mRNA Transcriptome Analysis of Bone in a Mouse Model of Implant-Associated Staphylococcus aureus Osteomyelitis

To investigate the molecular pathogenesis of bone with osteomyelitis, we developed implant-associated osteomyelitis (IAOM) models in mice. An orthopedic stainless pin was surgically placed in the right femoral midshaft of mice, followed by an inoculation of Staphylococcus aureus into the medullary cavity.

ABSTRACT

To investigate the molecular pathogenesis of bone with osteomyelitis, we developed implant-associated osteomyelitis (IAOM) models in mice. An orthopedic stainless pin was surgically placed in the right femoral midshaft of mice, followed by an inoculation of Staphylococcus aureus into the medullary cavity. Typical characteristics of IAOM, like periosteal reaction and intraosseous abscess, occurred by day 14 postinfection. By day 28 postinfection, necrotic abscess, sequestrum formation, and deformity of the whole femur were observed. Transcriptional analysis identified 101 and 1,702 differentially expressed genes (DEGs) between groups by days 3 and 14 postinfection, respectively. Gene Ontology and Kyoto Encyclopedia of Genes and Genomes analyses revealed the enrichment of pathways in response to the bacterium, receptor-ligand activity, and chemokine signaling by day 3 postinfection. However, by day 14 postinfection, the enrichment switched to angiogenesis, positive regulation of cell motility and migration, skeletal system development, and cytokine-cytokine receptor interaction. Furthermore, protein-protein interaction network analysis identified 4 cytokines (interleukin 6 [IL-6], Cxcl10, gamma interferon [IFN-γ], and Cxcl9) associated with IAOM at an early stage of infection. Overall, as the pathological changes in this mouse model were consistent with those in human IAOM, our model may be used to investigate the mechanism and treatment of IAOM. Furthermore, the data for transcriptome sequencing and bioinformatic analysis will be an important resource for dissecting the molecular pathogenesis of bone with IAOM.

INTRODUCTION

Implant-associated osteomyelitis (IAOM) is a serious deep bone infection that complicates orthopedic surgeries, with an incidence of 1 to 5% after closed fractures and 10 to 15% after open ones (1). With increasing surgical activity and use of metal implants, the potential of infection that may develop to IAOM accordingly increases (2). Typical treatment of osteomyelitis involves surgical debridement followed by local and long-term systemic antibiotic administration, which is often associated with prolonged hospitalization and even infection recurrence, leading to a substantial economic burden on family and society (3).

The most common causative pathogen of IAOM is Staphylococcus aureus (4, 5). In spite of the availability of different antimicrobial compounds, treatment of S. aureus osteomyelitis remains a major clinical challenge, as the pathogen has a diverse set of mechanisms to persist in host bone to cause a relapsing infection, such as formation of biofilms on the implant surface or within the osteocyte lacuno-canicular network (6, 7), formation of small-colony variants (SVCs) (8), internalization and survival within host cells (9, 10), and secretion of virulence factors to affect antibody functions (11). As IAOM involves complex host responses to both pathogen and implant, a better understanding of the pathological mechanisms of IAOM is essential to develop alternative therapies for this devastating disease.

The mouse osteomyelitis models previously reported involve mainly hematogenous infection and local exogenous infection. Most of the hematogenous osteomyelitis is created by injection of bacteria into the tail vein of a mouse, leading to formation of systemic infection in the mouse besides osteomyelitis in the bone. Although this model can simulate the features of clinical hematogenous osteomyelitis, it is difficult to control the optimal amount of bacteria for injection through the tail vein. Insufficient injection may not cause bone tissue infection, while overinjection may lead to infections in other important organs, like the liver, kidneys, and heart (12, 13). However, the local exogenous osteomyelitis mouse models that are created by direct bacterial injection or incubation of a foreign body with bacteria before implantation cannot replicate the onset or progression of IAOM (14–16).

In order to better understand IAOM development and host response, we established an IAOM mouse model and verified their bacteriological and histopathological effectiveness. Additionally, transcriptome sequencing (RNA-Seq) and bioinformatic analysis were used to identify the possible key genes in the progression of IAOM.

RESULTS

Establishment of IAOM mouse model and radiographic evaluation.

To establish a mouse model of IAOM, a 0.5-mm unicortical bone defect was made on the dorsal side of right femoral midshaft (Fig. 1A and B) and a sterilized stainless pin was inserted into the medullary cavity through the canal (Fig. 1C). Next, 2 μl of S. aureus solution at 1 × 10⁶ CFU/ml was injected into the intramedullary cavity through the defect (Fig. 1D). Finally, the bone defect was sealed with bone wax (Fig. 1E) and the incision was closed with a 5-0 suture (Fig. 1F).

FIG 1.

Establishment of a mouse model of implant-associated osteomyelitis. (A) A 5-mm incision was made to expose the dorsal side of right femoral midshaft. (B) A unicortical defect was made using the needle of a 27-gauge syringe. (C) A 2-mm sterile stainless pin (0.3 mm in diameter) was implanted into the medullary cavity. (D) S. aureus (1 × 10⁶ CFU/ml, 2 μl) or 2 μl of PBS (control) was injected into the medullary cavity using a microsyringe. (E) The canal cortical bone was sealed with bone wax. (F) The incision was closed with 5-0 silk suture.

