Differentiating Staphylococcus aureus from Escherichia coli mastitis: S. aureus triggers unbalanced immune-dampening and host cell invasion immediately after udder infection

The etiology determines quality and extent of the immune response after udder infection (mastitis). Infections with Gram negative bacteria (e.g. Escherichia coli) will quickly elicit strong inflammation of the udder, fully activate its immune defence via pathogen receptor driven activation of IκB/NF-κB signaling. This often eradicates the pathogen. In contrast, Gram-positive bacteria (e.g. Staphylococcus aureus) will slowly elicit a much weaker inflammation and immune response, frequently resulting in chronic infections. However, it was unclear which immune regulatory pathways are specifically triggered by S. aureus causing this partial immune subversion. We therefore compared in first lactating cows the earliest (1–3 h) udder responses against infection with mastitis causing pathogens of either species. Global transcriptome profiling, bioinformatics analysis and experimental validation of key aspects revealed as S. aureus infection specific features the (i) failure to activating IκB/NF-κB signaling; (ii) activation of the wnt/β-catenin cascade resulting in active suppression of NF-κB signaling and (iii) rearrangement of the actin-cytoskeleton through modulating Rho GTPase regulated pathways. This facilitates invasion of pathogens into host cells. Hence, S. aureus mastitis is characterized by eliciting unbalanced immune suppression rather than inflammation and invasion of S. aureus into the epithelial cells of the host causing sustained infection.

activating signaling from the toll-like-receptors (TLR) in the MEC 11,13,17 . As a consequence, challenging these cells with S. aureus will not substantially activate the NF-κB complex of transcription factors, those well-known master regulators of immune gene expression 18 .
Despite this large body of evidence regarding the pathogen-specific differentiation of the immune response of the MEC in vitro, the determinants are still elusive distinguish in vivo the very early immune response of the udder in a pathogen-specific fashion. However, those in vitro studies suggest that already the first pathogen contacts during the first hour of infection differentiate the immune response of those epithelial cells in a pathogen-specific fashion 17 .
We tried previously to elucidate in vivo pathogen-specific differences in the immune response and compared transcriptome alterations in udder samples collected in a randomized trial in which mid-lactating heifers had been infected with either E. coli 1303 or S. aureus 1027 pathogens. However, only as late as 24 h post infection (pi) did the E. coli infection result in statistically significant alterations of the expression of selected candidate genes in the milk-producing parenchyma, while the S. aureus infection had to last for 72 h or longer to yielding significant regulation of some candidate genes 19 . Global transcriptome profiling of S. aureus infected udder samples identified only 5 regulated gene loci as early as 12 h pi, but not any more from later time points. For comparison, E. coli infection had regulated the expression of 1048 loci, all at 24 h pi 4,20 . These longer term mastitis models had in addition the problem, that the udder response against E. coli infection was confounded by strong systemic responses 4,20 . Others profiled in cattle or goat the response of the udder after infecting with S. aureus only [21][22][23] . While these data show some upregulation of cyto-and chemokine encoding genes, the extent of their regulation cannot be evaluated against the full immune responsive capacity of the animals, since the direct comparison against an infection with E. coli under identical experimental settings had not been provided.
We therefore developed an alternative mastitis model by sequentially infecting udder quarters of healthy mid-lactating heifers with the same high dose (10 6 live bacteria/udder quarter) of defined E. coli 1303 or S. aureus 1027 pathogens and sampled the quarters at 1, 2 and 3 h pi. We described the model and its clinical aspects in a companion paper 24 and indicated that neither infection was accompanied by any signs of a systemic reaction (absence of fever, no alteration of blood leucocyte counts) or udder swelling and changes in the counts of somatic milk cells. We also validated, based on a limited set of inflammation-related candidate genes that infections with both pathogens had provoked significant modulations of immune gene expression and that expectedly E. coli had provoked a stronger inflammatory response than S. aureus.
We have now exploited those samples for global transcriptome profiling to get a naïve and unbiased view on the very early immune response of the udder against invading E. coli and S. aureus strains. We found that S. aureus, but not E. coli infection quickly triggers prevailing immune evasive mechanisms in the udder, through distinct immunosuppression and invasion of the pathogen into the epithelial cells of the host.

