A Cross‐Tissue Transcriptome‐Wide Association Study Identified Susceptibility Genes for Intervertebral Disc Degeneration

ABSTRACT

Background

Intervertebral disc degeneration (IDD) is a prevalent spinal condition frequently associated with pain and motor impairment, imposing a substantial burden on quality of life. Despite extensive investigations into the genetic predisposition to IDD, the precise pathogenic genes and molecular pathways involved remain inadequately characterized, underscoring the need for continued research to clarify its genetic underpinnings.

Methods

This study leveraged IDD data from the FinnGen R12 cohort and integrated expression quantitative trait loci data across 49 tissues from the Genotype‐Tissue Expression version 8 database to perform a cross‐tissue transcriptome‐wide association study (TWAS). The analytical framework incorporated functional summary‐based imputation (FUSION), unified test for molecular signatures (UTMOST), and gene‐level analysis via multi‐marker genome annotation (MAGMA). To substantiate the findings, Mendelian randomization (MR) and colocalization analyses were subsequently conducted.

Results

Through TWAS and MAGMA analyses, 33 susceptibility genes associated with IDD were identified. Subsequent MR and colocalization analyses refined this list to six candidate genes—ADD1, GFPT1, MAPRE3, MSANTD1, SLC30A6, and XBP1—which may contribute to the initiation and progression of IDD by modulating pathways implicated in the endoplasmic reticulum stress response.

Conclusion

Six susceptibility genes associated with the risk of IDD were identified in this study, offering novel insights into the genetic architecture and potential pathogenic pathways underpinning the development of IDD.

This study conducted a cross‐tissue transcriptome‐wide association study (TWAS) integrating data from FinnGen R12 and GTEx v8 to identify susceptibility genes for intervertebral disc degeneration (IDD). Through TWAS, MAGMA analyses, Mendelian randomization, and colocalization analyses, six candidate genes (ADD1, GFPT1, MAPRE3, MSANTD1, SLC30A6, and XBP1) were identified, which may influence IDD onset and progression by regulating pathways related to endoplasmic reticulum stress response. The findings offer new insights into the genetic mechanisms underlying IDD.

1. Introduction

Intervertebral disc degeneration (IDD) represents a primary contributor to chronic low back pain, a condition that imposes a significant burden on global healthcare systems and negatively impacts patients' quality of life [1]. IDD is characterized by the progressive deterioration of the intervertebral discs, leading to reduced mobility, pain, and, in severe cases, disability [2]. Epidemiological studies indicate that IDD affects a substantial proportion of the global population, with prevalence rates increasing markedly with age [3]. Beyond its clinical impact, IDD imposes a considerable economic burden on healthcare systems, highlighting the urgent need for a deeper understanding of its etiology and the development of effective therapeutic strategies [4].

Emerging evidence suggests that genetic factors play a critical role in the development and progression of IDD [5, 6]. Genome‐wide association studies (GWAS) have identified numerous genetic variants associated with IDD risk, highlighting the complex interplay between genetic predisposition and environmental factors [7, 8, 9, 10]. However, the functional implications of these variants remain largely unexplored. Translating these genetic associations into actionable biological insights is essential for deepening our understanding of IDD pathogenesis and facilitating the development of targeted therapeutic interventions.

Transcriptome‐wide association studies (TWAS) offer a powerful approach to bridge the gap between genetic associations and functional mechanisms [11]. By integrating GWAS data with expression quantitative trait loci (eQTL) datasets, TWAS can identify genes whose expression levels are associated with disease risk. The Unified Test for Molecular Signatures (UTMOST) further enhances this approach by leveraging cross‐tissue eQTL data to improve the detection of gene‐trait associations, which has been widely used in identifying susceptibility genes for multiple diseases, such as migraine, sleep apnea, and age‐related macular degeneration [12, 13, 14, 15].

In this study, we performed a cross‐tissue TWAS analysis by integrating IDD GWAS data from the FinnGen R12 cohort with eQTL data from the Genotype‐Tissue Expression Project (GTEx) v8. To assess tissue‐specific associations, we employed functional summary‐based imputation (FUSION) and validated our findings using multi‐marker analysis of genomic annotation (MAGMA). Candidate genes identified through these analyses were further investigated using Mendelian randomization (MR) and colocalization methods to establish causal relationships and shared genetic mechanisms. Subsequent bioinformatics analyses were conducted to explore the biological properties of these genes, offering new insights to the molecular basis of IDD.