To identify the radiographic characteristics of IAOM mice, X-ray examination was performed. By day 3 postinfection, no difference in the bone morphology or intramedullary bone mineral density (BMD) was observed between IAOM and control mice (Fig. 2A). Infected femurs in IAOM mice displayed obvious periosteal reaction by day 14 postinfection and the bone morphology became abnormal, with dramatically increased bone mineral density around the infected implant, indicating osteonecrosis in the infected bone. By day 28 postinfection, bone destruction developed progressively in IAOM mice, severe periosteal reactions were observed not only near the infected implant but also in the distal femur, and a large area of sequestra and osteolysis were also observed in the medullary cavity (Fig. 2A). In contrast, no change was observed in the implanted bone in control mice (Fig. 2A). The radiographic scores for IAOM mice were significantly higher than those for control mice by days 14 and 28 postoperation, respectively (Fig. 2B). Collectively, the above data demonstrated that the IAOM mouse model showed typical radiologic pathological changes of acute, subacute, and chronic osteomyelitis along with the time of infection.

FIG 2.

Radiographic evidence of bone destruction in infected femur. (A) Representative radiographic images of femoral bones from control and implant-associated osteomyelitis (IAOM) mice. Blue arrows indicate the periosteal response, red arrows sequestrum formation, and yellow arrows osteolysis. (B) Quantification of bone destruction using a modified radiographic scoring method. ns, no statistical difference between the two groups at the same time point. **, P < 0.01 versus control (n = 6/group). (C) Representative three-dimensional (3D), coronal, and transverse micro-CT images of the femurs from control and IAOM mice by days 14 and 28 postinfection. Quantitative analysis of the bone fraction (BV/TV) (D), trabecular thickness (Tb. Th) (E), trabecular number (Tb. N) (F), trabecular separation (Tb. Sp) (G), cortical bone mineral density (Ct. BMD) (H), cortical area (Ct. Ar) (I), cortical thickness (Ct. Th) (J), cortical periosteal perimeter (Ct. Pe. Pm) (K), and cortical endosteal perimeter (Ct. En. Pm) (L). *, P < 0.05 versus control; **, P < 0.01 versus control (n = 4 to 6/group).

To assess the development of bone destruction during S. aureus osteomyelitis, femurs of control and infected mice were subjected to micro-computed tomography (micro-CT) analysis. There was dramatic loss of trabecular bone in the distal femur and significantly increased periosteal new bone formation at day 14 that became more severe at day 28 postinfection (Fig. 2C). Quantitative data for trabecular bone in distal femur showed that the ratio of bone volume to tissue volume (BV/TV) was considerably decreased in infected mice, accompanied by a dramatic decrease of trabecular number (Tb. N) and a significant increase in trabecular separation (Tb. Sp) (Fig. 2D to G). For the changes in cortical bone, at days 14 and 28 after infection, BMD decreased significantly (Fig. 2H), and the cortical area (Ct. Ar), cortical thickness (Ct. Th), and cortical periosteal perimeter (Ct. Pe. Pm) were remarkably increased in infected femur due to periosteal new bone formation (Fig. 2I to K), but no change in cortical endosteal perimeter (Ct. En Pm) was observed (Fig. 2L). The above data demonstrated considerable destruction in microstructure and periosteal new bone formation of the mouse femur during S. aureus osteomyelitis.

Location and quantification of S. aureus in the bone with IAOM.

To determine the development of chronic infection and S. aureus loads in implanted femurs, the infected femurs of IAOM mice were homogenized and incubated in a tryptic soy broth (TSB) agar plate. The results showed that S. aureus could be detected at each time point of infection in IAOM mice. Bacterial loads in bone increased dramatically 3 days after infection and became stable 14 days later (Fig. 3A), whereas there was no significant change in bacterial loads of implants between days 3, 14, and 28 postoperation (Fig. 3B).

FIG 3.

Bacterial colonization of femoral tissue and implant pins in IAOM mice. (A) Quantification of Staphylococcus aureus in bone tissue (n = 8/group) by days 3, 14, and 28 postinfection. *, P < 0.05. (B) Quantification of S. aureus on implant (n = 5/group). (C) Immunohistochemistry staining of S. aureus in the femurs of control and IAOM mice (n = 6/group). Red stars indicate the position of the implant in the medullary cavity and red arrows the S. aureus-positive staining. Scale bar = 100 μm.

To further confirm the location of S. aureus in infected femurs, immunohistochemical staining was performed. By day 3 postinfection, S. aureus was observed mainly in bone marrow where the implant was located, and no obvious S. aureus was found in bone marrow of the distal femur (Fig. 3C). However, by days 14 and 28 postinfection, S. aureus was observed in the medullary cavity away from the implant (Fig. 3C). The above results suggest that S. aureus may not only attach to the implant but also spread in the medullary cavity of infected bone with progression of IAOM.

Histopathological characteristics of the bone with IAOM.

To observe the histopathological changes in infected bone, hematoxylin and eosin (H&E) staining was performed in the control and infected femurs on days 3, 14, and 28 postoperation. No significant change was observed in the bone morphology or histology in control mice. In contrast, progressive bone destruction was observed in the bone with IAOM. At a very early time point after infection (day 3 postinfection), the femur showed massive infiltration of neutrophils and some fibroblast-like cells in the marrow around the infected implant (Fig. 4A). By day 14 postinfection, lots of empty lacunae were observed in the cortical bone, indicating loss of osteocytes and progression of osteonecrosis. Some abscess in the bone marrow and a considerable amount of periosteal new bone formation were observed around the cortical bone close to the implant (Fig. 4A). When IAOM progressed into a chronic stage by day 28 after infection, the infected femur showed characteristics of bone destruction, such as extensive infiltration of neutrophils in the medullary cavity, formation of necrotic abscess in the medullary cavity, deformity of the whole femur, and sequestrum formation (Fig. 4A). Smeltzer’s scores confirmed progressive destruction of the bone with IAOM (Fig. 4B).

FIG 4.

