Utility of blood DNA methylation and immune cell-type estimates to identify clinically relevant links between immunity and health conditions in a large cohort

The immune system plays a central role in health and disease. Changes in the proportions of immune cell types have been linked with diverse pathological conditions. However, direct measurements of circulating immune cells are often not available, especially in large human cohorts. Here, whole blood DNA methylation (DNAm) array data might be a feasible solution to study the links between immune cells and human health. This study investigated associations between DNAm estimates of six immune cell types and fifteen circulating serum biomarkers related to health conditions among 1007 participants of the BiDirect Study and then performed follow-up analyses. We found an association between the granulocyte estimates (Gran) and serum levels of high-sensitivity C-reactive protein (hsCRP; beta=0.021, p=3.71x10^-5, false discovery rate-FDR=0.0038). This was further confirmed by polygenic score analysis (beta=0.0325, p=0.041). In follow-up analyses, we identified a statistical interaction between Gran and hsCRP linked to the diagnosis of chronic pain (beta = -5.65, p=0.0047, FDR = 0.047) and showed significant Gran-hsCRP interaction effects on pain perception as assessed with the pain sensitivity questionnaire (PSQ-minor; beta=0.43, p=0.0125). In addition, by analyzing DNAm levels in granulocyte-expressed genes, we pinpointed potential novel target genes for chronic pain, including PLBD1 and TAPBP. These findings exemplify the utility of blood DNAm and immune cell-type estimates to help identify clinically relevant links between immunity and health outcomes in large cohorts.

1 Introduction

The immune system plays a central role in health and disease. As different immune cell types have different roles, their distribution in the body determines different immune functions (1). Changes in the proportions of various immune cell types have been linked with pathological conditions, including somatic and psychiatric diseases, as well as with the exposure to disease risk factors (2, 3). Therefore, shifts in immune cell-type proportions might represent clinically relevant predictors for risk assessment, diagnosis and prognosis in the context of an array of physiological and mental health conditions. However, because these changes might be subtle in non-immune disorders, measurements of circulating immune cell counts are rarely available in human cohorts outside of this context. In such cases, a feasible and pragmatic alternative is presented by the advent of DNA methylation (DNAm) profiling in whole blood. Due to the high cell-type specificity of DNAm, current deconvolution algorithms can render accurate quantifications of cell-type fractions using DNAm array data. Such estimates have proven useful resources to interrogate the links between immune cell-type shifts and diverse phenotypes in large samples (3).

In the current study, we aimed to assess the utility of whole blood DNAm and derived estimates of primary immune cell types to explore the relationships between immune variability and health in the German BiDirect Study. To achieve this, we performed a screening analysis where we tested the associations between six cell types and fifteen health-related biomarkers. Next, we conducted data-driven analyses to follow-up on significant findings and explore potential implications for disease phenotypes. This approach revealed links between granulocytes, C-reactive protein (CRP) and chronic pain in BiDirect.

2 Methods

2.1 Study population

The study design and methods of the BiDirect Study have been described elsewhere (4). Briefly, the study recruited over 2300 participants in the district of Münster, Germany, that were allocated to one of three cohorts: 1) a depression cohort (patients hospitalized due to an acute depressive episode at the time of recruitment, n=1004), 2) a cardiovascular cohort (patients with an acute coronary event three months before recruitment, n=348), or 3) a population-based cohort (randomly sampled community-dwelling adults, n=965). BiDirect participants underwent extensive phenotyping within a 10-year span, including four examinations with face-to-face interviews, laboratory tests, neuropsychological assessments, genotyping and DNAm profiling, and self-administered questionnaires applied during examinations and intermediate follow-ups. The current investigation employed a subsample of 1007 BiDirect participants from the depression (cohort 1, n=442) and population-based (cohort 3, n=565) cohorts with DNAm data available from the baseline assessment (mean age: 53 ± 8; females: 536, 53%).

2.2 Immune cell-type estimates

DNAm levels were measured from whole blood samples in two batches using the Infinium MethylationEPIC v1.0 or v2.0 BeadChip (Illumina Inc. San Diego, USA) arrays. Cell-type estimations were made from raw DNAm beta-values for granulocytes (Gran), monocytes (Mono), B cells (Bcell), CD4⁺ T cells (CD4T), CD8⁺ T cells (CD8T) and natural killer (NK) cells according to the method by Houseman et al., 2012 (5) using the online DNA Methylation Age Calculator (https://dnamage.clockfoundation.org/) (6).

2.3 Health-related biomarkers

Fifteen laboratory-derived blood measurements were tested for association with the immune cell-type estimates. These included metals (iron, copper, zinc), markers of inflammation (high-sensitivity C-reactive protein [hsCRP], interleukin [IL]-6, IL-1beta, tumor necrosis factor [TNF]-alpha and interferon [IFN]-gamma), liver function (alanine aminotransferase [ALAT] and aspartate aminotransferase [ASAT]), thyroid function (free tri-iodothyronine [fT3], free thyroxine [fT4] and thyroid stimulating hormone [TSH]) and neurodegeneration (neurofilament light chain [NfL] and glial fibrillary acidic protein [GFAP]).

2.4 First-step follow-up

2.4.1 Confirmatory polygenic score analysis

Follow-up analyses of the identified immune cell-biomarker associations used polygenic scores (PGSs) to elucidate the relationship between DNAm estimates and the genetic determinants of immune cells and seek genetic support for the identified links with biomarkers. PGS weight files were obtained from the PGS Catalog (7) for the proportions of lymphocytes (Cat.ID: PGS003941), monocytes (Cat.ID: PGS003942), neutrophils (Cat.ID: PGS003943), eosinophils (Cat.ID: PGS003944) and basophils (Cat.ID: PGS003945), and for CRP (Cat.ID: PGS002164). PGSs were then calculated in BiDirect by applying allelic scoring with plink 1.9 (8). Genotyping, quality control (QC) and genotype imputation procedures in BiDirect have been described elsewhere (9). Genotype data was available for all individuals included in this study.

2.4.2 Implications for disease outcomes

To evaluate implications of the identified immune cell-biomarker associations for disease outcomes, we performed interaction analyses using data on (self-reported) lifetime physician’s diagnosis of the following diseases: depression (n_cases=543), anxiety (n_cases=256), addiction (n_cases=51), psychiatric disorder (other than depression and anxiety, n_cases=99), chronic pain (e.g. pain in back, legs or joints, migraine, fibromyalgia; n_cases=311), heart problems (e.g. heart failure, n_cases=117), hypertension (n_cases=522), kidney disease (n_cases=83), diabetes (n_cases=87) and thyroid disease (n_cases=315).

2.5 Second-step follow-up

2.5.1 Pain scales

Follow-up analyses of the identified association with disease outcomes used data on pain scales. In BiDirect, pain perception was measured using the pain sensitivity questionnaire (PSQ-minor). Pressure pain threshold (PPT) was measured using a handheld mechanic pressure algometer (upper limit=10) on the left and right hands (index finger pad) separately. Detailed procedures are described elsewhere (10).

2.5.2 DNAm analysis in cell-type gene sets

To test associations with DNAm levels of genes expressed in neutrophils and/or granulocytes, the “Granulocyte” and “Neutrophil” gene sets were obtained from Harmonizome 3.0 (11). These contained 43 and 31 genes, respectively, from which 31 overlapped between both gene sets. DNAm probes corresponding to 31 neutrophil-expressed genes, 12 non-neutrophil granulocyte-expressed genes and CRP were identified using the array’s annotation file. Data pre-processing and QC followed standard procedures as previously described (12). Briefly, these included 1) data conversion, 2) calculation of detection p-values, 3) removal of probes that failed detection, located on the sex chromosomes, contained polymorphisms, and/or are cross-reactive, 4) removal of outlier samples, 5) calculation of normalized beta values using the dasen method, and 6) transformation into M-values for use in statistical analyses. These procedures were performed using the bigmelon (13) R package. The M-values of 265 probes annotated to selected genes and available in both BiDirect batches after QC were used for statistical analysis.

