Antibacterial activity of human mesenchymal stem cells mediated directly by constitutively secreted factors and indirectly by activation of innate immune effector cells

Abstract Mesenchymal stem cells (MSC) have been shown to improve wound healing and suppress inflammatory immune responses. Newer research also indicates that MSC exhibit antimicrobial activity, although the mechanisms underlying this activity have not been fully elucidated. Therefore, we conducted in vitro and in vivo studies to examine the ability of resting and activated MSC to kill bacteria, including multidrug resistant strains. We investigated direct bacterial killing mechanisms and the interaction of MSC with host innate immune responses to infection. In addition, the activity of MSC against chronic bacterial infections was investigated in a mouse biofilm infection model. We found that MSC exhibited high levels of spontaneous direct bactericidal activity in vitro. Moreover, soluble factors secreted by MSC inhibited Staphylococcus aureus biofilm formation in vitro and disrupted the growth of established biofilms. Secreted factors from MSC also elicited synergistic killing of drug‐resistant bacteria when combined with several major classes of antibiotics. Other studies demonstrated interactions of activated MSC with host innate immune responses, including triggering of neutrophil extracellular trap formation and increased phagocytosis of bacteria. Finally, activated MSC administered systemically to mice with established S. aureus biofilm infections significantly reduced bacterial numbers at the wound site and improved wound healing when combined with antibiotic therapy. These results indicate that MSC generate multiple direct and indirect, immunologically mediated antimicrobial activities that combine to help eliminate chronic bacterial infections when the cells are administered therapeutically.

conventional antibiotics. 1 Methicillin-resistant Staphylococcus aureus (MRSA) in particular accounts for high mortality worldwide 2 and is responsible for many chronic implant infections. 3 The use of mesenchymal stem cell (MSC) to treat bacterial infections has received increasing attention in recent years, based on in vitro studies documenting direct bactericidal activity. [4][5][6] Thus, it is possible that administration of MSC may be a means of potentiating conventional antibiotic therapy, 6,7 it is therefore important to elucidate more fully the mechanisms underlying MSC antimicrobial activity, both in vitro and in vivo.
Dealing with biofilm formation is currently one of the principle challenges facing clinicians when dealing with chronic infections. Bacteria in biofilms live in an environment that favors bacterial persistence and evasion of host immune responses. 8 Within biofilms, bacteria elude killing from the immune system and antibiotics. 9 For example, S. aureus biofilms can influence macrophage polarization and inhibit bacterial phagocytosis. 10 The compact three-dimensional structures of biofilms also limit neutrophil recruitment and killing because of a decrease in surface receptor recognition. 11 Previous studies have shown that MSC have the ability to directly influence the immunological properties of macrophages and neutrophils by secreting factors such as PGE 2 , 12 IL-6, IL-8, or IFN-β. 13 Following exposure to MSC-secreted factors, macrophages develop increased phagocytosis, mediated in part by NADPH oxidase activation. 14 Neutrophils exposed to MSC conditioned medium are resistant to apoptosis and demonstrate increased migration. 15 Studies in animal models of infection have shown that human MSC can increase monocyte recruitment and decrease excessive neutrophil influx and neutrophil elastase production, particularly in murine models of cystic fibrosis and pulmonary Pseudomonas aeruginosa infection. 16 MSC also produce antimicrobial peptides (AMPs), which are short peptides commonly found in neutrophils or epithelial cells. 17 AMPs kill bacteria directly by disrupting the integrity of the microbial membrane, 17 or by inducing the release of proinflammatory cytokines and in turn the recruitment of immune cells. Human MSC have been shown to produce multiple AMPs, including the cathelicidin peptide LL-37 18 , hepcidin, 19 β-defensin 2, and lipocalin 2. 20 MSC-produced AMPs are thought to be one critical component regulating the ability of MSC administered therapeutically to control or eliminate bacterial infections, as explored in multiple animal models. 4,18,21 A number of in vivo mouse models have explored the effect of MSC on acute bacterial infections. For example, human MSC decreased bacterial burden in a mouse model of Escherichia coli pneumonia, 18 and also Klebsiella pneumoniae pneumosepsis. 22 In another study, human MSC also reduced mortality associated with P. aeruginosa in a mouse peritonitis and sepsis model. 23 MSC have also been shown to augment antibiotic treatment effects in murine cystic fibrosis, 24 in part by the secretion of LL-37. 16 Lastly, instillation of MSC into airways of explanted lungs have been shown to decrease E. coli bioburden, and ameliorated acute lung injury including alveolar fluid clearance and inflammation. 25 However, few studies to date have investigated whether MSC can also be used in the treatment of chronic bacterial infections. Notably, our group recently reported that activated murine MSC were effective for the treatment of chronic S. aureus biofilm infection in a mouse implant infection model. 4 In this model, we found that activated MSC delivered systemically demonstrated strong antibacterial activity and elicited resolution of wound infection when combined with antibiotic therapy.
Importantly, in the same study we also showed that in pet dogs with spontaneous multidrug resistant (MDR) wound infections, systemic administration of activated canine MSC cleared bacterial infection, even when administered to animals with infections that had persisted for months with antibiotic treatment alone. 4 Based on these compelling data from realistic animal models of

