated on FF-modified broth medium agar plates for 3 to 5 days under an atmosphere of 5% CO₂/95% air to determine the number of CFUs/g of tissue.

Samples for histological analysis were fixed in neutral buffered 4% PFA and embedded in paraffin (FFPE blocks), and sections were cut at a thickness of 5 μm and stained with hematoxylin and eosin (H&E) according to standard procedures. For immunofluorescent staining, FFPE sections were first incubated in a 70°C oven for 20 min, dewaxed in xylene and rehydrated by sequentially immersing slides in 100, 95, 80, and 70% ethanol, and then in DDW. Following rehydration, slides were placed in Tris/EDTA buffer, pH 9, for antigen retrieval (S236784-2, Agilent Dako) and boiled for 40 min. After cooling to room temperature, slides were washed twice in TBS-T solution (Tris-Buffered Saline, pH 7.6; Alfa Aesar, United Kingdom) with 0.1% Tween-20 (Fisher Scientific) and blocked by incubation in blocking solution containing 5% (v/v) donkey serum (Sigma) in TBS-T. FFPE sections were stained with DAPI (Sigma, Rehovot, Israel), and rat anti-mouse S100A9 (ab105472, Abcam), rabbit anti-mouse CD31 (Clone D8V9E, Ionpath), and rabbit anti-mouse Na/K-ATPase (Clone EP1845Y, Ionpath) primary antibodies. Next, sections were incubated with the following secondary antibodies; goat anti-rat IgG conjugated with Alexa Fluor 488 (a11006, Invitrogen), donkey anti-rabbit IgG conjugated with Alexa Fluor 647 (a31573, Invitrogen), and donkey anti-rabbit IgG conjugated with Alexa Fluor 694 (a21207, Invitrogen).

Fresh mammary tissues for fluorescence staining were fixed in 4% PFA overnight at room temperature, incubated with 15% (wt/vol) sucrose for 72 h at 4°C, and frozen in optimal cutting temperature compound (Sigma-Aldrich). Serial 15 μm cryosections were stained with DAPI and phalloidin (Sigma, Rehovot, Israel) and rabbit anti-human Ly6G (EPR3094; Abcam), rat anti-mouse ICAM-1 (Alexa Fluor 488 conjugated, 116,112, Biolegend), goat anti-mouse myeloperoxidase (AF3667, Biotest), rabbit anti-mouse Citrullinated histone H3 (ab5103, Abcam), rabbit ant-mouse ACKR1 (MBS7005916, MyBioSource), and rat anti-mouse CD90 (ab3105, Abcam) primary antibodies. Donkey anti-rabbit IgG conjugated with Alexa Fluor 546 (A10040, Sigma Thermo Fisher), and the above described secondary antibodies were used as described above. All immunofluorescence stainings were performed in blocking solution, and mounted with fluorescent mounting medium (S3023, Dako).

Pericytes were visualized in mammary tissues of transgenic mice expressing GFP under Nestin promoter (Mignone et al., 2004). Surplus tissues were kindly shared by Dr. Dalit Sela-Donenfeld, the Hebrew University. Telocytes were visualized in mammary tissues of transgenic mice expressing GFP under Foxol1 promoter (Shoshkes-Carmel et al., 2018). Surplus tissues were kindly shared by Dr. Michal Shoshkes-Carmel, the Hebrew University.

Microscopic analysis was performed using M1 Imager Axio epifluorescence microscope with an MRm Axio camera (Zeiss), and images were captured using Zen software V3.4.

Lymph node divested L4 or R4 mammary glands were dissected from unchallenged and challenged mice and RNA was extracted for quantitative real-time PCR (QPCR) and Nanostring nCounter profiling. The qPCR analysis of mammary tissues was performed as previously described by us (Salamon et al., 2020). Briefly, total RNA was isolated from mammary tissue using the GeneElute Mammalian Total RNA Miniprep Kit (Sigma-Aldrich, Rehovot, Israel) combined with on-Column DNase I Digestion Set (Sigma). Reverse transcription was performed using qScript cDNA Synthesis Kit (Quanta BioSciences, Gaithersburg, MD, United States) and cDNA was used for subsequent qPCR reactions. PCR was conducted on a StepOne Plus PCR instrument (Applied Biosystems) using the FAST qPCR Universal Master Mix (Kappa Biosystems, Boston, MA, United States). All reactions were performed in triplicates and the gene expression levels for each amplicon were calculated using the ΔΔCT method (Livak and Schmittgen, 2001) and normalized against Ptma mRNA. Melting curve analysis was performed on each primer set to confirm amplification of a single product and all amplicons were sequenced to ensure reaction specificity (data not shown).

