Genetically predict the association between 91 human blood cell perturbation phenotypes and IBD: A Mendelian randomization study Authors: Xin Yao, Song Wang, Xiao Li, Jieying Wen, Yunsi Huang, Weixuan Lan, Xuyu Huang, Hao Li, Yunlong Sun, Xiaoqian Zhao, Tao Zhang Published: 2024 ABSTRACT Inflammatory bowel disease (IBD), encompassing Crohn disease and ulcerative colitis, is a group of persistent and recurrent gastrointestinal disorders. Despite the prevalence of these conditions, no studies have been conducted to examine the connection between altered human blood cell phenotypes and the underlying mechanisms of IBD pathogenesis. By utilizing summary statistics from genome-wide association studies, we executed a systematic two-sample Mendelian randomization (MR) investigation on 91 genetically determined blood cell perturbation traits in relation to 3 separate IBD phenotypes. Our analysis sought to delineate the putative causal links between these blood cell perturbation phenotypes and IBD, thereby contributing to a more nuanced comprehension of the pathophysiological underpinnings and offering a foundation for the development of novel therapeutic approaches. The forward MR analysis identified 7 human blood cell perturbation phenotypes associated with various IBD outcomes, while the reverse MR analysis revealed that 9 human blood cell perturbation phenotypes were influenced by various IBD phenotypes. The study has uncovered human blood cell perturbation phenotypes associated with various IBD diseases, contributing to a deeper understanding of the pathogenesis of IBD. It also provides new insights for early clinical diagnosis, disease activity monitoring, immune surveillance, prognosis assessment, and personalized treatment. 1. Introduction The perturbation phenotype of human blood cells represents a novel hematological measurement, distinct from previous studies focusing solely on the quantity, volume, and distribution of blood cells. By subjecting whole blood or specific blood cells in vitro to various stressors or stimulatory compounds that mimic in vivo conditions, we can observe the differential behavioral and functional responses of blood cells under different stimuli. These differences not only uncover intracellular biological changes, including gene expression, signaling pathway activation, and metabolic alterations but also aid in the discovery of new biological biomarkers and targets associated with diseases. Due to the partial conservation of biological pathways among various tissues and organs, identical blood cell perturbation phenotypes may be pertinent to a diverse array of tissues and organs. For instance, neutrophils, when disrupted by Alhydrogel for 21 hours, have been significantly associated with a range of cardiometabolic diseases, including heart failure, type 2 diabetes, and chronic kidney disease.[1] Inflammatory bowel disease (IBD), characterized by chronic inflammation of the digestive tract, primarily includes ulcerative colitis and Crohn disease. The exact etiology of IBD remains unresolved; however, it is widely accepted that the condition is linked to a complex interplay of genetic predisposition, environmental triggers, and immune system dysfunctions.[2] The impact of IBD is pronounced in developed countries,[3] and its incidence is escalating annually in developing nations.[4] This acceleration may be correlated with the changing dynamics of factors such as population density, urbanization, dietary patterns, and lifestyle habits, which are concurrent with the rise of emerging industrialization.[5] Mendelian randomization (MR)[6] represents an analytical technique that exploits the random allocation of genetic variants during meiosis as instrumental variables (IVs) to probe and infer causal relationships between exposure risk factors and disease endpoints. This method is particularly advantageous over conventional observational studies, as it is more adept at circumventing the pitfalls of reverse causation and the biases introduced by uncontrolled confounders. We intend to apply MR to systematically exclude the impacts of both recognized and occult confounders, thereby revealing the genetic-level causal links between altered phenotypes of human blood cells and IBD. 2. Materials and methods 2.1. Study design In this investigation, we utilize single nucleotide polymorphisms (SNPs) as IVs to perform a bidirectional two-sample MR analysis, examining the association between the disrupted phenotypes of human blood cells and IBD. The IVs selected for this analysis are required to meet 3 core assumptions: (1) they must robustly proxy an individual’s exposure status; (2) they must be independent of any confounders that could influence the outcome; and (3) they should not have a direct impact on the outcome unless this effect is mediated through the exposure. Figure 1 delineates the conceptual schema and the analytical workflow of the study. The conceptual schema and the analytical workflow of the study. IVW: inverse variance-weighted. 