Uncovering lipid dynamics in Staphylococcus aureus osteomyelitis using multimodal imaging mass spectrometry

Abstract

Osteomyelitis occurs when Staphylococcus aureus invades the bone microenvironment, resulting in a bone marrow abscess with a spatially defined architecture of cells and biomolecules. Imaging mass spectrometry and microscopy are tools that can be employed to interrogate the lipidome of S. aureus-infected murine femurs and reveal metabolic and signaling consequences of infection. Here, nearly 250 lipids were spatially mapped to healthy and infection-associated morphological features throughout the femur, establishing composition profiles for tissue types. Ether lipids and arachidonoyl lipids were altered between cells and tissue structures in abscesses, suggesting their roles in abscess formation and inflammatory signaling. Sterols, triglycerides, bis(monoacylglycero)phosphates, and gangliosides possessed ring-like distributions throughout the abscess, suggesting a hypothesized dysregulation of lipid metabolism in a population of cells that cannot be discerned with traditional microscopy. These data provide insight into the signaling function and metabolism of cells in the fibrotic border of abscesses, likely characteristic of lipid-laden macrophages.

Keywords: Imging mass spectrometry; arachidonic acid; ether lipids; foam cells; lipidomics; macrophages; osteomyelitis; spatial biology; staphylococcus aureus.

Figure 1.. Microscopy and MALDI IMS are used to map and understand lipid distributions throughout a S. aureus-infected femur.

(A) Pre-IMS fluorescence microscopy captures autofluorescence signal from murine tissues and GFP fluorophore expression from S. aureus. (B, C) [PC(16:0_16:0) + H]⁺ and [PI(18:0_20:4) - H]⁻ are the two most abundant lipids detected in sequential MALDI IMS experiments. Data is acquired in positive and negative ion modes from the same section using a pitch offset strategy, yielding an effective spatial resolution of 20 μm. (D) Post-IMS histological staining with Masson’s Trichrome (MTC), or others, aids in tissue type identifications. (E) An adjacent section is fixed to prevent artificial bone marrow cracks and aids in cell type identifications throughout the intramedullary cavity. Annotations are provided where appropriate.

Figure 2.. Machine learning and histology are leveraged to interrogate the murine lipidome.

(A) k-Means clustering results and segmented images for separate (AI) positive and (AII) negative ion data reveal distinct pixel clusters that are associated with tissue types. A single section replicate from each infected and mock-infected femur are shown, and cluster labels on the right are supported by histology. (B, C) Positive ion clusters mark tissue types like the physis (yellow), new bone formation (purple), and periosteum/fibrous tissue (orange) in the (B) distal femur and (C) femoral diaphysis. Corresponding zooms of post-IMS histological stains are shown, one of which is a stain of a MALDI IMS section replicate (Rep). (D) Adjacent histology is used to help explain the molecular heterogeneity driving the bone marrow clusters in positive and negative ion modes. (E) Higher magnification of the H&E and MTC stains demonstrate the demarcation (dotted line) of degenerate or viable neutrophils to the left of the line, and fibrous tissue to the right of the line. The spatial resolution for all clustering datasets is 20 μm.

Figure 3.. Lipids specific to tissue types are discovered from the k-means clustering analysis.

(A) An overlaid image of eight positive ions displays discrete distributions to soft tissues surrounding and within the infected femur. The ions’ m/z (Top to Bottom) are 844.526, 786.601, 910.667, 704.524, 725.558, 758.548, 772.526, and 734.570. (B) An overlaid image of six negative ions highlight similar distributions. The ions’ m/z (Top to Bottom) are 909.550, 770.569, 850.572, 714.508, 913.580, and 847.570. The spatial resolution for the datasets is 20 μm, and the hotspot removal function in SCiLS for intensity scaling was employed due to the wide range of ion intensities. * in ion image labels denotes preliminary identifications.

Figure 4.. Imaging workflow enables targeted investigation of S. aureus lipids.

