Resident macrophages acquire innate immune memory in staphylococcal skin infection

Staphylococcus aureus (S. aureus) is a common colonizer of healthy skin and mucous membranes. At the same time, S. aureus is the most frequent cause of skin and soft tissue infections. Dermal macrophages (Mφ) are critical for the coordinated defense against invading S. aureus, yet they have a limited life span with replacement by bone marrow derived monocytes. It is currently poorly understood whether localized S. aureus skin infections persistently alter the resident Mφ subset composition and resistance to a subsequent infection. In a strictly dermal infection model we found that mice, which were previously infected with S. aureus, showed faster monocyte recruitment, increased bacterial killing and improved healing upon a secondary infection. However, skin infection decreased Mφ half-life, thereby limiting the duration of memory. In summary, resident dermal Mφ are programmed locally, independently of bone marrow-derived monocytes during staphylococcal skin infection leading to transiently increased resistance against a second infection.


Introduction
Staphylococcus aureus (S. aureus) is an important human pathogen, which causes severe invasive infections including osteomyelitis, endocarditis, pneumonia, and septicemia (Noskin et al., 2007). On the other hand, S. aureus stably colonizes the anterior nares and moist areas of the skin in up to 30% of healthy people with a high fidelity of clones (Verhoeven et al., 2014). Accordingly, S. aureus and its individual human host usually maintain a stable partnership with high adaptation at the mucocutaneous membranes. Specific and lasting immunity to virulent microorganisms has traditionally been assigned to the adaptive immune system, more specifically to long-lived memory T cells and B cells (Ahmed et al., 2009). However, in the case of S. aureus, high frequencies of S. aureus-specific antibodies and circulating T cells do not safely protect from invasive infections (Stentzel et al., 2015). Moreover, humans and mice with global defects in lymphocyte development do not exhibit an exquisite susceptibility to S. aureus infection, despite the frequent encounters outlined above (Schmaler et al., 2011). In contrast, infections with S. aureus are a primary indicator of a deficiency in cellular innate immunity, that is MyD88/IRAK4 deficiency and chronic granulomatous disease (Feuerstein et al., 2017).
Accordingly, several studies have identified Mj as being essential for the rapid and coordinated defense against S. aureus invading the skin (Feuerstein et al., 2015;Abtin et al., 2014). The concept of innate immune memory, where microbial effectors induce mechanisms in mononuclear phagocytes which increase their ability to protect against infections, offers an attractive solution to this conundrum Chan et al., 2017;Schrum et al., 2018;Yoshida et al., 2015;Kleinnijenhuis et al., 2012;Cheng et al., 2014;Netea et al., 2011). As an example, priming with LPS boosts resistance to S. aureus 3 weeks after injection (Yoshida et al., 2015). More specifically, it was very recently reported that staphylococcal soft tissue infections induce a more robust immune response to a second challenge (Chan et al., 2018). Mechanisms for the establishment of innate memory involve phosphorylation of the stress-response transcription factor ATF7, epigenetic programming through histone modifications leading to stronger gene transcription upon re-stimulation as well as metabolic changes of Mj via mTOR-and HIF-1a-mediated aerobic glycolysis (Yoshida et al., 2015;Cheng et al., 2014). These observations have indeed largely reshaped our understanding of immunological memory, which can currently best be conceptualized as multidimensional and gradual (Pradeu and Du Pasquier, 2018). Despite the increasing consensus that immunological memory in myeloid cells impacts on host defense, its contribution to infections in specific compartments remains unclear. As an example, certain organisms, for example mycobacteria, appear to alter transcriptional programs on the stem cell level (Kaufmann et al., 2018). This seems particularly important for the memory of tissue Mj resident in submucous tissues like the dermis, which are seeded prenatally and are then rapidly replaced by monocyte-derived Mj. In staphylococcal skin infection, the dermal Mj subset composition changes due to infiltrating Ly6C hi monocytes (Feuerstein et al., 2015). Scrutinizing whether, and if so how, these dynamic alterations impact on the memory response of the individual Mj and of the skin as an organ seems essential to unravel target cells and genes and to ultimately modify local host resistance. Thus, we have analyzed in an intradermal S. aureus infection model, whether, in which cells and how an 'S. aureus memory status' develops and how this impacts on Mj subset composition in the dermis, as well as on the response to a second infection with the same pathogen in vivo.
We found that staphylococcal skin infection directly induced a memory signature in individual dermal Mj, signified by the increased expression of STAT1 and CXCL9, but at the same time increased Mj turnover by naive bone-marrow derived cells, which eventually extinguished the memory. Since colonization only partially induced memory, it is tempting to speculate on minor penetrating infections to be a prerequisite for an efficient and lasting local secondary response.