| Flow cytometric assessment of bacterial killing
Bacterial killing assessment using flow cytometry was performed according to manufacturer's instruction using LIVE/DEAD BacLight Bacterial Viability and Counting Kit (Thermo Fisher Scientific). Histograms were generated using FlowJo 10.5 software.

| Direct bacterial killing assay (BKA)
BKA were performed as previously described. 4 Conditioned medium (CM) from human BM-MSC was generated by plating 5 × 10 5 cells per well in a 24 well plate with 500 μL per well of antibiotic free media containing: DMEM, 10%FBS, Glutamax 1x, NEAA 1x, Essential AA 1x and FGF-basic, then incubating at 37 C in a 5% CO 2 incubator.
Conditioned medium was collected 24 hours post plating and immediately frozen at −80 C. The CM was thawed prior to use and cellular debris was removed by centrifugation. For bactericidal activity assessment, 50 μL of log phase S. aureus cultures (OD600 of 0.6, corresponding to 7.5 Log10 CFU/mL) were inoculated in 500 μL of MSC CM in 24-well plates. Cocultures of bacteria and CM were incubated at 37 C in ambient air for 3 hours, then numbers of viable bacteria were determined by plating log 10 serial dilutions on LB agar 4 quadrant plates (Thermo Fisher Scientific) and manual counting of colonies 24 hours later according to previously published protocol. 26 The ability of MSC CM to augment the bactericidal activity of major antibiotic classes was determined by incubating bacteria in MSC CM with the addition of subtherapeutic concentrations of antibiotics. The antibiotic concentrations to be used were determined in advance by titration and elucidation of minimal bactericidal concentrations. 26 The antibiotic concentrations used in these assays were as follows: cefazolin (375 ng/mL), gentamicin (200 ng/mL), vancomycin (500 ng/ mL), enrofloxacin (2 μg/mL), imipenem (30 ng/mL), and daptomycin (50 ng/mL). All antibiotics were obtained from Sigma-Aldrich    Images (9 per well) were collected every 15 minutes using a ×10 objective, and analyzed using IncuCyte S3 Software (Essen BioScience Inc).

| Neutrophil extracellular trap assay
In all, 500 000 human neutrophils were plated on 0.01% poly-L lysine (Sigma-Aldrich) coated coverslips (Chemglass Life Sciences LLC) in 24-well cell culture plates, then incubated with 0.5 mL MSC CM for 3 hours. Studies were done using neutrophils obtained from three unrelated, healthy donors. IRB approval was obtained for collection of human blood samples with signed informed consent. After washing off the CM, S. aureus was added at an MOI of 1 for the indicated time points in HBSS containing calcium, magnesium and 10% autologous human serum. Staining for neutrophil extracellular trap (NET) formation was performed according to a published protocol. 28,30 Cells were immunostained with antibodies for detection of histone H3 using Anti Histone H3 (RM188) Antibody (Caymen Chemical, Ann Arbor, Michigan) and neutrophil elastase Anti-Neutrophil Elastase antibody (ab21595) (Abcam) with slight modifications for staining in a 24-well plate. Images were taken on Olympus IX83 spinning disk confocal microscope, by imaging 15 random fields per condition. The NET area was calculated using ImageJ 31 ; and the NET area (in pixels) was determined by merging channel 2 and 3 pixels (representing histone H3 and neutrophil elastase staining, respectively) then normalized to DAPI channel area.

