Integrating optical coherence tomography and bioluminescence with predictive modeling for quantitative assessment of methicillin-resistant S. aureus biofilms

Abstract

Significance: Methicillin-resistant Staphylococcus aureus (MRSA) biofilm infections present a critical challenge in orthopedic trauma surgery and are notoriously resistant to systemic antibiotic therapy. Noninvasive, quantitative imaging methods are urgently needed to assess biofilm burden and therapeutic efficacy, especially for emerging photodynamic therapy (PDT) strategies.

Aim: We aim to establish a quantitative framework using a combined bioluminescence and optical coherence tomography (OCT) imaging approach to correlate bioluminescent signal with viable MRSA burden in both planktonic and biofilm states and to determine how biofilm density and structure influence this relationship.

Approach: Bioluminescent MRSA (SAP231-luxCDABE) was cultured in planktonic and biofilm forms using in vitro growth models in 24-well plates and custom macrofluidic devices, respectively. Bacteria bioluminescence intensity (BLI), counted colony-forming units (CFU), and OCT-based biofilm thickness measurements were collected to construct linear regression models to evaluate how well BLI alone, or combined with biofilm density (CFU/volume), predicts bacterial counts across culture conditions.

Results: Bioluminescence strongly correlated with CFU in planktonic cultures ( $R^2=0.98$ ). In biofilms, BLI per CFU decreased with density, indicating metabolic downregulation, and BLI alone was less reliable ( $R^2=0.59$ ). Incorporating biofilm density (CFU/volume) improved prediction ( $R^2=0.84$ ). A joint model for both states showed excellent fit ( $R^2=0.985$ ), but the biofilm versus planktonic group remained a significant factor ( $p=0.002$ ), revealing systematic differences. This highlights the need for a mixed-model approach that segments subvolumes by morphological features to improve accurate, generalizable CFU estimation across both growth states.

Conclusions: Bioluminescence alone underestimates bacterial burden in dense, metabolically suppressed MRSA biofilms. The combination of BLI with OCT-derived structural metrics enables accurate, nondestructive quantification of viable bacterial load. This approach provides a robust toolset for preclinical evaluation of antimicrobial therapies, particularly for optimizing PDT dosimetry and assessing biofilm response in translational infection models.

Keywords: biofilm; bioluminescence imaging; dual-modality imaging; methicillin-resistant Staphylococcus aureus; optical coherence tomography.

Fig. 1

Planktonic MRSA bioluminescence in $500\mu \mathrm{L}$ TSB in 24-well plates, measured 2 min after pipetting into the wells. CFU counts per well are indicated at the top of each corresponding column (e.g., 0 CFU, $1\times {10}^3\mathrm{CFU}$, and $2.5\times {10}^3\mathrm{CFU}$). Bioluminescence units are in $\text{photon}/\sec /{\mathrm{cm}}^2/\text{steradian}$ (logarithmic scale).

Fig. 2

(a) Schematic of a macrofluidic model of MRSA biofilm growth. Black arrows indicate the direction of flow: the top pump pushes tryptic soy broth (TSB) through the microtubes to the wells of the macrofluidic device (MD), whereas the bottom pump simultaneously pulls the waste. (b) White light and corresponding bioluminescent image of a biofilm grown on a metal orthopaedic washer. Red rectangles indicate the area where the OCT image shown in panel (c) was acquired; (c) OCT 3D-rendered image of a $1.5\times 1.5{\mathrm{mm}}^2$ region of a biofilm; (d) cross sectional OCT image from the location labeled with red in panel (c); biofilm thickness was measured from the biofilm surface down to the metal surface as indicated with a white double arrow; (e) SEM image of MRSA biofilm. M, MRSA; E, extracellular polymeric substance.

Fig. 3

(a) Bioluminescence standard curve (red) for serial dilutions of MRSA culture of known number of CFUs in TSB; (b) schematic representation of one of the wells with planktonic NRSA from which BLI was measured, with indicated volume. Well depth was 0.7 cm, surface area was $1.77{\mathrm{cm}}^2$; (c) histogram categorizing prediction quality, summarizing ratio of accurate (green) versus highly inaccurate (orange) predictions corresponding to the data residuals.

Fig. 4

Modelling CFU of MRSA biofilm using bioluminescence: (a) Biofilm BLI versus CFU experimental data, with prediction model equation at the bottom right, showing very wide intervals for each prediction (for details see text). (b) Schematic of a metal washer with a small average biofilm volume ($0.002{\mathrm{cm}}^3$) compared with planktonic suspensions. The surface area of the metal washer was $0.94{\mathrm{cm}}^2$; (c) model performance bar plot indicates the spread of obtained biofilm data reduces the reliability of this model—most predictions made by it differ from actual, obtained CFU values by a factor of 3 or more.

Fig. 5

White light, bioluminescence, and cross-sectional OCT images of (a) high and (b) low density biofilms of comparable volumes but different densities and subsequent effects on produced bioluminescence signal; (c) sample volume versus BLI produced; (d) biofilm density versus its BLI; (e) biofilm density versus BLI per CFU plot.

Fig. 6

Improved CFU prediction models, using BLI and volume: (a) Biofilm predictor model, improved from $R^2=0.59$ to 0.84 by adding sample density. Model (gray plane of fit) is described by equation below graph; (b) similar model adjustment for planktonic bacteria; (c) joint model predicting CFU of both biofilm and planktonic MRSA; (d) prediction accuracy for panel (a) showing the proportion of points falling within defined absolute distances from the model-predicted mean CFU; (e) corresponding evaluation of model performance in panel (b) for planktonic MRSA; (f) accuracy of the joint biofilm–planktonic model. To account for differing sample sizes ($N_{\text{biofilm}}=57$, $N_{\text{planktonic}}=856$), a subset of $N=57$ planktonic samples was randomly selected to match the biofilm group for prediction ratio assessment.