A Membranome-Centered Approach Defines Novel Biomarkers for Cellular Subtypes in the Intervertebral Disc

Abstract

Objective

Lack of specific marker-sets prohibits definition and functional distinction of cellular subtypes in the intervertebral disc (IVD), such as those from the annulus fibrosus (AF) and the nucleus pulposus (NP).

Design

We recently generated immortalized cell lines from human NP and AF tissues; these comprise a set of functionally distinct clonal subtypes. Whole transcriptome analyses were performed of 12 phenotypically distinct clonal cell lines (4× NP-Responder, 4× NP-nonResponder, 2× AF-Sheet forming, and 2× AF-nonSheet forming). Data sets were filtered for membrane-associated marker genes and compared to literature.

Results

Comparison of our immortal cell lines to published primary NP, AF, and articular chondrocytes (AC) transcriptome datasets revealed preservation of AF and NP phenotypes. NP-specific membrane-associated genes were defined by comparison to AF cells in both the primary dataset (46 genes) and immortal cell-lines (161 genes). Definition of AF-specific membrane-associated genes yielded 125 primary AF cell and 92 immortal cell-line markers. Overlap between primary and immortal NP cells yielded high-confidence NP-specific marker genes for NP-R (CLDN11, TMEFF2, CA12, ANXA2, CD44) and NP-nR (EFNA1, NETO2, SLC2A1). Overlap between AF and immortal AF subtypes yielded specific markers for AF-S (COLEC12, LPAR1) and AF-nS (CHIC1).

Conclusions

The current study provides a reference platform for preclinical evaluation of novel membrane-associated cell type–specific markers in the IVD. Future research will focus on their biological relevance for IVD function in development, homeostasis, and degenerate conditions.

Introduction

Low back pain has an extensive socioeconomic impact.^1,2 Degeneration of the intervertebral disc is strongly associated with low back pain and disc herniation.³ Severe disc degeneration is estimated to occur in 10% of the general population at 50 years of age, and up to 60% in 70-year-old individuals.⁴ The intervertebral disc (IVD) degenerates more rapidly than other and surrounding musculoskeletal tissues.⁵ In particular, lumbar IVDs are susceptible to damage and injuries, underscoring important unique anatomical/mechanical features that contribute to the etiology of disc degeneration. IVD physiology is unique in that it is subject to a number of specific stressors. It is the most hypoxic tissue naturally occurring in the human body. Its nutrient supply to the nucleus pulposus (NP) and the surrounding annulus fibrosus (AF) occurs exclusively through the cartilaginous endplates (CEP; Fig. 1A and B ) and vessels surrounding the AF.^5,6 In addition, the IVD is subject to constant heavy loading, predominantly by muscle pressure.⁷ The osmolarity of the NP is higher compared to articular cartilage (AC), thereby retaining tissue-associated water to maintain disc height.⁸ The AF confines the NP to its central position within the IVD; this mediates vertical pressure absorbance within the spine. In addition, the AF endows the spinal column with a degree of flexibility.

Figure 1.

Intervertebral disc morphology and degenerative disc disease. (A) Schematic representation of an intervertebral disc. The IVD comprises 3 distinctive tissues: a central gelatinous nucleus pulposus (NP), surrounded by the lamellar annulus fibrosus (AF). The AF consists of multiple, concentrically arranged lamellae in which cells and ECM components interact to provide function and structure. In the AF, two distinct zones are distinguished primarily based on their collagen/proteoglycan content: the inner and outer AF. The NP and AF are flanked at each side by cartilaginous endplates (CEP), which mediate the contact between the IVD body and vertebra. (B) Sections of a healthy AF (15-year-old male donor) stained with Safranin O and Fast green. The outer AF predominantly contains collagens while the inner AF contains proteoglycans. The characteristic lamellar structure is clearly visible; bar represents 200 µm. (C) A nondegenerate L1/L2 intervertebral disc of 63-year-old male donor. The lamellar annulus fibrosus is clearly recognizable. The central nucleus pulposus displays no aberrant morphology; the cartilaginous endplate is relatively thin. (D) The L4/L5 intervertebral disc of the same donor. Degenerative changes cause apparent loss of lamellar structure in the AF. The NP is no longer distinguishable from the AF and additional ossification is found near the CEP (For interpretation of the references to colours in this figure legend, refer to the online version of this article).

Degenerative disc disease (DDD) is characterized by reduced proteoglycan content, loss of structural integrity, and height of the disc and CEP erosion ( Fig. 1C and D ). The CEPs are thought to be the weak link in the context of mechanical compression of the IVD. Loss of CEP integrity allows the NP to bulge into the vertebral body; this loss of structural integrity within the IVD in turn results in overloading of the AF. Consequential wear and tear may lead to AF rupture and NP protrusion (disc herniation). The exact nature of the processes that leads to irreversible disc degeneration are poorly understood.⁶ A number of cellular changes have been observed in the degenerate IVD: a decrease in cell numbers, the appearance of senescent cells, and cell cluster formation.^9,10 Concomitant production of inflammatory mediators and catabolic enzymes lead to a decrease in proteoglycan content in the NP and loss of structural integrity of the AF.^11,12 Finally, posttranslational modifications of ECM molecules^13-15 and alterations in nutrient supply have been reported to correlate with DDD.^16,17 Incomplete understanding of the underlying biology of functionally distinct cell populations in the IVD, however, hampers the identification of signaling pathways and factors important for IVD development, homeostasis, and disease.

We recently generated immortal clonal NP and AF cell lines: we identified multiple recurrent, functionally distinct cellular NP and AF phenotypes from independent young (13-15 years old) donors.^18-20 While these findings provided vital proof-of-concept that multiple relevant cellular functions can be distinguished in the NP and AF, a clear understanding of the contribution of these cells to IVD biology in vivo awaits further elucidation. Heterogeneity of cells in the NP was first described in the early 1980s.^21,22 Importantly, differences in cell phenotypes and cellular composition in current human IVD literature are likely to contribute to large variation in reported NP markers in various species (Suppl. Table S1).^23-32 In addition, variations in age, gender, or clinical stage of DDD inevitably introduce substantial variation in human datasets, which limits understanding of these markers in healthy human disc biology.^33-35 The availability of single cell–derived clonal NP and AF subpopulations provides an obvious advantage in the context of defining cellular heterogeneity in the IVD. Morphologically and functionally distinctive cell types provide an excellent starting point for gene expression profile-based, comparative phenotypic analysis. Biological specificity of cell types within diverse tissues is determined by specialized proteins, such as membrane-associated proteins, which have structural roles, determine morphology, and movement and mediate signaling from the microenvironment (e.g., cytokines, growth factors) and/or interaction with other cells and ECM components. A comprehensive analysis of molecules expressed at the cell surface of the different IVD cell types and a detailed understanding of underlying molecular signaling networks is expected to help identify specific cellular subtypes and to define cellular responses to their environment in development and disease. Definition of cancer cell membranomes have yielded novel druggable targets and biomarkers for tumor staging,^36-38 and provided important proof-of-principle that crucial advances can be made in translational research by using selective approaches to identify unique features of cells. To obtain an unbiased molecular description of our immortal NP and AF cellular phenotypes, we adapted this approach in the current study by combining unbiased comparative whole-transcriptomic analysis with a selective filtering strategy specifically tailored toward identification of novel membrane-associated markers for IVD cell subtypes. The herein outlined membranomics approach using our unique AF and NP models aims to generate a reference platform for comprehensive preclinical evaluation of novel cell type-specific markers in the IVD.

