Neutrophil swarms containing myeloid-derived suppressor cells are crucial for limiting oral mucosal infection by C. albicans

Oral mucosal colonization by C. albicans (Ca) is benign in healthy people but progresses to deeper infection known as oropharyngeal candidiasis (OPC) that may become disseminated when combined with immunosuppression. Cortisone-induced immunosuppression is a well-known risk factor for OPC, however the mechanism by which it permits infection is poorly understood. Neutrophils are the primary early sentinels preventing invasive fungal growth, and here we identify that in vivo neutrophil functional complexes known as swarms are crucial for preventing Ca invasion which are disrupted by cortisone. Neutrophil swarm function required leukotriene B4 receptor 1 (BLT1) expression, and swarms were further characterized by peripheral association of polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs) showing that OPC recruits PMN-MDSCs to this site of infection. Furthermore, PMN-MDSCs associated with Ca hyphae had no direct antifungal effect but showed prolonged survival times and increased autophagy. Thus in vivo neutrophil swarms are complex structures with spatially associated PMN-MDSCs that likely contribute immunoregulatory functions to resolve OPC. These swarm structures have an important function in preventing deep invasion by Ca within the oral mucosa and represent a mechanism for increased disease severity under immune deficient clinical settings.


INTRODUCTION
The oral epithelium provides both mechanical and immunological barriers against the opportunistic fungus Candida albicans that causes oropharyngeal candidiasis (OPC), also known as thrush.
Individuals with immune de ciencies or impaired immunity due to corticosteroid therapy 5 are especially susceptible to both OPC and deeper fungal infection.However, surprisingly little is known about the mechanisms by which corticosteroids such as cortisone initiate OPC, despite their widespread clinical use such as with oral inhalers.
Neutrophil recruitment is an important rst line to defense in host response to OPC [6][7][8] .Cortisone-treated mice infected with C. albicans form white tongue plaques consisting of an adherent bio lm that disrupts the tongue liform papillae and hyphae that penetrate deep into the corneum stratum (CS) 9,10 .Although fungal invasion recruits massive numbers of neutrophils that in ltrate throughout the epithelium and connective tissue (CT), they are unable to resolve C. albicans infection in cortisone-induced immune suppressed animals 9 .Similarly, mice de cient in IL-17 receptor signaling show persistent fungal infection despite extensive neutrophilic in ltration 6,10,11 .Thus, the function of neutrophils in resolving OPC has been questioned in the context of immunode ciencies.
Epithelial infection by C. albicans causes a temporal release of an array of cytokines and chemokines that recruit and activate neutrophils.IL-1 signaling induces expression of granulopoietic cytokines such as G-CSF in the connective tissue, leading to production and release of neutrophils from the bone marrow into the circulation 7 .Epithelial cells also produce neutrophil chemokines, including KC/CXCL1 and MIP-2/CXCL2, that induce in ltration of polymorphonuclear neutrophils into the oral mucosa [6][7][8] .
Glucocorticoids are known to inhibit neutrophil recruitment by affecting several steps in the extravasation cascade 16 .Furthermore, chronic exposure to glucocorticoids can shift the balance of the innate immune response resulting in dysregulated expression of chemokines and pro-in ammatory cytokines 17 .Thus, cortisone-induced OPC might be a consequence of delayed neutrophil recruitment, although the large numbers of neutrophils located within C. albicans infected epithelium suggests other functional defects.
Recently a novel function of neutrophils has been identi ed in which neutrophils switch from simple chemotaxis towards the site of bacterial, fungal or parasitic infection to a swarm-like migration pattern known as neutrophil swarming 12 .Neutrophil swarming is characterized by highly coordinated chemotaxis and accelerated neutrophil accumulation to form clusters around the invading microorganism, thus insulating the infected site from surrounding healthy tissue 13 .Neutrophil swarms can be transient (small diameter of < 150 neutrophils,) or permanent (large diameter of > 300 neutrophils) in nature 12,14 .Initiation of swarm formation is controlled by leukotriene B4 and interaction with its receptor BLT1 (also known as LTB4R1) 15 .Other mediators contributing to swarm formation in mice include CXCR2 ligands, such as CXCL2 (MIP-2).Hopke et al. (2020) showed that neutrophil swarming is a response to C. albicans infection in vitro, involving LTB4/BLT1 and G-CSF signaling for swarm formation, and MPO for C. albicans killing.However, there is limited in vivo evidence for the function of neutrophil swarms beyond those reported by intravital microscopy in bacteria infected murine skin and liver 12 .
