Intradiscal Cutibacterium acnes Sustains Modic Type 1‐Like Lesions Over Time in a Rat Lumbar Endplate Injury Model

ABSTRACT

Background

Modic changes (MC) are painful vertebral bone marrow lesions. Three interconvertible types represent different pathology stages with MC1 being the most painful, MC2 less painful, and MC3 asymptomatic. Bacterial (Cutibacterium acnes (C. acnes)–mediated) and non‐bacterial etiologies are both suggested to induce MC1, but it remains unclear whether MC conversion is etiology‐specific. This has important implications for patient management and the development of MC therapies.

Aims

To assess MC‐subtype lesion prevalence and pain‐like behavior over time in an etiology‐specific MC rat model.

Methods

Four‐to‐five‐month‐old Sprague–Dawley rats underwent sham (n = 12) surgery or endplate injury followed by an intradiscal injection of 2.5 μL of either TNF‐α (1000 ng/mL PBS) (n = 19) or C. acnes (3.2 × 10⁸ CFU/mL in PBS) (n = 18) to mimic aspects of bacterial versus non‐bacterial MC etiologies. MC‐subtype prevalence was evaluated by magnetic resonance imaging at 1, 8, and 14 weeks post‐injury. Disc degeneration, bone marrow lesion neutrophil elastase and CD19 immunoreactivity, spinal cord sensitization, and pain‐like behavior (von Frey) were assessed, and contributors to pain‐like behavior were identified (random forest).

Results

C. acnes injection caused higher MC1‐like lesion prevalence than TNF‐α and Sham at all‐time points and higher spinal cord substance P expression at 14 weeks post‐injury. MC2‐like lesions increased over time in TNF‐α‐injected discs and were more prevalent at 14 weeks post‐injury than C. acnes. C. acnes injection resulted in bone marrow lesions with higher neutrophil elastase‐ and CD19 immunoreactivity. MC1‐like lesions contributed strongest to pain‐like behavior, although von Frey did not differ between groups.

Conclusion

Intradiscal C. acnes injection sustained MC1‐like lesions in an etiology‐specific lumbar MC rat model, suggesting that patients with bacterial MC etiology may be more likely to persist in painful MC1.

Intradiscal C. acnes injection sustained MC1‐like lesions with persistent neutrophil sustenance in an etiology‐specific lumbar MC rat model. This suggests that patients with bacterial MC etiology may persist in painful MC1, highlighting C. acnes and neutrophils as potential targets to promote MC1‐to‐MC2 conversion and improve pain and disability.

1. Introduction

Chronic low back pain (cLBP) is one of the most disabling conditions worldwide, carrying an enormous socioeconomic burden [1, 2]. Vertebral endplate bone marrow lesions, called Modic changes (MC), are a common finding on T1‐weighted (T1w) and T2w magnetic resonance (MR) images in cLBP populations (43% in cLBP vs. 6% in the asymptomatic population) [3]. MCs are independent predictors of cLBP, exhibiting a characteristic inflammatory pain profile [3, 4, 5, 6, 7]. Three interconvertible MC subtypes have been defined based on their appearance on T1w and T2w MRI: MC1: T1w: Hypointense, T2w: Hyperintense; MC2: T1w and T2w: Hyperintense; MC3: T1w and T2w: Hypointense [7]. MC1 have been shown to be the most painful, with MC2 having a weaker pain association, and MC3 being asymptomatic and occurring very rarely [8, 9, 10]. Longitudinal studies suggest MC subtypes are interconvertible and represent different stages of the same pathology [11, 12]. MC1 may represent an early and dynamic stage, with approximately one‐third progressing to MC2 over 4 years in a population‐based cohort study. However, reverse transitions from MC2 to MC1 and de novo MC2 lesions have also been reported, indicating that a universal progression sequence cannot be clearly defined. MC3 was rare and appeared to represent a relatively stable endpoint. However, the biological mechanisms that drive MC subtype persistence and conversion remain largely unknown.

MC lesions usually occur above and below degenerated discs and colocalize with endplate (EP) damage [13, 14, 15, 16, 17, 18]. EP damage is believed to initiate MC development by promoting pro‐inflammatory and pro‐fibrotic disc‐marrow crosstalk [19]. Bone marrow lesions following acute EP injury can resolve spontaneously and the resulting pain is typically short‐lived [20]. Hence, MC appears to involve a persisting pro‐inflammatory stimulus that prevents an effective healing response. A growing body of evidence suggests that this pro‐inflammatory stimulus can either be of bacterial or non‐bacterial origin [21, 22, 23]. Although newer sequencing studies have detected additional bacterial taxa that might play a role in MC [24, 25, 26], there is evidence from human studies that implicate Cutibacterium acnes (C. acnes) as an important player in the bacterial MC etiology [23, 27, 28, 29]. In contrast, an autoimmune response of bone marrow leukocytes against disc cells and matrix is suggested in the non‐bacterial etiology. Recently, the first evidence of distinct bacterial (C. acnes‐mediated) and non‐bacterial subtypes in MC1 patients was provided, suggesting that there are at least two MC1 etiologies [23]. However, it is unclear if both MC1 etiologies promote conversion to other MC subtypes, or if one etiology is more likely to persist in a more painful MC1.

The pathobiology of different MC subtypes remains largely unknown. Independent of the etiology, inflammation, fibrosis, neutrophilic infiltrates, and activated neutrophils have been attributed to MC1 [7, 19, 30, 31]. MC2 is characterized by fatty marrow conversion, inflammation, absence of neutrophilic infiltrates, and signs of neutrophil clearance [7, 31, 32] and sclerosis is the hallmark of MC3 [7, 33]. The only study that examined etiology‐specific mechanisms in MC1 bone marrow reported a predominant innate (neutrophil‐driven) immune response in bacterial versus a predominant adaptive immune response (B‐cell‐driven) in non‐bacterial MC1 bone marrow [23]. Yet, the etiology‐specific mechanisms of MC conversion remain unclear. Understanding these mechanisms could inform novel therapies that promote the conversion of endplate bone marrow lesions to less painful MC subtypes, since complete resolution of bone marrow lesions appears unlikely once extensive marrow is involved [12].

While human MRI protocols can distinguish MC subtypes on T1w and T2w MRI and patient IVD samples can be used to assess bacterial load, current methods cannot non‐invasively determine the underlying etiology of MC lesions. Therefore, animal models represent an essential complementary approach that enables controlled, etiology‐specific studies of MC development and longitudinal subtype conversion in ways not possible in human studies alone. A rat model of a transcorporeal EP injury followed by a TNF‐α injection induced MC1‐like changes with pain‐like behaviors and spinal cord sensitization, although this model used a transcorporeal injury which disrupted the bone, and the model was not etiology‐specific [34]. Injecting C. acnes into rat tail discs or implanting disc surrogates into rat tail vertebrae led to MC1‐like lesions on MRIs after 8 weeks, providing evidence for at least two different MC1 etiologies [35, 36]. Consistent with findings from rat tail discs, injection of C. acnes into cervical rat discs led to disc degeneration and pain‐like behavior 6 weeks post‐injury [37]. However, these previous models lack longitudinal data on MC subtype development, particularly at chronic time points. Determining if MC subtype lesions change in bacterial versus non‐bacterial injected discs over time, and if these changes correspond to changes in pain‐like behavior remains an important open question.

