An in vitro platform for elucidating the molecular genetics of S. aureus invasion of the osteocyte lacuno-canalicular network during chronic osteomyelitis

Abstract

Staphylococcus aureus osteomyelitis is a devasting disease that often leads to amputation. Recent findings have shown that S. aureus is capable of invading the osteocyte lacuno-canalicular network (OLCN) of cortical bone during chronic osteomyelitis. Normally a 1 μm non-motile cocci, S. aureus deforms smaller than 0.5 μm in the sub-micron channels of the OLCN. Here we present the μSiM-CA (Microfluidic – Silicon Membrane – Canalicular Array) as an in vitro screening platform for the genetic mechanisms of S. aureus invasion. The μSiM-CA platform features an ultrathin silicon membrane with defined pores that mimic the openings of canaliculi. While we anticipated that S. aureus lacking the accessory gene regulator (agr) quorum-sensing system would not be capable of invading the OLCN, we found no differences in propagation compared to wild type in the μSiM-CA. However the μSiM-CA proved predictive as we also found that the agr mutant strain invaded the OLCN of murine tibiae.

Graphical Abstract Text:

We successfully developed the μSiM-CA (Microfluidic – Silicon Membrane – Canalicular Array) platform to distinguish the phenotype of mutant S. aureus strains based on their propagation through nanopores that mimic canaliculi invaded during chronic osteomyelitis. In doing so, we demonstrated the application of the μSiM-CA as a predictive tool for canalicular invasion by studying a putative virulence regulator of S. aureus, the accessory gene regulator, in the μSiM-CA and in an in vivo model of implant-associated osteomyelitis.

Introduction

Staphylococcus aureus is the most common pathogen isolated from chronic osteomyelitis,¹ with over 50% of cases caused by hard-to-treat methicillin resistant S. aureus (MRSA) strains.² Rigorous intervention studies (i.e. the Surgical Care Improvement Project (SCIP))³ have demonstrated that infection rates for elective surgery cannot be reduced below 1-2%,⁴ and rates of recurrent or persistent infection following a two stage revision surgery are as high as 33%.⁵ Despite strategies such as surgical site debridement, complete hardware exchange and aggressive long-term antimicrobial therapy, infection continues to recur. In total, the cost for treatment of implant-associated osteomyelitis is projected to exceed $1.62 billion by 2020.⁶

Given the high financial burden and patient morbidity caused by implant-associated osteomyelitis, it is important to elucidate the pathogenesis of S. aureus bone infection. The invasiveness of S. aureus infection is attributed to its arsenal of virulence factors and resistance mechanisms including: secreted toxins, immune evasion,⁷ intracellular persistence,^{8, 9} biofilm formation,^{10, 11} the creation of slow growing small colony variant (SCV) subpopulations,^{12, 13} and the development of antibiotic resistance.¹⁴ Recently, we have shown that S. aureus is also capable of invading the osteocyte lacuno-canalicular network (OLCN) of live bone, by deforming from spherical cocci into submicron rod-shaped cells, and propagating throughout the channels where they avoid the immune system. This phenomenon has been observed experimentally in a murine model for chronic osteomyelitis,¹⁵ and in infected foot bones of patients with diabetic ulcers.¹⁶ Additionally, S. aureus colonization of the OLCN was shown to be an active process that involves bacterial replication at the leading propagating cells, rather than dormant persistence.¹⁵ Previous studies have identified intracellular invasion of osteoblasts,^17-19 and osteocytes,⁸ as a possible mechanism for S. aureus persistence in chronic osteomyelitis. However, our transmission electron microscopy (TEM) studies of infected bone failed to substantiate these in vitro observations,^{15,16,20, 21} suggesting that colonization of the OLCN is the dominant mechanism by which S. aureus chronically infects bone. The mechanisms regulating S. aureus propagation through the OLCN are not known.

