No Causal Association Between C-Reactive Protein and the Risk of Type 1 Diabetes: A Bidirectional Mendelian Randomization Study

doi:10.21203/rs.3.rs-2346001/v1

Background

Accumulating evidence from observational studies has shown that circulating C-reactive protein (CRP) levels are correlated with Type 1 diabetes (T1D) appearing a potential predictive marker of intervention, yet are of unknown causality. To clarify, we introduce a bidirectional two-sample Mendelian randomization (MR) framework to investigate the causality between circulating CRP levels and T1D.

Methods

Based on aggregated statistics from large-scale genome-wide association studies (GWAS), we evaluated the pooled impact of CRP on the risk of developing T1D. We obtained 6 single nucleotide polymorphisms (SNPs) for CRP selected as instrumental variables from a recent GWAS (n = 204,402). The T1D related SNPs were from a large-scale T1D GWAS (n = 6,808 T1D cases; n = 12,173 controls). Subsequent inverse-variance weighted (IVW) method, simple median method, weighted median method were conducted to acquire the genetic correlation between CRP levels and T1D. In sensitivity analyses, MR-Egger, MR-PRESSO, and leave-one-out analysis were applied to exclude the potentially pleiotropic variants in this study.

Results

The results of IVW provided no causal evidence that genetically predicted circulating CRP levels on the risk of T1D, with OR of 0.922 (95% CI: 0.662–1.285, P = 0.631). Furthermore, we denoted 14 T1D-related SNPs as an instrumental variable in MR analyses and yielded no significant associations of T1D on CRP levels according to the IVW result (OR: 1.000, 95% CI: 0.990–1.010, P = 0.930). MR-Egger, MR-PRESSO, and leave-one-out analysis indicated no indication for potential directional pleiotropy effects.

Conclusion

Our findings failed to provide evidence to support the causal relationship between CRP levels and T1D.

Biological sciences/Genetics

Health sciences/Endocrinology

Health sciences/Health care

Type 1 diabetes

C-reactive protein

causality

Mendelian randomization

bidirectional

Type 1 diabetes (T1D) is a prevalent autoimmune disease, principally involving irreversible destruction of insulin-producing islet β-cell. Its onset precedes in childhood that develops throughout the life span, with elevated incidences¹. The vast majority of individuals with T1D are long-lived with the good quality of life, but even under modern medical administration, the increased morbidity and mortality appear inevitably. Furthermore, the lack of a current means to prevent T1D; exogenous application of insulin, to date, represents the only option for the treatment of T1D^2,3. Genome-wide association studies (GWASs) have revealed the typical complex (polygenic) trait of T1D which garners increasing amounts of considerable biological interest^4–8.

An association between inflammatory exposure and immune response may be a potential mechanism to trigger β-cell destruction in the initiation of T1D⁹. The crosstalk between β-cells and immune cells, an active process, contributes to increased inflammatory response at early stages that results in cytokine-inducible β-cell dysfunction. Subsequently, the apoptosis of β-cell occurs due to autoreactive T cells¹⁰. Homeostasis model assessment of β-cell function index (HOMA-β) is an accepted gauge to assess β-cell function, and its impairment leading to an absolute deficiency of insulin is a vital determinant of T1D.

C-reactive protein (CRP), an acute phase reactant, is characterized as an independent inflammation marker that reflects the systemic inflammatory status¹¹. CRP may be involved in the initiation and development of T1D; however, the outcomes from existing research are inconsistent. A small observational study in 135 Spanish children shows a elevation in the CRP levels among those who progressed to T1D compared to controls¹². Cross-sectional epidemiological studies indicated that individuals with T1D displayed a higher circulating level of CRP, associated with higher exposure to the risk of cardiovascular disease^13–15. However, this differential effects are absent among 284 adolescents with or without T1D¹⁶. Additionally, none of those studies take their linkages into consideration in the prospective cohort design, as well as the confounding factors and reverse causality. Together, the evidence for causality between CRP and T1D remains inconclusive. Identifying the impact of CRP levels on T1D risk through randomized controlled trials requires a large sample, long follow-up times, and enormous costs, thus based upon which further investigations present great challenges.

Mendelian randomization (MR) sheds a reliable path on the nature of the relationship between CRP levels and T1D risk. It utilizes genetic variants as the instrumental variables (IVs) of exposure (i.e., CRP levels) to determine whether an observational association between this exposure and clinical outcome (i.e., T1D) complies with genetic epidemiology for being considered causal. MR studies can insulate the observational association from confounding factors and reverse causation bias, attributing to randomly assorted genetic variants originating at conception. In this study, we sought to determine dominant single nucleotide polymorphisms (SNPs) in the CRP levels, and to explore their association with T1D risk within published GWAS datasets by using a bidirectional two-sample MR framework^8,17.

Assumptions of MR

MR study relies on three stringent assumptions to access the accurate implementation of instrumental variable (IV) analysis with genetic. First, the selected IVs and the exposure are required for close interlink. Second, the exposure predicted by SNPs should affect the outcome only through its impact that assumption satisfies the exclusion-restriction to eliminate horizontal pleiotropic effects rather than other pathways. Third, IVs should vary independently of confounders. Figure 1 here embodies the basic idea of profiling the impact of CRP levels on T1D risk.

Data On Circulating Amounts Of Crp

Selection of circulating CRP levels-associated SNPs was sourced from a large-scale meta-analysis of GWAS built above two large-scale available GWASs to characterize loci associated with CRP levels, derived from 88 studies in 204,402 participants of European ancestry after adjusting for age, sex, body mass index, and population substructure¹⁷ (https://gwas.mrcieu.ac.uk/datasets/ieu-b-35/).

Data On T1d

Summary data was drawn from the largest publicly available T1D GWAS, covering 138,229 SNPs ⁸ (https://www.ebi.ac.uk/gwas/studies/GCST005536), with imputation to the 1000 Genomes in 6,808 T1D cases and 12,173 controls in people of European descent. The case samples of T1D were sourced from UK Genetic Resource Investigating Diabetes (UKGRID) cohort. A total of controls were ascertained from the British 1958 Birth Cohort¹⁸, the UK National Blood Services collection¹⁹, and the NIHR Cambridge Biomedical Research Centre Cambridge BioResource²⁰. The corresponding information of the study has been detailed elsewhere⁸.

Generation Of Genetic Instruments

We introduced a bidirectional two-sample MR framework to investigate the causality between circulating CRP levels and T1D. SNPs associated with CRP levels were defined as IVs. All selected SNPs is required to fulfill following conditions: 1. all SNPs should bear on the premise of genome-wide significance (P < 5×10^− 8); 2. we constrained IVs (r² < 0.01) for the exclusion of SNPs with strong LD. In this study, we counted the F-statistic as the strength of the IV-exposure correlation. F-statistic was obtained by using an approximation²¹. Upon setting T1D as exposure, all SNPs were defined as IVs should meet the conditions which were identical to those described above.

Statistical analysis

The analyses of reconciled datasets between the exposure and outcome were performed with R software version 4.1.2 using the TwoSampleMR package. We conducted the Two-sample MR analysis with the inverse-variance weighted (IVW) method to acquire estimates between ‘SNPs-CRP’ and ‘SNPs-T1D’ in the primary investigation of the causative relationship of CRP on T1D. In brief, IVW represents a conventional approach facilitating the most accurate assessment owing to its inclusion of all selected SNPs in the analysis are valid IVs²². The heterogeneity unfolded from each SNP was assessed by Cochran’s Q statistic²³. A fixed-effects model was applied in the absence of heterogeneity, otherwise, the random-effects model results in a more conservative validation²⁴. Additionally, the simple median method and weighted median method were available as supplemental for IVW²⁵.

In sensitivity analyses, MR-Egger regression and MR-Pleiotropy RESidual Sum and Outlier (MR-PRESSO) were followed to evaluate the robustness of the causal estimates by exploring bias resulting from horizontal pleiotropic effects. MR-Egger regression is disposed to regression dilution bias, in which the intercept term provided a credible interpretation in the estimate of average pleiotropy across all SNPs²⁶. MR-PRESSO, a routine procedure of outlier detection, was also assessed to correct for horizontal pleiotropy²⁷. It is composed of three parts, including MR-PRESSO global test, MR-PRESSO outlier test, and MR-PRESSO distortion test. We adapted MR-PRESSO goal test for assessment of the overall horizontal pleiotropy. SNPs with P less than 0.05 counted as outlier instruments were excluded from further analysis. The SNP removal step was repeated until nonsignificance remained in the global test. We manually searched genome annotation with PhenoScanner for association with each SNP phenotyping (https://www.phenoscanner.medschl.cam.ac.uk/phenoscanner). Additionally, we adopted the funnel plot to depict individual causal effect estimates for each SNP. For the acquisition of the potential impact of SNPs on causal estimates, we applied the leave-one-out permutation validation to exclude each SNP one by one. The results of anterior-posterior fluctuation by culling SNPs are the embodiment of a stable association. Two-tailed tests are considered statistically significant at a prespecified threshold of P < 0.05.

Initially, we performed MR analysis upon setting CRP as exposure to estimate the genetic predicted association of the CRP variable with T1D risk. For use as IVs, we extracted 11 independent SNPs (Supplementary Table S1). Weak IV bias was avoided due to the F statistics of all SNPs being greater than 10. Then, we manually searched these SNPs through PhenoScanner database and removed 3 SNPs (i.e., rs1386821 and rs4129267: autoimmunity-related trait located in IL-6R; rs2852151: T1D-associated genetic variant located in PTPN2) owing to violating IV assumptions^28–30. Additionally, rs1490384 and rs4420638 were removed due to failing the global test. The MR estimates with remaining SNPs exhibited nonsignificant heterogeneity within Cochran’s Q test for the association of CRP levels on T1D (I² = 0.438, P = 0.413). Therefore, we used the fixed-effects IVW method providing evidence, indicating an absence of significant genetic correlation of CRP levels on T1D, with OR of 0.922 per natural log-transformed CRP unit, 95% confidence interval (CI) 0.662 to 1.285 (P = 0.631) (Fig. 2). Other analyses including simple median method (OR: 1.190, 95% CI: 0.713–1.987, P = 0.505), weighted median method (OR: 1.114, 95% CI: 0.713–1.741, P = 0.635), and MR-Egger (OR: 0.730, 95% CI: 0.364–1.463, P = 0.425) demonstrated little association of CRP levels with T1D. Moreover, there was no indication for potential directional pleiotropy effects, supported by the nonsignificant intercept (intercept = 0.015, P = 0.490) (Supplementary Table S3). As demonstrated from the results of the leave-one-out analysis, no single SNP could result in the deviation of estimates of CRP on T1D risk (Supplementary Figure S1), suggesting that the results could be reliable. The funnel plot appeared asymmetric due to too few SNPs (Supplementary Figure S2). Forest plots and scatter plots for CRP are shown in Supplementary Figure S3.

Next, we denoted T1D as exposure and CRP levels as the outcome to further validate the causality of T1D on CRP levels and then enrolled 18 SNPs with significance (P < 5×10^− 8) for T1D in this MR analysis (Supplementary Table S2). Among these 18 SNPs, 3 SNPs (i.e., rs1456988, rs2611215, rs9585056) with F statistics less than 10 were culled, and rs516246 which failed the global test was excluded from the analysis. The MR estimates among remaining SNPs exhibited no significant heterogeneity within Cochran’s Q test for the association of T1D on CRP levels (I² = -15.658, P = 0.591). Figure 2 showed that T1D risk was not correlated with CRP levels by employing the fixed-effects IVW method (OR: 1.000, 95% CI: 0.990–1.010, P = 0.930). Again, no suggestive of significant associations of T1D on CRP levels was provided from simple median method (OR: 1.010, 95% CI: 0.990–1.020, P = 0.480), weighted median method (OR: 1.010, 95% CI: 0.990–1.020, P = 0.420) and MR-Egger (OR: 1.010, 95% CI: 0.980–1.030, P = 0.580). No significantly horizontal across estimates of included SNPs was observed (intercept = -0.001, P = 0.565) (Supplementary Table S3). The leave-one-out analysis also remained the stable estimate of T1D risk on CRP levels after omitting one single SNP at a time (Supplementary Figure S4). The funnel plot showed a symmetrical distribution implying balanced pleiotropy (Supplementary Figure S5). The forest plot and scatter plot for T1D are shown in Supplementary Figure S6.

In this current research, we performed a bidirectional two-sample MR to determine whether the causal association between CRP levels and T1D. No indication supported a causal correlation of CRP levels on T1D risk. Besides, the genetic predisposition of T1D was independent of CRP levels. These findings failed to provide evidence for a substantial role for CRP in the pathogenic pathways of T1D.