Histopathological analysis of femurs in control and IAOM mice. (A) Representative images of hematoxylin and eosin (H&E) staining of femurs in control and IAOM mice by days 3, 14, and 28 postinfection. Red stars indicate the position of the implant in the medullary cavity, the red dotted box indicates infiltration of neutrophils, green arrows indicate neutrophils, the red solid box indicates fibroblast-like cells, yellow arrows indicate the empty lacunar of osteocytes, black arrows indicate the periosteal reaction, the yellow solid box indicates abscess, and red arrows indicate sequestra. (B) Quantitative analysis of histopathological changes using Smeltzer’s scoring method. ***, P < 0.001 versus control (n = 6/group).

To verify the infiltration of neutrophils and the formation of fibroblasts during infection, we performed immunohistochemistry to evaluate the expression of myeloperoxidase (MPO), a marker of neutrophils, and immunofluorescence for α-smooth muscle actin (α-SMA), a marker of fibroblasts. The results showed MPO-positive cells in bone marrow by day 3 after infection and an increase with the extension of infection (Fig. 5A). In addition, the expression of α-SMA around abscess was dramatically increased by day 14 and became even greater by day 28 after infection (Fig. 5B).

FIG 5.

Infiltration of neutrophils and formation of fibroblasts. (A) Representative images of immunohistochemical staining for myeloperoxidase (MPO) in the femurs of control and IAOM mice (n = 6/group). Scale bars = 1 mm for images at lower magnification and 50 μm for images at higher magnification. (B) Immunofluorescence of α-smooth muscle actin (α-SMA) (green) in a femoral section (n = 6/group). DAPI stains nuclei blue. Scale bars = 100 μm for images at lower magnification and 25 μm for images at higher magnification.

To confirm bone destruction in the femur with IAOM, Goldner’s trichrome staining and tartrate-resistant acid phosphatase (TRAP) staining were performed. Consistent with the H&E staining results, periosteal new bone formation, formation of necrotic abscess, deformity of the whole femur, and sequestrum formation were observed in the IAOM femur with progression of infection (Fig. 6A). As revealed by TRAP staining, the number of osteoclasts showed little change in the control mice at each time point after operation. However, along with development of infection, the number of TRAP⁺ cells in trabecular bones increased gradually in the mouse femur with IAOM (Fig. 6B and C). Collectively, the above data demonstrate that periosteal reactive bone formation and abundance of osteoclasts may be enhanced by prolonged infection.

FIG 6.

Changes in bone structure and osteoclastic abundance. (A) Representative images of Goldner’s trichrome staining of femurs in control and IAOM mice by days 3, 14, and 28 postinfection. High magnification of the area marked with the red dotted box shows peripheral new bone formation in femurs of IAOM mice. Blue arrows indicate filtration of neutrophils, red arrows sequestrum, and yellow arrows abscess. Scale bars = 500 μm for images at lower magnification and 50 μm for images at higher magnification. (B) Representative images of tartrate-resistant acid phosphatase (TRAP) staining of femurs in control and IAOM mice by days 3, 14, and 28 postinfection. Scale bars = 50 μm. (C) Quantification of TRAP⁺ cells per trabecular bone surface (n = 6/group). In the area from 0.2 to 0.7 mm above the distal growth plate, 2 microscopic fields (magnification, ×200) were randomly selected for TRAP⁺ cell count. **, P < 0.01 versus control; ***, P < 0.001 versus control.

(i) Screening of DEGs.

In order to characterize the molecular responses to IAOM, we performed transcriptome analysis of the IAOM and control femurs by days 3 and 14 postoperation using high-throughput sequencing. Results showed a total of 101 differentially expressed genes (DEGs), among which 96 were found to be upregulated and 5 downregulated in the femur when the IAOM and control groups were compared by day 3 postinfection (Fig. 7A). By day 14 after the operation, the number of DEGs significantly increased, i.e., up to 1,702, with 1,065 upregulated and 637 downregulated in the IAOM femurs compared with controls (Fig. 7B).

FIG 7.

Screening of DEGs and their KEGG analysis by days 3 and 14 postinfection. (A) Volcano map of DEGs by day 3 postinfection. (B) Volcano map of DEGs by day 14 postinfection. (C) KEGG analysis by day 3 postinfection. (D) KEGG analysis by day 14 postinfection.

(ii) GO analysis of DEGs.

Gene Ontology (GO) analysis was performed to predict biological functions of the above-mentioned DEGs. By day 3 after infection, biological process (BP) was mainly enriched in response to bacteria, response to molecule of bacterial origin, cytokine-mediated signaling pathway, and defense response to other organisms. For cellular component (CC), DEGs were enriched in the external side of the plasma membrane, extracellular matrix, and proteinaceous extracellular matrix. For molecular function (MF), they were enriched in receptor-ligand activity and receptor regulator activity (Table 1). For biological functions of DEGs on day 14 postinfection, BP was different from that on day 3 postinfection: it was enriched in angiogenesis, positive regulation of cell motility and migration, morphogenesis of an epithelium, and skeletal system development, whereas the enrichments of CC and MF were almost the same as for DEGs by day 3 after infection; the only distinction was that the number of counts was increased at day 14 (Table 1).

TABLE 1.