Intradermal S. aureus infection increases resistance to secondary infection
In order to analyze the impact of an S. aureus infection on immunity against a second infection, mice were intradermally (i.d.) infected with S. aureus (primed), were left untreated or received a vehicle injection. After 3 weeks, the mice were (re-)infected with the same inoculum ( Figure 1A  . Bacterial burden and myeloid cell composition of the skin were analyzed after 5 days of (re-)infection, when initial neutrophil (PML) recruitment ceased and the bacterial load was declining (Feuerstein et al., 2015). 5 days after infection incoming Ly6C hi dermal Mj can be distinguished from resident Ly6C lo cells (within the CD64 hi population) based on a 'waterfall' of Ly6C and MHC2 expression (Figure 1-figure supplement 1B). 3 weeks after primary S. aureus infection, that is immediately before the second infection, CD64 hi dermal Mj (gating strategy depicted in Figure 1-figure supplement 1A) were increased ( Figure 1C), whereas S. aureus could not be cultured and PML were virtually absent ( Figure 1H). Previous infection with S. aureus (primed) significantly increased bacterial killing and was associated with fewer neutrophils in the tissue 5 days after the 2 nd infection ( Figure 1B, Figure 1D). Next, we determined the kinetics of dermal myeloid cell recruitment during secondary infection with S. aureus in naïve and primed mice at five dpi. In addition to the reduced number of neutrophils, primed mice had less Ly6C hi skin macrophages compared to naïve mice ( Figure 1E). As shown in our previous work, Ly6C hi CD64 hi CX 3 CR1 int skin   Mj are progeny of recruited Ly6C hi monocytes and will eventually differentiate into resident Ly6C lo Mj (Feuerstein et al., 2015;Kolter et al., 2019). In previously infected mice, we observed a slight increase of Ly6C lo dermal Mj and an increase of Ly6C hi dermal Mj already 24 hr after (re-)infection while by day 5 Ly6C hi dermal Mj were lower in primed mice ( Figure 1E, Figure 1F). In contrast to Ly6C hi dermal Mj, PML counts were largely unaffected one day after secondary infection. However, increased bacterial killing was already apparent one day after reinfection ( Figure 1G). To further explore, whether the enhanced antimicrobial properties after (re-)infection were due to ongoing inflammation after the primary infection or rather recall phenomena, we analyzed bacterial load, neutrophil counts and histological alterations for three weeks after primary infection. We found that bacteria were no longer detectable and that skin infiltrating PML, Ly6C expression and tissue edema were back to base line 2 weeks after infection ( Figure 1H).
Infection-induced resistance against S. aureus does not depend on bone marrow-derived monocytes and on mature T, B, and NK cells To investigate the role of bone marrow-derived cells in the establishment of induced resistance against S. aureus, we analyzed the proportion of Ly6C hi Mj, which are putative monocyte progeny (Kolter et al., 2019), and of Ly6C lo dermal Mj during (re-)infection. During steady state, Ly6C hi dermal Mj only accounted for less than 1% of skin immune cells. After secondary infection periods of 1 or 5 days, the number of Ly6C hi dermal Mj increased to 2-3% of total skin cells in wt mice ( Figure 1F). While this increase was absent in Ccr2 -/mice, Ly6C lo Mj numbers were similar to wt mice ( Figure 2A and Figure 2-figure supplement 1A). This indicates a dynamic turnover process which is highly dependent on the respective timeline of infection. We next tested whether, despite significantly reduced numbers of inflammatory monocytes, enhanced resistance against S. aureus could be induced in Ccr2 -/mice in vivo. First, we found naïve Ccr2 -/mice to exhibit an overall higher bacterial burden after primary S. aureus infection. However, a rechallenge of previously infected Ccr2 -/mice revealed increased bacterial killing (as indicated by a reduction in S. aureus CFU), decreased infiltrating PML, and less skin IL-1b concentrations as compared to naïve counterparts 5 days after (re-)infection ( Figure 2B). To further investigate whether this memory phenotype could be transferred with bone marrow from primed to naïve mice, we injected CD45.1 mice i.d. with S. aureus or PBS. After 2 weeks, bone marrow from these mice was transferred into lethally irradiated naïve wt recipient mice. After an additional 2 weeks, recipient mice were infected with S. aureus for 5 days. In concordance with our previous finding that monocytes were not essential for innate memory after dermal S. aureus infection, we did not observe a significant difference in bacterial killing and neutrophil recruitment between recipient mice, which were reconstituted with bone marrow from infected mice, and those which were reconstituted with bone marrow from PBS treated mice ( Figure 2-figure supplement 1B).
Next we determined, whether S. aureus-induced immune memory was dependent on the adaptive immune system. Therefore, we analyzed Rag2 -/gc -/mice, which lack mature B, T and NK cells, but have similar skin Mj numbers compared to wt mice ( Figure 2D and Figure 1F), for enhanced protection against bacteria by re-exposure. Primed Rag2 -/gc -/mice showed a clear memory phenotype with increased bacterial killing, reduced PML and IL-1b tissue concentration at 5 days p.i. ( Figure 2C). Therefore, the lasting protective effects exerted by S. aureus did not depend on bone- infection and without secondary infection, data is gated on all CD64 hi macrophages. (n: three biological replicates representing three independent experiments). (E) Representative FACS plots of the dermal myeloid cell composition in naïve and primed mice treated as in (A) 5 days after secondary infection (numbers represent % of total skin cells). (F) Frequency of Ly6C lo and Ly6C hi dermal Mj 3 weeks after primary infection as well as after (re-) infection periods of 1 and 5 days in primed mice and mice infected for the first time (naïve). (G) Wt mice infected with S. aureus (10 7 CFU in 10 ml PBS) and (re-)infected 3 weeks later (primed) or for the first time (naïve) with the same amount of bacteria and analyzed for bacterial load and skin infiltrating PML after (re-)infection for 1 day (n: 7-8 biological replicates representing two independent experiments). (H) CFU counts, skin neutrophil and Ly6C hi dermal Mj numbers of WT mice up to 21 days after i.d. S. aureus infection and without secondary infection. HE staining of cryosections from wt mice 3 weeks after i.d. S. aureus infection. Data were analyzed using either two tailed Mann-Whitney test or two-way ANOVA with Bonferroni's multiple comparison test. Error bars are mean ± SEM. *p<0.05, **p<0.01, ***p<0.001. The online version of this article includes the following figure supplement(s) for figure 1:

Resident dermal M' mediate innate memory
To further demonstrate that dermal Mj were responsible for the enhanced resistance to secondary S. aureus infection, we treated previously infected mice with either liposomal clodronate or control liposomes ( Figure 3A). This method depletes about 70-80% of all CD64 hi dermal Mj within 2 days by induction of apoptosis (Feuerstein et al., 2015). Repopulation of dermal Mj occurred one week after depletion ( Figure 3B). We injected clodronate liposomes into mice 2.5 weeks after S. aureus infection to deplete resident dermal Mj. After an additional 1.5 weeks mice were (re-)infected. We found reduced bacterial killing, increased PML numbers and higher IL-1b concentrations in the skin of clodronate treated mice ( Figure 3C). This indicated that depletion of the resident dermal Mj pool abolished innate memory, although the Mj density was maintained due to invasion of monocytes. We next tested, whether infection of contralateral tissue (ear) could increase the protection as well. Interestingly, innate memory could only be observed after a re-challenge of the same ear ( Figure 3D), indicating that it was primarily mediated by local resident Mj without contribution of blood-or bone marrow-derived cells.

Innate memory alters the transcriptional signature in dermal M'
We next investigated the transcriptional changes associated with innate memory in dermal Mj. Resident dermal Mj were sorted from S. aureus infected and control Ccr2 -/mice, which were used in order to specifically analyze the memory program in resident dermal Mj, which -as shown aboveessentially mediate increased resistance to a secondary infection. Three weeks after S. aureus infection, 65 genes were differentially expressed as compared to Mj from vehicle control mice including the transcription factor Stat1 ( Figure 4A). Gene ontology analysis using the Panther bioinformatics platform revealed enrichment of terms associated with antigen presentation, response to interferons and positive regulation of cytokine production ( Figure 4B). Accordingly, intradermal S. aureus infection induced lasting changes in a distinct subset of genes which was associated with increased resistance to secondary infection.