This study evaluated the prevalence of MC subtype lesions and pain‐like behavior in an etiology‐specific MC rat model over time. The lumbar EP injury rat model recapitulated features of bacterial MC etiology with intradiscal C. acnes injection, and non‐bacterial MC etiology with intradiscal TNF‐α injection. It was hypothesized that (i) C. acnes injection would result in bacterial proliferation and a persisting pro‐inflammatory stimulus sustaining MC1‐like lesion prevalence over time, whereas the one time TNF‐α injection would induce an acute inflammatory response with MC1‐like lesions leading to an increase in MC2‐like lesion prevalence over time, and (ii) that an increase in MC2‐like lesion prevalence would correspond with decreased pain‐like behavior and spinal cord sensitization. It was further hypothesized that (iii) bone marrow lesions adjacent to C. acnes‐injected discs would show increased neutrophil elastase immunoreactivity (NE‐ir), whereas CD19‐ir (indicative of B‐cell presence) would be higher in TNF‐α–injected discs, consistent with clinical findings from MC1 patients with high and low intradiscal C. acnes load.

2. Materials and Methods

2.1. Study Design

All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC; PROTO202000031). Forty‐nine skeletally mature male Sprague–Dawley rats (4–5‐months old) were randomly assigned to 1 of 3 groups: Sham (n = 12), EP injury + TNF‐α (n = 19), or EP injury + C. acnes (n = 18). Male rats were selected to maintain consistency with previously validated behavioral assays that were developed in male animals only. Rats in the sham group underwent an anterior abdominal incision with exposure of the L4–L6 lumbar spine only. In the injury groups, an EP puncture was performed followed by intradiscal injection of either TNF‐α or C. acnes, such that the measured outcomes represent the injectate‐specific effects in the context of EP injury. Pain‐like behavior was assessed biweekly after surgery. Animals were euthanized at 1 week (acute), 8 weeks (early chronic), or 14 weeks (late chronic) post‐injury. Surgeries were performed in 6 cohorts, with 2 cohorts assigned to each time point. Following euthanasia, lumbar spines (L2–L6) and spinal cords from each animal were isolated. The spinal cords were processed for histology. The spines underwent ex vivo T1w and T2w MR imaging (MRI) prior to histological processing (Figure 1).

FIGURE 1.

Schematic overview of the study design and experimental timeline. The surgery involved (1) an anterior abdominal approach, followed by (2) EP injury, and (3) intradiscal injection of either TNF‐α or C. acnes, such that all outcome measures reflect the combined effects of both the EP injury and the respective injectate. Pain‐like behavior was evaluated biweekly using the von Frey assay to assess hind paw mechanical allodynia. Post‐euthanasia assessments included ex vivo spinal MRI using T1w and T2w sequences as well as histological analysis of spine and spinal cord. The timeline depicts the three experimental cohorts and the corresponding analyses time points. C. acnes, Cutibacterium acnes; EP, endplate; MRI, magnetic resonance imaging; T1w, T1‐weighted.

2.2. Materials

If not stated otherwise, materials were obtained from Fisher Scientific, Fair Lawn, NJ, USA.

2.3. C. acnes Culture

C. acnes was obtained commercially in the form of a lyophilized pellet (#6919, ATCC, Manassas, VA, USA). Prior to surgery, bacteria were cultured anaerobically in a liquid brain‐heart infusion (BHI) medium (#B11059) at 37°C for 10 days. The bacterial suspension was then plated on BHI agar plates (#NC1064252) and incubated anaerobically for an additional 3 days. The bacterial species was confirmed with 16S rRNA sequencing (Azenta) 1 day before surgery. On the day of surgery, the number of colony‐forming units (CFUs) was counted, and the C. acnes stock solution was diluted to a final concentration of 3.2 × 10⁸ CFUs/mL in phosphate‐buffered saline (PBS). This concentration was chosen based on a previous study by Li et al. [37], who demonstrated that this bacterial load induced disc degeneration, pain‐like behavior, and hyperintense bone marrow lesion signals, but not discitis, in male Sprague–Dawley rats. A separate dilution tube was prepared for each rat. Immediately after injection, the C. acnes suspension was plated on BHI agar plates again, cultivated for 3 days, and submitted for 16S rRNA sequencing to reconfirm the species' identity.

2.4. Surgical Procedure and EP Puncture Injury

The surgical procedures were performed under aseptic conditions and general anesthesia with 4% isoflurane and an oxygen flow rate of 1 L/min similar to previously reported [38]. The L4‐5 and L5‐6 discs were identified using a preoperative anterior–posterior X‐ray. The L3–6 lumbar spine was exposed via an anterior abdominal incision through the skin and peritoneal membrane. The disc levels were confirmed using anatomical landmarks, such as the aortic trifurcation and the iliac crest, as well as a C‐arm computed tomography (CT) image. The rats in the sham group underwent lumbar spine exposure only. In the two injury groups, EP injury was induced by puncturing through the annulus fibrosus (AF) through the caudal L4 and cranial L6 EPs into the vertebral body using a 21 G drill bit (~25%–30% of the whole EP) (#29045A824, McMaster‐Carr, Robbinsville, NJ, USA) containing a 3‐mm stopper. The aim of the EP puncture was to facilitate crosstalk between the disc and the bone marrow since patients with MC have crosstalk between the MC bone marrow and the adjacent disc [19]. EP punctures were followed by a 2.5 μL intradiscal injection of either TNF‐α (1000 ng/mL PBS) or C. acnes (3.2 × 10⁸ CFU/mL in PBS) using a calibrated microliter syringe (Hamilton Company, Reno, NV, USA) with a 26 G needle with a 3 mm stopper. Abdominal muscles were closed using 3–0 silk sutures (#501180836) and the skin was closed using 4‐0 nylon sutures (#502092799). The rats were allowed ad libitum access to food and water and were closely monitored for complications. All animals were allowed unrestricted movement in their cages for the entire experimental duration and were singly housed due to aggressive behavior.

2.5. Pain‐Related Behavior Assessment

Von Frey assay for hind paw mechanical allodynia was used to assess pain‐like behavior. A single blinded operator evaluated pain‐like behaviors biweekly in a dedicated behavioral analysis room with regular indoor lighting. Test cages with wire‐mesh floors were used to allow access to the rat hind paws from below. Before testing, the rats were acclimated to the blinded operator and the test cage for 5 consecutive days. On the day of testing, the animals were acclimated in the experimental room for 1 h and the testing cage for 20 min before assessment. This was followed by the application of calibrated von Frey filaments to the plantar surface of each hind paw in ascending order from 0.6 to 26 G (Stoelting, Wood Dale, IL, USA). Each filament was applied five times until it bent to ensure consistent sufficient pressure. The lowest force filament that induced nocifensive behaviors, including paw licking, extended paw withdrawal, and fanning/shaking of the paw in three out of five repetitions was identified as the paw withdrawal threshold. Paw withdrawal thresholds from right and left hind paws were averaged for statistical analysis. The baseline hind paw withdrawal threshold was determined using the average of two measurements taken 3 days apart, 1 week prior to surgery. Von Frey measurements post‐surgery were normalized to baseline.