GO enrichment analysis of DEGs on the day 3 and 14 postoperation (top 5 for each term)

Time postoperation	Term	GO IDa and description	Count	%	P (adjusted)
3 days	BP	GO:0009617, response to bacterium	28	29.79	4.53E−19
	BP	GO:0002237, response to molecule of bacterial origin	22	23.40	1.24E−16
	BP	GO:0019221, cytokine-mediated signaling pathway	22	23.40	1.75E−15
	BP	GO:0098542, defense response to other organism	22	23.40	9.12E−14
	BP	GO:0032496, response to lipopolysaccharide	21	22.34	2.85E−16
CC	GO:0009897, external side of plasma membrane	15	15.79	1.74E−08
CC	GO:0031012, extracellular matrix	8	8.42	1.65E−02
CC	GO:0005578, proteinaceous extracellular matrix	7	7.37	1.74E−02
CC	GO:0043230, extracellular organelle	5	5.26	1.34E−03
CC	GO:0020003, symbiont-containing vacuole	4	4.21	2.00E−05
MF	GO:0048018, receptor-ligand activity	22	0.13	2.36E−15
MF	GO:0030545, receptor regulator activity	22	0.13	5.55E−15
MF	GO:0005125, cytokine activity	20	0.12	4.20E−19
MF	GO:0005126, cytokine receptor binding	18	0.11	4.74E−14
MF	GO:0001664, G-protein-coupled receptor binding	12	0.07	2.74E−07
14 days	BP	GO:0001525, angiogenesis	104	7.19	1.23E−16
BP	GO:2000147, positive regulation of cell motility	101	6.98	9.07E−14
BP	GO:0030335, positive regulation of cell migration	99	6.85	9.40E−14
BP	GO:0002009, morphogenesis of an epithelium	93	6.43	1.70E−10
BP	GO:0001501, skeletal system development	90	6.22	3.77E−11
CC	GO:0031012, extracellular matrix	129	8.88	1.35E−37
CC	GO:0005578, proteinaceous extracellular matrix	109	7.50	7.96E−34
CC	GO:0009897, external side of plasma membrane	74	5.09	4.03E−14
CC	GO:0045177, apical part of cell	53	3.65	0.004719483
CC	GO:0044420, extracellular matrix component	48	3.30	7.66E−18
MF	GO:0030545, receptor regulator activity	77	5.43	5.12E−10
MF	GO:0048018, receptor ligand activity	74	5.22	5.12E−10
MF	GO:0000978, RNA polymerase II proximal promoter sequence-specific DNA binding	64	4.52	0.000102769
MF	GO:0000987, proximal promoter sequence-specific DNA binding	64	4.52	0.000192836
MF	GO:0000982, transcription factor activity, RNA polymerase II proximal promoter sequence-specific DNA binding	62	4.38	0.000102197

^aID, identifier.

(iii) KEGG analysis of DEGs.

The KEGG results showed that DEGs by day 3 postinfection were mainly enriched in cytokine-cytokine receptor interaction, tumor necrosis factor (TNF) signaling pathway, and chemokine signaling pathway (Fig. 7C), whereas DEGs were enriched in pathways in cancer, cytokine-cytokine receptor interaction, and phosphatidylinositol 3-kinase (PI3K)-Akt signaling pathway by day 14 postinfection (Fig. 7D).

(iv) PPI analysis of DEGs and verification at the transcriptional level.

The STRING database was used to analyze the hub genes in protein-protein interaction (PPI) of DEGs. The PPI network of DEGs by day 3 postinfection had a total of 44 nodes and 206 edges. Further analysis by MCODE plug-in showed that the interleukin 6 (IL-6), chemokine C-X-C motif ligand 10 (Cxcl10), gamma interferon (IFN-γ), and chemokine C-X-C motif ligand 9 (Cxcl9) genes were the most important genes, because they had higher scores in degree and closeness centrality among clusters (Table 2 and Fig. 8A). The DEG PPI network for samples collected by day 14 postinfection had 530 nodes and 2,395 edges, in which IL-6, tumor necrosis factor (TNF), fibronectin 1 (Fn1), and vascular endothelial growth factor A (VEGFA) had higher scores in degree and were mostly enriched in cluster 3 (Table 3). In contrast, in cluster 1, Cxcl10, chemokine C-X-C motif ligand 2 (Cxcl2), chemokine C-C motif ligand 4 (Ccl4), and Cxcl9 were the top proteins (Table 3 and Fig. 8B). These results suggest that IL-6, Cxcl10, IFN-γ, and Cxcl9 may play an important role in an early stage of IAOM.

TABLE 2.

PPI analysis of DEGs on day 3 postoperation (degree top 30)

Gene	Degree	Avg shortest path length	Betweenness centrality	Closeness centrality	MCODE cluster
IL-6	47	1.347826	0.233687	0.741935	1
Cxcl10	41	1.42029	0.116286	0.704082	1
IFN-γ	37	1.478261	0.101706	0.676471	1
Cxcl9	36	1.507246	0.108304	0.663462	1
Cd274	30	1.608696	0.044662	0.621622	1
Ccl2	30	1.623188	0.015756	0.616071	1
IL-1β	30	1.623188	0.025238	0.616071	1
Cd40	28	1.666667	0.017702	0.6	1
Cxcl1	28	1.652174	0.01159	0.605263	1
Arg1	27	1.681159	0.050902	0.594828	1
Ccl4	25	1.666667	0.01287	0.6	3
Ccr5	24	1.724638	0.01382	0.579832	1
Timp1	23	1.753623	0.042055	0.570248	3
Cxcr3	22	1.768116	0.005317	0.565574	1
Serpinb9	21	1.768116	0.003909	0.565574	1
Ccl22	20	1.782609	0.002549	0.560976	1
Ctla4	20	1.797101	0.003808	0.556452	1
Cxcl16	20	1.782609	0.032635	0.560976	1
Cd14	18	1.84058	0.012505	0.543307	1
Gzmb	18	1.84058	0.010053	0.543307	1
Socs3	17	1.855072	0.003087	0.539063	3
Ccl12	16	1.884058	0.005320	0.530769	1
Irg1	16	1.855072	0.008781	0.539063	3
Gbp2	16	2.028986	0.050080	0.492857
Cxcl3	16	1.840580	0.001181	0.543307	1
Mmp3	15	1.869565	0.012342	0.534884
Nos2	15	1.855072	0.010886	0.539063	1
Gbp5	12	1.927536	0.024663	0.518797	2
Saa3	12	1.971014	0.005245	0.507353
Pdcd1lg2	12	1.956522	0.000630	0.511111	4

FIG 8.

