Single-cell transcriptome reveals Staphylococcus aureus modulating fibroblast differentiation in the bone-implant interface

Human bone-implant interface consists of eight major cell types

Five fresh peri-implant tissues were collected from three patients with PJI and two patients with AL for scRNA-seq as illustrated in Fig. 1a, clinical characteristics and implant characteristics were presented in Additional file 1: Table S1. It is noteworthy that PJI11 was collected from patient with acute infection while PJI03, PJI04 were derived from chronic infection patients. After a series of quality control procedures (Additional file 2: Fig. S1a, b), a total of 36,466 cells were qualified for subsequent analysis with 21,982 and 14,484 cells from PJI and AL, respectively. Eight major cell types were manually annotated according to canonical cell marker genes and eleven cell clusters were confirmed with dimension reduction and unsupervised clustering (Fig. 1b, Additional file 3: Fig S2a). Specifically, the following cells were identified by their distinct marker genes: fibroblasts (PDGFRA, COL3A1), endothelial cells (SELE, VWF), smooth muscle cells (ACTA2, TAGLN), B cells (CD79A, MS4A1), T1-4 cells (CD3D, CD4), plasma cells (MZB1, SDC1), mast cells (CST3, KIT), myeloid cells (LYZ, CD14) (Fig. 1c, d). Gene ontology enrichment analysis revealed their corresponding functions in immune modulation and tissue construction (Additional file 2: Fig. S1d).

Fig. 1figure 1

Single-cell RNA sequencing reveals cell heterogeneity in the periprosthetic environment. a Schematic overview showing study workflow from sampling to single-cell sequencing and data analysis. b UMAP visualization of total cells from the periprosthetic tissue of patients with AL and PJI, single cells are colored by cluster annotation. c Colored UMAP plot showing marker genes for each type of cell. d Violin plots showing the selected canonical marker genes across the annotated cell clusters. Expression levels are normalized and log transformed. e Stacked bar plot showing proportions of annotated cell clusters in each sample

Compared with the PJI group, the AL group exhibited a higher proportion of stromal cells including fibroblasts (43.45 vs. 7.53%), endothelial cells (20.27 vs. 5.54%), and smooth muscle cells (7.92 vs. 4.33%), whereas the PJI group consists of a higher percentage of immune cells such as T cells (43.26 vs. 12.5%) and myeloid cells (20.68% vs 8.06%). It is noteworthy that plasma cells are specifically presented in the PJI group and rarely detected in the AL group (7.87% vs. 0.19%) (Fig. 1e, Additional file 2: Fig. S1c).

Fibroblasts, T_1 cells and myeloid cells are the major modulators in the bone-implant interface

A comparison of cell–cell interactions between PJI and AL groups showed stronger crosstalks among immune cells, but significantly downregulated fibroblasts–fibroblasts communication in the PJI group (Fig. 2a, b, Additional file 4: Fig. S3a–f). By computing the strengths of incoming and outgoing signals, we found that fibroblasts play a major cell–cell communication role in an aseptic environment (Fig. 2c, Additional file 4: Fig. S3j, k). In contrast, immune cells such as myeloid and T_1 cells contributed more frequently to cell communications in an infected environment (Fig. 2d, Additional file 4: Fig. S3g–i). Moreover, receptor-ligand enrichment indicated that downregulated communications in PJI mainly concentrated in the collagen pathway and PTN pathway which are strongly associated with fibroblast functions (Fig. 2e, f, i). By analyzing immune cell-dependent intercellular communications, much higher communication strength relating to myeloid was observed in the acute PJI group (Fig. 2g, j). The upregulated pathways are predominantly responsible for chemokines (IL10, IL16) and antigen processing (MHC-I, MHC-II). On the contrary, the chronic PJI group mainly dependent on T_1 cell interaction (Fig. 2h, k).

Fig. 2figure 2

Cell–cell interaction analysis reveals differences in the major mediators between AL and PJI. a, b Heatmap and circos plot shows differential interaction strength relative to PJI, red indicates increased interaction strength and blue indicates decreased strength in PJI. c, d Scatter plot shows incoming interaction strength (y-axis) and outcoming interaction strength (x-axis) for each cell cluster in AL (c) and PJI (d) samples. e The dotted heatmap shows major differentially expressed ligand-receptor pairs in AL and PJI groups. fk Bar plots list the relative strength of pathways from and target to Fibroblasts (f, i)/Myeloid cells (g, j)/T_1 cells (h, k)

