Fibrocyte enrichment and myofibroblastic adaptation causes nucleus pulposus fibrosis and associates with disc degeneration severity

Introduction

The intervertebral discs (IVDs) are the largest avascular connective tissues in the body, playing a central role in spine movement. IVD degeneration (IDD) is associated with fibrosis of the nucleus pulposus (NP), the innermost colloidal core of IVDs, along with a loss of hydration and swelling pressure. The degeneration involves a degradation of hyaline extracellular matrix (ECM), in particular aggrecan and collagen II, and accumulation of fibrotic components including collagen I (COL1) and small leucin-rich proteoglycans such as biglycan and fibronectin.¹ This leads to reduced tissue mechanical strength and hence disc prolapse under load, ultimately causing back pain and myelopathy.² Previous study has shown that transplantation of mesenchymal stromal cells (MSCs)³ can alleviate IDD via inhibiting NP fibrosis. Understanding and controlling the profibrotic events may therefore hold the key to modifying the degeneration process. However, the mechanism of NP fibrosis and the origin of the fibroblastic population in the IDD remains elusive.

The primitive NP is initially populated by forkhead transcription factor (FOXA2)- or brachyury (TBXT)-expressing vacuolated cells derived from the notochord.^4,5,6 In humans, these vacuolated cells are gradually replaced by the chondrocyte-like cells after birth.⁷ Tissue fibrosis relates to an excessive fibroblast and myofibroblast activity.⁸ Reports have shown an increased number of α-smooth muscle actin (αSMA), fibroblast-specific protein 1 (FSP1) and COL1 positive fibroblastic cells in the degenerative IVDs.^4,9 In fact, NP cells (NPC) have a capacity to become FSP1⁺ cells when exposed to excessive TGF-β^5,10 as well as undergo fibroblastic/myofibroblastic transition in an injury-induced IDD model.⁴ However, tracing study indicated that not all myofibroblastic cells are derived from local cells,⁴ implying alternative origins of effector cells for NP fibrosis. In line with this notion, single cell transcriptomic studies have revealed that the NP of human degenerative IVDs^{7,11,12,13,14} and animal model of IDD⁶ contain cell types other than IVD cells. These include endothelial cells, myeloid granulocytic suppressor cells, CCR2⁺ monocytes and derivatives of macrophages. Consistent with the finding, studies showed that bone marrow cells could migrate into mechanically overloaded IVD explants¹⁵ and that focal enrichment of fibroblastic cells could be found near the infiltrating blood vessels in the degenerative IVDs.¹⁶ These evidences support that access of extrinsic cells to the degenerated IVD may mediate NP fibrosis.

We hypothesized that cell sources other than resident disc cells can contribute to NP fibrosis. In this study, we interrogated the cell hierarchy in the NP of degenerative IVDs through an integrative analysis of the published single cell transcriptomes and characterization in human tissue and cells. Our findings indicated an emergence of a small hematopoietic population that possess a fibrocyte phenotype and myofibroblastic identity in the degenerative samples. Using a mouse model that labels or depletes myeloid-derived cells, we analyzed the time course of the disc fibrocytes appearance and identified their role in NP fibrosis and IDD progression. Our study implies a function of circulatory fibrocytes in NP fibrosis and as a target for modifying IDD.

