Blood from healthy human donors was supplied by the Dutch blood bank (Sanquin, Amsterdam, The Netherlands). Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats using Ficoll-Paque (Pharmacia, Uppsala, Sweden) density centrifugation. Monocytes were positively selected by magnetic-activated cell sorting (MACS) with anti-CD14 labelled microbeads (Miltenyi Biotec, 130050201) according to the manufacturer’s instructions.

Isolated monocytes were seeded in a 24-wells plate at a density of 3 x 10⁵ cells/well, except when stated otherwise. When monocytes were seeded on top of 13-mm diameter titanium disks (Alfa Aesar, 10385-HP), these were placed in the 24-wells plate before seeding. Monocytes were differentiated into macrophages by culturing for 7 days at 37°C, 5% CO₂ in α-Minimum Essential Medium (α-MEM, Gibco Paisley, 22561021) supplemented with 10% (v/v) hyFBSclone fetal bovine serum (hyFBS, Biowest, HYCLSV30160), 100 U/mL penicillin- streptomycin (1% p/s, Gibco, 15140122), and 40 ng/mL human recombinant M-CSF (Peprotech, 300-25). Culture media was refreshed after 3-4 days.

Mesenchymal stem cells (MSCs) were isolated from human bone marrow aspirates upon informed consent. The aspiration procedure was approved by the local medical research ethics committee, University Medical Center Utrecht, under the protocols METC 08-001/K and METC 07-125/C.

Aspirates were diluted in PBS, filtered through a 100 µm cell strainer and the mononuclear cell layer was collected after Ficoll-Paque density centrifugation. Approximately 2.5 x 10⁵ mononuclear cells were plated per cm² in MSC expansion medium consisting of α-MEM supplemented with 10% (v/v) heat-inactivated FBS (FBS, Biowest, S181H), 1% p/s, 0,2 mM L-ascorbic acid-2-phosphate (ASAP, Sigma-Aldrich, A8960) and incubated at 37°C, 5% CO₂. Cells starting from passage 3 were used in the experimental setups.

To build the macrophage-MSC co-culture the following steps were adopted. Monocytes were seeded in 24-wells plates at a density of 3 x 10⁵ cells/well and differentiated into macrophages by culturing for 7 days. Meanwhile, MSCs were cultured to reach ~70% confluency until monocytes fully differentiated into macrophages. Then, MSCs were fluorescently labelled with celltrace violet (Invitrogen, C34557) diluted in Hanks balanced salt solution (HBSS) according to the manufacturer’s instructions for labelling adherent cells. After staining, MSCs were detached with 0,25% trypsin/EDTA (Gibco, 25200056) and re-seeded together with macrophages at a density of 1 x 10⁵ cells/well according to the experimental setup. Cells were allowed to adhere for about ~24 h before readouts started or bacteria were added, as explained below in the section “Multicellular infection model to study intracellular survival of bacteria”.

Cell seeding densities and culture plate format were adjusted according to the experimental setup. For instance, when measuring cytokine production, monocytes and MSCs were combined in a 96-well plate at a density of 1.5 and 0.5 x 10⁵ cells/well, respectively. Moreover, when the identification of single cell types was not relevant for the experimental setup, fluorescence labelling was omitted.

All experiments used GFP-labelled Staphylococcus aureus (kind gift from Prof. Simon Foster) and Staphylococcus epidermidis (kind gift from Prof. Leo Koenderman) were transformed with a GFP-expressing plasmid pCM29 to constitutively express GFP, as previously described (Boero et al., 2021). Bacteria were grown overnight in Todd-Hewitt broth (THB) with 10 ng/mL chloramphenicol to reach stationary phase.

A solution of silver ions (AgNO₃) was prepared by dissolving silver nitrate (Sigma-Aldrich, S6506) in ultrapure water. Commercially available 20 nm and 100 nm silver nanoparticles (AgNP) (Alfa Aesar, J67067 and J67099), were used. Before each experiment, both sizes of nanoparticles were pelleted by centrifugation for 30 min at 4°C with 17000 x g (20 nm AgNP) or 300 x g (100 nm AgNP). A stock solution of 80 µg/mL was prepared for each type of nanoparticle in ultrapure water. AgNP and AgNO₃ were diluted to various concentrations in α-MEM with 10% FBS or in THB to test their effects on host or bacterial cells.

Monocytes and MSCs were seeded in a 96-wells plate at a density of 1.5 and 0.5 x 10⁵ cells/well, respectively. The same numbers of cells were combined in the co-culture, where macrophages were combined with MSCs as previously described.