Using 100 ng of RNA per sample, RNA was profiled using Nanostring nCounter Mouse Myeloid Innate Immunity V2 Panel of 754 genes, including 20 internal reference genes for data normalization.¹ Data was analyzed using nSolver™ analysis software version 4.0, GraphPad, and Excel softwares. Differential expression analysis was performed with p-values adjusted for multiple comparisons according to Benjamini and Hochberg, corresponding to a false discovery rate (Benjamini and Hochberg, 1995). Differentially expressed gene sets were mapped onto gene interaction networks from STRING v.11 (Szklarczyk et al., 2018).

Gene Set Enrichment Analysis (GSEA) (Subramanian et al., 2005) was done using the java-based GSEA program and curated molecular signature database (MSigDB) both provided by the BROAD institute.² The Hallmark gene set of 50 well-defined biological states or processes (Liberzon et al., 2015), c5GOBP gene set (MSigDB; c5 Gene ontology Biological Processes), and neutrophil extracellular trap formation gene set taken from Wu et al. (2020) were used to assess the effect of bacterial challenge on murine mammary transcript expression.

Bulk gene expression data of 734 murine immune genes was produced using Nanostring nCounter technology. Ligand-receptor communication scores were calculated using the expression correlation method as previously described by Armingol et al. (2021). For comprehensive systematic analysis of putative ligand-receptor interactions, we used CellTalkDB (Shao et al., 2020) which is a manually curated database of literature-supported ligand-receptor interactions in human and mouse. The communication score (CS) was computed for every ligand-receptor pair in the gene expression data. Correlation coefficients were calculated as CS for each ligand-receptor pair across all data samples. We assumed that a communication pathway is active only when ligand and receptor were both significantly (adjusted p-value <0.05) differentially expressed with fold change (FC) greater than 1 (see representation of data analysis using Volcano and correlation plots in Supplementary Figure 1). Ligand-receptor interactions and CS were visualized using heatmaps (GraphPad) and Sankey diagrams (SankeyMATIC).³

Experimental mastitis model systems were established in lactating BALB/c mice using intramammary (IMM) infusion of bacterial suspension to glands L4 and R4 (Figure 1A). We have used E. coli strain P4-NR and the type strain M. bovis PG45 as previously reported by us (Salamon et al., 2020; Schneider et al., 2022). Field strains of Strep. uberis (strain PS10) and Staph. epidermidis (strain PS20), isolated from clinical mastitis in dairy cows, are reported here for the first time. The disease in lactating mice was characterized by high bacterial counts in the gland tissue homogenates (Figure 1B), massive recruitment of blood neutrophils into tubular and alveolar milk spaces (Figure 1C), which is further supported by increased expression of the neutrophil marker Ly6G (Figure 1D), and increased expression of inflammatory markers such as Tnfɑ, Il1β, Cxcl1 (KC), Cxcl2 (Mip2), and IkBɑ (Figure 1D).

The above described murine mastitis model systems represent the most prevalent mammary pathogenic bacteria in dairy animals; Gram-negative coliforms, Gram-positive streptococci and staphylococci, and mycoplasma. The shared hallmark of the disease caused by acute mammary infection by all these pathogens is massive recruitment of blood neutrophils into the tubular and alveolar milk spaces (Figures 1C,D). Neutrophil trafficking requires transendothelial, parenchymal, and transepithelial neutrophil migration. The structural cellular elements putatively involved in this process are schematically outlined in Figure 1C. Using immunofluorescence staining for S100a9, myeloperoxidase (MPO), and Ly6G as neutrophil markers, we demonstrate here neutrophils at each of the above stages of recruitment (Figure 2 and Supplementary Figures 2, 3). The immensely arborized alveolar and tubular epithelial duct system of the lactating mammary gland is surrounded by elaborative honeycomb-like vascular system in close proximity to all epithelial elements (Supplementary Figure 2C). The site of neutrophil extravasation and the underlying molecular mechanisms involved are still unknown in the mammary gland. In mice, the capillaries branch from approximately 10 μm lumen diameter vessels to reach an average of <6 μm diameter, then converge into larger post-capillary venules with diameter > 6 μm, that project to ascending venules (Fujiwara and Uehara, 1984; Kucharz et al., 2021). Based on size and the pan-endothelial cell marker, CD31, the mammary vascular system can be visualized and identified (Figures 2B, 3A and Supplementary Figure 2C). Moreover, as previously reported, we show here steady state expression of the adhesion molecule ICAM-1 by these blood vessels (Figures 3A,B), and their surrounding networks of pericytes (Nestin⁺ and CD90⁺) and telocytes (Foxl1⁺) (Figures 3C,D; Fujiwara and Uehara, 1984; Gherghiceanu and Popescu, 2005; Klein et al., 2022). Post capillary venules, which are considered a specialized site of immune cells recruitment, were demonstrated using the ACKR1 marker (Figure 3E; Thiriot et al., 2017; Girbl et al., 2018).