Results

Identification of myofibroblastic NPC in IDD and NP fibrosis

We examined the cellular heterogeneity and fibroblastic cell entities by an integrative analysis of 4 published single-cell RNA sequencing (scRNA-seq) data derived from cells isolated from NP tissues^7,12,14 or recognized as NPC.¹¹ The datasets consist of 6 non-degenerative human IVDs (3 scoliosis, 1 spine fracture, 1 spinal injury and 1 brain death cases), as well as 13 mildly (Pfirrmann II–III) and 12 severely (Pfirrmann IV-V) degenerative human IVDs (23 disc herniation and 2 burst fracture cases) (Table 1). A number of existing NPC phenotypes were previously identified from these datasets (marker genes summarized in Table S1). The meta-analysis identified 12 cell clusters from a total of 192 184 cells and the cell types were annotated based on their cluster differentially expressed genes (DEGs, SI file 1 and Fig. 1a, d). These clusters were mostly present across the four datasets although their proportion and distribution varied (Fig. 1b). Clustered NPC were identified by their expression of ACAN, COL2A1, and SOX9 (SI file 1). The most abundant NPC population was the chondrogenic cluster (ChonNP) (51%) which expressed COL2A1, ACAN and CNMD. The regulatory cluster (RegNP) (21%) expressed MMP3, GPX3 and CHI3L2/1. These RegNP gene markers were also expressed in the inflammatory C1/3 chonNPC and regulatory NPC reported by Han et al.⁷ and Tu et al.¹² respectively (Table S1). The fibroblastic cluster (FibroNP; 8%) expressed COL1A1, COL3A1, POSTN and ASPN, which are known to associate with tissue fibrosis. We found that the proportion of FibroNP significantly increased with IDD severity and presented as a major NPC population (30%) in the severely degenerated samples (Fig. 1c and Fig. S1). The cycling cluster (CyclingNP) (1%) expressed the cell proliferation genes including TOP2A and STMN1, containing two distinct subsets expressing high level of either collagen I or II gene (COL1A1^high and COL2A1^high). The progenitor cluster (ProgNP) (1%), distinguished by the expression of PDGFRA and PLA2G2A, was unique to the non-degenerative samples as reported by Gan et al.¹⁴ In addition to 5 NPC clusters, we also identified neutrophils (LYZ⁺), granulocyte-like myeloid-derived suppressor cells (G-MDSC⁺; MPO⁺), macrophages (CD68⁺), T cells (TRBC2⁺), endothelial cells (CD34⁺ and PECAM1⁺), pericytes (RGS5⁺ and ACTA2⁺), and erythrocytes (HBB⁺) (Fig. 1d, e). Endothelial cells (2%) were also found, which highly expressed PECAM1 and CD34 (Fig. 1e and Fig. S2). Erythrocytes (1%) were defined by gene expression of HBB (Fig. 1e).

Table 1 Demographic data

Full size table

Contractile myofibroblasts are primary effector cells of tissue fibrosis.⁸ Fibroblast activation protein-α (FAPα, encoded by FAP) and FSP1 (encoded by S100A4) mark the active fibroblasts or myofibroblast precursors,¹⁷ and αSMA (encoded by ACTA2) marks the mature myofibroblasts.¹⁸ These three markers, which were used to indicate fibroblastic NPC in mouse IDD,^4,9 were found enriched in the FibroNP (Fig. 2a). A markedly higher abundance of ACTA2⁺, FAP⁺ and S100A4⁺ FibroNP cells were found in the severely degenerative NP (dNP) than the mildly degenerative or the non-degenerative NP (nNP) (Fig. 2b). More fibrillar collagens, specifically COL1 and thin collagen III, were predominantly deposited in human degenerative IVDs, which together with reduction of hyaline matrix aggrecan and collagen II, indicated NP fibrosis (Fig. S3). Immunofluorescence further verified a higher number of cells expressing αSMA (14.0% vs 48%), FAPα (31.4% vs 78.8%) and FSP1 (28.4% vs 65%) in the dNP tissues (from degenerative disc disease subjects) than in the nNP tissues (from scoliosis subjects) (Fig. 2c), indicating the augmented myofibroblast sprouting in dNP. In line with our single-cell transcriptome analysis, cells from either dNP or nNP tissues were found to contain a lower amount of αSMA⁺ cells than FAPα⁺ or FSP1⁺ cells, which was further supported by fluorescence-activated cell sorter (FACS) analysis of isolated NPC (Fig. 2d). FACS also demonstrated an increased myofibroblast population in the dNP compared to the nNP. Notably, nNP [(78.0 ± 11.7)/%] contained a considerable amount of cells expressing intracellular COL1 compared to dNP [(93.4 ± 5.5)/%] (Fig. S4). Under monolayer expansion, dNP cells were more flattened and expressed higher levels of myofibroblastic markers including ACTA2, FAP and COL1A1 (Fig. S5). These findings support that human IDD associates with an accumulation of NPC which have myofibroblastic characteristics.