Methods

Human IVD Tissue

Human IVD tissue from young idiopathic scoliosis patients (age 8-15 years) was obtained as surplus material during scoliosis correction surgery.^18,20 Degenerate and nondegenerate lumbar IVD tissue was obtained post-mortem from a 63-year-old donor.¹⁹ Acquisition and use of all human materials was approved by the MUMC-Medical Ethical Review Committee (Approval ID 08-4-021, July 11, 2012).^18-20,26

Cell Isolation, Culturing, and RNA Extraction

The isolation of primary NP and AF cells, the immortalization procedures, and the establishment and characterization of immortal NP and AF cell lines have been described in detail elsewhere.^18,20 Briefly, 12 immortal cell clones derived from one male donor (D5, 13 years old, male spina bifida patient) were used for RNA isolation and comparative transcriptome analysis. Culturing procedures are described in detail in the Supplemental Methods section. RNA isolation, quality assessment, and real-time polymerase chain reaction (PCR) analysis were essentially carried out as published.^18-20 Procedures are described in detail in the Supplemental Methods section.

Whole Transcriptome Expression Datasets

Raw datasets from published whole transcriptome expression measurements (Affymetrix U133 Plus 2.0 array) of primary human AF (3), NP (3), and AC (3) cell donors (age range: 46-60 years; n = 9) were derived from a previous study.²⁶ Processing of RNA samples, labeling, hybridization of 12 Illimuna humanHT-12v4 arrays, data extraction, and analyses are described in detail in the Supplemental Methods section. Briefly, an arbitrary cutoff of fold change >2.0 up or down and an average expression above 100 or higher (mean fluorescence intensity) in the cell type of interest was used to define and select expressed genes. Modified t tests were used to evaluate statistically significant differences between groups (Benjamini Hochberg FDR adjusted P value <0.05 was considered significant).

Results and Discussion

High-Throughput Expression Phenotyping of Novel Immortal NP and AF Clones

We previously defined functionally distinct immortal NP and AF cell lines based on their (in)ability to respond to established chondrocyte-differentiation conditions (cf. (non)Responder NP clones: NP-R and NP-nR^18,20 and based on their (in)ability to process procollagens in vitro (non) Sheet-forming AF clones: AF-S and AF-nS²⁰). Four representative clones for both the NP-R and NP-nR cellular phenotypes were chosen for large scale expression analyses, in addition to 2 AF-S and 2 AF-nS clones (Suppl. Fig. S1A). We first evaluated expression of previously reported marker-sets for NP-R cells (FOXF1, CA12,¹⁸ NP-nR (CD24), and AF cells (SFRP2, COL12A1²⁰) prior to whole array expression analyses. This initial marker analysis confirmed that FOXF1 and CA12 discriminated between NP-R, NP-nR, and AF cell clones (Suppl. Fig. S1B). Expression levels of COL12A1 and SFRP2 were higher in AF and NP-nR cell clones compared to NP-R (Suppl. Fig. S1B). NP-nR and AF cell clones showed similar levels of SFRP2 and COL12A1 mRNA. CD24 levels, however, were significantly higher in NP-nR compared to NP-R or AF cell clones. This initial combined distinctive marker expression profiling of clonal IVD cell subtypes provided a solid basis for comparative genome-wide transcriptomic analysis.

To determine the degree of transcriptome preservation of our immortalized AF and NP cell clones, we compared our immortal array dataset with published array datasets of primary isolates, comparing 3 donor-matched AC, AF, and NP primary cell isolates.²⁶ Principal component analysis (PCA) on the normalized datasets produced a high-resolution clustering overview. The first component strongly separated the 2 datasets, which were generated on different array platforms (see Supplemental Methods section). The second and third components clearly separated AC cells from all IVD cells, in line with their divergent tissues of origin ( Fig. 2A ). Notably, primary AF and immortal clonal AF cell lines, all from unrelated donors, were presented as a distinctive cluster based on gene expression profiling. Although primary NP isolates showed considerable variation (3 independent donors), the immortal NP-R and NP-nR clones all clustered close to the primary NP isolates. Hierarchical clustering of immortal clones resulted in the anticipated separation of the PCA clusters ( Fig. 2B ; Suppl. Fig. S2). In line with their earlier observed overlap in marker expression of SFRP2 and COL12A1, NP-nR clones clustered closer to the AF clones than the NP-R group. Thus, PCA analyses indicated significantly distinctive gene expression profiles among immortal clonal IVD cellular subtypes and supported considerable preservation of primary AF and NP transcriptomic phenotype.

Figure 2.

Principle component analysis reveals overlap between primary and immortalized NP and AF clones. Transcriptome comparison between immortal cell clone and primary cell datasets generated on Affymetrix and Illumina platforms (first component), respectively. PCA analysis was preceded by ranking genes (descending expression value; see Supplemental Methods for detailed description). (A) PCA plot of the second and third components of primary AC, NP, and AF samples from 3 independent healthy donors (circles) and immortalized NP-Responder (n = 4), NP non-Responder (n = 3), and AF (n = 3) cell clones from a single donor (squares); these clones have been described in previous reports.^18,20 The first component separated the dataset on platform (primary cells: Affymetrix platform, immortalized cell lines: Illumina platform). (B) Dendrogram showing the hierarchical clustering of immortalized NP-Responder (n = 4), NP non-Responder (n = 4), and AF (n = 4) cell clones; clustering was based on Pearson correlation. The NP-R, AF, NP-nR clusters clearly differ. NP-nR clone 102 fell outside the NP-nR cluster (Suppl. Fig. S2). As this effect persisted in subsequent analyses, this clone was omitted from further analysis. Analyses were performed and figures generated using the open source scripting language R (version 2.13.0) and R packages of Bioconductor 2.8.