Macrophages, MSCs, and the co-culture were incubated with fresh media containing 10 ng/mL LPS O111:B4 from Escherichia coli (Sigma-Aldrich), or various concentrations of AgNO₃, 20 nm AgNP, and 100 nm AgNP. After 24 h stimulation, the culture medium was harvested to measure cytokine production and cells were processed for viability evaluation. When measuring cell viability, the culture medium was replaced by α-MEM with 10% FBS and 10% Alamar Blue solution prepared by dissolving Resazurin sodium salt (Sigma-Aldrich, R7017) in PBS. Then cells were incubated at 37°C for 2-3 h, in the dark. Next, the supernatant was transferred to a new plate and fluorescence was measured at 530-10/580-10 nm (Ex/Em) with a Clariostar plate reader (BMG labtech). Background fluorescence values were subtracted, and metabolic activity was normalized to the control sample. Production of cytokine TNF-α and IL-6 was measured in the collected supernatant by ELISA (Duoset, R&D Systems, DY210 and DY217B), according to the manufacturer’s instructions. ELISA values are expressed as fold-change over non-stimulated controls. Samples were analyzed in triplicates and experiments repeated three times with different monocytes and MSCs donors.

The direct antimicrobial properties of silver were determined by measuring OD (600nm), combined with the broth microdilution method. Overnight bacterial culture was diluted in THB to reach a final inoculum of 5 x 10⁵ colony-forming units per mL (CFU/mL). In a flat-bottom 96-well plate, the bacterial suspension was mixed with AgNO₃, 20 nm AgNPs, or 100 nm AgNP in equal parts in triplicates, with a final volume of 200 μL. Bacterial growth was monitored at 37°C by measuring OD (600nm) every 5 min, for a total time of 12 h, on a Clariostar plate reader with gentle shaking before each measurement. Then, bacterial suspensions were serially diluted and plated on Todd-Hewitt agar (THA), and colonies counted after overnight incubation at 37°C. Samples were analyzed in triplicates.

An overnight bacterial culture was diluted in α-MEM to reach a final concentration of 1 x 10⁷ CFU/mL. To mimic the in vivo situation, bacteria were incubated with 5% normal human serum (NHS) for 15 min at 37°C, which coats them with antibodies and complement (opsonization) to enable their uptake by immune cells. Serum was collected from blood obtained from healthy donors after informed consent, as previously described (Boero et al., 2021). Approval from the Medical Ethics Committee of the University Medical Center Utrecht was obtained (METC protocol 07-125/C approved on March 1, 2010).

Opsonized bacteria were added to the co-culture at various multiplicity of infection (MOI). To synchronize bacterial uptake, plates were centrifuged for 5 min with 110 x g at RT, and then incubated at 37°C, 5% CO₂. To study intracellular bacterial survival, the cells were washed twice after 30 min of infection to remove free bacteria and cultured in media supplemented with 100 µg/mL gentamicin (Serva, 22185.02, to kill the bacteria) and 20 µg/mL lysostaphin (Bioconnect, MBS635842, to lyse the bacteria and lose the GFP signal) for 1 h. Afterwards, cells were washed twice and incubated in media with only 5 µg/mL gentamicin. All washing steps were performed with warm α-MEM. This treatment allows only intracellular bacteria to survive, as both gentamicin and lysostaphin are unable to penetrate mammalian cell membranes within short time periods (Hamza et al., 2013; Hamza and Li, 2014). To verify treatment efficacy in lysing bacteria, serum-opsonized S. aureus was incubated for 1 h at 37°C in α-MEM in presence of 100 µg/mL gentamicin and 20 µg/mL lysostaphin. Then, GFP expression was measured with a MACSquant VYB (Miltenyi Biotech) flow cytometer and data were analyzed with FlowJo (v.10.1., FlowJo LLC).

At the desired time points, co-culture samples were processed for flow cytometry analysis, microscopy observation, or quantification of intracellular bacteria via CFU count.

For flow cytometry, cells were detached by a combination of trypsin and eventually gentle scraping in 1 mM DPBS/EDTA if cells were still attached to the bottom of the culture plate. Cells were transferred to a 96-wells plate and stained with sytox orange dead cell stain for flow cytometry (Invitrogen, S34861) according to the manufacturer’s instructions. Samples were measured with a MACSquant VYB flow cytometer and data were analyzed with FlowJo. The gating strategy is summarized in Supplementary Figure 1 . Briefly, a total of 10000 events were collected for each sample gated based on forward scatter (FSC) and side scatter (SSC) parameters. The two cell types were selected based on the signal of CellTrace violet. Non-infected samples from the sytox negative population were used to set GFP fluorescence baseline and define the proportion of infected, GFP-positive cells.

