Modic type 2 changes are fibroinflammatory changes with complement system involvement adjacent to degenerated vertebral endplates

Abstract

Background

Vertebral endplate signal intensity changes visualized by magnetic resonance imaging termed Modic changes (MC) are highly prevalent in low back pain patients. Interconvertibility between the three MC subtypes (MC1, MC2, MC3) suggests different pathological stages. Histologically, granulation tissue, fibrosis, and bone marrow edema are signs of inflammation in MC1 and MC2. However, different inflammatory infiltrates and amount of fatty marrow suggest distinct inflammatory processes in MC2.

Aims

The aims of this study were to investigate (i) the degree of bony (BEP) and cartilage endplate (CEP) degeneration in MC2, (ii) to identify inflammatory MC2 pathomechanisms, and (iii) to show that these marrow changes correlate with severity of endplate degeneration.

Methods

Pairs of axial biopsies (n = 58) spanning the entire vertebral body including both CEPs were collected from human cadaveric vertebrae with MC2. From one biopsy, the bone marrow directly adjacent to the CEP was analyzed with mass spectrometry. Differentially expressed proteins (DEPs) between MC2 and control were identified and bioinformatic enrichment analysis was performed. The other biopsy was processed for paraffin histology and BEP/CEP degenerations were scored. Endplate scores were correlated with DEPs.

Results

Endplates from MC2 were significantly more degenerated. Proteomic analysis revealed an activated complement system, increased expression of extracellular matrix proteins, angiogenic, and neurogenic factors in MC2 marrow. Endplate scores correlated with upregulated complement and neurogenic proteins.

Discussion

The inflammatory pathomechanisms in MC2 comprises activation of the complement system. Concurrent inflammation, fibrosis, angiogenesis, and neurogenesis indicate that MC2 is a chronic inflammation. Correlation of endplate damage with complement and neurogenic proteins suggest that complement system activation and neoinnervation may be linked to endplate damage. The endplate-near marrow is the pathomechanistic site, because MC2 occur at locations with more endplate degeneration.

Conclusion

MC2 are fibroinflammatory changes with complement system involvement which occur adjacent to damaged endplates.

Abbreviations
• ABCA8
• ABC-type organic anion transporter ABCA8
• ACTZ
• alpha-centractin
• ANOVA
• analysis of variance
• ANXA1
• annexin 1
• ANXA2
• annexin 2
• APOA1
• apolipoprotein A-I
• APOH
• beta-2-glycoprotein 1
• ARH
• low density lipoprotein receptor adapter protein 1
• ARL8B
• ADP-ribosylation factor-like protein 8B
• BEP
• bony endplate
• BGH3
• Transforming growth factor-beta-induced protein ig-h3
• BGN
• biglycan
• BPI
• bactericidal permeability-increasing protein
• C1QT3
• complement C1q tumor necrosis factor-related protein 3
• C3
• complement component C3
• C5
• complement component C5
• C8A
• complement component C8 alpha chain
• C8B
• complement component C8 beta chain
• C8G
• complement component C8 gamma chain
• C9
• complement component C9
• CAP7
• azurocidin
• CATG
• cathepsin G
• CAVN1
• caveolae-associated protein 1
• CC110
• coiled-coil domain-containing protein 110
• CEAM6
• carcinoembryonic antigen-related cell adhesion molecule 6
• CEAM8
• carcinoembryonic antigen-related cell adhesion molecule 8
• CEP
• cartilage endplate
• cLBP
• chronic low back pain
• CO12A1
• collagen alpha-1 (XII) chain
• CO14A1
• collagen alpha-1 (XIV) chain
• CO3A1
• collagen alpha-1 (III) chain
• CO6A1
• collagen alpha-1(VI) chain
• CO6A2
• collagen alpha-2 (VI) chain
• CO6A3
• collagen alpha-3 (VI) chain
• DAG1
• dystroglycan
• DCN
• decorin
• DEP
• differentially expressed protein
• DPYL3
• dihydropyrimidinase like 3
• ECM1
• extracellular matrix protein 1
• ELANE
• neutrophil elastase
• F13A
• coagulation factor XIII A chain
• FDR
• false discovery rate
• FMOD
• fibromodulin
• FN1
• fibronectin
• GGT5
• glutathione hydrolase 5 proenzyme
• GPDA
• glycerol-3-phosphate dehydrogenase [NAD(+)], cytoplasmic
• GSEA
• Gene set enrichment analysis
• H14
• histone H1.4
• HSP72
• heat shock-related 70 kDa protein 2
• HSPB1
• heat shock protein beta-1
• HV692
• immunoglobulin heavy variable 1–69-2
• ISLR
• immunoglobulin superfamily containing leucine-rich repeat protein
• ITIH2
• inter-alpha-trypsin inhibitor heavy chain H2
• IVD
• intervertebral disc
• K1C17
• keratin, type I cytoskeletal 17
• K2C8
• keratin, type II cytoskeletal 8
• KCRB
• creatine kinase B-type
• KNG1
• kininogen-1
• LBP
• low back pain
• LEG7
• galectin-7
• LOXL2
• lysyl oxidase homolog 2
• LUM
• lumican
• MC
• modic changes
• MC1
• modic type 1 changes
• MC2
• modic type 2 changes
• MIME
• mimecan
• MPO
• myeloperoxidase
• MRI
• magnetic resonance imaging
• MS
• mass spectrometry
• MYO1C
• unconventional myosin-Ic
• MYPC1
• myosin-binding protein C, slow-type
• NUDC1
• NudC domain-containing protein 1
• ORA
• overrepresentation analysis
• PCA
• principal component analysis
• PEDF
• pigment epithelium-derived factor
• PERE
• eosinophil peroxidase
• PLCB3
• 1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase beta-3
• PLIN1
• perilipin-1
• POSTN
• periostin
• PRG4
• proteoglycan 4
• PRTN3
• myeloblastin
• PSMD8
• 26 S proteasome non-ATPase regulatory subunit 8
• RL30
• 60 S ribosomal protein L30
• ROCK2
• Rho-associated protein kinase 2
• S10A9
• protein S100-A9
• SAA4
• serum amyloid A-4 protein
• SEM3A
• semaphorin-3A
• SEM3C
• semaphorin-3C
• SNTB2
• beta-2-syntrophin
• SODC
• superoxide dismutase (Cu-Zn)
• STAT
• statherin
• T1w
• T1-weighted
• T2w
• T2-weighted
• TIGAR
• fructose-2,6-bisphosphatase TIGAR
• TPP1
• tripeptidyl-peptidase 1
• VEGFR1
• vascular endothelial growth factor receptor 1
• VPS11
• vacuolar protein sorting-associated protein 11 homolog
• ZN500
• zinc finger protein 500
• α-SMA
• actin, aortic smooth muscle