Enrichment of fibroblastic NPC with hematopoietic features in IDD

Previous histological studies suggested that immune cells such as CD68⁺ macrophages and neutrophils may infiltrate into the NP during disc degeneration.¹⁹ In the single cell analysis, immunocyte-like populations could be found at all degeneration grades, expressing markers for neutrophils (LYZ and HLA-DRA), macrophages (CD163 and CD68), G-MDSC (ITGAM and MPO) and T cells (TRAC and TRBC2) (SI file 1). Interestingly, we found that COL1A1, COL3A1 and POSTN were also expressed in the immunocytes, predominantly the neutrophil and macrophage populations, albeit at a considerably lower level compared with FibroNP (Figs. 3a and S6). These COL1A1⁺ immunocytes expressed myeloid lineage genes (e.g., S100A9, PTPRC, ITGAM and ITGB1), S100A4, and ECM genes VIM, COL1A1 and FNDC3A, but not macrophage marker CD115 or T-cell marker CD90 (THY1) (SI file 2), consistent with the classification of fibrocytes.²⁰ Fibrocytes are conventionally defined as a rare subpopulation of monocytic progenitors that co-expresses collagen I and III in addition to hematopoietic markers CD11b, CD45 or CD34.^21,22,23 While we could identify cells showing a co-expression of CD34 with COL1 or αSMA in the dNP tissue, the integrative scRNA-seq analysis indicated the CD34⁺COL1A1⁺ or CD34⁺ACTA2⁺ cells were in fact mostly enriched in the endothelial cell and pericyte clusters but not the immunocyte cluster (Fig. S7). We further identified COL1A1⁺ subpopulations co-expressing ITGAM (coding CD11b) (1 094 cells) or PTPRC (coding CD45) (1738 cells) (Fig. 3b). The abundance of these subpopulations increased with degeneration severity and occupied up to 0.1% cell population in the degenerative group (Fig. 3c). Immunostaining and flow cytometry analysis demonstrated the presence of CD45-expressing cells in the degenerative samples, all of which co-express COL1. These CD45⁺COL1⁺ cells were increased in the dNP tissue (16.4% vs 5.4%, Fig. 3d) or dNP-derived cell culture (0.80% vs 0.30%, Fig. 3e) compared to the nNP controls and adopted an elongated morphology (insert, Fig. 3d). The combined single cell analysis and in vitro validation therefore support the presence of fibrocytes in degenerative IVD.

Mapping of the PTPRC⁺COL1A1⁺ and ITGAM⁺COL1A1⁺ cells indicated the enrichment of fibrocytes within the NPC clusters, present mostly in the FibroNP (7.5%, as fib-fNP), ChonNP (4.2%, as fib-cNP) and RegNP (3.0%, as fib-rNP), in addition to the immunocyte clusters of macrophages (35.4%, as fib-M), neutrophils (34.2%, as fib-N), T cells (11.4%, as fib-T) and G-MDSC (3.0%, fib-G) (Table S2). By comparing these fibrocyte populations, we identified specific gene markers (SI file 3) and examined their expression (Fig. 3f). GNB1 and gluconeogenesis-related gene ALDOA, together with marker genes of COL1A1 and ITGAM, were enriched in both immunocytes- (fib-M/N) and NPC-clustered (fib-cNP/fNP) fibrocytes. Within immunocyte-clustered fibrocyte subpopulations (fib-M/N/T/G), a total of 47 genes (e.g., PTPRC and CAPZA1) were considerably enhanced and mainly linked to the actin organization and immune process (Table 2 and Fig. S8). RORA was dominantly expressed in fib-T. Fib-M-related genes LGALS1 and LGALS3BP have been linked to monocyte-derived fibrocyte differentiation.²⁴ 43 genes (e.g., COL3A1, ASPN and ITGB5) that are commonly expressed in the three NPC-derived fibrocyte subsets are related to the response to TGF-β (Fig. S8). This finding is consistent with the notion that, similar to other fibrotic diseases, disc fibrocyte maturation may involve TGF-β-mediated endothelial- mesenchymal transition.²⁵ We also performed GO analysis for DEGs specific to the NPC-fibrocytes (Table 3). 35 genes distinctly enriched in fib-fNP were mostly related to cell adhesion function. 26 genes enriched in the Fib-cNP were related to matrix organization. Two degeneration-featured genes, HTRA1²⁶ and ANGPTL4²⁷ were distinctly expressed in the NPC-fibrocytes (Fig. 3g). Fibroinflammatory molecules TGFBI and SPP1 from the disc fibrocytes were inferred to act on the FibroNP and stimulate the fibrotic process in CellChat analysis (Fig. S9). Fibrocyte-FibroNP communication may also rely on POSTN-ITGB5. Moreover, ACTA2 was found expressed in PTPRC⁺ FibroNP cells (Fig. 3h), supporting their potential in myofibroblastic differentiation.^24,25 Consistent with the scRNA-seq finding, co-immunostaining showed a markedly higher expression of CD45 and αSMA in the dNP compared to the nNP (Fig. 3h). We also observed αSMA⁺ cells negative for CD45.