PPI analysis of DEGs in cluster 1 by days 3 and 14 postinfection. (A) PPI analysis in cluster 1 by day 3 postinfection. (B) PPI analysis in cluster 1 by day 14 postinfection. The node size from small to large indicates the closeness centrality of proteins from minimum to maximum; the node color from blue to red indicates the degree of proteins from minimum to maximum. (C) mRNA expression of IL-6, Cxcl10, IFN-γ, and Cxcl9 in femurs from control and IAOM mice (n = 7 or 8/group). *, P < 0.05; **, P < 0.01; ***, P < 0.001 (versus control).

TABLE 3.

PPI analysis of DEGs on the 14 postoperation (degree top 30)

Gene	Degree	Avg shortest path length	Betweenness centrality	Closeness centrality	MCODE cluster
IL-6	247	2.221481	0.092156	0.45015	3
TNF	227	2.276296	0.068679	0.43931	3
Fn1	187	2.279259	0.07761	0.438739	8
VEGFA	155	2.344444	0.050202	0.42654	3
EGFR	147	2.32963	0.052896	0.429253	5
Ccl2	137	2.439259	0.013345	0.409961	3
IL-1β	126	2.460741	0.013466	0.406382	2
Jun	118	2.438519	0.029243	0.410085	8
IFN-γ	116	2.541481	0.010698	0.393471	3
Tlr2	114	2.542963	0.011955	0.393242	3
IL-17a	111	2.583704	0.009444	0.387041	3
Icam1	110	2.533333	0.011882	0.394737	2
Col1a1	105	2.536296	0.019608	0.394276	3
Timp1	103	2.525926	0.00827	0.395894
Cxcl10	103	2.655556	0.004837	0.376569	1
Ptgs2	97	2.501481	0.016158	0.399763	3
Cxcl2	92	2.68963	0.004749	0.371798	1
Cd19	88	2.726667	0.013125	0.366748	2
Col3a1	85	2.658519	0.008419	0.376149	3
Rps27a	83	2.619259	0.036351	0.381787	2
Cd80	83	2.734815	0.005129	0.365655	3
Ctgf	82	2.566667	0.011921	0.389610	5
Bgn	82	2.628148	0.014726	0.380496	5
Spp1	81	2.580741	0.005282	0.387486	8
Ccl4	80	2.725926	0.002030	0.366848	1
Cxcl9	80	2.802963	0.003688	0.356765	1
Mmp2	79	2.594074	0.009009	0.385494	7
Lox	78	2.557037	0.014874	0.391078	5
Socs3	77	2.677778	0.013166	0.373444	2
Cxcr5	77	2.873333	0.003347	0.348028	1

Next, we assessed the expression of the identified genes, including IL-6, Cxcl10, IFN-γ, and Cxcl9, in infected femurs of IAOM mice. The results showed that the mRNA expression of IL-6, Cxcl10, IFN-γ, and Cxcl9 was significantly upregulated on days 3 and 14 postinfection (Fig. 8C), which was consistent with the results of PPI network analysis.

DISCUSSION

In this study, we established a mouse model that mimics the pathological features of IAOM in clinic. Our data showed characteristic changes of infected bone from an acute stage to a chronic stage of infection, including infiltration of neutrophils, periosteal reactive new bone formation, formation of necrotic abscess, sequestrum formation, and deformity of the whole femur. Using this IAOM mouse model, whole-transcriptome sequencing and bioinformatic analysis were performed to study molecular responses of infected bone at an early stage of S. aureus IAOM. The genes for 4 cytokines (IL-6, Cxcl10, IFN-γ, and Cxcl9) were screened out as the most important genes associated with early-stage IAOM. This study may provide a clue to better understanding of the molecular mechanism of how S. aureus may trigger pathological changes in bone, thereby leading to new therapies aimed at prevention or treatment of IAOM at an early stage of infection.

Various methods have been used to establish IAOM mouse models, such as implantation of a steel pin coated with S. aureus through a transcortical hole at tibial metaphysis (15), fixation of the osteomized femur with a locking plate and inoculation of S. aureus into the fracture gap (17), implantation of intramedullary Kirschner wire into the femur through the knee and inoculation of S. aureus into joint space (18), and implantation of stainless Kirschner wire inoculated with bacteria through a hole drilled into the distal femur (19). To evaluate the pathogenesis of intramedullary osteomyelitis, we modified these mouse models by inserting the stainless implant into the bone marrow cavity, followed by inoculating S. aureus completely into the middle of the femur medullary cavity and sealing the canal with bone wax finally. This modified method may prevent the loss of the implant during the course of infection and avoid potential infection in joint or surrounding soft tissue. Additionally, since the time of bacterial incubation and drying of bacteria have a great impact on formation of biofilm (20), in our modified model, injection of S. aureus through the canal may ensure equal amounts of bacteria in each inoculation, which is a prerequisite for a stable and consistent infection model.