2.2. Data source The aggregated statistics for 91 perturbed phenotypes of human blood cells are derived from the study by Homilius, Max et al, which obtained dynamic gene expression phenotype data under various environmental conditions by batching perturbations of blood from 2685 individuals of European ancestry over different time intervals. To ensure rigorous analysis results, the researchers conducted stringent quality control on the genome-wide phenotype data, using a significance threshold of P < 5 × 10‐8 for phenotype significance, setting the clustering analysis standard to linkage disequilibrium (LD) r2 > 0.50, and ensuring that the physical distance of clustering variants was <250 kb.[1] The data from this study were obtained with the consent of the participants and approval from the local ethics committee, and the publicly available dataset can be accessed from the genome-wide association studies catalog website. A name correspondence table for the 91 perturbed phenotypes of human blood cells is provided in Table S5, Supplemental Digital Content, http://links.lww.com/MD/O32. We also obtained a publicly available dataset related to IBD from the FinnGen Biobank.[7] All data have been approved through ethical review. The name correspondence and detailed information for IBD can be found in Table S6, Supplemental Digital Content, http://links.lww.com/MD/O32. 2.3. Instrumental variable selection The selection process of instrumental variables serves the 3 major assumptions of MR,[8,9] with SNPs closely associated with perturbed phenotypes in human blood cells and IBD showing genome-wide significance with P < 1 × 10−5. Furthermore, to ensure the independence among instrumental variables, we excluded SNPs with significant LD and those in close proximity to each other, using a threshold of r2 = 0.001 and a genetic distance of 10,000 kb for LD. Given that genetic variants need to exhibit a certain degree of association with the exposure, only genes with an F-statistic > 10 were considered as eligible instrumental variables.[10] The reliable SNPs identified through this rigorous selection process will be employed in subsequent MR analyses. 2.4. MR analysis We harmonize the allelic directions at the genetic level for exposure and outcome to ensure consistency in the directionality of the analysis results, with ambiguous palindromic SNPs and incompatible SNPs excluded. The inverse variance weighted (IVW)[11,12] method stands as the linchpin in the analysis of causal associations between exposure and outcome, calculating the effect of exposure on outcome by weighting all instrumental variables through regression analysis and averaging the results, yielding precise and statistically significant average causal effect estimates. IVW provides the most accurate causal association effect estimates only when all genetic variants are valid. The weighted median[13] method weights and ranks each instrumental variable to ensure that the results remain reliable even when invalid genetic variants are within an acceptable range. The MR-Egger[14,15] regression employs weights inversely proportional to the outcome variance, thereby constructing a robust model. Complementing these primary approaches, both the simple mode and weighted mode methodologies are utilized to enhance the analytical rigor and ensure the dependability of the findings. 2.5. Sensitivity analyses The intercept from the MR-Egger regression model is initially used to assess the horizontal pleiotropy of genetic variants. The MR-PRESSO[16] method further identifies and removes any factors that may lead to horizontal pleiotropy, ensuring the reliability of repeated analysis results after correcting for anomalous SNPs. If the result of Cochran Q test[17] reaches statistical significance, it suggests the presence of heterogeneity. The leave-one-out[18] method is employed to create forest plots by sequentially removing each SNP to observe outliers. Additionally, funnel plots are used to assess symmetry for sensitivity analysis, ensuring the stability of the analysis results. All analytical methods are performed using the TwoSampleMR package[19] (version 0.6.6) in R software (version 4.4.1). 