2. Methods

2.1. Data Source

The study design was outlined in Figure 1. GWAS summary statistics for IDD were obtained from the FinnGen R12 cohort, comprising 51,683 cases and 353,224 controls of European descent [16]. EQTL data were sourced from the GTEx project version 8, including gene expression profiles across 49 distinct tissue types [17].

FIGURE 1.

The flowchart of this study. GTEx, genotype‐tissues expression project; GWAS, genome‐wide association; FUSION, functional summary‐based imputation; MAGMA, multi‐marker Analysis of GenoMic Annotation; TWAS, transcriptome‐wide association studies; UTMOST, unified test for molecular signatures.

2.2. TWAS Analyses in Cross‐Tissue and Single Tissue

To elucidate the genetic basis of IDD, a two‐stage TWAS framework was implemented, incorporating both cross‐tissue and single‐tissue analytical strategies. In the discovery phase, cross‐tissue association analysis was performed using the UTMOST algorithm, which integrated multi‐tissue eQTL data from the GTEx version 8 dataset spanning 49 tissues. This method improves imputation precision by accounting for tissue‐specific genetic regulation of gene expression and aggregates gene–trait associations through the generalized Berk–Jones (GBJ) test [18]. Statistical significance was defined by a false discovery rate (FDR) threshold of < 0.05, facilitating the identification of genes with systemic tissue‐level relevance to IDD.

For validation, single‐tissue TWAS was conducted using the FUSION framework. Gene expression weights were estimated by integrating multiple predictive models—including BLUP, BSLMM, LASSO, Elastic Net, and Top1—with the optimal model selected for each gene based on predictive performance. These weights were subsequently combined with GWAS summary statistics for IDD to assess gene‐trait associations. Statistical significance was determined using a FDR threshold of < 0.05. To ensure the robustness of gene identification, candidate genes were required to meet significance criteria in both cross‐tissue and single‐tissue analyses.

2.3. Conditional and Joint Analysis

To further delineate the genetic architecture underlying the TWAS‐identified associations, conditional and joint association analyses (COJO) were undertaken, enabling the resolution of independent association signals within genomic loci implicated in IDD [19, 20]. Utilizing GWAS summary statistics in conjunction with linkage disequilibrium (LD) patterns derived from the 1000 Genomes Project, we conditioned on lead variants within each locus to assess the persistence of secondary signals after accounting for the primary association. Joint modeling was subsequently employed to evaluate the aggregate effects of multiple variants, thereby identifying those exerting independent contributions to IDD susceptibility. Variants that remained statistically significant in both conditional and joint analyses were prioritized as putative causal candidates, offering a refined subset of targets for subsequent functional interrogation.

2.4. MAGMA Analysis

To further characterize the genetic landscape of IDD, gene‐based analyses were performed using MAGMA software (version 1.08). Leveraging GWAS summary statistics, SNP‐level associations were aggregated within defined gene boundaries to derive gene‐level p‐values [21]. This method accounts for LD and gene length, thereby enhancing the reliability of gene prioritization [22]. For comprehensive details regarding parameter settings and methodological specifications, readers are referred to the official MAGMA documentation [23].

2.5. Mendelian Randomization and Bayesian Colocalization

To investigate potential causal relationships and shared genetic architecture underlying IDD, we conducted MR and Bayesian colocalization analyses. MR was implemented using the TwoSampleMR package [24], with cis‐eQTL SNPs serving as instrumental variables, gene expression levels as the exposure, and IDD GWAS summary statistics as the outcome. Independent SNPs (r² < 0.001) were identified through LD clumping, and causal estimates were derived using the Wald ratio method, with statistical significance defined as p < 0.05. For Bayesian colocalization, the coloc R package was employed to assess whether IDD‐associated GWAS signals and eQTL signals shared common causal variants [25]. A posterior probability greater than 0.75 was interpreted as strong evidence supporting a shared causal variant [26].

2.6. GeneMANIA Analysis

To investigate the functional interrelationships among the identified genes, we conducted GeneMANIA analysis, integrating data on protein–protein interactions, co‐expression patterns, and shared biological pathways. A gene interaction network was constructed using the GeneMANIA web platform (https://genemania.org/) [27, 28].

3. Results

3.1. TWAS Analysis Results

In the cross‐tissue TWAS analysis, 439 genes demonstrated significance (p < 0.05) (Table S1), of which 111 remained statistically significant following FDR correction (Table 1). In the single‐tissue TWAS analysis, 1679 genes reached an FDR threshold of < 0.05 in at least one tissue type (Table S2). Notably, 62 candidate genes exhibited consistent statistical significance across both cross‐tissue and single‐tissue analyses (Table S3).

TABLE 1.