On day 3 after implant infection, no significant change was observed by X-ray examination. However, dense infiltration of neutrophils was observed by histological analysis, indicating characteristic histological changes of acute osteomyelitis (21). As the infection progressed over the course of 14 days, bacterial load increased dramatically with periosteal bone formation and intraosseous abscess. These features are typical of subacute osteomyelitis (22). Since inadequate treatment of acute or subacute infections may result in chronic osteomyelitis, the typical subacute osteomyelitis developed in our animal model. By day 28 postinfection, we found sequestra surrounded by intense fibrosis and inflammation in the infected femur. Additionally, we also observed abscess in bone marrow, dense marrow fibrosis, thick periosteal reactive bone formation from the surviving periosteum, and even deformity of the whole femur. These pathological changes are typical features observed in human chronic osteomyelitis (23, 24).

In the present study, bioinformatic analysis based on transcriptome data revealed information about global changes in genes in the bone modulated by S. aureus IAOM, which may help unravel regulatory pathways that control the pathogenesis of bone and progression of IAOM. It was reported that the transcriptional adaptation of S. aureus during acute osteomyelitis is mainly associated with genes of metabolic adaptation, immune evasion, and replication (25). Consistently, our work demonstrated transcriptional adaptation of infected bone that was mainly involved in the response to the bacteria and ligand-receptor activity and in the pathway in cytokine-cytokine receptor interaction at the early stage of osteomyelitis. Furthermore, IL-6, Cxcl10, IFN-γ, and Cxcl9 are screened out as the most important factors by PPI network analysis. These molecular responses are also consistent with the histological observation that neutrophils accumulated in bone marrow around the infected implant. Since the inflammatory factors play a protective role by recruiting immune cells (26–28), our data suggest that bone marrow may be engaged mainly in recognition of bacteria, activation of immune cells, and release of inflammatory factors at an early stage of S. aureus IAOM.

S. aureus is able to cope with the biological pressure within host microenvironments, and it can switch its transcriptional responses to a survival mode during chronic osteomyelitis (25). Our data demonstrated alterations of gene expression in host bone with the transition from acute to subacute or chronic infection. GO analysis of DEGs in bone revealed enrichment of the extracellular matrix, angiogenesis, and positive regulation of cell migration and motility by day 14 postinfection. In addition, the transcriptional response of bone cells during the subacute and chronic stages of IAOM paralleled the histological changes of increased periosteal bone formation and dense infiltration of neutrophils in bone marrow observed in our present mouse model and other mouse models (12, 15). Consistent with functional analysis, our KEGG analysis showed enrichment of cancer pathways and the PI3K-Akt signaling pathway, which has proved to be associated with cell proliferation, metabolism, and migration (29, 30). Thus, these data highlight increased metabolism and remodeling in infected bone, in which the PI3K-Akt signaling pathway may play a critical role in the immune response and bone metabolism (31).

In response to infection, proinflammatory cytokines, such as TNF-α, IL-1β, and IL-6, have been considered to be the major acute-phase regulators, promoting generation of macrophages and neutrophils (32). Immune cells, such as natural killer cells, T cells, and macrophages, may produce large amounts of IFN-γ, thereby stimulating production of Cxcl9 and Cxcl10, which are potent chemoattractants for recruiting more lymphocytes (33, 34). All these proinflammatory cytokines and chemokines are required for efficient antibacterial immune response (35, 36). However, the inflammatory response in infection is a double-edge sword due to its opposite effects on the host. Excessive IL-6, Cxcl10, IFN-γ, and Cxcl9 have a pathological effect on bone formation and resorption (37–39). Our present study showed persistent elevation of IL-6, Cxcl10, IFN-γ, and Cxcl9 in the bones of S. aureus IAOM mice by days 3 and 14 postinfection, indicating activation of the immune system and potential bone destruction induced by these inflammatory factors. Further elucidation of the process involved in the expression of these cytokines and chemokines may help clarify the pathogenesis and develop antibiotic adjuvants for IAOM.

This study has 2 limitations. First, it would be difficult to take the implant out if the bone required further processing for paraffin embedding because the implant had been inserted completely into the femoral marrow cavity in our IAOM model. It should be appropriate to carefully slice the sample paraffin embedded until the implant is exposed, melt the remainder of the paraffin block in liquid paraffin at 60°C, carefully take the sample out, and re-embed it in paraffin. Second, although the PPI network analysis of the transcriptional data has screened out several potential key genes in bone modulated by S. aureus IAOM, it is known that changes in mRNA expression do not always equate to alterations at the protein level, and further experiments are needed to evaluate the protein expression of these genes.

In conclusion, we have established a feasible model of implant-associated osteomyelitis in mice and assessed its effectiveness by radiographic, bacteriological, and histopathological methods. Using this model, our transcriptome analysis suggests that the IL-6, Cxcl10, IFN-γ, and Cxcl9 genes might be the key genes in an early stage of S. aureus IAOM and thus the potential target genes for early diagnosis or treatment of the disease.

MATERIALS AND METHODS

Preparation of bacteria.

S. aureus was isolated from a patient with chronic osteomyelitis, and methicillin-sensitive S. aureus was identified using PHOENIX 100 (Becton, Dickinson Microbiology Systems, USA). A frozen stock of S. aureus strains was routinely grown on tryptic soy broth (TSB) with shaking at 180 rpm at 37°C for 16 h and collected by centrifugation at 3,000 rpm for 10 min. The bacterial pellets were washed and resuspended in phosphate-buffered saline (PBS). The concentration of S. aureus was adjusted to an optical density (OD) of 0.5 at 600 nm, approximately equal to 1 × 10⁸ CFU/ml, and further adjusted to 1 × 10⁶ CFU/ml for bacterial injection into IAOM mice.

IAOM mouse model infected by S. aureus.

