Co-culturing mammalian host cells with bacteria in vitro is an important tool for examining the host-pathogen interaction in infectious disease modelling [1]. One of the essential readouts is the intracellular bacterial burden carried by the infected host cells. The most commonly used method for enumerating such bacterial load is counting the colony-forming-unit (CFU) number by agar plate culturing following the lysis of host cells and release of bacteria. However, the reliability of such measurements is impacted by various factors, including, but not limited to, host cell/tissue type, bacterial strain, bacterial load, and culturing conditions for CFU enumeration [2, 3]. Osteomyelitis (OM), an infectious disease with pathogen-mediated infection in bone tissues, features a considerable proportion of infected but culture-negative cases, ranging between 20-40% [4-6]. This clear deficiency in CFU evaluation create difficulties for the accurate diagnosis and therefore treatment of OM.

Staphylococci, including coagulase-negative species, such as S. epidermidis, and S. aureus, comprise the major pathogens in adult OM [7]. In this study, we employed a previously established osteocyte-like cell model, differentiated SaOS2 (SaOS2-OY) and the single most common causative pathogen in OM, S. aureus, to simulate the in vitro infection of osteocytes, the most abundant cell type in bone tissue [8, 9]. Two previously characterised S. aureus strains, a high virulence strain, WCH-SK2 (SK2), shown previously to establish an intracellular infection in human osteocytes [10] and a low virulence strain, WCH-SK3 (SK3) [11], were chosen for experimentation. In addition to CFU enumeration to quantify bacterial number, a polymerase chain reaction (PCR) approach measuring genomic DNA copies represented by a single copy gene within the S. aureus genome, was also performed in parallel. In the past, real time quantitative PCR (qPCR) assays have been widely used for the quantification of bacteria. However, the working principle of standard qPCR is dependent on the establishment of a standard curve using a known number of bacterial gene copies and the fitting of test samples into a linear regression to estimate the bacterial number for each assay [12, 13]. Here, we utilised digital droplet PCR (ddPCR) for this measurement. By its nature, the latter technique offers the advantage of counting the absolute target sequence copy number within the sample and removes the need for normalisation [14]. For all PCR-based assays, the quality of DNA containing the target sequence for amplification and quantification is critical for the sensitivity and reproducibility of readout. Therefore, methods for DNA preparation from both human and bacterial cells were examined. Here, we introduced the usage of DirectPCR™ Lysis Reagent (Direct buffer), a lysis buffer that maximises the release of genomic DNA from samples and is compatible with the downstream PCR analysis from raw cell culture lysate without any purification step. With the application of Direct buffer for DNA preparations in comparison to a commonly used DNA extraction spin column kit (DNA kit), the absolute bacterial genome copy number quantified by ddPCR was measured to be 5-fold higher in SaOS2-OY samples and 100-fold higher in samples from two strains of S. aureus, SK2 and SK3; better overall reproducibility among biological replicates was also achieved using Direct buffer over the standard approach (Fig. 1A-C). Serial 10-fold dilutions of bacterial log-phase suspension cultures were used to generate DNA lysates for testing the extent of DNA release. Plotting genome copy number against CFU from the same cultures showed near perfect linear regression relationships (Fig. 1D-E), confirming the complete and highly reproducible release of bacterial DNA. With this validating readout, we sought to determine if the elimination of purification steps by using the Direct buffer might minimise the sample loss associated with the utilisation of conventional column-based extraction methods [15]. Further, the release of genomic material by this reagent appeared to be complete and unbiased in our experiments, which led to higher consistency of inter- and intra-assays.

Fig. 1:

Using the combined approach of direct lysis DNA preparation and ddPCR quantification, we then examined the S. aureus persistence in the intracellular environment in infected SaOS2-OY cells. Following infection, the SaOS2-OY and S. aureus co-cultures were lysed to break down host cell structures for the release of intracellular S. aureus for CFU development on agar plates. Consistent with previous findings, significant reductions ranging between 10³-fold and 10⁶-fold in cultured CFU on agar plates over the 5-day co-culture time period were observed in both SK2 and SK3 infected SaOS2-OY groups (Fig 2. A-F). This trend was also found in low multiplicities of infection (MOI) (^∼10⁷ total CFU/well, Fig 2. C & E) and high MOI (^∼10⁸ total CFU/well, Fig 2. D & F) for initial bacteria-inoculated groups with either strain. However, when total DNA preparations from the whole culture lysates were taken for ddPCR quantification of bacterial genome copy number, differences in the dynamic changes of bacterial load were found. In the high virulence strain SK2-infected group, the genome copy numbers by 5 days post-infection were maintained at the levels of 10⁷ and 10⁸ cells/well for the low and high MOI groups, respectively (Fig 2. C & D), similar to the original input levels.