INTRODUCTION

Modic changes (MC) are vertebral endplate signal intensity changes visualized by T1-weighted (T1w) and T2w magnetic resonance imaging (MRI) that occur adjacent to degenerated intervertebral discs (IVDs) and colocalize with vertebral endplate damage.^1–8 They are specific for chronic low back pain (cLBP), which could result from increased nerve fiber density in MC endplates.^9–11 From the three interconvertible types of MC, Modic type 1 changes (MC1) represent an edema-like signal pattern on MRI (T1w: hypointense, T2w: hyperintense), which indicates bone marrow inflammation.¹ This is supported by histological findings of Modic et al.,¹ who described fibrovascular granulation tissue and fibrous replacement of normal bone marrow in MC1. Bone marrow inflammation in MC1 has also been reported in a semi-quantitative histomorphometric analysis of MC bone marrow, where more edema, inflammatory infiltrates, and connective tissue were found.¹²

MC2 are visualized as hyperintense signals by both T1w and T2w MRI.¹ Histological studies show fatty replacement of normal bone marrow in MC2.^1,12 Like MC1, MC2 bone marrow also shows signs of inflammation. For example, granulation tissue and edema was shown in MC2 biopsies.^1,12 Adipocytes in MC2 might contribute to the inflammatory environment. They can secrete pro-inflammatory adipokines and are important contributors to inflammation.¹³ However, in contrast to MC1 bone marrow, MC2 do not contain lymphatic and neutrophilic inflammatory infiltrates, and inflammation does not present as edema-like signal pattern on MRI.^1,12 This indicates distinct inflammatory pathomechanisms in MC2. Cytokine profiles from IVDs at levels with MC2 suggest a pro-inflammatory mechanism with osteoclastic and complement system activation.^14–16 However, the biological mechanisms taking place in MC2 are unknown.

Endplate damage is believed to trigger MC development by promoting disc/vertebra crosstalk, compromising the disc's immune privilege, and allowing comingling of disc material with bone marrow that could trigger an autoinflammatory response.^15,17 Hence, the bone-endplate-IVD junction is likely the important pathomechanistic site in MC.

Unfortunately, the endplate-near bone marrow and the adjacent endplates of MC patients are typically not accessible for research. Here, we investigated the protein expression in MC2 endplate-near bone marrow and assessed the adjacent endplate degeneration histologically in two human cadaveric spines with MC2. The aim of this study was to investigate inflammatory pathomechanisms in MC2 endplate-near bone marrow and their association with the degree of endplate degeneration. We hypothesized that (i) BEP and CEP are more degenerated in MC2 than control, (ii) that MC2 are chronic inflammatory changes with complement system involvement, and (iii) that these bone marrow changes correlate with the severity of endplate degeneration.

MATERIALS AND METHODS

The study was conducted in accordance with the Declaration of Helsinki and approved by the local Ethics Commission (#BASEC Nr. 2021-Req-2021-00887). Chemicals were purchased from Sigma-Aldrich, Buchs, Switzerland, if not stated otherwise.

Cadaveric spines

Two fresh frozen lumbar cadaveric spines (L1-S1) with multiple levels of MC2 and non-MC were used for this study. Specimens were obtained from Science Care (Phoenix, AZ). Postmortem, spines were stored within 5–7 days at −20°C.

Imaging, custom-made 3D guide design, and bone marrow biopsy collection

Before MRI, spines were thawed for 12 h at room temperature. T1w and T2w images were done with a 3T MRI device (Prisma, Siemens). After imaging, spines were frozen again. MC2 and control regions were identified by an experienced radiologist with over >13 years of experience including >6 years in musculoskeletal radiology. Vertebrae of the frozen spines were transversally separated by cutting through the center of the IVD (Figure 1A). To ensure that biopsies were collected from predefined regions, custom-made 3D printed guides for each vertebra were created. Based on computer tomography scans (SOMATOM Edge Plus), 3D models were generated (3D Slicer 4.8.1) and matched to the sagittal and coronal plane of the MRI sequences (Blender 2.79b), a critical step, as MC2 and control regions were determined based on MRI. Due to the higher resolution, CT scans and not MR images were used to create the guides. Each guide consisted of a cranial and two caudal 3D-printed casts (Figure 1B–D). The cranial part had a grid of biopsy channels, the caudal cast was designed to allow for unobstructed biopsy collection and to apply counter pressure which ensured that the caudal CEP did not break off. The specific biopsy guides for all vertebral bodies were 3D printed (Original Prusa i3 MK3, Prusa Research s.r.o., Czech Republic) using polylactic acid (PLA) filament (Figure 1B–D).

Axial biopsies spanning the entire vertebral body height including both CEPs were collected with Jamshidi bone marrow biopsy needles (HS HOSPITAL SERVICE S.P.A., Aprilia, Italy) (8G × 100 mm), which resulted in biopsy diameters of approximately 4 mm. For each region of interest, pairs (n = 58 pairs) of biopsies 5 mm apart were taken. One biopsy was assigned to mass spectrometry (MS), the other to paraffin histology (Figure 2A, B). Biopsies subjected to MS were separated into six pieces: (1) cranial CEP, (2) bone marrow adjacent to cranial CEP, (3) cranial mid bone marrow, (4) caudal mid bone marrow, (5) bone marrow adjacent to caudal CEP and (6) caudal CEP (Figure 2C). Bone marrow pieces 2 and 5 were added to lysis buffer (4% sodium dodecyl sulfate in 100 mM Tris/HCl pH 8.2) and immediately frozen at −20°C.