(A) Fluorescence microscopy highlights five established SACs (white arrows) that are expressing GFP in the autofluorescent bone marrow. Intensity scaling for channels is adjusted. B) Microscopy is registered with 20 μm MALDI IMS data and regions of interest are more accurately drawn around SACs (green) and surrounding bone marrow (red). An ion distribution that is associated with S. aureus overlays the microscopy with 35% transparency to demonstrate registration accuracy. (C) Overlaid average mass spectra from the green and red extracted pixels highlight the most abundant PGs in the SACs. Peak annotations are provided where appropriate, and representative ion images recapitulate their specificity to bacteria. (D) Spectra and ion images are also shown for another class of S. aureus lipids, lysyl-PGs. * in ion image labels denotes preliminary identifications.

Figure 5.. Clustering-assisted analysis of single abscesses reveals differences in lipid abundances between the outer and inner portions of bone marrow abscesses.

(A) Schematic of experimental design shows single abscesses chosen from each femur, which is represented by different shapes. An outlined shape corresponds to a section replicate (Rep) measurement of each abscess, and a solid shape corresponds to the average of section replicates for the abscess. (B) A pre-IMS fluorescence image and post-IMS H&E stain highlight visual changes between the two abscess regions contained within the drawn region of interest. (C, D) For (C) positive and (D) negative ion analysis each, four ion distributions (Top) contribute to the clustering algorithms, shown as individual and overlaid 20 μm ion images, and clustering results (Bottom) for each femur are reported. Scale bars represent 100 μm. (E, F) Bar graphs show the statistically significant differences in lipid abundances between the outside abscess pixels (red cluster) and inside abscess pixels (purple cluster) for each (E) positive and (F) negative ion analysis. The mean, with biological standard deviation, for three femurs is displayed as a bar with colors corresponding to the derived cluster. Section replicate measurements and replicate averages (one femur) are displayed on the graph. A paired T-test is used to test statistical differences, with *<0.05, **<0.01, ***<0.005 denoted below the graph on the x-axis labels.

Figure 6.. Principal component analysis (PCA) emphasizes the spatial heterogeneity of lipids detected throughout the intramedullary cavity.

(A, B) Pixels that are associated with bone marrow in all nine infected samples are used in the analysis. (AI) Positive ion lipids (+K) separate by Component 1 and Component 3 as shown in the loadings plot. (AII) Ion images from the annotated and colored loadings exemplify the variation in distributions explained by these components. (BI) Negative ion lipids (-H) separate by Component 1 and Component 4 as shown in the loadings plot, and (BII) select ion images are also displayed. Each Component 1 is shown to highlight the largest variation and Components 3 and 4 are selected to represent the ring-like distribution. The percentage of variance explained by each component is included in the axis labels. The spatial resolution for the datasets is 20 μm.

Figure 7.. Unique lipid classes possess similar ring-like distributions that are associated with cells at the abscess border.

(A) Pearson’s correlation tests highlight lipids (+K or -H) that correlate with the distribution of (AI) [PC O-(16:0_20:4) + K]⁺ and (AII) [PI O-(18:1_20:4) - H]⁻ within and surrounding a multi-SAC containing abscess (labeled in D & E). Black dots on the linear coefficient scale are supplemented to emphasize lipids with pronounced ring-like distributions as seen from their ion images. (B) Arachidonoyl ether glycerophospholipids are the most correlated lipids in both polarities. (C) Additional lipid classes like lyso lipids (L-), cholesterol esters (CEs), triglycerides (TGs), monosialo-gangliosides (GM1s) and bis(monoacylglycero)phosphates (BMPs) are also highly correlated. The spatial resolution for the datasets is 10 μm. (D) A pre-IMS fluorescence image shows autofluorescence signal (blue arrows mark hotspots), mostly in the GFP channel, derived from the abscess border morphology. (E) A post-IMS H&E stain with cracks (orange arrows) helps orient this autofluorescent signal mostly to the outer fibrotic region of the abscess. (F) A 50% transparent ion image of [PI O-(18:0_20:4) - H]⁻ is overlaid onto the fluorescence image to show their close association. (G) An overlaid ion image represents a more complete view of the lipid architecture of a bone marrow abscess. The ions’ m/z in reading order are 721.503, 766.539, 861.550, and 871.569. * in ion image labels denotes preliminary identifications. (H) H&E (Left) and F4/80 IHC (Right) staining is performed on a separate femur to confirm the presence of macrophages in the outer fibrotic region (white stars) surrounding the inner neutrophilic region (black stars).