Results
Validation of microarray data. We have exploited for this study tissue samples from the gland cistern (GC) which had been collected in a previous infection trial and from which the expression of a set of candidate genes had already been determined with RT-qPCR 24 . Hence, we compared for 8 genes (MX2, IL10, S100A9, TNF, IL8, IL6, CCL20, LCN2) the previously measured data with the extent of mRNA modulation as measured in the current analysis using microarray hybridization. We found a highly significant positive correlation (r, 0.74; P < 0.0001) between both methods based on 192 individual measurements conducted with each of both methods (Supplementary Figure 1). E. coli regulated faster and more genes than S. aureus. E. coli infection modulated the mRNA abundance of 351 locus specific transcripts during the first 3 h of udder infection. The Ingenuity program related 291 of them to known genes subsequently referred to as differentially expressed genes (DEGs). In contrast, S. aureus infection changed the expression of only 122 loci 98 of which could be associated with known genes (Fig. 1A). Only 20 loci including 16 DEGs were regulated by infection with both pathogen species (Table 1). Supplementary  Table 1 lists all genes regulated exclusively by either E. coli or S. aureus infection. Sorting the regulated genes according to the family of factors which they encode shows that E. coli regulated above average cytokines, nuclear receptors, micro RNAs and transmembrane receptors ( Table 2). The regulated genes included 30 transcription factors and regulators. They included NF-κB factors and their inhibitors, C/EBP factors and NFIL3. The latter factor has previously been highlighted as a major regulator during early E. coli induced mastitis 20 . Neither of them was regulated during the S. aureus infection albeit that this challenge also regulated the expression of 8 transcription regulators. Among these, only the factors JUNB and EGR1 were also regulated during the E. coli infection ( Table 1). The comparison of the extent of changes in the mRNA abundance derived from those 16 DEGs having been regulated by both types of infection shows that E. coli induced 8 of them more strongly, while the others were induced to approximately the same extent (Table 1).
Considering the time course of immune activation, we found that E. coli modulated the mRNA abundance from only 3 loci during the first h of infection (Supplementary Table 1) including the upregulated expression IL1RN (interleukin 1 receptor antagonist) known to antagonize IL1 signaling 25 . S. aureus, however very quickly modulated the mRNA abundance stemming from 8 loci already during the first h of infection (Supplementary Table 1). The upregulated DEGs included two chemokine ligands (CXCL10, CCL8) relevant for immune cell recruitment, but also NPR3 (natriuretic peptide receptor C). The ligand for this receptor is an antagonist of TNF-function and IL6 secretion 26 . Immune dampening is also indicated by the downregulated expression of TAGAP (T-cell activation GTPase-activating protein), a known activator of T-cell function 27  E. coli infection dominantly regulated pathways eliciting inflammation while S. aureus triggered cytoskeletal rearrangements. The top Canonical Pathways having solely been addressed by the E. coli infection regulated genes were all related to inflammation and pathogen related signaling (Fig. 1C). Highest significance-probabilities were attributed to the pathways 'signaling through TREM1' (triggering receptor expressed on myeloid cells), TNFR2 (tumor necrosis factor receptor subfamily, member 1B) and IL6. They also included 'signaling mediated through pattern recognition receptors' , such as TLR2 (toll-like receptor 2) and other receptors, such as CD40 and ICAM1 (intercellular adhesion molecule 1).
Acute Phase Signaling emerged as the only from the multitude of annotated pathways having been activated by those 16 genes found to be regulated during the infection with both pathogens. Six of them (e.g. 37.5%) contribute to this pathway. The respective DEGs included in the E. coli infected samples SOCS1, SOCS3, IL1A, NFKBIE, CP, (C) Ingenuity based identification of key signaling pathways as regulated by either E.coli or S. aureus infection or by both pathogens. The ordinate indicates the logarithm of the statistical significance (p-value). The z-score represents the likelihood that the pathway is regulated (red, up-; blue, down-regulated), with more intense colors reflecting higher significance. Ratio indicates the proportion of DEGs found in the current data set relative to the total number of genes contributing to that pathway. SERPINA3, IL6, NFKB2, NFKB1, FOS, NFKBIA, SOD2, IL1B, LBP, SERPINE1, and IL1RAP. In contrast, relevant DEGs found in those S. aureus infected tissues were IL18, HP, IL1RN, CFB, and TNF.
