Introduction
Cervical most cancers is the fourth commonest malignancy amongst ladies worldwide, accounting for an estimated 604,000 new instances and 342,000 deaths yearly.1 Persistent an infection with high-risk human papillomavirus (HPV), notably HPV16 and HPV18, is the required explanation for cervical carcinogenesis.2 Nevertheless, HPV an infection is frequent but solely a minority develop cervical most cancers or its precursors, strongly suggesting that host genetic elements play a essential function in illness susceptibility.3
Excessive-grade squamous intraepithelial lesion (HSIL) is the rapid precursor to invasive cervical most cancers, histologically comparable to cervical intraepithelial neoplasia grade 2 or 3 (CIN2/3).4 Present cervical most cancers screening methods primarily depend on cytological examination (Pap smear) and high-risk HPV testing. Whereas HPV testing provides excessive sensitivity for detecting CIN2+ (96.9%), its specificity is notably decrease (88.8%), resulting in extreme colposcopy referrals, overtreatment, and diminished affected person compliance.4 Conversely, cytology-based screening demonstrates considerably decrease sensitivity, starting from 58.8% to 73.4% for CIN2+ detection.5 Even high-quality cytology has been reported to overlook no less than 30% of HSIL and cervical most cancers instances.6 These limitations underscore the pressing want for novel biomarkers and a deeper understanding of the molecular mechanisms underlying HSIL pathogenesis.
Present screening tips, nevertheless, stay discordant on the optimum method. The US Preventive Providers Activity Pressure and the Girls’s Preventive Providers Initiative suggest co-testing (mixed HPV and cytology) each 5 years for girls aged 30–65 years.7 In distinction, the American Most cancers Society’s 2020 up to date guideline shifted towards main HPV testing each 5 years as the popular technique.8 Whereas HPV testing provides excessive sensitivity, research have proven that high-risk HPV testing alone would miss roughly 30% of cytologically detected dysplastic samples,9 and co-testing can obtain a sensitivity of as much as 97.5% for detecting cervical lesions.10 These ongoing debates spotlight the unresolved challenges in present screening and reinforce the necessity for novel biomarkers to enhance threat stratification.
Genome-wide affiliation research (GWAS) have recognized a number of susceptibility loci for cervical most cancers, most notably inside the human leukocyte antigen (HLA) area on chromosome 6p21.3.11,12 Nevertheless, GWAS findings alone can not distinguish whether or not related variants affect illness threat by way of results on gene expression, protein ranges, or metabolic pathways. Latest large-scale expression quantitative trait loci (eQTL) research, such because the eQTLGen consortium comprising 31,684 blood samples, have considerably improved our skill to map regulatory variants.13 Equally, complete GWAS of 1,400 circulating metabolites in 8,299 people have characterised the genetic structure of the human metabolome.14 Each gene expression and metabolite ranges have been implicated in cervical carcinogenesis,15,16 but whether or not these associations mirror causal relationships stay unclear resulting from potential confounding and reverse causation inherent in observational research.
Mendelian randomization (MR) makes use of genetic variants as instrumental variables to deduce causal relationships between exposures and illnesses, providing benefits over observational research by being much less inclined to confounding and reverse causation.17 Regardless of these methodological strengths, essential information gaps stay. The causal results of most genes and circulating metabolites on HSIL threat haven’t been systematically evaluated. Furthermore, whether or not circulating metabolites mediate the consequences of genetic variants on HSIL stays unclear, as few research have employed mediation MR frameworks to dissect these relationships.18 Elucidating these mechanisms might inform the event of novel preventive methods for cervical most cancers.
On this research, we carried out a complete two-sample Mendelian randomization evaluation to realize three aims: (1) to establish genes and circulating metabolites which might be causally related to HSIL threat utilizing a unidirectional two-sample MR framework; (2) to analyze whether or not circulating metabolites mediate the consequences of causal genes on HSIL utilizing a two-step mediation MR method; and (3) to discover the potential organic pathways concerned by way of practical enrichment evaluation. Our findings present novel genetic and metabolic insights into the etiology of HSIL and will inform future mechanistic research.
Strategies
Examine Design
This research was carried out in accordance with the Strengthening the Reporting of Observational Research in Epidemiology utilizing Mendelian Randomization (STROBE-MR) tips (Supplementary Table 1). Utilizing a two-sample MR design, we (1) recognized genes and metabolites causally related to HSIL, (2) assessed mediation by way of two-step MR, and (3) explored organic pathways by way of practical enrichment evaluation. The general research design is summarized in Supplementary Figure 1.
 |
Determine 1 Forest plots of causal associations. (A) Genetically predicted gene expression and HSIL threat. (B) Circulating metabolites and HSIL threat. Abbreviations: OR, odds ratio; CI, confidence interval; IVW, inverse-variance weighted; FDR, false discovery price. Notes: Circles symbolize the OR for every genetic instrument, with circle measurement proportional to the precision of the estimate (inverse variance). Horizontal strains point out 95% CIs. Crimson circles point out risk-associated exposures (OR > 1); blue circles point out protecting exposures (OR < 1). The vertical dashed line signifies the null impact (OR = 1.0). All associations proven have been statistically vital after FDR correction (q < 0.05).
