High-sensitivity c-reactive protein, low-grade systemic inflammation and type 2 diabetes mellitus: A two-sample Mendelian randomization study
Raymond Noordam1, Charlotte H Oudt1, Maxime M Bos1, Roelof AJ Smit1,2, Diana van Heemst1
1) Department of Internal Medicine, Section Gerontology and Geriatrics, Leiden University Medical Center, Leiden, the Netherlands
2) Department of Cardiology, Leiden University Medical Center, Leiden, the Netherlands
Address of correspondence Raymond Noordam PhD
Department of Internal Medicine, Section Gerontology and Geriatrics, Leiden University Medical Center,
PO Box 9600, 2300 RC Leiden, the Netherlands Email: r.noordam@lumc.nl
Fax: +31 (0) 71 526 66912
Running title: “Inflammation and type 2 diabetes mellitus”
Words abstract: 250
Words manuscript: 2,764
Number of tables: 3
Number of figures: 2
Number of supplementary tables: 2
Key words: C-Reactive Protein, inflammation, type 2 diabetes mellitus, Mendelian Randomization
Abstract
Background and aims: The role of inflammation in type 2 diabetes mellitus (T2D) remains unclear. We investigated the associations of high sensitivity c-reactive protein (hsCRP) concentration with T2D and glycemic traits using two-sample Mendelian Randomization.
Methods and Results: We used publically available summary-statistics data from genome-wide association studies on T2D (DIAGRAM: 12 171 cases; 56 862 controls) and glycemic traits (MAGIC: 46 186 participants without diabetes mellitus). We combined the effects of the genetic instrumental variables through inverse-variance weighting (IVW), and MR-Egger regression and weighted-median estimation as sensitivity analyses which take into account potential violations (e.g., directional pleiotropy) of the assumptions of instrumental variable analyses. Analyses were conducted using 15 known hsCRP genetic instruments among which 6 instruments are hsCRP specific and not involved in inflammatory processes beyond hsCRP
concentration regulation. Though we found no association between the combined effect of the genetic instrumental variables for hsCRP and T2D with IVW (odds ratio per 1 ln[hsCRP in mg/L]: 1.15; 95% confidence interval: 0.93, 1.42), we found associations for T2D with MR-Egger regression and weighted-median estimation (odds ratio with 95% confidence interval per 1 ln[hsCRP in mg/L], MR-Egger regression: 1.29; 1.08, 1.49; weighted-median estimator: 1.21; 1.02, 1.39). We found no association with T2D for the combination of hsCRP-specific genetic instruments nor did we found associations with glycemic traits in any of the analyses.
Conclusion: Evidence was provided for a potential causal association between hsCRP and T2D, but only after considering directional pleiotropy. However, hsCRP was not causally associated with glycemic traits.
Introduction
While many risk factors have been associated with an increased risk of developing type 2 diabetes mellitus (T2D) [1-3], the question remains whether such risk factors are causal. Mendelian Randomization (MR) analysis is a method to ascertain causality of associations using genetic variants as instrumental variables [4].
One of the risk factors that has frequently been associated in observational settings with an increased risk to develop T2D, is low-grade systemic inflammation [5, 6].
Notably, different inflammatory chemokines and cytokines (e.g., interleukin 6 [IL-6]
and tumor necrosis factor-α), have been identified as predictive markers for future T2D [7].
Accumulation of adipose tissue increases the expression of high sensitivity C- reactive protein (hsCRP) in the liver [8-11]. Higher concentrations of hsCRP have been consistently associated with an increased risk of developing T2D in multiple cohort studies [12]. Importantly, hsCRP is synthesized upon IL-6 stimulation [5], and is therefore frequently used in epidemiological studies as a proxy for a chronic low- grade systemic inflammatory state. However, three previously performed MR studies on the association between hsCRP and T2D showed mixed results as to whether the observational findings are causal [13-15]. However, two of these studies were based on genetic variation in the CRP gene only, and all studies addressing the association between genetically-determined hsCRP concentration and T2D generally contained a relatively low number of participants.