The DEGs regulated solely by the S. aureus infection can almost all be attributed to alteration of the actin-based cytoskeleton. These pathways are anchored around RhoA (ras homolog gene family, member A) and ILK (integrin-linked kinase) signaling (Fig. 1C). The Rho family of proteins (14 members) are guanosine triphosphatases. They are crucial for remodeling of the actin cytoskeleton 28 . ILK-signaling is not only known to mediating cell adhesion and interaction with the extracellular matrix but also for the orderly accumulation of F-actin inside of the cell membrane at around the sites of integrin attachment and the formation of stress fibers and focal adhesion 29 . Indeed, 33 from among the 98 DEGs (e.g. 33.7%) regulated only through the S. aureus infection were factors related to cytoskeletal structures or their regulators (Supplementary Table 3). In contrast, E. coli infection modulated only the expression of ACTC1 (actin, alpha, cardiac muscle 1) from among all of the cytoskeleton related factors. Different upstream regulators governed the udder response towards E. coli or S. aureus. We exploited the Ingenuity pathway analysis program to interrogate the lists of DEGs regarding key upstream regulators of the complex patterns of the pathogen-specific immune responses of the udders. Considering only the 3 top scoring potential regulators, we found ( Fig. 2A) that the udder response towards the E. coli infection was guided by the effects elicited through the bacterial cell wall component LPS (lipopolysaccharide) and the well-known master cytokines TNF (tumor necrosis factor α) and IL1B (interleukin 1B). More than 100 of the E. coli specific  DEGs were found to be associated with gene regulatory networks being governed by anyone of these 3 upstream regulators. The genes involved are listed in the Supplementary Table 3A. The graphical representation of the multitude of their regulatory interdependence can only superficially visualize the high complexity of those networks (Supplementary Figure 2). Abundance of the chemical compound LPS has not been measured, but E. coli infection had increased the mRNA concentrations encoding TNF by 8 fold and IL1B by almost 14 fold (Supplementary  Table 3). These data show that inducing inflammation characterized the very early response of the udder after an infection with E. coli. The regulatory networks of those genes having exclusively been regulated during the infection with S. aureus were less complex (Fig. 2). The networks activated by the top ranking upstream regulators TGFB1 (transforming growth factor beta 1), SRF (serum response factor) and ESR1 (estrogen receptor 1) comprised only 22 DEGs or less. The expression of none of these 3 master regulators was found to be regulated in our samples. However, the high statistical significance attributed to these pathways indicates that the expression of a considerable fraction of their downstream factors was modulated by the S. aureus infection ( Fig. 2A).
The 16 DEGs regulated in common during infection with either of the pathogens are known to belong to networks being regulated by peptidoglycan, interferon γ (IFNG) or the mitogen activated kinase p38 (P38MAPK). We did not measure the abundance of peptidoglycan. The mRNA concentration of the MAP3K8-encoding gene-a representative of the MAP kinases-was 3.6 fold upregulated in the E. coli infected udders (at 3 h pi) while that of IFNG had not been changed during the udder infections. However, the relevance of all these 3 upstream regulators during mastitis is well known (see below).

E. coli and S. aureus infection induced different immune dampeners.
The strong inflammation having been induced by the E. coli infection was counter-acted by the induction of a broad spectrum of immune-dampening factors. Such factors having exclusively been induced by the E. coli infection included several TNFα induced proteins (TNFAIP2 30 ; TNFAIP3 31 , also known as ubiquitin-editing enzyme A20; TNFAIP6 32 ) the suppressors of cytokine signaling SOCS1, SOCS3 33 and the NF-κB inhibitors NFKBIA, NFKBID, NFKBIE 34 . Also the dampener of inflammasome formation mir223 35 and the inhibitor of NF-κB p65 phosphorylation NR4A1 36 have only been induced during the E. coli infection. Induction of their expression can collectively be explained by the action of those upstream regulators.  Figure S2 shows the graphical display of the highly complex regulatory network during an E. coli infection.
