Role of C-reactive protein in the bone marrow of Modic type 1 changes

Abstract

Modic type 1 changes (MC1) are vertebral bone marrow lesions and associate with low back pain. Increased serum C-reactive protein (CRP) has inconsistently been associated with MC1. We aimed to provide evidence for the role of CRP in the tissue pathophysiology of MC1 bone marrow. From 13 MC1 patients undergoing spinal fusion at MC1 levels, vertebral bone marrow aspirates from MC1 and intrapatient control bone marrow were taken. Bone marrow CRP, interleukin (IL)-1, and IL-6 were measured with enzyme-linked immunosorbent assays; lactate dehydrogenase (LDH) was measured with a colorimetric assay. CRP, IL-1, an d IL-6 were compared between MC1 and control bone marrow. Bone marrow CRP was correlated with blood CRP and with bone marrow IL-1, IL-6, and LDH. CRP expression by marrow cells was measured with a polymerase chain reaction. Increased CRP in MC1 bone marrow (mean difference: +0.22 mg CRP/g, 95% confidence interval [CI] [−0.04, 0.47], p = 0.088) correlated with blood CRP (r = 0.69, p = 0.018), with bone marrow IL-1β (ρ = 0.52, p = 0.029) and IL-6 (ρ = 0.51, p = 0.031). Marrow cells did not express CRP. Increased LDH in MC1 bone marrow (143.1%, 95% CI [110.7%, 175.4%], p = 0.014) indicated necrosis. A blood CRP threshold of 3.2 mg/L detected with 100% accuracy increased CRP in MC1 bone marrow. In conclusion, the association of CRP with inflammatory and necrotic changes in MC1 bone marrow provides evidence for a pathophysiological role of CRP in MC1 bone marrow.

1 |. INTRODUCTION

Chronic low back pain (CLBP) is one of the most common and most costly chronic medical conditions, with a growing personal and socioeconomic burden.¹ A fundamental challenge for improving the lives of CLBP patients is the identification of specific CLBP phenotypes and the development of effective targeted treatments for these phenotypes. CLBP patients with disc degeneration and with adjacent bone marrow lesions, known as Modic changes (MC), have been suggested to be “an entity on its own” that “deserve to be diagnosed as having specific LBP.”² MC are magnetic resonance imaging (MRI) signal intensity changes of the vertebral bone marrow around a degenerated intervertebral disc³ and are an independent factor that associates with CLBP.⁴ Three interconvertible MC types exist, of which Modic type 1 changes (MC1) has the highest association with pain.⁴ Prevalence of MC1 in CLBP patients is 4%−69%.⁴ MC1 patients report a greater frequency and duration of LBP episodes, seek care more often, have a higher risk for a poor outcome of conservative treatment, and have an “inflammatory pain pattern.”^5–7 Larger lesions are more painful and have a positive predictive value for pain of up to 100%.^8,9 There is currently no approved treatment and no treatment consensus. Insufficient data impede to the conclusion of the effect of preoperative MC1 on the outcome of spinal fusion.¹⁰ Recently, basivertebral nerve radiofrequency ablation has shown to be an effective treatment option for MC1 patients.¹¹ Clinical guidelines recommend against routine spine MRI for CLBP because of the risk of patient catastrophizing and because of high costs.¹² Therefore, MC1 are generally not detected in primary care, and it often takes years until MC1 are identified on MRI. Serum or plasma biomarkers are desirable because (i) they do not require expensive MRI, (ii) they could already be prescribed by primary care physicians, (iii) they are quantitative, and (iv) they have no scanner-dependent sensitivity.