Although neutrophils were considered a homogeneous population when compared to macrophages and lymphocytes, several studies have now demonstrated intrinsic functional heterogeneity in the circulating neutrophil pool (reviewed in 18 ).A speci c subset of immunomodulatory cells expressing high levels of Arginase 1 (Arg1) 19 are known as myeloid-derived suppressor cells (MDSCs) and have been well described when in ltrating tumors and in hyperproliferative skin disorders 20 .In mice, two major MDSC subsets are identi ed by their differential expression of the Ly6G and Ly6C surface antigens and by their immunosuppressive activities, in part mediated by the expression of the enzyme Arg1.Monocytic (M) MDSCs are de ned as being CD11b + Ly6G neg Ly6C high , while granulocytic or polymorphonuclear (PMN) MDSCs are CD11b + Ly6G + Ly6C low 21 .Although they express similar surface markers as neutrophils, PMN-MDSCs can be differentiated from neutrophils by their ability to suppress T cell proliferation 22 .One recent study showed that systemic C. albicans infection induces mainly the PMN-MDSC subset where they have a protective role by limiting host hyperin ammatory responses 23 .Furthermore, adoptive transfer of MDSCs conferred protection against systemic fungal dissemination and increased murine survival 23 .However, it is not known whether MDSCs have any biological role in localized fungal mucosal infections such as OPC.
In this work we show that cortisone immunosuppression not only delayed cytokine / chemokine expression and recruitment of neutrophils in response to C. albicans oral infection but disrupted in vivo BLT1 signaling and neutrophil swarming.Furthermore, we found that neutrophils swarms contained spatially localized PMN-MDSCs that associated with C. albicans hyphae and are components of in vivo swarms.Our ndings provide a mechanistic explanation for how cortisone is a signi cant risk factor for OPC and supply in vivo evidence that neutrophil swarming is a crucial component for resolving C. albicans oral infection.

RESULTS
Cortisone both delayed in ammatory cell recruitment and altered their ability to control infection.To determine whether cortisone affects only neutrophil recruitment, we compared C. albicans oral infection between immunocompetent (IC) and cortisone immunosuppressed (IS) mice histologically on days postinfection (dpi) 1,3 and 5. IC and IS animals had similar total body weight loss (Fig. 1A) and infection levels as measured by Ca CFU (Fig. 1B).But at dpi1, we observed striking differences in in ammatory cell recruitment (Fig. 1C).IC animals had extensive in ammatory cell in ltration primarily in the cellular epithelium (cEp), but also extending into the corneum stratum (CS) and connective tissue (CT).These in ammatory cells (consisting of multi-lobulated nuclear morphology characteristic of neutrophils) were well organized and tightly clustered around and beneath invading fungal cells as in swarms (black boxes, Fig. 1C).In contrast, IS animals had no evidence of in ammatory cell recruitment despite having fungal invasion deep into the cEp that nearly reached the basement membrane (arrow).
By dpi 3, IC tongues showed a one log-fold reduction in fungal burden and fungal cells were mainly isolated to the uppermost CS layers (Fig. 1D, arrow).Histological H&E assessment of in ammatory cell in ltration showed that it was largely absent in the cEp and con ned to the CT (Fig. 1D).Interestingly, resolution of infection in IC animals was accompanied by increased hyperplastic thickening of the cEp (Fig. 1D black bars and Fig. 1F) as well as thickening of the CS (Fig. 1G).IS animals at dpi3 had signi cantly more weight loss than IC animals (Fig. 1A) along with a one log-fold increase in Ca CFU compared to dpi1 and profuse recruitment of in ammatory cells to the cEp (Fig. 1D, circle).However, there was a remarkable lack of organization in in ammatory cells at the infection site, and these cells were widely dispersed instead of forming clusters as found in IC animals.The thickened cEp response was not found in IS animals, however the CS layer was signi cantly increased (Fig. 1G) as a result of the mass of fungal hyphae localized in this area (Fig. 1D).At dpi5, IC animals weight returned to pre-infection levels and tongue tissues had histologically normal architecture and cEp thickness (Fig. 1E).However, the epithelium retained 10 3 CFU/g tissue (Fig. 1B) with little visible in ammatory cell in ltration (Fig. 1E).In contrast, IS animals had signi cant morbidity along with continued loss of body weight that required sacri ce (Fig. 1A).Ca infection levels in tongues climbed to 10 7 CFU/g (Fig. 1B) that was mainly found within the CS (Fig. 1E).In ammatory cells were proli c and widely distributed without the organization of swarms observed in IC tissues.These large numbers of in ammatory cells signi cantly increased the thickness of the CS that was three-fold more than that found in IC animals at any time point (Fig. 1G).The cEp in IS animals was only slightly thickened, but we observed several areas in which CS desquamation (containing large masses of fungi) from the underlying cEp occurred (Fig. 1E, arrow), resulting in exfoliation and further seeding of Ca into the oral environment.Thus, cortisone caused signi cant delays in in ammatory cell recruitment as well as loss of organized structure of in ammatory cells recruited by dpi5, resulting in ineffective reduction of Ca infection levels.
