Abstract
Disability and distress caused by chronic low back pain (LBP) lacking clear pathoanatomical explanations cause huge problems both for patients and society. A subgroup of patients has Modic changes (MC), identifiable by MRI as vertebral bone marrow lesions. The cause of such changes and their relationship to pain are not yet understood. We explored the pathobiology of these lesions using profiling of gene expression in blood, coupled with an edema-sensitive MRI technique known as short tau inversion recovery (STIR) imaging. STIR images and total RNA from blood were collected from 96 patients with chronic LBP and MC type I, the most inflammatory MC state. We found the expression of 37 genes significantly associated with STIR signal volume, ten genes with edema abundancy (a constructed combination of STIR signal volume, height, and intensity), and one gene with expression levels significantly associated with maximum STIR signal intensity. Gene sets related to interferon signaling, mitochondrial metabolism and defense response to virus were identified as significantly enriched among the upregulated genes in all three analyses. Our results point to inflammation and immunological defense as important players in MC biology in patients with chronic LBP.
Similar content being viewed by others
Introduction
Activity limitation and work absence caused by low back pain (LBP) has emerged as an enormous economic, societal, and social burden globally1,2. A large proportion of LBP patients (10–70%) will develop chronic LBP, defined as duration of pain for longer than 3 months3,4. Despite considerable research efforts, LBP remains poorly understood and has in most cases unknown pathoanatomical cause5.
Increasing attention, both in research and clinical settings, has been directed towards patients with Modic changes (MC) as a subgroup among LBP patients6. MC are vertebral bone marrow lesions identifiable by magnetic resonance imaging (MRI) first described in the late 80’s7,8. An association between MC and LBP has been suggested9,10, but conclusions after several studies are inconsistent11. Three types of MC are distinguished by their intensity on T1- and T2-weighted MRI. MC type I is considered to represent the most inflammatory state12,13 with possible association to LBP14. The extent of edema in MC of type I can be evaluated using short tau inversion recovery (STIR). This MRI technique suppresses the signals from fat, allowing for a more sensitive detection of edema signals, and has been suggested as a supplement to the routinely used T1- and T2-weighted scans15.
The etiology behind MC is unknown. Prevailing theories emphasize an immunological component, involving autoimmunity or low-grade bacterial infections in the intervertebral disc, triggering a detrimental cascade of inflammation in the vertebrae16. To advance in the search for biological explanations and treatment options to subjective conditions like unspecific back pain, unbiased endpoints can be useful. Analysis of gene expression profiles from blood reflects systemic processes, potentially shedding light on the immunological state and underlying pathology of a disorder. In this study, we aimed at improving the pathobiological understanding of MC type I related STIR signal increases in patients with chronic LBP, using whole blood gene expression profiling and pathway analysis.
Materials and methods
Study cohort
The study is based on baseline data from a sub-sample of patients with chronic LBP and MC included in a randomized controlled trial comparing amoxicillin to placebo (the Antibiotics in Modic changes (AIM) study, n = 180)17. The patients had suffered from LBP with a duration of more than 6 months and had MC type I or II in the vertebral body at the same disc level as a lumbar disc herniation verified on MRI within the preceding 2 years. Patients were excluded from the study if a specific diagnosis could explain their low back symptoms (e.g., tumor or spinal stenosis), if they had undergone surgery for disc herniation the previous 12 months or had former low back surgery for reasons other than disc herniation. Eligibility criteria and the study protocol for the AIM study is fully published elsewhere18.
Out of 118 patients with MC type 1, we included in this study patients of Caucasian ethnicity with successful blood sampling at screening (n = 96).
Blood samples and STIR images were collected for all patients before randomization and start of intervention in the AIM study. Data regarding age, gender, education, body mass index (BMI), ethnicity, smoking habits, work and physical activity habits were gathered by trial care givers (medical doctors, physiotherapists or study nurses). Additionally, information was collected on back pain history, current disability (Roland-Morris Disability Questionnaire (RMDQ), 0–24, where 0 corresponds to no disability and 24 corresponds to maximum disability19), emotional distress (Hopkins Symptom Checklist-25, where values ≥ 1.75 associate with psychiatric diagnosis20) and fear-avoidance beliefs (Fear-Avoidance Beliefs Questionnaire (FABQ), where higher values indicate more fear avoidance beliefs21).
Written informed consent was obtained from all patients in accordance with the Helsinki Declaration, and the study was approved by the Regional Ethics Committee in South East Norway (project 2014/158).
MRI
MRI was performed at the local study sites using the same MRI protocol and scanner type (Avanto 1.5 T, Siemens). The imaging included standard T1- and T2 weighted fast spin echo images and sagittal STIR images, as detailed previously22.
MC were defined as type I (hypointense T1 signal and hyperintense T2 signal), type II (hyperintense T1 signal and iso/hyperintense T2 signal) and type III (hypointense T1 and T2 signal). MC with diameter ≤ 5 mm or height < 10% of vertebral body height did not qualify the patient for inclusion. Patients with MC type I at the endplate superior and/or inferior to the previously herniated disc level were included in the present study also if they had other MC types at the same or a different endplate. Disc herniation was defined as a focal displacement of disc material, i.e. involved < 25% of the disc circumference23. In this study, disc herniation was further classified as protrusion (if its largest diameter was smaller than its base) or extrusion (base smaller than diameter—or sequestration)23. Disc degeneration (DD) was graded using Pfirrmann grades, which range from 1 (no DD) to 5 (highest degree of DD)24.