S. aureus infection had also triggered the expression of several factors counteracting inflammation. However, only two of them (IL1RN, CD200 37 ) had also been induced by the E. coli infection. S. aureus infection exclusively increased the levels of mRNAs encoding SLIT2 (slit homolog 2 protein), the micro-RNA miR-143 and C1QTNF3 (C1q and tumor necrosis factor related protein 3), all known to counteracting LPS induced inflammation. SLIT2 inhibits NF-κB activation 38 , miR-143 downregulates TLR2 expression 39 while C1QTNF3 is a direct antagonist of TLR4 signaling by prohibiting binding of its ligand LPS 40 . RSPO3 (R-spondin-3) 41 and CTHRC1 (collagen triple helix repeat containing 1) 42 complete the list of inhibitory factors having exclusively been induced during the S. aureus infection. Both factors are activators of the wnt/β-catenin-signaling pathway which, in turn is known to inhibit inflammation through confining NF-κB activation by the stabilizing the NF-κB inhibitory IκB-factors 43 . S. aureus, but not E. coli swiftly invaded MEC. The clear indication that S. aureus, but not E. coli infection triggered cytoskeletal rearrangement in the host cells shortly after udder infection could relate to the known high potential of S. aureus pathogens to invade udder cells. Hence, we examined if the invasion potential of the E. coli 1303 and S. aureus 1027 strains having been used for infection would be significantly different. We therefore co-cultured the mammary epithelial model cell MAC-T for 20 min or 1 h with live E. coli 1303 or S. aureus 1027 at a MOI of 10 and determined the fraction of internalized pathogens. Only a very small fraction of the E. coli pathogens was internalized after 1 h of co-culture (<1% of the inoculum) while almost 20% of the S. aureus pathogens was internalized after that time (Table 3). Indeed, already after 20 min of co-culture had 3% of the S. aureus 1027 inoculum been internalized, while no internalized E. coli 1303 pathogens could be detected after this short time of co-culture.
We compared these values to those collected from 3 more strains each of E. coli and S. aureus mastitis pathogens and found them to be representative. The efficiency of all S. aureus strains examined to invaded the MAC-T cells was higher by at least 1 or 2 orders of magnitude (internalization of approximately 10-40% of the inoculum) than that of all those E. coli strains (0.02-0.05% internalization; Table 3). The latter data are quantitatively very similar to the previously reported potential of the same E. coli strains (ECC-Z, ECA-0157, ECA-727) to invading MAC-T cells 44 . Re-calculation of the data from that report shows that no more than 0.06% of the inoculum of any of these strains had invaded the host cells after 1 h of co-culture. Hence, all the S. aureus strains very clearly invaded the MEC host cells while the capacity of E. coli strains to invade the host cells was much lower.

S. aureus infection modulated the actin-dependent cytoskeleton. We examined if S. aureus
internalization would structurally alter the actin-dependent cytoskeleton. We therefore stained actin fibers with fluorescent-labeled phalloidin in primary MEC cells having either been co-cultured for 1 h with GFP-tagged S. aureus 1027-158 or un-tagged E. coli 1303 and in untreated control cells. Laser-scanner images revealed no difference in the appearance of the actin-stress fibers in control and E. coli treated cells. These images dominantly revealed thick actin bundles in both ventral as well as transversal fibers (Fig. 3). The ventral fibers were very clearly terminated at either side by knobbed actin aggregates, conceivably indicating focal adhesion foci. S. aureus harboring cells however revealed finer and longer actin fibrils which had not so evidently been terminated by prominent focal adhesion points. These differences become quite apparent by consecutively running through optical serial sections from the ventral to the dorsal side of the cells (Supplementary video). Such serial sections reveal in particular thin dorsal fibers in the S. aureus infected cells but not in their adjacent uninfected neighbors. While these images are intriguing, it is clear that more and different experiments would be required to quantify and eventually elucidate the significance of these structural alterations.

Discussion
To the best of our knowledge, this is the first study globally profiling the very first transcriptional changes discriminating the E. coli vs S. aureus infection elicited immune responses of the udder. Hence, it gives an account of the pathogen-specific activation of innate immune mechanisms in the udder. It complements our initial description of the infection model reporting also the modulated expression of a limited set of candidate genes 24 . Our key result is that, in the absence of any recognizable systemic influences or cross-talk between infected and uninfected udder quarters, both pathogen species switched the udder response into entirely different, in some aspects Results may be representative for the immune reaction of the udder against a broader spectrum of S. aureus pathogens. The current study was based on samples from the gland cistern (GC).
Relevance of the distal udder compartments (including GC) for triggering the early immune response has already previously been highlighted by others 23,45 : It was found during udder infection with E. coli lasting for 12 h or 24 h that the immune response in this tissue occurred somewhat faster than in the lobulo-alveolar tissue but that in principle the same regulatory networks of immune gene expression had been triggered in both tissues 45 .