To obtain insight into genetic network activity underlying the preservation of immortal phenotypes in the NP, we first generated a list of putative NP, AC, and AF markers in rat, canine, and human IVDs. Rodents, in contrast to humans, retain notochordal cells (NC) throughout their life time³⁹; for this reason published NP markers in, for example, rat studies, include NC markers. These putative markers were based on whole transcriptome expression studies performed on freshly isolated cells.^23-26 We identified a total of 37 unique markers that showed differential expression between NP and AF, NP and AC, or NP and NC in one or more of these studies (Suppl. Table S1). Twenty-five positive NP markers (showing no overlap with AF- or AC-specific genes) were selected to build a literature-based NP gene-network ( Fig. 3A ; Suppl. Table S2); this network was then used to visualize clone-specific differences in gene expression between NP-R and NP-nR clones. NP-R cells showed relatively higher expression of CA12, CLEC2B, DSC2, KRT19, FOXF1, and PTN, while NP-nR cells expressed A2M, SNAP25, SOSTDC1, and VCAN ( Fig. 3A ). The differential expression of FOXF1, CA12, KRT19, and PTN is consistent with our previous qPCR-based analyses.¹⁸ A2M, CLEC2B, DSC2, SNAP25, and SOSTDC1 potentially provide additional functional definition of NP cell subtypes. The NP-nR specific Alpha 2 Macroglobullin (A2M) is a general protease inhibitor that functionally interacts with important cytokines like bFGF, PDGF, NGF, IL1β, and IL6 that are thought to be involved in DDD.^40-42 The C-type lectin domain family 2, member B (CLEC2B) prevents blood vessel formation by endothelial cells and may fulfill a similar function in the NP.⁴³ Desmocollin 2 (DSC2) is an ECM glycoprotein and component of the rat NP matrix.²³ Synaptosomal-associated protein 25 (SNAP25) is involved in intracellular membrane trafficking and has been extensively studied in the context of brain function.⁴⁴ Genes with known functions in the brain, such as axon guidance, have been reported before as NP markers.^45,46 Finally, Sclerostin domain-containing protein 1 (SOSTDC1) is a BMP antagonist. SOSTDC1 deficiency was recently shown to promote fracture healing through expansion of periosteal mesenchymal stem cells.⁴⁷

Figure 3.

Differential expression of published NP markers among NP cellular subtypes. (A) Selected marker genes (NP cell markers; Suppl. Tables S1 and S2) were used as input for GeneMANIA network building using Cytoscape visualization software. Gene expression differences between NP-R (red) and NP-nR (blue) clones are depicted as log2-based fold change (FC) in colored circles. Colored lines in the network represent established biological connections between markers; parentheses: percentage representation. The network is strongly interconnected based on co-expression and genetic interactions studies. Of note: established NP markers are not equally expressed by isolated cellular subpopulations derived from the NP. (B) Progenitor-associated gene expression in NP-Responder clones. Selected genes (NP progenitor markers; Table 1 , Suppl. Table S2) were used as input for network building using GeneMANIA and Cytoscape visualization software. Gene expression differences between NP-R (red) and NP-nR (blue) clones are depicted as log2-based FC in colored circles. Colored lines in the network represent established biological connections between markers; parentheses: percentage representation (For interpretation of the references to colours in this figure legend, refer to the online version of this article).

To visualize relative gene activity in this set of 25 markers in primary NP cells, the published transcriptome dataset²⁶ was plotted within the same gene network. This approach revealed that NP markers KRT18, FoxF1, and VCAN were relatively highly expressed in primary NP cells (Suppl. Fig. S3A). Comparative gene network analysis thus reveals a number of striking differences in activity profiles of previously proposed general NP markers between immortal NP cellular subtypes and primary isolates, and emphasizes the importance of functional definition of cellular heterogeneity in the NP.

Progenitor Cells in the Mature NP

Progenitor or stem cells in the mature IVD were first isolated and studied in 1995 and have since been studied in both NP and AF of various species.^48-54 Progenitor cells were first isolated from the IVD by fluorescence-activated cell sorting (FACS) using a discriminatory candidate cell surface marker-based approach.⁵⁵ NP progenitor cells were isolated by FACS from mouse, human, and recently cow intervertebral discs using a combination of membrane-associated Cluster of Differentiation 24 (CD24), Ganglioside G2 (GD2), and/or receptor tyrosine kinase, epithelial specific (Tie2) cell surface markers.^56,57 A summary of these reported findings shows little overlap in the markers deployed to identify progenitor cells in the mature IVD ( Table 1 ).

Table 1.

Overview of Literature Reports on Cell Heterogeneity and Progenitor Populations in the Intervertebral Disc.

Tissue Origin (Ref)	Reference	Isolation method	Markers
Humana; NP, AF	54	Flow cytometry	CD105, CD166, CD63, CD90, CD73 (NT5E), P75, CD133
Human; NP	48	IHC	HSP27, HSP72
Lapine, ovine, humana; NP	49	IHC	Notch1, Jag1, Delta4, Stro1, C-kit (CD117), CD105, CD166
Human; AF	50	Flow cytometry	CD24, CD49, CD51, NSE
Human; CEP	52	CFU assay; agarose	CD133, CD105, CD90, CD73 (NT5E), Stro1, CD166, CD44
Canine; NP	53	CFU assay	Sox2, Oct3/4, Nanog, CD133 (mRNA only)
Murine, human; NP	55	CFU assay; agarose	CD24, Tie2, GD2, CD44, (CD49f, CD56, CD73 (NT5E), CD90, CD105, CD166)b; (CD271, Flt1)c
Lapine; AF	51	CFU assay; fibroblast	CD29, CD44, CD166, Oct-4, Gnl3, SSEA-4

Various markers have been used to isolate multipotent cell populations from IVD tissues (as indicated in the corresponding column). Typical MSC markers (CD73, CD90, and CD105) are often used in addition to CD24 and CD44. We measured gene and protein expression of genes printed in bold in immortal cell clones of a single donor (NP and AF).

^aDegenerate tissue.

^bExpressed on all mouse and human NP cells.

^cCorrelated to progenitor phenotype.

In our immortal clonal NP cell lines, distinctive gene signatures were identified that correlated with morphological characteristics: wave-like NP-Responder (NP-R) clones expressed high levels of FoxF1 and CA12, whereas cobble stone-shaped NP-non-Responder (NP-nR) clones expressed high levels of membrane-associated CD24.¹⁸ NP-R clones exhibited spontaneous spheroid-forming capacity on specific culture substrates (Aggrecan [ACAN], Matrigel). This stemness-associated feature has previously been utilized to identify progenitor cells in multiple systems.^58,59 Hence, we hypothesized that the immortal NP-R clones harbor a progenitor-like phenotype; in contrast, our analyses suggested NP-nR clones maintain a more differentiated phenotype.^18-20

To obtain additional support for this hypothesis, we focused on the “cell heterogeneity and progenitor” genes listed in Table 1 and used these to generate a progenitor gene-network representing the NP progenitor cell phenotype. This network was then used to visualize relative expression differences between NP-R and NP-nR clones ( Fig. 3B ). Significantly higher expression of ANGPT1, CD44, CD63, FLT1 (VEGFR1), KIT, NT5E (CD73), and TEK was observed in NP-R clones. FLT1 (VEGFR1) and ANGPT1 are known as important signaling molecules mediating survival of NP cells.⁵⁶ In the same study TEK (Tie-2) was identified as an NP progenitor marker. The expression of these markers in NP-R clones strongly supports the progenitor-like nature of this cell type.