3. Results 3.1. SNP selection We identified genetic instrumental variables that were strongly associated with 91 perturbed human blood cell phenotypes and 3 IBDs, with P-values all <1 × 10‐5, indicating a significant correlation between these SNPs and the exposure factors, thereby meeting the association assumption. We excluded SNPs with weak associations to the exposure, and only instrumental variables with an F-statistic >10 were deemed statistically significant. Palindromic SNPs and those with indeterminate directionality were ultimately discarded. After merging the exposure and outcome data, SNPs with consistent directionality were incorporated into the final analysis. We have meticulously listed the positive results from the bidirectional MR analysis, while statistically nonsignificant negative results were not recorded. The detailed findings are presented in Tables S1 and S2, Supplemental Digital Content, http://links.lww.com/MD/O32 and http://links.lww.com/MD/O32. 3.2. MR analysis results The forward MR analysis results indicate the following: neutrophil perturbation response (side fluorescence standard deviation of neutrophil-1 in response to butyric acid perturbation measured by white cell differential channel by fluorescence [WDF] dye) is a protective factor for IBD [odds ratio (OR) = 0.98, 95% confidence interval (CI) = 0.96 to 0.99, P-value = .012]. Neutrophil perturbation response (forward scatter median of neutrophil-1 in response to Pam3CSK4 perturbation measured by WDF dye) is positively associated with an increased risk of IBD [OR = 1.06 (1.01–1.11), P-value = .009]. Platelet perturbation response (forward scatter standard deviation of platelet in response to ciproflaxin perturbation measured by platelet-F dye) is a risk factor for ulcerative colitis (UC) [OR = 1.07 (1.00–1.13), P-value = 0.034]. Neutrophil perturbation response (forward scatter coefficient of variation of neutrophil-4 in response to Pam3CSK4 perturbation measured by WDF dye) is a common protective factor for both IBD [OR = 0.93 (0.88–0.99), P-value = .018] and UC [OR = 0.91 (0.86–0.97), P-value = .003]. Neutrophil perturbation response (neutrophil-2/neutrophil-4 ratio in response to water perturbation measured by WDF dye) is a common protective factor for both IBD [OR = 0.95 (0.90–0.99), P-value = .014] and UC [OR = 0.92 (0.87–0.97), P-value = .003]. Platelet perturbation response (side fluorescence coefficient of variation of platelet in response to chloroform (1 hour) perturbation measured by white count and nucleated red blood cells dye) is a common protective factor for both IBD [OR = 0.98 (0.97–0.99), P-value = .003] and UC [OR = 0.98 (0.96–0.99), P-value < .001]. Neutrophil perturbation response (forward scatter median of neutrophil-4 in response to Pam3CSK4 perturbation measured by WDF dye) is a common risk factor for both IBD [OR = 1.05 (1.00–1.10), P-value = .031] and UC [OR = 1.07 (1.01–1.12), P-value = .016]. Figure 2 illustrates the aggregated results derived from the IVW method. The MR-Egger regression analysis revealed no instrumental variables with intercepts significantly differing from zero, as evidenced by P-values exceeding 0.05. Furthermore, the MR-PRESSO analysis did not yield a significant global test P-value below 0.05, suggesting the absence of appreciable horizontal pleiotropy and the exclusion of outliers. The Cochran Q test failed to identify any SNPs with significant heterogeneity (P < .05). Detailed sensitivity analyses of positive results from the forward analysis are provided in Table S3, Supplemental Digital Content, http://links.lww.com/MD/O32. Visual representations of the data, including forest plots, funnel plots, leave-one-out analyses, and scatter plots, are presented in Figures S1–S4, Supplemental Digital Content, http://links.lww.com/MD/O25, http://links.lww.com/MD/O26, http://links.lww.com/MD/O27, http://links.lww.com/MD/O28, respectively. Aggregated results of the IVW method in the forward MR analysis. IVW: inverse variance-weighted. Reverse MR analysis revealed that the risk of IBD [OR = 0.83 (0.70–0.99), P-value = .035] and UC [OR = 0.82 (0.70–0.96), P-value = .013] was associated with decreased levels of Red blood cell perturbation response (side fluorescence median of RBC in response to LiCl perturbation measured by platelet-F dye). The development of UC was associated with decreased levels of white blood cell perturbation response (side scatter median of WBC 1 at baseline measured by platelet-F dye) [OR = 0.92 (0.85–1.00), P-value = .046]. It was also associated with decreased levels of platelet perturbation response (side fluorescence coefficient of variation of platelet in response to colchicine perturbation measured by reticulocyte dye) [OR = 0.92 (0.84–1.00), P-value = .045], increased levels of reticulocyte perturbation response (side fluorescence coefficient of variation of reticulocyte 1 in response to rotenone perturbation measured by reticulocyte dye) [OR = 1.12 (1.00–1.25), P-value = 0.047], decreased levels of Eosinophil perturbation response (side scatter median of eosinophil 1 at baseline measured by WDF dye) [OR = 0.92 (0.85–0.99), P-value = 0.032], increased levels of Lymphocyte perturbation response (side scatter standard deviation of lymphocyte in response to colchicine perturbation