Because fibroblast, myeloid cells and T_1 cells were identified as the major regulators, we decided to further compare their differences between AL and PJI groups. Fibroblasts/myeloid cells were classified into 6 subtypes and T_1 cells were classified into 5 subtypes according to their transcriptome heterogeneity (Fig. 3a, c, e Additional file 3: Fig. S2b–d), Fibro_0,1,2, Myeloid_2,5,4 and T_1 subtype 2 were mainly presented in the AL group while Fibro_3,4, Myeloid_0,1,3,4 and T_1 subtype 0,1,3,4 predominantly existed in the PJI groups. The representative marker for each cell subtype were provided in Fig. 3b, d, f. The upregulated genes of fibroblasts in the PJI group were mainly enriched in inflammatory pathways and the downregulated genes were responsible for extracellular matrix production and copper ion detoxification (Fig. 3h, Additional file 5: Fig. S4a, b). Myeloid cells in the PJI group were upregulated in immune-modulating pathways (Fig. 3g, Additional file 5: Fig. S4c, d). T_1 cells in the PJI groups were upregulated in lymphocyte/leukocyte activation (Additional file 5: Fig. S4e, f). H&E staining results revealed that the PJI samples have more immune cell infiltration. Masson staining showed that samples from the AL group exhibited a higher level of collagen deposition (Fig. 3i).

Fig. 3figure 3

Fibroblast, myeloid cells and T_1 cells subtype analysis revealed cell functional differences. a, c, d UMAP visualization of fibroblasts (a)/myeloid cells (c)/T_1 cells d from the periprosthetic tissue of patients with AL and PJI, cells are colored by subtypes. b, d, f Heatmap displayed the highly expressed marker genes for each fibroblast (b)/myeloid cell (d)/T_1 cell (f) subclusters. g AUCell quantification of Gene ontology term for Cytokine production in myeloid cells. h AUCell quantification of Gene ontology term for extracellular matrix organization in fibroblasts. i Images of HE and Masson staining results for AL and PJI samples

Fibroblasts in the bone-implant interface are mainly CTHRC1 +

By integrating the publicly available datasets with our in-house generated fibroblast expression profiles as illustrated in Fig. 4a, we divided all of the analyzed fibroblasts into seven meta-clusters (Fig. 4b). Among the seven meta-clusters, we noticed fibroblast meta-cluster2 overexpresses collagen-related genes (COL1A, COL1A2, COL3A1, CTHRC1) (Fig. 4c, Additional file 6: Fig. S5a–e). This CTHRC1+ fibroblast was observed in AL, PJI, periodontitis, and synovium of osteoarthritis and rheumatoid arthritis(Fig. 4d). Gene ontology (GO) analysis and GSEA enrichment both showed that the overexpressed genes in CTHRC1+ fibroblasts were enriched in ossification and collagen organization (Fig. 4e-f). Comparing the expression pattern of CTHRC1 in PJI and AL samples, we found both groups were composed of a large portion of CTHRC1 + fibroblasts (81.5% in AL vs 86.7% in PJI). Hence, we believe that CTHRC1 + fibroblast is the major type of fibroblast in the human bone-implant interface. (Fig. 5a).

Fig. 4figure 4

Meta-analysis of human fibroblast scRNA-seq data. a Illustration of the selected scRNA-seq data from GEO publicly available database. scRNA-seq derived from different organs were acquired from the GEO database and integrated with the in-house generated scRNA-seq data, then fibroblast scRNA-seq was extracted for further meta-analysis. (b) UMAP visualization of the fibroblast, cells are colored according to cell meta-clusters. c Heatmap displayed the highly expressed marker genes for each fibroblast meta-clusters. d UMAP visualization of the fibroblast in each sample. e The heatmap shows the GSEA enrichment result for each of the fibroblast cell meta-clusters. GO:BP databases were used in this analysis. f GSEA results shows three GO terms (ossification, collagen trimer, collagen fibril organization) are enriched in the fibroblast meta-cluster 2

Fig. 5figure 5

In-depth fibroblast trajectory analysis reveals a bipolar mode of differentiation. a UMAP visualization of the fibroblast, cells are colored according to CTHRC1 expression level. b Pseudotime trajectory analysis shows fibroblasts differentiation, cells are colored by pseudotime. c Trajectory plots show fibroblasts differentiation in AL and PJI groups, cells are colored by cell states. d Stacked bar plot showing proportions of each cell state in AL and PJI groups. e UMAP visualization of the fibroblast, cells are colored according to subclusters. f Fibroblast subclusters are projected to pseudotime trajectory, cells are colored by subclusters. g Heatmap revealed pseudotime-dependent differentially expressed gene clusters for cell fate1, cell fate2, and pre-branched cells. h Bar plots show the enrichment result of gene ontology and KEGG pathway for each of the gene clusters identified in (g), both are colored with log10qvalues