Modic type 1 changes are considered as local inflammation of the bone marrow.^7,13 Increased serum C-reactive protein (CRP) has inconsistently been associated with MC1. In a prospective study of 36 CLBP patients, high-sensitivity CRP was significantly higher in MC1 (4.64 ± 3.09 mg/L) compared to control (1.33 ± 0.77 mg/L) and compared to Modic type 2 changes (1.75 ± 1.30 mg/L).¹⁴ Two other studies reported no difference in serum CRP between CLBP patients with any type of lumbar MC and controls without a history of CLBP and without lumbar MC.^15,16 Of three studies comparing CRP in CLBP patients without analyzing MC, one study reported significantly increased CRP in patients with CLBP.^17–20 High CRP was associated with a better outcome of conservative treatment,^20,21 whereas high CRP was a risk factor for poor surgical outcomes.²²

CRP is an acute phase protein, and its serum levels increase rapidly and strongly in response to inflammation and tissue damage. CRP is mainly produced by hepatic cells in response to interleukin (IL)-6, but can also be synthesized at sites of inflammation by smooth muscle cells, macrophages, endothelial cells, lymphocytes, and adipocytes.²³ Besides opsonizing bacterial cell wall components, CRP binds to phosphocholines and intracellular proteins of necrotic but not apoptotic cells based on pattern recognition.^23–25 The CRP-ligand-complex activates the classical complement pathway via binding of C1q or binds to Fc receptors on granulocytes, monocytes, and macrophages leading to the secretion of proinflammatory cytokines. As a consequence of this interaction, CRP is localized to sites of cell damage and initiates an acute inflammation that is often accompanied by granulocytic infiltrates.^24,26

It remains unclear if CRP is linked to local inflammation of the bone marrow in MC1 which would further justify studies investigating CRP as a biomarker for MC1. In this study, we aimed to provide evidence for a pathophysiological link between blood CRP and bone marrow changes in MC1.

2 |. METHODS

2.1 |. Patients

This is an explorative basic research study using tissue from human subjects. Ethical approval to collect bone marrow aspirates from patients undergoing spondylodesis at the Balgrist University Hospital or the University Hospital Zurich was obtained from the Ethics Commission of the Canton of Zurich (BASF 2017–00761). Each patient gave informed consent to participate in this study. The study was conducted in accordance with the Helsinki Declaration of 1975, as revised in 2000. Inclusion criteria were (i) spinal fusion at the level of MC1, confirmed independently by two radiologists, with an MC1 lesion large enough that a pedicle screw came to lie into the MC1 lesion, (ii) a second pedicle screw in a vertebral bone marrow region without MC from the same patient. Exclusion criteria were infectious diseases, tumors, prior spinal fusion, and juvenile scoliosis. Patients were identified preoperatively based on T1-weighted and T2-weighted lumbar MRI. MRI scanner information and MRI sequences are provided in Supporting Information: Material 1. Two experienced board-certified specialized radiologists with over 8 years of experience classified Modic change lesions (e.g., predominant MC1) and graded MC1 lesion size (0–4) of the aspirate level and of all lumbar levels. They further graded disc degeneration (DD) (Pfirrmann grad 0–5) and endplate score (EPS) to assess endplate damage according to Rajasekaran et al. (grade 0–6).^27,28 Interrater agreements were reported for DD, EPS, and MC1 lesion size as interclass correlation coefficient(3,k).²⁹ Blood CRP was measured during routine preoperative clinical examination 1–2 weeks before surgery. CRP was measured as high-sensitivity CRP. Patients reported preoperative back and leg pain on a 10-point visual analogue score for back and leg pain, respectively, and filled out the Oswestry-Disability-Index questionnaire (ODI, version 2.1). Epidural and periradicular steroid injections in the past 6 months and decompression surgery were recorded.

2.2 |. Collecting bone marrow aspirates

Aspirates were taken with Jamshidi needles through the pedicle screw trajectory shortly before the screws were placed.^13,30 Intraoperative X-ray confirmed the proper position of the needle before 2–3 ml aspirates were taken. Aspirates were immediately transferred into ethylenediaminetetraacetic acid tubes in the operating room. Plasma was isolated by centrifugation (700g, 10min, 4°C), aliquoted, and frozen at −80°C within 1 h after aspiration. Bone marrow cells were lyzed in Qiazol (Qiagen).

2.3 |. Analysis of bone marrow plasma

Total protein concentration in bone marrow plasma was measured with Pierce™ BCA Protein Assay Kit (Thermo Fisher Scientific) according to manufacturer’s protocol. Bone marrow plasma CRP was measured in duplicates with CRP Human ELISA Kit (Thermo Fisher Scientific) according to manufacturer’s protocol. One patient was excluded because the coefficient of variance was above recommended variance (>20%). For comparison of MC1 and control bone marrow CRP, CRP concentrations were normalized to total bone marrow protein. Necrosis/cell damage was assessed by quantifying lactate dehydrogenase (LDH) activity using CyQUANT™ LDH Cytotoxicity assay (Thermo Fisher Scientific). IL-1β and IL-6 were measured in duplicates in 9 of the 13 patients with MesoScale U-Plex (Meso Scale Diagnostics) and normalized to total bone marrow protein. IL-1β and IL-6 could not be measured for two patients because no bone marrow plasma was left.

2.4 |. CRP gene expression of bone marrow cells

From 8 of the 13 patients, RNA was isolated from bone marrow cells using miRNeasy Mini Kit (Qiagen) according to manufacturer’s protocol. One hundred nanogram RNA was used for complementary DNA (cDNA) synthesis using SensiFAST cDNA synthesis kit (Bioline) in a 20 µl reaction according to manufacturer’s protocol. The expression of CRP was measured with quantitative real-time PCR (RT-PCR) using TaqMan primer/probes (Hs00357041_m1; Thermo Fisher Scientific). Each reaction contained 0.5 μl primer/probe, 6.25 μl SensiFAST Probe no-ROX master mix (Bioline), 4.75 μl RNase-free water, and 1 μl cDNA. After an initial polymerase activation step (2 min, 95°C), 40 cycles (10 s denaturation, 30 s annealing/extension) were run on a Mic RT-PCR system (Labgene Scientific). The expression of hypoxanthine phosphoribosyltransferase 1 (HPRT1; forward primer: 5’-AGAATGTCTTGATTGTGGAAGA-3’, reverse primer: 5’-ACCTTGACCATCTTTGGATTA-3’) was quantified as a reference gene using SensiFast No-ROX Kit and 2.5% input of total cDNA and 40 cycles of 5 s 95°C, 20 s 60°C, 10 s 72°C followed by melting curve analysis. The absence of a CRP amplification signal but the amplification of HRPT1 was considered as the absence of CRP transcription.

2.5 |. Statistics

Statistical analysis was done in R version 3.6.2 (R Core Team). A p < 0.05 was considered significant. Normal distribution was tested with the Shapiro-Wilk test. Fisher’s Exact test was used to compare the number of aspirations per level between MC1 and control. Dependent t-test and Wilcoxon test were used to compare normal and nonnormal distributed values between MC1 and control, respectively. Pearson’s r and Spearman’s ρ correlations were calculated for normal and nonnormal distributed data, respectively. Correlations were calculated to test associations (i) between blood CRP and MC1 bone marrow CRP, (ii) between blood CRP and the difference in CRP between MC1 and intrapatient control CRP (ΔCRP), (iii) between bone marrow CRP, LDH, IL-1β, and IL6, and (iv) between CRP and patient demographics. A linear regression model with an interaction term for bone marrow CRP with blood CRP and the presence of MC1 (MC1, control) was calculated to test if bone marrow CRP depends on blood CRP and the presence of MC1. T-tests were calculated comparing blood CRP or bone marrow CRP from patients taking and not taking (i) antihypertensive, (ii) insulin/antidiabetics, (iii) nonopioid analgesics, and (iv) opioids. Spearman’s correlations of blood CRP with the number of total lumbar MC1 and with the cumulative MC1 lesion size score were calculated. A correlation coefficient >0.8 was considered very strong, between 0.6 and 0.79 moderate, between 0.3 and 0.59 fair, and below 0.3 as poor.³¹

2.6 |. Role of the funding source

This study was supported by the Balgrist Foundation, the VELUX Foundation, the Baugarten Foundation, the Center for Applied Biotechnology and Molecular Medicine of the University of Zurich, and the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health under Award Number U19AR076737 and the Clinical Research Priority Program of the University of Zurich (CRPP Pain). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. All funding sources were not involved in the study design, sample and data collection, analysis and interpretation of data; in the writing of the manuscript; and in the decision to submit the manuscript for publication.

3 |. RESULTS

Baseline patient demographics are summarized in Table 1 and radiological readouts in Table 2. Interrater agreements for radiologic readouts were 0.89 for disc degeneration, 0.97 for endplate score, and 0.50 for MC1 size.

TABLE 1.

Patient demographics and blood CRP at baseline

	All subjects (n = 13)
Age, year	68.8 ± 11.3
Female	11 (84.6)
Size, cm	151 ± 34
Weight, kg	97.1 ± 37.5
Smoker	4 (30.8)
VAS.back (0–10)	7.7 ± 1.6
VAS.leg (0–10)	5.6 ± 3.8
ODI (0%−100%)	49.7 ± 17.1
CRP blood (mg/L)	4.0 [0.5–52.9]

Note: Data are shown as mean values ± standard deviation, except for female sex and smokers, where a number of individuals are reported. Data in parentheses are percentages. Data in square brackets indicate range. Abbreviations: CRP, C-reactive protein; ODI, Oswestry Disability Index; VAS.back, visual analogue score for back pain; VAS.leg, visual analogue score for leg pain.

TABLE 2.

Radiological data at aspiration levels

	MC1	Control	p value
Aspiration levels			0.153
L2	0	1
L3	0	2
L4	3	6
L5	6	3
S1	4	1
DD (0–5)	4.8 ± 0.3	4.2 ± 0.9	0.018
EPS (0–6)	5.2 ± 0.6	3.2 ± 1.5	0.001
Size (0–4)	3.0 ± 0.8

Note: Data are shown as mean values ± standard deviation, except for aspiration levels, where the number of aspirates are reported. Fisher’s exact test was used to compare the difference between MC1 and control for aspiration levels. Dependent Wilcoxon tests were used to compare DD, EPS, and size between MC1 and control.

Abbreviations: DD, Pfirrmann grade of disc degeneration; EPS, Rajasekaran endplate score for endplate damage; MC1, Modic type 1 change.

3.1 |. Blood CRP correlates with bone marrow CRP in MC1

Median blood CRP concentration was 4 mg/L (range 0.2–52.9 mg/L). All values were below 8.0 mg/L except one patient with suspected inflammatory arthritis (HLA-B27 positive). This patient was excluded from further analysis. For one patient, no blood CRP measures were available. Bone marrow CRP concentration normalized to total protein tended to be higher in MC1 compared to intrapatient control (mean difference: +0.22 mg CRP/g protein, 95% confidence interval [CI] = −0.04 to 0.47 mg CRP/g protein, p = 0.088) (Figure 1A). Blood CRP concentration correlated very strongly with normalized CRP concentration in MC1 bone marrow plasma (ρ = 0.89, p < 0.001) and strongly with normalized CRP concentration in control bone marrow plasma (ρ = 0.63, p = 0.044) (Figure 1B). A linear regression model for bone marrow CRP with blood CRP and the presence of MC1 (MC1, control) as explanatory variables revealed that there is a significant dependence of bone marrow CRP from blood CRP (p < 0.001) and that CRP in MC1 bone marrow increases stronger with increasing blood CRP (interaction term, p = 0.054). The difference of CRP concentration in MC1 to intrapatient control also correlated strongly with blood CRP (r = 0.69, p = 0.018) (Figure 1C). A blood CRP threshold of 3.2 mg/L detected with 100% accuracy increased CRP concentrations in MC1 bone marrow (red line in Figure 1C). There was poor correlation of blood CRP with the number of lumbar MC1 (ρ = −0.19, p = 0.563) and with the cumulative size scores of all lumbar MC1 lesion (ρ = −0.36, p = 0.266).

FIGURE 1.

Blood and bone marrow CRP concentrations in MC1 patients. (A) CRP concentration is higher in Modic type 1 change bone marrow compared to intrapatient control bone marrow. CRP concentrations normalized to total protein concentrations are indicated. (B) Blood CRP concentration correlates with bone marrow CRP. (C) Blood CRP concentration correlates with increased CRP in Modic type 1 change bone marrow plasma. Gray shadow indicates 95% confidence interval. A blood CRP threshold of 3.2–3.8 mg/L (red line) detects with 100% accuracy increased CRP concentrations in MC1 bone marrow. CRP, C-reactive protein; MC1, Modic type 1 change.

The CRP gene was not expressed in bone marrow cells.

3.2 |. Tissue damage in MC1 without an increase of IL-1β and IL-6

Bone marrow LDH concentration was significantly higher in MC1 than in intrapatient control bone marrow (143.1%, 95% CI [110.7%, 175.4%], p = 0.014) indicating local cell damage. Normalized concentrations of IL-6 (MC1: 93.9 ± 178.8 × 10⁹, control: 34.6 ± 20.9 × 10⁹, p = 0.287) and IL-1β (MC1: 16.6 ± 10.1 × 10⁹, control: 18.6 ± 8.2 × 10⁹, p = 0.614) were not significantly increased in MC1 versus intrapatient control bone marrow, but bone marrow CRP correlated fairly with IL-1β (ρ = 0.52, p = 0.029) and with IL-6 (ρ = 0.51, p = 0.031) (Figure 2). When MC1 and control bone marrow were analyzed separately, IL-1 correlated moderately with CRP in MC1 (ρ = 0.64, p = 0.064) and fairly in control bone marrow (ρ = 0.35, p = 0.350) (Supporting Information: Figure S-2A). IL-6 correlated very strongly with CRP in control bone marrow (ρ = 0.80, p = 0.010) and fairly in MC1 (ρ = 0.50, p = 0.173) (Supporting Information: Figure S-2B). IL-1b correlate fairly (ρ = 0.47, p = 0.287) and IL-6 strongly with lesion size (ρ = 0.78, p = 0.039). LDH correlated fairly with IL-1β (r = 0.48, p = 0.033) but poorly with IL-6 (ρ = −0.21, p = 0.352).

FIGURE 2.

Correlation of bone marrow CRP with bone marrow (A) IL- 1β and (B) IL-6. Gray areas indicate 95% confidence intervals. CRP, C-reactive protein; IL, interleukin.

3.3 |. No correlation of CRP with demographic data and 1-year outcome

There were no significant correlations between demographic data and medication with blood or bone marrow CRP. Seven patients had steroid injections within 6 months before surgery and three patients had decompression surgery at any time before fusion. Previous steroid injection and decompression surgery had no effect on blood or bone marrow CRP.

4 |. DISCUSSION

Increased serum CRP has inconsistently been associated with lumbar MC1, yet its role in MC1 pathophysiology remains unknown. Therefore, we aimed to provide evidence for the role of CRP in the tissue pathophysiology of MC1 bone marrow. We showed that blood CRP in CLBP patients with lumbar MC1 correlates with increased CRP concentrations in MC1 bone marrow and that CRP is not expressed by bone marrow cells in MC1 lesions. This suggests that hepatic CRP accumulates preferentially in MC1. Increased LDH in MC1 bone marrow indicates that MC1 are local injuries with cell damage that might trigger inflammatory reactions. While we could not find increased IL-1β and IL-6 concentration in MC1 bone marrow, bone marrow CRP correlated with IL-1β and IL-6, indicating that CRP is linked to subtle inflammatory mechanisms in MC1 bone marrow.

Our data suggest that hepatic CRP accumulates stronger in MC1 compared to control bone marrow. CRP may accumulate in the plasma fraction of MC1 aspirates because (i) soluble CRP binds to necrotic cell components in MC1 and because (ii) cell-bound CRP is shed from necrotic cells in MC1.³² We confirmed increased necrosis in MC1 with increased LDH, an intracellular protein that is released from necrotic but not apoptotic cells. The mechanisms causing cell and tissue damage in MC1 remain unknown, yet experimental endplate fractures in a rabbit ex vivo explant model have been shown to induce the release of LDH, potentially from endplate and disc cells.³³ In MC1, endplates are damaged, and the disks are degenerated (Table 2).³⁴ Our findings indicate that tissue destruction in MC1 is not limited to an initiating event but is a chronic mechanism in MC1. Segmental instability with MC1 is common and could relate to increased cell and tissue damage.³⁵

Generally, tissue necrosis triggers inflammatory cascades in surrounding cells of the innate immune system and of other immune-competent cells like fibroblasts. To further substantiate that necrosis is an important mechanism in MC1, we tested if MC1 contain increased concentrations of the proinflammatory cytokines IL-1β and IL-6 that are typically released by immune-competent cells after exposure to necrotic cell debris. While we could not find increased concentrations in MC1, the correlation of IL-1β with LDH and of IL-1β and IL-6 with bone marrow CRP, as well as lesion size with IL-6 suggest that inflammatory mechanisms in MC1 are linked to CRP and necrosis. Since IL-1β and IL-6 concentrations were not increased in MC1, these cytokines have most likely no pathomechanistic relevance in MC1.