| RT-qPCR analysis of AMP expression
To assess mRNA expression for AMPs, RNA isolated using the

| Mesh implant biofilm animal model
All procedures involving live animals were approved by the Institutional Animal Care and Use Committee at Colorado State University.
S. aureus coated surgical mesh was implanted subcutaneously in nu/nu mice (Charles River Laboratories) as previously described. 4 The mesh was coated with S. aureus (Xen36 strain) engineered to express lucif-

| Statistical analyses
Statistical comparisons between two treatment groups were done using nonparametric t tests (Mann-Whitney test). Comparisons between three or more groups were done using one-way ANOVA, followed by Tukey multiple means post-test. Tests for synergy were performed using two-way ANOVA with significant P ≤ .05 interaction factors denoting synergistic interactions. 33

| MSC spontaneously produce factors with bactericidal activity
The first studies were done to confirm that MSC produce antimicrobial factors spontaneously, as has been reported previously. 5 We found that when S. aureus and E. coli were incubated with MSC CM, strong bactericidal activity was detected, and caused a decrease of greater than 2log CFU/mL of bacterial growth ( Figure 1A). Moreover, incubation of a drug-resistant strain of MRSA with MSC CM induced a 1.4-fold decrease in bacterial CFU ( Figure 1B). It should also be noted that the bactericidal activity of MSC CM was titratable (data not shown). This spontaneous bactericidal activity was observed using three different donor MSC (two male and one female; data not shown). Passage of MSC also did not alter bactericidal activity, at least for up to nine passages ( Figure 1C). For example, MSC CM from passage nine cultures exhibited the same degree of bactericidal activity as passage 1 cultures, with up to 66% killing.

| MSC CM acts synergistically with antibiotics to generate bactericidal activity
Previous studies have determined that the primary mediators of MSC secreted bactericidal activity are AMP. 5 It has also been reported previously that AMP such as LL-37 can synergize with conventional antibiotics for bacterial killing. 34,35 Therefore, we conducted studies to determine whether MSC CM contains factors that synergize with or exhibit additive bactericidal activity with conventional antibiotics, and to determine whether these effects are limited to only certain classes of common antibiotics.

| Expression of AMPs by MSC
Expression of AMPs by human MSC has been reported previously. 5,18 To confirm the expression by the MSC used in our system, we examined  (Figure 2A,B).

| MSC activation by innate immune pathways alters AMP expression and cytokine production
Previous studies have shown that activating MSC with innate immune stimuli increases their immune modulatory properties. 27 In attempt to define their direct antimicrobial properties, we conducted studies to determine whether MSC activation through major innate immune pathways (Toll-like receptors, NOD-like receptors, cytokines) leads to an increase in factors associated with bactericidal activity (Figure 3A-E).
We found that MSC stimulation by CpG oligonucleotides caused the F I G U R E 3 Legend on next page. greatest increase in expression of AMP genes ( Figure 3F). However, none of the stimuli evaluated produced an actual increase in direct bactericidal activity (data not shown). We concluded therefore that MSC secretion of bactericidal factors (ie, AMPs) was primarily a constitutive process, and largely independent of cell activation status. Secretion of cytokines related to innate immune cell recruitment (ie, MCP-1 and IL-8) was however responsive to MSC activation, particularly by the TLR3 agonist pIC (Figure 3G,H). Thus, MSC activation appears to be more relevant to their interaction with host innate immune cells.

| MSC factors trigger rapid bactericidal activity
We next asked the question of how rapidly MSC secreted factors could kill bacteria. Flow cytometry was used to assess bacterial cell membrane permeability (the first step in bacterial cell death). We observed that as soon as 15 minutes after exposure to MSC CM, bacterial cell death could be detected ( Figure 4A). Bacterial death continued to increase for up to 3 hours, at which time 98% of the culture consisted of bacteria with disrupted membrane integrity ( Figure 4B). Thus, MSC CM factors triggered bacterial killing extremely rapidly, consistent with what has been reported previously for AMP-mediated killing. 4,21

| MSC secreted factors disrupt biofilm formation
In addition to killing planktonic bacteria, factors secreted by MSC may also disrupt biofilm formation, as has been suggested previously. 4 To examine this effect in greater detail, we determined whether MSC CM could prevent biofilm formation, or disrupt bacterial biofilms once they had already formed. We observed that addition of MSC CM prevented initial adhesion and formation of MRSA biofilms ( Figure 4D). Importantly, addition of MSC CM also disrupted the biofilm when added to preformed biofilms ( Figure 4E