measured by WDF dye) [OR = 1.10 (1.02–1.20), P-value = .020], decreased levels of Neutrophil perturbation response (side scatter coefficient of variation of neutrophil-3 in response to nigericin perturbation measured by WDF dye) [OR = 0.81 (0.67–0.97), P-value = .024], and decreased levels of platelet perturbation response (forward scatter standard deviation of platelet in response to chloroform (8 hours) perturbation measured by white count and nucleated red blood cells dye) [OR = 0.64 (0.45–0.90), P-value = .010]. The development of Crohn disease was associated with increased levels of Neutrophil perturbation response (side fluorescence standard deviation of neutrophil-1 in response to butyric acid perturbation measured by WDF dye) [OR = 1.29 (1.05–1.58), P-value = .016]. Figure 3 delineates the consolidated findings obtained through the IVW methodology. The MR-PRESSO analysis yielded no indication of horizontal pleiotropy, and the Cochran Q test similarly did not detect any significant heterogeneity among the SNPs under consideration (P < .05). In the MR-Egger regression analysis, we noted the presence of weak pleiotropy. We conducted leave-one-out tests for pleiotropic outliers, but after sequentially removing each SNP for analysis, no significant outliers affecting the results were found. Sensitivity analyses of positive results in the reverse analysis are detailed in Table S4, Supplemental Digital Content, http://links.lww.com/MD/O32. Forest plots, funnel plots, leave-one-out sensitivity analyses, and scatter plots are depicted in Figures S5–S8, Supplemental Digital Content, http://links.lww.com/MD/O29, http://links.lww.com/MD/O30, http://links.lww.com/MD/O31, respectively. Aggregated results of the IVW method in the reverse MR analysis. IVW: inverse variance-weighted. 4. Discussion We are the first to use the bidirectional two-sample MR method to investigate the causal relationship between 91 human blood cell perturbation phenotypes and IBD. We have identified genetic correlations between these blood cell perturbation phenotypes and 3 different disease outcomes. Our findings suggest that within the context of a forward MR analysis, the responsiveness of neutrophil-1 to butyrate perturbation is positively correlated with a diminished risk of IBD. As a prevalent short-chain fatty acid, butyrate is synthesized by select gut microbiota through the fermentation process of consumed food, and it serves as a critical energy provider for the cells of the intestinal lining. It not only suppresses immune and inflammatory responses in gut cells by inhibiting NF-κB secretion[20] and modulating immune cell activity[21,22] but also plays a crucial role in maintaining the barrier function of the intestinal mucosa.[23] We found that the response of neutrophil-1 to Pam3CSK4 perturbation increases the risk of IBD, and the positive scatter median of neutrophil-4’s response to Pam3CSK4 perturbation is a risk factor for IBD and UC, whereas the scatter variability coefficient of this response is a protective factor. Pam3CSK4, a molecule that mimics bacterial lipoproteins, is involved in the activation of toll-like receptors 1/2,[24] inducing the expression of the pro-inflammatory cytokine NF-κB, thereby initiating and causing excessive inflammatory responses in the intestine.[25] However, experiments have shown that the absence of NF-κB expression can lead to the destruction of epithelial barrier function,[26,27] resulting in severe chronic intestinal inflammation. Thus, its role in the intestine is to maintain a dynamic balance of inflammation. This difficult-to-quantify personalized indicator may lead to inconsistencies in our research results. We also found that an increased response level of the ratio of neutrophil-2 to neutrophil-4 to water perturbation is a protective factor for IBD and UC. Unlike neutrophil-1 in a healthy baseline state, neutrophil-2 represents a subset of neutrophils undergoing apoptosis.[1] Upon stimulation with Pam3CSK4, neutrophil-1 demonstrates an elevated production of reactive oxygen species (ROS) and the formation of neutrophil extracellular traps (NETs).[28] Moreover, a sustained inflammatory state retards the apoptosis of neutrophil-1, which consequently results in a decline in the ratio of neutrophil-2 to neutrophil-4, a trend that is in concordance with our data analysis. Our results show that the response of platelets to chloroform (1 hour) perturbation is a protective factor for IBD and UC. Chloroform was once widely used for surgical anesthesia, but due to its toxicity and adverse effects on the liver and heart, it is now widely used in industry as a solvent for organic synthesis. Khobjai, Warachate, et al believe that platelets usually play a beneficial role in the early stages of the disease. The researchers employed a chloroform extract derived from plants to augment platelet adhesion activity. This approach may facilitate wound healing, as activated platelets are capable of secreting diverse cytokines and vascular endothelial growth factors. These factors are instrumental in promoting cell recruitment, tissue repair, and matrix reconstruction.