A clinico-molecular classification scheme for ossification of ligamentum flavum: Based on 1034 cases

Abstract

Introduction

Thoracic ossification of the ligamentum flavum (TOLF) is characterized by ectopic bone formation within the ligamentum flavum. It progresses gradually over an extended period, often leading to severe spinal cord damage, resulting in significant myelopathy that typically necessitates surgical intervention. While various factors have been suggested to contribute to its development, the exact cause of TOLF remains unclear.

Objective

To develop a novel classification system for TOLF by integrating clinical, radiological, and molecular characteristics to better understand and explore its pathogenesis.

Methods

We conducted a retrospective analysis of 1034 surgically treated TOLF patients at our institution from January 2010 to December 2023. The study incorporated four clinical dimensions: consensus clustering (n = 76), epidemiological distribution (n = 1034), imaging characteristics (n = 399), and clinical feature assessment (n = 586). Molecular analysis was performed on surgical specimens from 11 representative cases.

Results

Our analysis revealed two distinct TOLF subtypes with characteristic profiles: (1) The degenerative type (75 % of cases) primarily affected elderly patients (mean age 60 years) with normal body mass index(BMI), with localized involvement (≤3 levels) in the upper/lower thoracic spine. This subtype demonstrated significant biomechanical alterations, including changes in C2–7 SVA and LL, as supported by finite element analysis showing stress changes in the ligamentum flavum. Molecular analysis revealed the activation of wound-healing, ossification, and inflammatory pathways. (2) The metabotropic type (25 %) predominantly affected younger patients with abdominal obesity and metabolic syndrome, who exhibited multilevel involvement with preserved spinal curvature but elevated blood lipid levels and distinct dysregulation of lipid/energy metabolism pathways.

Conclusion

A novel TOLF classification system consisting of two subtypes, namely, degenerative and metabotropic, each with distinct clinical, biomechanical, and molecular profiles, was developed. The degenerative subtype is primarily associated with mechanical stress and the ageing process, whereas the metabotropic subtype is linked to systemic metabolic dysfunction.

Translational potential statement

The translational potential of this article lies in its innovative TOLF classification system, which offers a comprehensive approach to understanding the disease from clinical, biomechanical, and molecular perspectives. By identifying two distinct TOLF subtypes—degenerative and metabotropic—this study provides valuable insights into their unique pathophysiological mechanisms. This classification can facilitate more accurate diagnosis and enable the development of targeted, personalized treatment strategies tailored to the specific subtype, whether through biomechanical interventions for the degenerative form or metabolic therapies for the metabotropic type. Additionally, the identified molecular pathways could lead to novel therapeutic targets for TOLF, improving patient outcomes and reducing the need for invasive procedures.

Graphical abstract

1. Introduction

Thoracic ossification of the ligamentum flavum (TOLF) is characterized by ectopic bone formation within the ligamentum flavum and is relatively common in East Asia, with reported prevalence rates ranging from 3.8 % to 63.9 %, particularly among Japanese and Chinese people [1]. This condition progresses gradually over an extended period, often leading to severe spinal cord damage, resulting in significant myelopathy that typically necessitates surgical intervention. Since its initial description in 1920, extensive genetic and epidemiological research has demonstrated that TOLF is a complex, multifactorial disorder influenced by the interplay of genetic and environmental factors. Numerous factors have been implicated in its pathogenesis, including genetic predisposition, ageing, gender, hormonal influences, inflammatory mediators, and mechanical stress [[2], [3], [4]]. However, the precise aetiology of TOLF remains elusive.

Current classification systems for TOLF focus primarily on morphological characteristics, with the Sato classification (lateral, extensive, enlarged, fusion, and nodular types) and Kuh classification (beak and round types) being the most widely adopted in clinical practice [5,6]. However, the ability of these systems to elucidate the underlying mechanisms of ossification is limited. Some scholars have also reported that single-level and multilevel lesions have different clinical characteristics [7]. Single-level lesions occur predominantly in the lower thoracic spine (T10–T12), a mobile transition zone where axial mechanical overload induces repetitive tensile stress on the ligamentum flavum—a phenomenon supported by both cellular studies and animal models [[8], [9], [10]]. In contrast, multilevel TOLF typically involves relatively immobile segments and demonstrates different clinical progression patterns. Furthermore, molecular studies have revealed variations in osteogenic differentiation potential and genetic predisposition among different TOLF subtypes [11,12]. These findings collectively suggest the existence of distinct pathogenic subtypes within TOLF, highlighting the need for a more mechanistic classification system that integrates morphological, biomechanical, and molecular characteristics.

In this study, we propose a TOLF classification across three levels with five sequential steps, spanning from clinical phenotypes to molecular mechanisms. By establishing this multidimensional classification system, we aim to standardize TOLF characterization and foster more targeted investigations into its pathophysiology and treatment.

2. Materials and methods

2.1. Study design

The local ethics committee approved the study (M2021317). While informed consent was waived for the retrospective data analysis, written consent was obtained from all participants involved in sample acquisition. The study included all patients diagnosed with TOLF since medical record information became available (between January 2010 and December 2023). The overall design is illustrated in Fig. 1.