|
Knowledge Supply and Individuals
Knowledge assortment for all included GWAS research was accomplished previous to June 3, 2026. Abstract-level information for cis-expression quantitative trait loci (cis-eQTLs) have been obtained from the eQTLGen consortium (31,684 blood samples from 37 European-ancestry cohorts) (https://www.eqtlgen.org/).13 GWAS abstract statistics for 1,400 circulating metabolites have been obtained from Chen et al (8,299 European-ancestry people from the Canadian Longitudinal Examine on Growing old) (https://doi.org/10.1038/s41467-022-31175-4).14 HSIL GWAS abstract statistics have been obtained from FinnGen launch 12 (293,218 Finnish people, together with 8291 HSIL instances) (https://www.finngen.fi/),19 with HSIL outlined utilizing ICD-10 codes D06, N87.1, and N87.2 (comparable to CIN2/3). All unique GWAS research acquired approval from their respective institutional assessment boards, and knowledgeable consent was obtained from all individuals, as detailed within the unique publications.13,14,19 Particularly, the FinnGen research was authorised by the Coordinating Ethics Committee of the Hospital District of Helsinki and Uusimaa (HUS/990/2017).19
Instrumental Variable Choice
For each eQTL and metabolite analyses, instrumental variables (IVs) have been chosen utilizing a constant method. Impartial SNPs have been recognized utilizing a genome-wide significance threshold of P < 1×10−5 (reasonably than the standard P < 5×10−8) to make sure adequate statistical energy for the two-step mediation evaluation, which requires an ample variety of instrumental variables. This threshold has been broadly utilized in eQTL-based MR research. All IVs had F-statistics > 10, confirming the absence of weak instrument bias.
Mendelian Randomization Evaluation
Two-sample MR was carried out utilizing the inverse-variance weighted (IVW) technique as the first evaluation. Sensitivity analyses included weighted median, MR-Egger, Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO), and leave-one-out strategies. Heterogeneity was assessed utilizing Cochran’s Q statistic; vital heterogeneity (P < 0.05) prompted the usage of random-effects IVW. Horizontal pleiotropy was evaluated utilizing the MR-Egger intercept take a look at (P > 0.05 indicating no pleiotropy).
Colocalization Evaluation
Colocalization evaluation was carried out for genes exhibiting vital MR associations utilizing the coloc.abf technique.20 For every gene, we extracted abstract statistics for all SNPs inside ±500 kb of the gene heart from each eQTL and HSIL GWAS. The posterior chance of speculation 4 (PP.H4), indicating that each traits share a causal variant, was calculated. PP.H4 > 0.8 was thought-about sturdy proof of colocalization.
Practical Enrichment Evaluation
To discover organic pathways, we carried out GO and KEGG enrichment evaluation utilizing the clusterProfiler R bundle. Genes considerably correlated with COL11A2 expression (|r| > 0.4, P < 0.05) within the TCGA cervical most cancers dataset have been recognized. Enrichment was thought-about vital at adjusted P < 0.05 utilizing the Benjamini-Hochberg technique.
Statistical Evaluation and Visualization
All statistical analyses have been carried out utilizing R software program (model 4.4.3). MR analyses have been carried out utilizing the TwoSampleMR (v0.6.15) and MRPRESSO (v1.0) R packages. Visualization of forest plots, volcano plots, bubble plots, and regional affiliation plots was generated utilizing ggplot2 (v3.4.0), ggrepel, and coloc (v5.2.3). For the first gene-HSIL and metabolite-HSIL analyses, FDR correction was utilized throughout all assessments (11,000+ genes and 1,400 metabolites, respectively). For the eQTL-metabolite evaluation, FDR correction was utilized per gene throughout the 11 candidate metabolites. A q-value < 0.05 was thought-about statistically vital.
Outcomes
Genome-Broad MR Evaluation Recognized 11 Genes Causally Related to HSIL
Of the 11,000+ genes examined, 11 confirmed vital causal associations with HSIL after FDR correction (q < 0.05) (Figure 1A and Supplementary Figure 2).
Amongst these, 5 genes have been recognized as threat elements (OR > 1): VWA7 (OR = 1.77, P = 3.73×10−11), PAX8 (OR = 1.17, P = 9.92×10−11), GUSBP1 (OR = 1.47, P = 3.76×10−5), IKZF3 (OR = 1.23, P = 4.90×10−8), and PAX8-AS1 (OR = 1.16, P = 3.90×10−8). 5 genes have been protecting (OR < 1): ERBB2 (OR = 0.49, P = 1.24×10−5), COL11A2 (OR = 0.60, P = 3.40×10−5), SKIV2L (OR = 0.79, P = 2.76×10−10), TCF19 (OR = 0.78, P = 1.38×10−6), and PGAP3 (OR = 0.84, P = 7.94×10−6). NFKBIL1 was additionally considerably related to elevated threat (OR = 1.79, P = 7.05×10−14), however leave-one-out evaluation recognized an influential SNP; this discovering ought to due to this fact be interpreted with warning. (Table 1 and Supplementary Table 2).
 |
Desk 1 Mendelian Randomization Estimates for Genetically Predicted Gene Expression Related to HSIL Threat
|
Sensitivity analyses have been carried out to evaluate the robustness of the causal estimates. All IVs had F-statistics > 10 (vary: 20.8–2577.9), indicating no weak instrument bias (Supplementary Table 3). The MR-Egger intercept take a look at confirmed no proof of horizontal pleiotropy for any gene (all P > 0.05), indicating that horizontal pleiotropy was unlikely to bias the first IVW estimates (Supplementary Table 4). Heterogeneity evaluation revealed various levels of heterogeneity throughout genes. Particularly, COL11A2 confirmed vital heterogeneity (IVW Q = 19.42, P = 2.23×10−4) and VWA7 demonstrated nominally vital heterogeneity (IVW Q = 15.97, P = 0.014), whereas the remaining 9 genes confirmed no proof of serious heterogeneity (all IVW P > 0.05) (Supplementary Table 4). To account for potential heterogeneity, random-effects IVW fashions have been utilized for COL11A2 and VWA7. Importantly, MR-PRESSO international assessments confirmed the absence of serious outlier SNPs for all genes (all International P > 0.05), and no outliers have been detected or corrected for any publicity (Supplementary Table 5). Go away-one-out sensitivity evaluation recognized rs113381230 as an influential SNP for NFKBIL1. When this SNP was excluded, the causal estimate for NFKBIL1 decreased from β=0.581 to β=0.385, indicating that the affiliation could also be pushed by this single genetic variant (Supplementary Figures 3–7).
Eleven Circulating Metabolites Have been Causally Related to HSIL
After FDR correction (q < 0.05), 11 of the 1,400 circulating metabolites examined remained considerably related to HSIL, comprising seven threat metabolites (OR vary: 1.05–1.14) and 4 protecting metabolites (OR vary: 0.90–0.97) (Figure 1B and Supplementary Figure 8a).
Phospholipids represented the biggest class (5/11) (Supplementary Figure 8b). Threat-associated metabolites included 4 glycerophosphocholines (OR vary: 1.05–1.14, all P < 0.001), the arachidonate ratio (OR = 1.10, P = 9.59×10−5), sphingomyelin (OR = 1.10, P = 3.17×10−4), and 1-arachidonylglycerol (OR = 1.11, P = 3.26×10−4). Protecting metabolites included X-18921 (OR = 0.90, P = 3.35×10−5), the oleoyl-linoleoyl-glycerol ratio (OR = 0.93, P = 1.49×10−5), and two bilirubin degradation merchandise (OR = 0.93 for each, P < 0.001), according to their recognized antioxidant properties (Table 2 and Supplementary Table 6).
 |
Desk 2 Mendelian Randomization Estimates for Circulating Metabolites Related to HSIL Threat
|
All instrumental variables had F-statistics starting from 19.5 to 1893 (all > 10), indicating no weak instrument bias. Sensitivity analyses confirmed the robustness of all 11 metabolite-HSIL associations. No horizontal pleiotropy was detected (all Egger P > 0.05). Ten metabolites confirmed no heterogeneity, whereas sphingomyelin exhibited nominal heterogeneity (IVW P = 0.034), warranting a random-effects mannequin. MR-PRESSO recognized no outliers (all International P > 0.05) (Supplementary Tables 7–9). Go away-one-out evaluation revealed no influential SNPs for any metabolite (Supplementary Figures 9–14).
COL11A2 is the Solely Gene with Detectable Metabolic Mediation — However the Impact is Remarkably Small
Among the many 11 causal genes, we examined whether or not circulating metabolites mediated their results on HSIL. First, we examined every gene in opposition to the 11 HSIL-associated metabolites. Notably, COL11A2 was the one gene exhibiting vital associations with any of those metabolites after FDR correction (q < 0.05) (Supplementary Figure 15). Particularly, COL11A2 expression was considerably related to two phospholipid metabolites: 1-stearoyl-2-arachidonoyl-gpc (GCST90200685) (β = −0.143, P = 0.0017, q = 0.0094) and 1-palmitoyl-2-arachidonoyl-gpc (GCST90200692) (β = −0.125, P = 0.0069, q = 0.0190) (Supplementary Tables 10 and 11). For the remaining 10 genes, no vital eQTL-metabolite associations have been detected (all q > 0.05), suggesting they could act by way of direct mechanisms impartial of circulating metabolites. We due to this fact carried out two-step MR mediation evaluation for COL11A2 utilizing these two metabolites as candidate mediators. The 2 phospholipid metabolites mediated just one.46% (95% CI: 0.44%-3.49%) and 1.45% (95% CI: 0.30%-3.71%) of the entire protecting impact of COL11A2 on HSIL, respectively (Figure 2, Supplementary Figure 16a and Supplementary Table 12).