Table 2 Analysis of DEGs for immunocytes-clustered fibrocytes

Full size table

Table 3 Analysis of DEGs for NPC-clustered fibrocytes

Full size table

Given the fibrocytes being a lineage of monocytes expressing CD11b, we studied their temporospatial distribution in puncture injury-induced disc degeneration in CD11b-DTR mice, in which the expression of EGFP-diphtheria toxin (DT) receptor fusion protein is controlled by the ITGAM promoter, thereby directing transgene expression and expressing membrane-localized green fluorescence protein (GFP) in monocytes/macrophages.^23,28 Normal mouse NP, which contains predominantly notochordal cells,⁴ was negative for GFP expression (Fig. S10). Annulus puncture injury led to disc degeneration with NP fibrosis.^3,4 We observed an emergence of GFP⁺ cells in the NP (~15% of NPC population) by 3 days after puncture, which continued to peak at roughly 35% by 2 weeks post-puncture (wpp) and gradually subsided thereafter (Fig. 4a). GFP⁺ cells were also observed in the annulus fibrosus (AF) (Fig. S10c). We could observe GFP⁺ cells that barely expressed CD11b in the NP at 2wpp (Fig. 4b), possibly because of a longer turnover of GFP compared to CD11b.²⁸ Up to 95% of the GFP⁺ NPC were CD45-positive, indicating their myeloid origin (Fig. 4c). In line with previous report of fibrocyte differentiation into myofibroblast, we observed αSMA expression in nearly 60% GFP⁺ cells at 4 wpp (Fig. 4d). The abundance of these dual-positive cells decreased significantly by 8 wpp, indicating that the emergence of fibrocytes might be transient or they underwent differentiation. Furthermore, we detected CD45⁺COL1⁺ and CD45⁺αSMA⁺ cells in the punctured disc (Fig. 5b, c), and the amount of CD45⁺αSMA⁺ NPC showed a two-fold reduction by 8 wpp compared to 4 wpp. These findings collectively support the notion of fibrocyte infiltration into the degenerating discs and adopting a myofibroblastic phenotype (CD45⁺αSMA⁺) during NP fibrosis.

Fibrocyte depletion alleviates puncture-induced NP fibrosis

We tested whether depletion of CD11b-expressing cells by diphtheria toxin (DT) administration²⁸ could lead to a reduction of fibrocytes in the puncture induced IDD. We first examined the frequency and dosage of DT injection to enable extended depletion while minimizing lethality due to compromised immunity (Table S3). The number of CD11b⁺ monocytes were largely reduced in peripheral blood (>50%) after twice administration of DT at 10 μg/kg (Fig. S11). For the IDD model, the first injection was performed 3 days prior to the puncture surgery (Fig. 5a). We found a marked decrease of the GFP⁺ cells in the NP at 2 and 8 wpp, supporting the effect of DT on the transgene expressing cells (Fig. S11). Interestingly, a lower number of CD45⁺COL1⁺ cells was observed in the NP after the second dose of DT injection by 4 wpp when compared to vehicle control (Fig. 5b). The amount of CD45⁺αSMA⁺ NPC also tended to decrease although the difference is not statistically significant (Fig. 5c). These findings indicate that DT administration caused a depletion of monocytic fibrocytes in the NP which emerged during injury-induced IDD.

We questioned if the loss of the myeloid-derived fibrocytes leads to a modification of IDD and NP fibrosis. Puncture induced IDD showed a significant disc height loss by 40.4% at 2 wpp and 28.4% at 8 wpp (Figs. 5d and S12). DT administration could induce a gradual recovery and reduce the disc height loss to only 17.2% at 8 wpp. Disc integrity assessed by FAST staining³ showed a prominent improvement with DT administration at both 4 and 8 wpp, including reduced lacune formation, preserved NP-AF boundary, and increased alcian blue intensity in the NP (Fig. 5e). Moreover, alcian blue staining in the NP of the DT-injected mice was largely maintained, implying that the loss of proteoglycans was prevented. The amount of myofibroblastic NPC positive for αSMA steadily increased after puncture and took up to 76% of the whole population by 8 wpp (Fig. 5c). DT administration could lower its positivity to 41%. It is noteworthy that the majority of αSMA⁺ cells detected in the punctured discs, in particular in the DT-treated group, was negative for CD45 (>90%). Picrosirius red staining and immunostaining of COL1 further consolidated a decrease in fibrillar collagens (Fig. 5f–h). Pericellular expression of COL1 was noted in the NP after DT administration. Collectively, these findings demonstrated that IDD progression is dependent on a function of myeloid-derived cells where the CD45⁺COL1⁺ fibrocytes and αSMA⁺ myofibroblastic cells may mediate NP fibrosis.