Protocols for animal experiments were approved by the Animal Care and Use Committee at Nanfang Hospital, Southern Medical University. All the mice were housed in facility with a 12-h light/dark cycle, 24 ± 2°C room temperature, and ad libitum access to water and food. Male C57BL/6 mice aged 10 to 12 weeks were randomly divided into an IAOM group and a control group. Prior to surgery, mice were anesthetized by intraperitoneal injection of tribromoethanol (125 mg/kg of body weight). After the right hind leg was shaved followed by disinfection with iodine, a 5-mm incision was made on the dorsal side of the leg. After the femoral midshaft was exposed by blunt separation of the muscles, a canal was drilled by a 27-gauge syringe needle, with care not to penetrate the contralateral cortex of the femur. Next, a 2-mm sterile stainless pin (0.3 mm in diameter) was inserted into the bone marrow cavity through the canal. Finally, 2 μl of S. aureus solution at 1 × 10⁶ CFU/ml was slowly injected into the bone marrow cavity using a microsyringe for IAOM mice, while an equal volume of sterile PBS was used for the controls. The hole was then sealed with bone wax and the incision was closed with a 5-0 suture. Right femurs were collected on days 3, 14, and 28 after the operation for further analysis.

X-ray and micro-computed tomography (micro-CT) analysis of the femur.

After anesthesia with tribromoethanol, mice were placed with the ventral side up for X-ray imaging with an X-ray imaging system (Fx Pro in vivo imaging system; Bruker, Germany). Quantitative evaluation of the IAOM was performed using a modified scale based on the radiographic parameters previously reported (40), which include periosteal reaction, osteolysis, deformity, sequestrum formation, and spontaneous fracture. Parameters 1 to 3 were scored from 0 (absent) to 1 (mild) to 2 (severe), and parameters 4 and 5 were scored from 0 (absent) to 1 (present). A total score assigned for each sample was the sum of all the scores made for the 5 parameters by two blinded observers independently.

Femurs from mice of IAOM and control groups were dissected and cleaned of adherent soft tissue, followed by fixation in 70% for 24 h. Before micro-CT scanning, implants were removed from the femurs by expanding the hole through which the implant was inserted. Femurs were imaged using a high-resolution micro-CT (Scanco Medical, Wangen-Bruttisellen, Switzerland). All images were obtained with a voxel size of 12 μm, a voltage of 55 kV, a current of 145 mA, and an integration time of 400 ms and were averaged 2 times. Quantitative analysis was carried out using SkyScan CT-Analyser version 1.13 software (Bruker, Germany). The region of interest for evaluating cortical bone and trabecular bone were at distance of 1.2 to 2.4 mm from the distal growth plate of femur. Parameters such as bone mineral density of cortical bone (Ct. BMD), cortical area (Ct. Ar), cortical thickness (Ct. Th), cortical periosteal perimeter (Ct. Pe. Pm), and cortical endosteal perimeter (Ct. En. Pm) were used to evaluate cortical bone. Bone fraction (BV/TV), trabecular number (Tb. N), trabecular thickness (Tb. Th), and trabecular separation (Tb. Sp) were used to evaluate cancellous bone.

Bacterial quantification.

To quantify the number of S. aureus organisms in the infected bones of IAOM mice, the right femur was removed immediately after sacrifice and crushed with a tissue homogenizer (JXFSTPRP-24; Shanghai JingXin, China). The homogenized femurs were weighed (88.1 ± 23.3 mg) and suspended in sterilized PBS to achieve a mass per volume of 0.1 mg/μl. Next, the homogenates were taken for a series of dilutions and cultured on a TSB agar plate at 37°C for 24 h. Quantification of S. aureus organisms on the implant surface was performed according to the methods described previously (41). Briefly, the implant was removed, immersed in 1 ml of PBS, and eluted by sonication at 40 KHz for 5 min (SB25-12YDTD; Scientz, China). Next, elution solution was diluted and incubated on a TSB agar plate at 37°C, and bacterial colonies were counted 24 h later.

Histological analysis.

The right femur was fixed with 4% paraformaldehyde for 24 h, decalcified in 0.5 M EDTA for 7 days, and embedded in paraffin. Sagittal sections (4 μm) were obtained for staining. Sections were stained with hematoxylin-eosin (H&E) for histopathological scoring. Smeltzer’s scoring methods (42) were used to evaluate the histopathological changes by two blinded observers. Each section was assigned a score, which was the sum of intraosseous acute inflammation (0 to 4), intraosseous chronic inflammation (0 to 4), periosteal inflammation (0 to 4), and bone necrosis (0 to 4). To detect the abundance of osteoclasts, tartrate-resistant acid phosphatase (TRAP) staining was performed on deparaffinized and rehydrated sections using a leukocyte acid phosphatase kit (catalog no. 387A-1KT; Sigma-Aldrich, USA). The number of TRAP⁺ cells per millimeter of the trabecular bone surface in the area from 0.2 to 0.7 mm above the distal growth plate was quantified. To observe bone destruction and periosteal new bone formation, Goldner’s trichrome staining (catalog no. G3550; Solarbio, China) was performed.

Immunohistochemistry and immunofluorescence.

For immunohistochemical analysis, after deparaffinization and rehydration, antigen retrieval was performed by incubating the section in protease K solution (1 mg/ml) at 37°C for 15 min before endogenous peroxidase activity was quenched in 3% H₂O₂ for 15 min. After being blocked with 10% goat serum for 1 h at room temperature, sections were incubated with the rabbit anti-S. aureus antibody (catalog no. ab20920; Abcam) or anti-myeloperoxidase (anti-MPO) antibody (catalog no. ab208670; Abcam) overnight at 4°C. Next, sections were incubated with biotinylated goat anti-rabbit secondary antibody for 1 h at room temperature, followed by avidin-conjugated horseradish peroxidase (HRP) complex according to the manufacturer’s protocol (Vectastain ABC HRP kit; Vector Laboratories, USA). Finally, the peroxidase activities of sections were revealed with a 3,3’Diaminobenzidine (DAB) substrate kit (Vector Laboratories).