Dermal M' show a specific memory signature and increased responsiveness after re-infection ex vivo
Upregulation of the transcription factor Stat1, as observed in the gene expression array, and its target Cxcl9, was confirmed by quantitative PCR to be increased in sorted CD64 hi Mj from wt mice previously infected with S. aureus as compared to naïve mice ( Figure 5A and B). These genes have been shown to be involved in innate memory in LPS-primed Mj, which also show an enhanced resistance to S. aureus after 3 weeks (Yoshida et al., 2015). Accordingly, phosphorylated STAT1 was increased in dermal Mj from whole tissue lysates of S. aureus infected skin ( Figure 5C). Furthermore, Mj of previously infected mice had increased expression of Stat1 and Cxcl9 1 and 5 days after (re-)infection ( Figure 5D). In addition, increased concentrations of CXCL9 were detectable 5 days after (re-)infection in skin lysates of primed mice compared to naïve controls ( Figure 5E). Notably, CXCL9 has been reported to exhibit direct antimicrobial activity independent of its chemotactic activity (Reid-Yu et al., 2015). We confirmed this observation by an in vitro S. aureus killing assay showing the direct killing capacity of CXCL9 on the S. aureus strain Newman ( Figure 5-figure supplement 1). Notably, in dermal Mj sorted from primed mice that had received clodronate (as in Figure 3A), the enhanced expression of Stat1 and Cxcl9 was lost (Figure 5-figure supplement 2A). We then analyzed the response of previously infected dermal Mj to a second challenge with S. aureus ex vivo. Accordingly, CD64 hi dermal Mj were sorted from mice 21 days after S. aureus infection or from naive mice and then stimulated ex vivo with fixed S. aureus. We found that dermal Mj from primed mice showed higher pro IL1B expression and more TNF-a production than Mj from naïve mice ( Figure 5F). This finding suggested that in vivo priming leads to increased responsiveness to the same pathogen at the single Mj level.
We have previously shown that rather subtle differences in eliciting inflammation in the earliest stages of staphylococcal infection translate into altered skin immunity for days. In these early stages, the Mj response in vivo is similar regardless of whether live or fixed S. aureus is used (Feuerstein et al., 2015). Thus, the early Mj response relies on recognition of the bacterial particle, rather than on secreted factors or active proliferation. The usage of fixed bacteria allows a focus on the Mj activation status irrespective of the bacterial killing capacity. Upon challenge of primed mice for 4 hr with fixed S. aureus ( Figure 5G), we observed significantly more skin-infiltrating PML and Ly6C hi Mj ( Figure 5H) and visible skin erythema in previously infected mice (Figure 5-figure supplement 2B). In full agreement with the above results, primed mice had increased concentrations of CXCL1 in whole tissue lysates compared to naïve counterparts ( Figure 5I). This strongly suggested that the initial response to S. aureus primed Mj for an increased responsiveness to the second infection resulting in the more rapid recruitment of peripheral blood cells. While enhanced bacterial killing occurs in the absence of inflammatory monocytes ( Figure 2B) we aimed to investigate whether Ly6C hi cells invading the infected dermal microenvironment were primed as well. As Ly6C is downregulated by macrophages upon differentiation (Kolter et al., 2019), we sorted dermal Mj from Cx3cr1 gfp/+ reporter mice 3 weeks after primary infection (without secondary infection). We sorted two populations, namely CX 3 CR1 lo Ly6C lo resident Mj and CX 3 CR1 int Mj which are partly Ly6C hi and derived from incoming monocytes. We found both subsets to exhibit increased expression of Stat1 and Cxcl9 compared to cells sorted from naïve mice. Interestingly, the expression of both genes was higher in the Ly6C hi subset, indicating that, while innate memory occurs in the absence of  incoming Ly6C hi Mj, these cells can also be primed in the microenvironment of the infected ear (Figure 5-figure supplement 3). Surprisingly, we did not find S. aureus to prime Mj from different sources (bone-marrow, peritoneal cavity) for an enhanced secondary response in vitro as established for other effectors (Schrum et al., 2018;Cheng et al., 2014;Ifrim et al., 2014). Stimulation of bone marrow-derived Mj with fixed S. aureus followed by re-stimulation 24 hr later did not enhance cytokine production. Likewise, priming of peritoneal Mj for 24 hr with fixed S. aureus followed by a re-stimulation 5 days later failed to enhance intracellular TNF production (data not shown). This indicated that the induction of innate memory in vivo requires distinct cues from the infected tissue, although monocytes and adaptive immune cells are not involved (see above).