2.6. Euthanasia and Tissue Collection

At the respective endpoints (1, 8, or 14 weeks post‐injury), rats were anesthetized (5% isoflurane with an oxygen flow rate of 1 L/min) for 5 min and subsequently transcardially perfused with 10% neutral buffered formalin phosphate. Lumbar spinal cords (corresponding to T12‐L1 vertebral level) and lumbar spines (L2–L6) were harvested and fixed in 10% neutral buffered formalin phosphate at room temperature for 48 h.

2.7. MRI and MRI Grading

Imaging: For imaging, spines were transferred after fixation to PBS and MRI was performed on a 9.4 T vertical‐bore micro‐MRI system (Bruker Avance III 400) using a 20‐mm quadrature birdcage RF coil (Rapid Biomedical). T1w images were acquired using 3D MP‐RAGE (139 μm isotropic resolution, TR = 4 s, TI = 1.1 s) and T2w images were acquired using 3D RARE (139 μm isotropic resolution, TR = 2 s, TE = 17 ms).

MRI feature grading: An experienced MC researcher and a spine surgeon independently evaluated degenerative features and MC‐like lesions in a blinded manner. The L4‐5 and L5‐6 discs were evaluated for the degree of disc degeneration (grades 0–4, adapted from Pfirrmann grading [39]) (Figure S1A), and the scores were averaged between readers. MRI disc degeneration scores were analyzed on a per‐disc basis, with each disc treated as an independent observation.

The L6 cranial, L5 cranial and caudal, and L4 caudal EPs were assessed separately for (i) transverse EP disruption, (ii) presence and area of disc extrusion into the marrow, and (iii) presence and area of MC1‐, MC2‐, and MC3‐like lesions. The MRI slice in which each feature was most pronounced was selected for analysis.

1. For the extent of transverse EP disruption, the length of the disrupted EP was calculated as a percentage of the total EP length and graded as follows: Grade 0 = not affected, grade 1 = < 25%, grade 2 = 25%–50%, and grade 3 = > 50%. The grades were averaged between readers (Figure S1B).

2. The presence or absence of disc extrusion into the marrow was rated as 0 or 1, and the extrusion area was rated as a percentage of the L5 vertebral body area. Cases in which there was disagreement about the presence of extrusion were discussed between readers and agreed upon. One reader then performed area measurements after the discussion process (Figure S1C).

3. As for the disc extrusion, the presence (0,1) and the area of all three MC‐subtype lesions were evaluated. In cases of disagreement regarding MC subtype classification or presence/absence, the cases were discussed and agreed upon. Area measurements were subsequently performed by one reader post‐discussion (Figure S1D).

2.8. Spine Histology and Immunohistochemistry (IHC)

Disc degeneration: After MRI, spines were decalcified for 5 days at room temperature using formic acid. Subsequently, they were dehydrated in increasing concentrations of ethanol, embedded in paraffin, and mid‐sagittal sections containing the EP puncture injury were identified and sectioned at a thickness of 5 μm. To assess disc morphology, the sections were deparaffinized, rehydrated, and stained with Safranin‐O/Fast Green/Hematoxylin (SO/FG/H). Three experienced disc histology reviewers scored disc degeneration according to Lai et al. [40] and the scores were averaged across reviewers [40].

IHC: Deparaffinized and rehydrated mid‐sagittal sections were incubated for 20 min with 2.5% horse serum (Vector Laboratories Inc., Burlingame, CA, USA) to prevent non‐specific binding. Sections were then incubated for 1 h at room temperature with one of the following primary antibodies: (i) rabbit polyclonal antibody against rat TNF‐α (1:500 dilution, #NBP1‐19532, Novus Biologics, MN, USA), (ii) rabbit recombinant multiclonal antibody against rat NE (1:500 dilution, #ab314916, Abcam, Waltham, MA, USA), or (iii) rabbit recombinant polyclonal antibody against rat CD19 (1:150 dilution, #27949‐1‐AP, ThermoFisher, Waltham, MA, USA). Normal rabbit serum (#S‐500‐20, Vector Laboratories Inc., Burlingame, CA, USA) served as a negative control. After incubation with a horseradish peroxidase (HRP)‐conjugated anti‐rabbit secondary antibody (#MP‐7061, Vector Laboratories Inc., Burlingame, CA, USA), antibody‐ir was visualized using a diaminobenzidine (DAB)‐based peroxidase substrate (#MP‐7601, Vector Laboratories Inc., Burlingame, CA, USA). Sections were counterstained with toluidine blue to visualize spinal morphology, dehydrated, mounted, and examined under a bright‐field microscope (Axio Imager Z1, Zeiss, Thornwood, NY, USA).

IHC grading: Intradiscal TNF‐α‐ir was evaluated using a semi‐quantitative grading scale ranging from 0 to 3. Grade 0 indicated an absence of TNF‐α‐ir and grade 3 represented the strongest TNF‐α‐ir (Figure S2A). TNF‐α‐ir was assessed in the following regions of the disc: the anterior longitudinal ligament (ALL) including granulation tissue, anterior and posterior AF, nucleus pulposus (NP), cranial and caudal EPs, and posterior surrounding tissue at the L4‐5 and L5‐6 levels. Values from each region were averaged between the two readers. The overall mean TNF‐α‐ir was analyzed on a per‐disc basis, with each disc treated as an independent observation. NE‐ir and CD19‐ir were evaluated only within visible bone marrow lesions. Each lesion was graded on a scale from 0 to 3, where 0 indicated an absence of NE/CD19‐ir and 3 represented the strongest staining intensity (Figure S2B,C). Time points were combined for statistical analysis because NE‐ir and CD19‐ir were not hypothesized to change with time.

2.9. Immunohistochemical Analysis for Spinal Cord Sensitization and Neuroinflammation

Spinal cord IHC was performed as previously described [38]. Formalin‐fixed spinal cords were dehydrated in increasing ethanol concentrations, halved transversally (resulting in two spinal cord sections per animal), embedded in paraffin, and sectioned into 6 μm thick sections [38, 41]. For IHC staining, sections were deparaffinized and rehydrated. Antigens were retrieved for 5 min with antigen‐retrieval buffer (#H3292, Sigma‐Aldrich Inc., St. Louis, MO, USA) followed by 20 min blocking for non‐specific binding with 2.5% normal horse serum (Vector Laboratories Inc., Burlingame, CA, USA). Sections were incubated overnight at 4°C in a dark, humidified chamber with a mouse monoclonal primary antibody against rat Substance P (SubP) (1:1000 dilution, #ab14184, Abcam, Cambridge, MA, USA) or rabbit polyclonal primary antibody against rat Glial Fibrillary Acidic Protein (GFAP) (1:2000, #ab7260, Abcam, Cambridge, MA, USA). Normal mouse serum (#NC494H, Biocare Medical LLC, Pacheco, CA, USA) or normal rabbit serum, respectively, was used as a negative control. A secondary antibody for SubP was carried out with a VectaFluor Excel Amplified anti‐mouse IgG, Dylight 488 antibody kit (#DK‐2488, Vector Laboratories Inc., Burlingame, CA, USA), and for GFAP with a VectaFluor Excel Amplified anti‐rabbit IgG, Dylight 488 antibody kit (#DK‐1488, Vector Laboratories Inc., Burlingame, CA, USA), according to manufacturer's instructions. Sections were then stained with NeuroTrace 530/615 red fluorescent Nissl stain (1:200 dilution, #N21482, ThermoFisher Scientific, Waltham, MA, USA) to visualize the neurons and glial cells. The stained slides were mounted using a ProLong Gold Antifade Mount with DAPI (#P36931, Thermo Fisher Scientific, Waltham, MA, USA), and imaged using a Leica DM6 B microscope (Leica Microsystems Inc., Deerfield, IL, USA) at 20× magnification with standardized microscope settings. Threshold was set based on negative control and images were analyzed using ImageJ. Percentage SubP‐ and GFAP‐ir relative to area of spinal dorsal horn was quantified and then averaged between the left and right dorsal horn as well as across the two sections from each animal.