[29] However, other studies have shown that platelets can exhibit a “bipolar” feature in inflammation. On one hand, they promote inflammatory signal transduction and further exacerbate intestinal inflammatory responses and tissue damage by interacting with white blood cells (such as neutrophils, monocytes) and releasing various inflammatory mediators (such as IL-1β, platelet factor 4, β-thromboglobulin, etc).[30,31] On the other hand, platelets can limit the infiltration of white blood cells to the site of inflammation in the early stage of the disease by expressing the enzyme HYAL2, which degrades hyaluronic acid on the surface of endothelial cells, reducing hyaluronic acid’s support for white cell adhesion. This mechanism has been verified in a dextran sodium sulfate-induced experimental colitis model.[32] Furthermore, platelets can prevent complications such as bleeding by restoring vascular integrity and sealing the sites of white cell exudation.[33] Our reverse MR analysis revealed that an elevated risk of IBD and UC is correlated with a diminished responsiveness of red blood cells to lithium chloride perturbation. Lithium chloride is an effective glycogen synthase kinase 3 (GSK3) inhibitor, which indirectly mimics the negative regulatory effect of the PI3K/Akt pathway on GSK3 by inhibiting its activity.[34] This reduces the positive feedback of GSK3 on the pro-apoptotic pathway, including preventing Bax conformational changes, reducing apoptosis, and maintaining the proliferation and differentiation capacity of erythroid progenitor cells. These actions collectively promote erythropoiesis and the survival of erythroid progenitor cells. In IBD patients, due to the combined effects of impaired intestinal absorption, chronic inflammation, and reduced iron utilization, the number and quality of red blood cells typically show a downward trend.[35–37] Our findings further indicate that an elevated risk of UC is linked to lower baseline levels of white blood cells and eosinophils. Extensive prior research has established a correlation between heightened eosinophil levels in the tissues and serum of UC patients and increased disease activity. Such patients typically experience suboptimal clinical outcomes and responses to therapeutic interventions. However, some studies have failed to confirm this positive correlation, and some reports have even shown a negative correlation, which may be attributed to factors such as the focus of the study not being on eosinophils, small sample sizes, and concurrent treatments (such as corticosteroid therapy).[38,39] For example, during the inactive phase, eosinophil activity persists at levels higher than those observed during active inflammation. This suggests that eosinophils may contribute to the resolution of inflammation and the repair of compromised intestinal tissue. Such a reparative function is analogous to the role eosinophils play in other conditions, including allergies and asthma,[40] but research on similar mechanisms is currently limited. Additionally, an elevated risk of UC correlates with a diminished neutrophil-3 responsiveness to nigericin-induced perturbation. Nigericin facilitates the efflux of potassium ions and the influx of calcium ions, which in turn activates the NLRP3 inflammasome.[41,42] This activation leads to the secretion of NETs, serving as a mechanism for antimicrobial defense. However, long-term exposure to pro-inflammatory stimuli and the release of NETs may instead lead to chronic inflammation and tissue damage. Our analysis results differ somewhat from previous studies, and the specific mechanisms of white blood cells, neutrophils, and eosinophils in the development of inflammation and their impact on different stages of UC still require further exploration. The increased risk of UC is associated with a reduced level of platelet response to chloroform perturbation (8 hours) and colchicine perturbation. Chloroform perturbation activates the adhesive function of platelets, which may limit the infiltration of leukocytes to the site of inflammation in the early stage of the disease. However, the transition of platelets from a pro-inflammatory to a resolution-promoting phenotype may be restricted during the conversion from acute to chronic inflammation, leading to a “fatigued” phenotype characterized by reduced platelet activation and aggregation.[33] Colchicine interferes with platelet morphology and activation by disrupting microtubulin and regulating the phosphorylation state of cytoskeletal proteins, thereby reducing platelet aggregation.