To better understand the molecular mechanisms underlying the above described inflammatory process and neutrophil recruitment, we next performed immune profiling of mammary glands challenged with E. coli, M. bovis, and Strep. uberis bacteria. These strains were chosen as representatives of mammary pathogenic Gram-positive, Gram-negative, and mycoplasma species, while Staph. epidermidis was not further analyzed in this study. We used the Nanostring nCounter technology to profile the expression of 734 immune genes in challenged glands in comparison with normal unchallenged glands. Total and differential gene expression (log₂ fold-changes) in E. coli, M. bovis, or Strep. uberis challenged glands are presented and compared in Figure 4. Cluster analysis showed higher similarity in gene expression following challenge with M. bovis and Strep. uberis than following E. coli challenge (Figure 4A). MA (Figures 4B–D) and Volcano (Figures 4E–G) plots present normalized expression and relative expression for each challenge organism. In the E. coli challenged glands, 255 genes (159 upregulated and 96 downregulated) were significantly differentially expressed in comparison with the normal unchallenged glands (adjusted p-value <0.05). In the M. bovis challenged glands, 141 genes (107 upregulated and 34 downregulated) were significantly differentially expressed in comparison with the normal unchallenged glands (adjusted p-value <0.05). In the Strep. uberis challenged glands, 167 genes (139 upregulated and 28 downregulated) were significantly differentially expressed in comparison with the normal unchallenged glands (adjusted p-value <0.05). Next, we used correlation scatter plots (Figures 4H–J) and Venn diagram analysis (Figures 5A,B) of upregulated and down regulated differentially expressed genes to analyze the similarities and differences in mammary immune gene response to each pathogen. Scatter plots present all significantly changed genes (P_adj-value <0.05; |log₂ fold-change| >1) in either challenged gland, while non-significant log2 fold-changes are assigned the value of 1 and are located on the x-axis or on the y-axis (log2 = 0 in Figures 4H–J). Taken together, our analysis revealed a core of 100 genes which are similarly regulated in E. coli, M. bovis, and Strep. uberis challenged glands (Figures 5A–D) and based on STRING analysis functionally and/or physically interacting (Figure 5C) in response to the challenge with each pathogen.

To further explore the immune processes associated with these differentially expressed genes, we next performed Gene Set Enrichment Analysis (GSEA) of our expression data comparing bacterial infections to control samples (Figure 6). Significantly enriched gene sets common to the three pathogens are presented in Figure 6A and include TNFɑ signaling via NFkB, Interferon gamma and alpha response, IL1-mediated signaling, myeloid leukocyte response, and IL6-JAK-STAT3 signaling. Moreover, since neutrophil extracellular trap (NET) genes set is not available in MSigDB, we created a new gene set based on previously published data (Wu et al., 2020). GSEA revealed significant enriched expression of genes associated with NET in glands challenged by all three pathogens (Figure 6B). This finding was further validated using immunofluorescence staining for citrullinated histone 3 (CitH3; Figure 6C), MPO and DNA staining (Supplementary Figure 4) of mammary tissues infected by each of the three pathogens. CitH3, combined with MPO and DNA staining, is a specific marker of neutrophils forming NET (Radermecker et al., 2022; Sanders et al., 2022). We show here that some, albeit not all, neutrophils recruited from blood into the alveolar and tubular milk spaces form NET in response to IMM infection with E. coli, M. bovis, and Strep. uberis.

To better understand the molecular mechanisms unveiled by our expression data, we applied computational analysis of ligand-receptor (L-R) gene pairs (Figure 7). While many putative L-R interactions were pathogen-specific (Figures 7A–C), our analysis revealed interactions shared by the three pathogens (Figure 7D). These include interaction of the cytokines IL1β, IL1ɑ, TNFɑ, and Csf3 with their receptors (Il1r2, Tnfrsf1b, and Csf3r, respectively), chemokines CCL3 and CCL4 (with CCR1), and complement C3 and ICAM1 (with Itgam). Normalized gene expressions of C3, Icam1, and Itgam (Cd11b) are presented in Figure 7E. The high expression of ICAM-1 in blood vessels of infected mammary glands was further validated using immunostaining and epifluorescence microscopy (Supplementary Figure 5).