2.10. Statistical Analysis

All statistical analyses were performed using GraphPad Prism (version 10.6.1) or RStudio (version 2024.12.1+563). Data normality was assessed using the Shapiro–Wilk test. Depending on data distribution, appropriate statistical tests were applied to evaluate group differences. For normally distributed data, results are presented as mean ± standard deviation (SD); for non‐normally distributed data, values are shown as median with interquartile range (IQR). A p value < 0.05 was considered statistically significant.

To assess the effect of time, group, and interaction, mixed‐effects models were used for repeated measures (von Frey), whereas two‐way ANOVAs (MC subtype lesion prevalence, disc degeneration, intradiscal TNF‐α‐ir, spinal cord SubP and GFAP) were applied for non‐repeated measures. Tukey post hoc comparisons were performed when main effects were significant.

A random forest model was used to identify which MRI predictors were associated most with pain‐like behavior. MRI features averaged across the 4 EPs were used as predictors, and pain‐like behavior scores served as the outcome. Each forest consisted of 1000 trees, with one‐third of the predictors randomly selected at each split. Variable importance was assessed using 1000 outcome permutations to estimate pseudo p values for each predictor. To evaluate model performance, a leave‐one‐out cross‐validation (LOOCV) approach was implemented, where for n samples, n models were trained, each leaving out one sample for testing and using the remaining n–1 for training. Predictions for the left‐out samples were aggregated to compute the coefficient of determination (R ²), root mean square error (RMSE), and mean absolute error (MAE). Variable importance was quantified by the mean percentage increase in mean squared error (%IncMSE) across all folds. To assess the significance of each predictor, permutation‐based pseudo p values were calculated (1000 permutations of the outcome variable) and adjusted for multiple comparisons using the Benjamini–Hochberg false discovery rate (FDR) correction.

3. Results

3.1. Sham and EP Injury Procedures Did Not Affect Rat General Health

The procedures of rats in all 3 groups were well‐tolerated and there were no significant surgical complications. Postoperatively, the rats were monitored for distress or behavioral changes, including altered mobility, eating behavior, bowel movements, and response to handling. No apparent differences were observed among all 3 groups. Rats from all groups gained weight over time, but there was no difference in body weight between groups at any time point (p = 0.31) (Figure S3). Taken together, weight gain and no apparent behavioral differences indicated that this procedure did not have an effect on general rat health.

3.2. Discs of Both EP Injured Groups Were More Degenerated Than Sham With No Differences Between Groups

Degree of disc degeneration was assessed with MRI and histology. MRI disc degeneration scores did not differ among groups at 1‐week post‐injury. At 8‐ and 14‐weeks post‐injury, degeneration scores were significantly higher in the TNF‐α (week 8: p = 8.00 × 10⁻⁴; week 14: p = 9.40 × 10⁻³) and C. acnes (week 8: p = 4.00 × 10⁻⁴, week 14: p = 7.00 × 10⁻⁴) groups compared to Sham (Figure 2A,B). Moreover, degeneration scores significantly increased in C. acnes‐injected discs over time (week 1–8: p = 0.03; week 1–14: p = 5.90 × 10⁻³). However, no significant differences were detected between TNF‐α‐ and C. acnes‐injected discs at any of the time points assessed (week 1: p = 0.69; week 8: p = 0.88; week 14: p = 0.59).

FIGURE 2.

Degree of disc degeneration was increased in both EP injury + C. acnes and EP injury + TNF‐α discs compared to Sham without difference between injury groups. (A) Representative T1w and T2w MRI of discs per group and timepoint. (B) Both EP injury + C. acnes and EP injury + TNF‐α discs were significantly more degenerated than Sham at 8‐ and 14 weeks post‐injury, with no differences between the two injury groups at any time point. Disc degeneration increased over time in the EP injury + C. acnes group. Bars represent median with IQR. Significance bars represent results of Tukey's post hoc analysis of pairwise comparisons after detecting main effects with a two‐way ANOVA. Timepoints week 1, 8, and 14: Sham: N = 8 discs, EP injury + TNF‐α: N = 12–14 discs, EP injury + C. acnes: N = 10–12 discs. (C) Representative histological images of whole discs per group and timepoint, SafO/F/H staining. (D) Quantification of histological disc degeneration score. Discs from both EP injury groups were significantly more degenerated than Sham at all‐time points (weeks 1, 8, and 14). No significant differences were observed between EP injury + TNF‐α versus EP injury + C. acnes groups at any time point. Bars represent median with IQR. Significance bars represent results of Tukey's post hoc analysis of pairwise comparisons after detecting main effects with a two‐way ANOVA. Timepoints week 1, 8, and 14: Sham: N = 6–7 discs, EP injury + TNF‐α: N = 10–12 discs, EP injury + C. acnes: N = 9–12 discs. ANOVA, analysis of variance; C. acnes, Cutibacterium acnes; EP, endplate; IQR, interquartile range; SafO/F/H, Safranin‐O/fastgreen/hematoxylin; T1w, T1‐weighted.

Histological disc degeneration scores of TNF‐α (p = 3.00 × 10⁻⁴) and C. acnes (p = 1.00 × 10⁻⁴) showed already significantly higher degeneration scores at 1‐week post‐injury compared to Sham. Consistent with the MRI findings, both TNF‐α (week 8: p < 1.00 × 10⁻⁴; week 14: p < 1.00 × 10⁻⁴) and C. acnes (week 8: p < 1.00 × 10⁻⁴; week 14: p < 1.00 × 10⁻⁴) groups showed significantly higher disc degeneration scores than Sham at chronic time points (Figure 2C,D). In line with MRI, there were no significant differences between TNF‐α‐ and C. acnes‐injected discs at any of the time points assessed (week 1: p = 0.94; week 8: p = 0.81; week 14: p = 0.99). These findings were also consistent when stratifying for NP, AF, NP‐AF border, or EP regions between (Figure S4A–D).

Taken together, the overall degree of disc degeneration was similar in C. acnes and TNF‐α groups.