The notion that MC1 are inflammatory changes is based on the edema-like signal on MRI and the analysis of disks adjacent to MC1. However, the edema-like MRI signal relates to increased vascularization and interstitial fluid and does not indicate an increased concentration of proinflammatory cytokines.³⁶ The increased secretion of IL-6, IL-8, and PGE2 from cultured disc cells adjacent to MC1 has been reported in one study,³⁷ but could not be replicated with transcriptomic analysis of uncultured disc cells from MC1 levels.^13,38 Inconsistent results could also be attributed to the dynamic nature of MC and the timepoint of analysis; inflammatory phases in MC1 could be followed by consolidating phases with healing attempts and fibrotic mechanisms.³⁰ Recent in-depth analyses of MC1 bone marrow provide now new evidence that the hematopoietic elements in MC1 are indeed in an inflammatory state. Activation and accumulation of neutrophils and T-cells have been reported.³⁹ If these mechanisms include the release of proinflammatory cytokines and if they relate to CRP and necrosis remains unknown. Together, there is currently limited molecular evidence for inflammation in MC1.

Not surprisingly, there is no evidence for systemically increased inflammatory cytokines in MC1.^16,40 In addition, any subtle change of proinflammatory cytokines in MC1 may be overlaid with other local and systemic proinflammatory mechanisms in CLBP patients.⁴¹ It might be a more promising strategy to search for MC1 serum biomarkers that relate to pathophysiological changes in MC1 which are not present in other CLBP phenotypes. For example, serum biomarkers that relate to fibrosis were reported to be increased in CLBP patients with MC1 compared to CLBP patients without MC.⁴² Metabolic factors have also been identified as a causative factors for MC.⁴³ This indicates that MC not only relates to local pathomechan- isms but that also systemic factors play an important role.

In our cohort, a blood CRP concentration of >3.2 mg/L was specific for increased CRP in MC1 bone marrow compared to intra-patient control bone marrow, but we could not find a correlation of blood CRP with the total number or with the cumulative size score of lumbar MC1 lesions. This indicates that CRP may play a role in pathophysiology within the MC1 bone marrow but less on a systemic level. Many factors, that we could not control for, may have influenced the blood CRP level, for example, fasting, activity, smoking, BMI, and obscured an association of blood CRP with MC1 lesion size. Rannou et al.¹⁴ reported that MC1 patients had an average CRP concentration of 4.64 ±3.09 versus 33 ± 0.77 mg/L in CLBP patients without MC1. Two other studies reported no difference in blood CRP between patients with and without MC. However, they did not stratify MC for subtypes and they did not report lesion size.^15,16 Larger validation studies including appropriate control groups are required to test and validate CRP as a biomarker for MC1. It should be investigated which blood CRP concentration has the highest predictive power for MC1, how robust CRP is to changes over time and with respect to other comorbidities (e.g., osteoarthritis), and if clinical parameters correlate with CRP and MC1. In addition, longitudinal studies showing a correlation of CRP with MC1 lesion size are necessary to prove the specificity of CRP for MC1.

A limitation of this study is the limited sample size. Taking bone marrow biopsies from MC1 patients requires lesions that are large enough to place the bone marrow needle with high certainty into the MC lesion. We only took biopsies from lesions where we were sure that the needle came to lie in MC lesions. This excluded many patients. Biological variation in the analyzed samples increased type II errors and may have caused insignificant results of true tissue changes. For example, increased CRP in MC1 compared to intrapatient control bone marrow did not reach significance (p = 0.089) at significance level of α = 0.05. Another limitation of this study is that blood CRP did not correlate with pain and disability and hence does not reflect the clinical severity of MC1. The lack of a non-MC control group in this study impedes any conclusion about the utility of CRP as a biomarker. Despite these limitations, the association of CRP with inflammatory and necrotic changes in MC1 bone marrow provides evidence for a pathophysiological role of CRP in MC1 bone marrow.