Bipolar differentiation of fibroblast in the bone-implant interface

To further investigate fibroblast responses to infection in the bone-implant interface, we carried out a single-cell pseudotime trajectory analysis for fibroblasts in the AL and PJI groups, the result showed that fibroblasts could be classified into two continuous cell lineages: cell fate1/2 (Fig. 5b, c). The PJI group comprised many more cells that underwent fate1 than the AL group (77.48% vs. 12.82%), whereas the AL group mainly comprised cells in fate2 and pre-branched cells (Fig. 5d). Fibroblast subcluster1, 2, 4 were at pre-branched state, cell fate2, and cell fate1 respectively. Other subclusters were at intermediate state (Fig. 5e, f). We further classified pseudotime-dependent genes into 3 gene sets (geneset1–geneset3) according to their mode of expression (Fig. 5g, Additional file 7: Table S2). Geneset1 is composed of 317 genes and is responsible for cell fate2 regulation, GO term enrichment results showed its relation to extracellular matrix organization and stress response to copper ions (Fig. 5h). Geneset2 consists of 422 genes that contributed to fate1 differentiation. Its KEGG enrichment results were enriched in Staphylococcus infection and the phagosome pathway, GO term enrichment suggested its immune modulation functions such as positive regulation of leukocyte activation. Genset3 comprised 609 genes involved in extracellular structure organization, regulation of vascular development, epithelial cell proliferation, and ossification. Given the above findings, we termed cell fate1, cell fate2, and pre-branched cells as inflammatory, matrix-producing, and pre-branched fibroblasts, respectively.

Identification of key driver transcription factors determining fibroblast differentiation

To identify crucial regulators responsible for infection-induced fibroblast differentiation, we performed regulatory network analysis (SCENIC) for a total number of 388 regulons. The resultant 337 regulons with significant differences between PJI and AL groups were selected and intersected with genes critical for fibroblast differentiation (1348 genes identified during pseudotime analysis), thus resulting in 45 transcription factors. To exclude the effects of over-representation of a dataset due to high cell number contribution, we removed 34 transcription regulators whose expression levels were not consistent among individuals with the same diagnosis (Fig. 6a). The AUCell scores for each of the 11 regulons were shown in Fig. 6b, a distinct difference could be observed between PJI and AL. Furthermore, We found that inflammatory fibroblasts are strongly correlated with TFEC and NPAS2, whereas matrix-producing fibroblasts are mainly modulated by HMX1, SOX5, SOX9, ZIC1, ETS2, and FOXO1. The remaining TFs (NFATC2, KLF4, and EGR2) are responsible for cell differentiation in the pre-branched state (Fig. 6c). By selecting three representative TFs in each type of cell fates (NPAS2, SOX5, NFATC2), we verified the above-mentioned differences at the protein level using immunofluorescence (Fig. 6d, e, Additional file 8: Fig. S6). The results showed that NPAS2 was highly expressed in the PJI group while SOX5 and NFATC2 were mainly expressed in the AL group.

Fig. 6figure 6

Identification for both pseudotime-dependent and disease-dependent regulators. a Workflow for identification of 11 key regulators. SCENIC analysis result in 337 regulons with significant differences between PJI and AL groups then were intersected with genes critical for fibroblast differentiation (1348 genes identified during pseudotime analysis), resulting in 45 transcription factors. We further removed 34 transcription regulators whose expression levels were not consistent among individuals with the same diagnosis. b SCENIC analysis result for the selected eleven regulons visualized with heatmap. c Heatmap shows the eleven transcription factor expression changes along pseudotime differentiation. Immunofluorescent assays display NPAS2 d SOX5 e expression in the AL and PJI groups, sections were labeled with anti-collagen III (green) which shows the distribution of collagen

Forskolin is a potential treatment strategy by targeting fibroblast differentiation

To identify potential compounds for regulating fibroblast differentiation, we carried out a CMap analysis. Differentially expressed genes (DEGs) of fibroblasts in the PJI group were filtered by setting the adjusted p-value < 0.01 and log2Foldchange at ± 1.4. A total number of 24 genes were upregulated and 45 were downregulated (Fig. 7a, Additional file 9: Table S3). The most upregulated genes in the PJI group were IGKC, CXCL13, APOE, and IGLC3 whereas the most downregulated genes were PRG4, MGP, PLA2G2A, and CXCL14 (Fig. 7b). The DEGs were used as an input for CMap analysis, the top 70 compound hits are shown in Fig. 7c. Only forskolin, an adenylyl cyclase activator, has an estimated score < − 90.

Fig. 7figure 7

CMap analysis screening for potential fibroblast differentiation regulators. a Scatter plot showing differentially expressed genes between PJI and AL group. The most significant hits are highlighted with color rectangles. b Lolipop displays the top ten upregulated and downregulated genes in the PJI group. c The heatmap shows correlation scores for each compound (columns) when treated to a specific cell line (rows). The bar plot displays the average scores for each tested compound

留言 (0)

沒有登入
gif