MC related STIR signal increase was assessed at 12 endplates (Th12-S1)22. Three endpoint variables were defined prior to analysis: (I) Total volume of STIR signal increase (henceforth referred to as “STIR volume”). Each endplate was given a score based on the volume of increased STIR signal in percent out of the total vertebral body volume (0: no STIR signal, 1: < 10%, 2: < 25%, 3: 25–50%, 4: > 50%). The total sum score for the 12 endplates was reported (possible values 0–48). (II) Maximum STIR signal intensity (henceforth “STIR intensity”). The maximum intensity of MC related STIR signal increase at any endplate, reported as a percentage on an intensity scale ranging from normal vertebral body intensity (0%) to cerebrospinal fluid intensity (100%) (possible values 0–100%). (III) «STIR composite» was pre-constructed to categorize the patients according to the abundancy of edema at the level of the previously herniated disc25. STIR composite 1 represented MC related STIR signal increase with largest volume of < 25% and maximum intensity < 25%. STIR composite 3 represented STIR signal increase with volume ≥ 25% and height > 50% of vertebral body and maximum intensity ≥ 25% and presence at both sides of the disc. Patients in neither group were categorized as STIR composite 2.
Two radiologists independently evaluated the MRI images. Their inter-rater agreement was mostly very good for presence of MC type I (kappa 0.78–0.85 depending on level) and presence of STIR signal increase (0.83–0.90), and moderate for volume of STIR signal increase (0.47–0.61)22 and DD grades 3–4 vs 5 (0.44; grades 1–2 not reported). In the case of disagreements, they negotiated a conclusive finding (for presence of MC type I and DD grade) or a third radiologist was consulted, and the majority rating was used (for STIR findings). For the intensity measurements, the mean of two radiologists’ values was used. To further describe the samples, one experienced radiologist classified disc herniations as protrusions or extrusions.
The correlation between the STIR variables was examined using Pearson’s correlation coefficient (r).
Blood sample collection and treatment
Peripheral blood was sampled from each patient in Tempus Blood RNA Tubes (Thermo Fisher Scientific, Waltham, MA, USA), on average 14 days before the MRI (max 48 days). Simultaneously, blood samples were collected in EDTA tubes for 5-part differential blood cell count using flow cytometry locally at each study center, giving the relative content of neutrophils, lymphocytes, basophils, eosinophils, and monocytes in the samples. The relative blood cell counts were tested for correlation with the STIR variables.
RNA isolation, preparation and sequencing
Total RNA was isolated from the Tempus Blood RNA Tubes using the Preserved Blood RNA Purification Kit I (Norgen Biotek, Thorold, Canada) according to the manufacturer’s instruction. DNAse treatment was carried out as recommended. The quality and concentration of the RNA were measured using the BioAnalyzer 6000 Nano kit (Agilent Technologies, Santa Clara, CA, USA) and Qubit RNA HS (Thermo Fisher Scientific, Waltham, MA, USA).
Ribosomal RNA (rRNA) and globin transcripts dominate the RNA pool of whole blood samples. rRNA provides negligible relevant information about the transcriptome, and the globin transcripts were not expected to be relevant in this study. The total RNA samples were therefore depleted for rRNA and globin transcripts to maximize the number of informative reads retrieved from high-throughput sequencing. The first 82 of the 96 RNA samples were depleted for rRNA and globin transcripts using the Globin-Zero Gold rRNA Removal Kit (Illumina, San Diego, CA, USA), before cDNA library preparation using the strand-specific TruSeq RNA-seq library prep (Illumina, San Diego, CA, USA). The depletion kit was discontinued and replaced by a kit combining the ribosomal- and globin depletion with the cDNA library preparation. The last 14 samples were therefore depleted and prepared using the new TruSeq Stranded Total RNA with Ribo-Zero Globin kit (Illumina, San Diego, CA, USA).
The depleted RNA samples were sequenced with 2 × 75 bp paired-end configuration on the HiSeq3000 platform (Illumina). The quality of the sequencing was assessed using FastQC and Qualimap26,27.
Bioinformatics and analysis of differential gene expression
The raw sequencing data was trimmed for low quality reads and adapter sequences using Trim Galore! v0.4.528. The reads were mapped to the human genome (GRCh38.p10) using HISAT2 v2.1.029 and counted with featureCounts from Subread v1.6.330, using gene coordinates from Ensembl 8831. Only autosomal genes with a read count of more than three in at least five subjects were kept for further analyses.
Analysis of differential expression was performed with the R package DESeq2 v1.32.032. The DESeq2 pipeline includes normalization of the samples by sequencing depth and RNA composition, using the implemented median of ratios method. The differential expression analysis is done by fitting a negative binomial model to each gene and applying a Wald test to the log2 fold change between groups.
We tested for associations between gene expression and the three STIR variables separately (STIR volume, STIR intensity and STIR composite), adjusting for sequencing batch effects, age and gender. The analyses were repeated, adjusting for white blood cell proportions. The p-value distributions were estimated empirically from the z-scores reported by DESeq2, using the R package fdrtool v1.2.1533, controlling the false discovery rate (FDR) with the Benjamini–Hochberg method34.
Gene set enrichment analysis
To explore whether the patients’ gene expression profiles reflect the deregulation of certain biological pathways, we performed gene set enrichment analysis on the ranked gene lists from the differential expression analysis. We used publicly available enrichment map gene sets from the Bader lab (version from March 01, 2020) containing both manually curated and computationally generated gene sets35. 23 custom autoimmunity gene sets were created from Genome-wide association study (GWAS) data (Table S1), and a total of 18,577 gene sets were used as input to the GSEA analyses. 5474 gene sets were left for the analyses after gene set size filtering.
The GSEA v4.1.0 software from UC San Diego and Broad Institute36,37 was used for the gene set enrichment analyses. GSEA searches for pathways whose genes are significantly enriched at the top or bottom of a given ranked gene list. The input gene lists contained all genes from the differential expression analyses, ranked according to significance and odds ratio: The most significantly upregulated genes found on top of the list, and the most significantly downregulated genes on the bottom. Following general recommendations, gene sets containing less than 15 or more than 200 genes were filtered out from the analyses38. The GSEA results were visualized using Cytoscape v3.8.239 with apps EnrichmentMap v3.2.135 and AutoAnnotate v1.3.340. The enrichment map was created with parameters FDR q < 0.05, and Jaccard Overlap combined coefficient > 0.375 with combined constant = 0.5.