OralC.albicansinfection leads to recruitment of neutrophils but not macrophages.To quantitate in ammatory cell recruitment to the site of oral fungal infection, epithelial and CT compartments were measured histologically for Ly6G + neutrophils and CD163 + macrophages.In ammatory cells localized in the Ep following fungal infection in both IC and IS tongues were predominantly Ly6G + con rming their identity as neutrophils (Fig. 2A).We also observed these cells to be highly organized in Ep tissues on dpi1 in IC animals in contrast to the diffuse in ltrate in IS Ep tissues (Fig. 2A).Furthermore, a small subset of granulocytes adjacent to and within the neutrophil swarms were Ly6G + Arg1+, also consistent with the accumulation of granulocytic or polymorphonuclear (PMN) MDSCs (Fig. 2A).Surprisingly, macrophages were localized exclusively to CT and did not enter the Ep in either IC or IS animals (Fig. 2C).
Quanti cation of neutrophils showed that these cells were recruited primarily to the Ep at dpi1 in IC animals, as well as a small yet signi cant increase in neutrophil in ltration in the CT at dpi1 (Fig. 2B).As we observed histologically, IS animals had delayed recruitment of neutrophils, but the number of cells recruited at dpi3 was like that of IC animals at dpi1 and persisted in Ep through dpi5.In contrast, macrophages were absent from the Ep and localized in low numbers within the CT compartment in both IC and IS tongues at all time points (Fig. 2C).However, the numbers of macrophages cells in IC and IS animals were not signi cantly different than naive mice showing that oral fungal infection did not elicit further recruitment of macrophages to oral CT (Fig. 2D).
To determine systemic in ammatory cell production in response to Ca oral infection, we examined numbers of granulocytic cells (CD11b + Ly6G + Ly6C low ) which may include neutrophils and PMN-MDSCs, and monocytic cells (CD11b + Ly6G − Ly6C high ) which may include M-MDSCs and macrophage precursors.
Flow cytometry analyses of bone marrow and peripheral blood showed that systemic production of in ammatory cells was skewed towards granulocytic cells in response to oral Ca infection (Fig. 2E).
Granulocytes were signi cantly elevated in both bone marrow (dpi1 and dpi3) and blood (dpi3) in IS animals compared to IC animals (Fig. 2F).By dpi5, IS animals had a four-fold higher expansion (p < 0.001) of peripheral granulocytic cells compared with IC animals that had returned to basal levels (Fig. 2F).Thus, cortisone increased granulocyte expansion in response to oral Ca infection in both the bone marrow and peripheral blood.Surprisingly, monocyte levels were unchanged in the bone marrow and peripheral blood in IC and IS animals, showing that oral Ca infection had little impact on monocyte/macrophage production and recruitment.Thus, based upon frequencies, monocytic cell (macrophage or dendritic cell progeny) expansion and recruitment does not appear to have a role in the response to oral fungal infection, with the predominant response being granulocytic cell accumulation including neutrophils and possibly PMN-MDSCs.
Expansion of PMN-MDSCs in response to oral candidiasis was con rmed by the immunosuppressive activity on CD4 + T cells by Ly6G + Arg1 + cells.