Fig. 2:

However, the low virulence strain SK3-infected group demonstrated a decrease in genome copy counts. Interestingly, regardless of initial input amount, by 5 days post-infection, the remaining SK3 genome counts dropped to approximately 10⁶ cells/well for both groups (Fig 2. E & F). The quantification of human genome copy number was also performed in all groups to estimate the remaining host cells in culture and this also served as an indicator of host cell viability. With the low MOI co-cultures, the human genome copy counts in the presence of either strain of S. aureus were not statistically different from uninfected controls, across 1- and 5-days post-infection. When the initial bacterial inoculum was increased 10-fold to 10⁸ cells per well, reductions in the human genome copies of ∼30% and 50% were measured in 1- and 5-day post-infection groups, respectively; such a trend was found to be similar in both SK2- and SK3-infected groups. Overall, our observations demonstrated profound differences in the readout of the presence of intracellular bacteria when using CFU and ddPCR quantification methods. Such inconsistency between these readouts in our model was as high as 10⁶-fold, with interaction of factors including initial bacterial load and strain variation. In contrast, comparison of the same quantification methods performed on bacteria suspension cultures yielded differences under 2-fold, indicated by the respective slope factor (less than 2) in the linear regression curves for both SK2 and SK3 strains (see Fig 1. D & E). Together, our results indicated the dramatic change in bacterial culturability when comparing growth in ideal microbial suspension culture conditions and the growth limiting intracellular environment. Therefore, we propose that the interaction between the host cell response to infection and pathogen adaptation to the intracellular environment will compromise the reliability of the standard CFU counting method alone for the evaluation of bacterial persistence in a bacteria/host cell co-culture experiment. However, phenotypic variations were also observed in the recovered colonies from agar plating, with variations in haemolytic activity and small colony variant (SCV) formation, particularly in the SK2 group (Fig. 2A). SCV variants of S. aureus feature reduced metabolic activity and growth but are an indication of adaptation and the establishment of a chronic infection. This is consistent with the findings of reductions in CFU recovery but maintained levels of detectable genome copies, at least in the SK2 infected group.

To examine the applicability of this workflow to a clinical setting, we next extended the tests to clinical human bone specimens. Three clinically culture-negative prosthetic joint infection (PJI) cases were chosen for analysis. All three cases were confirmed infected by clinical observation for typical PJI symptoms by clinicians, using the current gold-standard Musculoskeletal Infection Society (MSIS) criteria [16]; neither the culturing of the synovial membrane biopsies nor synovial fluid around the prosthesis returned any positive cultures from a clinical laboratory. The bone tissue sections from the 3 patients were examined by Masson ‘s trichrome staining in comparison to tissue section from an osteoarthritis (OA) patient as non-infection control, to examine the bone matrix degradation, according to our previous study [17]. With this method, the red colour associates to intact bone matrix collagen and blue colour to degraded collagen. Within all the histological observations, matrix degradation was found in sections of all three PJI patients (Fig. 3 B-D), whereas the staining of the control OA patient bone indicated intact collagen (Fig. 3 A), consistent with infection in the 3 PJI patient specimens [17]. In addition to the gold-standard diagnostic approach and histological staining, we performed CFU analysis using homogenised bone tissue, and negative culture results were also recorded. In this study, the samples containing total genomic DNA were prepared from histological sections in conjunction with the use of the DirectPCR™ Lysis reagent. The prepared DNA samples were put through a PCR amplification using primer set targeting a highly conserved bacterial genomic region, elongation factor Tu (tuf) [18], for the enrichment of bacterial signal within host/pathogen genomic mixture. The generated amplicons were then sequenced using an Oxford Nanopore Technology (ONT) Minion sequencer for the identification of bacterial species. All three patient bone DNA samples were confirmed to be bacterial genomic positive, with combinations identified of the coagulase-negative staphylococcal species S. haemolyticus, S. hominis and S. epidermidis (Fig. 3 E-G). The respective bacterial load for each of the samples was then quantified by ddPCR. Each bone sample examined contained between 2 x 10⁴ and 1 x 10⁶ bacterial genomic copies per million human genomic copies (Fig. 3 E-G). To confirm that the detected bacteria species were not introduced during the handling procedures, bone specimens from five primary total hip replacement (non-infected) patients were processed at the same time along with the PJI specimens; PCR reactions with negative tuf amplification for these bone samples indicated that operational contamination was unlikely (Supplementary Fig. 1).