Histology

Biopsies subjected to histology were directly added to 4% buffered formalin solution and fixed for 60 h at room temperature. To decalcify the samples, biopsies were added to rapid decalcification (RDF) solution (Biosystems, Muttenz, Switzerland) at room temperature for 30 h. Samples were rinsed for 20 min under running tap water and consequently added to 70% ethanol until processing. Biopsies were then dehydrated in increasing EtOH concentration, embedded in paraffin, and cut into 5 μm sections. Sections were stained with hematoxylin/eosin (HE) Masson trichrome (MT) and Safranin O (Chroma-Gesellschaft Schmid GmbH & Co, Stuttgart, Germany). Histological readouts were performed by an experienced musculoskeletal pathologist with >13 years of experience. In a first step, the general quality of the biopsy was assessed, and it was evaluated, whether vertebral endplates are histologically scoreable. All vertebral endplates with processing artifacts that made it impossible to rate were excluded. BEP and CEP degenerative features were assessed according to the standardized endplate degeneration scoring system for human IVD degeneration (0–9; 0–3 = no degeneration, 4–6 = moderate degeneration, 7–9 = severe degeneration).¹⁸ Endplate scores between MC2 and control were compared with Mann–Whitney U tests (Graphpad Prism 9.3.1.).

Proteomic analysis Sample preparation

Bone marrow biopsies were boiled at 95 °C for 1 h in lysis buffer. Proteins were mechanically extracted with a tissue homogenizer (TissueLyser II, Qiagen) by applying 2 × 2 min cycles at 30 Hz followed by another incubation at 95 °C for 30 min. Protein concentration was determined using the Lunatic UV/Vis polychromatic spectrophotometer (Unchained Labs). For each sample, 50 μg of protein was reduced with 2 mM tris(2-carboxyethyl)phosphine and alkylated with 15 mM chloroacetamide at 30 °C for 30 min. Samples were processed using the single-pot solid-phase enhanced sample preparation (SP3). The SP3 protein purification, digest and peptide clean-up was performed using a KingFisher Flex System (Thermo Fisher Scientific) and Carboxylate-Modified Magnetic Particles (GE Life Sciences; GE65152105050250, GE45152105050250).¹⁹ Beads were conditioned following manufacturer's instructions, consisting of three washes with water at a concentration of 1 μg/μl. Samples were diluted with 100% ethanol to a final concentration of 60% ethanol. The beads, wash solutions and samples were loaded into 96 deep well- or micro-plates and transferred to the KingFisher. Following steps were carried out with a robot: collection of beads from the last wash, protein binding to beads, washing of beads in wash solutions 1–3 (80% ethanol), protein digestion (overnight at 37°C with a trypsin:protein ratio of 1:50 in 50 mM Triethylammoniumbicarbonat) and peptide elution from the magnetic beads using MilliQ water. The digest solution and water elution were combined and dried completely and re-solubilized in 40 μl 0.1% formic acid. For loading the Evotip (Evosep, Odense, Denmark), samples were 1:10 diluted in 0.1% formic acid and 5 μl was loaded on the tip following the manufacturer's instructions, except that the samples were washed with 100 μl of 0.1% formic acid.

Liquid chromatography tandem mass spectrometry (LC–MS/MS)

MS analyses were performed on a timsTOF Pro (Bruker, Daltonics, Bremen, Germany) coupled to an Evosep One (Evosep, Odense, Denmark). Samples were separated with the extended Evosep method “30 samples/day” keeping the analytical column (PSC-15-100-3-UHPnC, ReproSil C18 3 m 120 Å 15 cm ID 100 μm, PepSep) at 50°C. For the dual timsTOF, MS spectra were scanned from m/z 100 to m/z 1700 in ddaPASEF mode (data dependent acquisition Parallel Accumulation Serial Fragmentation). For the ion mobility settings, the inversed mobilities from 1/K0 0.60 to 1.60 Vs/cm² were analyzed with ion accumulation and ramp time of 100 ms, respectively. One survey TIMS-MS scan was followed by 10 PASEF ramps for MS/MS acquisition, resulting in a 1.1 s cycle time. Singly charged ions were excluded using the polygon filter mask and isolation windows for MS/MS were set to m/z 2.0 for precursor ions below m/z 700, and m/z 3.0 for ions above. The MS proteomics data were handled using the local laboratory information management system²⁰ and all relevant data have been deposited to the ProteomeXchange Consortium via the PRIDE (http://www.ebi.ac.uk/pride) partner repository with the data set identifier PXD038008.

Peptide identification and quantification

The acquired MS raw data were processed for identification and quantification using FragPipe (version 16.0), MSFragger (version 3.4) and Philosopher (version 4.4.1).²¹ Spectra were searched against a Uniprot Homo sapiens reference proteome (taxonomy 9606, canonical version from January 26, 2022), concatenated to its reversed decoyed fasta database and common protein contaminants. For the closed search settings, strict trypsin digestion with a maximum of one missed cleavage was set. Carbamidomethylation of cysteine was set as fixed modification, while methionine oxidation and N-terminal protein acetylation were set as variable. Label free quantification and match between run option were enabled. For differential expression analysis, the “Razor Intensity” reported in the combined_protein.tsv was used.

Differential expression analysis

Differential protein expression analysis was performed using the R-package prolfqua.²² Only proteins identified by at least two peptides were included into the analysis. First, protein abundances were log₂ transformed. Next, a linear model with a single factor (Modic changes) was fitted to each protein, and protein log₂fold-changes (log₂fc) were estimated and tested using the model parameters. To increase the statistical power, the variance estimates were moderated using the empirical Bayes approach, which exploits the parallel structure of the high throughput experiment.²³ Finally, the p-values were adjusted using the Benjamini and Hochberg procedure to obtain the false discovery rate (FDR).