We found that recruitment and localization of Ly6G + Arg1 + cells to infected tongue Ep tissues was affected by cortisone.In IC animals, Ly6G + Arg1 + cells were localized to the periphery of neutrophil swarms on dpi1 (Fig. 3A red arrows), then were reduced in the Ep at dpi3 and by dpi5 all were con ned to the CT compartment.In contrast IS animals had no epithelial recruitment of Ly6G + Arg1 + until dpi3 and those were randomly scattered throughout disorganized neutrophilic areas.Total numbers of Ly6G + Arg1 + cells recruited to Ep were signi cantly higher in IC animals at dpi1 than at any time point in IS animals (Fig. 3B), although the proportion of Ly6G + Arg1 + to total granulocytes was equivalent between IC dpi1 and IS dpi5 (Fig. 3C).Thus, IC animals recruited higher total numbers of Ly6G + Arg1 + cells to the Ep in response to Ca infection while IS animals had reduced and delayed Ly6G + Arg1 + recruitment.
Next, we examined Ly6G + cells from the bone marrow and spleen as proxy sites for the myeloid response in IC and IS mice for their ability to inhibit T-cell proliferation to con rm their identity as PMN-MDSCs (Fig. 3D).Bone marrow-derived CD11b + Ly6G + Ly6C low cells collected from both IC and IS mice dpi3 caused 60% suppression in CD4 + T cell proliferation, con rming the functional expansion of PMN-MDSCs in the granulocyte pool (Fig. 3D).On dpi5, the immunosuppressive effects of CD11b + Ly6G + Ly6C low cells, as measured from the spleen, only persisted in IS mice, suggesting a protracted expansion of PMN-MDSCs in IS mice as a result of continued Ca infection.Thus, PMN-MDSCs are recruited to the tongue as a part of the systemic granulocytic response to Ca infection.
Neutrophil swarming and BLT1 expression are reduced by cortisone.If swarming is affected by cortisone, we expected to see disrupted swarm morphology and BLT1 expression in IS animals.Indeed, IC animals produced neutrophil swarms that were readily visualized within the infected Ep (Fig. 4A, left panel, circle) in which neutrophils tightly surrounded the invading Ca hyphae and yeast (enlarged view), while IS animals had a diffuse neutrophilic in ltration that sometimes produced very small swarm-like nodules (transient swarms) (Fig. 4A, right panel, small circles).We next measured mean swarm areas from multiple tongue sections from IC and IS animals.Quanti cation of these areas showed that IC swarms were 8-fold larger at dpi1 than IS animals produced at any time point during Ca infection (Fig. 4B).A key factor required for swarming is localized expression of the chemoattractant leukotriene B4 that is recognized by the high a nity leukocyte receptor BLT1 15 .Therefore, we examined expression of BLT1 in neutrophils recruited to the infection site.Neutrophils were visualized by Ly6G staining in Ca infected tissues in IC and IS animals, then probed for BLT1 expression.We found that neutrophils forming organized swarms in IC animals also highly expressed BLT1 (Fig. 4C upper panel), while none of the neutrophils in IS animals were found to express BLT1 at levels detectable by immunohistochemistry (Fig. 4C lower panel).This absence of BLT1 expression was consistent in IS animals from dpi3 through dpi5.Thus, one mechanism by which cortisone disrupts neutrophil swarm formation is by reduction of leukotriene B4 / BLT1 signaling.
Cortisone induced immunosuppression delays expression of proin ammatory cytokines and chemokines in response to fungal infection.We next asked whether cortisone altered oral tissue speci c expression of key in ammatory cell recruitment cytokines that may skew neutrophil recruitment or swarming.We examined Ep and CT expression levels of the cytokines IL-1β and G-CSF, known to expand and activate neutrophils during OPC, and the chemokines KC/CXCL1 and MIP-2/CXCL2, which recruit neutrophils in response to oral candidiasis in tongue tissue [6][7][8] (Fig. 5B-E).We also examined tissue levels of IL-17A that may have a role in neutrophil recruitment (Fig. 5F).We examined Ep and CT myeloperoxidase (MPO) as a marker of neutrophil recruitment and found maximal levels at dpi1 in the Ep of IC animals that were delayed until dpi3 in IS animals (Fig. 5A).Among the neutrophil recruitment/activation chemokines and cytokines examined, all four (IL-1β, G-CSF, KC/CXCL1 and MIP-2/CXCL2) demonstrated temporal expression levels that very closely paralleled neutrophil recruitment in both IC and IS animals (Fig. 5B-E), suggesting that defective expression among these cytokines is unlikely to be the reason for loss of swarming behavior.However, cortisone induced immunosuppression delayed the expression of these cytokines and chemokines at dpi1 accounting for delayed neutrophil recruitment.We found that MIP-2/CXCL2 and IL-1β expression to be mainly localized to the Ep, while G-CSF and KC/CXCL1 were expressed in both Ep and CT.MIP-2/CXCL2 expression was the most upregulated chemokine tested, being 3-to 10-fold higher than others measured at dpi1 in IC and at dpi 3-5 in IS animals.In contrast, while endogenous IL-17A production was restricted to the CT (not detectable in Ep) in IC tissues, its expression was completely suppressed by cortisone at dpi1 and dpi3 (Fig. 5F), suggesting that CTlocalized innate immune cells expressing IL-17 play some biological role in the process of neutrophil recruitment.