Fig. 3:

In this study, by the analysis of genomic material, we have addressed the discrepancy between the ineffectiveness of the classical bacterial CFU development approach in determining a confirmed infection by both an in vitro experimental model and by examining human clinical cases. With the introduction of a direct DNA release approach, less handling is required while delivering increased accuracy. The addition of ddPCR to this approach provided the advantage of achieving absolute quantification without the need for a PCR standard curve, which is both cost and time effective. For the purposes of unknown pathogen diagnosis in clinical cases, the exact bacterial species readout is required from sequencing the generated amplicons. Here, the utilisation of Oxford Nanopore technology dramatically lowered the equipment and skill demands for performing the analysis, since the MinION sequencer is a portable device and the sequencing results are analysed by the on-board software with little bioinformatics requirement. Together, this workflow is potentially applicable as a point-of-care diagnostic method plus the prospect of a rapid turnover: within hours comparing to the current procedure in the magnitude of days.

To conclude, for the purpose of quantifying the intracellular bacterial load in a bacteria/host co-culture setting, we have developed a workflow with the advantages of: a) rapid and labour-saving with the elimination all the DNA extraction steps by virtue of using the direct lysis approach; b) minimisation of potential sample loss encountered with conventional DNA isolation methods, enabling better accuracy and reproducibility; c) absolute number quantification by ddPCR for achieving superior consistency of intra-assay experiments; d) further sequencing analysis by ONT MinION sequencer has the potential to become a point-of-care diagnostic approach when performing unknown pathogen identification in a clinical setting. Nevertheless, we suggest that the CFU plating method should not be ignored for the evaluation of bacterial phenotypic adaptation in such experimentation.

Materials and methods

SaOS2 cell culture and differentiation

The human osteosarcoma cell line SaOS2 was employed for performing the bacteria/host co-culture experiments, as previously described [9]. Cells were seeded at a density of 2 x 10⁴ cells per well in 48-well multi-well plates and cultured at 37°C/5% CO₂, in growth media consisting of αMEM (Thermo-Fisher, VIC, Australia) supplemented with 10% v/v foetal calf serum (FCS) (Thermo-Fisher), 2mM L-Glutamine (Thermo-Fisher), 1U/ml penicillin/streptomycin (Thermo-Fisher). At 90% confluence, SaOS2 cultures were switched to differentiation media comprising αMEM supplemented with 5% v/v FCS, 100mM ascorbate 2-phosphate (Sigma-Aldrich, St Louis, USA), 1.8 mM potassium di-hydrogen phosphate (Sigma-Aldrich), 1mM HEPES (Thermo-Fisher), 2mM L-Glutamine and 1U/ml penicillin/streptomycin. Cells were then maintained under differentiation conditions for 28 days to achieve osteocyte-like phenotype (SaOS2-OY) [8].

S. aureus preparation

Two pre-characterised S. aureus strains, a methicillin-resistant S. aureus (MRSA), WCH-SK2 (SK2) [11] and a methicillin-sensitive S. aureus (MSSA), WCH-SK3 (SK3), were grown in terrific broth (Thermo-Fisher) on a 37°C /200 rpm rocking platform to achieve log-phase suspension cultures individually. Bacteria were pelleted by 3000g for 5 min centrifugation and then resuspended in sterile phosphate-buffered saline (PBS) to estimate cell number by the optical density (OD) at 600nm light absorption. The two bacteria suspensions were then taken to undergo 10-fold serial dilutions and the diluents were individually plated on blood agar plates containing 10% v/v defibrinated sheep blood (Thermo-Fisher) to determine the colony-forming-unit (CFU) number from original bacterial suspension preparations. The calculated CFU numbers from agar plates were compared to the ddPCR quantified genome copy numbers.

Host cell infection by co-culturing of SaOS2-OY and S. aureus

For host cell infection, bacteria suspension cultures were reconstructed in sterile PBS as described above to achieve the density for low and high multiplicities of infection (MOI). The differentiation media for SaOS2-OY was removed from cells and the cells were washed with PBS twice. The resulting bacterial inoculums were added to SaOS2-OY cells to allow invasion for 1 hour at 37°C/5% CO₂. Post infection, SaOS2-OY cells were washed twice with PBS and incubated at 37°C/5% CO₂with 20 μg/ml lysostaphin (Sigma-Aldrich) in antibiotic-free media for 24h to eliminate extracellular bacteria before supplied with normal differentiation media. Supernatants were cultured on an agar plate to verify that extracellular bacteria were absent at 24 hours and 120 hours post-infection.

Measurements of intracellular bacterial number by CFU counting

The intracellular bacteria numbers were estimated at 24-hour and 120-hour post-infection time points by spreading cell lysates on blood agar plates. For optimal CFU reading, cell lysates were serially diluted and individually plated on blood agar plates for 37°C /5% CO₂incubation. CFU numbers were recorded after 48 hours of colony development.