Discussion

NPC are conventionally defined as collagen II-, aggrecan- and Sox9-expressing cells. Recent studies of single-cell transcriptome have indicated cell heterogeneity in the human NP, which contains cell populations expressing notochordal, chondrogenic, and fibroblastic phenotypes.^{7,11,12,13,14} Our meta-analysis of the human disc scRNA-seq datasets has outlined the commonly featured NPC hierarchy and stratified their fibroblastic subtypes. COL1A1⁺ FibroNP is previously defined as adhesion/fibroNPC¹² or C4 NPC.⁷ The FibroNP population was not reported in Wang et al scRNA-seq dataset, possibly because their cells were derived from the whole disc tissue without NP isolation and therefore may have dominated by COL1A1⁺ AF cells.¹¹ Nonetheless, they defined the FibroNP genes (POSTN, COL1A1, COL3A1 and TMSB4X) in MK167⁺ progenitor and NP progenitor cells, implying the fibroblast presence within the two NP subtypes. Han et al found a small fraction of fibrochondrocyte progenitors in the NP, which partially constitute the aforementioned two NP subsets. The regulatory NPC and C1/3 chonNPC subtypes reported by Tu¹² and Han⁷ are merged to the RegNP cluster in this study, sharing the expression of MMP3, CHI3L2/1 and GPX3. The CyclingNP cluster is a pool of TOP2A⁺ cells fraction from the COL2A1⁺ ChonNP or COL1A1⁺ FibroNP subsets, implying they may be derived from a common replicative progenitor origin. Pseudotime ordering of the NP clusters and velocity analysis may help construct their relationship.

The combination of collagen production, in particular collagen I, and the expression of leukocytic common antigen (CD45) or one of the hematopoietic or myeloid antigens (e.g., CD34 or CD11b) is considered as a sufficient criterion for fibrocyte identity.²² Our data show that CD34⁺COL1A1⁺ cells are mainly enriched in the endothelial cells and pericytes subpopulations. ITGAM (CD11b) or PTPRC (CD45) is therefore preferable to CD34 as the marker of fibrocytes in human NP tissues. Notably, the PTPRC⁺COL1A1⁺ and ITGAM⁺COL1A1⁺ cells express many common genes, and that PTPRC⁺ cells are more abundant and have a wider UMAP distribution than ITGAM⁺ cells. This is consistent with the suggestion that pan-hematopoietic marker CD45 is a more representative fibrocyte marker compared to CD11b.²² Our data indicate that the fibrocytes constitute 10%–20% of the NPC in the degenerative samples, in agreement with the low abundance of circulating fibrocytes (CD45⁺COL1⁺) found in lung (1.3 ± 1.6 cells/mm²) and kidney fibrosis (80 cells/mm² or 12% positivity).^20,29 Pronase digestion was commonly deployed for releasing single cells from the NP in scRNA-seq and flow cytometry assays.³⁰ However, the digestion can also result in extensive membrane protein degradation, and thereby affect the antibody recognition of surface molecules.³¹ This may in part explain the discrepancy in measurements of CD45⁺COL1⁺ cells between in situ immunohistochemistry and flow cytometry.

In addition to the NP clusters, the ITGAM ⁺COL1A1⁺ or PTPRC⁺COL1A1⁺ cells are mostly located within the macrophages, neutrophils and T cells clusters, implying the fibrocytes may be derived from multiple sources or differentiation routes of the immune system. A potential causal relationship between blood immune cells and IDD have been suggested.³² Interestingly, differentiation of CD11b⁺CD115⁺Gr1⁺ monocytes into fibrocytes could be dependent on CD4⁺ T cells.³³ Immune cells can infiltrate through focal defects in cartilaginous endplate (EP) and AF and advance the fibroinflammatory process.³⁴ The granulocyte-like myeloid-derived suppressor cells (G-MDSCs), shown in cardiac fibrosis with myofibroblast activation capabilities, may contribute to T-cell suppression and ROS production in IDD.¹² Monocyte-derived inflammatory macrophages could differentiate into myofibroblasts^35,36 and were identified in the NP of degenerative discs.^19,37 Interestingly, CD11b⁺CD45⁺ monocytes were previously reported to support the proliferation and colony forming ability of cultured MSCs.³⁸ Whether the monocytic cells infiltrated into the degenerating discs may exert a similar effect on local mesenchymal progenitors to promote fibrogenesis or repair is not clear. Moreover, CCR2⁺ monocytes have been identified in the degenerative IVDs from herniation patients and constitute the majority of CD11b⁺ and F4/80⁺ cells.³⁹ However, as reported in myocardial infarction model,⁴⁰ loss of CCR2 expression is required for fibrocyte differentiation of the infiltrated monocytes. This is in line with our observation that the disc fibrocytes are negative for CCR2. In a mouse model of renal fibrosis where CCR2 is depleted, migration of circulating fibrocytes to the kidney was interfered.²⁸ Whether CCR2 plays a role in monocyte infiltration and hence differentiation into disc fibrocytes await to be determined.