For immunofluorescence analysis, sections were deparaffinized, rehydrated, and subjected to antigen retrieval by incubating the section in Tris-EDTA (pH 9) solution at 95°C for 15 min. After being blocked with 10% donkey serum for 1 h at room temperature, sections were incubated with rabbit anti-α-smooth muscle actin (anti-α-SMA) antibody (catalog no. 14395-1-AP; Proteintech, China) overnight at 4°C. Alexa Fluor 488-conjugated donkey anti-rabbit IgG (H+L; 1796375; Thermo Fisher Scientific, Waltham, MA) was used as the secondary antibody. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI; E607303-0002; BBI Life Science, Shanghai, China). All sections were observed under a BX53 microscope (Olympus, Tokyo, Japan).

RNA preparation, transcriptome sequencing, and quantitative real-time PCR (qRT-PCR).

IAOM and control mice were sacrificed on days 3 and 14 after operation. The right femur was collected and pulverized in liquid nitrogen prior to extraction of RNA. Total RNA of the femur was extracted with TRIzol reagent according to the manufacturer’s instructions (Invitrogen, USA). RNA degradation and contamination were monitored on 1% agarose gels. RNA integrity was assessed using a RNA Nano 6000 assay kit (5067-1511; Agilent Technologies, USA) on the Bioanalyzer 2100 system (Agilent Technologies, USA).

For transcriptome sequencing, sequencing libraries were generated using the NEBNext Ultra RNA library prep kit for Illumina (New England BioLabs [NEB], USA) following the manufacturer’s protocol, and index codes were added to attribute sequences for each sample. PCR products were purified using the AMPure XP system (Beckman Coulter, USA), and library quality was assessed on the Agilent Bioanalyzer 2100 system. The clustering of the index-coded samples was performed on a cBot cluster generation system using TruSeq PE cluster kit v3-cBot-HS (Illumina) according to the manufacturer’s instructions. After cluster generation, the library preparations were sequenced on an Illumina sequencing platform and 150-bp paired-end reads were generated.

For qRT-PCR analysis, 5× Evo M-MLV RT master mix (catalog no. AG11706; Accurate Biotechnology, China) was used to reverse transcribe mRNA into cDNA. Following this, real-time quantitative PCR was performed using the SYBR green Premix Pro Taq HS qPCR kit (catalog no. AG11701; Accurate Biotechnology, China). The primer sequences for the genes used were as follows: for glyceraldehyde-3-phosphate dehydrogenase (GAPDH), 5′-TGTCGTGGAGTCTACTGGTG-3′ (forward) and 5′-GCATTGCTGACAATCTTGAG-3′ (reverse); for IL-6, 5′-CCCCAATTTCCAATGCTCTCC-3′ (forward) and 5′-CGCACTAGGTTTGCCGAGTA-3′ (reverse); for Cxcl10, 5′-CCAAGTGCTGCCGTCATTTTC-3′ (forward) and 5′-GGCTCGCAGGGATGATTTCAA-3′ (reverse); for IFN-γ, 5′-TGAACGCTACACACTGCATCTTGG-3′ (forward) and 5′-CGACTCCTTTTCCGCTTCCTGAG-3′ (reverse); and for Cxcl9, 5′-GGAGTTCGAGGAACCCTAGTG-3′ (forward) and 5′-GGGATTTGTAGTGGATCGTGC-3′ (reverse). The relative expression of each target gene was calculated using the threshold cycle (2^−△△CT) method with GAPDH for normalization.

Differential expression analysis.

The DESeq2 R package (1.16.1) (43) was used for differential expression analysis between IAOM and control (Ctrl) groups on days 3 and 14 after operation (IAOM_3D versus Ctrl_3D and IAOM _14D versus Ctrl_14D, respectively). The resulting P values were adjusted using Benjamini and Hochberg’s approach to control the false-discovery rate. The genes with an adjusted P value of <0.05 and |log₂ (fold change)| of >1 were recognized as differentially expressed genes (DEGs).

GO and KEGG enrichment analysis of DEGs.

Gene Ontology (GO) is a comprehensive database that can classify and describe genes according to their essential functions, which can be divided into three parts: biological process (BP), cellular component (CC), and molecular function (MF) (44). Kyoto Encyclopedia of Genes and Genomes (KEGG) is a database involving pathway information, which can help understand the advanced functions of biological systems from the molecular level (45). GO and KEGG enrichment analysis of DEGs was implemented by the clusterProfiler R package (46). GO and KEGG terms with adjusted P values of <0.05 were considered significant enrichment.

PPI analysis of DEGs.

The status of a target gene or protein in the protein network was identified through the protein-protein interaction (PPI) network. The DEGs were imported into the STRING database (https://string-db.org) and converted into the protein name before the PPI network graph and its interaction information file were obtained (47). The information of the protein interaction file was imported into Cytoscape software to create a visual network (https://www.cytoscape.org), in which the node, degree, closeness centrality and edge represented the protein, the number of interacted proteins, the degree of centrality in the network, and the lines connecting two nodes, respectively. Finally, the MCODE (molecular complex detection) plug-in was used for cluster analysis. The following were set as analysis parameters: node score cutoff of 0.2, degree cutoff of 2, k-core of ≥2, and maximum depth of 100.

Statistics.

All statistics were analyzed using GraphPad Prism 7.0. Two groups were compared using the Mann-Whitney U test. Multiple comparisons were assessed by the Kruskal-Wallis test. A P value of <0.05 was considered a significant difference.

Data availability.

The raw RNA-Seq data have been deposited in the GEO database under accession no. GSE166522.