Similarly, long lasting udder infections with S. aureus (2-30 days) induced in both these tissues the expression of key cytokines, pathogen receptors and acute phase response factors in similar directions, albeit to different extents 23 . The lobulo-alveolar tissue comprises the mammary epithelial cells and the pathogen specific reaction norm of this cell type governs the immune response of the udder 11 . Hence, the immune response pattern of the GC tissue allows predicting with some confidence the overall pathogen-specific immune response of the udder. We have only used a single S. aureus strain for udder infections. However, different commonalities of S. aureus 1027 with other S. aureus strains together suggest that our results may be representative for a wider spectrum of S. aureus strains. We have seen in many experimental udder infections with this particular strain S. aureus 1027 that it elicited in >90% of the cases (35/38; W. Petzl, unpublished observation) a weak and subclinical mastitis just as it is commonly observed in S. aureus caused mastitis. Moreover, this strain shares with other widely used S. aureus strains (N305 Niewbold; RF122) the failure to activate in MEC model cells the IκB/NF-κB signaling cascade 46 , representing a key lesion in the activation of immune functions. Last, we have seen in the present study that S. aureus 1027 can invade MEC model cells just as well as other S. aureus strains.

E. coli infection drives inflammation through TLR-signaling. It is long known that E. coli infection of
the udder swiftly induces a strong inflammatory response 3, 19 triggered by the activation of a battery of pathogen receptors, including members of the TLR-family and other pathogen recognition receptors (PRRs) 2,13,17,19,20 . Our current data fully comply with the previous findings. The list of the E. coli infection specific DEGs shows the typical vigorous induction of cytokine gene expression 6 triggered through a prominent activation of the TLR signaling cascade. The latter is revealed by the upregulated expression of TLR2, for example and the modulated expression of several NF-κB factors and their inhibitors. All TLRs (except TLR3) and the NOD family of pathogen receptors are known to ultimately activate the NF-κB complex of transcription factors 47 . These factors are master regulators of immune functions known to regulating a wealth of immune genes 18 . The knowledge based bioinformatics analysis highlighted LPS (the authentic ligand for TLR4), TNF and IL1 as upstream key regulators governing the very early immune response in the udder. Also this aspect is entirely consistent with our current knowledge regarding E. coli mastitis 2 . Hence, our transcriptome data as related to the E. coli infection recapitulate that the key features of the E. coli specific immune response of the udder are dominated by innate immune functions of the mammary epithelial cells 11 residing in the GC. Our data add as new information that this response was triggered without a significant contribution of a generalized systemic response. However, the increased IL10 mRNA concentration in the GC tissue at 3 h pi shows that the response of some professional immune cells had been triggered, either from those residing in the udder or having already been recruited into the gland. Cells other than mammary epithelial cells, such as macrophages are the source of IL10 in the udder 11 .

S. aureus infection does not trigger classical PRR-signaling and NF-κB activation in the gland cistern. S. aureus infection did not modulate the expression level of any of the NF-κB factors or their inhibi-
tory IκB regulators. It has frequently been observed in MEC that S. aureus infection or stimulation does not lead to an activation of the NF-κB factor complex 11, 13, 17, 46 albeit it does so in professional immune cells 11,13,46 . Our study now validates that NF-κB activation is lacking in vivo in the GC tissue of the udder during an infection with live S. aureus pathogens indicating the lack of any productive TLR-signaling. We recognize the failure to activating the TLR signaling cascade in the udder during an S. aureus infection as a first crucial determinant discriminating S. aureus from E. coli mastitis. It explains in part the notorious belated and low intensity immune response of this gland to that pathogen species 2, 6 .
Upstream regulators indicate activation of immune dampening quickly after S. aureus infection. We have identified TGFB1, ESR1 and SFR as upstream regulators dominating the response of the GC tissue against the S. aureus infection. Considering the immune dampening properties of these factors we note that the immune modulating effects of the pleiotropic cytokine TGFB1 have in many instances-albeit not exclusively-been shown to be immune suppressive 48 . The immune dampening properties of this factor rely in part on down-regulating the MyD88 dependent branch of TLR-signaling through destabilizing the MyD88 protein rather than reducing its mRNA concentration 49 . Hence, high threshold levels of TGFB1 in epithelial cells 50 constantly confine TLR2 or TLR4 signaling, as clearly evidenced by the endotoxin hypersensitivity of TGFB1 null model mice 51 . We did not record any modulation of the TGFB1 mRNA levels during our short term infection in this study. However, the TGFB1 factor is secreted as an inactive pro-factor requiring proteolytic activation 52 , a process which we did not monitor in the current study. Moreover, it was recently shown in longer term experimental mastitis trials of cows that the un-infected healthy control cows featured high TGFB1 mRNA levels and these were maintained during massive S. aureus infection of the udder, whilst the TGFB1 mRNA levels were down-regulated in the udder during E. coli mastitis 4 . In support, prolonged (>48 h) S. aureus infection of the udder increases the concentration of TGFβ1 in the milk 53 .