Fibrocytes could be derived not only from local monocyte differentiation but also migrated from peripheral blood in the form of partially differentiated collagen-producing cells.^22,28 Recruitment of monocytes and peripheral blood fibrocytes may rely on disc neo-vascularization. Normally, blood vessels recede from the outer AF and the EP in adults. However, under degenerative conditions the NP can become re-vascularized and blood vessel infiltration is frequently observed and spatially associated with macrophage infiltration.⁴¹ Most of the dNP tissues in this study for scRNA-seq analysis were dissected from extruded NP of herniated discs that normally show pronounced neo-vascularization.⁴² The increased number of GFP⁺ cells found in the AF and EP of the punctured CD11b-DTR discs may indicate the infiltration through the AF and EP routes. CD34⁺αSMA⁺ cells, presumably marking the endothelial cells and pericytes, could in fact be detected in the human dNP, supporting the presence of neo-vascularization. Studying disc vascularization and its relation to the emergence of disc fibrocytes may provide further insights into the pathway of NP fibrosis.

Trafficking fibrocytes express chemokine receptors CCR5,⁴³ CCR7²² and CXCR4.^22,44 Studies have reported increased CCR5 expression in AF and NP cells⁴⁵ and chemokines CCL2/7 in dNP.⁴⁶ Bone marrow cells, such as CD146⁺ MSCs,⁴⁷ may have a potential to migrate into the NP, possibly by control of SDF1.⁴⁸ CCL5 could interact with CCR1/3/5 in NPC and was up-regulated under inflammatory stimuli.⁴⁹ CXCR4 is a receptor of SDF1⁵⁰ and CXCL12,⁴⁴ which was found upregulated in degenerated discs and implicated in disc angiogenesis⁵⁰ and cellular apoptosis⁵¹ via PI3K/NF-κB pathway. Notably, the combination of CXCR4 and COL1 expression can identify fibrocytes in idiopathic pulmonary fibrosis.²⁹ As the disc fibrocytes highly express CXCR4 and CCR1 (Table 2), they might therefore communicate with the ProgNP population via the CXCL12-CXCR4 axis. Future study may test if CXCL12 has a function to recruit disc fibrocytes.

Tissue fibrosis is mediated by myofibroblast from local or extrinsic sources⁸ and fibrocytes may serve as one of the progenitors of myofibroblasts.^22,25 Expression of αSMA is the hallmark of mature myofibroblasts and essential to their contractility.¹⁸ Expression of αSMA in COL1A1-expressing FiboNP cells could therefore indicate their myofibroblastic property.²⁵ In addition to myofibroblasts, expression of ACTA2/αSMA, in combination with TAGLN (encoding transgelin) and MCAM (encoding CD146), can also mark pericytes. We find that the fibrocytes are not associated with the pericyte cluster, therefore the expression of αSMA found in the CD45⁺ and CD11b⁺ cells in the dNP is more likely related to myofibroblastic conversion rather than detection of pericytes. However, infiltrating pericytes can in fact contribute to scar-forming myofibroblasts in kidney and subretinal fibrosis.^8,52 Lineage tracing model can be used to delineate the contribution of pericytes to myofibroblasts in NP fibrosis in future.

FAPα is a well-known biomarker of (myo-)fibroblasts, and together with FSP1 it can label most of the COL1-producing cells in the bone.⁵³ FSP1 marks an active fibroblast population and represents a non-overlapping fibroblast entity to αSMA⁺ subtype in heart, kidney and skin fibrosis disease.⁵⁴ Our scRNA-seq analysis show that FAP⁺ or S100A4⁺ cells are not only found in the FibroNP but also in the ChonNP, although their overall abundance is not associated with degeneration severity (Fig. S13). We speculate that the ChonNP cells might be primed to become the FibroNP cells in the degeneration process. This notion is consistent with the finding of NPC transformation into a chondrogenic and later fibroblastic phenotype in a previous notochordal cell tracing model study.⁴

In this study, a vast number of αSMA⁺ cells in both human degenerative IVDs and mouse punctured discs are not positive for CD45 and CD11b. Also in the CD11b-DTR mice where the monocytes and their derived lineages (including fibrocytes) were depleted,²⁸ αSMA expressing cells could still be observed (>55%, 2 wpp in Fig. 5c) in the degenerative NP. This may be due to the existence of non-hematopoietic origins of the myofibroblast pool, such as those derived from resident NPC.⁴ Alternatively, fibrocytes may lose CD45 and CD11b expression during fibrocyte differentiation.³⁵ We also note the presence of residual CD45⁺COL1⁺ cells and GFP⁺ cells after DT administration. This likely arises from the incomplete monocyte ablation and a continuation of fibrocyte recruitment.