NP-nR clones were characterized by increased CD24, JAG1, PROM1 (CD133), and THY1 (CD90) expression. Primary NP cell isolates (mixed cell population) had high expression of CD24, ENO2, ITGB1 (CD29), POU5F1 (OCT3/4), and VEGFA, whereas TEK1 and ANGPT1 were expressed at relatively low levels (Suppl. Fig. S3B). Of note, the expression level of progenitor cell marker genes in the primary NP cell dataset was expected to be low, as progenitor cells are not likely to represent a predominant phenotype in the mature human IVD.

Although we evaluated a significant proportion of these cell markers, it is currently not clear how the 2 distinctive cellular NP phenotypes, NP-R and NP-nR, can be positioned on an imaginary differentiation scheme that was proposed based on membrane expression of CD24, CD44, Tie2, and GD2 ( Table 2 ).⁵⁶ Our findings fit this model to certain extent: our progenitor cells (NP-R) are indeed CD24 negative (CD24−), whereas the more differentiated NP-nR cells are CD24+. None of our immortal clones were Tie2+ however, which was proposed as a hallmark of NP progenitor cells. Although NP-R clones were initially negative for GD2, GD2-expression could be induced using chondrogenic differentiation stimuli.¹⁸ In summary, NP-R clones displayed a gene signature compatible with NP progenitor characteristics.

Table 2.

Positioning of Immortal NP-Phenotypes in NP-Based Progenitor/Differentiation Model.

Marker	Progenitor		Differentiation			Immortal Clones
	Dormant	Activated	1	2	3	NP-R	NP-nR
Tie2	+	+	−	−	−	−	−
GD2	−	+	+	+	−	−	+
CD24	−	−	−	+	+	−	+
CD44	+/−	+	+	+	+	+	+
CD271	+	+	+/−	−	−	ND	ND
FLT1	+	+	+/−	−	−	ND	ND

A recently postulated progenitor/differentiation model based on expression (+) or absence (−) of markers⁵⁵ is represented in the columns marked Progenitor and Differentiation. The right most columns show the measurements we performed on our immortal NP cell clones. ND = analysis not done.

Novel Cell Surface Markers in the NP

A previous study identified membrane-associated genes specific for the NP compared with AC, by using an in silico screening of differentially expressed genes for the presence of transmembrane domains.²⁶ To test whether these results could be reproduced in an independent study, we compared 90 previously identified membrane-associated genes²⁷ with primary NP/AC dataset²⁶ ( Fig. 4A ). Thirty-four overlapping NP membrane-associated marker genes (NP/AC) were identified between these 2 studies; the expression of these markers was at least 2-fold higher in primary NP cells in comparison to AC ( Table 3 ). CA12 remained the top hit (3 probes in the top 10; relative expression level: >20-fold). In addition expression of ABCG1, SLC27A2, TMEM71, FLRT2, CLDN11, and DNER was more than 10-fold higher in primary NP isolates. As these genes may define functional differences between NP cells and ACs, their potential involvement in IVD biology is briefly discussed below. ATP binding cassette subfamily G member 1 (ABCG1) is expressed in articular cartilage and downregulated in osteoarthritis and has a regulatory function in cholesterol and phospholipid transport.⁶⁰ Solute carrier family 27 member2 (SLC27A2) is a fatty acid transport protein. TMEM71 function is unknown; it is regulated by the candidate tumor suppressor gene inhibitor of growth 2 (ING2).⁶¹ Fibronectin-like domain-containing leucine-rich transmembrane protein (FLRT2) interacts with fibronectin and has distinct expression patterns in mouse development that are linked to FGF signaling.^62,63 Claudin 11 (CLDN11) is a tight junction protein originally identified in oligodendrocytes and is known to be expressed in a small cellular subpopulation in the human IVD.^64,65 Delta- and notch-like epidermal growth factor-related receptor (DNER) plays an important role in brain development.⁶⁶ Taken together, 34 membrane-associated marker genes that show a relatively higher transcriptional activity in NP compared to AC were identified in 2 independent datasets, further emphasizing the unique phenotype of NP cells. With the exception of CA12 these markers have not been studied in the IVD.

Figure 4.

Identification of membrane-associated NP marker genes. Comparison strategy of whole transcriptome datasets of published primary IVD cellular isolates and clonal cell lines. (A) Marker genes significantly more highly expressed in primary NP compared to AC from 2 independent published datasets^26,27 were compared. (B) Definition of primary NP-specific membrane marker genes compared to AF and AC²⁶ and definition of NP cell clone (R/nR) specific marker genes.¹⁸ (C) Subsequently both primary and clonal NP membrane marker gene sets were compared to each other. (D) Definition of primary AF-specific membrane marker genes compared to NP and definition of AF cell clone (S/nS) specific marker genes. (E) Subsequent comparison of primary and clonal AF membrane marker genes.

Table 3.

Common Human NP and AC Markers.

Gene	IDa	AC Meanb	NP Meanb	FCc
ABCG1	9619	20.3	739.7	36.4
AGPAT9	84803	675.2	2611.6	3.9
C1orf9	51430	1166	3352.1	2.9
CA12	771	71.3	2726.9	38.2
CDH19	28513	761.7	2760.6	3.6
CDS1	1040	76.9	249.5	3.2
CLDN11	5010	19.3	204.2	10.6
CLEC2B	9976	53.4	303.3	5.7
DNER	92737	365.6	3671.4	10
FAM62B	57488	78.5	196.5	2.5
FKBP11	51303	1102.8	2438.5	2.2
FLRT2	23768	21.1	240.5	11.4
FNDC3B	64778	385.9	1598.9	4.1
HS6ST3	266722	54.2	494.7	9.1
INSIG1	3638	885.9	2010.5	2.3
JPH3	57338	50.3	252.9	5
KCNS3	3790	61.2	175.8	2.9
MFAP3L	9848	49.2	128.8	2.6
MXRA7	439921	238.6	844.3	3.5
NETO2	81831	49.3	323	6.6
ODZ3	55714	36.3	217.6	6
ORAI3	93129	431.5	2239.2	5.2
PPAPDC1B	84513	774.6	2823.4	3.6
REEP1	65055	42.2	129.8	3.1
SEL1L	6400	190.6	1166.5	6.1
SGCG	6445	24.3	75.5	3.1
SLC27A2	11001	19.2	350	18.3
SLC44A1	23446	86.6	477.2	5.5
SLC7A1	6541	99.9	237	2.4
TMEM27	57393	19.3	44.2	2.3
TMEM71	137835	29.5	336.6	11.4
TSPAN31	6302	574	1177.6	2.1
WFS1	7466	484.1	1093.4	2.3
WNT11	7481	179.9	406.2	2.3

A comparison between 90 transmembrane cell surface markers²⁷ and primary tissue-specific markers²⁶ extracted from human micro-array data (NP vs. AC).