Regarding the immune dampening properties of both other upstream regulators, it is well known that ESR1 mediated effects of estrogen are often immune suppressive 54 . SRF, signaling through the co-factor ELK4 may be immune suppressive as well, since this signaling pathway induces IL10 55 and ERG1 expression 56 . ERG1, in turn is long known to directly stimulate TGFB1 expression 57 .
Pathogen-dependent differentiation in the activation of mechanisms attenuating the immune response. The comparison of all the immune dampening features as they had been induced by either of the pathogen species reveals a pathogen-specific differentiation. E. coli induces a battery of countermeasures to directly dampen cytokine-and pathogen-signaling. The three dominant upstream regulators of the response against E. coli (LPS, IL1, TNF) have in common that they are all activating the IκB/NF-κB signaling cascade (GO ID:0007249). LPS-signaling through TLR4-and IL1-signaling through IL1 receptor-will both activate their TIR-domains and thereby trigger NF-κB activation 46 . TNF-binding to a member of the tumor necrosis factor superfamily of receptors-activates the intracellular death domain (DD) of the receptor and thereby also ultimately activates NF-κB 58 . Hence, it is not surprising to find that most of the E. coli infection specific dampeners of immune activation are interfering with the IκB/NFκB-dependent downstream signaling of those cytokine receptors and PRRs. They establish negative feedback loops of their signaling and these all are operating in the cytoplasm of the target cell. However, these immune dampening mechanisms are juxtaposed and overwhelmed by a plethora of factors and pathways promoting inflammation. Consequently, inflammation prevails during early E. coli mastitis.
S. aureus, in contrast triggers immune dampening pathways swiftly after infection which are unrelated to cytokine-and PRR-signaling. This reflects and underscores the fact that these receptors are not activated by that pathogen in the epithelial cells. These immune dampeners may either directly down-regulate PRR abundance (miR-143 related destruction of TLR2 mRNA) or quench the immune responsiveness through enhanced secretion of the immune suppressive factors RSPO3, CTHRC1 and C1QTNF3 (Fig. 4). These factors (i) not only trigger immune suppression through mechanisms being completely unrelated to those as triggered by E. coli infection, but (ii) their functioning is not restricted to the target cell. Rather, these factors will also repress the immune responsiveness of surrounding cells. Most significant for the physiology of S. aureus mastitis is the fact that their effects are not outweighed by a strong stimulation of inflammation due to diminished cytokine-and absent PRR-signaling. Hence, a key aspect of S. aureus mastitis is unbalanced activation of immune dampening mechanisms and this represents a second major discriminator from E. coli mastitis.

Rearrangement of the cytoskeleton in the host cells is a dominant feature of S. aureus infection.
Considering the actin-cytoskeleton modulating properties of the S. aureus infection-specific upstream regulators, it has been shown that SRF, signaling through the co-factors MKL1/2 (myocardin-related transcription factors) is a key regulator of cytoskeletal rearrangements in immune cells 56 . This aspect is fully supported by the current analysis of the key signaling pathways activated by the S. aureus infection and uncovering them provides deeper insights into the responsible mechanisms. Regarding rearrangement of the actin-dependent cytoskeleton, it is known that the activity of the RhoA GTPases is of pivotal importance for regulating the structure of the actin-dependent cytoskeleton 28 . Pathogenic bacteria developed different strategies to facilitate invasion into the host cell through manipulating the activity of Rho GTPases 59 . S. aureus may adhere to endothelial cells via the fibronectin binding protein 60 . This induces local β1-integrin aggregation triggering reorganization of the actin cytoskeleton via Rho GTPase activation and this allows for S. aureus invasion 61 . Augmenting this, we identified signaling through the integrin-linked kinase (ILK) also as a major player of the host defense in response to the S. aureus infection. This kinase is involved in the control of actin accumulation and rearrangement of the cytoskeleton 29 . Hence, the bioinformatics analysis of the transcriptome data lines up well with our respective experimental data. They together reveal very clearly that S. aureus induces quickly after infection rearrangement of the actin cytoskeleton and conceivably exploits it to invade the epithelial cells of the host. In contrast, E. coli infection did not induce any of these signaling pathways or structural alterations. Moreover, our comparison of the differential capacity of E. coli and S. aureus pathogens to invade the MAC-T model cell for MEC clearly validated the long known facts that (i) E. coli has a much weaker potential to invade those cells than enteric bacterial pathogen species 62 and (ii) that many S. aureus strains have a genuine capacity to invading MEC 63 . Hence, invasion of S. aureus into the host cells very quickly after infection is a third crucial discriminator between S. aureus and E. coli mastitis.