Animal models of implant-associated S. aureus infections have provided important insights on S. aureus biofilm formation, maturation, and bacterial emigration to perpetuate chronic infections.^{20, 21} However, due to the immense time and cost associated with studying the invasion of the OLCN in vivo, the development of an in vitro canalicular invasion assay is imperative. We previously used ultrathin silicon nitride (SiN) membranes with precisely patterned 0.5 μm pores to confirm the ability of S. aureus to propagate through canalicular-sized openings half the normal size of the cocci.¹⁵ Here we advance and validate this tool as a platform for discovering the mechanisms underlying S. aureus propagation within cortical bone.

First developed in 2007,²² silicon ‘nanomembranes’ also provide the optical transparency, ultrathin dimensions, and high permeability useful for building multicompartmental models of tissue.^23,24 We have found these properties of silicon nanomembranes key to the development of in an in vitro platform for assessing S. aureus migration during bone infection we call the μSiM-CA (Microfluidic – Silicon Membrane – Canalicular Array; Figure 1). With facile integration into a transchamber device, and precise control over pore size, the μSiM-CA will define the bounds of S. aureus propagation and deformation, while allowing high quality live cell imaging and quantification of bacterial cell propagation through nanopores.

Figure 1. Modeling S. aureus invasion of OLCN in vitro with the μSiM-CA.

Cortical bone contains a complex matrix of lacunae connected by submicron canaliculi, which house osteocytes and their cell processes (A). TEM images of S. aureus invading and propagating in canaliculi of murine tibiae. Tibiae were infected for 14 days with wild-type UAMS-1 S. aureus using an established model of implant-associated osteomyelitis. Note the invasion site (arrow) at the canaliculi orifice on the endosteal surface facilitating evasion of neutrophils (noted by (N)), and nearby lacuna (noted by (L); B). Deformability is evident when comparing S. aureus at their normal diameter of ~1 μm to nearby bacteria at 25% of this size. (C). A schematic illustration of this is provided to demonstrate in vitro modeling with μSiM-CA (B). The nanoporous membrane is designed to provoke the pathogenic mechanisms associated with canalicular invasion by mimicking the sub-micron morphology and rigidity of canalicular openings in cortical bone.

To illustrate the pursuit of genes responsible for the OLCN invasive phenotype of S. aureus in chronic osteomyelitis, our lab identified the accessory gene regulator (agr) as a potential candidate. The agr system is the quorum sensing system of S. aureus, and is known to be the primary modulator of pathogenesis during infection in response to the autoinducing peptide AgrD.^25-27 As the concentration of AgrD reaches a threshold, the agr system activates most temporally expressed virulence genes including phenol soluble modulins (PSMs), alpha-toxin, beta-hemolysin, TSST-1, and leukotoxin, as it represses genes associated with colonization including protein A, coagulase, and fibronectin binding protein.²⁸ The agr system has been shown to be intimately associated with biofilm production as agr mutants are capable of forming robust biofilms but are incapable of activating biofilm dispersal shown both in vitro,²⁹ and in vivo. ^{20, 30} While the agr system is not known to play a role in cell shape determination, we hypothesized that the agr system may be required for the invasion of the OLCN because of its involvement in a variety of pathogenic mechanisms. Here we employ the μSiM-CA to quantify the propagation of S. aureus wild-type (WT) and agr mutants through submicron pores in vitro. Surprisingly, we find that the agr mutant strain propagates between compartments of the μSiM-CA at a similar rate to WT. We then confirm the ability of the agr mutants to invade canaliculi in a bacterial pin mouse model. While these findings invalidate our hypothesis that the agr system is required for invasion, they do demonstrate the predictive power of the μSiM-CA propagation assay and affirm its value for further screening of essential genes.

Given the utility of the μSiM-CA device in predicting the propagation phenotypes of S. aureus genetic variants, we replaced hand-made production with high-throughput manufacturing to improve reproducibility and efficiency of larger screening in future studies (Supplemental Figure S2).