^aENTREZ Gene ID.

^bAdjusted P value <0.05, fold change >2.0, and expression in NP >100 arbitrary units; see Methods section for details.

^cFold change (NP/AC).

Comparative Membranome Analysis of primary NP, AF, and AC Isolates

To identify additional NP- and AF-specific cell surface markers, we first analyzed the primary NP, AF, and AC cell datasets²⁶ and filtered for differentially expressed genes (FC > 2.0; adjusted P value <0.05; see Methods section for additional expression criteria). This yielded 898 genes of which expression was significantly higher in NP versus AF and 226 genes that were more highly expressed in NP versus AC. From these 2 gene lists 143 NP-specific markers (significantly higher in NP compared to AF and AC) were extracted. Using published meta-analyses,³⁶ 6134 membrane-associated proteins were defined. Of the 143 NP-specific markers, 46 were found to be cell membrane-associated ( Fig. 4B and C ). This marker set definition in NP and AF cell-lines showed overlap with our previous analyses in published primary NP cells and AC datasets (compare Table 3 with Table 4 ): CA12, CLDN11, Heparan Sulfate 6-0-sulfotransferase 3 (HS6ST3), Neuropilin- and tolloid-like 2 (NETO2), Calcium Release-Activated Calcium Modulator 3 (ORAI3, TMEM142C), and Choline transporter-like protein 1 (SLC44A1) ( Table 4 ; bold print). The latter 4 genes have thus far not been studied in the musculoskeletal system and appear highly specific for the NP in 2 independent human IVD cell studies.

Table 4.

NP-Specific Membrane-Associated Markers Compared to AF and AC in Primary Cell Expression Datasets.

Gene	IDa	AC Meanb	AF Meanb	NP Meanb	FCc	FCd
ANKRD10	55608	687.4	909.1	2267.0	3.3	2.5
ANXA2	302	18.5	23.5	380.8	20.6	16.2
ATP6V0E1	8992	35.4	37.7	124.4	3.5	3.3
C11orf1	64776	50.3	48.5	178.3	3.5	3.7
C9orf91	203197	68.0	75.3	276.3	4.1	3.7
CA12	771	71.3	134.4	2726.9	38.2	20.3
CLDN11	5010	19.3	47.2	204.2	10.6	4.3
CLIP4	79745	98.1	63.9	207.2	2.1	3.2
DKFZp451A211	400169	505.2	989.2	2828.0	5.6	2.9
DUSP6	1848	533.6	680.5	2172.2	4.1	3.2
DYRK2	8445	238.8	247.3	746.3	3.1	3.0
EFNA1	1942	909.3	2144.1	6752.4	7.4	3.1
F3	2152	68.8	55.4	263.1	3.8	4.8
FAM134A	79137	173.5	188.9	408.8	2.4	2.2
FGFR2	2263	25.9	26.5	109.7	4.2	4.1
FLRT1	23769	79.5	90.6	243.3	3.1	2.7
FUS	2521	43.2	41.3	105.2	2.4	2.5
GPR153	387509	319.2	310.1	686.7	2.2	2.2
GRB10	2887	179.9	154.4	651.0	3.6	4.2
HS6ST3	266722	54.2	183.9	494.7	9.1	2.7
KHDRBS3	10656	164.8	289.8	1349.9	8.2	4.7
KIAA1217	56243	39.9	184.3	529.1	13.2	2.9
LDLR	3949	140.3	148.1	339.9	2.4	2.3
LRRFIP1	9208	398.6	499.9	1150.3	2.9	2.3
NCAM1	4684	49.2	43.7	116.8	2.4	2.7
NETO2	81831	27.7	48.6	122.4	4.4	2.5
NNMT	4837	89.5	52.7	263.0	2.9	5.0
ORAI3	93129	431.5	1058.3	2239.2	5.2	2.1
PGK1	5230	39.1	54.9	471.6	12.1	8.6
PSMB4	5692	1084.0	800.1	2256.8	2.1	2.8
RP1-21O18.1	23254	198.8	153.9	422.4	2.1	2.7
RPL38	6169	1114.4	863.3	6333.9	5.7	7.3
RPLP2	6181	234.9	194.1	739.6	3.1	3.8
RRAD	6236	234.2	247.8	745.4	3.2	3.0
SLC16A3	9123	33.5	42.1	169.9	5.1	4.0
SLC2A1	6513	353.9	851.8	2506.6	7.1	2.9
SLC2A3	6515	410.9	437.5	1330.5	3.2	3.0
SLC44A1	23446	86.6	208.1	477.2	5.5	2.3
SLC6A6	6533	264.9	185.9	569.8	2.2	3.1
STX3	6809	356.4	327.7	746.2	2.1	2.3
TMEFF2	23671	271.5	321.8	2501.6	9.2	7.8
TMEM200A	114801	61.7	53.1	183.0	3.0	3.4
TNS1	7145	2142.0	3537.6	7840.3	3.7	2.2
TRPM8	79054	31.6	65.0	269.2	8.5	4.1
ZNF395	55893	1956.3	4013.4	9607.8	4.9	2.4
ZNF791	163049	711.1	1397.7	3029.9	4.3	2.2

^aENTREZ Gene ID.

^bDifferentially expressed genes were determined by adjusted P value <0.05, fold change >2.0, and expression in NP >100 (arbitrary units; see Methods section for details).²⁶ Marker genes overlapping with those in Table 4 are printed in bold.

^cFold change (NP/AC).

^dFold change (NP/AF).

NP Clone-Specific Membrane-Associated Markers

In parallel, we identified unique membrane-associated markers for the NP clonal subtypes. The immortal cell clone dataset was filtered using a published membranome dataset.³⁶ Of 161 differentially expressed membrane-associated genes, 79 were more highly expressed in the NP-R and 82 in the NP-nR ( Fig. 4B ; Suppl. Tables S3 and S4). These genes and their possible relevance for the NP cell phenotype are discussed below.