2.6 Statistical analysis

Cell-type estimates, laboratory measurements and pain scores were log-transformed to approximate normality. In all instances, individuals with missing values and outliers at ±3 standard deviations were removed for both cell-type and test phenotype prior to running the test, and robust regression models (R package “robustbase” (14)) adjusting for age, sex, cohort assignment, current smoking status, years of education and body-mass index (BMI) were employed. Sensitivity analyses using the original values of cell-type estimates were conducted. For significant findings, sensitivity analyses also adjusted for batch/array version with and without adjustment for years of education. For PGS analysis, the first five genomic principal components were added to the set of covariates. In the biomarker and PGS analyses, cell-type estimates were used as the dependent variable in the regression models. Interaction tests employed logistic regression for the dichotomous disease outcomes or linear regression for the pain scales as the dependent variables and a cell-type*biomarker interaction element. For both algometer variables, tobit models for right-censored data were applied using the R package “VGAM” (15). Regression results are reported as beta (i.e. coefficient estimate), standard error (SE) and p-value. For DNAm data, the first five DNAm principal components were added to the set of covariates and the analysis was performed using the R package “limma” (16). Results were considered statistically significant at false discovery rate (FDR)<0.05 in the screening analyses (i.e. association with laboratory measurements and interaction tests for disease outcomes). All other tests were considered significant at the nominal threshold (p<0.05).

3 Results

The highest and lowest cell-type estimates were found for Gran and NK, respectively (Figure 1A). After excluding individuals with missing data and outliers, the effective sample sizes for each association test in the initial screening phase ranged between 702 and 963 individuals. Results from this analysis showed a statistically significant association between Gran and hsCRP (beta=0.021, SE = 0.005, p=3.7x10^-5, FDR = 0.0038; n=931). Follow-up PGS analysis confirmed strong associations of Gran with the PGSs for the proportion of neutrophils (beta=0.1, SE = 0.017, p=6.72x10^-9) and lymphocytes (beta = -0.089, SE = 0.017, p=3.81x10^-7), and modest associations with the PGSs for the proportion of monocytes (beta = -0.039, SE = 0.019, p=0.0382) and CRP (beta=0.0325, SE = 0.016, p=0.041). Although only the association between Gran and hsCRP remained significant after correction for multiple comparisons in the initial screening analysis, nominal associations (p<0.05) were observed for Gran with fT3 and copper, for CD4T with hsCRP, copper, zinc and TSH, and for CD8T with hsCRP. Results from this screening analysis are shown in Figure 1B.

Figure 1

Figure 1. (A) DNAm-derived immune cell-type estimates (raw values representing relative proportions). (B) Results from the association analysis between DNAm-estimated immune cell-type proportions and health related biomarkers measured in serum/plasma. Coefficient estimates (Beta) are shown in color gradient. Significance levels are depicted by size; n.s.: non-significant. (C) Interaction between the proportion of granulocytes and hsCRP levels in chronic pain. The plot shows log-transformed values of Gran and hsCRP, stratified by chronic pain diagnosis. (D) Interaction between the proportion of granulocytes and hsCRP levels in pain perception. The plot shows log-transformed values of Gran and hsCRP, stratified by the level of PSQ-minor (categorized as low, medium and high according to the calculated tertils). (E) Interaction between the proportion of CD4⁺ T cells and hsCRP levels in chronic pain. The plot shows log-transformed values of CD4T and hsCRP, stratified by chronic pain diagnosis.

Next, we investigated whether the association between Gran and hsCRP might be clinically relevant by conducting analyses that tested interaction effects of Gran and hsCRP on disease diagnoses. This analysis identified a significant interaction between Gran and hsCRP in the context of lifetime chronic pain diagnoses (beta = -5.65, SE = 1.9, p=0.0047, FDR = 0.047; Figure 1C). The independent effects on chronic pain diagnosis of both Gran (beta = -7.9, SE = 2.3, p=7x10^-4) and hsCRP (beta = -1.36, SE = 0.5, p=0.0094) were also significant. Sensitivity analyses showed robustness of this finding. Also, about 32% of the study sample had history of a physician’s diagnosis of (general) chronic pain by the end of the BiDirect Study, with 219 individuals reporting this at baseline examination (i.e. prevalent chronic pain) and 92 reporting it afterwards (i.e. incident chronic pain). Therefore, we considered this relationship to be a valid observation and conducted further analyses.

Following-up on the Gran-hsCRP link to chronic pain, we found that the Gran-hsCRP interaction was slightly more significant for incident (p=0.017) than prevalent (p=0.029) chronic pain diagnoses. However, given that the relationship was significant in both instances, with effects following the same direction, we chose to continue using unstratified analyses. First, we tested the effects of Gran and hsCRP on pain sensitivity (PPT) and perception (PSQ-minor). All pain measures tested were associated with chronic pain (negative estimates for sensitivity and positive for perception). However, the Gran-hsCRP interaction was only significant for the PSQ (beta=0.43, SE = 0.17, p=0.0125; Figure 1D). The independent effects of Gran (beta=0.49, SE = 0.22, p=0.023) and hsCRP (beta=0.14, SE = 0.05, p=0.0023) on PSQ measures were significant. In addition, because CD4T and CD8T showed nominally significant associations with hsCRP, we also tested these interactions in chronic pain and found that CD4T, but not CD8T, positively interacted with hsCRP (beta=2.5, p=0.00175; n=820; Figure 1E) in chronic pain. However, CD4T showed no interaction effects with hsCRP on pain sensitivity or perception. Second, we explored associations of chronic pain diagnosis and hsCRP with DNAm levels in granulocyte-expressed genes and the CRP gene. Results of this analysis are shown in Table 1. After correction for multiple comparisons, we found that two methylation probes (cg22488164 in PLBD1 and cg24777950 in CTSG) were associated with hsCRP (n=926), while none remained significant after correction for multiple comparisons in the chronic pain analysis (n=961, 318 cases and 643 controls). Nominal (p=0.05) findings included another eight probes corresponding to six genes in the hsCRP analysis, and 12 probes corresponding to seven genes in the chronic pain analysis. Both analyses showed the same top probe (cg22488164 in PLBD1). Another three overlapping genes (GCH1, PDGFRA and TAPBP) coming from different probes were observed within the nominal-level findings.

Table 1

Table 1. Nominal-level findings from the gene set DNAm analysis.

4 Discussion

This study found associations of DNAm granulocyte estimates with CRP levels and CRP-PGS, as well as an interaction between Gran and CRP that is relevant for pain phenotypes. The finding that Gran showed a strong positive association with the genomic determinants of neutrophil proportion of leukocytes suggests that, in agreement with the literature, the major constituent of Gran are neutrophils, the most abundant type of granulocytes in the human body (1). It is known that CRP contributes to toxic granulation in neutrophils (17). Neutrophils are key in chronic inflammation (18) and chronic widespread pain in fibromyalgia (19). Moreover, higher CRP levels in individuals with chronic pain and higher pain sensitivity have been previously reported (20–22). These observations validate the utility of DNAm-derived immune cell-type estimates to study relationships between immunity and disease phenotypes in large cohort studies such as BiDirect. Beyond this, our study provides evidence of a significant interaction between immune cell proportions and CRP in the context of pain phenotypes which, to our knowledge, has not been reported before. This interaction was observed in both prevalent and incident chronic pain diagnoses in the BiDirect Study, suggesting an involvement in the development of chronic pain and potential for risk assessment. Of note, even though the diagnosis of chronic pain was associated with both medical and psychiatric diagnoses, including depression, anxiety, addiction, and heart and kidney disease, the Gran-CRP interaction was not associated with these other diagnoses, suggesting a direct relationship with pain phenotypes.

Associations with CRP were also suggested for CD4⁺ and CD8⁺ T cell estimates derived from DNAm data in our study. Although CRP is well-known for its role in inflammation, contributions to adaptive immunity via binding of CD4⁺ T cell subsets have been proposed (23). The evidence of association between CD4T and CRP was strengthened by our observation of a significant interaction of these variables in chronic pain in opposite direction to that found for Gran. Although a number of past studies found no changes in T cell counts in individuals with chronic pain, it is believed that specific CD4⁺ T cell subsets might regulate the transition from acute to chronic pain (24). In addition, the strong negative relationship between Gran and the PGS for lymphocytes in our study is in accordance with previous knowledge, as the neutrophil-to-lymphocyte ratio (NLR) has been used as a marker of inflammation involved in several diseases and overall mortality (25).

Finally, DNAm analysis identified methylation sites in PLBD1 and CTSG associated with CRP levels. Both genes have been previously linked to pain phenotypes (26, 27). Although our DNAm analysis did not reveal strong relationships between DNAm levels in granulocyte genes and a chronic pain diagnosis, nominal findings highlighted specific genes that might be relevant for pain phenotypes and worth of further investigation. In particular, PDGFRA (28), GCH1 (29, 30) and S100A8 (31, 32) have been previously associated with pain in different contexts and are targets of approved drugs. Similarly, although DNAm analyses stratified by prevalent or incident chronic pain were fairly underpowered and, therefore, results from these were not shown here, it is worth noting that the most significant DNAm probes for incident chronic pain located in promoter regions of SNAP23 (cg26348910, p=0.0045, effect=0.08; cg14205524, p=0.01, effect=0.074), whose protein product has been linked to neuropathic pain (33). Therefore, this gene might be an interesting candidate for follow-up studies that investigate its role and potential utility as predictive marker for the development of chronic pain.

Our study has some limitations. First, DNAm data was generated only for a fraction of BiDirect participants, limiting the sample size accessible for this investigation. Second, in order to avoid high burden of multiple testing, we limited the number of variables tested during the initial screening analysis. Nevertheless, our aim was not to conduct a phenome-wide analysis. Instead, we were interested in utilizing representative blood biomarkers associated with health status for this initial phase of our study. Third, chronic pain is broadly conceptualized here. However, we believe that this broad conceptualization might represent an advantage if we consider the importance of understanding the general mechanisms underlying the development of chronic pain phenotypes. Fourth, our study showed low power to detect associations with DNAm levels. Even if the literature seems encouraging, more research will be needed to determine if and how the suggested genes might be involved in the pathobiology of chronic pain and could be targeted for the management of this condition.

Overall, we found that individuals with chronic pain in the BiDirect Study showed higher pain perception but lower sensitivity to pain, as well as a shift in the relationships between serum CRP and the proportions of granulocytes and CD4⁺ T cells. Lower Gran levels appeared to add to the effects of higher CRP levels on increasing pain perception without affecting pain sensitivity, while CD4T showed no combined effects with CRP on pain ratings. Moreover, DNAm in specific granulocyte-expressed genes appeared to correlate with both CRP and chronic pain. These observations may have translational value for the management of pain phenotypes, demonstrating the utility of DNAm and derived cell-type estimates to identify relevant associations between immunity and pathological conditions in the BiDirect Study and, more generally, in large human cohorts.

Data availability statement

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.

Ethics statement

The studies involving humans were approved by the ethics committee of the University of Münster and the Westphalian Chamber of Physicians in Münster, North-Rhine-Westphalia, Germany. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent for participation was signed by the participants in accordance with the national legislation and institutional requirements.

Author contributions

MH-R: Conceptualization, Formal Analysis, Methodology, Writing – original draft. JH: Formal Analysis, Writing – review & editing. CL: Resources, Writing – review & editing. MN: Investigation, Writing – review & editing. WL: Resources, Writing – review & editing. JK: Resources, Writing – review & editing. HM: Resources, Writing – review & editing. MH: Resources, Writing – review & editing. KB: Funding acquisition, Supervision, Writing – review & editing.

Funding

The author(s) declared that financial support was received for the research and/or publication of this article. The BiDirect Study is supported by grants of the German Ministry of Research and Education (BMBF) to the University of Münster (01ER0816 and 01ER1506). CL was supported by the Heisenberg program of the DFG (DFG; LI 2654/4-1). Further support was provided by a grant from the EU Joint Program -Neurodegenerative Disease Research (JPND2021-650-289, coordinator: CL).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declare that no Generative AI was used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

References

1. Sender R, Weiss Y, Navon Y, Milo I, Azulay N, Keren L, et al. The total mass, number, and distribution of immune cells in the human body. Proc Natl Acad Sci U S A. (2023) 120:e2308511120. doi: 10.1073/pnas.2308511120

PubMed Abstract | Crossref Full Text | Google Scholar

2. Yang J, Wang Q, and Jiang W. Association between immune cells in peripheral blood and psychiatric symptoms. Front Psychiatry. (2023) 14:1198734. doi: 10.3389/fpsyt.2023.1198734

PubMed Abstract | Crossref Full Text | Google Scholar

3. Luo Q, Dwaraka VB, Chen Q, Tong H, Zhu T, Seale K, et al. A meta-analysis of immune-cell fractions at high resolution reveals novel associations with common phenotypes and health outcomes. Genome Med. (2023) 15:59. doi: 10.1186/s13073-023-01211-5

PubMed Abstract | Crossref Full Text | Google Scholar

4. Teismann H, Wersching H, Nagel M, Arolt V, Heindel W, Baune BT, et al. Establishing the bidirectional relationship between depression and subclinical arteriosclerosis–rationale, design, and characteristics of the BiDirect Study. BMC Psychiatry. (2014) 14:174. doi: 10.1186/1471-244X-14-174

PubMed Abstract | Crossref Full Text | Google Scholar

5. Houseman EA, Accomando WP, Koestler DC, Christensen BC, Marsit CJ, Nelson HH, et al. DNA methylation arrays as surrogate measures of cell mixture distribution. BMC Bioinf. (2012) 13:86. doi: 10.1186/1471-2105-13-86

PubMed Abstract | Crossref Full Text | Google Scholar

6. Horvath S. DNA methylation age of human tissues and cell types. Genome Biol. (2013) 14:R115. doi: 10.1186/gb-2013-14-10-r115. Erratum in: Genome Biol. 2015 16:96. doi: 10.1186/s13059-015-0649-6.

PubMed Abstract | Crossref Full Text | Google Scholar

7. Lambert SA, Gil L, Jupp S, Ritchie SC, Xu Y, Buniello A, et al. The Polygenic Score Catalog as an open database for reproducibility and systematic evaluation. Nat Genet. (2021) 53:420–5. doi: 10.1038/s41588-021-00783-5

PubMed Abstract | Crossref Full Text | Google Scholar

8. Chang CC, Chow CC, Tellier LC, Vattikuti S, Purcell SM, and Lee JJ. Second-generation PLINK: rising to the challenge of larger and richer datasets. Gigascience. (2015) 4:7. doi: 10.1186/s13742-015-0047-8

PubMed Abstract | Crossref Full Text | Google Scholar

9. Herrera-Rivero M, Garvert L, Horn K, Löbner M, Weitzel EC, Stoll M, et al. Berger K. A meta-analysis of genome-wide studies of resilience in the German population. Mol Psychiatry. (2025) 30:497–505. doi: 10.1038/s41380-024-02688-1

PubMed Abstract | Crossref Full Text | Google Scholar

10. Hermesdorf M, Berger K, Baune BT, Wellmann J, Ruscheweyh R, and Wersching H. Pain sensitivity in patients with major depression: differential effect of pain sensitivity measures, somatic cofactors, and disease characteristics. J Pain. (2016) 17:606–16. doi: 10.1016/j.jpain.2016.01.474

PubMed Abstract | Crossref Full Text | Google Scholar

11. Diamant I, Clarke DJB, Evangelista JE, Lingam N, and Ma’ayan A. Harmonizome 3.0: integrated knowledge about genes and proteins from diverse multi-omics resources. Nucleic Acids Res 53(D1):D1016–D1028. doi: 10.1093/nar/gkae1080

PubMed Abstract | Crossref Full Text | Google Scholar

12. Sommerer Y, Dobricic V, Schilling M, Ohlei O, Sabet SS, Wesse T, et al. Entorhinal cortex epigenome-wide association study highlights four novel loci showing differential methylation in Alzheimer’s disease. Alzheimers Res Ther. (2023) 15:92. doi: 10.1186/s13195-023-01232-7

PubMed Abstract | Crossref Full Text | Google Scholar

13. Gorrie-Stone TJ, Smart MC, Saffari A, Malki K, Hannon E, Burrage J, et al. Bigmelon: tools for analysing large DNA methylation datasets. Bioinformatics. (2019) 35:981–6. doi: 10.1093/bioinformatics/bty713

PubMed Abstract | Crossref Full Text | Google Scholar

14. Maechler M, Rousseeuw P, Croux C, Todorov V, Ruckstuhl A, Salibian-Barrera M, et al. robustbase: Basic Robust Statistics. R package version0.99-4-1 (2024). Available online at: http://robustbase.r-forge.r-project.org/ (Accessed January 15, 2025).

Google Scholar

15. Yee TW. VGAM: Vector Generalized Linear and Additive Models. R package version 1.1-13 (2025). Available online at: https://CRAN.R-project.org/package=VGAM. (Accessed January 15, 2025)

Google Scholar

16. Ritchie ME, Phipson B, Wu D, Hu Y, Law CW, Shi W, et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. (2015) 43:e47. doi: 10.1093/nar/gkv007

PubMed Abstract | Crossref Full Text | Google Scholar

17. Kabutomori O, Kanakura Y, and Iwatani Y. Movement of toxic granulation neutrophils and C-reactive protein in inflammatory processes. Am J Med. (2002) 112:595–6. doi: 10.1016/s0002-9343(02)91026-0

PubMed Abstract | Crossref Full Text | Google Scholar

18. Soehnlein O, Steffens S, Hidalgo A, and Weber C. Neutrophils as protagonists and targets in chronic inflammation. Nat Rev Immunol. (2017) 17:248–61. doi: 10.1038/nri.2017.10

PubMed Abstract | Crossref Full Text | Google Scholar

19. Caxaria S, Bharde S, Fuller AM, Evans R, Thomas B, Celik P, et al. Neutrophils infiltrate sensory ganglia and mediate chronic widespread pain in fibromyalgia. Proc Natl Acad Sci U S A. (2023) 120:e2211631120. doi: 10.1073/pnas.2211631120

PubMed Abstract | Crossref Full Text | Google Scholar

20. Huang C, Tong Q, and Tong Q. Association between C-reactive protein and chronic pain in US adults: A nationwide cross-sectional study. PloS One. (2025) 20:e0315602. doi: 10.1371/journal.pone.0315602

PubMed Abstract | Crossref Full Text | Google Scholar

21. Farrell SF, Armfield NR, Cabot PJ, Elphinston RA, Gray P, Minhas G, et al. C-reactive protein (CRP) is associated with chronic pain independently of biopsychosocial factors. J Pain. (2024) 25:476–96. doi: 10.1016/j.jpain.2023.09.008

PubMed Abstract | Crossref Full Text | Google Scholar

22. Afari N, Mostoufi S, Noonan C, Poeschla B, Succop A, Chopko L, et al. C-reactive protein and pain sensitivity: findings from female twins. Ann Behav Med. (2011) 42:277–83. doi: 10.1007/s12160-011-9297-6