For confocal imaging, cells were collected as described for flow cytometry analysis and fixed in 1.5% paraformaldehyde. Cell membranes were stained with 3 µg/mL Alexa Fluor 647-conjugated Wheat Germ Agglutinin (WGA, Invitrogen, W32466) for 10 min at RT, on a shaking plate. Then, samples were transferred to CELLview slide (Greiner Bio-One, 543079) previously coated with poly-L-lysine (Sigma-Aldrich, P4707), and imaged on a Leica TCS SP5 microscope with a HCX PL AP CS 63x/1.40-0.60 OIL objective (Leica Microsystems). Images were adjusted for publication using Image J Fiji.

Finally, to quantify the number of intracellular bacteria, cells were lysed with 0.1% Triton X-100 and plated on THA plates in serial dilutions. Plates were incubated overnight at 37°C after which colonies were counted.

A variation of this model was used to study silver effects. Briefly, cells were incubated for 24 h with different concentrations of AgNO₃ and 20 nm AgNP before adding S. aureus at a MOI=10. Samples were analyzed by flow cytometry and CFU counting at 30 min and 4 h after infection. In this setup, when cells were incubated with silver, gentamicin and lysostaphin treatment was not employed for the 4 h time point. Samples were analyzed in triplicates and experiments repeated three times with different monocytes and MSCs donors.

GraphPad Prism 9 (version 9.3) was used to create the graphs and determine statistical significance via a two-way or one-way ANOVA, or t-test. p<0.05 was considered statistically significant. Illustrations were created with BioRender.com.

To build a comprehensive, multicellular in vitro model that mimics the main players in IAI, we co-cultured primary MSC and monocyte-derived human macrophages as summarized in Figure 1 . First, monocytes were isolated from human blood and seeded on a surface to differentiate into macrophages ( Figures 1A, B ). Cells were seeded either on a cell culture plate or on any biomaterial mimicking an implant. When comparing plastic with titanium, the surface itself seemed to have minimal impact on the phagocytic capacity of the macrophages ( Supplementary Figure 2A ). After differentiation, macrophages were re-seeded together with MSCs, which can differentiate into osteoblasts, to create an IAI-like environment ( Figure 1C ). In order to distinguish each element in the co-culture, we labeled at least one cell type before mixing with the second one. The co-culture was viable for up to several days, enabling the verification of cytotoxicity, modulation of cell functions, and antibacterial effects over a relatively long-time span.

The co-culture of macrophages and MSCs was then exposed to GFP-expressing bacteria ( Figure 1D ). We verified that this model could be used to study infections caused by different bacterial species relevant in the orthopedic field, such as S. aureus and S. epidermidis. Although macrophages were equally capable of associating with both bacterial species, we observed that MSCs were more susceptible to S. aureus than S. epidermidis infection ( Supplementary Figure 2B ). To assess the efficiency of antibacterial treatments in both macrophages and MSCs, all further experiments were conducted with S. aureus.

At the desired time points, the co-culture was processed for (confocal) microscopy ( Figure 1 ), flow cytometry ( Supplementary Figure 2 ), or bacterial enumeration via CFU plating. Flow cytometry offered the possibility to track intracellular bacterial survival per host-cell type over time by including gentamicin and lysostaphin treatment. These two compounds are not membrane-permeable and therefore selectively kill and lyse extracellular bacteria (Hamza et al., 2013; Hamza and Li, 2014). We confirmed by flow cytometry that the treatment efficiently lysed bacteria, as no GFP signal could be detected from S. aureus already after 1 h incubation with gentamicin and lysostaphin ( Supplementary Figure 2C ). Moreover, microscopy images confirmed that all GFP signal detected by the flow cytometer originated from intracellular bacteria and not from membrane-bound or extracellular S. aureus ( Figure 1D ). These results confirm the validity of the model we developed to study intracellular bacteria.

As expected, after 30 min of direct contact between bacteria and host cells, almost all macrophages had engulfed at least one S. aureus bacterium. In comparison, less than half of the MSCs population had engulfed bacteria. After 24 h, the proportion of infected macrophages was significantly reduced both when cultured alone, and in the co-culture. Interestingly, the proportion of infected MSCs was only reduced when cultured alone, but not in co-culture ( Supplementary Figure 2D ). This highlights distinct behaviors between cells cultured alone or together.