Interferon-regulated genes
We investigated whether the significantly differentially expressed genes from our analyses were recognized as interferon-regulated genes using the Interferome database v2.0141. The database contains data from microarray experiments of murine or human cells treated with interferon in vivo or in vitro, summarizing observed effects of interferon stimulation on the expression of genes. The search was restricted to experiments on normal blood (i.e., without a known genetic condition, disease state or pre-treatment), using the differentially expressed genes significant at FDR < 0.05 from all analyses.
Results
Patient characteristics and STIR variables
The present study included 96 patients with LBP and MC type I (age between 27 and 63 years, 61 women). The patients had an average LBP intensity of 6.5 on a 0–10 numerical rating scale (NRS) and LBP lasting for a median of 3 years (Table 1). Most patients (92 of 96) had DD grade 4 or 5. The distribution of the STIR volume, STIR intensity and STIR composite variables in the study sample is illustrated in Fig. 1. The measured ranges of STIR volume and intensity were 1–8 and 9–73, respectively. There was weak to moderate correlation between the variables, with the strongest correlation found between STIR volume and STIR composite (0.68, p = 1.6 × 10–14).
Cell type proportions
All patients had either normal range or minutely increased or decreased blood cell counts, with no suspicion of underlying infection or pathology/comorbidities (Figure S1). None of the cell type proportions correlated significantly with any of the three STIR variables (Figure S2).
Quality control of gene expression data
The isolated total RNA had a mean RNA integrity number (RIN) of 9.2 and concentration of 157 ng/μl. After RNA sequencing, 96.6% of the reads were aligned to the reference genome, and 53.4% of these were assigned to genes. The unassigned proportion was mostly multimapped or ambiguously mapped reads.
Lowly expressed genes were filtered out, removing less credible counts and reducing the number of tests in the analyses. Reads from 25,944 genes (58% coding for proteins) represented by more than three counts in at least 5 samples were retained. The two different globin depletion kits used during the sequencing library preparation caused a small batch effect in the gene expression profiles (Figure S3). This was accounted for through the flow cell batch effect adjustment in the differential expression analyses.
Differential expression analyses
In total 42 genes were found to be significantly up- or downregulated in association with the STIR variables (Fig. 2). The majority (45%) were protein coding genes. The rest were mainly long non-coding RNA (29%), other non-coding RNAs (14%) and pseudogenes (10%) (Figure S4).
We found 37 genes with expression levels significantly associated with STIR volume. One gene was associated with STIR intensity, while ten genes had differential expression associated with the STIR composite variable (FDR < 0.05, Table S2–4). The corresponding numbers of significantly associated genes using FDR < 0.10 were respectively 87, 3 and 21.
The most significantly differentially expressed protein coding genes resulting from the analysis of STIR volume, were GLYATL2 (adjusted p-value (padj) = 6.3e−04), IFI27 (padj = 9.9e−04), MYO18B (padj = 5.6e−03) and APOLD1 (padj = 7.7e−03). GLYATL2 correlated negatively, while IFI27, MYO18B and APOLD1 correlated positively with STIR volume. In the analysis of the composite STIR variable, GLYATL2 (padj = 2.4e−08), NPTX1 (padj = 6.3e−07) and MYL9 (padj = 3.8e−06) were the most significantly differentially expressed genes, all three correlating negatively with increasing STIR composite value. The only gene significantly differentially expressed in the analysis of STIR intensity was GLYATL2 (padj = 0.028).
There were overlapping results across the analyses (Fig. 3). GLYATL2 was significant in all three analyses, while IFI27, RNU1-2, RSAD2 and RAVER2 were significantly differentially expressed in both the analysis of STIR volume and STIR composite (Figure S5).
Gene set enrichment analyses
We next investigated whether certain biological pathways were enriched (FDR < 0.05) in the differentially expressed genes. Genes from 59 gene sets were found significantly enriched among the STIR volume associated upregulated genes (Table S5). The most strongly enriched pathways were related to interferon signaling and response (Fig. 4). Additionally, pathways involved in mitochondrial metabolism and responses to virus were among the enriched gene sets.
In the analysis of maximum STIR signal intensity, 775 gene sets were found significantly enriched among the up- and downregulated genes (Table S6). 235 of these were enriched among the upregulated genes. The strongest enrichment was again observed for interferon response and mitochondrial metabolism, as seen in the analysis of STIR volume, in addition to cell cycle related gene sets. 540 gene sets were enriched among the downregulated genes. Interleukin and cytoskeleton related pathways were among the most strongly enriched gene sets.
In the gene set enrichment analysis of STIR composite, 144 gene sets were enriched among the upregulated genes, none among the downregulated genes (Table S7). Immune-related gene sets predominated the list, together with mitochondrial metabolism (as observed in the previous two analyses), cell cycle and antigen presentation.
Exploring the overlap between the different STIR variables, 36 gene sets were found significantly enriched in all three gene set enrichment analyses (Fig. 4). More details can be found in Table S8.
Cell type proportion adjustment
To investigate whether differences in the patients’ white blood cell proportions affect our results, we repeated all analyses adjusting for the flow cytometry-measured white blood cell proportions. The analysis results from the two approaches were mostly overlapping, albeit with fewer significant findings when adjusting for cell type proportions (Figure S5).
Interferon-regulated genes
As interferon signaling was the pathway most significantly enriched among the differentially expressed genes, we investigated whether the significantly differentially expressed genes from our analyses were recognized as interferon-regulated genes in the Interferome database.
Nine of the in total 42 significantly differentially expressed genes (FDR < 0.05) were reported as observed up- or downregulated by interferon-α, β or γ with a fold change of at least 2 (Table 2).
Discussion
To our knowledge, this is the first study exploring how MC-related STIR findings correlate with gene expression levels in patients with chronic LBP and MC type 1. Overall, the results implicated gene sets relevant to inflammation and immunological defense.
We identified multiple genes with expression levels significantly associated with the STIR signals. Several of the significant genes encode proteins involved in processes of potential importance to the development of MC.
Across all analyses, GLYATL2 was significantly lower expressed with increasing STIR signals. This gene codes for Glycine N-acyltransferase-like 2, a protein involved in the synthetizing of N-acyl glycines42. Among the wide range of functions of these signaling molecules, they are believed to contribute to the resolution of inflammation and have pain-reducing effects43,44. A lower expression of this gene might therefore be connected to the chronicity of inflammation observed in MC patients. It is however to be noted that the association observed in our data could be driven by a very high expression of this gene in a few patients.
Three protein coding genes were additionally significantly differentially expressed both in the STIR volume and STIR composite analyses (RAVER2, IFI27 and RSAD2). This overlap could partly be explained by the moderate correlation observed between the two STIR variables (Pearson's r = 0.68), but might also indicate that these are more robust findings. RAVER2, coding for Ribonucleoprotein, PTB Binding 2, is involved in alternative splicing regulation45. SNPs in this gene have been associated with Ulcerative colitis, an inflammatory bowel disease that causes long-lasting inflammation and ulcers in the digestive tract, suggesting a role for this gene in autoimmunity46. IFI27, coding for Interferon Alpha Inducible Protein 27, and RSAD2, coding for Radical S-Adenosyl Methionine Domain Containing 2, are both reported to be interferon-regulated genes. IFI27 is implicated in cell apoptosis induced by DNA damage47, while RSAD2 is demonstrated to be involved in a wide range of innate immune and antiviral responses48.
Among the non-overlapping results, the most significant finding in the differential expression analysis of STIR volume was the negative correlation of CASC20. This is a long non-coding RNA which function is still unknown, but variants in this gene have through GWAS been associated with height and body shape49,50. This is interesting, as related characteristics like BMI has been associated to MC51. Other genes with expression levels associated to STIR volume were MYO18B, which have been reported as relevant for muscle homeostasis and movement52, and APOLD1, coding for Apolipoprotein L domain containing 1, a protein thought to regulate endothelial cell signaling and vascular function in association with angiogenesis53. This finding agrees with the concept of vascular development in the subchondral bone marrow being a hallmark of MC type 154.
In addition to GLYATL2, the gene with expression most significantly associated with the STIR composite variable was NPTX1, coding for Neuronal Pentraxin 1, a secreted protein involved in neural development and remodeling55. Neuronal Pentraxin 1 has been shown to regulate mitochondria-driven neuronal injury in mice, while knocking down the gene provide neuroprotection56. This might be noteworthy, as an increased nerve-ingrowth has been demonstrated in vertebral endplates with MC compared to normal endplates57.
There were no other significantly differentially expressed genes aside from GLYATL2 associated to the STIR intensity variable. This low number of significant associations could be explained by the nature of the variable. This variable is a point value representing maximum observed STIR intensity and does therefore not reflect overall spread or amount of edema.
It should be emphasized that the genes implicated through expression levels associated with STIR variables in the present study need, in addition to replication, to be functionally studied, also at the protein level, in order to establish a biological role in MC type 1.
In addition to the previously mentioned interferon-regulated genes, the GSEA analyses revealed genes relevant to pathways of both interferon alpha (IFN-α) and interferon gamma (IFN-γ) responses to be significantly enriched among the differentially expressed genes. The interferons are signaling molecules with diverse roles in immunological defense and inflammation. Interferons have shown to mediate the activation and development of a range of immune cells and are involved in several autoimmune diseases58. Additionally, interferons have emerged as important players in bone metabolism, as studies from the field of osteoimmunology demonstrate how these molecules interfere with osteoclast differentiation59. Our study of MC affected vertebrae join a range of studies emphasizing the role of interferon dysregulation in spine conditions. In disc samples from patients with degenerated and herniated discs, a significant correlation was found between Pfirrmann DD grade (ranging from 2 to 5) and the gene expression of IFN-α160. Through gene expression profiling of degenerative versus non-degenerative disc cells, interferon signaling has been shown to be the primarily dysregulated pathway. Upregulation of IFN-α-induced genes is hypothesized to accelerate disc degeneration through a negative regulation of the cell cycle leading to decreasing disc cell numbers61,62. When mechanically inducing degeneration of bovine intervertebral discs, signaling molecules including IFN-γ were found upregulated. These molecules were thought to drive activation of microglia in the spinal cord, increasing the upregulation of other neuroinflammatory markers, driving further disc degeneration63.
One of the principal theories concerning the development of MC involve autoimmunity. Disc material is considered immune privileged, and it is shown that nucleus pulposus cells in contact with immune cells from the bone marrow can trigger inflammatory autoimmune processes and MC64. Interferons and interferon-regulatory factors have been associated to a range of autoimmune conditions58, and interestingly, IFN-γ has been found highly expressed in nucleus pulposus tissue after disc herniation65, and in higher concentration in herniated discs with extruded nucleus pulposus tissue than in intact herniated discs66. MRI has moderate accuracy compared to surgery for whether disc herniations extrude through the annulus fibrosus and the posterior longitudinal ligament67. Still, it should be noted that many disc herniations in our sample were shaped as extrusions on MRI (Table 1).
Interferons might also be of importance for the development of chronic pain, which is suggested both for sciatica68 and neuropathic pain69. Levels of in vitro production of IFN-γ from blood have been positively correlated with pain scores in patients with nonspecific acute, but not chronic LBP58. Similarly in another study, IFN-γ immunoreactivity was the marker most consistently identified in discs from patients with suspected discogenic pain compared to control subjects70.
Interferons recognized by their cognate receptors trigger signaling pathways leading to the activation of transcription of interferon-regulated genes. The outcome of the interferon response is determined by the nature and combination of these genes. All our patients had a disc herniation verified during the last two years before inclusion. Considering the aforementioned earlier findings of increased interferon expression after disc herniation, our observations may therefore be a result of (1) an increased interferon expression after disc herniation (that is not observable in our data as we did not have a control group without disc herniation) and (2) a subsequent deregulation of interferon-regulated genes, stimulating further inflammation, MC development and STIR signal increases. Nine of the in total 42 significantly differentially expressed genes found in our analyses were reported as interferon-regulated in the Interferome database. Seven of these showed the same direction in expression difference as expected after an interferon stimulation, further supporting this proposed mechanism.
The other gene sets that were found significantly enriched among the differentially expressed genes across all STIR variables were pathways related to defense response to virus and mitochondrial metabolism. The involvement of viruses in the degeneration and herniation of discs have been suggested, but studies show inconclusive results71. However, the genes underlying the defence to virus pathway have rather broad immune and inflammatory functions also occurring without a viral trigger.
The involvement of mitochondria, however, have more support. The dynamic structure and metabolism of mitochondria are known to affect the fate of cells, including regulating immune cell fate during immune responses72. Mitochondrial fission and fusion events where single mitochondria are separated into several structures or joined to larger ones, are of importance for T cell development toward memory or effector phenotypes73. Dysfunction in the dynamics of these events can also lead to the accumulation of damaged mitochondria, inducing oxidative stress and overproduction of reactive oxygen species (ROS). Such ROS are intracellular mediators that regulate antimicrobial and inflammatory responses. When dysregulated they can aggravate immune reactions and cause chronic inflammation74. Additionally, disc degeneration is thought to involve mitochondrial dysfunction, as fine-tuned cellular bioenergetics is fundamental to maintain matrix homeostasis in the intervertebral disc75.
Different cell types have different gene expression profiles, and thus the cell type proportions in blood samples could highly impact downstream gene expression analysis. As there were no correlations between the cell type proportions and the STIR variables, we did as expected not observe huge changes in the results when repeating the analyses adjusting for cell type proportions. Correspondingly, the GSEA results were also similar, but with fewer overlapping pathways between the three different STIR variable analyses. Adding more variables to the analysis model reduces the statistical power to detect significant associations, explaining the general reduction in significant findings.
Although the findings of this study were performed under fairly strict FDR control, we acknowledge the need for replication in other cohorts and in-depth studies to elucidate the biological mechanisms. In the present study, we chose to include only patients with MC type I, as this MC type is considered to mirror an active inflammatory state. The results would likely differ in a cohort with other MC types. This study was also restricted to patients with prior disc herniation, most of whom had marked DD. Our study did not include a control group without LBP, MC or disc herniation, as the participants were derived from a randomized controlled trial. A control group could have provided insight into whether genes were up-or down- regulated compared to a healthy state, and also whether our findings were driven by specific spine pathologies or were generalizable to the presence of STIR edema regardless of diagnosis.
In summary, we have demonstrated an association between STIR signals and systemic gene expression in patients with LBP and MC. Our results suggest that immunological processes, particularly interferon signaling, are important players in the underlying pathobiology of these patients.
Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request in accordance with local registration and ethical approval.
References
Global, regional, and national disability-adjusted life-years (DALYs) for 315 diseases and injuries and healthy life expectancy (HALE), 1990–2015: A systematic analysis for the Global Burden of Disease Study 2015. Lancet (London, England) 388, 1603–1658, doi:https://doi.org/10.1016/s0140-6736(16)31460-x (2016).
Hagen, K. B. & Thune, O. Work incapacity from low back pain in the general population. Spine 23, 2091–2095 (1998).
Manchikanti, L., Singh, V., Falco, F. J., Benyamin, R. M. & Hirsch, J. A. Epidemiology of low back pain in adults. Neuromodulation 17(Suppl 2), 3–10. https://doi.org/10.1111/ner.12018 (2014).
Itz, C. J., Geurts, J. W., van Kleef, M. & Nelemans, P. Clinical course of non-specific low back pain: A systematic review of prospective cohort studies set in primary care. Eur. J. Pain (Lond., Engl.) 17, 5–15. https://doi.org/10.1002/j.1532-2149.2012.00170.x (2013).
Maher, C., Underwood, M. & Buchbinder, R. Non-specific low back pain. Lancet (Lond., Engl.) 389, 736–747. https://doi.org/10.1016/s0140-6736(16)30970-9 (2017).
Brinjikji, W. et al. MRI findings of disc degeneration are more prevalent in adults with low back pain than in asymptomatic controls: A systematic review and meta-analysis. AJNR Am. J. Neuroradiol. 36, 2394–2399. https://doi.org/10.3174/ajnr.A4498 (2015).
de Roos, A., Kressel, H., Spritzer, C. & Dalinka, M. MR imaging of marrow changes adjacent to end plates in degenerative lumbar disk disease. AJR Am. J. Roentgenol. 149, 531–534. https://doi.org/10.2214/ajr.149.3.531 (1987).
Modic, M. T., Steinberg, P. M., Ross, J. S., Masaryk, T. J. & Carter, J. R. Degenerative disk disease: Assessment of changes in vertebral body marrow with MR imaging. Radiology 166, 193–199. https://doi.org/10.1148/radiology.166.1.3336678 (1988).
Zhang, Y. H., Zhao, C. Q., Jiang, L. S., Chen, X. D. & Dai, L. Y. Modic changes: A systematic review of the literature. Eur. Spine J. 17, 1289–1299. https://doi.org/10.1007/s00586-008-0758-y (2008).
Jensen, T. S., Karppinen, J., Sorensen, J. S., Niinimaki, J. & Leboeuf-Yde, C. Vertebral endplate signal changes (Modic change): A systematic literature review of prevalence and association with non-specific low back pain. Eur. Spine J. 17, 1407–1422. https://doi.org/10.1007/s00586-008-0770-2 (2008).
Herlin, C. et al. Modic changes-Their associations with low back pain and activity limitation: A systematic literature review and meta-analysis. PLoS ONE 13, e0200677. https://doi.org/10.1371/journal.pone.0200677 (2018).
Braithwaite, I., White, J., Saifuddin, A., Renton, P. & Taylor, B. A. Vertebral end-plate (Modic) changes on lumbar spine MRI: Correlation with pain reproduction at lumbar discography. Eur. Spine J. 7, 363–368 (1998).
Kuisma, M. et al. A three-year follow-up of lumbar spine endplate (Modic) changes. Spine 31, 1714–1718. https://doi.org/10.1097/01.brs.0000224167.18483.14 (2006).
Mera, Y. et al. Association between types of Modic changes in the lumbar region and low back pain in a large cohort: The Wakayama spine study. Eur. Spine J. 30, 1011–1017. https://doi.org/10.1007/s00586-020-06618-x (2021).
Fields, A. J. et al. Measuring and reporting of vertebral endplate bone marrow lesions as seen on MRI (Modic changes): Recommendations from the ISSLS Degenerative Spinal Phenotypes Group. Eur. Spine J. 28, 2266–2274. https://doi.org/10.1007/s00586-019-06119-6 (2019).
Albert, H. B. et al. Modic changes, possible causes and relation to low back pain. Med. Hypotheses 70, 361–368. https://doi.org/10.1016/j.mehy.2007.05.014 (2008).
Braten, L. C. H. et al. Efficacy of antibiotic treatment in patients with chronic low back pain and Modic changes (the AIM study): Double blind, randomised, placebo controlled, multicentre trial. BMJ 367, l5654. https://doi.org/10.1136/bmj.l5654 (2019).
Storheim, K. et al. Antibiotic treatment In patients with chronic low back pain and Modic changes (the AIM study): Study protocol for a randomised controlled trial. Trials 18, 596. https://doi.org/10.1186/s13063-017-2306-8 (2017).
Grotle, M., Brox, J. I. & Vollestad, N. K. Cross-cultural adaptation of the Norwegian versions of the Roland-Morris disability questionnaire and the Oswestry disability index. J. Rehabil. Med. 35, 241–247 (2003).
Sandanger, I. et al. Concordance between symptom screening and diagnostic procedure: The Hopkins Symptom Checklist-25 and the Composite International Diagnostic Interview I. Soc. Psychiatry Psychiatr. Epidemiol. 33, 345–354. https://doi.org/10.1007/s001270050064 (1998).
Waddell, G., Newton, M., Henderson, I., Somerville, D. & Main, C. J. A fear-avoidance beliefs questionnaire (FABQ) and the role of fear-avoidance beliefs in chronic low back pain and disability. Pain 52, 157–168. https://doi.org/10.1016/0304-3959(93)90127-b (1993).
Kristoffersen, P. M. et al. Short tau inversion recovery MRI of Modic changes: A reliability study. Acta Radiol. Open 9, 2058460120902402. https://doi.org/10.1177/2058460120902402 (2020).
Fardon, D. F. et al. Lumbar disc nomenclature: version 2.0: Recommendations of the combined task forces of the North American Spine Society, the American Society of Spine Radiology and the American Society of Neuroradiology. Spine J. 14, 2525–2545. https://doi.org/10.1016/j.spinee.2014.04.022 (2014).
Pfirrmann, C. W., Metzdorf, A., Zanetti, M., Hodler, J. & Boos, N. Magnetic resonance classification of lumbar intervertebral disc degeneration. Spine 26, 1873–1878. https://doi.org/10.1097/00007632-200109010-00011 (2001).
Kristoffersen, P. M. et al. Oedema on STIR modified the effect of amoxicillin as treatment for chronic low back pain with Modic changes-subgroup analysis of a randomized trial. Eur. Radiol. https://doi.org/10.1007/s00330-020-07542-w (2020).
Okonechnikov, K., Conesa, A. & Garcia-Alcalde, F. Qualimap 2: Advanced multi-sample quality control for high-throughput sequencing data. Bioinformatics (Oxford, England) 32, 292–294. https://doi.org/10.1093/bioinformatics/btv566 (2016).
Andrews, S. FastQC: A Quality Control Tool for High Throughput Sequence Data [Online]. Available online at: https://www.bioinformatics.babraham.ac.uk/projects/fastqc/. (2010).
Andrews, S. Trim Galore! Available online at: https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/, 2012).
Kim, D., Langmead, B. & Salzberg, S. L. HISAT: A fast spliced aligner with low memory requirements. Nat. Methods 12, 357–360. https://doi.org/10.1038/nmeth.3317 (2015).
Liao, Y., Smyth, G. K. & Shi, W. featureCounts: An efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics (Oxford, England) 30, 923–930. https://doi.org/10.1093/bioinformatics/btt656 (2014).
Cunningham, F. et al. Ensembl 2019. Nucleic Acids Res. 47, D745–D751. https://doi.org/10.1093/nar/gky1113 (2018).
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550. https://doi.org/10.1186/s13059-014-0550-8 (2014).
Strimmer, K. fdrtool: A versatile R package for estimating local and tail area-based false discovery rates. Bioinformatics (Oxford, England) 24, 1461–1462. https://doi.org/10.1093/bioinformatics/btn209 (2008).
Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: A practical and powerful approach to multiple testing. J. Roy. Stat. Soc.: Ser. B (Methodol.) 57, 289–300. https://doi.org/10.1111/j.2517-6161.1995.tb02031.x (1995).
Merico, D., Isserlin, R., Stueker, O., Emili, A. & Bader, G. D. Enrichment map: A network-based method for gene-set enrichment visualization and interpretation. PLoS ONE 5, e13984. https://doi.org/10.1371/journal.pone.0013984 (2010).