Observational studies intended to elucidate the clinical relevance of CRP levels in T1D have given rise to discrepant results^12–16. In observational studies such as these, the emergence of bias arises inevitably owing to various confounders that could affect observational studies. A previous derivation drew no causal association between CRP levels and diabetes by using MR analysis³¹. In that study, 3 SNPs were included as IVs in the pooled analysis, leaning on the ‘SNPs-CRP’ association derived from 5,274 individuals and the ‘SNPs-diabetes’ association observed among 1,923 cases and 2,932 controls. It is worth noting that multiple types of diabetes, including type 2 diabetes (T2D), T1D, and others, were placed within the case collection. Meanwhile, the complicated mechanism of T2D incidence involves multiple phenotypic factors. Thus, the association comes from the broad inclusion criteria that might be not readily available for clinical application. In contrast, the current study incorporated more participants from both exposure and outcome datasets and targeted T1D only, of which the corresponding conclusion appeared more reliable.

In autoimmune scenery, the etiology of T1D remains still poorly elucidated involving cross interactions of multiple genetic loci and environmental impacts. T cell-mediated attacks targeting β-cells, once initiated, trigger uncontrolled pro-inflammatory reactions or spontaneous definition of inflammation driving metabolism to dysregulation³². A proteomic analysis from a cohort study including 50 T1D subjects and 100 controls, arrives at the consistent conclusion that dysregulated innate immune responses contribute to the development of T1D³³. Innate immune responses exert their effects through the controlled release of soluble mediators including CRP³⁴. The inflammation in T1D is confined to islet and draining lymph nodes, of which high local concentrations of inflammatory products are diluted excessively once in the peripheral circulation³⁵. It may contribute to inconsistent correlations between CRP and T1D from clinical trials.

CRP, a cyclic pentameric protein, elevates in response to inflammation progression. Its gene expression occurring mainly in hepatocytes is regulated by interleukin-6 (IL-6) derived from adipocytes and immune cells³⁶. In combination with our current results, we speculate that there is a mediator closer to the onset of the inflammatory cascade in T1D compared to CRP. IL-6 contributes to the augmentation of the inflammatory process and β-cell apoptosis³⁷, and its inhibition contributes to the remission of autoimmune diabetes³⁸. Moreover, in our initial extraction of SNPs, culled SNPs (i.e., rs1386821 and rs4129267) are located in IL-6R genetic loci which is a distinct causal variant for multiple autoimmune diseases encompassing T1D^28,29. Therefore, IL-6 might be a practical alternative to explore the correlation between inflammation and T1D, while it is difficult to obtain open access to IL-6 databases.

In addition, gender differences may underlie the impacts of genetically amplified CRP levels on the pathophysiology of T1D in this case. A previous study found that the CRP levels of female T1D patients are higher than controls while there is no distinction of CRP levels among overall participants³⁹. This difference may be attributable to the confounders (i,e. obesity and leptin) that give rise to the independence of CRP in females rather than in males^40,41. However, we failed to rule out the possibility that gender differences could interfere with the causal association of CRP levels on T1D risk, owing to the constraints of databases.

Our MR approach possessing several strengths over observational study weakens the possibility of bias inherent to observational studies discernibly²⁷. First, since the T1D stages may be compartmentalized according to preclinic, eliminating the reverse causal effect could support to assess the role of CRP preceding the pathological changes to the pancreatic β-cells. Second, we sought to determine the cumulative lifetime risk of T1D due to genetical CRP level; however, single CRP measurements would seem unlikely to predict accurately a disease manifested in the later stages of life. Last, the two-sample MR approach was employed for direct evaluation of the role of CRP based on the large cohort of T1D encompassing 6,808 cases and 12,173 controls. Statistical power exerted by this analysis coincides with an approach building on individual-level data²²; however, clinical cohorts rarely enable to reach such desired sample sizes.

Our study also has intrinsic limitations. This analysis is vulnerable to the pleiotropic effects of genetic instruments, which may access relatively confusing estimates and potentially lead to causalities accompanying bias⁴². Despite multiple steps taken for assessing pleiotropy, residual bias may still be present. Even under the conditions in the acquisition of improper pleiotropic description from MR Egger, therefore we used two additional pleiotropy robust MR methods following the elimination of SNPs with pleiotropic effects in sensitivity analyses, thus obtaining consistent results which are reassuring. The selected GWAS databases in our study enroll on European ancestry only, our findings could not be generalized to other descents, despite potential consistency with different populations. Meanwhile, we could not rule out the potential impact of gender differences in the association bewteen CRP levels and T1D, due to the restriction of publicly available database.

In summary, our results shed light on no significant impact of CRP levels dominated by genes on T1D risk, and this important insight facilitates the understanding of the complicated pathophysiology of T1D. As such, the role of CRP participating in the progression and occurrence of T1D needs to be established by a large randomized controlled trial, or further MR corroboration based upon a larger scaled GWAS database.

Author Contributions:

QH came up with the idea and contributed to the design of this study. FT and HS have conducted the main analysis and drafted the manuscript. All authors have revised the manuscript and approved thefinal version.

Funding:

This work was supported by the National Key Research and Development Program of China (2018YFC2000200), Traditional Chinese Medicine Science and Technology Plan of Zhejiang Province (2020ZZ009), the Huangqi Famous Traditional Chinese Medicine Doctor Inherit Workstation Project of Zhejiang Province of China (GZS2020021), the National Natural Science Foundation of China (82174131), Zhejiang Provincial Natural Science Foundation of China (LY21H030002), and the Medical Health Science and Technology Project of Zhejiang Provincial Health Commission (2019RC229).

Acknowledgments:

We thank all the researchers involved in this MR study. We also thank all the investigators for sharing these data.

Data availability:

The datasets come from large-scale genome-wide association studies (i,e. C-reactive protein and Type 1 diabetes) and are publicly available. Analysis code and intermediate results are available from the corresponding author on reasonable request.

Lind, M. et al. HbA1c level as a risk factor for retinopathy and nephropathy in children and adults with type 1 diabetes: Swedish population based cohort study. BMJ 366, l4894, doi:10.1136/bmj.l4894 (2019).
Magliano, D. J., Harding, J. L. & Shaw, J. E. Glycemic control and excess mortality in type 1 diabetes. N Engl J Med 372, 880, doi:10.1056/NEJMc1415677 (2015).
Rawshani, A. et al. Excess mortality and cardiovascular disease in young adults with type 1 diabetes in relation to age at onset: a nationwide, register-based cohort study. Lancet 392, 477-486, doi:10.1016/S0140-6736(18)31506-X (2018).
Qu, H. Q. et al. Genetic architecture of type 1 diabetes with low genetic risk score informed by 41 unreported loci. Commun Biol 4, 908, doi:10.1038/s42003-021-02368-8 (2021).
Robertson, C. C. et al. Fine-mapping, trans-ancestral and genomic analyses identify causal variants, cells, genes and drug targets for type 1 diabetes. Nat Genet 53, 962-971, doi:10.1038/s41588-021-00880-5 (2021).
Zhu, M. et al. Identification of Novel T1D Risk Loci and Their Association With Age and Islet Function at Diagnosis in Autoantibody-Positive T1D Individuals: Based on a Two-Stage Genome-Wide Association Study. Diabetes Care 42, 1414-1421, doi:10.2337/dc18-2023 (2019).
Chiou, J. et al. Interpreting type 1 diabetes risk with genetics and single-cell epigenomics. Nature 594, 398-402, doi:10.1038/s41586-021-03552-w (2021).
Onengut-Gumuscu, S. et al. Fine mapping of type 1 diabetes susceptibility loci and evidence for colocalization of causal variants with lymphoid gene enhancers. Nat Genet 47, 381-386, doi:10.1038/ng.3245 (2015).
Kulkarni, A. et al. 12-Lipoxygenase governs the innate immune pathogenesis of islet inflammation and autoimmune diabetes. JCI Insight 6, doi:10.1172/jci.insight.147812 (2021).
Eizirik, D. L., Colli, M. L. & Ortis, F. The role of inflammation in insulitis and beta-cell loss in type 1 diabetes. Nat Rev Endocrinol 5, 219-226, doi:10.1038/nrendo.2009.21 (2009).
Koenig, W. High-sensitivity C-reactive protein and atherosclerotic disease: from improved risk prediction to risk-guided therapy. Int J Cardiol 168, 5126-5134, doi:10.1016/j.ijcard.2013.07.113 (2013).
Perez-Segura, P. et al. Children with type 1 diabetes have elevated high-sensitivity C-reactive protein compared with a control group. BMJ Open Diabetes Res Care 8, doi:10.1136/bmjdrc-2020-001424 (2020).
Purohit, S. et al. Serum Levels of Inflammatory Proteins Are Associated With Peripheral Neuropathy in a Cross-Sectional Type-1 Diabetes Cohort. Front Immunol 12, 654233, doi:10.3389/fimmu.2021.654233 (2021).
Sousa, G. R. et al. Glycemic Control, Cardiac Autoimmunity, and Long-Term Risk of Cardiovascular Disease in Type 1 Diabetes Mellitus. Circulation 139, 730-743, doi:10.1161/CIRCULATIONAHA.118.036068 (2019).
Castelblanco, E. et al. Low-grade Inflammatory Marker Profile May Help to Differentiate Patients With LADA, Classic Adult-Onset Type 1 Diabetes, and Type 2 Diabetes. Diabetes Care 41, 862-868, doi:10.2337/dc17-1662 (2018).
Tommerdahl, K. L. et al. Impact of Obesity on Measures of Cardiovascular and Kidney Health in Youth With Type 1 Diabetes as Compared With Youth With Type 2 Diabetes. Diabetes Care 44, 795-803, doi:10.2337/dc20-1879 (2021).
Ligthart, S. et al. Genome Analyses of >200,000 Individuals Identify 58 Loci for Chronic Inflammation and Highlight Pathways that Link Inflammation and Complex Disorders. Am J Hum Genet 103, 691-706, doi:10.1016/j.ajhg.2018.09.009 (2018).
Power, C. & Elliott, J. Cohort profile: 1958 British birth cohort (National Child Development Study). Int J Epidemiol 35, 34-41, doi:10.1093/ije/dyi183 (2006).
Wellcome Trust Case Control, C. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature 447, 661-678, doi:10.1038/nature05911 (2007).
Dendrou, C. A. et al. Cell-specific protein phenotypes for the autoimmune locus IL2RA using a genotype-selectable human bioresource. Nat Genet 41, 1011-1015, doi:10.1038/ng.434 (2009).
Chen, H. et al. No Causal Association Between Adiponectin and the Risk of Rheumatoid Arthritis: A Mendelian Randomization Study. Front Genet 12, 670282, doi:10.3389/fgene.2021.670282 (2021).
Burgess, S., Butterworth, A. & Thompson, S. G. Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 37, 658-665, doi:10.1002/gepi.21758 (2013).
Greco, M. F., Minelli, C., Sheehan, N. A. & Thompson, J. R. Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome. Stat Med 34, 2926-2940, doi:10.1002/sim.6522 (2015).
Greenland, S. Quantitative methods in the review of epidemiologic literature. Epidemiol Rev 9, 1-30, doi:10.1093/oxfordjournals.epirev.a036298 (1987).
Bowden, J., Davey Smith, G., Haycock, P. C. & Burgess, S. Consistent Estimation in Mendelian Randomization with Some Invalid Instruments Using a Weighted Median Estimator. Genet Epidemiol 40, 304-314, doi:10.1002/gepi.21965 (2016).
Bowden, J., Davey Smith, G. & Burgess, S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. Int J Epidemiol 44, 512-525, doi:10.1093/ije/dyv080 (2015).
Verbanck, M., Chen, C. Y., Neale, B. & Do, R. Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 50, 693-698, doi:10.1038/s41588-018-0099-7 (2018).
Smyth, D. J. et al. Shared and distinct genetic variants in type 1 diabetes and celiac disease. N Engl J Med 359, 2767-2777, doi:10.1056/NEJMoa0807917 (2008).
Fortune, M. D. et al. Statistical colocalization of genetic risk variants for related autoimmune diseases in the context of common controls. Nat Genet 47, 839-846, doi:10.1038/ng.3330 (2015).
Hundhausen, C. et al. Enhanced T cell responses to IL-6 in type 1 diabetes are associated with early clinical disease and increased IL-6 receptor expression. Sci Transl Med 8, 356ra119, doi:10.1126/scitranslmed.aad9943 (2016).
Brunner, E. J. et al. Inflammation, insulin resistance, and diabetes--Mendelian randomization using CRP haplotypes points upstream. PLoS Med 5, e155, doi:10.1371/journal.pmed.0050155 (2008).
Pugliese, A. Autoreactive T cells in type 1 diabetes. J Clin Invest 127, 2881-2891, doi:10.1172/JCI94549 (2017).
Zhang, Q. et al. Serum proteomics reveals systemic dysregulation of innate immunity in type 1 diabetes. J Exp Med 210, 191-203, doi:10.1084/jem.20111843 (2013).
Toubi, E. & Vadasz, Z. Innate immune-responses and their role in driving autoimmunity. Autoimmun Rev 18, 306-311, doi:10.1016/j.autrev.2018.10.005 (2019).
Kaldunski, M. et al. Identification of a serum-induced transcriptional signature associated with type 1 diabetes in the BioBreeding rat. Diabetes 59, 2375-2385, doi:10.2337/db10-0372 (2010).
Del Giudice, M. & Gangestad, S. W. Rethinking IL-6 and CRP: Why they are more than inflammatory biomarkers, and why it matters. Brain Behav Immun 70, 61-75, doi:10.1016/j.bbi.2018.02.013 (2018).
Linnemann, A. K. et al. Interleukin 6 protects pancreatic beta cells from apoptosis by stimulation of autophagy. FASEB J 31, 4140-4152, doi:10.1096/fj.201700061RR (2017).
Jorns, A. et al. Remission of autoimmune diabetes by anti-TCR combination therapies with anti-IL-17A or/and anti-IL-6 in the IDDM rat model of type 1 diabetes. BMC Med 18, 33, doi:10.1186/s12916-020-1503-6 (2020).
Harms, R. Z. et al. Confirmation and Identification of Biomarkers Implicating Environmental Triggers in the Pathogenesis of Type 1 Diabetes. Front Immunol 11, 1922, doi:10.3389/fimmu.2020.01922 (2020).
Gaines, J. et al. Gender differences in the association of sleep apnea and inflammation. Brain Behav Immun 47, 211-217, doi:10.1016/j.bbi.2014.11.013 (2015).
Beasley, L. E. et al. Inflammation and race and gender differences in computerized tomography-measured adipose depots. Obesity (Silver Spring) 17, 1062-1069, doi:10.1038/oby.2008.627 (2009).
Burgess, S., Bowden, J., Fall, T., Ingelsson, E. & Thompson, S. G. Sensitivity Analyses for Robust Causal Inference from Mendelian Randomization Analyses with Multiple Genetic Variants. Epidemiology 28, 30-42, doi:10.1097/EDE.0000000000000559 (2017).

Table 1 Mendelian Randomisation estimates between CRP levels and T1D risk.
Full analysis with all SNPs (CRP on T1D)
Analysis	SNPs	Method	Odds ratio	95% CI	P-value
	11	MR-Egger	1.660	0.823, 3.327	0.191
	11	Inverse-variance weighted	1.015	0.606, 1.701	0.954
	11	Weighted median	1.126	0.929, 1.364	0.226

Excluding SNPs
Analysis	SNPs	Method	Odds ratio	95% CI	P-value
	6	Simple median	1.190	0.713, 1.987	0.505
	6	Weighted median	1.114	0.713,1.741	0.635
	6	MR-Egger	0.730	0.364, 1.463	0.425
	6	Inverse-variance weighted	0.922	0.662, 1.285	0.631
	6	MR-PRESSO	0.992	0.662, 1.285	0.652

Full analysis with all SNPs (T1D on CRP)
Analysis	SNPs	Method	Odds ratio	95% CI	P-value
	18	MR-Egger	1.004	0.975, 1.033	0.810
	18	Inverse-variance weighted	1.003	0.994, 1.013	0.498
	18	Weighted median	1.006	0.992, 1.020	0.436

Excluding SNPs
Analysis	SNPs	Method	Odds ratio	95% CI	P-value
	14	Simple median	1.010	0.990, 1.020	0.480
	14	Weighted median	1.010	0.990, 1.020	0.420
	14	MR-Egger	1.010	0.980, 1.030	0.580
	14	Inverse-variance weighted	1.000	0.990, 1.010	0.930
	14	MR-PRESSO	1.000	0.991, 1.010	0.931

No competing interests reported.

SupplementaryMaterials.docx

No Causal Association Between C-Reactive Protein and the Risk of Type 1 Diabetes: A Bidirectional Mendelian Randomization Study

Status:

Version 1

Abstract

Background

Methods

Results

Conclusion

Figures

Introduction

Materials And Methods

Assumptions of MR

Data On Circulating Amounts Of Crp

Data On T1d

Generation Of Genetic Instruments

Statistical analysis

Results

Discussion

Declarations

References

Table

Additional Declarations

Supplementary Files

Status:

Version 1