Multiple genetic variants for hsCRP blood concentrations have been identified in genome-wide association studies (GWAS) [16]. MR analyses can be conducted using publically available summary statistics datasets from already published genome- wide association studies. This strategy has the advantage of being highly efficient and generally contains a high number of study participants [17]. Within the present study,
we aimed to investigate whether hsCRP is causally associated with T2D and glycemic traits using publically available data. To provide evidence whether high hsCRP itself or the systemic inflammation that drives hsCRP expression is causally associated with any of the study outcomes, we performed sensitivity analyses with genetic instruments specific for hsCRP concentration rather than genetic instruments also involved in other inflammatory processes based on published literature.
Materials and Methods
Selection of the genetic instrumental variables
Genetic instrumental variables for hsCRP serum concentration for the present Mendelian randomization study were selected from a GWAS meta-analysis of Dehghan et. al. [16]. This study is currently the largest study examining genetic variants in relation to hsCRP concentration in European-ancestry cohorts. Within this GWAS meta-analysis, a total of 66 185 participants from fifteen population-based studies were included. We selected the lead single nucleotide polymorphisms for genome-wide significant loci (p-value < 5e-8) from the discovery analyses, which also attained a genome-wide association in the fixed-effects meta-analysis of the discovery and replication panels. A total of 15 independent genetic loci were identified and used in the present study as genetic instrumental variables. We additionally defined two groups of genetic instrumental variables, notably one group of genetic instrumental variables specifically affecting hsCRP serum concentration and one group of genetic instrumental variables that also affect other inflammatory processes in addition to the regulation of hsCRP concentration. For this, a literature search (MedLine; Google Scholar), using the common research terms “inflammation” and “inflammatory process” in combination with the gene name, was conducted independently by two researchers (RN and MMB) to allocate the 15 genetic instrumental variables to one of the two groups. In case of inconsistencies in allocation of the genetic instrumental variables, consensus was reached in work discussions between the two researchers.
Data sources on the study outcomes
For the present study, we used T2D and measures of glucose-insulin
homeostasis as outcomes. We used publically available summary statistics datasets of two large GWAS conducted by the DIAGRAM and MAGIC consortia, which
identified multiple genetic loci associated with T2D and glycemic traits (notably fasting glucose and insulin, the homeostatic model assessment for insulin resistance (HOMA-IR) and beta-cell function (HOMA-B), and HbA1c), respectively. These datasets contain the summary level meta-analysis data of these GWAS, which includes the per-allele beta estimates of the SNPs on the outcomes, as well as the accompanying standard errors and effect alleles. Similar to the GWAS meta-analysis on hsCRP, all participating cohorts in the meta-analyses were of European ancestry.
The stage-1 meta-analysis on T2D from the DIAGRAM consortium contained data from 12 different cohorts [18], which includes a total of 12 171 cases of T2D and 56 862 controls. T2D was defined as fasting glucose >6.9 mmol/L, treatment with glucose-lowering agents and/or diagnosis by a general practitioner or medical specialist. The meta-analyses on fasting glucose, fasting insulin, HOMA-IR and HOMA-B from the MAGIC consortium, comprised data from 21 different cohorts [20]. In these meta-analyses, 46 186 individuals without diabetes mellitus were included. Data on fasting insulin, HOMA-IR and HOMA-B were log-transformed to approximate a normal distribution. The meta-analysis on HbA1c from the MAGIC consortium comprised data from 23 different cohorts of European ancestry. After excluding participants with diabetes mellitus, the meta-analysis on HbA1c was conducted using data from 46 368 individuals [21].
Statistical analyses
For all individuals genetic instruments, we calculated the F-statistic of the individual genetic instruments as a measure of strength [22].
Methods for Mendelian randomization analyses based on summary-level data, which are different from the methods used when calculating per-participant genetic risk scores, have been described in more detail elsewhere [23-25]. For the Mendelian
Randomization analyses, we combined the 15 genetic instruments for hsCRP to obtain a genetically determined (causal) association between hsCRP concentration and the study outcomes. Consequently, all estimates for the analyses can be interpreted as the change in outcome per genetically determined 1 ln(hsCRP in mg/L) increase. The statistical power for the Mendelian randomization analyses for the combination of the 15 hsCRP genetic instruments on T2D was calculated using a publically available power calculator [26]. Considering an alpha of 0.05 and an explained variance of 5%
(from Dehghan et al [16]), we had a power of 0.81 to observe significant odds ratios of 1.12.