PubMed Abstract | Crossref Full Text | Google Scholar

23. Zhang L, Liu SH, Wright TT, Shen ZY, Li HY, Zhu W, et al. C-reactive protein directly suppresses Th1 cell differentiation and alleviates experimental autoimmune encephalomyelitis. J Immunol. (2015) 194:5243–52. doi: 10.4049/jimmunol.1402909

PubMed Abstract | Crossref Full Text | Google Scholar

24. Laumet G, Ma J, Robison AJ, Kumari S, Heijnen CJ, and Kavelaars A. T cells as an emerging target for chronic pain therapy. Front Mol Neurosci. (2019) :216. doi: 10.3389/fnmol.2019.00216

PubMed Abstract | Crossref Full Text | Google Scholar

25. Buonacera A, Stancanelli B, Colaci M, and Malatino L. Neutrophil to lymphocyte ratio: an emerging marker of the relationships between the immune system and diseases. Int J Mol Sci. (2022) 23:3636. doi: 10.3390/ijms23073636

PubMed Abstract | Crossref Full Text | Google Scholar

26. Liu X, Tian Y, Meng Z, Chen Y, Ho IH, Choy KW, et al. Up-regulation of cathepsin G in the development of chronic postsurgical pain: an experimental and clinical genetic study. Anesthesiology. (2015) 123:838–50. doi: 10.1097/ALN.0000000000000828

PubMed Abstract | Crossref Full Text | Google Scholar

27. Zorkoltseva IV, Elgaeva EE, Belonogova NM, Kirichenko AV, Svishcheva GR, Freidin MB, et al. Multi-trait exome-wide association study of back pain-related phenotypes. Genes (Basel). (2023) 14:1962. doi: 10.3390/genes14101962

PubMed Abstract | Crossref Full Text | Google Scholar

28. Barkai O, Puig S, Lev S, Title B, Katz B, Eli-Berchoer L, et al. Platelet-derived growth factor activates nociceptive neurons by inhibiting M-current and contributes to inflammatory pain. Pain. (2019) 160:1281–96. doi: 10.1097/j.pain.0000000000001523

PubMed Abstract | Crossref Full Text | Google Scholar

29. Cronin SJF, Rao S, Tejada MA, Turnes BL, Licht-Mayer S, Omura T, et al. Phenotypic drug screen uncovers the metabolic GCH1/BH4 pathway as key regulator of EGFR/KRAS-mediated neuropathic pain and lung cancer. Sci Transl Med. (2022) 14:eabj1531. doi: 10.1126/scitranslmed.abj1531

PubMed Abstract | Crossref Full Text | Google Scholar

30. Sadhu N, Jhun EH, Yao Y, He Y, Molokie RE, Wilkie DJ, et al. Genetic variants of GCH1 associate with chronic and acute crisis pain in African Americans with sickle cell disease. Exp Hematol. (2018) 66:42–9. doi: 10.1016/j.exphem.2018.07.004

PubMed Abstract | Crossref Full Text | Google Scholar

31. Parisien M, Lima LV, Dagostino C, El-Hachem N, Drury GL, Grant AV, et al. Acute inflammatory response via neutrophil activation protects against the development of chronic pain. Sci Transl Med. (2022) 14:eabj9954. doi: 10.1126/scitranslmed.abj9954

PubMed Abstract | Crossref Full Text | Google Scholar

32. Blom AB, van den Bosch MH, Blaney Davidson EN, Roth J, Vogl T, van de Loo FA, et al. The alarmins S100A8 and S100A9 mediate acute pain in experimental synovitis. Arthritis Res Ther. (2020) 22:199. doi: 10.1186/s13075-020-02295-9

PubMed Abstract | Crossref Full Text | Google Scholar

33. Wang X, Tian S, Wang H, Liu P, Zheng H, Wu L, et al. Botulinum toxin type A alleviates neuropathic pain and suppresses inflammatory cytokines release from microglia by targeting TLR2/MyD88 and SNAP23. Cell Biosci. (2020) 10:141. doi: 10.1186/s13578-020-00501-4

PubMed Abstract | Crossref Full Text | Google Scholar