NP Responder-Specific Membrane-Associated Markers

Among NP-R-specific membrane markers, 3 genes showed a larger than 10-fold change: (1) plasticity related gene 1 (LPPR4), (2) transient receptor potential cation channel, subfamily V, member 2 (TRPV2), and (3) protocadherin 10 (PCDH10). LPPR4 is known for its role in brain plasticity⁶⁷ and has not been studied in the context of the IVD or articular cartilage. TRPV2 is known to be expressed in chondrogenic high-density cultures of limb bud cells in chickens and mice and was recently associated with genetic skeletal disorders.^68,69 PCDH10 is a negative regulator of the WNT pathway.⁷⁰ Platelet-Derived Growth Factor Receptor Alpha (PDGFRA) has a fold change of 8.2 (NP-R/NP-nR) and is required for the formation of chondrogenic mesoderm and interacts with CAV1, which inhibits its signaling.^71,72 Interestingly, the NP-marker PAX1 is a transcriptional regulator of PDGFRA.⁷³ The PDGFRA ligand PDGF partakes in a complex autocrine feedback with its receptor and modulates the effect of TGFβ-stimulation.⁷⁴ In addition, the PDGFR ligand PDGF-BB was recently shown to delay disc degeneration in a rabbit model.⁷⁵ In contrast, PDGF was shown to potentiate the chondrocyte’s response to IL1β in rheumatoid arthritis.⁷⁶ Thus, differences in a cell’s ability to activate the PAX1-PDGFRA-PDGF signaling axis may be relevant to local inflammatory responses and contribute to disc degeneration. CLDN11 (fold change 6.9 NP-R/nR) expression was recently inversely correlated with SDC2 expression.⁷⁷ Of note: NP-nR clones showed a higher expression of SDC2 than CLDN11 expressing NP-R clones. RGM (repulsive guidance molecule) domain family member B (RGMB) was restricted to NP-R clones (fold change 5.5). RGMa and RGMb are both highly expressed in the developing chick notochord.⁷⁸ RGMb functions as a co-receptor for BMP and acts as an antagonist of noggin, which in turn antagonizes BMP-signaling and has an important role in the embryonic development of the NP.⁷⁹ In the NP, RGM expression is thought to prevent nerve in-growth in the IVD.^45,80 As qPCR analysis independently confirmed the micro-array-based differences (Suppl. Fig. 4A), these markers are likely to have potential with regards to progenitor isolation from the NP.

NP nonResponder-Specific Membrane-Associated Markers

Among the 82 NP-nR specific membrane genes, protein tyrosine phosphatase, receptor type, F (PTPRF), tumor necrosis factor (ligand) superfamily, member 7 (CD70), chromosome 13 open reading frame 13 (SHISA2), G protein-coupled receptor, family C, group 5, member C (GPRC5C), chemokine orphan receptor 1 (CXCR7), and prostaglandin F2 receptor negative regulator (PTGFRN) all have more than 10-fold higher expression in NP-nR clones than in NP-R clones. Strikingly, multiple genes could be traced to the WNT signaling pathway. PTPRF shares an extracellular domain with NCAM and controls beta-catenin signaling.⁸¹ NCAM was previously suggested as a NP marker.^23,30 SHISA2 modulates both WNT and FGF signaling in the developing chicken embryo⁸²; in Xenopus it promotes somatic maturation of precursor cells.⁸³ Low-density lipoprotein Receptor-related Protein, 5 (LRP5; fold change 5) is part of the canonical WNT signaling. Induced expression of LRP5 mediates cartilage destruction in ostheoarthritis.^84,85 PTGFRN is involved in cell migration by selective interaction with Collagen type I and Fibronectin. This effect is modulated by the expression of the tetraspanins CD9 and CD81.⁸⁶ Differential expression of these genes in NP-nR (vs. NP-R and AF cells) was confirmed by qPCR (Suppl. Fig. 4B).

Markers for NP Cellular Heterogeneity In Vivo

To begin to verify the potential usefulness of the clone-specific marker sets to distinguish between NP cell subtypes, we lastly compared the set of 46 primary NP cell-specific membrane-associated genes to NP-R and NP-nR-specific membrane-associated markers ( Fig. 4C ). A subset of 8 genes was identified both as primary NP-specific cell surface markers and as being differentially expressed between immortalized NP clonal cell types. NP-R cells were marked by CLDN11, TMEFF2, CA12, ANXA2, CD44 expression; NP-nR by EFNA1, NETO2, and SLC2A1 ( Table 5 ). TransMembrane protein with EGF-like and two Follistatin-like domains 2 (TMEFF2) regulates non-canonical BMP4/activin signaling and is involved in PI3K and RAS/ERK1/2 signaling.⁸⁷ Annexin A2 (ANXA2) regulates membrane organization and trafficking including endo- and exocytosis.⁸⁸ CD44 is the receptor for hyaluronan, known to be expressed by notochordal and mature NP cells.^56,89 Finally, Solute carrier family 2 member 1 (SLC2A1; also known as GLUT1) is a known NP marker⁹⁰ and was expressed at substantially higher levels in NP-nR clones. Importantly, expression of SLC2A1 is reported to be elevated in the context of human disc degeneration.⁹¹

Table 5.

Differentially Expressed NP-Specific Membrane-Associated Markers in NP-R and NP-nR Cell Clones.

Gene	IDa	FCb	P c	Gene Description
CLDN11	5010	6.90	0.011	Claudin 11 (oligodendrocyte transmembrane protein)
TMEFF2	23671	5.46	0.005	Transmembrane protein with EGF-like and 2 follistatin-like domains 2
CA12	771	2.72	0.025	Carbonic anhydrase XII
ANXA2	302	2.51	0.009	Annexin A2
CD44	960	2.16	0.042	CD44 antigen (homing function and Indian blood group system)
EFNA1	1942	−3.24	0.003	Ephrin-A1
NETO2	81831	−3.64	0.002	Neuropilin (NRP) and tolloid (TLL)-like 2
SLC2A1	6513	−4.55	0.022	Solute carrier family 2 (facilitated glucose transporter), member 1

^aENTREZ Gene ID.

^bFold change (NP-R/NP-nR).

^cAdjusted P value <0.05, fold change >2.0, and expression in NP >100 (arbitrary units; see Methods section for de tails).

Cellular Heterogeneity in the AF

Cellular heterogeneity in the inner (iAF) and outer AF (oAF) has been suggested based on morphological differences: oAF cells have an elongated appearance while iAF cells are more rounded and showed morphological similarities with NP cells. A recent study demonstrated differences in ECM expression after prolonged culturing of iAF and oAF cells on a polycarbonate urethane substrate: oAF cells expressed higher amounts of COL1A1, while iAF cells expressed more COL2A1, ACAN, and VCAN.⁹¹ COMP and Glypican 3 (GPC3) have been reported, although not consistently, as AF cell markers in vivo (Suppl. Table S1). GPC3 is a heparan sulfate proteoglycan involved in WNT, BMP and FGF signaling. Relevantly, GPC3 is an important inhibitor of SHH signaling⁹²; as SHH is expressed in the embryonic notochord, GPC3 might be an important regulator of NP/AF boundary demarcation in IVD development. Secreted Frizzled Related Protein 2 (SFRP2) is an inhibitor of WNT signaling and was identified as an AF marker in primary bovine IVD cells,²⁵ which we validated as an mRNA marker of human AF cells in vitro.²⁰ Finally, differential expression of mitochondrial genes was suggested to differentiate between senescent and non-senescent cells in the AF.^93,94 Overall, this limited gene list stresses the need for identification of additional AF-specific markers ( Table 6 ).