| Treatment with activated MSC CM increases NET area
NETs are produced by neutrophils following contact with bacterial products and certain pro-inflammatory cytokines; and are an important mechanism by which neutrophils can kill bacteria and prevent their spread into tissues. We observed that when neutrophils were incubated with CM from activated MSC, there was a significant increase in the total NET area produced per cell after contact with S. aureus ( Figure 6A). The activated MSC also induced significantly greater NET area formation than CM from nonactivated MSC. The effects of MSC CM on NET formation were detected as soon as 30 minutes after exposure ( Figure 6B,C). Thus, induced NET formation is another mechanism by which MSC could stimulate the innate immune system to increase bacterial elimination and chronic infection control. Error bars represent mean and SD from three replicates. C, Bacteria (USA-300) grown as biofilms were incubated with control medium and MSC CM for the indicated time points, then stained with LIVE/DEAD BacLight kit as described in the Materials and Methods section to identify live and dead bacterial colonies, as revealed by immunohistochemical staining and evaluation by confocal microscopy. Green (SYTO9) represents live bacterial clusters, whereas red clusters represent dead bacteria stained with propidium iodide. Merged channels show yellow color as red and green overlap. Right column "MSC-CM" shows MRSA biofilm incubated for 2 (top) or 24 (bottom) hour with MSC conditioned medium. Left column "MRSA biofilm" grown in DMEM medium only with additives matched to MSC-CM. Images taken with ×10 objective. 4D, Prevention of biofilm formation by MSC CM (compared with control medium) as assessed using S. aureus biofilm assays, as noted in the Materials and Methods section. The Y axis depicts bacterial colonies, quantitated using crystal violet staining after 72 hours in culture. Blue shows the biofilm grown in DMEM media with all additives, red shows biofilm with the addition of MSC CM. E, Effects of MSC CM on pre-formed biofilms following 2 or 24 hours of exposure. Bars depict the ratio of live vs dead bacteria in biofilms, as quantitated using ImageJ software, as described in the Materials and Methods section. Statistical significance was determined for *P ≤ .05, **P ≤ .01, ***P ≤ .001, ****P ≤ .0001 as assessed by one-way ANOVA and Tukey multiple means post-test. Each experiment was conducted using CM from three different donor MSC. The figures depicted are representative of the results obtained in three independent experiments

| Treatment with activated MSC decreases bacterial burden in a mouse model of chronic biofilm infection
We previously reported that activated mouse MSC were able to effectively control and eliminate biofilm infection in a mouse model when coadministered with a beta-lactam antibiotic (amoxi-clav). 4 However, it has not been determined previously whether human MSC exhibit similar activity against chronic S. aureus biofilm infections. To address this question, nu/nu mice were implanted with S. aureus coated surgical mesh as previously described 4 Figure 7A). Quantitation of bacterial burden using luciferase imaging gives similar results ( Figure 7B). In addition, the overall wound surface area was significantly smaller in the MSC-treated group compared with control animals ( Figure 7C), decreasing from an average diameter of 29.8 mm 2 to 22.5 mm 2 . There was also less purulent material within the abscesses of MSC treated animals, rendering the mesh F I G U R E 6 Effects of mesenchymal stem cells (MSC) conditioned medium (CM) on neutrophil extracellular trap (NET) formation. A, Neutrophils were incubated with MSC CM or medium, then incubated with live S. aureus for 30 minutes or 2 hours, as noted in the Materials and Methods section, then fixed and immunostained for detection and quantitation of NET formation, using confocal microscopy. Total NET area was normalized to DAPI cell count, and was digitized and quantitated using ImageJ software, as described in the Materials and Methods section. Bars depict the total area at 30 minutes (black) or 2 hours (gray) following exposure to S. aureus. *** denotes P < .0005 as assessed by ANOVA and Tukey multiple means post-test. Each experiment was conducted using MSC CM obtained from three different donor MSC, and neutrophils were collected from three unrelated healthy donors. B, Representative ×10 magnification images of NET formation by neutrophils 30 minutes (top row) or 2 hours (bottom row) after exposure to S. aureus. Red, green, and blue depict histone H3, neutrophil elastase, and DAPI expression, respectively. The upper right corner of each image depicts shows NET total area, calculated by Image J software, with colors inverted for clarity. C, Representative ×40 magnification images of neutrophil NETs, imaged under same conditions as described for (B) material more readily visualized ( Figure 7D). These results indicate therefore that activated human MSC are also effective in reducing chronic bacterial infections, similar to what has been reported previously for activated canine and murine MSC. 4