3.3. Intradiscal C. acnes Injection Induced MC1‐Like Lesion Persistence Over Time, Whereas TNF‐α Injection Increased MC2‐Like Lesions

Both EP injury groups developed all 3 MC subtype lesions on MRI, as determined by a qualitative evaluation of 4 EPs (L6 cranial, L5 cranial and caudal, and L4 caudal) for MC1‐like (T1w: hypo‐ or isointense; T2w: hyperintense), MC2‐like (T1w and T2w: hyperintense), and MC3‐like (T1w and T2w: hypointense) lesions (Figure 3A). Consistent with MRI, all corresponding MC1‐3 histological features described in MC patients, namely cellular infiltrates (MC1), fatty marrow (MC2), and sclerotic bone (MC3), were observed in both injury groups (Figure 3B) [7, 31, 33].

FIGURE 3.

Intradiscal injectate following EP injury determined MC subtype prevalence over time. (A) Representative T1w and T2w MR images showing MC1‐ (top), MC2‐ (middle), and MC3‐like (bottom) lesions in EP injury + TNF‐α (left) or EP injury + C. acnes (right) discs. All 3 different MC subtypes developed in both EP injury + TNF‐α and EP injury + C. acnes groups. MC1‐like lesions: T1w: Hypo‐, or isointense; T2w: Hyperintense. MC2‐like lesions: T1w and T2w: Hyperintense. MC3‐like lesions: T1w and T2w: Hypointense. Red arrows surround bone marrow lesions. Images are from time points 8‐ and 14‐weeks post‐injury. (B) Representative histological images of features associated with MC1 (cellular infiltrates), MC2 (fatty replacement of normal bone marrow), and MC3 (sclerotic bone) in EP injury + TNF‐α and EP injury + C. acnes groups. In both injury groups, all 3 histological features of MC subtypes were found. #: Normal bone marrow region. 1: Cellular infiltrates; 2: Fatty replacement of normal bone marrow; 3: Increased bone structure. SafO/F/H staining. (C) Quantification of MC1‐like lesion prevalence (% of total number of EPs: L6 cranial, L5 caudal, L5 cranial, L4 caudal) revealed that EP injury + intradiscal C. acnes injection resulted in a significantly higher prevalence of MC1‐like lesions across all 3 time points compared to both EP injury + TNF‐α and Sham. The EP injury + TNF‐α group tended to have significantly more MC1‐like lesions compared to Sham. (D) MC2‐like lesions increased in the EP injury + TNF‐α group over time and were significantly higher than EP injury + C. acnes and Sham at 14‐weeks post‐injury. (E) MC3‐like lesions increased over time in the EP injury + C. acnes group. Bars represent median with IQR. Significance bars represent pairwise comparisons from Tukey's post hoc analysis following two‐way ANOVA. Timepoints week 1, 8, and 14: Sham: N = 4 rats, EP injury + TNF‐α: N = 5–7 rats, EP injury + C. acnes: N = 5–6 rats. *p < 0.05, **p < 0.01, ***p < 0.001. ANOVA, analysis of variance; C. acnes, Cutibacterium acnes; EP, endplate; IQR, interquartile range; MC, modic change; SafO/F/H, Safranin‐O/fastgreen/hematoxylin; T1w, T1‐weighted.

To test the hypothesis that intradiscal injectate would influence the prevalence of MC subtype lesions over time, the prevalence of each MC subtype lesion across the 4 evaluated EPs per group and time point was quantified. MC1‐like lesion prevalence in C. acnes–injected discs was significantly higher than TNF‐α (week 1: p = 0.05; week 8: p = 0.02; week 14: p = 6.00 × 10⁻⁴) and Sham (week 1: p = 3.00 × 10⁻⁴; week 8: p < 1.00 × 10⁻⁴; week 14: p < 1.00 × 10⁻⁴) at all‐time points assessed (Figure 3C). MC1‐like lesions in TNF‐α‐injected discs trended higher than Sham at 1‐week (p = 0.15) and 8‐week (p = 0.05) post‐injury. This showed that both intradiscal injectates in combination with EP injury induced MC1‐like lesions detectable on MRI, but C. acnes injection resulted in a higher and more sustained MC1‐like lesion prevalence over time.

MC2‐like lesion prevalence in the TNF‐α group was significantly increased with time (week 1 vs. 14: p = 2.00 × 10⁻⁴, week 8 vs. 14: p = 5.60 × 10⁻³). At 14‐weeks post‐injury, MC2‐like lesions were significantly higher in TNF‐α compared to C. acnes (p = 4.9 × 10⁻³) and Sham (p = 2.9 × 10⁻³) (Figure 3D). In contrast, MC2‐like lesion prevalence was only slightly elevated in C. acnes at 1‐week post‐injury (p = 0.05), but not at later time points (week 8: p = 0.61; week 14: p = 0.85).

Overall, MC3‐like lesions were the least prevalent of the 3 subtypes. No significant differences in MC3‐like lesion prevalence were observed among groups at any time point. However, in the C. acnes group, the prevalence of MC3‐like lesions increased over time (week 1–8: p = 0.02; week 1–14: p = 2.4 × 10⁻³). Even though some MC3‐like lesions were observed in the TNF‐α group, there was no statistical difference over time (Figure 3E).

In summary, the intradiscal injectate was a key determinant of MC subtype lesion prevalence, with C. acnes sustaining MC1‐like lesion prevalence and TNF‐α resulting in significant increases in MC2‐like lesions over time.

3.4. Origin of Intradiscal Pro‐Inflammatory Stimulus Determines Bone Marrow Immune Cell Response

Since C. acnes and TNF‐α injectates led to different MC subtypes on MRI over time, it was investigated whether these subtype‐differences were driven by the overall intradiscal inflammatory burden. To quantify intradiscal inflammatory burden, TNF‐α‐ir, a cytokine commonly associated with chronic intradiscal inflammation [42], was assessed. No difference in intradiscal TNF‐α‐ir between C. acnes‐ and TNF‐α–injected discs at any time point assessed was found (Figure 4A,B). This finding was consistent if stratifying for different disc regions (Figure S5), suggesting no difference in TNF‐α‐intradiscal inflammatory burden between injury groups.

FIGURE 4.

Intradiscal C. acnes versus TNF‐α injection following EP injury determined adjacent bone marrow lesion immune cell response. (A) Representative images of intradiscal inflammatory burden measured as TNF‐α‐ir in Sham (left), TNF‐α (middle) and C. acnes (right)‐injected discs. Upper right images represent magnified areas of the overview section. Arrows indicate TNF‐α positive cells. (B) Both EP injury + TNF‐α and EP injury + C. acnes groups led to increased discal TNF‐α‐ir at all 3 time points without a difference between injury groups. Bars represent median with IQR. Significance bars represent pairwise comparisons from Tukey's post hoc analysis. Timepoints week 1, 8, and 14: Sham: N = 6–8 discs, EP injury + TNF‐α: N = 8–10 discs, EP injury + C. acnes: 10–12 discs. (C) Representative images of NE‐ir (top) and CD19‐ir (bottom). Red arrows indicate NE‐positive cells, orange arrows indicate CD19‐positive cells. (D) Bone marrow lesions adjacent to EP injury + C. acnes‐ versus EP injury + TNF‐α‐ discs showed increased NE‐ir and CD19‐ir. Bars represent median with IQR. Significance bars represent results from Mann–Whitney U‐test. All time points combined: EP injury + TNF‐α: N = 12 bone marrow lesions, EP injury + C. acnes: N = 16 bone marrow lesions. *p < 0.05, **p < 0.01, ***p < 0.001. C. acnes, Cutibacterium acnes; IQR, interquartile range; NE, neutrophil elastase.