In summary, our data regarding the S. aureus specific features of the very early events after infecting the gland cistern show that the expression of three extracellular factors (C1QTNF3, CTHRC1, RSPO3) is enhanced (Fig. 4). They inhibit inflammation through blocking NF-κB activation, either directly through prohibiting pathogen recognition (C1QTNF3) or through stimulating the wnt/β-catenin signaling cascade (CTHRC1, RSPO3). The latter will modulate the activity of the Rho GTPases which, in turn will lead to rearranging the actin-cytoskeleton facilitating invasion of S. aureus into the epidermal cells of the host.
Taken together, we have first time deciphered the fundamentally different host response patterns triggered in the udder quickly after an E. coli or S. aureus infection (Fig. 5).
I. We recognize as a first crucial determinant discriminating S. aureus from E. coli mastitis the failure to activating PRR signaling cascades in the udder during an S. aureus infection. This results in a diminished inflammatory response which is not strong enough to outweigh immune dampening mechanisms. II. Hence, unbalanced activation of immune dampening mechanisms during S. aureus mastitis represents a second major discriminator from E. coli mastitis. III. The capacity of S. aureus to invading the epithelial cells of the host very quickly after infection is a third crucial discriminator between S. aureus and E. coli mastitis.

Material and Methods
Bacterial strains. The bacterial strains used for udder infusion (E. coli 1303 64 and S. aureus 1027 ) have originally been isolated from clinical cases of mastitis and their cultivation and asseveration has been described 19 . We have used them ever since in many experimental udder infections of healthy first lactating heifers. That particular E. coli strain 1303 had always elicited severe clinical mastitis within 24 h after infusing 500 CFU into a single udder quarter (n, 38). Infusing 10,000 CFU of S. aureus 1027 had caused subclinical mastitis in 35 cows while causing severe mastitis in 3 cows. That S. aureus strain has in addition been characterized at the genomic and proteomic level 65 together with the S. aureus strains RF122 (originally isolated from clinical mastitis), B and D. The latter two strains had been isolated from subclinical cases of mastitis. Prof. Susanne Engelmann kindly provided the green fluorescent protein (GFP) expressing strain S. aureus 1027-158 having been established by her through transforming Experimental setting and samples. The study was designed to compare and discriminate the earliest modulations of the transcriptome of the udder in response to E. coli 1303 or S. aureus 1027 infection based on global transcriptome profiling. It complements our initial description of the infection model including its clinical aspects, sample preparation and the modulated expression of a limited set of candidate genes 24 . The infection model featured infusing sequentially a high number (5 × 10 6 CFU/udder quarter) of either E. coli 1303 or S. aureus 1027 pathogens into three of four udder quarter of healthy HF heifers at mid lactation (n = 6 animals for E. coli and S. aureus, respectively). The trial started by infusing the pathogens into the left hind quarter. The left front and right hind quarters were infused 1 h and 2 h later, respectively. The right front quarter remained untreated and served as baseline control. Cows were culled at 3 h pi, hence well before any clinical symptoms (fever, leukopenia, clots in milk, increased counts of somatic milk cells counts [SSC], leukopenia) were detectable. Those parameters had been monitored at 1 h intervals. The fact that SSC values remained largely unchanged indicated by corollary absence of considerable cross-talk between infected and control quarters that shortly after infection. It has frequently been observed during longer lasting udder infections (24 h) with E. coli, but not with S. aureus, that the infection of a single udder quarter had increased the expression of immune genes (cytokines, bactericidal genes) in the neighboring uninfected control quarter 4,20 . Increased gene expression in the controls causes in tendency to underestimate the infection related extent of induced immune gene expression during E. coli mastitis in such experimental settings. Immediately after culling (<5 min), samples from the teat cistern, gland cistern and milk-producing parenchyma were snap frozen in liquid nitrogen and stored herein. Based on that initial profiling of the udder responses we selected for the current study the samples collected from the gland cisterns (GC) of the infected udders. Infections with both pathogen species had induced in this compartment clear and significant changes in the expression profiles of our candidate genes while little changes had been recorded in proximal locations of the lobulo-alveolar milk-producing parenchyma.