Fibrocyte-to-myofibroblast conversion acquires activation of TGF-β/SMAD signaling,⁵⁵ and interleukin-18 receptor 1⁵⁶ and muscarinic receptor M3⁵⁷ are shown to promote the fibrocyte formation and contractile function. Moreover, fibrocytes can secret soluble factors such as periostin⁵⁸ and TIMP1⁵⁹ and semaphorin-7A⁶⁰ to regulate myofibroblast activity and collagen expression in pulmonary and intestinal fibrosis.²¹ Class 3 semaphorins expression are found associated with disc innervation and angiogenesis.⁶¹ Interestingly, TIMP1⁶² and POSTN⁶³ are up-regulated in the degenerated discs and enriched in the FibroNP. Our CellChat data suggest that POSTN may interact with ITGB5 of disc fibrocytes. Whether disc fibrocytes regulate NP fibrosis through a paracrine manner awaits to be addressed. Furthermore, the NPC-enriched disc fibrocytes express HTRA1 and ANGPTL4. HTRA1 is a proteolytic enzyme for degrading the matrix either on its own⁶⁴ or by upregulating ADAMTS5²⁶ or matrix metalloproteinases, whereas expression of ANGPTL4 is positively associated with IDD severity.²⁷ Notably, our data suggest that ANGPTL4 is a major mediator of the intercellular communication between the disc fibrocytes and the CyclingNP, RegNP and FibroNP subpopulations. Their regulatory role in the fibrocytes and NP fibrosis worths to be investigated.

In conclusion, this study implicates a function of myeloid-derived fibrocytes in NP fibrosis, shedding new insight into the cell fate control of disc immunocytes and their modulatory effect on IDD progression. Understanding the effects and regulation of the fibrocytes may potentiate the development of new diagnostic and therapeutic strategies for modifying IDD.

Materials and methods

scRNA-seq data acquisition and downstream analysis

Four single cell RNA sequencing data sets (GSE160756, GSE165722, CNP0002664, PRJCA014236) were downloaded and processed for next-step analysis. 31 individual samples were rearranged according to Pfirrmann grade, where grade I samples were classified as non-degenerative NP, Pfirrmann grade II/III as mildly degenerated NP, and Pfirrmann grade IV/V as severely degenerated NP (Table 1). The degenerated NP samples were mainly obtained from patients with disc herniation (n = 23), with the rest from burst fracture subjects (n = 2). Seurat package (v4.3.0) (was used for comprehensive single-cell analysis with default settings, except those addressed below. This encompassed processes like quality control, normalization, scaling, dimensional reduction, unsupervised clustering, and the recognition of cluster gene markers. FastMNN (mutual nearest neighbor) algorithm implemented in the SeuraWrappers package was used to remove batch effect across samples, resulting in a correction matrix with 50 principal components (PCs). The top 40 principles were used to run the Uniform Manifold Approximation and Projection (UMAP) and cluster cells at a resolution of 0.4. Marker genes were determined with P < 0.05, min.pct = 0.25, pct.1 > pct.2, and log(fold-change) > 0.25 by the FindAllMarkers function. CellChat analysis was conducted to identify potential intercellular interactions, and communication probability values for each interaction were calculated.

Human samples

All procedures described in the present study were approved by the Institutional Ethics Review Board of The University of Hong Kong. IVD tissues were obtained from patients under informed consent. Non-degenerative and degenerative IVDs were harvested from adolescent idiopathic scoliosis (n = 7) and lumbar degenerative disc disease patients (n = 9) respectively. The age of the patients ranged from 14 to 72 years, and tissue was evaluated with the Pfirrmann classification system (Table S4).

Cell isolation and culture

NP tissues were carefully dissected away from the cartilaginous EP and AF, identified by their transparent or pale colloidal appearance. The tissues were then rinsed by saline buffer to remove blood contamination and unwanted debris. In severely degenerated IVDs (grade IV–V), where the NP tissues were fibrotic and not distinguishable from the inner AF, only the most central regions of the NP were harvested. Primary NP cells (NPC) were extracted by sequential enzyme digestion with pronase (0.25%, Roche Diagnostics) and collagenase II (600 U/mL, Worthington Biochemical),³⁰ and maintained in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum (FBS, Biosera) and 1% penicillin/streptavidin (P/S, Invitrogen) at 37 °C and 5% CO₂.

Flow cytometry

Newly extracted NPC were washed by phosphate buffered saline (PBS) and re-suspended in PBS. Cells were fixed in 4% paraformaldehyde (PFA) and blocked by 5% bovine serum albumin (BSA). Permeabilization with 0.1% triton X-100 (Sigma-Aldrich) was applied where necessary. Cells were then treated with primary and secondary antibodies, and signals from cells labeled with conjugated fluorophores were detected by using BD FACS CantoII Analyzer (BD Biosciences). The antibodies used for different flow cytometry analysis are listed in Table S5. Appropriate IgG control fluorescence compensation was applied to avoid false positive signals. Data were further analyzed by BD FACS Diva software (BD Bioscience).