The significant genes for intervertebral disc degeneration risk in cross‐tissue UTMOST analysis.

Gene symbol	CHR	Ensemeble ID	Location (hg38)	Test score	p_value	FDR
TGFA	2	ENSG00000163235	70 447 284–70 554 193	28.93	6.00853E‐13	2.2532E‐09
NAT8L	4	ENSG00000185818	2 059 327–2 069 089	26.96	5.38281E‐12	8.15473E‐09
TACC3	4	ENSG00000013810	1 712 858–1 745 171	24.54	8.96616E‐12	8.40578E‐09
FGFR3	4	ENSG00000068078	1 793 293–1 808 872	543.24	2.50849E‐11	1.57273E‐08
PCBP1	2	ENSG00000169564	70 087 477–70 089 203	482.35	2.51638E‐11	1.57273E‐08
RNF212	4	ENSG00000178222	1 056 250–1 113 564	459.96	3.16966E‐11	1.69803E‐08
SLBP	4	ENSG00000163950	1 692 731–1 712 344	23.72	4.52389E‐11	2.12057E‐08
HAUS3	4	ENSG00000214367	2 227 464–2 242 133	23.11	8.13544E‐11	3.05079E‐08
NFU1	2	ENSG00000169599	69 396 113–69 437 628	19.9	6.37671E‐10	2.17388E‐07
DCC	18	ENSG00000187323	52 340 197–53 535 899	20.67	7.84567E‐10	2.45177E‐07
MFSD10	4	ENSG00000109736	2 930 561–2 934 834	18.11	8.27208E‐09	2.06802E‐06
GFPT1	2	ENSG00000198380	69 319 780–69 387 250	15.42	5.52848E‐08	1.29574E‐05
MXD1	2	ENSG00000059728	69 897 688–69 942 945	16.02	7.77228E‐08	1.71447E‐05
MSANTD1	4	ENSG00000188981	3 244 369–3 271 738	14.93	9.79815E‐08	2.04128E‐05
FAM53A	4	ENSG00000174137	1 617 915–1 684 313	14.87	1.03737E‐07	2.04744E‐05
PAIP2B	2	ENSG00000124374	71 182 738–71 227 103	159.65	1.59852E‐07	2.99722E‐05
KHK	2	ENSG00000138030	27 086 772–27 100 762	14.38	1.75303E‐07	3.13042E‐05
AGBL5‐IT1	2	ENSG00000229122	27 061 038–27 061 815	14.35	1.88737E‐07	3.21711E‐05
HTT	4	ENSG00000197386	3 041 363–3 243 957	14.8	2.91228E‐07	4.74828E‐05
MAPRE3	2	ENSG00000084764	26 970 637–27 027 219	14.13	3.07826E‐07	4.80978E‐05
CGREF1	2	ENSG00000138028	27 098 889–27 119 128	13.16	6.8883E‐07	0.000103324
EMILIN1	2	ENSG00000138080	27 078 615–27 086 403	13.88	1.05294E‐06	0.000143279
NOP14‐AS1	4	ENSG00000249673	2 934 882–2 961 738	12.81	1.06982E‐06	0.000143279
ELOCP21	2	ENSG00000226186	70 955 899–70 956 234	13.23	2.40619E‐06	0.000311146
TEX261	2	ENSG00000144043	70 968 325–70 994 873	11.77	2.54739E‐06	0.000318424
FGFRL1	4	ENSG00000127418	1 009 936–1 026 898	12.52	3.68453E‐06	0.00044571
PREB	2	ENSG00000138073	27 130 756–27 134 666	181.26	4.92197E‐06	0.000563962
SERHL	22	ENSG00000172250	42 500 671–42 512 237	10.95	5.02517E‐06	0.000563962
PCBP1‐AS1	2	ENSG00000179818	69 960 104–70 103 220	11.22	5.11325E‐06	0.000563962
NRP2	2	ENSG00000118257	205 681 990–205 798 133	12.07	6.22072E‐06	0.000666505
MXD4	4	ENSG00000123933	2 247 432–2 262 109	11.31	7.96421E‐06	0.000829605
SLC5A6	2	ENSG00000138074	27 199 587–27 212 958	10.61	9.95988E‐06	0.001009447
SNRPG	2	ENSG00000143977	70 281 362–70 293 740	11.32	1.09413E‐05	0.001079731
TMEM175	4	ENSG00000127419	932 387–958 656	11.08	1.15261E‐05	0.001108275
LETM1	4	ENSG00000168924	1 811 479–1 856 156	10.68	1.77056E‐05	0.001659899
BCS1L	2	ENSG00000074582	218 658 764–218 663 443	10.34	2.32957E‐05	0.002130701