Hence, it was hypothesized that the origin of the intradiscal pro‐inflammatory stimulus (C. acnes vs. TNF‐α), rather than intradiscal inflammatory burden itself, determined whether MC1‐like lesions persisted or converted over time. Unfortunately, repeated attempts by multiple IHC experts to detect the injected C. acnes strain using commercially available antibodies were unsuccessful. These antibodies showed similar staining patterns in confirmed positive controls (C. acnes cultures) and bacterial‐negative tissues such as brain, indicating insufficient specificity (Figure S6). One antibody detected C. acnes in human IVDs in a previous study [43], and this discrepancy may reflect species‐specific differences, differences in tissue processing, or batch differences with the commercially available antibodies.

Therefore, it was evaluated whether C. acnes and TNF‐α intradiscal injections elicited distinct bone marrow immune responses, indirectly assessing whether the injectate's origin influenced bone marrow responses. It was found that bone marrow lesions adjacent to C. acnes–versus TNF‐α‐injected discs exhibited significantly higher NE‐ir (p = 0.02) and CD19‐ir (p = 0.04), indicative of increased neutrophil and B‐cell presence, respectively (Figure 4C,D).

Taken together, our findings suggest that the origin of the intradiscal injectate (C. acnes vs. TNF‐α) dictated the adjacent bone marrow immune cell response, despite comparable intradiscal inflammatory burden.

3.5. Spinal Cord SubP Was Elevated in the C. acnes Group, but Pain‐Like Behavior Did Not Differ Between Injury Groups

MC1 lesions are more painful than MC2 lesions in humans, and a conversion from MC1 to MC2 has been linked to pain reduction in cLBP patients with MC [8, 9, 11]. Because C. acnes injections induced sustained MC1‐like lesions and TNF‐α injections increased MC2‐like lesions over time, it was hypothesized that the C. acnes group would exhibit higher pain‐like behavior than the TNF‐α group. Both TNF‐α and C. acnes groups had significantly lower hind paw withdrawal thresholds than Sham at any time point assessed, indicating increased mechanical allodynia suggestive of similarly increased pain‐like behavior (Figure 5A). Counter to our hypothesis, there was no difference between the injury group at any time point assessed. Sham animals maintained stable thresholds across the duration of the study showing sensitivity of the assay (Figure 5A).

FIGURE 5.

Pain‐like behavior and spinal cord sensitization was increased in both EP injury groups and SubP was significantly higher in the EP injury + C. acnes group. (A) Normalized hind paw withdrawal thresholds (% baseline) measured using von Frey testing over 13 weeks following EP injury with TNF‐α (blue squares) or C. acnes (pink triangles) injection, or sham surgery (black circles). Both EP injury groups demonstrated a significant and sustained reduction in mechanical thresholds compared to Sham, indicating long‐lasting mechanical hypersensitivity. No significant difference was observed between the EP injury + TNF‐α and EP injury + C. acnes groups. Data are shown as mean ± SD. Mixed‐effect analysis with Tukey post hoc analysis. *p < 0.05 for EP injury + TNF‐α versus Sham; #p < 0.05 for EP injury + C. acnes versus Sham. Time points 1–13 weeks: Sham: N = 4–12 rats, EP injury + TNF‐α: 6–19 rats, EP injury + C. acnes: 6–18 rats. (B) Random forest model showed that MC1‐like presence and area contributed strongest to pain‐like behavior. Graph shows %IncMSE. p values are FDR‐adjusted permutation‐based values. Black bars: p < 0.05. (C) Representative images of SubP expression spinal cord dorsal horn. (D) Quantification of SubP‐ir in the spinal cord dorsal horn area at 1‐, 8‐, and 14‐weeks. SubP levels were significantly increased in EP injury + TNF‐α and EP injury + C. acnes groups compared to Sham at 8‐ and 14‐weeks. SubP was significantly higher in EP injury + C. acnes versus EP injury + TNF‐α groups at 14‐WKs post‐injury. Significance bars represent results from Tukey's post hoc analysis. Time points weeks 1, 8, 14: Sham: N = 4, EP injury + TNF‐α: N = 6–7, EP injury + C. acnes: N = 5–6 spinal cords. (E) Representative images of GFAP expression in spinal cord dorsal horn. (F) Quantification of GFAP‐ir in the spinal cord dorsal horn area at 1‐, 8‐, and 14‐weeks. GFAP expression increased significantly in both EP injury + TNF‐α and EP injury + C. acnes groups compared to Sham, with a progressive increase from week 1 to week 14. There was no difference between injury groups. Significance bars represent results from Tukey post hoc analysis. Time points weeks 1, 8, 14: Sham: N = 4, EP injury + TNF‐α: N = 6–7, EP injury + C. acnes: N = 5–6. Bars represent mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001. C. acnes, Cutibacterium acnes; EP, endplate; GFAP, glial fibrillary acidic protein; IncMSE, percent increase in mean standard error; IQR, interquartile range; MC, modic change; SubP, substance P.

The random forest model, which identified MRI features contributing most to pain‐like behavior, demonstrated moderate predictive accuracy under LOOCV (R ² = 0.48, RMSE = 29.11, MAE = 22.45). MC1‐like lesion area (20.10%, p = 0.01) and presence (15.61%, p = 0.02) were the strongest predictors of pain‐like behavior in our model based on the %IncMSE. However, EP injury–related features like transverse EP defect (14.78%, p = 0.05) and disc extrusion area (13.68%, p = 0.03) also tended to significantly contribute to pain‐like behavior in this model (Figure 5B). These results indicated that MC1‐like lesions were the strongest contributors to pain‐like behavior, although our model showed EP injury–related changes also played a significant role.

Spinal cord sensitization and neuroinflammation were assayed as biochemical measures of pain since both are known to be associated with pain‐like behavior in rats [41]. Consistent with our hypothesis, SubP‐ir was significantly higher in spinal cord dorsal horn at 14 weeks post‐injury in the C. acnes compared to the TNF‐α group (p = 6.4 × 10⁻³). However, results were mostly in line with the behavioral outcomes, with significantly increased SubP and GFAP for both TNF‐α– and C. acnes compared to Sham, and no differences between injury groups (Figure 5C–F).

In summary, results showed that MC1‐like lesions were most predictive of pain‐like behavior and spinal cord SubP was increased in C. acnes at 14‐WKs post‐injury. Pain‐like behavior did not statistically differ between injury groups at any time point assessed and our findings indicated that EP‐injury had a significant impact on pain‐like behavior itself.

4. Discussion

This study evaluated the prevalence of MC subtypes and pain‐like behavior in an etiology‐specific MC rat model over time. It was found that intradiscal C. acnes injection following EP injury sustained MC1‐like prevalence over time, while TNF‐α injection led to an increase in MC2‐like lesions. Furthermore, our results indirectly indicated etiology‐specific differences in pain‐like behavior in our model: MC1‐like lesions were the strongest predictors of pain‐like behavior, and the C. acnes group showed higher MC1‐like lesion prevalence and increased spinal cord SubP levels. Nevertheless, mechanical allodynia did not differ between groups. This study provides the first evidence for potential etiology‐specific differences in MC subtypes, suggesting that MC patients of a bacterial etiology might be more likely to remain in a painful MC1 rather than convert to MC2.