Table 6.

Expression of AF-Specific Membrane-Associated Markers Genes Compared to NP.

Gene	Gene IDa	AF Meanb	NP Meanb	FCc
SLC40A1	30061	1241.8	148.2	8.4
RNFT1	51136	271.2	38.0	7.1
C18orf55	29090	369.5	66.6	5.5
COLEC12	81035	257.7	47.9	5.4
DAD1	1603	3141.1	584.2	5.4
LPAR1	1902	1546.4	294.7	5.2
IFITM2	10581	1411.7	297.8	4.7
ALG10B	144245	307.2	68.5	4.5
IFITM1	8519	995.7	222.5	4.5
THBD	7056	388.0	88.4	4.4
HEATR5A	25938	286.0	66.2	4.3
LRP6	4040	704.8	173.6	4.1
SLC38A9	153129	249.9	62.3	4.0
ATP11C	286410	397.1	100.6	3.9
FLRT2	23768	688.3	175.1	3.9
GPR125	166647	374.9	96.5	3.9
SLC30A5	64924	115.6	30.1	3.8
GPR137B	7107	232.0	61.0	3.8
TMEM133	83935	331.6	88.7	3.7
CD55	1604	3992.5	1074.7	3.7
EFHA2	286097	149.8	41.6	3.6
LRIG1	26018	1224.4	340.5	3.6
SLC35A5	55032	486.7	135.8	3.6
ANKRD46	157567	778.0	220.9	3.5
CD58	965	237.3	68.8	3.4
TMEM27	57393	151.3	44.2	3.4
EMP3	2014	868.2	254.0	3.4
CD59	966	333.2	97.9	3.4
FAM69A	388650	348.7	102.9	3.4
TACSTD2	4070	2031.0	603.1	3.4
C5orf28	64417	195.1	58.0	3.4
SCNN1A	6337	513.8	153.3	3.4
SLC10A7	84068	388.9	116.1	3.4
LRIG2	9860	524.8	158.6	3.3
SMC3	9126	957.8	296.4	3.2
ATP2C1	27032	122.2	38.4	3.2
FLJ20160	54842	236.1	76.0	3.1
LIFR	3977	447.5	144.3	3.1
FLJ11171	55783	216.9	72.3	3.0
OMA1	115209	327.8	111.3	2.9
C9orf123	90871	212.5	72.9	2.9
ITM2C	81618	1483.2	514.1	2.9
ABHD2	11057	854.9	297.1	2.9
EMP2	2013	999.4	350.1	2.9
CD99	4267	606.0	213.8	2.8
IL13RA1	3597	169.0	59.7	2.8
C20orf199	441951	1288.8	466.2	2.8
LOC440993	440993	276.4	100.4	2.8
ADAM9	8754	229.3	85.2	2.7
MGC24039	160518	131.7	49.0	2.7
TNFSF10	8743	290.6	109.2	2.7
PCNX	22990	168.4	63.7	2.6
TMEM200B	399474	300.5	113.7	2.6
ZMYM6	9204	220.5	83.9	2.6
SEPSECS	51091	131.7	50.6	2.6
FAT4	79633	213.8	82.8	2.6
LPAR4	2846	212.2	82.3	2.6
LOC493869	493869	206.8	80.9	2.6
AGTRAP	57085	136.9	54.0	2.5
CHIC1	53344	385.5	152.1	2.5
CLCN3	1182	174.6	69.1	2.5
STX7	8417	194.0	77.0	2.5
SLC39A7	7922	782.0	312.4	2.5
ACVR1	90	3337.6	1337.0	2.5
DISP1	84976	178.9	73.0	2.4
REEP1	65055	316.9	129.8	2.4
SLC35B3	51000	1395.7	572.9	2.4
SLC39A14	23516	344.9	142.4	2.4
ROBO1	6091	606.3	251.1	2.4
SLC41A2	84102	272.7	113.4	2.4
FDFT1	2222	593.3	247.6	2.4
MIA3	375056	362.0	151.4	2.4
HIGD1A	25994	2291.2	960.2	2.4
ITGB1	3688	221.4	92.9	2.4
SLC33A1	9197	1725.7	731.3	2.4
UPK1B	7348	160.2	68.0	2.4
SLC16A3	9123	183.0	77.6	2.4
FAS	355	192.3	82.1	2.3
C20orf30	29058	217.5	92.9	2.3
SLC30A6	55676	365.7	156.2	2.3
PTPLAD2	401494	108.2	46.3	2.3
LRP12	29967	129.7	55.8	2.3
STEAP2	261729	312.0	134.9	2.3
FAM73A	374986	365.2	158.3	2.3
LAG3	3902	330.2	144.4	2.3
KCNA1	3736	167.5	73.6	2.3
ULBP2	80328	184.1	81.2	2.3
ATL3	25923	1223.0	540.4	2.3
ATP11B	23200	236.3	104.9	2.3
PROCR	10544	568.0	252.3	2.3
C7orf44	55744	237.1	105.6	2.2
ASAM	79827	195.6	87.3	2.2
C14orf101	54916	230.7	103.0	2.2
TSPAN4	7106	115.5	51.7	2.2
SLC31A2	1318	545.7	246.3	2.2
SACM1L	22908	1277.7	577.6	2.2
FZD2	2535	197.5	89.6	2.2
TM2D2	83877	1033.3	469.6	2.2
USMG5	84833	3457.8	1573.8	2.2
MFSD8	256471	161.8	73.7	2.2
JAM2	58494	318.1	144.9	2.2
ARMCX5	64860	273.9	125.2	2.2
AXL	558	1232.8	566.4	2.2
STX17	55014	421.2	194.1	2.2
LNPEP	4012	215.2	99.5	2.2
DLEU2	8847	406.6	188.7	2.2
LOC219854	219854	395.2	183.6	2.2
C9orf105	401505	649.6	303.2	2.1
GHITM	27069	649.2	303.7	2.1
SLC16A7	9194	237.6	112.6	2.1
TMEM154	201799	172.1	81.9	2.1
EGFR	1956	253.3	121.0	2.1
TMEM19	55266	113.9	55.0	2.1
CAPRIN1	4076	308.4	149.0	2.1
SEL1L	6400	889.5	433.1	2.1
IFNAR1	3454	197.7	96.8	2.0
SPA17	53340	125.9	61.8	2.0
GLIPR1	11010	850.1	417.6	2.0
CNIH4	29097	2894.7	1424.1	2.0
KCNK1	3775	582.0	287.8	2.0
LOC203547	203547	1509.4	747.3	2.0
IL17RD	54756	149.7	74.3	2.0
INTS7	25896	102.8	51.2	2.0
DPY19L4	286148	283.4	141.6	2.0
SLC13A4	26266	1083.4	541.6	2.0

^aENTREZ Gene ID.