We tested the validity of our co-culture model by using it to determine a possible therapeutic window for silver as an antibacterial treatment. For this purpose, silver must be at a concentration that is antibacterial, but not toxic to human cells. Therefore, we incubated S. aureus, macrophages, and MSCs with several concentrations of AgNO₃, 20 nm AgNP, and 100 nm AgNP.

We observed that at the lowest concentration tested, AgNO₃ reduced bacterial growth and significantly decreased the number of viable S. aureus,while complete inhibition of growth and killing was achieved after exposure to AgNO₃ concentrations higher than 0.0017 µg/mL ( Figures 2A, B ). Although bacterial growth speed was affected by increasing the concentrations of 20 nm AgNP ( Figure 2D ), no bacteriostatic or bactericidal effects were observed after incubation with 20 nm AgNP and 100 nm AgNP ( Figures 2D, E, G, H ).

On the other hand, AgNO₃ completely abolished the metabolic activity of host cells starting at 0.017 µg/mL ( Figure 2C ), while exposure to AgNP, regardless of size or concentration, partially affected the metabolism of both cell types. For instance, concentrations of 20 nm AgNP higher than 1 µg/mL reduced the total metabolic activity of the macrophage population by up to ~25% ( Figure 2F ). Incubation with any concentration of 100 nm AgNP reduced the viability of macrophages, while the metabolic activity of MSCs was affected only at concentrations higher than 10 µg/mL ( Figure 2I ). Nonetheless, we exclude a toxic effect derived from exposure to nanoparticles as more than 75% of the total metabolic activity remained for both the macrophages and MSCs populations.

In addition, we assessed whether silver triggered an inflammatory response by measuring the secretion of TNF-α and IL-6. Overall, non-toxic silver concentrations did not elicit any inflammatory response in monocultured cells compared to control samples ( Figure 3 ). However, from this assay we could observe the added value of co-culturing different cell types. Addition of MSCs reduced LPS-mediated activation of macrophages, as observed by decreased production of TNF-α. In addition, including immune cells reduced IL-6 production by MSCs. Moreover, exposure to 100 nm AgNP at 20 µg/mL and at concentrations higher than 5 µg/mL significantly increased TNF-α ( Figure 3C ) and IL-6 ( Figure 3F ) levels respectively in the co-culture.

Once we established the effects of silver in simple in vitro models, we could further explore the consequences of its use in an IAI-like environment thanks to our multicellular model to study infection. Besides defining the antibacterial, non-toxic, and non-inflammatory concentrations of AgNO₃ and AgNPs, we explored the impact of silver on the antibacterial functions of host cells ( Figure 4A ).

Due to their toxicity to host cells, we excluded AgNO₃ at concentrations of 0.17 µg/mL and 1.7 µg/mL from the assays. Only three representative concentrations of 20 nm AgNP were selected, as similar cytotoxicity and antibacterial activity were observed. Analysis of samples by flow cytometry showed a remarkable shift in the side-scatter (SSC) values for macrophages in the co-culture when exposed to AgNP for 24 h ( Supplementary Figure 3A ). This increase in cellular granularity was concentration dependent and particularly evident in the presence of 100 nm AgNP. The effect was probably caused by the internalization of nanoparticles by macrophages, as shown by bright field microscopy ( Supplementary Figure 3B ). For this reason, 100 nm AgNPs were excluded from the analysis with the multicellular model.

Despite the previously observed antibacterial activity of AgNO₃ ( Figures 2A, B ), this was not strong enough to counteract bacterial growth within the time frame tested in the multicellular model. In spite of the presence of AgNO₃ or 20 nm AgNP, after 4 h incubation, the bacteria had already overtaken and killed most macrophages and MSCs in culture as observed by the increased amount of cells positive for sytox staining ( Supplementary Figure 4 ).

Following S. aureus introduction into the co-culture, silver treatment did not influence bacterial uptake by host cells. We confirmed the presence of intracellular S. aureus in both cell types. As observed before, macrophages phagocytosed more bacteria than MSCs ( Figure 4B ). Depending on the donor, silver treatment had either no impact on phagocytosis or eventually led to an increased uptake of bacteria by both cell types. Moreover, exposure to silver did not reduce the amount of S. aureus able to survive intracellularly compared to the control samples ( Figure 4C ). Confocal microscopy imaging confirmed the flow cytometry results. Higher numbers of intracellular S. aureus were observed within macrophages than MSCs, while exposure to AgNO₃ or AgNP did not influence bacterial uptake ( Figure 4D ).