Subramanian, A. et al. Gene set enrichment analysis: A knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. 102, 15545–15550. https://doi.org/10.1073/pnas.0506580102 (2005).
Mootha, V. K. et al. PGC-1α-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes. Nat. Genet. 34, 267–273. https://doi.org/10.1038/ng1180 (2003).
Reimand, J. et al. Pathway enrichment analysis and visualization of omics data using g:Profiler, GSEA Cytoscape and EnrichmentMap. Nat. Protoc. 14, 482–517. https://doi.org/10.1038/s41596-018-0103-9 (2019).
Shannon, P. et al. Cytoscape: A software environment for integrated models of biomolecular interaction networks. Genome Res. 13, 2498–2504. https://doi.org/10.1101/gr.1239303 (2003).
Kucera, M., Isserlin, R., Arkhangorodsky, A. & Bader, G. D. AutoAnnotate: A Cytoscape app for summarizing networks with semantic annotations. F1000Res 5, 1717–1717. https://doi.org/10.12688/f1000research.9090.1 (2016).
Rusinova, I. et al. Interferome v2.0: An updated database of annotated interferon-regulated genes. Nucleic Acids Res. 41, D1040-1046. https://doi.org/10.1093/nar/gks1215 (2013).
Waluk, D. P., Schultz, N. & Hunt, M. C. Identification of glycine N-acyltransferase-like 2 (GLYATL2) as a transferase that produces N-acyl glycines in humans. Faseb J. 24, 2795–2803. https://doi.org/10.1096/fj.09-148551 (2010).
Burstein, S. H., McQuain, C. A., Ross, A. H., Salmonsen, R. A. & Zurier, R. E. Resolution of inflammation by N-arachidonoylglycine. J. Cell. Biochem. 112, 3227–3233. https://doi.org/10.1002/jcb.23245 (2011).
Battista, N., Bari, M. & Bisogno, T. N-acyl amino acids: Metabolism, molecular targets, and role in biological processes. Biomolecules https://doi.org/10.3390/biom9120822 (2019).
Bartoletti-Stella, A. et al. Messenger RNA processing is altered in autosomal dominant leukodystrophy. Hum. Mol. Genet. 24, 2746–2756. https://doi.org/10.1093/hmg/ddv034 (2015).
Bouzid, D. et al. Association of the RAVER2 gene with increased susceptibility for ulcerative colitis. Hum. Immunol. 73, 732–735. https://doi.org/10.1016/j.humimm.2012.04.018 (2012).
Rosebeck, S. & Leaman, D. W. Mitochondrial localization and pro-apoptotic effects of the interferon-inducible protein ISG12a. Apoptosis 13, 562–572. https://doi.org/10.1007/s10495-008-0190-0 (2008).
Rivera-Serrano, E. E. et al. Viperin reveals its true function. Ann. Rev. Virol. 7, 421–446. https://doi.org/10.1146/annurev-virology-011720-095930 (2020).
Sakaue, S. et al. A cross-population atlas of genetic associations for 220 human phenotypes. Nat. Genet. 53, 1415–1424. https://doi.org/10.1038/s41588-021-00931-x (2021).
Christakoudi, S., Evangelou, E., Riboli, E. & Tsilidis, K. K. GWAS of allometric body-shape indices in UK Biobank identifies loci suggesting associations with morphogenesis, organogenesis, adrenal cell renewal and cancer. Sci. Rep. 11, 10688. https://doi.org/10.1038/s41598-021-89176-6 (2021).
Arana, E. et al. Modic changes and associated features in Southern European chronic low back pain patients. Spine J. 11, 402–411. https://doi.org/10.1016/j.spinee.2011.03.019 (2011).
Zhao, S. et al. Myosin-18B promotes mechanosensitive CaMKK2-AMPK-VASP regulation of contractile actin stress fibers. iScience 23, 100975. https://doi.org/10.1016/j.isci.2020.100975 (2020).
Regard, J. B. et al. Verge: A novel vascular early response gene. J. Neurosci. 24, 4092–4103. https://doi.org/10.1523/jneurosci.4252-03.2004 (2004).
Modic, M. T. Modic type 1 and type 2 changes. J. Neurosurg. Spine 6, 150–151. https://doi.org/10.3171/spi.2007.6.2.150 (2007).
Boles, N. C. et al. NPTX1 regulates neural lineage specification from human pluripotent stem cells. Cell Rep. 6, 724–736. https://doi.org/10.1016/j.celrep.2014.01.026 (2014).
Al Rahim, M., Thatipamula, S. & Hossain, M. A. Critical role of neuronal pentraxin 1 in mitochondria-mediated hypoxic–ischemic neuronal injury. Neurobiol. Dis. 50, 59–68. https://doi.org/10.1016/j.nbd.2012.10.003 (2013).
Ohtori, S. et al. Tumor necrosis factor-immunoreactive cells and PGP 9.5-immunoreactive nerve fibers in vertebral endplates of patients with discogenic low back Pain and Modic Type 1 or Type 2 changes on MRI. Spine 31, 1026–1031. https://doi.org/10.1097/01.brs.0000215027.87102.7c (2006).
Crow, M. K., Olferiev, M. & Kirou, K. A. Type I interferons in autoimmune disease. Annu. Rev. Pathol. 14, 369–393. https://doi.org/10.1146/annurev-pathol-020117-043952 (2019).
Takayanagi, H., Sato, K., Takaoka, A. & Taniguchi, T. Interplay between interferon and other cytokine systems in bone metabolism. Immunol. Rev. 208, 181–193. https://doi.org/10.1111/j.0105-2896.2005.00337.x (2005).
Sadowska, A. et al. Inflammaging in cervical and lumbar degenerated intervertebral discs: Analysis of proinflammatory cytokine and TRP channel expression. Eur. Spine J. 27, 564–577. https://doi.org/10.1007/s00586-017-5360-8 (2018).