Fig. 1.

Flow chart of the article.

2.2. Pilot clustering analysis

2.2.1. Consensus clustering

Patients who were diagnosed with TOLF on the basis of clinical symptoms and radiological findings between January 2010 and June 2010 were included in this analysis. The exclusion criteria were as follows: (1) the presence of ossification of the posterior longitudinal ligament (OPLL), diffuse idiopathic skeletal hyperostosis (DISH), or thoracic disc herniation; (2) a history of spinal surgery; (3) incomplete clinical data; and (4) the presence of spinal infections, trauma, tumours, ankylosing spondylitis, spinal deformities, or autoimmune diseases.

Seventy-six patients with TOLF were enrolled in this step. Age, BMI, triglyceride level, total cholesterol level, low-density lipoprotein level, high-density lipoprotein level and involved segments were recorded. Consensus clustering analysis is described in the supplementary methods and was performed to identify distinct subtypes, with the optimal cluster number determined by evaluating average silhouette values across potential cluster solutions (k = 2–6).

2.3. Large-sample verification

2.3.1. Epidemiology

A cross-sectional survey was conducted between January 2010 and December 2023. Patients diagnosed with TOLF on the basis of clinical symptoms and radiological findings were enrolled in the study. The exclusion criteria were consistent with those described in Step 1.

A total of 1034 patients with TOLF, consisting of 519 men and 515 women, were included in this part of the study. The mean age was 54.20 ± 10.52 years, and the mean BMI was 27.42 ± 4.48 kg/m². TOLF levels were recorded for all participants.

2.3.2. Image features

Patients who were diagnosed with TOLF and who underwent whole-spine radiography via X-rays were included. The exclusion criteria were consistent with those described in the pilot clustering analysis. A total of 399 patients (mean age: 55.87 ± 9.78 years) were included in this part of the study.

To create a comparable control group matched for age and gender, patients without TOLF were selected from a database of individuals who underwent whole-spine radiography and sagittal MRI for low back pain during the study period. The inclusion criteria were limited to patients diagnosed with lumbar spondylosis and without thoracic spine compressive lesions, as confirmed by radiological examinations. The same exclusion criteria used for the TOLF group were applied during the recruitment of the control cohort.

Methods for measuring the sagittal plane parameters and establishing finite element models are provided in the supplementary materials.

2.3.3. Clinical features

Patients who were diagnosed with TOLF and whose clinical data were complete were included. The clinical data included gender, age, BMI, waist circumference, comorbidities, routine blood test results, bone turnover markers, blood fat and CT values of the thoracic spine. The exclusion criteria were consistent with those described in the pilot clustering analysis. A total of 586 TOLF patients, comprising 338 males and 248 females, were included.

The medical data were divided into baseline characteristics, laboratory test results, and imaging characteristics and recorded. The measurement methods are described in the supplementary material.

2.3.4. Metabolomics and proteomics

Ligamentum flavum specimens were collected from 8 patients with TOLF and 3 patients with thoracic disc herniation. Proteomic and metabolomic analyses were subsequently performed and are described in the supplementary material. Patient information is provided in Supplementary Table 1.

2.4. Statistical analysis

The data were analysed using SPSS version 29.0. Quantitative data are expressed as the mean ± standard deviation and were assessed by one-way ANOVA and t-tests. Categorical data are presented as percentages and were compared using the X² test. Differences were considered significant if the p value was <0.05.

3. Results

3.1. Identification of two distinct TOLF subtypes with discriminative features

The results of the consensus clustering analysis strongly supported a two-cluster classification model (Fig. 2a), revealing two distinct phenotypic patterns with significant clinical and biochemical differences.

Fig. 2.

Consensus clustering and clinical features of the two types of TOLF a. Heatmap of consensus clustering and average silhouette width plots. b. Heatmap depicting the clinical features of the two clusters. HTN: hypertension; DM: diabetes mellitus; CVD: cardiovascular disease; CeVD: cerebrovascular disease; Op: osteoporosis; TG: triglyceride; LDLC: low-density lipoprotein; HDLC: high-density lipoprotein; TC: total cholesterol. Disease state: 0: nondiseased; 1: diseased. c. Differences in age between the two clusters. d. Differences in BMI between the two clusters. e. Differences in high-density lipoprotein between the two clusters. f. Differences in segment between the two clusters. g. ROC curves for age, BMI, high-density lipoprotein level and segments.

Cluster 1 exhibited predominantly degenerative features, characterized by the following.

• •Advanced age
• •Mildly elevated BMI
• •Increased incidence of osteoporosis
• •Few diseased segments

Cluster 2 exhibited systemic metabolic involvement, featuring the following.

• •Younger age at presentation
• •Significant obesity
• •Multilevel spinal involvement
• •Frequent comorbidities (cardiovascular disease and diabetes mellitus) (Fig. 2b–f)

Considering that factors such as age, BMI, and other variables can be influenced by a range of elements, the number of involved segments has emerged as a key discriminator between subtypes because of its stability. Through integrated mean comparison and cut-off value analysis, we determined that ≤3 involved segments served as the optimal discriminative threshold for subtype classification (Fig. 2d). On the basis of distinct clinical profiles, we designated the following:

Degenerative TOLF (Cluster 1): Characterized by age-related pathogenesis.

Metabotropic TOLF (Cluster 2): Defined by systemic metabolic dysfunction.

3.2. Degenerative TOLF is associated with mechanical stress

The distribution of TOLF cases is illustrated in Fig. S1: 66.68 % were located in the lower thoracic spine (T8–T12), 18.09 % were in the middle thoracic spine (T4–T8), and 15.23 % were in the upper thoracic spine (T1–T4). In the degenerative TOLF group, 98.77 % of the cases were in the junctional area, whereas only 1.23 % were in the middle thoracic region. In contrast, among the metabotropic TOLF cases, 33.93 % were located in the middle of the thoracic spine (Fig. S1).

On the basis of the radiographic findings, compared with the control group, the degenerative-type upper TOLF group had a significantly greater C7 SVA (1.74 ± 1.95 cm vs. −0.37 ± 2.74 cm, P = 0.002), C2–7 SVA (2.92 ± 1.09 cm vs. 2.08 ± 1.10 cm, P = 0.003), and SS (39.39 ± 9.69° vs. 34.31 ± 6.91°, P = 0.009). Additionally, the T1 pelvic angle (4.19 ± 6.07° vs. 8.77 ± 6.79°, P = 0.007) and PT (8.45 ± 6.13° vs. 13.95 ± 7.06°, P = 0.002) were significantly smaller. Degenerative-type lower TOLF demonstrated significant differences in the C2–T1 angle (1.94 ± 1.05° vs. 2.42 ± 0.96°, P < 0.01), C7 SVA (1.30 ± 4.30 cm vs. −0.37 ± 2.74 cm, P = 0.002), T9 slope (3.51 ± 8.55° vs. 8.08 ± 5.54°, P < 0.01), T12 slope (8.92 ± 8.27° vs. 13.31 ± 6.00°, P < 0.01), T1 SPI (3.20 ± 4.20° vs. 5.17 ± 2.82°, P < 0.01), TLK (13.50 ± 13.68° vs. 8.79 ± 9.87°, P = 0.006), and LL (38.50 ± 22.71° vs. 49.69 ± 9.85°, P < 0.01). In contrast, compared with the control group, the metabotropic TOLF group did not show any significant differences in sagittal alignment parameters (Table 1).

Table 1.

Comparison of sagittal alignment among OLF subgroups and control group.

	Degenerative type		Metabotropic type (n = 94)	Control group (n = 75)	P1	P2	P3
	Upper thoracic (n = 20)	Lower thoracic (n = 277)
Sagittal parameters
Cervical spine
C2T1 (°)	2.35 ± 0.82	1.94 ± 1.05	2.18 ± 1.16	2.42 ± 0.96	0.79	<0.01	0.15
C2slope (°)	10.28 ± 8.43	6.61 ± 8.74	7.10 ± 9.93	6.83 ± 8.25	0.11	0.85	0.86
C7slope (°)	17.19 ± 9.29	20.72 ± 7.93	20.84 ± 8.15	19.61 ± 7.91	0.25	0.28	0.33
CL (°)	9.06 ± 8.32	14.20 ± 11.38	15.13 ± 11.18	13.31 ± 10.57	0.10	0.55	0.29
C7 SVA (cm)	1.74 ± 1.95	1.30 ± 4.30	0.43 ± 2.86	−0.37 ± 2.74	0.002	0.002	0.07
C2-7 SVA (cm)	2.92 ± 1.09	1.95 ± 1.17	2.25 ± 1.25	2.08 ± 1.10	0.003	0.41	0.36
Thoracic spine
T1slope (°)	17.97 ± 10.20	21.85 ± 8.59	22.25 ± 9.04	19.78 ± 8.03	0.40	0.06	0.07
T1pelvic (°)	4.19 ± 6.07	8.56 ± 8.25	6.78 ± 8.25	8.77 ± 6.79	0.007	0.84	0.09
T9slope (°)	7.47 ± 5.42	3.51 ± 8.55	6.55 ± 6.78	8.08 ± 5.54	0.66	<0.01	0.12
T12slope (°)	12.85 ± 6.73	8.92 ± 8.27	11.66 ± 7.47	13.31 ± 6.00	0.76	<0.01	0.12
TK (°)	28.43 ± 10.03	28.60 ± 11.49	30.51 ± 11.81	30.27 ± 9.56	0.45	0.25	0.8
T1SPI (°)	4.26 ± 4.15	3.20 ± 4.20	4.43 ± 2.90	5.17 ± 2.82	0.25	<0.01	0.10
TLK (°)	8.57 ± 8.81	13.50 ± 13.68	10.18 ± 10.38	8.79 ± 9.87	0.93	0.006	0.38
Lumbar spine
L4S1 (°)	36.91 ± 6.01	36.41 ± 9.82	36.10 ± 7.28	35.12 ± 8.17	0.43	0.31	0.43
LL (°)	50.98 ± 10.11	38.50 ± 22.71	49.80 ± 12.10	49.69 ± 9.85	0.61	<0.01	0.95
LPA (°)	1.67 ± 5.88	2.53 ± 7.68	3.85 ± 6.01	3.96 ± 6.29	0.15	0.14	0.90
L1S1 (°)	5.47 ± 8.34	7.34 ± 8.66	6.18 ± 7.72	7.59 ± 6.03	0.20	0.81	0.20
Pelvic
PI (°)	47.84 ± 7.71	45.67 ± 12.20	45.76 ± 11.66	48.26 ± 9.26	0.85	0.09	0.13
PT (°)	8.45 ± 6.13	11.84 ± 8.74	12.27 ± 7.52	13.95 ± 7.06	0.002	0.06	0.14
SS (°)	39.39 ± 9.69	34.04 ± 10.66	35.89 ± 9.60	34.31 ± 6.91	0.009	0.84	0.23

Values in bold indicates P value < 0.05, which is considered as significance difference. P1 is a comparison between the degenerative type-upper thoracic and control group. P2 is a comparison between the degenerative type-lower thoracic and control group. P3 is a comparison between the metabotropic TOLF and control group.

In the degenerative-type upper TOLF group (Fig. 3a), the C2–7 sagittal vertical axis (C2–7 SVA) strongly correlated with the occupation rate (r = 0.79, P < 0.01). In the degenerative-type lower TOLF group (Fig. 3b), although the C7 SVA (r = 0.173, P = 0.004), T9 slope (r = 0.17, P = 0.007), T12 slope (r = 0.31, P < 0.01), and lumbar lordosis (LL) (r = −0.64, P < 0.01) were significantly correlated, only LL was strongly correlated with the occupation rate (Supplementary Table 2).

Fig. 3.

Different finite element models a. Correlations between C2–7 and the occupation rate in upper TOLF. b. Correlations between LL and the occupation rate in lower TOLF. c. Healthy spine model and the average stress on the ligamentum flavum (flexion). d. Upper TOLF spine model and the relative compressive stress on the ligamentum flavum (flexion). The X axis represents the fold change in stress, with the healthy spine set to 1.0. e. Lower TOLF spine model and the relative compressive stress on the ligamentum flavum (flexion). The X-axis represents the fold change in stress, with the healthy spine set to 1.0.

The parameters of the components are presented in Supplementary Table 3. In the healthy model (Fig. 3c), the lower thoracic segments exhibited the highest ligamentum flavum stresses across the entire thoracic region. In the upper TOLF model (Fig. 3d), stress elevation was observed in the upper thoracic ligamentum flavum compared with that in the healthy model. In the lower TOLF model (Fig. 3e), the lower thoracic segments demonstrated increased stresses relative to those in the healthy model.

3.3. Associations of the two distinct TOLF subtypes with clinical features (Table 2)

Table 2.

Clinical Features of two types of OLF.

	Degenerative type (n = 435)	Metabotropic type (n = 151)	P value
Baseline characteristics
Sex (male/female)	241/194	97/54	0.06
Age	58.74 ± 8.30	49.08 ± 10.91	<0.01
18–44	16 (3.7 %)	51 (33.8 %)
45–59	214 (49.2 %)	68 (45.0 %)
Over 60	205 (47.1 %)	32 (21.2 %)
BMI	23.2 ± 3.06	30.34 ± 5.07	<0.01
18.5–24.9	318 (73.1 %)	19 (17.9 %)
25.0–29.9	105 (24.1 %)	64 (60.4 %)
>30.0	12 (2.8 %)	23 (21.7 %)
Waist circumference	89.8 ± 16.42	101.7 ± 15.45	<0.01
Coexistence of hypertension or diabetes (n%)	134 (30.8 %)	87 (57.6 %)	<0.01
Coexistence of Metabolic Syndrome (n%)	91 (20.9 %)	110 (72.8 %)	<0.01
Laboratory test results
Routine blood test
NLR	5.41 ± 4.55	5.90 ± 4.78	0.010
PLR	166.41 ± 99.51	174.37 ± 108.08	0.043
MLR	0.40 ± 0.32	0.40 ± 0.33	0.38
SII	1126.05 ± 979.83	1348.94 ± 1192.25	<0.01
Bone turnover marker
Osteocalcin (14–46 ng/ml)	18.47 ± 10.41	18.28 ± 8.92	0.44
β-CTX (0–0.704ngml)	0.60 ± 0.33	0.77 ± 1.62	0.042
TP1NP (15.3–52.7 ng/ml)	55.58 ± 27.07	64.84 ± 67.69	0.05
VD (≥20 ng/ml)	17.12 ± 8.75	19.22 ± 8.95	0.037
PTH (15–65 pg/ml)	38.94 ± 14.14	43.28 ± 22.78	0.059
Blood fat
Total cholesterol (<5.18 mm/L)	4.88 ± 1.29	5.42 ± 1.24	0.12
Triglyceride (<1.7 mmol/L)	1.70 ± 0.71	2.58 ± 1.24	0.02
High density lipoprotein (≥1.04 mmo/L)	1.12 ± 0.50	0.85 ± 0.39	<0.01
Low density lipoprotein (<3.64 mmol/L)	2.99 ± 1.16	3.13 ± 1.03	0.24
Imaging characteristics
Number of involved segment	1.86 ± 0.78	5.39 ± 2.13	<0.01
Average CT value (T1-12)	186.55 ± 48.98	237.51 ± 53.09	<0.01

Values in bold indicates P value < 0.05, which is considered as significance difference.

The mean age for the degenerative type was 58.74 ± 8.30 years (with only 3.7 % aged between 18 and 44 years), and the mean BMI was 23.2 ± 3.06 kg/m² (73.1 % within the normal BMI range). In contrast, the mean age for the metabotropic type was 49.08 ± 10.91 years (33.8 % aged between 18 and 44 years), and the mean BMI was 30.34 ± 5.07 kg/m² (only 17.9 % within the normal BMI range). With respect to waist circumference and comorbidities, the metabotropic type was associated with a greater waist circumference and a greater incidence of hypertension, diabetes, and metabolic syndrome.

In terms of inflammation-related variables, the NLR (5.90 ± 4.78 vs. 5.41 ± 4.55; P = 0.01), PLR (174.37 ± 108.08 vs. 166.41 ± 99.51; P = 0.043), and SII (1348.94 ± 1192.25 vs. 1126.05 ± 979.83; P < 0.01) were significantly greater in the metabotropic type group than in the degenerative type group. With respect to bone turnover markers and blood lipid profiles, the metabotropic type group had significantly higher levels of β-CTX (0.77 ± 1.62 ng/ml vs. 0.60 ± 0.33 ng/ml, P = 0.042), TP1NP (64.84 ± 67.69 ng/ml vs. 55.58 ± 27.07 ng/ml, P = 0.05), VD (19.22 ± 8.95 ng/ml vs. 17.12 ± 8.75 ng/ml, P = 0.037), and triglycerides (2.58 ± 1.24 mmol/L vs. 1.70 ± 0.71 mmol/L, P = 0.02), along with significantly lower levels of HDL (0.85 ± 0.39 mmol/L vs. 1.12 ± 0.50 mmol/L, P < 0.01).

The measurement of CT Hounsfield unit (HU) values is straightforward and involves the use of the tissue density of vertebral trabecular bone mass to estimate bone mineral density (BMD). The average CT value of the thoracic spine (T1–T12) was significantly greater for the metabotropic type (237.51 ± 53.09 HU) than for the degenerative type (186.55 ± 48.98 HU).

3.4. Molecular features

Ligamentum flavum specimens were collected from 8 patients with TOLF and 3 patients with thoracic disc herniation. Among them, 4 TOLF specimens were degenerative (Group A), 4 TOLF specimens were metabotropic (Group B), and the 3 thoracic disc herniation specimens comprised Group C.

A total of 6960 proteins were identified, of which 3398 were selected on the basis of a P value < 0.05 and a fold change ≥2.0 or ≤0.5. Compared with those in Group C, 634 proteins were upregulated and 322 were downregulated in Group A, while 1240 proteins were upregulated and 250 were downregulated in Group B. Additionally, 801 proteins were upregulated and 151 were downregulated in Group B compared with those in Group A.

A total of 2750 metabolites were identified across the three groups. Among these, 1140 metabolites were selected on the basis of a P value < 0.05 and a fold change ≥1.2 or ≤0.83. Specifically, 350 metabolites were upregulated and 134 were downregulated in Group A compared with those in Group C; 381 metabolites were upregulated and 136 were downregulated in Group B compared with those in Group C; and 93 metabolites were upregulated and 46 were downregulated in Group B compared with those in Group A.

3.5. Bioinformatics analysis reveals distinct pathogenic mechanisms among TOLF types

With respect to proteomics and metabolomics, the three groups were in three distinct physiological states (Fig. 4a and b). After the two types of omics data were combined, this finding was confirmed through an unsupervised consensus clustering algorithm (Fig. 4c).

Fig. 4.

Molecular features of the two subtypes of TOLF a. PCA of the proteome across the three groups. b. PLS‒DA of the metabolome across the three groups. c. Consensus clustering of the three groups. d. Network plot of significantly up- and downregulated pathways identified by the GSEA algorithm in Group A, with criteria of an |NES| > 1 and a p value < 0.01. e. Network plot of significantly up- and downregulated pathways identified by the GSEA algorithm in Group B, with criteria of an |NES| > 1 and a p value < 0.01. f. GSEA network plot of significantly differentially expressed proteins between Group B and Group A, with thresholds set as a |log2FC| ≥ 1 and a q value < 0.05. g. ssGSEA based on hallmark genes. h. K-means clustering of differentially expressed proteins and metabolites revealed distinct expression patterns. Differentially expressed proteins and metabolites were identified using ANOVA with a significance threshold of p < 0.05. i. ORA network plot of significantly differentially expressed proteins between Group B and Group A, with thresholds set as a |log2FC| ≥ 1 and a q-value <0.05.