| DISCUSSION
MSC have been widely evaluated experimentally and clinically for their ability to stimulate wound healing and suppress inflammation. [36][37][38] However, there is comparatively less research regarding the use of MSC for anti-infective therapy. 4,21 In particular, even less is known regarding the use of MSC for treatment of chronic bacterial infections, which typically involve the formation of bacterial biofilms and high levels of antimicrobial resistance. 4,5,39 Therefore, in the present study, which follows a prior publication by our group demonstrating the utility of activated MSC for treating chronic MDR bacterial infections 4 ; we have now examined in detail the direct and indirect antimicrobial properties of human MSC. Notably, these studies confirmed and extended our understanding of the direct antibacterial activity of MSC reported previously, and also elucidated new indirect mechanisms involving activation of host innate immune defenses. These host defenses included MSC-induced augmented NET formation by neutrophils. Although much of the direct antibacterial activity of MSC involving production of AMPs has been previously reported, little has been noted previously regarding the ability of MSC to disrupt preexisting bacterial biofilms. 39 Here, we also report that newly forming or already formed biofilms can be disrupted by MSC, highlighting the potential for MSC to be used as a treatment for chronic infections, including those often associated with biofilms. 40, 41 We also found that MSC produced factors that F I G U R E 7 Treatment of chronic biofilm infection by activated mesenchymal stem cells (MSC). Mice (n = 6 per group) were implanted with S. aureus infected mesh, then treated with activated MSC and amoxi-clav, or amoxi-clav only, as described in the Materials and Methods section. A, Bacterial bioburden in wound tissues (CFU/wound tissue) at the end of the 12 day study. Results pooled from four independent experiments. B, Luminescent imaging of wound bioburden, determined using an IVIS unit, and converted to area under the curve (AUC) Results pooled from three independent experiments. C, Mean measured wound area (mm 2 ) for each treatment group of mice. Results pooled from four independent experiments. D, Representative digital camera images of wound tissues immediately following euthanasia, obtained from treated and control mice. Lower two images showing representative IVIS imaging from treated and control mice. With red circle showing the field used to calculate radiance in ROI, radiance color scale shown in right bar. * denotes P < .05 as assessed by two tailed nonparametric t test and Mann-Whitney post-test enhance the bactericidal activity of all major classes of antibiotics evaluated, including enhancement of antibiotic activity against MDR bacteria.
Production of AMP is one of the most widely investigated direct antimicrobial mechanisms of MSC. 5,18,20,21,42 Previous studies have reported spontaneous production of several different AMP by human and animal MSC, as revealed by bactericidal assays. 18,21 Although most studies have reported that MSC constitutively produce antimicrobial factors, there are conflicting reports as to whether MSC activation with TLR ligands or cytokines enhances AMP production and bactericidal activity. 5,42 In our studies, which investigated the effects of MSC activation on transcription of AMP genes and induction of bactericidal activity, we observed no net increase in in vitro bactericidal activity over that generated by nonactivated MSC. Thus, we concluded that the overall net effect of MSC activation with innate immune stimuli was not reflected by increased bactericidal activity; and production of antimicrobial factors appears to be a constitutive property of MSC.
We also discovered several important indirect mechanisms by which MSC may generate antibacterial activity in vivo, including induction of NET formation by neutrophils and increased neutrophil phagocytosis. It is also important to note that activation of MSC with TLR ligands, particularly the TLR3 ligand pIC, enhanced production of factors secreted by MSC which augmented host innate immune responses to bacterial infections.
For example, exposure of neutrophils to activated MSC CM elicited significantly greater bacterial phagocytosis than CM from nonactivated MSC ( Figure 5). These results are consistent with those of Brandau et al, who also reported that neutrophil phagocytosis was enhanced following exposure to CM from LPS-activated MSC. 43 The same group also reported increased survival of neutrophils following exposure to MSC CM. 44 We also discovered that CM from activated MSC triggered a significant increase in neutrophil NET area formation, compared with CM from nonactivated MSC or to control neutrophils exposed only to S. aureus. This is an important effect of MSC, because formation of NETs is an important mechanism by which neutrophils can contain and eliminate bacterial infections. 11 Thus, activation of MSC appears to be an essential step in maximizing the interaction of MSC with host innate immune defenses to increase bacterial killing. One possible mediator of increased NET formation in response to MSC CM is IL-8 45 , which has been reported as a key inducer of NET. 46 Thus, our finding that pIC activation triggered increased secretion of IL-8 by MSC is consistent with a possible IL-8 mechanism for NET formation.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.