Differentially expressed proteins (DEPs) were defined as log₂fc > ±1 and false discovery rate (FDR) < 0.1. Gene set enrichment analysis (GSEA) was performed with Web-based GSEA Toolkit (WebGestalt)²⁴ with proteins ranked based on t-statistic. Overrepresentation analysis (ORA) of all upregulated DEPs was performed with WebGestalt using the complete identified protein list as background input. ORA and GSEA terms were considered significantly enriched for FDR <0.1. ORA of DEPs that correlated with histological BEP and CEP scores was also done with WebGestalt using all upregulated DEPs as background input. Due to low protein numbers, no FDR threshold was applied, and the top 10 terms ranked based on the enrichment ratio are presented.

Correlation of endplate scores with proteins

BEP and CEP scores were correlated with normalized protein abundances of all DEPs using Spearman correlation (Graphpad Prism 9.3.1.). Correlations were calculated for MC2 and control samples. Correlation coefficients from ±0.21 to 0.40 were considered as weak correlation, ±0.41 to 0.60 as moderate, ±0.61 to 0.80 as strong, and ±0.81 to 1.00 as very strong correlation.²⁵ Protein/histology pairs with fewer than 10 samples were excluded from the correlation analysis.

RESULTS Donor characteristics

Vertebrae of two donors with MC2 and intra-patient control regions were used for this study. Donor 1 had MC2 regions in vertebrae L2–L5. The major MC2 regions were adjacent to the L2/3 and L4/5 bone-disc junction. MC2 regions of donor 2 were found adjacent to the L1/2 and L4/5 bone-disc junction. The major MC2 regions were in L1 caudal, L2 cranial, L4 caudal, and L5 cranial. Donor 1 was female, donor 2 was male. In average, they were 53.5 years ±3.5 old, 171.5 cm ± 12.0, 109.0 kg ± 12.7, and had a body mass index (BMI) of 38.0 ± 9.9 (Table 1). Donors tested negative for Hepatitis B and C, and human immunodeficiency virus 1 (HIV-1) and HIV-2.

TABLE 1 Donor information

	Donor information
	Age	Height (cm)	Weight (kg)	BMI	Sex
Donor 1	56.0	163.0	118.0	45.0	F
Donor 2	51.0	180.0	100.0	31.0	M
Mean	53.5	171.5	109.0	38.0	50% F
SD	3.5	12.0	12.7	9.9

MC2 occur adjacent to more degenerated vertebral endplates with lost CEP/BEP boundaries

From 58 collected biopsies (containing each a cranial and caudal vertebral endplate) subjected to histological processing and analysis, five had to be excluded from evaluation, because they were not interpretable due to technical processing issues. Degenerative features of 53 biopsies could be evaluated for at least one endplate score. Total 79 CEPs and 83 BEPs were evaluated. CEP (control [n = 56]: median = 0, IQR [0, 0]; MC2 [n = 23]: median = 2, IQR [0, 4]; p < 0.001) and BEP degeneration scores (control [n = 59]: median = 0, IQR [0, 0]; MC2 (n = 24): median = 2, IQR [2, 4], p < 0.001) were significantly increased in MC2 compared to control (Figure 3). Furthermore, CEP/BEP boundaries were often completely lost in MC2 (Figure 4C, D). All control endplates with a CEP (15.22%) and BEP (11.11%) score higher than 1 were collected right adjacent to larger MC2 lesions (lesions that affected minimally one third of the entire vertebral body).

MC2 and donor explain most of the protein variance

From 58 collected biopsies, a total of 99 endplate-near bone marrow pieces were analyzed (control: n = 69; MC2: n = 30; Table 2). Five endplate-near bone marrow pieces had to be excluded due to sampling issues and from the 12 biopsies collected from the S1 levels, only cranial pieces were available. By proteomic analysis, 2534 proteins were detected. Principal component analysis (PCA) of the entire dataset showed that PC1 explains 37.7% and PC2 24.5% of the variance (Figure 5). ANOVA revealed that most of the observed variance can be explained by MC and by donor, while anatomical location (cranial/caudal) and spinal segment (L1-S1) were not significant (Figure S1).

TABLE 2 Collected samples subjected to proteomic analysis

	Control	MC2	Total
(2) Cranial BM adjacent CEP	40	13	53
(5) Caudal BM adjacent CEP	29	17	46
Total	69	30	99

Abbreviations: BM, bone marrow; CEP, cartilage endplate; MC2, Modic type 2 changes.

MC2 endplate-near bone marrow contains more complement system components, extracellular matrix (ECM) proteins and angiogenic factors

Differential expression analysis comparing MC2 to control revealed 318 DEPs (182 upregulated and 136 downregulated DEPs; Figure 6A). Top upregulated DEPs are found in Table 3, top downregulated are listed in Table 4.

TABLE 3 Top upregulated differentially expressed proteins (DEPs) comparing MC2 to control in endplate-near bone marrow