The unadjusted source codes of the statistical methods for summary-level data Mendelian randomization for R statistical computing [27], as provided online by the authors [28, 29], were used for the calculations of the combined effect of the
individual genetic instrumental variables on the study outcomes. Using inverse- variance weighting (IVW), which is analogous to pooling estimates from different studies in conventional meta-analyses, we weighted the combined estimate by the inverse of the variance of the causal per-allele effect on the study outcomes. However, the estimates from IVW could be biased from the inclusion of individual genetic instruments with pleiotropic effects. To take into account potential violations of the assumptions for two-sample Mendelian Randomization analyses (e.g., directional pleiotropy), we conducted two sensitivity analyses, notably MR-Egger regression [28]
and weighted-median estimator analyses [29]. With MR-Egger analyses, we were able to formally test the presence of directional pleiotropy [28]. Such sensitivity analyses have not been used before to address the causal association between hsCRP and T2D using summary-level data [15]. The I2gx was calculated to test the presence of bias from weak instruments in the MR-Egger analyses; an I2gx > 0.90 was considered sufficient for instruments in the MR-Egger analyses [30]. Additionally, these analyses
were repeated for the two separate groups of hsCRP genetic instruments (hsCRP- specific instrumental variables and immunological instrumental variables).
As a second step, we investigated the causal effect of the individual genetic instruments (per 1 mg/L ln[hsCRP] increase) on the study outcomes. This was performed by dividing the effect of the genetic instrument on the study outcome by the effect of the genetic instrument on ln[hsCRP]. These causal effects were presented in funnel plots, and in these funnel plots we explored whether any of the individual genetic instruments were associated with any of the study outcomes to identify potential pleiotropic genetic variants, which could bias the results of the Mendelian Randomization analyses. For these analyses, we corrected for multiple testing with Bonferroni based on the 15 independent hsCRP genetic instruments (α = 0.05 / 15 = 0.0033).
All results are presented as odds ratio (OR; in case of T2D) or beta estimate (in case of the glycemic traits) with a 95% confidence interval (95%CI) per 1 ln(hsCRP in mg/L) increase.
Results
Individual genetic instrumental variables for hsCRP
All considered genetic instrumental variables for hsCRP had a F-statistic above 10 (median: 61; range: 28.4 – 711), and had therefore sufficient strength for use as genetic instruments in the present study (Table 1). Of the different genetic
instrumental variables, we found no support in literature for 6 genes (notably CRP, LEPR, GCKR, SALL1, ASCL1, and PABPC4) describing any other immunological
involvement than hsCRP concentration. These genetic instruments were therefore considered to be specific for hsCRP concentration; the other genetic instruments for which support was found were considered as instruments potentially involved in general immunological processes in addition to drive hsCRP expression.
Mendelian Randomization analyses
Using IVW to combine the causal estimates of the individual genetic instruments (Table 2; Figure 1), we found no association between genetically
determined ln(hsCRP) and T2D (OR per 1 ln(hsCRP in mg/L) increase: 1.15; 95%CI:
0.93 – 1.42). However, in both sensitivity analyses, we found that a high genetically determined ln(hsCRP) was associated with a higher risk of T2D (MR-Egger: 1.29;
1.08 – 1.49; weighted-median estimator: 1.21; 1.02 – 1.39). Genetically determined ln(hsCRP) was not associated with any of the studied glycemic traits (e.g., fasting glucose and fasting insulin) in any of the used statistical analyses. None of the observed intercepts from the MR-Egger regression analyses deviated significantly from zero, hence, there was no evidence in these analyses for directional pleiotropy.
We found no or minimal association of the 6 combined genetic instruments specific for hsCRP concentration and T2D, dependent on the analysis method used for the analyses (Table 3). For example, using the inverse-variance weighing, we found
no association between genetically-determined hsCRP concentration and T2D (1.08;
0.78 – 1.48). Though none of the associations were statistically significant, similar results were specifically observed with weighted-median estimator analyses (1.12;
0.87 – 1.45), but somewhat more pronounced associations were observed MR-Egger regression analyses (1.26; 0.87 – 1.83). Using the genetic instruments not specific of hsCRP concentration, we found similar results as compared with the overall analysis comprising all genetic instruments using MR-Egger (1.30; 1.02 – 1.67) and weighted median estimator (1.66; 1.23 – 2.22; Supplementary Table 1).