[43,44] The changes in platelet aggregation response caused by these perturbations may be related to different stages of UC and the process of transition from acute to chronic inflammation. We also found that the increased risk of UC is associated with an increased level of response of reticulocytes to rotenone perturbation. Rotenone is a mitochondrial respiratory chain inhibitor that primarily blocks electron transport by inhibiting complex I (NADH dehydrogenase) activity, leading to increased ROS and superoxides, subsequent mitochondrial dysfunction and cell apoptosis.[45,46] As precursors of red blood cells, reticulocytes still retain some mitochondria, suggesting that rotenone may indirectly affect the generation, maturation, and function of reticulocytes through mitochondrial dysfunction and oxidative stress, which may be related to the mechanism of oxidative stress induced by persistent inflammation in ulcerative colitis. The increased risk of UC is associated with heightened lymphocytic response to colchicine disturbance. Colchicine effectively inhibits lymphocyte proliferation and growth through various mechanisms, including suppression of DNA synthesis, disruption of microtubules, alteration of cellular structure and function, and prevention of mitosis.[47] Lymphocyte changes during the progression of ulcerative colitis exhibit distinct phase-specific characteristics, with acute phases marked by the activation and proliferation of inflammatory cells, while chronic phases are characterized by the redistribution of lymphocyte subsets and polarization of immune responses.[48] A multicenter randomized controlled clinical trial demonstrated that elevated platelet-to-lymphocyte ratio levels are usually associated with more severe disease activity in UC patients,[49] closely related to the recruitment of lymphocytes to the intestines by chemokines. Our analysis indicates that the increased risk of Crohn disease is associated with a heightened neutrophil response to butyrate disturbance. Butyrate, one of the most studied short-chain fatty acids, has been shown to possess significant anti-inflammatory properties. Studies suggest that butyrate exerts its anti-inflammatory effects by inhibiting TNF-α release, enhancing mucosal barrier function, selectively modulating inflammatory factors, and suppressing the NF-κB signaling pathway. However, it has no significant effect on LPS-induced interleukin-8 (IL-8) release,[50] and the persistent release of IL-8 promotes neutrophil aggregation at the inflammation site, potentially leading to excessive inflammatory responses in chronic inflammatory diseases such as Crohn disease. In summary, the perturbed phenotypes of various blood cells and their level changes are of multifaceted significance in the monitoring of IBD, spanning from basic research to clinical applications. This not only aids in the deeper understanding of the pathogenesis of IBD but also provides new insights for early clinical diagnosis, disease activity monitoring, immune surveillance, prognostic assessment, and personalized treatment. Despite its contributions, our study is not without limitations. Firstly, in an effort to mitigate the impact of uncontrolled confounders on the analytical outcomes, we restricted our dataset to IBD patients of European descent. This approach, while reducing variability, may compromise the broader applicability of our results across different ethnic groups. Secondly, the potential influence of pleiotropy and heterogeneity cannot be entirely dismissed, which may affect the interpretation of our findings. Additionally, due to the constraints of the original data, we were unable to further explore the causal effect of factors such as age and gender. 5. Conclusions Our forward MR analysis results indicate that 7 perturbed phenotypes of human blood cells are associated with various IBD outcomes, while the reverse MR analysis results suggest that 9 perturbed phenotypes of human blood cells are influenced by different IBDs. These blood cell biomarkers elucidate the mechanisms of action of inflammation and immune microenvironments in various IBD diseases and also provide new insights for the development of potential therapeutic targets. Acknowledgments We would like to express our gratitude to the FinnGen Database for providing the online datasets, as well as to Homilius, Max et al for making their publicly available research data accessible. Author contributions Conceptualization: Xin Yao. Data curation: Song Wang. Formal analysis: Xiao Li. Investigation: Jieying Wen. Methodology: Weixuan Lan. Project administration: Hao Li. Resources: Yunsi Huang. Software: Yunlong Sun, Xiaoqian Zhao. Supervision: Xuyu Huang. Writing – original draft: Xin Yao. Writing – review & editing: Tao Zhang. Supplementary Material