Upregulated DEPs in MC2 bone marrow	Uniprot ID	Abbreviation	Description	Log2fc	FDR
1	Q9Y4K0	LOXL2	Lysyl oxidase homolog 2	3.560	0.000
2	Q9NQ88	TIGAR	Fructose-2,6-bisphosphatase TIGAR	3.128	0.001
3	Q6NZI2	CAVN1	Caveolae-associated protein 1	2.939	0.000
4	P36955	PEDF	Pigment epithelium-derived factor	2.871	0.000
5	O00159	MYO1C	Unconventional myosin-Ic	2.755	0.000
6	P62736	α-SMA	Actin, aortic smooth muscle	2.726	0.000
7	Q14563	SEM3A	Semaphorin-3A	2.720	0.065
8	O14498	ISLR	Immunoglobulin superfamily containing leucine-rich repeat protein	2.640	0.001
9	P19823	ITIH2	Inter-alpha-trypsin inhibitor heavy chain H2	2.593	0.000
10	Q99985	SEM3C	Semaphorin-3C	2.566	0.032
11	P02647	APOA1	Apolipoprotein A-I	2.512	0.000
12	P00488	F13A	Coagulation factor XIII A chain	2.488	0.000
13	Q16610	ECM1	Extracellular matrix protein 1	2.408	0.042
14	P20774	MIME	Mimecan	2.387	0.000
15	O60240	PLIN1	Perilipin-1	2.370	0.000
16	P12277	KCRB	Creatine kinase B-type	2.333	0.000
17	P21810	BGN	Biglycan	2.265	0.000
18	P01042	KNG1	Kininogen-1	2.178	0.000
19	Q15063	POSTN	Periostin	2.175	0.002
20	P51884	LUM	Lumican	2.163	0.000
21	P07585	DCN	Decorin	2.152	0.000
22	P12109	CO6A1	Collagen alpha-1(VI) chain	2.144	0.000
23	Q15582	BGH3	Transforming growth factor-beta-induced protein ig-h3	2.125	0.000
24	Q06828	FMOD	Fibromodulin	2.098	0.000
25	P21695	GPDA	Glycerol-3-phosphate dehydrogenase [NAD(+)], cytoplasmic	2.090	0.006
26	P02749	APOH	Beta-2-glycoprotein 1	2.080	0.000
27	Q92954	PRG4	Proteoglycan 4	2.062	0.003
28	P04792	HSPB1	Heat shock protein beta-1	1.998	0.000
29	P02808	STAT	Statherin	1.946	0.045
30	P10412	H14	Histone H1.4	1.916	0.000

Note: Ranked based on decreasing log₂fc. Proteins were considered differentially upregulated if false discovery rate (FDR) was <0.1 and log₂fc >1.

Abbreviation: FC, fold change.

TABLE 4 Top downregulated differentially expressed proteins (DEPs) comparing MC2 to control in endplate-near bone marrow

Downregulated DEPs in MC2 bone marrow	Uniprot ID	Abbreviation	Description	Log2fc	FDR
1	P62888	RL30	60 S ribosomal protein L30	−6.649	0.000
2	O60304	ZN500	Zinc finger protein 500	−5.934	0.001
3	Q13425	SNTB2	Beta-2-syntrophin	−5.565	0.000
4	Q9H270	VPS11	Vacuolar protein sorting-associated protein 11 homolog	−5.198	0.000
5	Q96RS6	NUDC1	NudC domain-containing protein 1	−5.004	0.000
6	Q00872	MYPC1	Myosin-binding protein C, slow-type	−4.819	0.000
7	P05787	K2C8	Keratin, type II cytoskeletal 8	−4.134	0.001
8	Q8TBZ0	CC110	Coiled-coil domain-containing protein 110	−3.759	0.000
9	A0A0G2JMI3	HV692	Immunoglobulin heavy variable 1–69-2	−3.643	0.000
10	P35542	SAA4	Serum amyloid A-4 protein	−3.564	0.002
11	P48556	PSMD8	26 S proteasome non-ATPase regulatory subunit 8	−3.047	0.001
12	Q9NVJ2	ARL8B	ADP-ribosylation factor-like protein 8B	−3.005	0.000
13	Q9BXJ4	C1QT3	Complement C1q tumor necrosis factor-related protein 3	−2.804	0.001
14	P47929	LEG7	Galectin-7	−2.803	0.001
15	Q01970	PLCB3	1-phosphatidylinositol 4,5-bisphosphate phosphodiesterase beta-3	−2.568	0.000
16	O94911	ABCA8	ABC-type organic anion transporter ABCA8	−2.421	0.015
17	Q04695	K1C17	Keratin, type I cytoskeletal 17	−2.345	0.002
18	Q5SW96	ARH	Low density lipoprotein receptor adapter protein 1	−2.310	0.034
19	P00441	SODC	Superoxide dismutase [Cu-Zn]	−2.305	0.000
20	P40199	CEAM6	Carcinoembryonic antigen-related cell adhesion molecule 6	−2.114	0.000
21	P11678	PERE	Eosinophil peroxidase	−1.921	0.000
22	P08311	CATG	Cathepsin G	−1.908	0.000
23	P20160	CAP7	Azurocidin	−1.862	0.000
24	P17213	BPI	Bactericidal permeability-increasing protein	−1.852	0.000
25	P06702	S10A9	Protein S100-A9	−1.836	0.000
26	O75116	ROCK2	Rho-associated protein kinase 2	−1.779	0.009
27	P54652	HSP72	Heat shock-related 70 kDa protein 2	−1.777	0.019
28	O14773	TPP1	Tripeptidyl-peptidase 1	−1.741	0.000
29	P24158	PRTN3	Myeloblastin	−1.731	0.000
30	P31997	CEAM8	Carcinoembryonic antigen-related cell adhesion molecule 8	−1.703	0.000

Note: Ranked based on increasing log₂fc. Proteins were considered differentially expressed if false discovery rate (FDR) was <0.1 and log₂fc was <−1.

Abbreviation: FC, fold change.

ORA of the upregulated DEPs in MC2 revealed enriched “complement and coagulation cascades” (FDR = 0.00) and “complement activation” (FDR = 0.00) as the top overrepresented pathways and “extracellular structure organization (FDR = 0.00) and “angiogenesis” (FDR = 0.00) as the top enriched biological processes (Figure 6B, C). Most of the significantly upregulated biological processes and pathways identified by ORA in MC2 related to complement system, inflammation, ECM organization/cell adhesion, or angiogenesis/wounding (Table 5).

TABLE 5 Results from overrepresentation analysis (ORA) of upregulated differentially expressed proteins (DEPs) in MC2

Umbrella term	Pathway/biological process
Complement system	Complement and coagulation cascades
	Complement activation
	Human complement system
Inflammation	Inflammatory response pathway
Extracellular matrix organization and cell adhesion (fibrosis)	Extracellular structure organization
	Connective tissue development
	Cell substrate adhesion
	Focal adhesion-PI3K-Akt-mTOR-signaling pathway
	Focal adhesion
Angiogenesis and wound healing	Angiogenesis
	Regulation of vasculature development
	Platelet degranulation

Note: Most of the significantly upregulated biological processes related to complement system, inflammation, ECM organization/cell adhesion, or angiogenesis/wounding.