Treatment with anti-Ly6G antibody caused disruption of neutrophil swarming and increased tissue invasion of Ca.To determine whether cortisone had additional effects beyond delaying recruitment of neutrophils, we treated animals with anti (α)-Ly6G antibody to speci cally reduce early neutrophil recruitment then measured effects on swarming and animal morbidity (CFU, weight loss and tissue invasion).We expected that if cortisone solely functioned to reduce swarm formation, then α-Ly6G treatment in IC animals (IC α-Ly6G) would have equivalent infection outcomes as IS or ISα-Ly6G control animals.α-Ly6G treatment reduced total tissue granulocytic recruitment in IC animals at dpi1 (Fig. 6A).However, by dpi3 total granulocyte numbers in tissues did not differ between groups with or without α-Ly6G treatment (and IS compared to IS α-Ly6G).Ca infection levels (CFUs) were signi cantly increased in IC α-Ly6G compared to IC animals at both dpi1 and dpi3 (Fig. 6B), however IC α-Ly6G, IS, and IS α-Ly6G groups all had equally elevated CFUs despite different total numbers of recruited granulocytes.Strikingly, animal morbidity as measured by weight loss was most severe in α-Ly6G treated animals (IC α-Ly6G and IS α-Ly6G animals lost signi cantly more weight than IS animals) requiring early sacri ce by dpi3 (Fig. 6C).However, total numbers of infecting Ca tongue CFUs and total granulocyte numbers did not account for this increased weight loss.
We next measured neutrophil swarming behavior among these groups expecting that α-Ly6G treatment would reduce numbers of neutrophils and have impaired swarming (Fig. 6D).As expected, IC α-Ly6G mice at dpi1 with reduced granulocyte numbers showed nearly complete loss of swarms.At dpi3, α-Ly6G treated animals (IC α-Ly6G and IS α-Ly6G) had signi cantly smaller swarm areas than IS mice, that was consistent with weight loss in each group.IC α-Ly6G tissues still contained some small mini-swarms with detectable BLT1 staining (Fig. 6E, circles); however, IS α-Ly6G animals showed in a complete loss of swarms (despite having abundant numbers of neutrophils) and undetectable BLT1 expression.Thus, cortisone caused more profound loss of BLT1 expression than α-Ly6G treatment although both have the same target.
To assess the impact of impaired swarm formation, we examined severity of infection by assessing the depth of Ca invasion relative to the basement membrane (Ca-BM separation, BM shown in dotted lines Fig. 6F), since invasion into the CT is permissive for dissemination.There was a signi cant increase in Ca invasion depth in IC α-Ly6G animals at dpi 1, with many animals showing CT invasion (Fig. 6H, black bars show CT invasion depth).At dpi3, both IC α-Ly6G and IS animals had signi cantly increased invasion compared to IC (Fig. 6F), although increased cEp thickening in the IC group may have mitigated apparent invasion depth.Strikingly, Ca invaded up to 200 µm into CT in IS α-Ly6G animals (Fig. 6H, lower panel), despite having equal numbers of neutrophils to IS animals in which the invasion depth was mainly restricted to the Ep.Thus, animals that exhibited the greatest CT invasion depth were those with the greatest loss of swarming behavior (IC α-Ly6G dpi1 and IS α-Ly6G dpi3), that could not be accounted for by absolute differences in neutrophil numbers or CFUs.