Ethics Statement.
Our previous report about this infection trial 24 included also the statement that the animal experimentation complied with the pertinent ethical standards as outlined in the German Tierschutzgesetz (TierSchG) and had been approved by the responsible ethics committee of the regional government of Upper Quantitative real-time PCR (RT-qPCR). Preparation of cDNA from total RNA with Superscript II (Invitrogen, Karlsruhe, Germany) and mRNA copy number quantification using the LightCycler and the SYBR Green plus reagent (both Roche, Basel, Switzerland) was conducted just as previously described 10 . Sequences of the oligonucleotide primers are given in the Supplementary Table 2. Spearman rank correlations between microarray and RT-qPCR results were calculated with the GraphPad Prism Version 5 software package (GraphPad Software, Inc., La Jolla, CA, USA).

Bacterial invasion assays. Quantitative assay.
Pre-cultures were started by inoculating a single colony of the pathogen into 1.5 ml Brain Heart Infusion (BHI) broth and shaking the cultures overnight at 37 °C. Next, they were diluted 1:100 and grown at 37 °C on a shaker to a density of 0.5 at OD 600 nm. Plating validated that this was equivalent to approximately 2.5 × 10 8 /ml of the S. aureus 1027 and E. coli 1303 pathogens. We used the established MAC-T mammary epithelial model cell to analyze the invasiveness of E. coli and S. aureus pathogens. Maintenance cultures of MAC-T cells were cultivated at 37 °C in DMEM medium, supplemented with 10% FCS and 4 mM L-Glutamine as described 11 . For infection experiments, the medium was changed to HEPES (25 mM) buffered DMEM without antibiotics added. The cells were grown in 24 well plates to near confluency resulting in approximately 10 5 cells/well. They were infected with 10 6 pathogens, resulting in a multiplicity of infection (MOI) of 10. Pathogens were washed off with PBS (5x) after 20 min or 60 min and the cells were incubated in growth medium for another 2 h in the presence of 100 µg/ml of gentamycin to kill any external bacteria. Subsequently, cells were lysed for 20 min through the addition of an equal volume of medium containing 0.1% Triton X100. Dilutions hereof (1:10-1:1000) were plated to determine the number of internalized pathogens. Controls monitored the adherence of the bacteria to the plastic of the culture plates as well as external adherence to the MAC-T cells.
Qualitative assay. Primary bovine mammary epithelial cells (pbMEC) were prepared from udder tissue and cultivated RPMI 1640 medium on collagen type I coated tissue plates (Greiner bio-one) as described 9 . The strain S. aureus 1027-158 , expressing the green fluorescent protein (GFP) under the control of a tetracyclin stimulated promoter was grown up overnight at 37 °C, diluted by 1:100 and then grown up to an OD 600 of 0.5 in tryptic soy broth supplemented with 10 µg/ml tetracycline. Bacteria were washed with-and re-suspended in-RPMI prior to applying them for infection. For conducting infection experiments, pbMEC were grown on collagen coated cover slips in HEPES buffered RPMI medium supplemented with 10 µg/ml tetracycline and infected with S. aureus 1027-158gfp at a MOI of 100 for 1 h. Subsequently, the cells were washed three times with PBS to remove any unbound S. aureus. The pbMEC were cultivated for another two hours in medium containing 5 µg/ml lysostaphin (Sigma Aldrich, prepared from Staphylococcus staphylolyticus) to lyse any external S. aureus. For control experiments, cover slips with pbMEC were left unstimulated or were similarly infected for 1 h with E. coli 1303 , washed with PBS and cultivated for another 2 h in RPMI medium. At the end of the experiment all cells were fixated for 10 min with 3.7% paraformaldehyde and permeabilized with 0.5% Triton X-100. Stress fibers were labelled with the red fluorescing ActinRed 555 ReadyProbes Reagent (ThermoFisher) following the standard protocol of the manufacturer. Finally, cover slips were mounted with DABCO antifading reagent (Carl Roth, Karlsruhe, Germany) and examined with a confocal laser scanning microscope (TCS SP2, Leica, Germany).