Our etiology‐specific MC rat model recapitulated key features of MC patients. Consistent with clinical evidence that MC1 is the more painful subtype, MC1‐like lesions were the strongest contributors to pain‐like behavior [8, 9, 44]. Furthermore, MC‐like lesions in our model coincided with disc degeneration and EP damage, with no differences between C. acnes– and TNF‐α–injected discs, which aligns with findings from MC1 patients of different etiologies [13, 14, 15, 16, 17, 18, 23]. Moreover, consistent with findings from MC1 patients of different etiologies, our model exhibited distinct bone marrow immune cell profiles. C. acnes–injected discs showed increased bone marrow lesion NE‐ir, indicating elevated neutrophil presence and mirroring findings from patients with a bacterial MC1 etiology [23]. Predominant adaptive (B cell–driven) responses have been attributed to MC1 patients with a non‐bacterial etiology [23]. However, the elevated CD19‐ir in our model is not inconsistent with clinical findings, as neutrophil predominance does not preclude additional B‐cell involvement, and an adaptive immune response to C. acnes is biologically plausible. Taken together, this model reproduces key aspects of the MC pathology and allows meaningful parallels to be drawn with the human condition.

MC1‐like lesions may be sustained by persistent C. acnes‐mediated recruitment and activation of bone marrow neutrophils. C. acnes is an aerotolerant anaerobic commensal skin bacterium frequently found in discs of MC patients [23, 45]. The disc's low oxygen tension and pH, and limited immune surveillance create an environment that favors ongoing intradiscal C. acnes proliferation and virulence factor production [46]. These bacterial factors may provide a persistent pro‐inflammatory stimulus that drains through damaged EPs into the adjacent bone marrow, thereby recruiting and activating bone marrow neutrophils. This is supported by the increased NE‐ir in bone marrow lesions adjacent to C. acnes–injected discs, despite comparable levels of intradiscal inflammatory burden between groups. While TNF‐α served as our primary readout of discal inflammatory burden, other cytokines involved in disc degeneration, especially such as interleukin‐1β (IL‐1β), IL‐6, or IL‐8 may diverge between groups and contribute to distinct downstream immune responses [42, 47, 48, 49, 50, 51, 52, 53]. However, injectate‐dependent differences in adjacent bone marrow immune responses suggest that intradiscal inflammation alone is insufficient to sustain MC1‐like lesions and that neutrophil‐attracting and ‐activating stimuli may be critical for MC1 persistence. This aligns with clinical observations showing that neutrophilic infiltrates are specific to MC1, whereas MC2 is characterized by neutrophil clearance [31, 32]. C. acnes–mediated neutrophil recruitment and activation is plausible, as C. acnes has been shown to recruit and activate neutrophils both in vitro and in vivo [54]. Due to the high oxygen tension and immune surveillance in the bone marrow, it is unlikely that C. acnes colonizes the bone marrow and directly attracts and activates neutrophils in MC. However, it has been shown that in response to C. acnes stimulation in vitro, NP cells secrete elevated levels of IL‐8 [55], a key chemoattractant and activator of neutrophils. Moreover, C. acnes released factors like lipase and propionic acid can attract neutrophils [56, 57]. Thus, the elevated NE‐ir in bone marrow lesions adjacent to C. acnes–injected discs may reflect neutrophilic infiltration driven by disc‐derived cytokines produced in response to the bacteria, or by bacterial products themselves. This chronic neutrophil recruitment and activation may reinforce tissue damage and help maintain MC1‐like inflammatory lesions [30, 58, 59]. Our results, aligned with existing literature, suggest that C. acnes sustained MC1‐like lesions by continuously stimulating neutrophil recruitment and activation in adjacent bone marrow, thereby maintaining chronic neutrophil‐mediated inflammation and preventing MC conversion.

MC1 persistence may be driven not only by bacterial etiology but also by other intradiscal pro‐inflammatory stimuli capable of recruiting and activating neutrophils. It should be noted that likely a key aspect of the bacterial MC etiology was mimicked by injecting C. acnes into the discs. However, the non‐bacterial etiology is likely more complex and may involve an autoimmune reaction of bone marrow leukocytes against disc fragments. Injecting TNF‐α into the disc mimics a pro‐inflammatory stimulus, but this likely does not fully recapitulate the non‐bacterial MC etiology. Recently, it was shown that MC discs contain a higher abundance of pro‐inflammatory matrix fragments generated by cleavage through the high‐temperature requirement serine protease A1 (HTRA1) [60]. These cartilaginous components are potential danger‐associated molecular patterns (DAMPs) that bind to Toll‐like receptors 2 (TLR2) and TLR4 on innate immune cells, such as neutrophils, and activate them [61, 62]. Therefore, although our data indicated that C. acnes sustained MC1‐like lesions through a neutrophil‐mediated mechanism, similar neutrophil‐driven pathways may also arise in the absence of bacteria, for example through continuous activation by DAMPs released from degenerated disc tissue. Thus, neutrophil persistence itself, rather than bacterial etiology alone, may prevent lesion conversion. These findings underscore the need for future mechanistic studies investigating the drivers of MC conversion.

Understanding MC conversion has implications on patient pain management and development of disease‐modifying MC treatments. MC1 seems to be the most painful lesions and have the highest positive predictive value for discography concordant pain [3, 8, 9, 10, 11, 63, 64]. This was consistent with our finding that MC1‐like lesions contributed strongest to pain‐like behavior. Moreover, conversion from MC1 to MC2 lesions has been associated with a decrease in pain in MC patients [11]. Although differences were observed in MC1‐ and MC2‐like prevalence between injury groups, the differences in pain‐like behavior were less definitive. This discrepancy between MC patients and our model could result from the acute, large injury that was induced, which was demonstrated to be painful itself. Notably, changes in spinal cord SubP, which is associated with pain‐like behavior in other rat models [41], showed group differences at 14 weeks post‐injury, suggesting that extended timepoints may be needed to detect significant differences in mechanical allodynia between groups. However, the finding of our study that discal C. acnes injection led to sustained MC1‐like lesions over time may have important prognostic implications for MC patients of different etiologies and underscores the importance to develop non‐invasive etiology‐specific diagnostic approaches. Emerging approaches, such as magnetic resonance spectroscopy and blood‐based biomarkers, show great promise for future non‐invasive etiology‐specific diagnostics [23, 65]. Similarly, there are no disease‐modifying MC treatments available yet. Jensen et al. [12] studied the progression of MC subtypes using MRI over a 4‐year period in a prospective, observational, population‐based cohort study. They demonstrated that the greater the bone marrow involvement, the less likely the MC lesions were to resolve. These findings suggest that accelerating the conversion of painful MC1 to less painful MC2 subtypes might be a more promising treatment approach than attempting complete resolution [12]. Eliminating the persistent pro‐inflammatory stimulus may reduce the recruitment of adjacent neutrophils and activate and promote the conversion of MC1–MC2. This study, showing that intradiscal C. acnes sustains MC1‐like lesions over time, therefore supports considering antibiotics as a potential treatment strategy. The use of oral antibiotics for the treatment of MC patients has shown to be somewhat effective in some patients; however, the side effects of these long‐term treatments can be very severe and it is unclear how systemic antibiotics target the disc [66, 67, 68]. Moreover, the use of antibiotics for all MC patients is controversial, specifically when considering that bacterial and non‐bacterial etiologies exist. Nevertheless, injecting antibiotics into MC discs could be an effective strategy for MC patients of a bacterial etiology, thereby eliminating the persistent pro‐inflammatory stimulus that sustains MC1 and hinders conversion. A small clinical trial is currently investigating the effect of intradiscal antibiotic injection in MC1 patients, which might be a promising conversion treatment approach [69]. However, a growing body of evidence supports that the disc itself harbors a microbiome [24, 25, 43]. Therefore, antibiotic use should be approached with caution. Persisting bone marrow lesion neutrophils may also represent an additional therapeutic target for promoting the transition from MC1 to MC2. Since neutrophil modulation has emerged as a novel treatment approach for several rheumatic and inflammatory diseases, similar strategies could be effective for treating MC1 lesions that are characterized by persistent neutrophilic activity [70]. Therefore, interventional studies targeting intradiscal C. acnes and/or bone marrow neutrophils in this preclinical model would provide valuable information for testing the proposed pathomechanisms and for screening potential MC conversion therapies. In addition, both EP‐injury groups, particularly the discs injected with C. acnes at 14 weeks, showed evidence of central sensitization. This suggests that central mechanisms may sustain pain even after peripheral inflammation resolves, making spinal cord sensitization an important therapeutic target. Taken together, investigating MC conversion mechanisms may provide valuable insights for patient management and help identify key therapeutic targets to improve outcomes in MC‐related pain.