 |
Determine 2 Mediation evaluation of the COL11A2-HSIL protecting impact by way of two phospholipid metabolites. Notes: GCST90200685, 1-stearoyl-2-arachidonoyl-gpc; GCST90200692, 1-palmitoyl-2-arachidonoyl-gpc. Arrows symbolize causal paths within the two-step Mendelian randomization framework. Numbers adjoining to arrows point out the proportion of the entire impact of COL11A2 on HSIL mediated by every metabolite (%), with 95% confidence intervals proven in parentheses. GCST90200685 and GCST90200692 mediated 1.46% (95% CI: 0.44%–3.49%) and 1.45% (95% CI: 0.30%–3.71%) of the entire protecting impact, respectively. The remaining proportion represents the direct impact of COL11A2 on HSIL not mediated by the measured circulating metabolites. Stable arrows point out direct results; dashed arrows point out mediated results. All estimates have been derived utilizing the two-step MR technique with FDR correction utilized for a number of testing.
|
All instrumental variables for COL11A2 had F-statistics > 10 (vary: 26.3–409.3), and MR-PRESSO detected no outlier SNPs for both metabolite (each International P > 0.05) (Supplementary Table 13). Go away-one-out evaluation additional demonstrated that the mediated proportion remained steady after excluding any particular person SNP, starting from 1.39% to 1.63% for GCST90200685 and from 1.32% to 1.55% for GCST90200692 (Supplementary Figure 16b and Supplementary Table 14). Taken collectively, these sensitivity analyses affirm the reliability of the mediation estimates. This remarkably small mediated proportion (<2%) signifies that the protecting impact of COL11A2 on HSIL is basically impartial of circulating phospholipid metabolites, suggesting different mechanisms.
COL11A2 Exhibits Weak Colocalization Proof
To additional examine the genetic structure underlying these associations, we carried out colocalization evaluation for all 11 genes. Eight genes confirmed sturdy proof of shared causal variants with HSIL (PP.H4 > 0.98), together with NFKBIL1, VWA7, PAX8, SKIV2L, PAX8-AS1, IKZF3, TCF19, and PGAP3 (Table 3, Figure 3A and Supplementary Figures 17–20). This means that the eQTL and GWAS indicators for these genes are pushed by the identical causal variants.
 |
Desk 3 Colocalization Evaluation of HSIL-Related Genes
|
In marked distinction, COL11A2 confirmed weak colocalization proof (PP.H4 = 1.58×10−15), with PP.H3 approaching unity (Table 3). The regional affiliation plot (Figure 3B) visually demonstrates that the 4 COL11A2 eQTL SNPs (highlighted in purple) weren’t the lead GWAS indicators at this locus. Notably, rs213197 confirmed reverse directional results: it was negatively related to COL11A2 expression (β = −0.194, P = 6.66×10−44) however positively related to HSIL threat (β = 0.115, P = 6.69×10−11).
These findings point out that the eQTL and GWAS indicators on the COL11A2 locus are pushed by distinct causal variants, additional supporting the conclusion that COL11A2 influences HSIL by way of mechanisms impartial of the first GWAS indicators at this locus.
 |
Determine 3 Colocalization and practical enrichment analyses. (A) Colocalization of HSIL-associated genes. (B) Regional affiliation plot for the COL11A2 locus on chromosome 6p21.3. (C) Gene Ontology (GO) enrichment evaluation of COL11A2-correlated genes. (D)KEGG pathway enrichment bar plot exhibiting prime enriched pathways for COL11A2-correlated genes. Notes: Bar plot exhibiting the posterior chance of speculation 4 (PP.H4) for every of the 9 genes with out there colocalization information. PP.H4 represents the chance that each gene expression and HSIL share the identical causal variant. The horizontal dashed line signifies the edge for sturdy colocalization proof (PP.H4 > 0.8). Genes with PP.H4 > 0.98 are proven in blue; COL11A2 (PP.H4 = 1.58 × 10−15) is proven in purple, indicating weak colocalization proof with distinct causal variants. PP.H4 values have been calculated utilizing the coloc.abf technique. Higher panel: eQTL indicators for COL11A2 gene expression; decrease panel: GWAS indicators for HSIL. The x-axis represents genomic place (hg19 coordinates); the y-axis represents −log10(P-value) for every SNP. The 4 COL11A2 eQTL SNPs used as instrumental variables are highlighted in purple; all different SNPs are proven in blue. Bubble plot exhibiting the highest enriched organic processes amongst genes considerably correlated with COL11A2 expression (|r| > 0.4, P < 0.05) within the TCGA cervical most cancers dataset. The x-axis represents the gene ratio (variety of enriched genes / complete genes within the pathway); the y-axis represents the enriched organic processes. Bubble measurement is proportional to the variety of genes enriched in every course of; bubble coloration represents the adjusted P-value (Benjamini-Hochberg technique), with purple indicating greater significance (smaller P and blue indicating decrease significance (bigger P). All proven pathways had adjusted P < 0.05. Bar plot exhibiting the highest enriched KEGG pathways amongst COL11A2-correlated genes. The x-axis represents the gene rely (variety of genes enriched in every pathway); the y-axis represents the KEGG pathway names. Bar coloration represents the adjusted P-value (Benjamini-Hochberg technique), with darker purple indicating greater significance. All proven pathways had adjusted P < 0.05.