Real-time quantitative PCR

Total RNA was prepared using RNeasy mini kit (Invitrogen) and cDNA synthesis was performed using High-capacity RNA-to-cDNA kit (Applied Biosystems). PCR was performed using PowerUp SYBR green master mix (Applied Biosystems), and data are presented as expression levels relative to GAPDH using the 2^-ΔΔCt method. The primer sequences were listed in Table S6.

Mouse disc puncture model

Tg (ITGAM-DTR/eGFP)34Lan/J mice which express membrane localized green fluorescent protein (GFP) with diphtheria toxin (DT) receptor under the control of CD11b promoter were used for disc puncture. The animal experiments were approved by The University of Hong Kong Committee on the Use of Live Animals in Teaching and Research (CULATR). CD11b-DTR mice (female, 12 weeks of age, n = 5 per group and time points) received needle puncture at levels of C5/6 and C7/8 of tail IVDs as described previously.⁴ At 3 days prior to the surgery and 4 weeks after the surgery, these mice were subjected to intraperitoneal injection of saline or DT (10 μg/kg body weight) (Sigma-Aldrich). At assigned end points, mice were euthanized by intravenous injection of sodium pentobarbital (1.2 g/kg), and intact discs with attached vertebral bones were harvested for histological study. Peripheral blood was harvested at day 0, 3, and 7 after DT injection, and number of GFP⁺ monocytes was measured by fluorescence-activated cell sorting (FACS).

Disc height measurement

Anterior-to-posterior radiographs of tail discs of CD11b-DTR mice or C57BL/6J were taken using a digital radiography machine (Siemens) with an exposure at 25 kV, for 5 s. The radiographs were obtained biweekly in all groups up to 8 weeks. Disc height index was calculated as previously described⁴ and presented as percent relative to the unoperated level of tail disc C6/7.

Histology and immunostaining

Samples from human and CD11b-DTR mice tissues were fixed with 4% paraformaldehyde in phosphate buffer for 24 h. Decalcification was performed for mouse IVDs using mouse solution. Paraffin-embedded tissue blocks of human NP and mice IVDs were sectioned at 7 μm. For histological staining, sections were deparaffinized and rehydrated, and subjected to FAST staining (Alcian blue, Safranin O, Fast green and Tartrazine stain) (Sigma-Aldrich) as previously described,⁶⁵ or Masson’s trichrome stain and Picrosirius red stain following standard procedures. Polarized microscopy was combined with picrosirius red staining to observe the collagens network in human NP tissue and mouse IVDs.⁶⁶ According to the histomophological changes, disc degeneration in CD11b-DTR mice was scored with reference to Tam’s scoring system.⁶⁵

For immunofluorescence, antigen retrieval was performed following rehydration by incubating with 0.8% hyaluronidase (Sigma-Aldrich) at 37 °C for 30 min and antigen retrieval buffer (100 mmol/L Tris, 5% EDTA, pH 9.5) at 95 °C for 10 min. The sections were blocked in protein block solution (DAKO) for 60 min before incubating with antibodies. Fluorophore conjugated antibodies were used to enable fluorescence detection. Information for these primary and secondary antibodies were summarized in Table S5. The sections were mounted with VECTASHIELD Mounting Medium with DAPI for nuclei staining. Isotype-matched mouse, rabbit or rat IgG was used as negative controls. Tissue sections were imaged by fluorescence microscope (Nikon Eclipse E600), and fluorescent positive cells in five fields of view (FOV) were manually counted in ImageJ with reference to total number of cells.

Data analysis

Data are presented as mean ± S.D. of four to six independent experiments, as indicated in respective figure legends. Unpaired t-test (two tailed) was used to determine significant differences. Comparisons among multiple groups were assessed using one-way ANOVA with Bonferroni post-hoc test. A P < 0.05 was considered statistically significant.

Introduction

Results

Identification of myofibroblastic NPC in IDD and NP fibrosis

Enrichment of fibroblastic NPC with hematopoietic features in IDD

Fibrocyte depletion alleviates puncture-induced NP fibrosis

Discussion

Materials and methods

scRNA-seq data acquisition and downstream analysis

Human samples

Cell isolation and culture

Flow cytometry

Real-time quantitative PCR

Mouse disc puncture model

Disc height measurement

Histology and immunostaining

Data analysis

Related Articles

Responses