PMN-MDSCs are differentially localized within swarms in IC and IS mice, although both express COX-2.Since we observed PMN-MDSCs to be components of swarms, we questioned whether PMN-MDSCs recruitment was also affected by α-Ly6G treatment that might account for differences in Ca tissue invasion.Numbers of PMN-MDSCs were not signi cantly different in IC α-Ly6G treated tongue Ep compared to IC animals at dpi1 (Fig. 7A).However, by dpi3 IS α-Ly6G animals with the most severe disease had signi cantly higher numbers of PMN-MDSCs compared to all other groups (Fig. 7A).Since MDSCs are known to have an immune-suppressive role in the context of tumor biology, we compared localization and proximity to Ca of PMN-MDSCs within IC α-Ly6G and IS α-Ly6G tissues that have equal CFUs but differ in Ca invasion depth (dpi3, Fig. 6E) by image deconvolution (Fig. 7B).PMN-MDSCs in IC α-Ly6G tissues showed peripheral positioning around Ca with little direct contact (Fig. 7B, upper panel).In contrast, IS α-Ly6G tissues showed dense foci of PMN-MDSCs (as shown by Arg1 + staining) that they were closely associated with Ca hyphae (Fig. 7B, lower panel).PMN-MDSCs from either group of α-Ly6G treated animals also showed expression of COX-2, suggesting that they might be undergoing autophagy 24 .
Human PMN-MDSCs have little antifungal activity in vitro, but association with Ca hyphae increased their autophagy and survival.Since PMN-MDSCs appear to be localized with Ca hyphae in vivo, we questioned whether they have any direct antifungal activity as do neutrophils.We compared Ca killing and phagocytic activity in vitro using freshly isolated human peripheral neutrophils (since they have higher activity than murine neutrophils) and human neutrophil-derived PMN-MDSCs.In contrast to previous reports 23 , PMN-MDSCs had signi cantly lower (10-fold) phagocytic index (Fig. 7C) and had no fungicidal activity compared with that of neutrophils (Fig. 7D).Surprisingly, hyphal form Ca cells strongly attracted PMN-MDSCs (Fig. 7E) as we found in infected tissues in vivo.When co-incubated together, Ca yeast cells showed very little association with neutrophil-derived PMN-MDSCs in vitro, while Ca hyphae were found to be tightly associated with the surfaces of PMN-MDSCs (Fig. 7E).Since we observed COX-2 expression in vivo, this suggested that PMN-MDSCs-Ca hyphal association might induce autophagy in order to promote PMN-MDSC survival.Indeed, Ca hyphae increased PMN-MDSC autophagy by two-fold over yeast cells, and these levels were 70% of the positive control with rampamycin (Fig. 7F).Since autophagy inhibits MDSC apoptosis and promotes survival in the tumor microenvironment 25 , we examined MDSC survival following exposure to Ca. PMN-MDSC survival was increased by 50% after 60-90 min exposure to Ca hyphae compared to unexposed cells (Fig. 7G), and was statistically equal to rampamycin treatment.In total these data show that PMN-MDSCs, although they have no direct antifungal activity, can strongly associate with Ca hyphae that promote their survival time.

DISCUSSION
While neutrophils are essential rst responders to clear fungal infection, our understanding of their role has been that pro-in ammatory signaling recruits an optimal quantity of neutrophils to the infection site.This work shows that a much more complex three-dimensional recruitment of neutrophils, but not macrophages, are needed for functional clearance of Ca in tissues as shown by the ability of cortisone to disrupt this morphology.We are the rst to identify in vivo neutrophil swarms (previously described in vitro 15,26 ) that are spatially associated with PMN-MDSCs within the oral mucosa.The mere presence of high numbers of neutrophils is not su cient for C. albicans clearance, as observed following cortisone treatment in which elevated numbers of in ltrating neutrophils in the mucosa are unable to control infection.Although one previous work showed that MDSCs are found in peripheral blood following systemic infection by Ca 23 , we are the rst to discover that localized mucosal infection by Ca recruits PMN-MDSCs that are part of the neutrophilic swarm complex.Furthermore, we found no role for macrophages in the host response to oral candidiasis, unlike Ca infection in other tissues such as lung or kidney.