Methods

Strains and Growth Conditions

S. aureus UAMS-1 strain and its derivatives were used for all experiments. S. aureus strains were grown on tryptic soy agar (TSA) or in tryptic soy broth (TSB) at 37°C. The S. aureus UAMS-1 agr-null strain (UAMS-1Δagr.:tetM) was constructed by phage 80α transduction of the Δagr.:tetM allele from S. aureus RN6911 into UAMS-1.³¹ Integration of the Δagr.:tetM allele and replacement of agr was verified by polymerase chain reaction (PCR) amplification and sequencing of chromosomal region flanking agr in UAMS-1. S. aureus strains UAMS-1 and UAMS-1 Δagr.:tetM were transformed with pCM29-sarA::gfp and pCM29-sarA::rfp reporter plasmids, respectively. Antibiotics ampicillin and chloramphenicol were added to the growth medium at concentrations 100 μg/ml and 10 μg/ml as needed. Scanning electron microscopy (SEM) methods are provided in Supplementary data.

Laboratory and High-throughput Fabrication of μSiM-CA

The μSiM-CA platform features a 400 nm thick SiN membrane with an array of 500 nm pores fabricated by SiMPore Inc. (West Henrietta, NY, USA). Additionally, SiN membranes were modified to create pore sizes of 600, and 700nm (methods provided in Supplementary data) and native 3 μm pore and non-porous membranes were purchased from SiMPore Inc. to act as positive and negative controls, respectively (Figure 3).

Figure 3. Increasing pore sizes allows for increased S. aureus propagation into the basal chamber.

SiN membranes were precisely modified by reactive ion etching to achieve pores of 500, 600 and 700 nm, visualized by TEM at low (4500X, scale bar = 0.5 μm) and high (26500X, scale bar = 100 nm) magnifications. 3 μm pore membranes were used as a positive control for S. aureus propagation (visualized at 2300X, scale bar = 1 μm and 6300X, scale bar = 0.5 μm). The membranes were assembled into μSiM-CA devices and loaded with WT UAMS-1 S. aureus for 6 hours, then S. aureus CFU’s were quantified from the basal chamber. Devices featuring non-porous membranes showed no detectable CFU’s following incubation (not detected (ND), n = 3). 500 and 600 nm membranes showed significantly less S. aureus propagation relative to the 700 nm and 3 μm groups. Significance was evaluated by ANOVA with Tukey post hoc for multiple comparison. # indicates p < 0.001 for 500 and 600 nm groups compared to the 3 μm control group. * indicates p < 0.05, for 500 and 600 nm groups compared to 700 nm. Data presented as mean with standard deviation (n = 3-5).

The nanoporous membrane was layered between silicone sealing gaskets (100 μm and 300 μm thick; Trelleborg, Trelleborg, Sweden) and irreversibly bonded together following UV ozone exposure and thermal incubation as previously described.^{32, 33} The bottom channel is constructed from 50 μm thick biocompatible pressure sensitive adhesive (3M 467MP) to minimize optical working distance for high resolution microscopy. The uppermost polydimethylsiloxane (PDMS) layer acts as the top well for bacterial culture with aspiration pores that penetrate through the bonding and chip layers for access to the bottom channel. The entire system is mounted on a No. 1 glass coverslip, again to reduce the optical working distance.

We used a contract manufacturer (ALine Inc., Rancho Dominguez, CA) to achieve high-throughput manufacturing and enable larger screening in future studies. Methods are provided in the supplement (Supplemental Figure S2) and resulted in ~300 devices per week compared to ~12 per week by hand.