|
COL11A2-Associated Genes are Enriched in ECM and PI3K-Akt Pathways
To discover the potential organic mechanisms by way of which COL11A2 could affect HSIL, we carried out GO and KEGG enrichment evaluation on genes correlated with COL11A2 expression (|r| > 0.4, P < 0.05) within the TCGA cervical most cancers dataset. GO evaluation revealed vital enrichment in extracellular matrix (ECM)-related organic processes, together with “extracellular matrix group” (adjusted P = 8.2×10−10), “collagen fibril group” (adjusted P = 3.1×10−7), and “cell adhesion” (adjusted P = 2.5×10−6) (Figure 3C and Supplementary Table 15). KEGG pathway evaluation recognized vital enrichment in “ECM-receptor interplay” (adjusted P = 1.2×10−6), “Focal adhesion” (adjusted P = 3.8×10−6), and “PI3K-Akt signaling pathway” (adjusted P = 4.5×10−5) (Figure 3D and Supplementary Table 16). The highest 5 enriched KEGG pathways and their interrelationships are offered as a community plot (Supplementary Figure 21).
These findings counsel that COL11A2 could exert its protecting impact on HSIL by way of regulating ECM reworking and cell adhesion, doubtlessly by way of the PI3K-Akt signaling pathway — a speculation that aligns with the small mediated proportion noticed by way of circulating metabolites.
Dialogue
This research systematically recognized genetically predicted genes and circulating metabolites related to HSIL threat utilizing two-sample MR and additional explored potential mediation relationships by way of two-step MR. Amongst 11 genes and 11 metabolites causally related to HSIL, we discovered that the protecting impact of COL11A2 was minimally mediated by circulating phospholipid metabolites (mediated proportion <2%), suggesting that COL11A2 would possibly exert its protecting perform primarily by way of different mechanisms, similar to ECM-receptor interplay and PI3K-Akt signaling pathways. These findings present novel genetic and metabolic insights into HSIL etiology, and COL11A2 emerged as a candidate protecting gene meriting additional mechanistic investigation.
Our discovering that NFKBIL1, VWA7, and IKZF3—all situated inside the HLA area on chromosome 6p21.3—are risk-associated genes for HSIL is according to earlier GWAS findings implicating this area in cervical most cancers susceptibility.21 The HLA class II area has lengthy been acknowledged as a key determinant of HPV persistence and cervical carcinogenesis, possible by way of its function in antigen presentation and immune response modulation. Our MR evaluation offers causal proof supporting that genetically predicted expression of those immune-related genes instantly influences HSIL threat, extending past affiliation indicators recognized by GWAS.
PAX8 and its antisense transcript PAX8-AS1 have been additionally recognized as risk-associated genes for HSIL. PAX8 is a transcription issue important for Müllerian duct growth, and its continued expression in grownup cervical epithelium could affect susceptibility to HPV-mediated transformation. PAX8-AS1, the antisense transcript of PAX8, could regulate PAX8 expression by way of cis-acting mechanisms. Practical research have demonstrated that PAX8-AS1 suppresses cervical most cancers development by way of the miR-675-3p/DCN axis, with decrease expression correlating with superior FIGO stage and lymph node metastasis.22 The concordant threat associations of each PAX8 and its antisense transcript help a causal function for this locus in HSIL susceptibility.
The potential protecting affiliation of ERBB2 with HSIL (OR = 0.49) is sudden given its established oncogenic function in breast and gastric cancers. ERBB2 (HER2) amplification drives tumor development in these malignancies, but its function in cervical carcinogenesis seems distinct. Though direct proof linking ERBB2 to cervical most cancers threat is proscribed, earlier MR research have demonstrated context-dependent genetic results on cervical most cancers susceptibility, suggesting that genetic determinants could function by way of tissue-specific mechanisms.23 The protecting affiliation we noticed might mirror ERBB2’s function in epithelial differentiation and restore mechanisms within the regular cervical epithelium. Nevertheless, on condition that ERBB2 had solely two instrumental variables (precluding formal pleiotropy testing), this discovering must be interpreted with acceptable warning and requires experimental validation.
Our metabolite findings are broadly according to present lipidomic research in cervical lesions. A shotgun lipidomic research demonstrated that glycerophospholipids—notably phosphatidylcholines and phosphatidylethanolamines—are considerably altered throughout HPV-associated cervical transformation, with over 90% of marker lipids positively correlating with the diploma of cervical neoplasia.24 Equally, built-in microbiome-metabolome analyses have revealed vital alterations in lipid and natural acid metabolites throughout cervical lesion development, with distinct metabolic profiles distinguishing cervical most cancers (CC) from precancerous lesions.25 Our MR evaluation extends these observational findings by offering genetic proof that circulating phospholipids are causally related to HSIL threat, lowering considerations about confounding that plague observational metabolomics research. The protecting affiliation of bilirubin degradation merchandise (OR = 0.93 for each) aligns with the recognized antioxidant properties of bilirubin. Oxidative stress has been implicated in HPV-mediated cervical carcinogenesis, and endogenous antioxidants could mitigate DNA injury and irritation. This discovering is according to a latest three-stage MR research demonstrating that iron standing impacts cervical most cancers threat by way of metabolic pathways involving cortisone and carnitine,26 additional supporting the function of metabolic and oxidative stress mechanisms in cervical carcinogenesis.
COL11A2 emerged as a novel protecting gene for HSIL (OR = 0.60), with remarkably minimal mediation by circulating phospholipid metabolites (<2%). COL11A2 encodes one of many three alpha chains of kind XI collagen, a minor fibrillar collagen that regulates collagen fibril diameter and group. Whereas COL11A2 mutations are classically related to inherited connective tissue problems similar to Stickler syndrome and otospondylomegaepiphyseal dysplasia, its function in most cancers is rising. Our practical enrichment evaluation revealed that COL11A2-correlated genes are considerably enriched in ECM-receptor interplay and PI3K-Akt signaling pathways -two interconnected pathways that govern cell-matrix communication and intracellular signaling. The ECM-receptor interplay pathway is essential for cell adhesion, migration, and survival; dysregulation of ECM reworking is a trademark of most cancers development.27,28 The PI3K-Akt signaling pathway is ceaselessly activated in cervical most cancers and promotes cell proliferation, survival, and angiogenesis.29,30 The minimal mediated proportion (<2%) by way of circulating phospholipid metabolites means that COL11A2’s protecting impact is basically impartial of systemic metabolic modifications, pointing as a substitute to native mechanisms involving ECM reworking and cell signaling.
This discovering is especially intriguing given the weak colocalization proof (PP.H4 = 1.58×10−15), which signifies that the eQTL and GWAS indicators on the COL11A2 locus are pushed by distinct causal variants. The commentary that the lead eQTL SNPs for COL11A2 aren’t the first GWAS indicators at this locus, along with the alternative directional impact of rs213197—negatively related to COL11A2 expression however positively related to HSIL threat—additional helps the protecting function of upper COL11A2 expression by way of mechanisms impartial of the first GWAS indicators. Various mechanisms could embrace direct results on ECM composition and stiffness within the cervical stroma, influencing epithelial cell habits, modulation of collagen signaling by way of integrin and discoidin area receptors, or regulation of immune cell infiltration and performance within the cervical microenvironment. A plasma proteomic research of HSIL and cervical most cancers recognized complement and coagulation pathways, ldl cholesterol metabolism, and IL-17 signaling as enriched pathways,31 elevating the likelihood that COL11A2 could intersect with these processes, although direct proof linking COL11A2 to those pathways stays to be established.
The identification of 11 causal genes and 11 circulating metabolites offers potential targets for biomarker growth. Present cervical most cancers screening methods—cytology and HPV testing—have well-documented limitations in sensitivity and specificity. Metabolite-based biomarkers supply a number of theoretical benefits: they’re quantifiable in blood, mirror systemic physiological states, and could also be modifiable by way of life-style interventions. For example, a mixed panel of metabolites similar to phospholipids and bilirubin degradation merchandise has been prompt to have potential as an adjunctive blood-based take a look at to triage HPV-positive ladies, notably these with cytological abnormalities. Nevertheless, the small impact sizes noticed for many metabolites point out that particular person metabolites are unlikely to function standalone diagnostic biomarkers. As a substitute, metabolite panels or built-in genetic-metabolic threat scores could have higher utility. Such multi-marker approaches might assist refine present screening algorithms, particularly in settings the place colposcopy sources are restricted. The mix of key microbiota and metabolites has proven excessive diagnostic worth for CC in latest research,25 suggesting that multi-omics approaches—integrating genomics, metabolomics, and microbiome information—warrant additional investigation for cervical precancer threat stratification.
A number of limitations of this research must be acknowledged. First, the eQTL information have been derived from blood samples, which can not totally seize cervical tissue-specific gene expression. Second, our findings have been based mostly on European-ancestry populations, limiting generalizability to different ethnic teams. Third, the small variety of instrumental variables for ERBB2 and TCF19 (two SNPs every) and the influential SNP recognized for NFKBIL1 warrant cautious interpretation of those particular associations. Fourth, whereas FDR correction was utilized to attenuate false positives, the potential for kind I errors can’t be excluded. Moreover, the modest impact sizes noticed for many metabolites (OR < 1.15) counsel that particular person metabolites are unlikely to function standalone scientific biomarkers—a key consideration for future translational efforts. Lastly, given the exploratory nature of MR research, all findings require replication in impartial cohorts and practical validation. Regardless of these limitations, the convergence of our MR estimates with latest multi-omics and practical research helps the credibility of the general findings and offers a basis for future mechanistic and biomarker analysis.
Conclusions
This MR research recognized 11 genes and 11 circulating metabolites related to HSIL threat, extending earlier GWAS findings by making use of a unidirectional two-sample MR framework and exploring metabolic mediation. COL11A2 emerged as a protecting issue whose affiliation was minimally mediated by circulating phospholipid metabolites, suggesting that native mechanisms—doubtlessly involving ECM-receptor interplay and PI3K-Akt signaling—could underlie its impact. These findings present candidates for future mechanistic and replication research.
Knowledge Sharing Assertion
The eQTL abstract statistics have been obtained from the eQTLGen consortium (https://www.eqtlgen.org/). The GWAS abstract statistics for circulating metabolites have been obtained from Chen et al (https://doi.org/10.1038/s41467-022-31175-4). The HSIL GWAS abstract statistics have been obtained from FinnGen launch 12. (https://www.finngen.fi/). All information analyzed on this research are publicly out there.
Ethics Approval and Consent to Take part
This research used solely publicly out there summary-level GWAS information from beforehand printed research (eQTLGen consortium, Chen et al, and FinnGen). All unique research obtained moral approval and knowledgeable consent from their individuals. The current research didn’t contain any main information assortment or human participant contact. In response to Article 32, Objects 1 and a couple of of the Measures for Moral Evaluate of Life Science and Medical Analysis Involving Human Topics (collectively issued by the Nationwide Well being Fee, the Ministry of Schooling, the Ministry of Science and Expertise, and the Nationwide Administration of Conventional Chinese language Drugs of the Individuals’s Republic of China on February 18, 2023), analysis utilizing legally obtained publicly out there information or anonymized data information, and which doesn’t trigger hurt to human topics, contain delicate private data, or contain business pursuits, is exempt from moral assessment. Due to this fact, this research was exempt from extra institutional moral assessment. The exemption was confirmed by the Division of Gynecological Oncology, Most cancers Hospital of Gansu Province, below the provisions of the above-mentioned nationwide laws.
Acknowledgments
The authors thank all consortia and investigators who made the summary-level information publicly out there.
Creator Contributions
All authors made a major contribution to the work reported, whether or not that’s within the conception, research design, execution, acquisition of knowledge, evaluation and interpretation, or in all these areas; took half in drafting, revising or critically reviewing the article; gave last approval of the model to be printed; have agreed on the journal to which the article has been submitted; and conform to be accountable for all facets of the work.
Funding
This work was supported by Scientific Analysis Program Challenge of Well being Trade in Gansu Province (No. GSWSKY2021-027).
Disclosure
The authors declare no competing pursuits.
References
1. Sung H, Ferlay J, Siegel RL, et al. International most cancers statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 nations. CA Most cancers J Clin. 2021;71(3):209–13. doi:10.3322/caac.21660
2. Molina MA, Steenbergen RDM, Pumpe A, Kenyon AN, Melchers WJG. HPV integration and cervical most cancers: a failed evolutionary viral trait. Traits Mol Med. 2024;30(9):890–902. doi:10.1016/j.molmed.2024.05.009
3. de Souza ATB, de Sousa Santos DL, Medeiros FS, et al. Comparability of IL-10 gene promoter polymorphisms and haplotypes between high-grade squamous intraepithelial lesions or cervical most cancers and unfavourable cervical cytology. Sci Rep. 2025;15(1):28330. doi:10.1038/s41598-025-12851-5
4. Yin J, Zhang S, Wang S, et al. Age-specific scientific efficiency of HPV-based vs cytology-based cervical most cancers screening in China. Nat Commun. 2025;17(1):521. doi:10.1038/s41467-025-67215-4
5. Zhou Q, Li T, Zhu L, et al. The scientific efficiency of various cervical most cancers screening methods in premenopausal and postmenopausal ladies. Menopause. 2025;32(7):640–647. doi:10.1097/GME.0000000000002546
6. Nanda Ok, McCrory DC, Myers ER, et al. Accuracy of the Papanicolaou take a look at in screening for and follow-up of cervical cytologic abnormalities: a scientific assessment. Ann Intern Med. 2000;132(10):810–819. doi:10.7326/0003-4819-132-10-200005160-00009
7. Curry SJ, Krist AH, Owens DK, et al. Screening for cervical most cancers: us preventive companies process power advice assertion. Jama. 2018;320(7):674–686. doi:10.1001/jama.2018.10897
8. Fontham ETH, Wolf AMD, Church TR, et al. Cervical most cancers screening for people at common threat: 2020 guideline replace from the American Most cancers Society. CA Most cancers J Clin. 2020;70(5):321–346. doi:10.3322/caac.21628
9. Salibindla D, Jug R. Benchmarking high-risk human papillomavirus testing in opposition to cytology and biopsy leads to the detection of cervical dysplasia to tell scientific screening tips. J Am Soc Cytopathol. 2025;14(2):86–90. doi:10.1016/j.jasc.2024.11.003
10. Woo CG, Son SM, Hwang HK, Bae JS, Lee OJ, Sensitivity LHC. Specificity, and cost-benefit impact between main human papillomavirus testing, main liquid-based cytology, and co-testing algorithms for cervical lesions. Diagn Cytopathol. 2025;53(1):35–42. doi:10.1002/dc.25410
11. Beckhaus T, Kachuri L, Nakase T, et al. Genome-Broad Affiliation Analyses of HPV16 and HPV18 Seropositivity Establish Susceptibility Loci for Cervical Most cancers. J Med Virol. 2025;97(2):e70195. doi:10.1002/jmv.70195
12. Masuda T, Ito H, Hirata J, et al. Tremendous mapping of the most important histocompatibility advanced area and affiliation of the HLA-B*52:01 allele with cervical most cancers in Japanese ladies. JAMA Community Open. 2020;3(10):e2023248. doi:10.1001/jamanetworkopen.2020.23248
13. Võsa U, Claringbould A, Westra HJ, et al. Massive-scale cis- and trans-eQTL analyses establish 1000’s of genetic loci and polygenic scores that regulate blood gene expression. Nat Genet. 2021;53(9):1300–1310. doi:10.1038/s41588-021-00913-z
14. Chen Y, Lu T, Pettersson-Kymmer U, et al. Genomic atlas of the plasma metabolome prioritizes metabolites implicated in human illnesses. Nat Genet. 2023;55(1):44–53. doi:10.1038/s41588-022-01270-1
15. Bowden SJ, Bodinier B, Kalliala I, et al. Genetic variation in cervical preinvasive and invasive illness: a genome-wide affiliation research. Lancet Oncol. 2021;22(4):548–557. doi:10.1016/S1470-2045(21)00028-0
16. Yang C, Qin L-H, Li L, et al. Blood metabolites, mitochondrial biofunction, and cervical most cancers: a bidirectional Mendelian randomization research. BMC Womens Well being. 2025;25(1):363. doi:10.1186/s12905-025-03909-7
17. Sanderson E, Levin MG, Walker V, et al. Challenges and future instructions for Mendelian randomization. Nat Genet. 2026;58(5):984–994. doi:10.1038/s41588-026-02546-6
18. Lepik Ok, Auwerx C, Sadler MC, van der Graaf A, Ojavee SE, Kutalik Z. Lima: sturdy inference of molecular mediation from abstract statistics. Am J Hum Genet. 2026;113(1):202–220. doi:10.1016/j.ajhg.2025.12.005
19. Kurki MI, Karjalainen J, Palta P, et al. FinnGen offers genetic insights from a well-phenotyped remoted inhabitants. Nature. 2023;613(7944):508–518. doi:10.1038/s41586-022-05473-8
20. Giambartolomei C, Vukcevic D, Schadt EE, et al. Bayesian take a look at for colocalisation between pairs of genetic affiliation research utilizing abstract statistics. PLoS Genet. 2014;10(5):e1004383. doi:10.1371/journal.pgen.1004383
21. Chen D, Enroth S, Liu H, et al. Pooled evaluation of genome-wide affiliation research of cervical intraepithelial neoplasia 3 (CIN3) identifies a brand new susceptibility locus. Oncotarget. 2016;7(27):42216–42224. doi:10.18632/oncotarget.9916
22. Lu X, Track P, Chen Q. lncRNA PAX8-AS1 suppresses cervical most cancers growth by regulating miR-675-3p/DCN axis. Int J Biol Markers. 2025;3936155251386471. doi:10.1177/03936155251386471
23. Fukazawa EM, Baiocchi G, Soares FA, et al. Cox-2, EGFR, and ERBB-2 expression in cervical intraepithelial neoplasia and cervical most cancers utilizing an automatic imaging system. Int J Gynecol Pathol. 2014;33(3):225–234. doi:10.1097/PGP.0b013e318290405a
24. Starodubtseva NL, Chagovets VV, Nekrasova ME, et al. Shotgun lipidomics for differential prognosis of hpv-associated cervix transformation. Metabolites. 2022;12(6):503. doi:10.3390/metabo12060503
25. Zhai Q, Zhao L, Wang M, et al. Built-in evaluation of microbiome and metabolome reveals insights into cervical neoplasia aggravation in a Chinese language cohort. Entrance Cell Infect Microbiol. 2025;15:1556153. doi:10.3389/fcimb.2025.1556153
26. Wang Z, He Y, Wang Y, et al. Iron standing and most cancers threat throughout 14 cancers: proof from cross-sectional analyses and a three-stage mendelian randomization. Int J Surg. 2025;2025:1. doi:10.1097/JS9.0000000000004245
27. Li J, Chen X, Li J. Exploration and validation of the prognostic worth of mitophagy and mitochondrial dynamics-related genes in cervical most cancers. Sci Rep. 2025;15(1):24950. doi:10.1038/s41598-025-09310-6
28. Lu P, Weaver VM, Werb Z. The extracellular matrix: a dynamic area of interest in most cancers development. J Cell Biol. 2012;196(4):395–406. doi:10.1083/jcb.201102147
29. Pan Z, Li L, Fang Q, et al. Built-in bioinformatics evaluation of grasp regulators in anaplastic thyroid carcinoma. Biomed Res Int. 2019;2019:9734576. doi:10.1155/2019/9734576
30. Luo W, Li Y, Li G, Yang M. RING field protein-1 promotes the metastasis of cervical most cancers by way of regulating matrix metalloproteinases by way of PI3K/AKT signaling pathway. J Mol Histol. 2025;56(2):120. doi:10.1007/s10735-025-10396-5
31. Han S, Zhang J, Solar Y, et al. The plasma DIA-based quantitative proteomics reveals the pathogenic pathways and new biomarkers in cervical most cancers and excessive grade squamous intraepithelial lesion. J Clin Med. 2022;11(23):7155. doi:10.3390/jcm11237155