Neutrophil swarming is an emergent behavior triggered when neutrophils encounter targets that are signi cantly larger than individual neutrophils such as damaged tissues 26 , clusters of bacteria 27 , fungal hyphae 28 , or parasites 29 .During swarming, neutrophils communicate with each other, engaging in positive feedback signaling loops 30 driven exclusively by LTB4 in mice 31 , and by LTB4 and other cytokines in human neutrophils 26,30 .Several features of the swarming process are truly unique and could not have been expected from previous knowledge of neutrophil fundamental activities e.g., chemotaxis, phagocytosis, enzyme and reactive oxygen release.These features, which were revealed in the past few years, include the transcellular synthesis of LTB4 during swarming 31 , which acts as a relay between neutrophils 32 , complement activation by neutrophils 33 and coordinated cytokine release 30,34 .Cytokines such as G-CSF and IL-1 are critical for C. albicans clearance as demonstrated by IL-1R KO studies 7 .IL-17 has also been shown to contribute to C. albicans clearance during OPC, however, its role appears less relevant for neutrophil recruitment 6 .We showed that C. albicans infection elicited a cytokine cascade involving IL-1 and G-CSF which resulted in the local production of neutrophil-speci c chemokines such as CXCL-1 and CXCL-2.Disruption of this cytokine cascade by cortisone at an early stage of infection resulted in delayed and defective swarm formation, reduced PMN-MDSC recruitment and C. albicans overgrowth and bio lm formation.
Correlations have been also demonstrated between neutrophil swarming and both sustained 35 and transient 36 calcium uxes in individual neutrophils, and the repetitive calcium waves centered on the target during the initiation of swarming 37 .Moreover, factors limiting the size of the swarms have also been identi ed, including lipid mediators like lipoxin A4 30 and resolvin D1 38 , intracellular signaling pathways that desensitize key receptors 27 , and the activation of NADPH oxidase terminating the calcium waves 37 .The swarming behavior of neutrophils is often perturbed in patients, in a broad range of conditions that associate with higher frequency of infections including cirrhosis, solid organ transplant, stem cell transplant, chronic granulomatous disease (CGD) 13,39,40 .We observed that cortisone treatment downregulated the expression of BLT1 leading to defective swarm formation.However, it is possible that cortisone also impacts expression of lipoxins or resolvins resulting in sustained in ammation in the site of infection.Thus, restoration of components of the LTB4/BLT1 pathway could be a therapeutic intervention, as shown in a patient where restoring the swarming activity of neutrophils correlated with a reduced rate of infections 14 .
We show for the rst time that PMN-MDSCs are induced systemically in the bone marrow and recruited into the oral epithelium in response to localized C. albicans oral infection where they locate to the periphery of the neutrophil swarms.PMN-MDSCs were shown to have a protective role during murine systemic candidiasis by limiting host hyperin ammatory responses 23 .However, this study found that adoptive transfer of neutrophilic MDSCs did not reduce kidney CFUs and had no protective effect in immunosuppressed mice 23 .Our results suggest that Ca kidney infection (resulting from systemic candidiasis) may also require neutrophil swarming and localized PMN-MDSCs that may explain these ndings.
The complete role of PMN-MDSCs in oral C. albicans clearance remains to be de ned.Although we found anti-Ly6G treatment signi cantly depleted neutrophils (as previously shown in response to C. albicans infection 41 ), PMN-MDSCs were not affected and were found surrounding C. albicans.Despite such PMN-MDSC in ltration, C. albicans invaded beyond the basement membrane, and this invasion was signi cantly deeper (up to 200 µm) when neutrophil depletion was combined with cortisone.Thus, PMN-MDSCs might contribute immunosuppressive functions that permit deeper Ca invasion without exerting any direct fungicidal activity.However, C. albicans appears to communicate with PMN-MDSCs through co-localization and by inducing expression of COX-2, a marker of autophagy 24 .Thus, there is likely bidirectional signaling between PMN-MDSCs and neutrophil swarms as well as between PMN-MDSCs and C. albicans cells.
Thus, we nd that in vivo neutrophil swarms containing PMN-MDSCs are essential for clearing oral C. albicans infection, and disruption of this function by cortisone is a mechanism for increased disease severity under immune de cient clinical settings.We show that effective host defense against C. albicans requires complex interactions among neutrophil swarms and PMN-MDSCs as key mechanisms of innate immunity.

MATERIALS and METHODS
Murine model of oral candidiasis.Immunocompetent (IC) or Immunosuppressed (IS) C57BL/6J, 6-8 week-old female mice were infected sublingually as previously described 9 .Animal protocols were approved by the University at Buffalo Institutional Animal Care and Use Committee (ORB06042Y).Each experimental group consisted of 5-8 mice, and experiments were repeated at least twice.Mice were immunosuppressed using 225 mg/kg of body weight cortisone 21-acetate (Sigma Aldrich C3 130-5G).