The upregulated DEPs of the complement system were complement component 1qb (C1QB), C1QC, C5, C8A, C8B, C8G, C9, and complement factor B (CFB) and CFH. Furthermore, many different collagen subtypes such as collagen type VI alpha 1 chain (CO6A1), CO6A2, CO6A3, CO3A1, CO12A1, CO14A1, and other ECM proteins and proteoglycans such as fibromodulin (FMOD), biglycan (BGN), decorin (DCN), and fibronectin (FN1) were upregulated in MC2 (Table 3, not shown). Additionally, alpha smooth muscle actin (α-SMA), a key marker for excessive matrix producing myofibroblasts, was also increased in MC2. The top upregulated protein was lysyl oxidase homolog 2 (LOXL2), an important pro-angiogenic factor. Further angiogenic associated proteins upregulated in MC2 were annexin 1 (ANXA1) and ANXA2, extracellular matrix protein 1 (ECM1), periostin (POSTN), and vascular endothelial growth factor receptor 1 (VEGFR1) (Table 3). Among the top upregulated proteins were also axon guiding proteins such as semaphorin-3A (SEM3A) and (SEM3C) and positive regulators of neurogenesis like dystroglycan 1 (DAG1) and dihydropyrimidinase like 3 (DPYL3) (Table 3, not shown).

GSEA confirmed enrichment of the complement system (“complement and coagulation cascades”: normalized enrichment score [NES] = 2.32, FDR = 0.00; “complement activation”: NES = 2.12, FDR = 0.00), processes associated with ECM remodeling (“extracellular structure organization”: NES = 2.14, FDR = 0.00; “connective tissue development”: NES = 1.99, FDR = 0.00), and of inflammatory processes (“inflammatory response pathway”: NES = 1.99, FDR = 0.00; “acute inflammatory response”: NES = 1.96, FDR = 0.00) (Figure 6D, E). The inflammatory processes did not seem to be driven by neutrophils because “neutrophil mediated immunity” (NES = −1.94, FDR = 0.02) and “granulocyte activation” (NES = −1.94, FDR = 0.04) (Figure 6D, E) were downregulated. Neutrophil associated proteins were also found among the downregulated DEPs, which included azurocidin (CAP7), cathepsin G (CATG), neutrophil elastase (ELANE), and myeloperoxidase (MPO) (Table 4, not shown).

Complement proteins and neurogenic factors correlate with endplate degeneration

To identify whether certain proteins can be linked to endplate degeneration, the 318 DEPs between MC2 and control were correlated with histological BEP and CEP scores. All upregulated DEPs had a positive correlation with BEP and CEP scores, all downregulated DEPs a negative correlation. BEP and CEP scores correlated very strongly (ρ = 0.91, p = 2.58 E-21). Sixty-seven DEPs had a positive correlation >0.41 with BEP scores, 35 with CEP scores. Eleven downregulated DEPs had a negative correlation <−0.41 with CEP score, and none of the downregulated DEPs had a negative correlation <−0.41 with BEP scores. Glutathione hydrolase 5 proenzyme (GGT5) correlated strongest positively with BEP scores (ρ = 0.81, p = 3.31 E-04), alpha-centractin (ACTZ) strongest with CEP scores (ρ = 0.81, p = 3.31 E-04).

From the DEPs that had a moderate to very strong positive correlation with BEP and CEP scores, 32 overlapped, 35 were specific for BEP, and three were specific for CEP (Figure 7A). ORA of DEPs specific for BEP revealed enriched complement system related terms like “complement activation” and “complement and coagulation cascades”, processes involved in wounding (“coagulation”; “blood blotting cascade”), and “negative regulation of cell adhesion” (Figure 7B, C). ORA of DEPs that overlapped between CEP and BEP were complement system, wounding, and specifically neurogenesis related terms (“response to axon injury”, “positive regulation of neurogenesis” and “regulation of neutron projection development”) (Figure 7D, E). Due to low protein numbers, ORA was not performed for proteins that negatively associated with endplate scores. However, neutrophil related proteins cathepsin G (CATG), protein S100-A8 (S100A8), S100A9 (CEP: ρ = −0.416, p = 0.001) correlated negatively with CEP scores. Complement system activation was represented by complement factors C8B, C8A, and C1QC. Neurogenesis was represented by neurogenic proteins DPYL3, DAG1, galectin-1 (LGALS1), microtubule associated protein 1B (MAP1B), and serpin family F member 1 (SERPINF1). Control bone marrow of biopsies that had a CEP and BEP score >1 also had trend towards more C1QC and C8B compared with control marrow with CEP and BEP scores = 0 (Table 6).

TABLE 6 Median protein abundances of complement component 1CQC (C1QC) and C8B in endplate-near control bone marrow with cartilage endplate (CEP) and bony endplate (BEP) scores > and <1

	Control
	CEP score					BEP score
	<1	>1	Difference Δ > 1 − < 1	p value	n	<1	>1	Difference Δ > 1 − < 1	p value	n
C1QC	15.98, [15.76, 16.24]	16.28, [15.70, 17.12]	0.30	0.14	38; 7	15.97, [15.75, 16.25]	16.81, [15.91, 17.81]	0.84	0.08	39; 5
C8B	15.55, [14.66, 15.98]	16.24, [14.78, 16.74]	0.69	0.18	23; 3	15.55, [14.66, 16.01]	16.65, [14.78, 16.74]	1.10	0.19	24; 3

Note: Values represent median with interquartile range.

DISCUSSION

Here, we investigated inflammatory processes in MC2 bone marrow and linked bone marrow changes with endplate degeneration. We found that MC2 occur adjacent to stronger degenerated vertebral endplates and endplate-near MC2 bone marrow contains increased complement system and ECM proteins, as well as angiogenic and neurogenic factors. Complement system and neurogenic proteins correlated with endplate damage. This gives evidence for fibroinflammatory processes in MC2 and indicates that the pathomechanistic site in MC2 is at the bone-disc junction.