Operation of in vitro propagation assays

The μSiM-CA device was loaded by adding 10 μl of TSB to the bottom chamber, below the membrane. Then 80 μl of either WT S. aureus UAMS-1/sarA::gfp, UAMS-1 Δagr::TetM/sarA::rfp or a 1:1 mix of UAMS-1/sarA::gfp and UAMS-1 Δagr::TetM/sarA::rfp, grown in TSB culture media was added to the top chamber of the μSiM-CA device, and statically incubated at 37°C for 6 hours (Figures 3 and 5) or 3, 6, 12 and 24 hrs (Figure 2).

Figure 5. The S. aureus agr system is not required for propagation through nanopores.

A representative fluorescent image of an agar plate used to quantify CFUs shows equivalent numbers of GFP WT and RFP agr mutant S. aureus aspirated from the bottom well of the μSiM-CA following 6 hours of co-culture (A, n = 4; serial dilutions from 10⁻¹ to 10⁻⁶). This observation was confirmed with qPCR by amplifying strain specific sequences (B, n = 4). Recovered CFU/mL and GU/mL of WT and agr mutant are reported as means ± standard deviation. Monocultures and co-culture assays were performed to assess differences in propagation, adherence and aggregation of WT vs. agr mutant strains. Representative SEM images of the underside of μSiM-CA membranes following media aspiration illustrate the marked differences in bacterial aggregation and adherence to the basal membrane surface between WT and agr mutant (C, E, G, 1000X scale bar = 10 μm; D, F, H, 5000X scale bar = 2 μm). Live cell CLSM confirms that the large aggregates of bacteria on the underside of the membrane in co-cultures are exclusively RFP agr mutant cells (I, J, scale bar = 2μm). Above the membrane, both cell populations (GFP WT and RFP agr mutant) can be seen occupying the surface of the membrane at 5 hours of incubation (I). Significance was evaluated by Mann-Whitney test.

Figure 2. The μSiM-CA device provides a more efficient platform for S. aureus propagation compared to commercially available track-etched PET membranes.

Commercially available devices with 0.4 μm porous PET membranes were compared to the μSiM-CA devices for S. aureus propagation assays. PET-based devices are assembled as a two-component insert/plate system (A), whereas the μSiM-CA transchamber system is assembled in a layer-by-layer fashion as a monolithic device, mounted upon a coverslip for high resolution microscopy (B). Representative SEM images of PET and SiN membranes reveal the contrasting membrane thicknesses of 10 μm and 0.4 μm, respectively (C, F), and the infrequent and randomly dispersed pores found in a commercial PET membrane (D) compared to the patterned pore matrix SiN (E). The utility of each device to serve as a tool to evaluate S. aureus nanopore propagation was tested by adding WT UAMS-1 to the devices and quantifying the bacterial load in the bottom well at various timepoints. Propagation was observed as early as 3 hours in the μSiM-CA device featuring the ultrathin SiN membrane (H), whereas propagation was first observed at 72 hours in the PET-based device (G). Representative data is shown for two independent replicate experiments.

Commercially available 24 well, 0.4 μm pore diameter, polyethylene terephthalate (PET) cell culture inserts (Corning, Inc., cat. no. 353095) were loaded with the manufacturer’s recommended volume and statically incubated at 37°C for 24, 48 and 72 hrs (Figure 2).

SEM and Confocal Laser Scanning Microscopy (CLSM) methods are provided in Supplementary data.

CFU and qPCR Quantification

Quantification of bacterial cells in the bottom channel of the μSiM-CA device was performed by aspirating media from the bottom well following incubation. Samples were plated on TSA, incubated overnight at 37°C before and resultant colonies were counted for colony forming unit (CFU) enumeration. Additionally, quantitative polymerase chain reaction (qPCR) was performed to quantify genome units of S. aureus. qPCR was performed on a Rotor-Gene Q (Qiagen, Hilden, Germany) using SYBR green (Quanta Biosciences, Gaithersburg, MD). The primer sequences used were the following: gyrB forward, 5’-CCAGGTAAATTAGCCGATTGC-3’; gyrB reverse, 5’-AAATCGCCTGCGTTCTAGAG-3’: tetM forward, 5’-TGGGCTTCCATTGGTTTATC-3’; tetM reverse, 5’-ATCCGTCACATTCCAACCAT-3’. Genomic units (GU’s) were calculated by running a series of standard samples with known S. aureus WT or agr mutant gDNA concentration alongside each reaction. The calculated tetM GU’s was subtracted from the gyrB GU’s to determine relative amounts of the mutant and WT strains.