Mice were anesthetized with a dose of Ketamine (100 mg/kg) and Xylazine (10 mg/kg) and infected sublingually with cotton balls carrying 1x10 7 C. albicans cells (CAI4-URA + with URA3 replaced at the RPS1 locus using a CLP10 plasmid or SC5314) for 1 h.Mice were monitored daily for weight loss.On days 1, 3 and 5 post-infection (dpi) mice were weighed, and sacri ced by cervical dislocation under anesthesia and tongues were harvested immediately.To quantify C. albicans, one-half of the tongue tissue was weighed, homogenized in phosphate buffer and plated on yeast extract-peptone dextrose (YPD) agar plates for 48 h to obtain the number of CFU/g of tissue.The other half of the tongue was xed in 10% formalin and embedded in para n, to obtain 4µm sections that were processed for periodic acid Schiff (PAS) or Hematoxylin & Eosin stain for histological analysis and for immunohistochemistry. Whole tongues were obtained for epithelial and connective tissue dissection and protein extraction assays.

Protein extraction, MPO ELISA and Multiplex ELISA assays
Freshly extracted whole tongues were treated with dispase at 37°C for 1 h to dissect the epithelium from the underlying connective tissue as previously described 42 .Total protein was isolated from tongue epithelial or connective tissue using NP-40 buffer containing protease inhibitor cocktail (Thermo Scienti c, Rockford, IL).Connective and epithelial tissues were homogenized in NP-40 buffer and sonicated.After centrifugation, total protein concentrations were determined using the BCA (bicinchoninic acid) protein assay (Thermo).Dissected tongue tissues were analyzed for the presence of MPO (Mouse myeloperoxidase DuoSet ELISA, R&D systems) and IL-1β, IL-1α, TNFα, G-CSF, CXCL1/ KC and CXCL2 / MIP-2, IL-17 (Bio-Plex™ Pro mouse custom assays, BioRad, Hercules CA, USA).The analysis was performed in duplicates on a Luminex 200 (Millipore), located at the Department of Flow and Image Cytometry, Roswell Park Cancer Institute.

Collection of cells from murine blood, spleen and bone marrow
Blood was collected by heart puncture and collected in K2-EDTA tubes (BD) on the day of the assay under anesthesia, as described above.At least 200 µl of blood was obtained from each mouse.Femurs and spleens were harvested from mice after euthanasia by cervical dislocation under anesthesia.Bone marrow was ushed from femurs using fresh α-MEM media, passed through a 70 µm strainer using icecold HBSS without Ca+, pelleted by centrifugation, and resuspended in assay buffer.Spleens were homogenized using complete RPMI media and passed through a 70 µm lter.Erythrocytes were lysed with cell lysis buffer (BioLegend®, San Diego, CA).

Flow cytometry
All antibodies were from BioLegend (San Diego, CA), unless stated otherwise.Blood (100 µl) and bone marrow cells (1 x 10 6 in 100 µl of FCM buffer) (Leinco Technologies, St. Louis, MO) were rst incubated with Fc block (TruStain FcX) and then stained with Live / Dead Aqua uorescent reactive dye (Invitrogen), FITC conjugated anti-CD45 (30-F11), APC conjugated anti-CD11b (M1/70), PE conjugated anti-Ly6G (1A8), and PerCP/ Cy5.5 conjugated anti-Ly6C for 20 min at 30°C.Forward scatter versus side scatter was set to include singlets.By gating on CD11b + cells, the numbers of neutrophilic (Ly6G + Ly6C low ) and monocytic (Ly6G − Ly6C hi ) cells were determined.Cells were run on a LSR Fortessa ow cytometer and data were analyzed by FlowJo (BD).Cytometry services were provided by the Flow and Image Cytometry Shared Resource at the Roswell Park Comprehensive Cancer Center, supported in part by the NCI Cancer Center Support Grant 5P30 CA016056.

Figures
Figures

Figure 7 MDSCs
Figure 7 1 microscope connected to a Zeiss Axiocam 105 color, and using ZEN 2011 Software.Number of cells / mm 2 were obtained for Ly6G + MPO + granulocytes / neutrophils, CD163 + macrophages, and arginase + MDSCs in each tissue compartment (epithelium and connective tissue).Image deconvolution of PAS and Arg1 stained slides was performed at the Optical Imaging and Analysis facility of the University at Buffalo School of Dental Medicine.Research pathology services for this study were provided by the Pathology Network Shared Resource, which is funded by the National Cancer Institute (NCI P30CA16056) and is a Roswell Park Comprehensive Cancer Center Cancer Center Support Grant shared resource.