DNA extraction from host cells and bacteria

For the quantification of the genome, the total DNA was isolated using either DNeasy Blood & Tissue Kits (Qiagen Inc., VIC, Australia), or DirectPCR™ Lysis Reagent (Direct buffer) (Viagen Biotech Inc., CA, USA), as per the manufacturer ‘s instructions. For this new approach using the Direct buffer, 500 μl buffer together with 200 μg/ml proteinase K (Thermo-Fisher) was used for the lysis of one SaOS2-OY sample in one well from 48-well tissue culture plate or one pelleted bacterial sample prepared from the serial dilutions from suspension culture. Individual cell lysates were transferred to 1.5 ml tubes and digested at 55°C for 35 min and then heat inactivated at 85°C for 15 min to terminate enzyme digestion. The processed lysates were ready for PCR analyses.

Digital droplet polymerase chain reaction (ddPCR)

The ddPCR assay was performed using QX200 Digital Droplet PCR System (Bio-Rad Laboratories, USA) (Bio-Rad Laboratories, CA, USA) as per manufacturer ‘s instruction. Primer sets targeting a human genome specific sequence within type X collogen (COL10A1) gene and a S. aureus genome specific sequence within sigma factor B (sigB) gene, were used for the amplification of PCR products from DNA samples from SaOS-OY and S. aureus, respectively. The sequences of primer sets are listed as below.

COL10A1: forward 5 ‘-ccaccaggtcaagcagtcat-3 ‘, reverse 5 ‘-gttggcactaacaagaggggt-3 ‘

sigB: forward 5 ‘-ggggcaacaagatgaccatt-3 ‘, reverse 5 ‘-tgccgttctctgaagtcgtg-3 ‘

Analysis of human bone tissues

All human studies received institutional research ethics approval (Royal Adelaide Hospital Human Research Ethics Committee Approval No. 14466). Bone biopsies were collected from patients undergoing either primary total hip replacement or revision surgery for PJI, with informed written patient consent. Each bone specimen was separated into two parts. One of these was homogenised in a bead-beating tissue homogeniser (Bead Ruptor Elite, Omni International, Kennesaw, GA, USA), with 2 cycles of beating at 3m/s for 30s to disassociate the bone structure for the release of potential viable bacteria. Bone homogenates were then sent to a clinical pathology laboratory (SA Pathology, Adelaide, Australia) for CFU analysis. Clinical soft tissue and fluid samples were sent to SA Pathology directly from surgery for pathogen analysis. The second piece of bone biopsy was processed and de-mineralised using OSTEOSOFT® (Sigma-Aldrich) solution for standard paraffin embedding procedure. Post-embedded, the bone tissues were sectioned under DNase/RNase-free conditions. One 5μm section from each bone specimen was used for Masson ‘s Trichrome staining, as described in our previous study [17] and six sections were pooled together in DirectPCR™ Lysis Reagent for the isolation of total genomic DNA, as described above. Prepared DNA samples were then subjected to 1) conventional PCR amplification for an ONT sequencing readout (MinION sequencer, Oxford Nanopore Technology, Oxford, UK), according to the manufacturer ‘s instructions; 2) ddPCR for the absolute quantification for both human and bacterial genome copy number. The primer sets targeting the human COL10A1 genomic sequence, as listed above, and broad bacterial tuf sequence (forward 5 ‘-ttctcaatcactggtcgtgg-3 ‘, reverse 5 ‘-ggagtatgacgtccaccttc-3 ‘) were used for measuring human and bacterial genome copies, respectively. The output sequencing data were analysed by Epi2ME software within the WIMP workflow (Oxford Nanopore Technologies) for bacterial taxonomical analysis.

Statistics

The results were graphed as means ± standard errors of the means (SEMs). The significance between two treatment groups was evaluated using two-tailed T-test by GraphPad Prism 9.5.1 (GraphPad Software, MA, USA), where applicable.

Acknowledgements

This work was supported by a National Health and Medical Research Council of Australia (NHMRC) Ideas Grant scheme (ID 2011042) awarded to G.J.A. and D.Y. and an Australian Orthopaedic Association Research Grant awarded to G.J.A. and L.B.S.. Oxford Nanopore sequencing was supported by a University of Adelaide Faculty of Health and Medical Sciences infrastructure grant with technical assistance provided by Ms Thessa Kroes and Dr. Mark Corbett. Q.S. was supported by a University of Adelaide Faculty of Health and Medical Sciences Postgraduate Research Scholarship.

Author contributions

D.Y., G.J.A. and L.B.S. conceived the study. D.Y. adapted key protocols and developed the workflow; Q.S. performed most of the experimental work; K.H. contributed to molecular analysis and D.M. contributed to histological analysis; N.J.G and A.R.Z validated the methods developed in this work; Q.S. and D.Y. drafted the manuscript; G.J.A and D.Y provided scientific insight and supervised the study; L.B.S. consented patients and performed biopsy retrieval; all authors contributed to manuscript preparation and data presentation and agreed on the submitted version of the manuscript.

Competing interest

The authors declare no competing interests.