Kazezian, Z. et al. Gene expression profiling identifies interferon signalling molecules and IGFBP3 in human degenerative annulus fibrosus. Sci. Rep. 5, 15662. https://doi.org/10.1038/srep15662 (2015).
Guo, W. et al. Gene expression profile identifies potential biomarkers for human intervertebral disc degeneration. Mol. Med. Rep. 16, 8665–8672. https://doi.org/10.3892/mmr.2017.7741 (2017).
Navone, S. E. et al. Mechanical loading of intervertebral disc modulates microglia proliferation, activation, and chemotaxis. Osteoarthr. Cartil. 26, 978–987. https://doi.org/10.1016/j.joca.2018.04.013 (2018).
Dudli, S. et al. Modic type 1 change is an autoimmune response that requires a proinflammatory milieu provided by the “Modic disc”. Spine J. 18, 831–844. https://doi.org/10.1016/j.spinee.2017.12.004 (2018).
Moen, G.-H., Moen, A., Schistad, E. I. & Gjerstad, J. Local up-regulation of interferon-γ (IFN-γ) following disc herniation is involved in the inflammatory response underlying acute lumbar radicular pain. Cytokine 97, 181–186. https://doi.org/10.1016/j.cyto.2017.06.005 (2017).
Park, J. B., Chang, H. & Kim, Y. S. The pattern of interleukin-12 and T-helper types 1 and 2 cytokine expression in herniated lumbar disc tissue. Spine 27, 2125–2128. https://doi.org/10.1097/00007632-200210010-00009 (2002).
Oh, K. J. et al. Comparison of MR imaging findings between extraligamentous and subligamentous disk herniations in the lumbar spine. AJNR Am. J. Neuroradiol. 34, 683–687. https://doi.org/10.3174/ajnr.A3258 (2013).
Scuderi, G. J. et al. Epidural interferon gamma-immunoreactivity: A biomarker for lumbar nerve root irritation. Spine 34, 2311–2317. https://doi.org/10.1097/BRS.0b013e3181af06b6 (2009).
Tsuda, M. et al. IFN-gamma receptor signaling mediates spinal microglia activation driving neuropathic pain. Proc. Natl. Acad. Sci. U.S.A. 106, 8032–8037. https://doi.org/10.1073/pnas.0810420106 (2009).
Cuellar, J. M. et al. Cytokine evaluation in individuals with low back pain using discographic lavage. Spine J. 10, 212–218. https://doi.org/10.1016/j.spinee.2009.12.007 (2010).
Walker, B. F. et al. Are viruses associated with disc herniation? A clinical case series. BMC Musculoskelet. Disord. 21, 27. https://doi.org/10.1186/s12891-020-3052-8 (2020).
Scott, I. & Youle, R. J. Mitochondrial fission and fusion. Essays Biochem. 47, 85–98. https://doi.org/10.1042/bse0470085 (2010).
Chen, Y., Zhou, Z. & Min, W. Mitochondria, oxidative stress and innate immunity. Front. Physiol. 9, 1487–1487. https://doi.org/10.3389/fphys.2018.01487 (2018).
Geto, Z., Molla, M. D., Challa, F., Belay, Y. & Getahun, T. Mitochondrial dynamic dysfunction as a main triggering factor for inflammation associated chronic non-communicable diseases. J. Inflamm. Res. 13, 97–107. https://doi.org/10.2147/jir.S232009 (2020).
Gruber, H. E. et al. Mitochondrial bioenergetics, mass, and morphology are altered in cells of the degenerating human annulus. J. Orthopaed. Res. 31, 1270–1275. https://doi.org/10.1002/jor.22361 (2013).
Acknowledgements
We would like to extend our thanks to the patients who volunteered for this study, the hospitals wherefrom the patients were recruited (Drammen Hospital, Haukeland University Hospital, Oslo University Hospital, St Olavs Hospital, University Hospital of North Norway, Østfold Hospital Trust) and to all members of the AIM study group (clinicians, radiologists, coordinators and patient representative) for their involvement in implementing the study.
Funding
This work was supported by governmental organisations Helse Sør-Øst (Grant number 2015090); Helse Vest (Grant numbers HV 911938, HV 911891); and Klinbeforsk (Grant number 2017201). The funding authorities were not involved in the study design, collection or analysis of data, interpretation of the results or drafting of the manuscript.
Author information
Authors and Affiliations
Consortia
Contributions
M.D.V., S.T.F., J.A.Z., K.S., L.M.P. and B.A.L. designed the study. M.W., L.C.H.B., E.G., E.I.S., A.J.H., A.F., J.S.S. and the A.I.M. study group participated in data collection. A.E., P.M.K. and N.V. performed the MRI assessment. S.T.F. supervised the lab work. M.D.V. performed the lab work, analyzed the data and wrote the first draft of the manuscript. M.D.V., S.T.F., M.D.V., J.A.Z., K.S., L.M.P. and B.A.L. participated in the interpretation of the data. J.A.Z. and K.S. procured the resources. M.D.V., L.M.P. and B.A.L. supervised the study. All authors participated in drafting, critical revision and final approval of the article.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Vigeland, M.D., Flåm, S.T., Vigeland, M.D. et al. Correlation between gene expression and MRI STIR signals in patients with chronic low back pain and Modic changes indicates immune involvement. Sci Rep 12, 215 (2022). https://doi.org/10.1038/s41598-021-04189-5
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-021-04189-5
This article is cited by
-
A multidimensional nomogram combining clinical factors and imaging features to predict 1-year recurrence of low back pain with or without radicular pain after spinal manipulation/mobilization
Chiropractic & Manual Therapies (2023)
-
Modic changes as seen on MRI are associated with nonspecific chronic lower back pain and disability
Journal of Orthopaedic Surgery and Research (2023)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.