Comprehensive bioinformatics analyses revealed fundamental differences in the pathogenic underlying mechanisms associated with Groups A and B. GSEA demonstrated that Group A was characterized by significant upregulation of inflammatory responses and negative regulation of wound-healing pathways (Fig. 4d–f). In contrast, Group B a exhibited marked disruptions in RNA metabolism, translation processes, and methylation pathways. Further investigation using ssGSEA of hallmark gene sets revealed additional molecular distinctions between the groups. Group A showed pronounced activation of hypoxia-related pathways (hypoxia, angiogenesis, and glycolysis), suggesting an adaptive response to ischaemic conditions (Fig. 4g). Conversely, Group B demonstrated predominant upregulation of fatty acid metabolism and oxidative phosphorylation pathways, indicating enhanced mitochondrial energy production.

Pattern analysis of protein expression profiles revealed three major clusters (Fig. 4h):

C5 (shared in both groups): Characterized by wound healing, inflammatory response, and tissue mineralization proteins.

C4 (Group B-specific): Dominated by RNA metabolism and translation-related factors.

C2 (Group A-specific): Characterized by markers of impaired wound-healing resolution.

Gene Ontology analysis further corroborated these findings, with mitochondrial and ribosomal pathways strongly associated with Group B (vs. Group A), highlighting fundamental differences in energy metabolism and protein synthesis capabilities between the groups (Fig. 4i).

These results collectively suggest that the pathogenesis associated with Group A is driven by chronic inflammation and failed tissue repair mechanisms. Group B exhibits disturbances in fundamental cellular processes, including RNA processing and energy metabolism.

3.6. Identification of key molecular pathways in TOLF subtypes

To elucidate the molecular networks underlying the distinct TOLF subtypes, we conducted a comprehensive weighted gene coexpression network analysis (WGCNA) of the proteomic data (Fig. 5a). The results revealed that the MEred and MEbrown modules were highly significantly correlated with Group A, whereas the MEturquoise and MEblue modules were significantly correlated with Group B. Through overrepresentation analysis (ORA) of genes that were differentially expressed (|log2FC| ≥ 1, p < 0.05) of these modules, we identified distinct signalling pathway activation patterns (Fig. 5b and c):

Fig. 5.

Key pathogenesis and key regulatory genes of the two distinct TOLF subtypes a. WGCNA identified modules significantly associated with Groups A and B (p < 0.05). b. Enrichment network of significantly up- and downregulated WGCNA module genes (GS > 0.2, MM > 0.8) related to Group A based on ORA. c. Enrichment network of significantly up- and downregulated WGCNA module genes (GS > 0.2, MM > 0.8) related to Group B based on ORA. d. KDA revealed significantly up- and downregulated driver genes in modules significantly associated with Group A. e. KDA revealed significantly up- and downregulated driver genes in modules significantly associated with Group B. f. MEGENA analysis revealed metabolite modules with similar expression patterns, with only secondary modules selected. g. GeneOverlap analysis identified modules associated with significantly different metabolites in Groups A and B. h. ORA of significantly different (p < 0.05) metabolite modules in Group A. i. ORA of significantly different (p < 0.05) metabolite modules in Group B.

Group A signature pathways:

Upregulated: TRAIL signalling, the TLR10 cascade, and the Wnt/β-catenin pathway.

Downregulated: NOTCH2 signalling.

Group B signature pathways:

Upregulated: KIT receptor signalling, the TCA cycle, and oxidative phosphorylation.

PPI interaction analysis revealed that the driving genes in Group A were associated mainly with angiogenesis, inflammation, tissue repair, and osteogenesis, whereas those in Group B were more strongly related to cellular stress, lipid metabolism, and the cell cycle (Fig. 5d and e).

The molecular landscape was further refined through MEGENA network analysis (Fig. 5f) and GeneOverlap testing of secondary nodes with differentially expressed metabolites (Fig. 5g). Metabolomic pathway enrichment revealed fundamentally distinct metabolic programs: Group A was dominated by taurine/hypotaurine metabolism, and Group B was characterized by phenylalanine/tyrosine/tryptophan biosynthesis (Fig. 5h and i).

4. Discussion

In this study, through integrative analysis of clinical and molecular data, we identified and characterized two distinct subtypes of TOLF, establishing a novel classification system that reflects their unique pathophysiological mechanisms (Fig. 6).

Fig. 6.

Summary of characteristics of the two distinct TOLF subtypes. The major epidemiology, imaging, clinical and molecular features of the two distinct TOLF subtypes are summarized.

The degenerative type of TOLF accounts for 75 % of TOLF, typically affecting the upper and lower thoracic spine (≤3 levels), with an onset age of approximately 60 years and a patients with normal BMI. This type is characterized by upregulated wound-healing, ossification, and inflammation signalling pathways, which may be related to mechanical stress and ageing. In contrast, the metabotropic type of TOLF accounts for 25 % of TOLF, is often multilevel, occurs in younger individuals with abdominal obesity, and is associated with metabolic syndrome. This type is associated with upregulated energy metabolism, lipid metabolism, and inflammation pathways, which may be linked to systemic metabolic factors. Our results suggest that a single pathological mechanism is insufficient to fully account for the complexity of TOLF. To the best of our knowledge, this is the first article to present a new subtype of TOLF from multiple perspectives, offering a critical theoretical foundation for understanding its aetiology and paving the way for the development of precision treatments in the future.

4.1. Epidemiology

TOLF primarily affects one or two levels of the spine, with predominant involvement of the lower thoracic region (T8–T12), a finding supported by numerous previous studies. For example, an article on the epidemiology of ossification of spinal ligaments from Japan revealed that TOLF most commonly occurred at T11, followed by T10 and T3 [13]. Another review revealed that TOLF was most frequent in the lower thoracic spine, primarily affecting the T10/11 and T11/12 levels. Guo et al. reported similar results in a study involving Chinese volunteers [14]. These findings align with our own epidemiological data, which revealed that among 699 patients with TOLF, 73.00 % had lesions in the lower thoracic spine (T8–T12), 13.64 % had lesions in the middle thoracic spine (T4–T8), and 13.36 % had lesions in the upper thoracic spine (T1–T4). This pattern may be attributed to the thoracic–lumbar region being a mobile transition zone, with the presence of floating ribs, making it more prone to degeneration because of significant tensile forces acting on the posterior column [15]. The axial mechanical overload and the subsequent increase in repetitive tensile strain on the ligamentous tissues are key contributors to the development of TOLF. In contrast, the middle thoracic region is protected by the ribs and is more stable, and it is difficult for the mechanical stress factor to explain the occurrence of TOLF in this area. Our results revealed that almost all cases of TOLF in the middle thoracic region were multilevel, suggesting that the aetiology of this type of TOLF may differ from that in other regions.

4.2. Image features

To investigate whether mechanical stress contributes to TOLF development, we analysed sagittal alignment across different TOLF distribution patterns and identified the key factors by correlating TOLF occupation rates with sagittal alignment. The most influential factor was used to construct a finite element model. The results revealed significant differences in sagittal alignment in upper and lower TOLF compared with that in the control group but no significant differences in middle and multilevel TOLF. LL and the C2–7 SVA were the most strongly correlated factors. When LL decreases or the C2–7 SVA increases, mechanical stress at the junction increases, akin to a lever effect, to restore neutral sagittal balance. Two finite element models confirmed these findings, showing increased mechanical stress on the ligamentum flavum in the upper and lower thoracic regions. Only three previous studies have investigated changes in sagittal alignment in patients with TOLF. Liang et al. reported that TOLF patients exhibited significantly greater thoracic kyphosis (TK), lumbar lordosis (LL), and sacral slope (SS), whereas Chang et al. reported that TOLF patients had significantly lower SS and LL values [16,17]. Choi et al. reported that patients with TOLF had a lower PT, and the majority of the patients had a false type 2 spine, which evolved pathologically from the Roussouly type 3 [18]. However, their small sample sizes, lack of segment classification, and absence of finite element models limit the robustness of their conclusions.

4.3. Clinical features

Numerous studies have identified TOLF as a degenerative condition, as its incidence is greater than 20 % in the Asian population over 65 years old. In a CT-based cross-sectional study, Kim et al. reported that the mean age of patients with TOLF was 60.9 years [19], with the prevalence increasing significantly with age. Similarly, Kang et al. investigated the clinical features of TOLF in a Korean population and reported that the average age of onset was 60.9 years [7]. The results of Fujimori et al. supported these findings in Japanese patients, who reported that TOLF typically occurs at approximately 59 years of age [13]. However, several studies have come to different conclusions. Kudo S et al. reported that the rate of TOLF decreased among women over 50 years of age [20]. Liang et al. reported that the mean age of TOLF was 49 years on the basis of a cross-sectional study of 2000 consecutive individuals [16].

In addition to age, BMI has been identified as a significant characteristic associated with TOLF. Multiple studies have established a strong correlation between obesity and TOLF, from being an independent risk factor to influencing its development and severity. Chang et al. demonstrated that TOLF patients consistently exhibited higher BMIs than nonaffected individuals did and that even asymptomatic subjects with TOLF had higher BMIs than healthy controls did [17]. Zhao et al. also reported a positive correlation between BMI and TOLF score [3,21]. Moreover, Tang et al. reported that newly onset TOLF was usually observed in patients with a high BMI; notably, a reduced BMI was observed in patients whose TOLF disappeared within the 5-year follow-up period [22]. Some studies have explored the underlying mechanism of obesity in the pathogenesis of TOLF. Li et al. reported that hyperinsulinaemia could stimulate the BMP-2-induced osteogenic differentiation of ligament cells via the PI3K/Akt and ERK pathways [4]. Fan et al. demonstrated that leptin could promote the osteogenesis of TOLF via STAT3, Runx2 and SRC-1 [8]. In addition, adipokines accelerate the osteogenic process [23].

In this study, we comprehensively investigated the clinical characteristics of TOLF. Our findings revealed that the mean age of onset in the TOLF cohort was approximately 55 years, with patients exhibiting slightly elevated BMI values, which is consistent with previous reports. Through detailed subgroup analysis, we identified distinct clinical profiles between the two types of TOLF. Patients with the metabotropic type of TOLF presented with a younger age, more pronounced abdominal obesity, a higher incidence of metabolic syndrome, elevated systemic inflammatory markers, and increased triglyceride levels. In contrast, the majority of degenerative-type TOLF patients were characterized by a normal BMI range and appeared to present an age-related degenerative condition. These findings suggest that these two types of TOLF may represent two distinct disease entities with different clinical phenotypes and potentially diverse underlying pathophysiological mechanisms. This may explain the inconsistent conclusions drawn in previous studies regarding the pathogenesis and clinical characteristics of TOLF. On the basis of these observations, we recommend that future research should differentiate between these two clinical presentations and focus on elucidating their distinct mechanistic pathways through targeted investigations.

4.4. Molecular features

In our earlier investigations, we discerned variations in osteogenic differentiation potential and epigenomic landscapes between single-level and multilevel TOLF [11,12]. The present study further revealed distinct clinical phenotypes between different types of TOLF, potentially attributable to divergent pathogenic mechanisms and genetic underpinnings. This difference may account for the heterogeneous findings in previous pathogenesis studies. Different disease entities clearly manifest in disparate research outcomes concerning their pathogenesis. Factors such as mechanical stress, genetic predisposition, endocrine and metabolic influences, inflammatory mediators, and miRNAs have been implicated in TOLF development [2,[24], [25], [26], [27], [28]]. Our proteomic and metabolomic analyses revealed significant differences between the two conditions. At the molecular level, the degenerative type of TOLF exhibited perturbations in inflammatory pathways and impaired wound-healing processes, whereas the metabotropic type of TOLF was characterized by alterations in RNA metabolism, protein translation, and methylation pathways.

On the basis of pathophysiological distinction, it is easy to provide a plausible explanation for the observed differences in clinical manifestations between the two types of TOLF: The localized ossification process in degenerative TOLF is initiated by altered stress distribution, which leads to microstructural damage in the ligamentum flavum (sagittal parameter changes and a degenerate state). This mechanical stress triggers the activation of wound-healing and inflammatory pathways, subsequently inducing localized ossification. In contrast, metabotropic TOLF develops through a distinct mechanism involving systemic metabolic disturbances (abdominal obesity combined with metabolic syndrome). These disturbances activate lipid and energy metabolism pathways, promoting the differentiation of fibroblasts into osteoblasts, which drives the ossification process. In metabotropic TOLF, the systemic nature of metabolic alterations leads to widespread ossification affecting multiple segments of the ligamentum flavum. Conversely, degenerative TOLF represents a secondary response to localized mechanical stress, resulting in ossification specifically at sites of concentrated stress.

Several limitations should be considered when interpreting our results. First, the molecular analysis was conducted using data from 8 TOLF patients and 3 controls, which is a relatively small sample size and limited further analysis. The collection of a large patient cohort is a major limitation in this field because of the rarity of the disease [29,30]. Further studies with larger sample sizes could be considered, especially those focusing on the metabolites of the ligament tissue and protein validation. Second, all molecular analyses in this study were conducted using bioinformatics methods; therefore, further cell and animal experiments are necessary to validate our findings. Third, the study design was retrospective and single centre, which may introduce some biases. A multicentre, prospective study would be beneficial. Fourth, the study was only conducted in Asian groups and was based on a single-centre cohort; as such, it may have limitations in terms of generalizability to other populations. While TOLF is indeed rarer in non-Asian groups, there are documented cases, suggesting that the disease is not entirely absent in these populations. Previous studies have reported the highest incidence of TOLF in China and Japan, whereas the incidence is relatively low in Europe and America, which may be due to dietary and sleeping habits, geographical regions and genetic differences between races and ethnicities. Therefore, future studies should aim to include larger, more heterogeneous cohorts, including both Asian and non-Asian populations, to assess whether the classification holds true across different groups and to determine whether ethnic-specific factors need to be considered.

Despite these limitations, this study presents the first proposed classification of TOLF. Our findings suggest that a single pathological mechanism is insufficient to explain TOLF complexity, offering new insights into its pathogenesis and identifying promising therapeutic targets.

5. Conclusion

In summary, we systematically identified two distinct subtypes of TOLF and characterized their clinical features and molecular signatures. It is anticipated that the current “one-size-fits-all” research strategy will shift towards a subtype-based approach, making TOLF research more rational and conducive to future advancements.

Declaration of Generative AI in scientific writing

I am writing to confirm that no generative AI or AI-assisted technologies were used in the preparation of our manuscript.

Author contribution

Longjie Wang: Data curation; Writing - original draft; Shuo Tian: Data curation; Jiangliang Lin: Methodology; Yinhao Liu: Investigation; Zhongqiang Chen: Supervision; Chuiguo Sun: Investigation; Weishi Li: Conceptualization.

Funding

This work was supported by AO Spine Asia Pacific National Research Grant 2024 (AOSRG2024010), the Postdoctoral Fellowship Program of CPSF (GZC20240078) and National Natural Science Foundation of China (82172480).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article.