Murine model for Osteomyelitis

All animal studies were performed in accordance with protocols approved by the University of Rochester’s Committee on Animal Resources and with the Animal Welfare Act. Animal surgeries were performed as previously described.^{35, 36} Briefly, female Balb/c mice (Jackson Research Labs, Bar Harbour, ME) were anesthetized prior to surgery with xylazine/ketamine and were administered preoperative buprenorphine. A flat stainless-steel wire (MicroDyne Technologies, Plainville, CT) was bent into an L-shaped implant and sterilized, then inoculated with UAMS-1 Δagr.:TetM S. aureus for 20 minutes (approximately 5.0 × 10⁶ CFU/mL). The right hind-limb was shaved and washed with 70% ethanol then a 5 mm incision was created on the medial surface of the tibia. Next, the tibia was drilled with 30- and 26- gauge needles before inserting the infected pin. Finally, the muscle and skin were closed, and mice were left to heal for 14 days post infection (n =5). TEM methods are provided in Supplementary data.

Results

To illustrate the structural advantages of silicon nanomembranes applied in a transmigration platform, we compared the propagation of WT S. aureus UAMS-1 through μSiM-CA devices to commercially available track-etched (TE) membrane inserts (Figure 2A, B). The membranes used in the commercial devices are composed of PET and were measured to be ~10 μm thick (Figure 2C). The pores in TE membranes exist at random locations and at low density (4 × 10⁶ pores/cm², 0.5% porous) (Figure 2D). By comparison the SiN membranes used for these studies were patterned at a density of 1.2 × 10⁸ pores/cm² to give a uniform porosity of 23% (Figure 2E) and are only 400 nm thick (Figure 2F). While the pores in SiN membranes do not model the depth of canaliculi, the high pore density and short propagation distance allow for measurable propagation through the μSiM-CA in only a few hours (Figure 2G). By comparison, the assay time for TE membranes was measured in days (Figure 2H). Measuring the speed of S. aureus emergence in the μSiM-CA platform informed future studies by setting an acceptable time period where we expect propagation of WT S. aureus and will be used as a baseline for phenotype genetic variants. Control experiments with E. coli showed that these rod-shaped bacteria could not propagate through nanopores, indicating that the observed propagation phenotype may be strain specific (Supplemental Figure S1).

As our ability to precisely tune the μSiM-CA membrane allows for elucidation of the physical parameters needed for S. aureus invasion, we modified SiN membranes to obtain a gradient of pore sizes and measured WT S. aureus propagation into the basal chamber. SiN membranes manufactured with 500 nm pores were etched to achieve 600 and 700 nm pore membranes (Figure 3A). It is important to note that the total number of pores does not change between 500, 600 and 700 nm pore membranes. SiN membranes manufactured with 3 μm pores were used as a positive control that does not require S. aureus deformation for passage into the basal chamber and non-porous SiN membranes used as negative controls and to test for device leakage. The μSiM-CA devices with each type of membrane were loaded with WT S. aureus for 6 hours, and S. aureus propagation into the basal chamber was measured by CFU enumeration. We found that the 500 and 600 nm membrane groups had significantly less propagation compared to the 700 nm devices and the 3 μm controls (p < 0.05). Direct comparison of 500 and 600 nm pores showed no statistical difference. These results suggest that membrane pores that are 600 nm in diameter, or lower, involve a similar mechanism of propagation. For further experiments we selected 500 nm pores, as this represents the average diameter of canaliculi in vivo and requires S. aureus cells to deform to approximately half of their native size.