Innate memory is a transient effect and partially independent of classical Toll-like receptor signaling and skin microbiota
MyD88 serves as a common adapter for Toll-like receptors (TLR), which are putative sensors for S. aureus and other skin commensals in dermal Mj (Feuerstein et al., 2015). In order to establish the role of Myd88 in the innate memory model, we infected Myd88 -/mice for 3 weeks and analyzed transcriptional priming (without secondary infection), as well as killing capacity and PML recruitment 5 days after re-challenge. Strikingly, primed Myd88 -/mice had reduced bacterial burden and increased expression of Stat1 and Cxcl9 compared to naïve mice ( Figure 6-figure supplement 1A and B). Furthermore, dermal Mj from primed mice exhibited a robust pro-IL1b response after S. aureus ex vivo stimulation ( Figure 6-figure supplement 1C). The above results indicate that innate S. aureus memory seems to be, at least partially, independent of MyD88.
Next we assessed the role of skin flora in macrophage priming. All mice tested so far were bred and kept under specific pathogen free conditions. These conditions did not exclude S. aureus colonization. We therefore tested our mouse strains for S. aureus colonization on a random basis and found S. aureus colonization in about one third of all tested mice (data not shown). This raised the question whether S. aureus colonization alone would be able to induce innate memory. We therefore monocolonized a cohort of sDMDMm2 mice (Uchimura et al., 2016;Brugiroux et al., 2016) -harboring a defined physiological intestinal microbiota but devoid of S. aureus -under axenic conditions with S. aureus by smearing a bacterial suspension onto their ear skin. Successful colonization was confirmed one and two weeks after initial ear coating. After three weeks, mice were infected intradermally with fixed bacteria for 4 hr. Skin colonization with S. aureus led to a 50% increase in the frequency of Ly6C hi Mj after infection of sDMDMm2 mice while it did not affect the abundance of infection-induced PML ( Figure 6A). S. aureus-infected SPF mice generally revealed higher numbers of PML and Ly6C hi Mj compared to infected sDMDMm2 mice with or without S. aureus colonization. Importantly, the increase in both Ly-6C hi Mj and PML induced by infection (SPF mice) was much greater as compared to the innate memory effect induced by skin colonization (sDMDMm2). Thus, S. aureus colonization slightly increases the response to subsequent infection, but does not induce an immunological memory similar to a dermal infection with the same organism.
A key feature of the adaptive immune system is the establishment of long-term memory, which protects the host from similar infections for months to years. In contrast, we hypothesized that innate Figure 5 continued independent experiments). (D) RNA expression of sorted dermal Mj from primed or naïve wt mice following (re-)infection periods of 1 and 5 days (n: three biological replicates from one independent experiment). (E) CXCL9 ELISA of skin lysates of primed or naïve wt mice after 5 days of (re-)infection (n: eight biological replicates representing two independent experiments). (F) Dermal Mj sorted either from S. aureus infected mice (primed) or naïve wt mice, were re-challenged ex vivo with fixed S. aureus. Expression of pro-IL-1b was determined by qPCR. Intracellular (i.c) TNF concentrations were determined by antibody staining and analysis by flow cytometry. (G-H) Wt mice were infected i.d. with S. aureus (10 7 CFU in 10 ml PBS) and (re-)infected i.d. 3 weeks later (primed) or for the first time (naïve) with fixed bacteria (10 8 CFU in 10 ml PBS) and analyzed for skin infiltrating PML and Ly6C hi dermal Mj (n: eight biological replicates from two independent experiments). (I) CXCL1 ELISA of skin lysates from mice treated as in G. Data were analyzed using either two tailed unpaired T test (B, D and F)   memory is intrinsic to the single cell and transient, that is it is non-clonal and not passed on to the next cell generation and lost, when the single trained cell is replaced. To test this, we analyzed dermal Mj from naïve and primed mice after 6 and 12 weeks as well as the innate memory status of mice which were (re-)infected after 6 weeks. Already after 6 weeks, the innate memory signature, that is the increased expression of Stat1 and its target gene Cxcl9 were lost ( Figure 6B). Furthermore, increased bacterial killing was lost in wt mice 6 weeks after infection ( Figure 6E, left panel).
To further investigate this phenotype, wt mice were intravenously treated with the PKH26 red fluorescent cell linker, which labels resident dermal Mj for up to four weeks in vivo, whereas Ly6C hi incoming monocytes remain unlabeled ( Figure 6-figure supplement 2). One week after labeling, mice were i.d. infected with S. aureus and one and three weeks after infection, PHK26 positive dermal Mj were analyzed ( Figure 6C). Compared to uninfected mice, S. aureus infection decreased the half-life of dermal Mj from an estimated 35 days to 10 days, that is by around 70%. The increased turnover of dermal Mj was confirmed by a complementary approach employing chimeras harboring bone marrow from congenic mice (B6-CD45.2 bone marrow in B6-CD45.1 mice), which allowed to discriminate previously resident Mj from Mj derived from the (transplanted) bone-marrow, combined with S. aureus infection. Eight weeks after transplantation of 10 7 wt bone marrow cells, mice were infected i.d. with S. aureus. Compared to uninfected mice, an increased replacement of the resident dermal Mj pool by donor derived cells was observed up to six weeks after S. aureus infection ( Figure 6D). We speculated that immune memory would be prolonged in Ccr2 -/mice, in which monocyte input after infection is largely absent (Figure 2A). Indeed, increased bacterial killing could be observed in Ccr2 -/mice following secondary infection in mice that had been primed 6 weeks previously ( Figure 6E, right panel), while no difference in PML numbers was observed between primed and naïve mice (data not shown). Taken together, these data indicate that S. aureus-induced innate memory is a transient phenomenon, which is lost within 6 to 12 weeks due the replacement of resident Mj by monocyte-derived Mj (model depicted in Figure 7).