Individual hsCRP genetic instruments in relation to study outcomes
After correction for multiple testing (Supplementary Table 2), a total of 3 genetic instruments were associated with one or more of the investigated study outcomes (notably rs1260326 in GCKR, rs4420638 in APOC1 and rs9987289 in PPP1R3B). Funnel plots, presenting the causal effect of the individual genetic
instruments on the outcomes (x-axis) and the inverse of their standard errors (y-axis) is presented in Figure 2. Of the different genetic instruments, the rs4420638-A variant in APOC1 (Figure 2A) was associated with higher risk of T2D (per-allele odds ratio (OR): 1.14; 95% confidence interval (CI): 1.09, 1.20). However, this genetic
instrument was not associated with any of the other study outcomes (Figures 2B-F).
In contrast, rs2794520-C in CRP, which was the strongest genetic instrument (F- statistic = 711), was not associated with a higher risk of T2D (per-allele OR: 1.03;
95%CI: 0.99, 1.06).
Discussion
In the present study, we assessed whether a high hsCRP circulation, as a reflection of high chronic systemic low-grade inflammation, is causally associated with T2D and/or glycemic traits using MR analyses. Within the present study, which made use of publically available summary statistics data of the DIAGRAM and MAGIC consortia, we found evidence that the association between elevated hsCRP and T2D risk is causal, but only in the MR-Egger and median-weighted estimator sensitivity analyses. In the IVW and median-weighted estimator analyses, these results attenuated when we restricted to the hsCRP genetic instrumental variables that
according to current literature solely contribute to hsCRP concentration and are not involved in other immunological processes (of any kind). These results suggest that hsCRP concentration, as a reflection of elevated systemic inflammation, could be a causal risk factor for T2D.
Based on the results from the Mendelian Randomization analyses, we found evidence that a high hsCRP blood concentration, as a reflection of elevated systemic inflammation, might be causally associated with T2D risk, although this statement is only supported by the sensitivity analyses MR-Egger and median-weighted estimator that take into account directional pleiotropy. However, when we restricted to genetic instruments specific for hsCRP concentration, we observed attenuated results for both IVW and median-weighted estimator analyses. Results from the MR-Egger regression analyses showed a more pronounced association, but this was not statistically
significant which could be the result of a limited statistical power due to a reduced number of genetic instruments. This observation should therefore be studied in more detail and confirmed in future studies. Previously, in observational studies, an 1.26 times increased risk for T2D was observed per 1 ln(mg/L) increase in hsCRP [12], which was similar to the causal estimate observed in the present study within
sensitivity analyses accounting for directional pleiotropy. No evidence was found that the CRP gene, which was used in the previous MR analyses on hsCRP and T2D [13, 14], was associated with any of the study outcomes.
In addition to the combined effect of the selected genetic instrumental variables on the study outcomes, we also found genetic variation in APOC1, GCKR and PPP1R3B to be associated with 1 or more of the investigated study outcomes, which might be of interest for follow-up studies. Genetic variation in APOC1, which encodes Apolipoprotein C1, was associated with the risk for T2D. Genetic variation in this gene has been associated with multiple phenotypes, including cardiovascular disease outcomes and longevity [31-33] as well as with an increased risk of diabetic nephropathy [34, 35]. In-vitro experiments showed that APOC1 has been related to LPS-induced tumor necrosis factor-α expression in macrophages [36]. Therefore, it could be hypothesized that APOC1 affects T2D risk through inflammation. Common genetic variation in GCKR, which encodes the glucokinase regulatory protein that is involved in glucose metabolism, was associated with lower fasting glucose and lower insulin resistance (as measured with the HOMA-IR). GCKR has been previously related to, among others, glucose-insulin traits [20], triglyceride concentration [37], and gallstone disease [38]. As the effect of GCKR on hsCRP and glycemic traits are opposite, variation in GCKR likely involves different biological mechanisms
indicative of biological pleiotropy [39]. PPP1R3B has been predominantly identified in GWAS on serum lipid concentrations [37]. Nevertheless, there is some evidence that PPR1R3B is involved in inflammatory and immune processes as well [40].
Therefore, we hypothesize the PPP1R3B might affect insulin resistance through inflammation.
The present study is the largest MR study addressing the causal association between hsCRP and T2D to date, and contains almost twice as much T2D cases,
originated from 4 additional cohorts [18, 19], than the sample used by the previous study addressing the causal association between hsCRP and T2D using summary-level data [15]. The present study has also a few limitations. Because we used available summary statistics datasets, stratified analyses or analyses adjusted for other covariates were not possible. Furthermore, the cohorts participating in the GWAS studies on hsCRP and glycemic traits overlapped. In case of weak genetic
instrumental variables, this might give results in the direction of the confounded associations [41]. However, since our genetic instruments were relatively strong (all F-statistics ≥ 28), this was not likely the case in the present study. In addition, the 15 genetic instrumental variables explained approximately 5% of the total variation in hsCRP [16], which might be low for the use as an instrumental variable. Nevertheless, additional efforts to identify more loci associated with hsCRP would give loci with smaller effect sizes, and smaller weights in the combined analyses. Also, our study samples comprised participants of European ancestry, which means that results could be different for analyses conducted in participants from other ancestries. And last, we assumed that the effect sizes on hsCRP are similar to the effect sizes with other markers of low-grade systemic inflammation, especially in case of the analysis restricted to instrumental variables associated with aspects of the immune response in addition to the regulation of hsCRP concentration. For this reason, additional efforts are required to estimate the precise effect sizes as well as to identify genetic variants associated with specific immunological markers to provide additional evidence.
In conclusion, the results of our study provide evidence for a possible causal association between hsCRP, as a measure of elevated systemic inflammation, and T2D risk, although this observation was only supported by the sensitivity analyses in which directional pleiotropy was considered. However, this observation was not supported by evidence for a causal association between low-grade systemic
inflammation and higher fasting glucose concentrations or increased insulin resistance.
Conflict of Interest Statement
All authors declare to have no conflict of interest.
Acknowledgments
We would like to thank the DIAGRAM and MAGIC consortia for making the summary statistics data files of their meta-analyses available. This study was funded by the European Commission funded project HUMAN (Health-2013-INNOVATION- 1-602757).
References
[1] Coope A, Torsoni AS, Velloso LA. MECHANISMS IN ENDOCRINOLOGY:
Metabolic and inflammatory pathways on the pathogenesis of type 2 diabetes.
European journal of endocrinology / European Federation of Endocrine Societies.
2016;174:R175-87.
[2] Hameed I, Masoodi SR, Mir SA, Nabi M, Ghazanfar K, Ganai BA. Type 2 diabetes mellitus: From a metabolic disorder to an inflammatory condition. World journal of diabetes. 2015;6:598-612.
[3] Liu H, Cao MM, Wang Y, Li LC, Zhu LB, Xie GY, et al. Endoplasmic reticulum stress is involved in the connection between inflammation and autophagy in type 2 diabetes. General and comparative endocrinology. 2015;210:124-9.
[4] Lawlor DA, Harbord RM, Sterne JA, Timpson N, Davey Smith G. Mendelian randomization: using genes as instruments for making causal inferences in epidemiology. Statistics in medicine. 2008;27:1133-63.
[5] Donath MY, Shoelson SE. Type 2 diabetes as an inflammatory disease. Nat Rev Immunol. 2011;11:98-107.
[6] Kohlgruber A, Lynch L. Adipose tissue inflammation in the pathogenesis of type 2 diabetes. Current diabetes reports. 2015;15:92.
[7] Spranger J, Kroke A, Mohlig M, Hoffmann K, Bergmann MM, Ristow M, et al.
Inflammatory cytokines and the risk to develop type 2 diabetes: results of the prospective population-based European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study. Diabetes. 2003;52:812-7.
[8] Hotamisligil GS, Shargill NS, Spiegelman BM. Adipose expression of tumor necrosis factor-alpha: direct role in obesity-linked insulin resistance. Science (New York, NY). 1993;259:87-91.
[9] Weisberg SP, McCann D, Desai M, Rosenbaum M, Leibel RL, Ferrante AW, Jr.
Obesity is associated with macrophage accumulation in adipose tissue. The Journal of clinical investigation. 2003;112:1796-808.
[10] Lumeng CN, DeYoung SM, Bodzin JL, Saltiel AR. Increased Inflammatory Properties of Adipose Tissue Macrophages Recruited During Diet-Induced Obesity.
Diabetes. 2007;56:16-23.
[11] Harford KA, Reynolds CM, McGillicuddy FC, Roche HM. Fats, inflammation and insulin resistance: insights to the role of macrophage and T-cell accumulation in adipose tissue. The Proceedings of the Nutrition Society. 2011;70:408-17.
[12] Wang X, Bao W, Liu J, Ouyang YY, Wang D, Rong S, et al. Inflammatory markers and risk of type 2 diabetes: a systematic review and meta-analysis. Diabetes Care. 2013;36:166-75.
[13] Brunner EJ, Kivimaki M, Witte DR, Lawlor DA, Davey Smith G, Cooper JA, et al. Inflammation, insulin resistance, and diabetes--Mendelian randomization using CRP haplotypes points upstream. PLoS Med. 2008;5:e155.
[14] Dehghan A, Kardys I, de Maat MP, Uitterlinden AG, Sijbrands EJ, Bootsma AH, et al. Genetic variation, C-reactive protein levels, and incidence of diabetes. Diabetes.
2007;56:872-8.
[15] Prins BP, Abbasi A, Wong A, Vaez A, Nolte I, Franceschini N, et al.
Investigating the Causal Relationship of C-Reactive Protein with 32 Complex Somatic and Psychiatric Outcomes: A Large-Scale Cross-Consortium Mendelian
Randomization Study. PLoS medicine. 2016;13:e1001976.
[16] Dehghan A, Dupuis J, Barbalic M, Bis JC, Eiriksdottir G, Lu C, et al. Meta- analysis of genome-wide association studies in >80 000 subjects identifies multiple loci for C-reactive protein levels. Circulation. 2011;123:731-8.
[17] Burgess S, Scott RA, Timpson NJ, Davey Smith G, Thompson SG, Consortium E-I. Using published data in Mendelian randomization: a blueprint for efficient identification of causal risk factors. European journal of epidemiology. 2015.
[18] Morris AP, Voight BF, Teslovich TM, Ferreira T, Segre AV, Steinthorsdottir V, et al. Large-scale association analysis provides insights into the genetic architecture and pathophysiology of type 2 diabetes. Nat Genet. 2012;44:981-90.
[19] Voight BF, Scott LJ, Steinthorsdottir V, Morris AP, Dina C, Welch RP, et al.
Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat Genet. 2010;42:579-89.
[20] Dupuis J, Langenberg C, Prokopenko I, Saxena R, Soranzo N, Jackson AU, et al.
New genetic loci implicated in fasting glucose homeostasis and their impact on type 2 diabetes risk. Nature genetics. 2010;42:105-16.
[21] Soranzo N, Sanna S, Wheeler E, Gieger C, Radke D, Dupuis J, et al. Common variants at 10 genomic loci influence hemoglobin A(1)(C) levels via glycemic and nonglycemic pathways. Diabetes. 2010;59:3229-39.
[22] Pierce BL, Ahsan H, Vanderweele TJ. Power and instrument strength
requirements for Mendelian randomization studies using multiple genetic variants.
International journal of epidemiology. 2011;40:740-52.
[23] Bos MM, Smit RA, Trompet S, van Heemst D, Noordam R. Thyroid signaling, insulin resistance and type 2 diabetes mellitus: a Mendelian randomization study. The Journal of clinical endocrinology and metabolism. 2017.
[24] Noordam R, Smit RA, Postmus I, Trompet S, van Heemst D. Assessment of causality between serum gamma-glutamyltransferase and type 2 diabetes mellitus using publicly available data: a Mendelian randomization study. International journal of epidemiology. 2016.
[25] Burgess S, Scott RA, Timpson NJ, Davey Smith G, Thompson SG. Using published data in Mendelian randomization: a blueprint for efficient identification of causal risk factors. European journal of epidemiology. 2015;30:543-52.
[26] Brion MJ, Shakhbazov K, Visscher PM. Calculating statistical power in Mendelian randomization studies. International journal of epidemiology.
2013;42:1497-501.
[27] R Core Team. A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2016.
[28] Bowden J, Davey Smith G, Burgess S. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression.
International journal of epidemiology. 2015;44:512-25.
[29] Bowden J, Davey Smith G, Haycock PC, Burgess S. Consistent Estimation in Mendelian Randomization with Some Invalid Instruments Using a Weighted Median Estimator. Genetic epidemiology. 2016;40:304-14.
[30] Bowden J, Del Greco MF, Minelli C, Davey Smith G, Sheehan NA, Thompson JR. Assessing the suitability of summary data for two-sample Mendelian
randomization analyses using MR-Egger regression: the role of the I2 statistic.
International journal of epidemiology. 2016.
[31] Allen NB, Lloyd-Jones D, Hwang SJ, Rasmussen-Torvik L, Fornage M,
Morrison AC, et al. Genetic loci associated with ideal cardiovascular health: A meta- analysis of genome-wide association studies. American heart journal. 2016;175:112- 20.
[32] Souza MB, Araujo GS, Costa IG, Oliveira JR. Combined Genome-Wide CSF Abeta-42's Associations and Simple Network Properties Highlight New Risk Factors for Alzheimer's Disease. Journal of molecular neuroscience : MN. 2016;58:120-8.
[33] Deelen J, Beekman M, Uh HW, Broer L, Ayers KL, Tan Q, et al. Genome-wide association meta-analysis of human longevity identifies a novel locus conferring survival beyond 90 years of age. Hum Mol Genet. 2014;23:4420-32.
[34] Mooyaart AL, Valk EJ, van Es LA, Bruijn JA, de Heer E, Freedman BI, et al.
Genetic associations in diabetic nephropathy: a meta-analysis. Diabetologia.
2011;54:544-53.
[35] Bouillet B, Gautier T, Blache D, Pais de Barros JP, Duvillard L, Petit JM, et al.
Glycation of apolipoprotein C1 impairs its CETP inhibitory property:
pathophysiological relevance in patients with type 1 and type 2 diabetes. Diabetes Care. 2014;37:1148-56.
[36] Westerterp M, Berbee JF, Pires NM, van Mierlo GJ, Kleemann R, Romijn JA, et al. Apolipoprotein C-I is crucially involved in lipopolysaccharide-induced
atherosclerosis development in apolipoprotein E-knockout mice. Circulation.
2007;116:2173-81.
[37] Global Lipids Genetics C, Willer CJ, Schmidt EM, Sengupta S, Peloso GM, Gustafsson S, et al. Discovery and refinement of loci associated with lipid levels. Nat Genet. 2013;45:1274-83.
[38] Joshi AD, Andersson C, Buch S, Stender S, Noordam R, Weng LC, et al. Four Susceptibility Loci for Gallstone Disease Identified in a Meta-analysis of Genome- wide Association Studies. Gastroenterology. 2016.
[39] Solovieff N, Cotsapas C, Lee PH, Purcell SM, Smoller JW. Pleiotropy in complex traits: challenges and strategies. Nat Rev Genet. 2013;14:483-95.
[40] Lu YC, Yao X, Li YF, El-Gamil M, Dudley ME, Yang JC, et al. Mutated
PPP1R3B is recognized by T cells used to treat a melanoma patient who experienced a durable complete tumor regression. J Immunol. 2013;190:6034-42.
[41] Pierce BL, Burgess S. Efficient design for Mendelian randomization studies:
subsample and 2-sample instrumental variable estimators. American journal of epidemiology. 2013;178:1177-84.
[42] von Wnuck Lipinski K, Sattler K, Peters S, Weske S, Keul P, Klump H, et al.
Hepatocyte Nuclear Factor 1A Is a Cell-Intrinsic Transcription Factor Required for B Cell Differentiation and Development in Mice. Journal of immunology.
2016;196:1655-65.
[43] Yamagata K, Oda N, Kaisaki PJ, Menzel S, Furuta H, Vaxillaire M, et al.
Mutations in the hepatocyte nuclear factor-1alpha gene in maturity-onset diabetes of the young (MODY3). Nature. 1996;384:455-8.
[44] Kang MJ, Jo SG, Kim DJ, Park JH. NLRP3 inflammasome mediates IL-1beta production in immune cells in response to Acinetobacter baumannii and contributes to pulmonary inflammation in mice. Immunology. 2016.
[45] Yamagata K, Furuta H, Oda N, Kaisaki PJ, Menzel S, Cox NJ, et al. Mutations in the hepatocyte nuclear factor-4alpha gene in maturity-onset diabetes of the young (MODY1). Nature. 1996;384:458-60.
[46] Shavva VS, Mogilenko DA, Dizhe EB, Oleinikova GN, Perevozchikov AP, Orlov SV. Hepatic nuclear factor 4alpha positively regulates complement C3 expression and does not interfere with TNFalpha-mediated stimulation of C3 expression in HepG2 cells. Gene. 2013;524:187-92.
[47] Dussault I, Fawcett D, Matthyssen A, Bader JA, Giguere V. Orphan nuclear receptor ROR alpha-deficient mice display the cerebellar defects of staggerer.
Mechanisms of development. 1998;70:147-53.
[48] Chi H, Bogwald J, Dalmo RA, Zhang W, Hu YH. Th17 master transcription factors RORalpha and RORgamma regulate the expression of IL-17C, IL-17D and IL- 17F in Cynoglossus semilaevis. Developmental and comparative immunology.
2016;55:169-78.
Legends to Figures
Figure 1. Scatter plot and causal effect of hsCRP on T2D. The X-axis presents the effect of the genetic instrument on log-transformed hsCRP concentration (in mg/L);
the Y-axis presents the causal effect of the genetic instrument on T2D. The solid regression line represents the IVW estimate; the dashed regression line represents the MR-Egger estimate.
Figure 2. Association between the causal effect of the individual genetic instruments with T2D and investigated glycemic traits. The X-axis presents the causal effect per 1 mg/L log(hsCRP); the Y-axis depicts the inverse of the standard error of the causal estimate of the genetic instrument. A) Type 2 diabetes (T2D) B) HbA1c C) fasting glucose level D) fasting insulin level E) Homeostatic model assessment for insulin resistance (HOMA-IR) F) Homeostatic model assessment for pancreatic beta cell function (HOMA-B). Gene names are noted to show statistically significant associations.
Table 1: Overview individual hsCRP genetic instrumental variables
Gene SNP Chromosome Effect size on
ln(hsCRP in mg/L)
F-statistic Involved in other immunological processes?
Examples of described immunological processes involved
CRP rs2794520 1 0.193 711 No N/A
APOC1 rs4420638 19 0.240 688 Yes LPS-induced tumor necrosis factor-α expression
in macrophages [36]
HNF1A rs1183910 12 0.152 617 Yes B-cell differentiation [42]; causing MODY [43]
LEPR rs4420065 1 0.111 324 No N/A
IL6R rs4129267 1 0.094 250 Yes Encodes for the Interleukin-6 receptor
GCKR rs1260326 2 0.089 207 No N/A
IL1F10 rs6734238 2 0.047 69 Yes Encodes for interleukin 1 theta
NLRP3 rs12239046 1 0.048 61 Yes Mediates IL-1β production [44]; NLRP3
inflammasome
SALL1 rs10521222 16 0.110 48 No N/A
ASCL1 rs10745954 12 0.043 42 No N/A
PABPC4 rs12037222 1 0.047 41 No N/A
PPP1R3B rs9987289 8 0.079 39 Yes Involved in tumor-infiltrated lymphocytes [40]
BCL7B rs13233571 7 0.054 36 Yes B-Cell Tumor Suppressor 7A
HNF4A rs1800961 20 0.120 34 Yes Causing MODY [45], involved in complement
system [46]
RORA rs340029 15 0.044 28 Yes Involved in regulation of different inflammatory
responses and lymphocyte development [47, 48]
Abbreviations: hsCRP, high sensitivity c-reactive protein; MODY, Maturity-onset diabetes of the young; SNP, single nucleotide polymorphism.
Genetic instrumental variables are ordered on instrumental strength (F-statistic). N/A denotes genes for which we did not find evidence in the literature that they influence immunological processes in addition to the regulation of CRP concentration.