Based on the role of the agr system in pathogenic infection, we hypothesized that an active agr system would be necessary for pathogenic invasion into the OLCN and, in turn, necessary for propagation through canalicular sized nanopores. An agr null mutant was created by phage transduction of the Δagr.:tetM allele from RN6911 into the prototypic laboratory strain UAMS-1, originally isolated from a patient with osteomyelitis.³⁰ The agr mutant was characterized by SEM to evaluate its morphology and adherent phenotype. After culturing WT and agr mutant strains on coverslips in vitro, the coverslips were washed with phosphate buffered saline (PBS) to remove any non-adhered cells and prepared for SEM. We found that the agr mutant strain exhibits increased adherence compared to WT, with increased cell abundance on washed coverslips and obvious cell-cell aggregation (Figure 4A, B). To rule out increased growth rate as a confounding factor, a growth curve was produced for the two strains from 0 to 5 hours, measured by optical density. No differences were observed between the two strains at all timepoints (Figure 4C), verifying that our SEM observations are not a product of growth rate.

Figure 4. The S. aureus agr mutant exhibits increased cell aggregation and adherence independent of growth rate in vitro.

Representative SEM images at 5000X (scale bar = 2 μm) show S. aureus cultured in vitro on a poly-L-lysine coated coverslip and washed with PBS. WT UAMS-1 shows relatively low abundance and appears to be arranged as a monolayer (A). Agr knockout strain exhibits an adherent phenotype and large aggregates (B). No differences in overall cell morphology are observed. Growth curves measured by optical density at 600 nm from 0 to 5 hours revealed no significant differences at all timepoints (C), thereby eliminating growth rate as a confounding factor. Significance was evaluated by two-way ANOVA Sidak post hoc for multiple comparison.

We then evaluated the propagation of S. aureus WT and agr mutant strains in the μSiM-CA device with 500 nm pores. Guided by our prior work with the WT strain (Figure 2, 3), we used a 6-hour time point as the basis for this comparison. Surprisingly, when equal amounts of WT and agr mutant S. aureus, expressing GFP and RFP, respectively, were added to the same μSiM-CA devices no significant difference in nanoporous membrane propagation was observed as determined by CFU’s or qPCR (Figure 5A, B). SEM imaging of the underside of the nanoporous membrane following bottom well aspiration demonstrated the differences between the two strains in adherence to the underside of the membrane (Figure 5C-H). In isolated cultures, few WT cells remained on the membrane following aspiration (Figure 5C, D), while large clusters of agr mutant cells remained adhered to the membrane (Figure 5E, F). This result is expected based on the loss of critical dispersal factors in the agr mutant and is consistent with the behavior on washed coverslips (Figure 4). In a co-culture of WT and agr mutant strains, we again observed bacterial cells adhered to the underside of the membrane as individuals and large aggregates (Figure 5G, H). These observations were confirmed by live cell CLSM at 5 hours of incubation. Remarkably, it was relatively easy to identify bright RFP expressing agr mutant clusters adhered to the underside of the nanoporous membrane (Figure 5I, J), but few WT cells were found in the same field-of-view, presumably because they readily disassociate from the membrane after propagation.

To test if our in vitro findings with the μSiM-CA predict the in vivo ability of the agr mutant to invade the OLCN in vivo, we used the strain in our standard model for implant-associated osteomyelitis. In this study, a trans-tibial pin infected with the S. aureus agr mutant strain was implanted for 14 days to establish a chronic-stage infection.²⁰ X-rays taken at Day 0 and Day 14 show osteolysis surrounding the implanted pin (Figure 6A), comparable to osteolysis observed in WT S. aureus infection (data not shown). Following the infection period, tibiae were harvested and sectioned to identify regions of bone colonized by Gram positive stained S. aureus (Figure 6B). Representative TEM images show regions of dense cortical bone, whose intrinsic canaliculi are fully occupied by cocci bacterial cells (Figure 6C). During S. aureus invasion of the OLCN, widened canaliculi are frequently observed as a result of bacterial erosion of the bone.^{15, 16} Narrow canaliculi, representing earlier stages of invasion, reveal S. aureus cells elongated and deformed to 50% of their typical size (Figure 6D), also matching the exact dimensions mimicked in the μSiM-CA model described here. These results suggest that the S. aureus agr system is not necessary for pathogenic invasion of the OLCN of cortical bone and, importantly, suggest the feasibility of the μSiM-CA to be applied as a predictive tool for OLCN invasion in vivo for continued genetic screening experiments.