This exploratory study has some relevant limitations to consider. First, consecutive imaging of the same animals was not possible because a sufficiently large coil capable of achieving the required resolution for our MRI readouts was unavailable. Therefore, statements were made about MC subtype prevalence since it was not possible to conclude about lesion conversion over time, although a sufficient sample size to compare across time points without a repeated measures analysis was provided. Second, MRI was performed ex vivo following fixation, which may alter tissue appearance on MRI. However, Sham images acquired before and after fixation showed no major differences on our sequences, and imaging was conducted at a standardized time point after fixation. Thus, our overall conclusions would not likely be altered had imaging been conducted in vivo. Nevertheless, future studies are required to confirm these findings in vivo. Third, the EP injury was intentionally designed to promote disc–bone marrow crosstalk but was very severe compared to clinical conditions. Future studies employing milder injury models may therefore better capture clinically relevant tissue responses and pain‐like behavior. Moreover, an EP injury–only was not included, or groups with injections and no EP injury due to sample size constraints related to ethical use of animals. Such groups would provide valuable insights into the isolated effects of EP injury or injectate only yet considered beyond the scope of the present study. The groups and time points priorities were designed to evaluate our main objective of characterizing etiology‐specific MC subtype lesions over time. Because only male rats were included in this study, sex‐specific differences in immune responses, pain‐like behavior, and disc biology were not evaluated, which may limit translational relevance. In addition, although rats are a well‐established model for inducing MC‐like lesions, species‐specific differences in spinal anatomy, biomechanics, and immune responses limit direct translation, and these findings will require validation in humans. Moreover, a laboratory reference C. acnes strain purchased from ATCC (ATCC #6919; isolated from facial acne/skin) was used, and no additional phylotype/lineage characterization was performed. Given evidence that distinct C. acnes phylotypes vary in inflammatory potential and can elicit different Modic‐like MRI signal patterns in vivo, these findings may not fully reflect responses to clinical spine/IVD isolates [71, 72, 73]. Finally, we injected substantially higher concentrations of C. acnes than the ones reported from human studies [23, 74]. Therefore, it remains to be determined whether the lower bacterial concentrations reported in MC1 patients would produce comparable effects.

In conclusion, intradiscal C. acnes sustained MC1‐like lesions in an etiology‐specific lumbar EP injury MC rat model, accompanied by sustained neutrophil‐driven inflammation in the surrounding bone marrow and greater spinal cord SubP. This suggests that bacterial activation of neutrophils may be a key determinant of MC1 persistence. Therapeutic strategies aimed at reducing neutrophil activity or eradicating C. acnes could thus represent promising approaches to promote MC1‐to‐MC2 conversion and hence improve pain and disability in cLBP patients.

Author Contributions

Study conception and design: Irina Heggli, Timothy D. Jacobsen, and James C. Iatridis. Data acquisition: Irina Heggli, Alon Lai, Denise Iliff, Marco D. Burkhard, Niklas Koehne, Alan C. Seifert, Harsev Singh, Jonathan J. Huang, Damien M. Laudier, Noah Bonnheim, Levon Rodriguez, and Timothy D. Jacobsen. Data analysis: Irina Heggli, Levon Rodriguez, Timothy D. Jacobsen, and James C. Iatridis. Manuscript drafting: Irina Heggli. All co‐authors edited and reviewed the manuscript. Funding for this project was provided by James C. Iatridis and Irina Heggli.

Funding

This work was supported by the National Institute of Arthritis and Musculoskeletal and Skin Diseases, R01AR078857 and R01AR080096; Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung, P500PB_217823.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: MRI grading scheme. (A) Degree of disc degeneration (0–4). (B) Transverse EP disruption grade (0–3). (C) Disc extrusion into marrow presence (0,1) and area (% of L5 vertebral body). (D) MC subtype lesions presence (0,1) and area (% of L5 vertebral body).

Figure S2: Histology grading key. (A) Intradiscal TNF‐α was graded from 0 to 3 in ALL + granulation tissue, AF, NP, EP, and posterior tissue. (B) Bone marrow lesions were graded from 0 to 3 for NE‐ir (top) and CD19‐ir (bottom).

Figure S3: Rat weight over time. Rats from all 3 groups gained weight over time without any differences at each time point between groups. Mixed‐effect analysis. Time points: 0–1 weeks: n = 49, TPs 2–8 weeks: n = 33, TPs: 9–14 weeks: n = 16.

Figure S4: Histological disc degeneration separated by region. Nucleus pulpous (NP), annulus fibrosis (AF), NP‐AF border, and EP regions from both EP injury groups were significantly more degenerated compared to sham at all 3 time points assessed (week 1, 8, and 14). There were no significant differences in any of the regions between discs injected with TNF‐α versus C. acnes at any timepoint assessed. Bars represent mean ± standard deviation (SD). Two‐way analysis of variance (ANOVA), significance bars represent pairwise comparisons from Tukey's Post hoc analysis. Time points weeks 1, 8, and 14: Sham: n = 6–7 discs, TNF‐α: n = 10–12, C. acnes: n = 9–12 discs. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure S5: Regional differences in TNF‐α‐ir. Like for the whole disc, there was no difference between both injury groups in any region at all timepoints assessed.

Figure S6: C. acnes immunohistochemistry in C. acnes positive culture (positive control) and brain tissue (negative control). Both commercially available anti‐C. acnes antibodies did not show specific staining for C. acnes.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.