STK36	2	ENSG00000163482	218 672 069–218 702 716	9.96	2.41228E‐05	0.00215382
TMEM129	4	ENSG00000168936	1 715 952–1 721 358	9.16	3.03823E‐05	0.002649617
PLCD4	2	ENSG00000115556	218 607 855–218 637 184	10.18	3.29476E‐05	0.002745631
UCN	2	ENSG00000163794	27 307 400–27 308 445	10.25	3.28754E‐05	0.002745631
RAB1A	2	ENSG00000138069	65 070 696–65 130 331	10.29	3.48695E‐05	0.002842624
BCL11A	2	ENSG00000119866	60 450 520–60 554 467	10.35	4.25559E‐05	0.003324682
CDK5R2	2	ENSG00000171450	218 959 666–218 962 155	9.73	4.51997E‐05	0.003459163
CENPA	2	ENSG00000115163	26 764 289–26 801 067	9.75	4.80478E‐05	0.003603584
RN7SKP80	22	ENSG00000202058	42 565 048–42 565 330	9	5.50028E‐05	0.003966545
GRK4	4	ENSG00000125388	2 963 571–3 040 760	8.35	7.08686E‐05	0.00492143
ANTXR1	2	ENSG00000169604	69 013 176–69 249 327	9.77	7.07364E‐05	0.00492143
ME2	18	ENSG00000082212	50 879 080–50 954 257	9.07	9.80814E‐05	0.006687367
ADD1	4	ENSG00000087274	2 843 844–2 930 076	9.22	0.000107261	0.006934998
EXOC6B	2	ENSG00000144036	72 175 984–72 826 041	8.52	0.000104881	0.006934998
OPRPN	4	ENSG00000171199	70 397 931–70 410 195	8.8	0.000107142	0.006934998
FAM13A	4	ENSG00000138640	88 725 955–89 111 398	8.74	0.000109158	0.006938014
LINC01816	2	ENSG00000231327	70 124 034–70 132 923	181.16	0.000118688	0.007417972
CRYBB1	22	ENSG00000100122	26 599 278–26 618 027	9.27	0.000128108	0.007875466
PDK1	2	ENSG00000152256	172 555 373–172 608 669	8.14	0.000135048	0.0081682
ADRA2C	4	ENSG00000184160	3 766 348–3 768 526	8.76	0.000150377	0.008862368
MCEE	2	ENSG00000124370	71 109 684–71 130 239	8.01	0.000155277	0.008958291
CMKLR2	2	ENSG00000183671	206 175 316–206 218 047	8.77	0.000173143	0.009837662
GALNT14	2	ENSG00000158089	30 910 467–31 155 202	8.29	0.000178762	0.009858195
CFAP65	2	ENSG00000181378	219 002 846–219 041 527	8.77	0.000176937	0.009858195
GALNT7	4	ENSG00000109586	173 168 811–173 323 967	8.87	0.000194542	0.01057296
CNOT9	2	ENSG00000144580	218 568 580–218 597 080	8.46	0.000210869	0.01129653
ZNF142	2	ENSG00000115568	218 633 329–218 659 655	7.16	0.000223008	0.011778599
C22orf31	22	ENSG00000100249	29 058 672–29 061 831	7.61	0.000250524	0.013048134
RAPGEF4‐AS1	2	ENSG00000228016	172 677 141–172 736 206	8.27	0.000259358	0.013323182
SLC30A6	2	ENSG00000152683	32 165 841–32 224 379	8.07	0.000273619	0.013680975
OTOP1	4	ENSG00000163982	4 188 726–4 226 929	8.44	0.000273615	0.013680975
OTOF	2	ENSG00000115155	26 457 203–26 558 756	7.71	0.000395861	0.019278935
HADHA	2	ENSG00000084754	26 190 635–26 244 672	7.47	0.000425553	0.019947802
PCGF3	4	ENSG00000185619	705 748–770 089	6.45	0.000422516	0.019947802
RRP7A	22	ENSG00000189306	42 508 344–42 519 796	6.87	0.000418886	0.019947802
SH3BP2	4	ENSG00000087266	2 793 023–2 841 291	7.31	0.000437736	0.020265552
ZNF638	2	ENSG00000075292	71 276 561–71 435 069	6.92	0.000500631	0.021543172