Figure 6. In a murine model for implant associated chronic osteomyelitis, UAMS-1 Δagr successfully invades and colonizes the OLCN of cortical bone.

14 days following infected tibial pin implantation, tibias show osteolysis surrounding the implanted pin similar to that observed during infection by WT UAMS-1 (A). After tibias were harvested, bone sections were osmicated, and gram stained to identify regions infected with S. aureus (B). Following S. aureus identification, regions of interest were imaged by TEM to evaluate colonization of the OLCN. Representative micrographs show S. aureus invasion of the OLCN, where specific canaliculi are widened due to S. aureus mediated degradation (C, n = 5). Canaliculi representing earlier stages of invasion show sub-micron sized S. aureus cells deformed within the narrow canaliculi (D).

Discussion

S. aureus osteomyelitis is one of the most serious risks associated with orthopaedic implants and fracture fixation. Since the discovery of S. aureus invasion of the OLCN, there is a profound need for research exploring the genetics basis of S. aureus propagation through sub-micron spaces to understand the pathogenesis of chronic osteomyelitis and to inform future drug therapies. This work describes and validates the μSiM-CA platform to elucidate S. aureus genes responsible for invasion of the OLCN. While preliminary work in a precoursor system has shown that WT S. aureus cells can propagate through 0.5 μm sized pores,¹⁵ observed by SEM and light microscopy, the current work advances the technique as a platform and most importantly demonstrates its predictive value as a genetic screen of S. aureus OLCN invasion of live bone.

The μSiM-CA platform features silicon nanomembranes that show distinct structural advantages compared to standard PET-based membranes. The 400 nm thick, highly porous (23%) silicon membranes allow for S. aureus propagation measured as early as 3 hours, in contrast to the 72 hrs in standard transwell systems that utilize thicker low porosity (0.5%) membranes. Further, the silicon membranes provide a starting point for a variety of modifications including tunable pore sizes achieved by reactive ion etching of the starting material. While TE membranes are commercially available in varying pore sizes (eg. 0.4, 3, 8, and 12 μm) our platform uniquely allows for nano-scale manipulation of pore dimensions.

Quantification of WT S. aureus propagation through a gradient of pore sizes suggest that membrane pores that are 600 nm in diameter, or lower, involve a similar mechanism of propagation. The fact that significantly more propagation was observed through 700 nm compared to 600 nm or 500 nm pores may indicate that canalicular openings only slightly smaller than S. aureus are more readily invaded. While we found less propagation through 700 nm pores compared to 3 μm pores on average, we cannot conclude these values are statistically different. A study with higher statistical power may reveal the values to be different, however we cannot rule out the possibility of a subset of S. aureus that are physically small enough to simply pass through 700 nm pores as they can with 3 μm pores. This possibility is not unreasonable as S. aureus cell volume is known to change slightly over the cell cycle.³⁷ Nonetheless, our results do suggest the existence of a size-cutoff within 100 nm for the selection of S. aureus cells that must deform to pass through the membrane versus cells that can simply pass through, and that the need for deformation slows transmembrane propagation.