Discussion
Mucocutaneous resistance against potentially pathogenic bacterial colonizers largely relies on cellular innate immunity. It seems inherently plausible that evolution has equipped this system with means to adapt to repeated microbial encounters, since they are likely to reoccur. Accordingly, it has been appreciated for over 50 years that Mj downmodulate the response to endotoxin after exposure, a phenomenon called tolerance (Greisman and Hornick, 1975). More recently, the opposite, for example propagation of cytokine transcription by a second microbial challenge, as compared to the first, has been linked to epigenetic changes and metabolic programming. Notably, much of the groundwork was done in vitro, and microbial components like b-glucan and LPS, and microorganisms, such as yeast, vaccine strain mycobacteria and herpes viruses have been found to reprogram Mj Yoshida et al., 2015;Kleinnijenhuis et al., 2012;Saeed et al., 2014;Barton et al., 2007). Very recently S. aureus has also been shown to induce a Mj dependent protective immunity in a penetrating skin and skin structure infection (Chan et al., 2018).
Systemic infections including those with focal onset, which are typically conceived by the host as sickness, and diffusible effectors are conceptually challenging, since they induce a cascade of multicellular intermediate events, including cytokines and soluble mediators. These, in turn, have the potential to change the activation state of distant immune cells and organs. In contrast, most challenges, which are imposed on the mucocutaneous innate immune system, include very local ground work, for example removal of bacteria or confinement to small abscesses, efferocytosis, tissue repair etc. They do not involve activation of systemic immunity. Here, we aimed to mimic this daily grind of dermal Mj with a discrete and strictly intradermal staphylococcal infection model, in order to better grip mechanisms underlying cell autonomous priming by bacteria residing in immediate proximity. This model recapitulates the initiation of the immune response by resident dermal Mj, the recruitment of PML and monocytes to the site of infection, the killing of bacteria and the contraction of inflammation without measurable circulation of bacteria and cytokines, and without effect on the mouse wellbeing (Feuerstein et al., 2017;Feuerstein et al., 2015). Thus, it differs sharply from subcutaneous S. aureus infections, which lead to strong inflammation, large abscesses, dermonecrosis and bacterial dissemination to organs like the spleen, kidney and liver (Chan et al., 2018). Another caveat of severe infection models with bacterial dissemination, for example to the bone-marrow, is the difficulty in excluding long term survival of bacteria in body niches. In the case of S. aureus, bacteria have been found to survive in osteoblasts from mice for months (Tuchscherr et al., 2015).
We found here that mice showed stronger PML and monocyte recruitment and bacterial killing early in dermal staphylococcal infection, if it had been preceded by an intradermal staphylococcal infection three weeks before, indicating that lasting immune priming indeed occurred at the local level. The memory effect was strictly dependent on skin resident dermal Mj, and did not require mature T, B and NK cells, as well as monocytes. The transcriptional effects of dermal Mj memory in situ were rather subtle, since intradermal S. aureus infection led to persistent transcriptional changes in only 65 genes compared to vehicle control. Yet, the circumscribed transcriptional changes acquired in vivo, which translated into an altered inflammatory response to S. aureus ex vivo, further substantiated the interpretation that Mj are primed for an enhanced secondary response on the single cell level.
It seems notable that the S. aureus-induced memory signature in dermal Mj comprised activation of STAT1, whereas type I inflammatory genes were not increased. It appears that the immediate innate immune response is fully resolved 2 weeks after infection in our model, as indicated by the absence of culturable bacteria, of skin infiltrating PML and of type I cytokines in the tissue at this stage. Among the signaling intermediates, metabolic and epigenetic changes which have been previously implicated in Mj memory Yoshida et al., 2015;Cheng et al., 2014;Saeed et al., 2014) STAT1 appears to play a dominant role (Yoshida et al., 2015;Saeed et al., 2014). It is intriguing that the STAT1-dependent chemokine CXCL9 possess direct antimicrobial activity independent of its chemotactic activity (Reid-Yu et al., 2015;Yang et al., 2003). However, whether this CXCL9 property directly impacts on the defense against S. aureus in the dermis remains to be established. A major observation was that substantial memory lasted for less than six weeks in vivo. This finding is in line with data from Garcia-Valtanen et al., who observed a decline in the b-glucan induced memory in vivo in a similar time span (Garcia-Valtanen et al., 2017). Several lines of evidence suggest that 'naive' bone marrow-derived cells dilute out the resident Mj population and thereby terminate immunological memory in local S. aureus infection. First, monocyte-deficient mice exhibited somewhat longer lasting immunological memory as compared to wt mice. Next, we did not find S. aureus to convincingly prime monocytes, bone marrow-derived Mj or peritoneal Mj in vitro. In other words, innate memory appears to require terminally differentiated Mj in the tissue environment. This is in contrast to what has been described for soluble TLR ligands and other microorganisms (Schrum et al., 2018;Cheng et al., 2014;Ifrim et al., 2014). Moreover, the memory phenotype could not be transferred by bone-marrow transplantation. However, the transfer experiments showed that S. aureus infection led to a strong redistribution of the skin myeloid cell subsets with high numbers of infiltrating bone marrow-derived monocytes. This indicated a strong dilution of resident dermal Mj within 6 weeks. Of note, recipient mice were irradiated without shielding of the heads, a method to protect resident cells of the ear skin from irradiation damage. This could hinder proliferation of resident dermal Mj after transplantation and after S. aureus infection, when it is normally high (own unpublished data) and could putatively maintain innate memory conditions. We therefore simulated S. aureus-induced changes of the dermal Mj compartment with an additional in vivo model where we labelled specifically resident dermal Mj, and not bone marrow-derived monocytes (own observations), with the phagocytic cell dye PKH26. We found that the normal half-life of dermal Mj (35 days in homeostasis) was shortened to approximately 10 days after S. aureus infection.
The skin constitutes a remarkable host-microbe interface, where micrometers decide between harmless colonization by S. aureus, which occurs in up to 30% of humans (Noskin et al., 2007) and invasive infection. However, the physical barrier of the epidermis appears to be less tight than commonly thought, since nucleic acids from S. aureus can be found in the non-lesional dermis of up to 30% of people (Crisp et al., 2015). Accordingly, we wondered, whether S. aureus, as a regular component of the skin microbiota, would subject dermal Mj to constant innate memory, thereby compensating for rapid replacement of programmed Mj. Analysis of transiently S. aureus colonized sDMDMm2 mice hinted at innate memory effects by colonization, yet the effects were modest. However, it remains an attractive hypothesis that S. aureus gets access to the dermis via disruption of the epidermis by minimal mechanical trauma and through preformed gaps, for example hair follicles. This may propagate the dynamic and flexible nature of the dermis as a large immune organ with high adaptation to colonizing bacteria.
In summary, tissue resident dermal Mj are programmed independently of bone marrow-derived monocytes during staphylococcal skin infection to mediate increased resistance against a second infection. However, the infection-induced turnover of resident skin Mj limits the duration of this innate memory response.
We propose that reprogramming of Mj serves different purposes related to pathogen-specific infection biology. In the case of colonizing or minimally invasive S. aureus, innate memory is restricted to the dermal niche and wanes over weeks. This local recall phenomenon is reminiscent of the recently reported priming of alveolar Mj in local adenovirus infections (Yao et al., 2018). On the other hand, systemic spread of staphylococci like in sepsis is relatively rare and carries a high risk of death. Thus, under these conditions, innate memory is likely futile. In contrast, microorganisms with an intracellular lifestyle and limited virulence, for example mycobacteria, may induce innate memory of cells in central components like hematopoietic stem cells in order to allow for individual adaptations in host resistance (Kaufmann et al., 2018).