CCDC117	22	ENSG00000159873	28 772 674–28 789 301	7.73	0.000498982	0.021543172
DNAJC5G	2	ENSG00000163793	27 275 433–27 281 499	7.57	0.000500039	0.021543172
ZDBF2	2	ENSG00000204186	206 274 663–206 314 427	6.66	0.000505546	0.021543172
ZNF407	18	ENSG00000215421	74 597 870–75 065 671	7.69	0.000490452	0.021543172
NICOL1	4	ENSG00000243449	2 041 995–2 043 964	7.81	0.000484826	0.021543172
RMC1	18	ENSG00000141452	23 503 496–23 531 822	6.73	0.000555382	0.023400932
AAMP	2	ENSG00000127837	218 264 129–218 270 178	6.95	0.000568744	0.023697646
ZFYVE28	4	ENSG00000159733	2 269 582–2 418 651	6.79	0.000627578	0.025861742
EWSR1	22	ENSG00000182944	29 268 009–29 300 525	8.17	0.000634526	0.025863852
TTC27	2	ENSG00000018699	32 628 032–32 821 051	6.56	0.000703841	0.028380699
CTDSP1	2	ENSG00000144579	218 397 136–218 405 941	6.99	0.000720623	0.028748262
TRIM60	4	ENSG00000176979	165 016 458–165 041 749	3.37	0.000743414	0.029345289
AAK1	2	ENSG00000115977	69 457 997–69 674 349	7.02	0.000806744	0.031513448
ARK2C	18	ENSG00000141622	46 326 809–46 463 140	7.01	0.000854202	0.033023281
XDH	2	ENSG00000158125	31 334 321–31 414 742	7.03	0.000873565	0.033427217
SPATA31H1	2	ENSG00000221843	27 537 386–27 582 722	7.16	0.000898199	0.034022707
GLI2	2	ENSG00000074047	120 735 623–120 992 653	6.66	0.000917153	0.034393245
MAPK4	18	ENSG00000141639	50 560 087–50 731 826	6.61	0.000927926	0.034452699
BLOC1S4	4	ENSG00000186222	6 716 174–6 717 664	6.93	0.000983609	0.036162108
SLC49A3	4	ENSG00000169026	681 829–689 271	5.99	0.001011793	0.036837133
DPY30	2	ENSG00000162961	31 867 809–32 039 805	6.17	0.001036298	0.037366526
HGFAC	4	ENSG00000109758	3 441 968–3 449 486	5.95	0.001051694	0.037560496
ARHGAP25	2	ENSG00000163219	68 679 601–68 826 833	6.34	0.001098225	0.038381145
GAS2L1	22	ENSG00000185340	29 306 618–29 312 787	5.9	0.001105938	0.038381145
ADH4	4	ENSG00000198099	99 123 657–99 157 792	6.1	0.001102865	0.038381145
KREMEN1	22	ENSG00000183762	29 073 035–29 168 333	6.45	0.00113566	0.038715668
XBP1	22	ENSG00000100219	28 794 555–28 800 597	6.33	0.001169824	0.039521064
RPL6P5	2	ENSG00000214064	145 337 230–145 338 057	3.24	0.001193711	0.039968018
OCIAD2	4	ENSG00000145247	48 885 019–48 906 937	6.7	0.001208153	0.040093572
RGPD4	2	ENSG00000196862	107 826 892–107 892 544	6.66	0.001246964	0.041018561
YIPF4	2	ENSG00000119820	32 277 904–32 316 594	6.21	0.001295261	0.04187267
TRAPPC12	2	ENSG00000171853	3 379 675–3 485 094	5.5	0.001292905	0.04187267
SMIM21	18	ENSG00000206026	75 409 476–75 427 703	3.21	0.001333275	0.042733182
METAP1D	2	ENSG00000172878	171 999 943–172 082 430	5.35	0.001474017	0.046843772
EEF1A1P9	4	ENSG00000249264	105 484 698–105 486 080	6.83	0.001489833	0.046948529
SMCHD1	18	ENSG00000101596	2 655 726–2 805 017	6.29	0.001565918	0.048934945
CYP27A1	2	ENSG00000135929	218 781 749–218 815 293	6.21	0.001606788	0.049388982
MEIS1‐AS3	2	ENSG00000226819	66 426 735–66 433 470	6.76	0.001602398	0.049388982

3.2. COJO Analysis

The 62 candidate genes were predominantly localized to chromosomes 2, 4, 18, and 22. To mitigate the potential confounding effects of LD and reduce false‐positive associations, COJO analysis was conducted within the corresponding tissue contexts (Table S4). In Adipose_Subcutaneous, conditioning on the predicted expression of MAPRE3 markedly attenuated the TWAS signals associated with CENPA and PREB (Figure S1A). In Muscle_Skeletal, adjustment for the predicted expression of RNF212, FGFR3, and MSANTD1 substantially diminished the signals linked to HTT, TMEM129, and FAM53A (Figure S1B). In Testis, conditioning on AAK1 and TGFA reduced the TWAS signals of LINC01816 and NFU1 (Figure S1C). Similarly, in Whole_Blood, conditioning on CGREF1 weakened the associations observed for EMILIN1 and KHK (Figure S1D). Given that CENPA, PREB, HTT, and EMILIN1 achieved significance exclusively in single‐tissue TWAS analyses and were likely influenced by LD, these genes were excluded from subsequent investigations.

3.3. Gene Analysis of MAGMA

MAGMA analysis identified 1249 genes significantly associated with IDD at a FDR of less than 0.05 (Table S5). To enhance the robustness of our findings, we integrated the UTMOST cross‐tissue results with the significant genes detected by both FUSION and MAGMA, culminating in the identification of 33 putative candidate genes (Figure 2).

FIGURE 2.

Venn diagram. MAGMA identified 1,249 significant genes, FUSION identified 1,679, and UTMOST cross‐tissue analysis identified 111, of which 33 were common.

3.4. MR And Colocalization Results

MR analysis identified significant putative causal associations between IDD and 27 genes (Figure S2, Table S6), with subsequent colocalization analyses substantiating robust causal links for six genes: ADD1, GFPT1, MAPRE3, MSANTD1, SLC30A6, and XBP1 (Table S7, Figure 3). ADD1, located at chromosome 4p16.3, demonstrated an odds ratio (OR) of 1.16 (95% CI: 1.08–1.25) with a posterior probability of shared causal variant (PP.H4) of 0.89; the lead colocalized variant was rs58115360. GFPT1, mapped to chromosome 2p13.3, yielded an OR of 1.19 (95% CI: 1.09–1.29) and a PP.H4 of 0.81, with rs77753566 identified as the most strongly colocalized variant. MAPRE3, also on chromosome 2 (2p23.3), exhibited a protective association with an OR of 0.89 (95% CI: 0.84–0.94) and a PP.H4 of 0.83, with rs6705514 as the top variant. MSANTD1, residing at 4p16.3, showed an OR of 1.06 (95% CI: 1.03–1.10) and a PP.H4 of 0.81, with rs10011326 representing the lead signal. SLC30A6, on chromosome 2p22.3, was associated with an OR of 1.16 (95% CI: 1.08–1.25) and a PP.H4 of 0.83, with rs6747488 identified as the top variant. Lastly, XBP1, located on chromosome 22q12.1, was linked to IDD with an OR of 1.10 (95% CI: 1.05–1.16) and a PP.H4 of 0.85; the most significant colocalized variant was rs5762877.

FIGURE 3.

The colocalization results confirmed the causal associations between six candidate genes and intervertebral disc degeneration.

3.5. Validation of Candidate Genes Expression

To validate the expression patterns of six candidate genes, we acquired transcriptional profiles of degenerated and non‐degenerated human nucleus pulposus cells from the Gene Expression Omnibus database (GSE245147) [29]. Differential expression analysis was subsequently performed using DESeq2 [30]. The results demonstrated statistically significant differences in five genes (excluding MSANTD1). Specifically, GFPT1, XBP1, and SLC30A6 exhibited downregulation in degenerated nucleus pulposus cells, whereas MAPRE3 and ADD1 showed upregulation in the degenerative condition (Table S8 and Figure S3).

3.6. GeneMANIA Analysis

The potential gene interaction network, centered on the six genes (ADD1, GFPT1, MAPRE3, MSANTD1, SLC30A6, and XBP1), is illustrated in Figure 4. The results revealed significant enrichment in pathways including the response to unfolded proteins, the response to endoplasmic reticulum stress, UDP‐N‐acetylglucosamine metabolic processes, amino sugar biosynthetic processes, and nucleotide‐sugar biosynthetic processes (Table S9). These findings suggest that the identified genes and their interactions play critical roles in key biological processes related to IDD pathogenesis.

FIGURE 4.

GeneMania gene network.

4. Discussion

We conducted a systematic evaluation of the association between genetic predisposition related to gene expression and IDD, using both IDD GWAS and the GTEx V8 eQTL data. Through cross‐tissue TWAS analysis, single‐tissue TWAS analysis, and MAGMA analysis, supplemented by MR analysis and colocalization analysis of inflammation, we ultimately identified six IDD susceptibility genes (ADD1, GFPT1, MAPRE3, MSANTD1, SLC30A6, and XBP1). These genes may play a pivotal role in the development and progression of IDD by modulating pathways associated with endoplasmic reticulum stress (ERS) response.

ERS refers to the cellular stress response that occurs when the function of the endoplasmic reticulum is impaired or overloaded [31]. Although the pathogenesis of IDD remains incompletely elucidated, research has demonstrated that inflammation, oxidative stress, excessive mechanical load, metabolic disorders, and extracellular matrix dysregulation mediate the occurrence of ERS, thereby promoting the degeneration of the intervertebral disc [32].

Oxidative stress and inflammatory responses are two critical factors in the development of IDD [33]. The accumulation of reactive oxygen species (ROS) and pro‐inflammatory cytokines can induce the degradation of the extracellular matrix and promote intervertebral disc cell apoptosis [34]. On one hand, ROS can interfere with normal protein modification and disulfide bond formation within cells, thereby activating ERS [35]. On the other hand, the crosstalk between ER, mitochondria, and Ca²⁺ homeostasis plays a key role in ROS‐mediated ERS [36]. Oxidative stress may lead to calcium homeostasis imbalance. In cases of severe or prolonged ER dysfunction, Ca²⁺ is released from the ER into the mitochondria, triggering a series of apoptotic signaling pathways [37]. This translocation is regulated by the IP3‐RGRP75‐VDAC1 signaling pathway, and the imbalance in calcium homeostasis contributes to the progression of IDD by initiating apoptosis in nucleus pulposus cells [37]. Studies have found that advanced glycation end products can elevate intracellular calcium levels, deplete ER luminal Ca²⁺ stores, disrupt Ca²⁺ homeostasis, and induce nucleus pulposus cell apoptosis through ERS, thereby contributing to the progression of IDD in rats [38]. Additionally, pro‐inflammatory cytokines such as TNF‐α, IL‐β, and IL‐1 can also activate ERS and regulate intervertebral disc cell apoptosis [39, 40]. In summary, the formation of the intervertebral disc inflammatory microenvironment is closely related to the secretion of downstream inflammatory mediators mediated by ERS, which in turn can feedback and exacerbate the ERS response, ultimately creating a vicious cycle. Therefore, inhibiting or blocking ERS‐related inflammatory outbreaks could be a potential therapeutic strategy for protecting intervertebral disc cell function.

Excessive mechanical load represents a pivotal etiological factor in the pathogenesis of IDD. The mechanosensitive calcium channel Piezo1 exhibits significant upregulation in response to shear or tensile stress [41]. Studies have shown that excessive mechanical stress mediates the activation of Piezo1, which accelerates nucleus pulposus cell senescence and apoptosis while stimulating the expression of ERS markers such as GRP78 and CHOP in cells [42]. Knockdown of Piezo1 inhibited ERS activation and delayed nucleus pulposus cell degeneration [43]. Furthermore, the apoptosis rate of annulus fibrosus cells transfected with CHOP shRNA was markedly diminished in response to cyclic tensile stress [44]. In summary, excessive mechanical stress may regulate intervertebral disc cell activity by mediating ER stress, thereby influencing the progression of IDD. However, other mechanosensitive channels, such as integrins, TRPV4, and N‐cadherin adhesion molecules, are also implicated in the evolution of IDD [45, 46]. Whether these channels are regulated by ER stress and the detailed mechanisms remain unclear.

Metabolic disorders can stimulate the development of IDD. Dysregulated glucose metabolism has been identified as a significant risk factor in the pathogenesis of IDD [47]. Intriguingly, glucose deprivation has been shown to trigger apoptosis in nucleus pulposus cells, a process that can be mitigated through the autophagy‐dependent p‐eIF2α/ATF4 signaling pathway [48]. Furthermore, hypercholesterolemia has emerged as a potential contributory factor in the progression of IDD [49]. In a hypercholesterolemia rat model, cholesterol induces ERS through the mSREBP1 pathway, leading to cell apoptosis and extracellular matrix degradation. Similarly, atorvastatin, a cholesterol‐lowering pharmacological agent, has been shown to attenuate the adverse effects of elevated cholesterol levels on intervertebral disc integrity [50].

In recent years, extensive research efforts have been directed toward the advancement and therapeutic application of pharmacological interventions targeting ERS to improve the intervertebral disc cell microenvironment, and significant progress has been made. Tauroursodeoxycholic acid, an FDA‐approved hydrophilic bile acid, can alleviate mechanical stress‐induced necrotic apoptosis of nucleus pulposus cells by inhibiting ERS [51]. Similarly, berberine, another FDA‐approved compound extracted from traditional Chinese medicines such as Coptis and Phellodendron, is an isoquinoline alkaloid widely used in clinical practice. Research has shown that in an oxidative stress‐mediated nucleus pulposus cell injury model, berberine can maintain intracellular Ca²⁺ homeostasis, inhibit ERS response, and delay nucleus pulposus cell apoptosis [52]. Eicosapentaenoic acid (EPA), an endogenous omega‐3 fatty acid, induces nucleus pulposus cell‐specific autophagy, thereby inhibiting ERS, delaying nucleus pulposus cell apoptosis, and further confirming its protective effect in a mouse acupuncture model [53]. As mentioned above, although numerous studies have substantiated the viability of targeting ERS to improve IDD, the underlying mechanisms have not yet been fully elucidated, and the specificity and efficacy still require further investigation.

XBP1, as a key transcription factor in the ERS response, is involved in the regulation of cellular stress responses and apoptosis [54]. Although there is no direct study on the role of XBP1 in IDD yet, its critical position in ER stress suggests that it may influence the survival of intervertebral disc cells by regulating cellular stress responses and apoptosis pathways [55, 56]. Moreover, the zinc transporter protein encoded by the SLC30A6 gene may play a potential protective role in IDD by regulating intracellular zinc ion levels, thereby enhancing the antioxidant capacity of intervertebral disc cells [57]. As for ADD1, GFPT1, MAPRE3, and MSANTD1, no studies have yet been reported linking them to IDD. Further research is needed to explore the roles these genes play in the pathological mechanisms of IDD.

This study has some limitations. First, the study population is limited to the European population, thereby limiting the extrapolation of our findings to broader demographic groups. Secondly, experiments are needed to further validate the proposed pathophysiological mechanisms of IDD. Moreover, the absence of intervertebral disc‐specific eQTL data in the GTEx database may preclude direct inference of gene expression effects within the disc microenvironment. However, our cross‐tissue framework leverages shared genetic regulation across tissues to identify genes involved in systemic or conserved pathways (e.g., ER stress, metabolic dysregulation) that may indirectly influence intervertebral disc homeostasis. These findings align with prior evidence linking ER stress and oxidative stress to IDD pathogenesis, supporting the biological plausibility of our results and providing new perspectives regarding the potential pathophysiological mechanisms of IDD.

In conclusion, this study, employing a cross‐tissue TWAS, identified six novel genes associated with susceptibility to IDD, thereby offering new perspectives on the molecular mechanisms underpinning its pathogenesis.

Author Contributions

All authors contributed to the study conception and design. W.Z. and L.Z. conceived the study and drafted the manuscript. W.Z. and L.Z. contributed equally to this work. H.Y., T.D., and Y.L. collected and analyzed the data, participated in the data preparation, and provided important comments on the manuscript. T.D. and Y.L. revised and approved the final version of the manuscript. The corresponding authors attest that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Regional association of TWAS hits. (A), (C), and (D) Chromosome 2 regional association plot. (D) Chromosome 4 regional association plot. The top panel highlights all genes in the region. The marginally associated TWAS genes are shown in blue, and the jointly significant genes are shown in green. The bottom panel shows a regional Manhattan plot of GWAS data before (gray) and after (blue) conditioning on the predicted expression of the green genes.

Figure S2: Forest plot of mendelian randomization analysis results.

Figure S3: Validation of candidate genes expression via Gene Expression Omnibus database.

Table S1: The results of TWAS analyses in cross‐tissue.

Table S2:. FUSION identified significant genes associated with intervertebral disc degeneration.

Table S3:. FUSION identified genes that were significant in the cross‐tissue UTMOST analysis.

Table S4:. The results of conditional and joint analysis.

Table S5:. MAGMA gene‐based test identified significant genes associated with intervertebral disc degeneration.

Table S6:. The results of Mendelian randomization analysis.

Table S7:. The results of Bayesian colocalization analysis.

Table S8:. Differential expression analysis of six genes.

Table S9:. The functional pathway analysis of genemania.

Zhang L., Zhao W., Yang H., Deng T., and Li Y., “A Cross‐Tissue Transcriptome‐Wide Association Study Identified Susceptibility Genes for Intervertebral Disc Degeneration,” JOR Spine 8, no. 3 (2025): e70109, 10.1002/jsp2.70109.

Data Availability Statement

The intervertebral disc degeneration GWAS data were obtained from the FinnGen R12 dataset (https://www.finngen.fi/en/access_results). Gene expression and eQTL data are freely available at https://ftp.ebi.ac.uk/pub/databases/spot/eQTL/imported/GTEx_V8. FUSION tool (http://gusevlab.org/projects/fusion/).