In this work the global regulatory system, agr, was identified as a potential regulator of S. aureus invasion of the OLCN in chronic osteomyelitis because of its intimate role in biofilm dispersal and regulation of virulence genes. It was described by Horswill, et al that an inactive agr system promotes the attachment and development of a robust biofilm, but an active agr system is required for detachment and dispersal from a biofilm.²⁹ In addition, previous studies have described the role of agr in S. aureus motility-like mechanisms such as agr-dependent expression of surfactant-like molecules,³⁹ phenol soluble modulins, for S. aureus passive motility across a surface known as colony spreading.^40-42 Thus, we originally hypothesized that an active agr system would be necessary for invasion into the OLCN, as it must continue its invasion of the host beyond its biofilm into deeper tissue.

The primary defect observed in the agr mutant was the inability to disaggregate, which is to be expected based on the upregulation of cell surface-associated proteins with the loss of agr function,⁴³ thereby increasing cohesion with neighboring cells and adhesion to substrates. When the agr knock-out mutant was evaluated in the μSiM-CA platform, we found that it is capable of propagating through the nanopores with no significant differences compared to WT S. aureus. These results were surprising given the observed agr mutant phenotype and known global down-regulation of virulence-associated genes. Taken together, these results show that S. aureus propagation through nanopores is not dependent on activation of the agr system and predict that the agr system is not required for invasion of the OLCN in vivo. Our in vivo model of implant associated chronic osteomyelitis confirmed that the S. aureus agr system is not necessary for invasion of the OLCN of cortical bone. Again, these findings were not originally expected given that agr mutants display decreased pathogenesis in in vivo models of osteomytelitis.³⁰ Futher, this result implies that all genes downstream of agr activation, including PSMs, toxins and degradative exoenzymes, are not required for chronic OLCN infection.

It has been described that S. aureus maintains an inactive agr system during early stages of host invasion, upregulating cell surface-associated proteins in order to establish secure attachment to a substrate and facilitate colonization. It is also during this phase of inactivated agr that S. aureus is capable for forming a robust biofilm to confer resistance to antimicrobial and immune cell attack. Following an increase in cell density, S. aureus employs AIP-mediated quorum sensing to activate the agr system where extracellular proteins, including various toxins, are produced. In light of our results both in vivo and in vitro, we speculate that the increased cell adherence resulting from agr inactivation may improve the bacteria’s ability to invade the OLCN of cortical bone. Additionally, Sordelli, et al recently reported that loss of agr function is associated with S. aureus adaptation and survival in chronic osteomyelitis.⁴⁴ However, to date, there is no reason to believe that the agr system would be involved in the regulation of any cell shape determining machinery, which would allow for invasion of the sub-micron sized OLCN. Thus, additional genetic screening studies using the μSiM-CA platform are warranted to determine the role of specific agr-regulated cell surface-associated proteins in OLCN invasion.

In summary, we formally demonstrate the utility of a novel μSiM-CA platform to distinguish the phenotype of mutant S. aureus strains based on their propagation through nanopores that mimic canaliculi. In so doing, the μSiM-CA allowed us to empirically establish that the agr system is not required for propagation through nanopores in vitro, as well as into the OLCN in vivo. The agr mutant strain displays the predicted dispersal defect, where it fails to disassociate from material surfaces as well as from neighboring cells following cell division. However, this does not impair its ability to invade host cortical bone in vivo, as demonstrated in a murine model for implant associated chronic osteomyelitis. These results suggest the feasibility of the μSiM-CA to be applied as a predictive tool for OLCN invasion in vivo. Future studies to screen various mutant strains using the manufactured μSiM-CA (Supplemental Figure S2) are warranted to identify the genetic mechanisms that mediate the haptotaxis and durotaxis in this process. Given the high socioeconomic burden and patient morbidity caused by S. aureus osteomyelitis, it is of utmost importance to elucidate the pathogenesis of persistent S. aureus bone infections. Ultimately by elucidating the genetic basis of S. aureus OLCN invasion, we could identify a new class of targets for antimicrobial development.