Bacterial strains
S. aureus strain Newman was used. Bacteria were grown in Luria-Bertani broth medium to exponential growth phase, washed with PBS and resuspended in Dulbecco's Phoshate Buffered Saline (PBS). Bacterial concentrations were determined using an optical spectrophotometer (600 nm, OD 600 ). Colony forming units (CFUs) of the inoculum were verified by dilution series plated onto blood agar plates (COS, Biomerieux) overnight at 37˚C. In some experiments, bacteria were heat-fixed (80˚C, 30 min) after adjusting the cell number to 10 9 CFUs/ml as previously described (Feuerstein et al., 2015). For colonization studies S. aureus suspension (10 9 CFUs/ml) was smeared onto both ears of the mice under axenic conditions using sterile cotton swab. Ears of control mice were smeared with a sterile PBS solution. Successful colonization of mice was tested by swabbing the ears of mice with a PBS-soaked cotton swab and streaking directly onto mannite sodium chloride plates.

Depletion of resident skin M' in vivo
Resident skin Mj were depleted by liposomal clodronate (Clodrosome Encapsula NanoSciences) as previously described (Feuerstein et al., 2015). Approximately, 10 ml of liposomal clodronate or control liposomes were injected i.d. into the ear pinna of anaesthetized mice. S. aureus was injected i.d. into the same ear pinna 10d later. Labeling of dermal macrophages with PKH26 Red Fluorescent Cell Linker wt mice were intravenously injected with 100 ml PKH26/Diluent B mix (0.1 mM PKH26 in 100 ml Diluent B). Labeling efficiency of resident dermal Mj was analyzed by FACS and stable up to 4 weeks (data not shown). 1 week after labeling mice were infected with 10 7 CFU S. aureus i.d. into both ears and labeling or rather turnover of resident Mj followed and analyzed for additional 3 weeks by FACS.
Mouse bone marrow chimera B6-CD45.1-mice (Ly5.1) were fractionally irradiated with 2 Â 4.5 Gy and reconstituted with 10 7 nucleated bone marrow cells from congenic B6-CD45.2 mice. After 8 weeks a > 99% donor chimerism for blood monocytes was confirmed by flow cytometry analysis using CD45.1 and CD45.2 antibodies. At this timepoint, untreated mice were sacrificed and chimerism of dermal macrophages was analyzed using CD45.1 and CD45.2 antibodies (d0). The other groups of mice were infected with 10 7 CFU S. aureus in 10 ml PBS intradermally into the left ear (as described above). Treated mice were sacrificed at the indicated time points and dermal Mj were isolated and analyzed by flow cytometry as described above to determine Mj chimerism in the infected ear (S.a.) and the uninfected ear (-).