^bAdjusted P value <0.05, fold change >2.0, and expression: >100 (arbitrary units; see Methods section for details).

^cFold change (AF/NP).

Novel AF Subclone-Specific Membrane-Associated Markers

To identify AF-specific membrane markers, we first aimed to define AF-specific genes by comparing primary cell isolates from AF and NP²⁶; this analysis yielded 1161 genes that showed a significantly higher expression in AF than in NP. Application of the membranome-filter to this gene collection produced a subset of 125 AF-specific genes encoding membrane-associated proteins ( Fig. 4D , Table 7 ). We have limited discussing the potential relevance of these markers in regard to AF cell type–specific biology to a selection of marker genes. Of interest, we identified 2 receptors from the canonical WNT pathway (LRP4 and FZD2). LRP4 was recently shown to facilitate chondrogenic differentiation and articular cartilage homeostasis.^95,96 In addition, expression of the ligand WNT16 was significantly higher in primary AF. WNT16 was shown to repress osteoclastogenesis and to protect articular cartilage in a murine model for osteoarthritis.^97,98 Combined with our earlier observation that the SFRP2 is an AF marker in vitro, these results suggest that regulation of WNT signaling separates AF cells from NP cells. Of note, none of these factors have thus far been investigated in the IVD. WNT signaling is known to be involved in IVD development, decreases in the adult mouse IVD and is reactivated in old IVDs.

Table 7.

Differentially Expressed AF-Specific Membrane-Associated Markers in AF-S and AF-nS Cell Clones.

Gene	Gene IDa	FCb	P c	Gene Description	FCd
CHIC1	53344	2.3	0.0017	Cysteine-rich hydrophobic domain 1	2.5
COLEC12	81035	−2.3	0.0432	collectin sub-family member 12	5.4
LPAR1	1902	−2.0	0.0232	lysophosphatidic acid receptor 1	5.2

^aENTREZ Gene ID.

^bFold change (AF2/AF1).

^cAdjusted P value <0.05, fold change >2.0, and expression in AF >100 (arbitrary units; see Methods section for details).

^dFold change (AF/NP).

To expose AF clone-specific markers, we next compared the 2 AF clone subtypes that were distinguished on their (in)ability to process procollagens in vitro: Sheet-forming (AF-S) and nonSheet-forming (AF-nS) AF clones.²⁰ A total of 120 differentially expressed genes remained after membranome-filtering ( Fig. 4D , Suppl. Tables S5 and S6). CLDN11, identified as an NP-R marker gene (cf. Suppl. Fig. S2), has also been implicated as a polarity gene in human AF cells in situ.⁶⁴ CLDN11 was preferentially expressed in AF-S clones. Of interest, AF-nS clones showed conspicuous Notch activity, as NOTCH3, JAG1, and downstream target gene HES1 were more highly expressed in AF-nS cells compared to AF-S cells (Suppl. Fig. S5C). Notch/Jagged/Hes1 activity was shown to be differentially regulated in NP an AF in response to hypoxia.⁹⁹ In addition, polymorphisms in the NOTCH3 gene have been associated with lumbar disc herniation and Notch signaling has been implicated in degenerative disc disease.^100,101 The exact biological implication of increased Notch-signaling in AF cells is currently unknown. It has been suggested that Notch is important for maintaining proliferative capacity of disc cells.⁹⁹ Our data add to these observations and for the first time identifies a cellular AF sub-phenotype in which Notch signaling has a distinctive function.

Last, to determine the validity of these clone-specific markers, we compared primary AF cell specific genes to AF subtype-specific markers ( Fig. 4E ). This identified 3 genes that were expressed both in primary as well as immortal monoclonal AF cell extracts: Cysteine-rich hydrophobic domain 1 (CHIC1), Collectin sub-family member 12 (COLEC12), and Lysophosphatidic acid receptor 1 (LPAR1; Table 7 ). CHIC1 (also known as AKAP13) is an anchoring protein involved in Protein Kinase A signaling. In addition to its involvement in cardiac hypertrophy, it is known to regulate osmotic responses and integration of inflammatory signals in T lymphocytes.^102,103 Little is known about COLEC12 in the context of IVD biology. LPAR1 was shown to play a role in synovial fibroblast activation during arthritis and plays a role in osteoclastogenesis.^104,105 Taken together, our analyses identify a substantial set of novel AF membrane-associated markers. Subsets among these markers, such as WNT receptors LRP4, LRP5 (NP-nR and AF marker), FZD2, and additional genes described above, discriminate between our previously established immortal IVD cell types and may prove useful to functionally define cellular heterogeneity in the AF.

Summary and Perspectives

In this study, we present a meta-analysis of molecular markers in relation to cellular NP and AF phenotypes in the IVD. We here provide independent evidence of whole-transcriptome-based clustering of our novel immortal NP and AF cell lines with their respective IVD-derived primary isolates; in addition, PCA analysis clearly separates the NP clonal subtypes from each other and from AF cellular subpopulations. Importantly, the data accrued confirms cellular heterogeneity in NP and AF cell populations and supports the notion that cellular phenotypes are epigenetically preserved during the immortalization process.

Our membranome-based transcriptomics analysis aimed to identify novel cell surface markers for NP and AF cellular (sub)types. PCA analysis and validation of membrane and non–membrane-associated marker expression using qPCR analysis consistently reveal a closer similarity of NP-nR with AF clones than with NP-R clones. Besides being lineage specific, a cell’s transcriptomic phenotype is also determined through interaction with its microenvironment.¹⁰⁶ Hence, it is conceivable that despite their dissimilar ontogeny, original NP-nR and AF cells were derived from areas with a relatively similar microenvironment in their respective tissues of origin.

The novel membrane markers identified in this study will be crucial for the isolation, systematic analysis, and functional characterization of NP cellular subtypes and their implication in the context of disc homeostasis and disease. Investigation of specific cell subtypes in the IVD requires histological studies focused on expression analysis of markers identified herein in IVD tissue sections to define their relative locations and to understand their ontogenic relationships. The availability of a functionally defined set of NP and AF cells enables systematic analysis of altered activity within gene networks, for instance, as a function of age, and is expected to significantly contribute to understanding gene-microenvironment interaction in the IVD. It is very likely that microenvironmental alterations (e.g., inflammation, oxygen tension, osmolarity) and accompanying physiological changes related to ageing or degeneration override physiological responses required to maintain cell and tissue homeostasis. Such knowledge will also be pivotal for defining appropriate conditions for isolation, culturing, and maintaining cells from the NP and AF and for application of IVD-directed differentiation of stem cells in regenerative procedures.