Endplate damage and complement system activation in MC2

Vertebral endplate damage is critical for the development of MC.⁶ We semi-quantitatively assessed CEP and BEP degeneration with histology in human MC2 and intra-patient control vertebral regions and found that MC2 lesions occur adjacent to degenerated CEPs and BEPs with lost CEP/BEP boundaries. This is indicative that vertebral endplate damage is a component of the MC2 pathology. That endplate damage leads to MC2 is supported by the fact that endplate damage is predictive for MC development.^8,26 Interestingly, all control endplates with CEP and BEP scores higher than 1 were collected right adjacent to larger MC2 lesions (lesions that affected minimally one third of the entire vertebral body). This suggests that the boarders of existing endplate damages in MC2 regions are degenerating and that endplate defects propagate from existing MC2 regions. This is plausible because boarders of existing endplate defects are likely under increased mechanical loads, which could cause further endplate damages.²⁷ Importantly, anchoring of the endplate regions adjacent to MC2 might also be weakened by biological reactions; the inflammatory molecules from MC2 may diffuse to adjacent non-MC regions and trigger degenerative changes. A role of MC2 inflammatory processes in endplate damaging is supported by (i) upregulated MC2 complement system proteins that correlate with vertebral endplate scores, and by (ii) elevated complement components (C1QC, C8B) in control marrow adjacent to MC2 lesions with CEP and BEP scores >1. Complement factors are known to enhance osteoclast formation, especially in an inflammatory environment.²⁸ Therefore, diffusion of complement factors from MC2 to adjacent non-MC regions could be a mechanism that extends endplate damage and MC2 lesions.

MC2 endplate-near bone marrow contains increased complement system proteins (C1QB, C1QC, C5, C8A, C8B, C8G, C9, and CFB, and CFH) from the classical and alternative pathway.²⁹ These pathways converge at the cleavage of C5 to 5a and 5b. C5a attracts immune cells and C5b assembles with the complement factors C6, C7, C8, and C9 to form the membrane attack complex (MAC).³⁰ C5, C8, and C9 are directly involved in the formation of the MAC. Increased C5, C8, and C9 in MC2 give indications for increased MAC formation in MC2. This is in agreement with reports of increased amounts of C8B, an important MAC complex protein, in MC2 discs, and with more MAC complexes in degenerated CEPs and nucleus pulposus.^16,31 MAC can either be embedded in membranes resulting in membrane lysis of pathogens, or it activates leukocytes as a consequence of autoantibody-antigen complex formation in autoimmune disorders such as rheumatoid arthritis and systemic lupus erythematosus.^32,33 In MC, occult infection of the disc as well as autoimmune reactions of the bone marrow against disc material are potential etiologies.^2,17,34–36 Hence, involvement of MAC in MC pathophysiology is plausible.

The reason for complement activation in MC2 remains unknown, yet complement activation is closely linked to tissue damage.^28,37 Intracellular proteins of necrotic/apoptotic cells can activate the complement system.³⁷ Also, ECM-derived proteins can activate the complement system. Cartilage-derived proteoglycans that we found upregulated in MC2 can activate the complement system.^38–40 For example, FMOD can directly bind to C1Q and consequently activate the classical pathway of the complement system.³⁸ BGN, DCN, and ECM1 are cartilage-derived proteoglycans that were among the top upregulated proteins in MC2 endplate-near bone marrow that can also activate the complement system.^39–42 These proteoglycans may have been produced within MC2 bone marrow as part of fibrotic changes, or they may have been expelled from the disc and CEP into the bone marrow as a consequence of degenerative changes. Lost CEP/BEP boundary in MC2 could facilitate their drainage into the bone marrow.

Taken together, activated complement system in MC2 indicate inflammation and may be linked to resorptive processes of the endplate. The activators of the complement system and its role in the bacterial and autoimmune etiologies of MC need to be investigated. A vicious cycle of endplate damaging and complement activation is plausible.

Evidence for chronic inflammation

Tissue damage can activate immune cells like neutrophils and macrophages to release pro-inflammatory cytokines such as tumor necrosis factor alpha (TNF-α), interleukin 1β (IL-1β), and IL-6.⁴³ These inflammatory stimuli can promote vascular permeability, which results in the infiltration of more pro-inflammatory cytokine producing immune cells.⁴⁴ The pro-inflammatory cytokines also stimulate endothelial cells to proliferate and migrate in order to form new blood vessels. This ensures nutrient supply and further facilitates immune cell migration. After tissue damage, fibroblast and mesenchymal stromal cells differentiate into myofibroblasts, which try to repair the damage.⁴⁵ However, failure to remove the initial inflammatory trigger prevents the clearance of myofibroblasts and of newly formed blood vessels. Therefore, fibrosis and pathological angiogenesis are signs of chronic inflammation.⁴⁴

MC2 endplate-near bone marrow contains a set of upregulated proteins that are linked to fibrosis and angiogenesis and hence indicate chronic inflammation. ORA identified “angiogenesis” and “regulation of vasculature development” among the top enriched processes in MC2. In addition, the top upregulated DEP in MC2 is LOXL2, a protein that promotes endothelial cell proliferation and sprouting and is highly associated with angiogenesis.⁴⁶ Increased angiogenesis is further supported by upregulated pro-angiogenic proteins like POSTN, ECM1, and VEGFR1.^47,48

We found several indications for fibrosis in MC2. First, ORA revealed “extracellular structure remodeling” and “connective tissue development” in MC2. Second, cell matrix adhesion, an important fibrotic mechanism, was upregulated. Third, several fibrosis-associated collagen subtypes were increased, that is, CO3A1 and CO6A1-3.^45,49,50 And fourth, α-SMA, the characteristic protein of fibrosis-associated myofibroblasts was upregulated. In a previous study, we showed that MC1 bone marrow contains an overrepresented population of leptin receptor high (LEPR^high) expressing bone marrow stromal cells (BMSCs).⁵¹ LEPR^high BMSCs are highly susceptible to undergo myofibroblast differentiation. Interestingly, this population has also an increased adipogenic differentiation capacity. It could be speculated that the LEPR^high expressing BMSC population in MC1 becomes the source of the adipocytes and myofibroblasts when MC1 converts to MC2.

Role of neutrophils and macrophages in MC2

Even though the complement system can attract and activate granulocytes, the inflammatory process in MC2 does not seem to be neutrophil-mediated. GSEA revealed a downregulation of “neutrophil-mediated immunity” and “granulocyte activation”, and key neutrophil derived proteins such as the neutrophil marker CEAM8, CAP7, ELANE, and MPO were significantly downregulated. Furthermore, neutrophil associated proteins had a negative correlation with endplate damage. As neutrophils are the most abundant leukocytes in humans, the downregulation of neutrophil associated proteins in MC2 could simply reflect replacement of hematopoietic cells with fibrous tissue. However, it could also be the consequence of a transitioning mechanism from MC1 to MC2. In MC1, neutrophilic infiltrates were found,¹⁵ maturation of neutrophils is dysregulated,¹⁵ and neutrophils are activated.³⁶ In contrast, in MC2, no neutrophilic infiltrates were found and the neutrophil progenitor pool is depleted.^12,15 Hence, in the transition from MC1 to MC2, neutrophils must be cleared. This is typically done by macrophages. GGT5, the protein with the strongest correlation with the BEP score, is highly expressed in macrophages and its tissue expression correlates with macrophage infiltration.⁵² Cluster of differentiation 14 (CD14), an important macrophage surface marker, is also elevated in MC2 endplate-near bone marrow. Therefore, our data point at increased numbers of macrophages in MC2. A role of macrophages in MC2 is further supported by previous studies, where it has been shown that discs adjacent to MC2 secrete more macrophage colony stimulating factor (M-CSF) and monocyte chemoattractant protein 1 (MCP1).¹⁴ These cytokines could attract macrophages to MC2 endplates.

Evidence for neurogenesis at degenerated endplates

We found that the neurogenic proteins SEM3A, SEM3C, DAG1, DPYL3, SERPINF1, and LGALS1 were increased in MC2 endplate-near bone marrow. Neurogenic proteins also correlated with BEP and CEP scores, suggesting that neurogenesis occurs at places of damaged endplates. This is in agreement with previous findings, where CEP damage has been shown to coincide with increased nerve fiber density, and where MC1 and MC2 CEPs were shown to have significantly more protein gene product (PGP) 9.5 nerve fibers.^9,53 Neoinnervation of CEPs may be responsible for the high specificity of MC for vertebrogenic LBP.^10,34,54 MC1 seem to be more painful than MC2.⁵⁵ Distinct inflammatory pathomechanisms in MC1 and MC2 (e.g., neutrophil vs. complement system mediated) could explain it. This study gives more evidence that the pain in MC may originate from the endplate region, which underscores that the pathomechanistic active site in MC is at the bone-disc junction. The identified proteins provide a first biological mechanism for neoinnervation in MC2. Neoinnervation and neo-angiogenesis often occur in chronically inflamed tissue.⁵³ Therefore, upregulation of neurogenesis along with increased angiogenesis and inflammation provides compelling evidence that MC2 is a chronic inflammation.

Relevance for future studies

Targeted treatments do not exist for MC2, mainly due to a lack of knowledge about the pathobiology. With upregulated complement factors, fibrosis, angiogenesis, and neurogenesis in MC2, we provide evidence that MC2 are chronic inflammatory processes. Therefore, the complement system in MC2 might be an interesting upstream target to control inflammation and progressive tissue damage. Understanding the cellular involvement and the transition mechanisms between MC subtypes could provide novel therapeutic options by blocking or accelerating conversions. In addition, by showing that MC2 occur adjacent to damaged vertebral endplates that correlate with complement proteins, we give evidence that the pathomechanistic active site of MC2 is at the bone marrow/endplate/disc junction. This underscores the importance to perform further cross compartmental MC2 studies including endplate-near bone marrow, endplates, and IVDs.

Limitations

The results of this study stem from cadaveric spines from donors with unknown clinical representation and not from patients. Even though multiple MC2 regions and adjacent vertebral endplates were analyzed, the data was collected from two donors. In addition, the enriched pathways may reflect changes in tissue and cellular composition and not necessarily activation of the pathways. However, this is the only study that quantified and compared protein expression in MC2 endplate-near bone marrow and correlated it with histological endplate degeneration.

In conclusion, concurrent inflammation, fibrosis, angiogenesis, and neurogenesis provide compelling evidence that MC2 is a chronic inflammation that occurs adjacent to more degenerated endplates. Correlation of MC2 inflammatory processes with endplate damage is indicative for the bone-disc junction as pathomechanistic active site in MC2 (Figure 8).

AUTHOR CONTRIBUTIONS

The initial study conception and design were proposed by Irina Heggli and Stefan Dudli. Marco D. Burkhard and Jonas Widmer performed MRI and CTs of the spines, Nadja Farshad-Amacker did the radiological readout. Irina Heggli, Christoph J. Laux, and Frédéric Cornaz optimized sample collection techniques with the custom-made 3D printed guides, Tamara Mengis and Sibylle Pfammatter optimized the proteomic sample processing and the MS protocol. Irina Heggli, Christoph J. Laux, Tamara Mengis, Nick Herger, Borbala Aradi-Vegh, and Stefan Dudli collected all biopsies and pre-processed them. Sibylle Pfammatter performed the MS of the samples and Witold E. Wolski did the statistical analysis. Histological readout was performed by Agnieszka Karol. Histological data analysis, bioinformatic enrichment analysis, and the first draft was done by Irina Heggli and Stefan Dudli. All co-authors edited and reviewed the manuscript. Funding for this project was provided by Stefan Dudli, Christoph J. Laux, Mazda Farshad, Florian Brunner, and Oliver Distler.

ACKNOWLEDGMENT

This work was supported by the “Stipendienfonds Swiss Orthopaedics”, the Balgrist foundation, the Velux Stiftung, and the Baugarten Stiftung.

CONFLICT OF INTEREST

The authors declare no conflict of interest.