Intracellular phosphorylated Stat1 staining
Mouse ears were subjected to enzymatic digestion, filtered, washed and cells finally resuspended in 100 ml RPMI medium supplemented with 10% FBS and antibiotics (ciprofloxacin, 10 mg/ml). Then cells were stimulated for 15 min at 37˚C with 200 ng/ml IFNg and subsequently fixed with 100 ml fixation buffer (Cytofix Fixation Buffer, BD Biosciences) for 10 min at 37˚C. After centrifugation and permeabilization with 400 ml ice cold Perm Buffer III (BD Phosflow) for 30 min at 4˚C cells were washed twice with PBS containing 1% BSA. Then cell surfaces were stained with anti-mouse CD45 eFluor450 (eBioscience), anti-mouse CD11b PE-Cy7 (eBioscience), anti-mouse CD11c APC (BD Biosciences) and intracellular phosphorylated Stat1 stained with mouse anti-Stat1 (pY701) Alexa Fluor 647 (BD Biosciences) 30 min at RT and subsequently analyzed by flow cytometry.

Tissue embedding and staining
Ear skin was collected, bisected and ear halves were embedded in Tissue-Tek O.C.T. compound (Sakura Finetek Europe B.V.) and subsequently frozen inliquid nitrogen. 8 mm cryosections were taken with a Leica CM 1850 (Leica Biosystems, Nussloch, Germany) and stained with H and E. Photographs were taken on a Zeiss Axioskop 50 (Zeiss, Jena, Germany) with the Olympus LC20 camera (Olympus Germany, Hamburg, Germany).

Enzyme-linked immunosorbent assay (ELISA)
Ears were mechanically homogenised by Tissue Lyser, Qiagen. IL-1b, CXCL9 and KC concentrations in the homogenised skin tissue were quantified by ELISA according to the manufacturer's instructions (R and D Systems).

Quantification of Colony Forming Units
Mice were sacrificed. Then ears were cut off at the hair line and homogenized in a Tissue Lyser, Qiagen. CFUs in the homogenized skin tissue were determined by serial dilutions on blood agar plates (COS, Biomerieux).

CXCL9 killing assay
To determine CXCL9 antimicrobial activity, S. aureus was grown to OD 600 (0.4) and then incubated at 37˚C, with shaking, for a further 2 hr in the presence of CXCL9 (4 mg/ml or 40 mg/ml) or PBS as control. Bacteria survival was determined by serial dilutions on blood agar plates.

GeneChip microarray assay
Ccr2 -/mice were i.d. infected with S. aureus or PBS into both ear pinnae. Three weeks later dermal Mj (CD45 hi CD11b hi CD64 hi CD3 lo CD11c lo ) were sorted by FACS (MoFlow Astrios), and directly resuspended in RLT lysis buffer (Qiagen). Total RNA was extracted using the RNAeasy micro kit, according to instruction manual (Qiagen). Further sample processing was performed at an Affymetrix Service Provider and Core Facility, 'KFB -Center of Excellence for Fluorescent Bioanalytics' (Regensburg, Germany; www.kfb-regensburg.de). In brief, 1 ng of total RNA was used for reverse transcription to synthesize single-stranded (ss) cDNA with a T7 promoter sequence at the 5' end. Next a 3' adaptor was added and the ss cDNA was converted to double-stranded cDNA, which acted as a template for a pre-in vitro transcription (IVT) amplification by a nine cycle PCR. Afterwards antisense RNA (complimentary RNA or cRNA) was synthesized and linearly amplified by an in vitro transcription (IVT) of the double-stranded cDNA template. 20 mg of cRNA was purified and reverse transcribed into double-stranded (ds) cDNA, whereat unnatural dUTP residues were incorporated at a fixed ratio relative to dTTP. Purified ds cDNA was fragmented using a combination of uracil DNA glycosylase (UDG) and apurinic/apyrimidinic endonuclease 1 (APE 1) at the dUTP residues followed by a terminal labeling with biotin. 5,5 mg fragmented and labeled ds cDNA were hybridized to Affymetrix Clariom S mouse arrays for 16 hr at 45˚C in a GeneChip hybridization oven 640. Hybridized arrays were washed and stained in an Affymetrix Fluidics Station FS450, and the fluorescent signals were measured with an Affymetrix GeneChip Scanner 3000 7G. Fluidics and scan functions were controlled by Affymetrix GeneChip Command Console v4.1.3 software.

Microarray data analysis
Summarized probe set signals in log2 scale were calculated by using the GCCN-SST-RMA algorithm with the Affymetrix GeneChip Expression Console v1.4 Software. After exporting into Microsoft Excel, average signal values, comparison fold changes and significance P values were calculated. Probe sets with a fold change above 2.0 fold and a student's t test P value lower than 0.05 were considered as significantly regulated. GEO accession number: GSE145094. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Data availability
All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided.