Integration of Genetic and Immune Infiltration Insights Into Data Mining of Multiple Sclerosis Pathogenesis

doi:10.21203/rs.3.rs-642302/v1

Download PDF

Research Article

Integration of Genetic and Immune Infiltration Insights Into Data Mining of Multiple Sclerosis Pathogenesis

https://doi.org/10.21203/rs.3.rs-642302/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 26 Jun, 2022

Read the published version in Computational Intelligence and Neuroscience →

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Background: Multiple sclerosis (MS) is an immune-mediated demyelinating disease of the central nervous system. Pathogenesis is closely related to environment, genetic and immune system, but the underlying interactions are still not clearly elucidated. Our study aims to uncover the genetic basis and immune landscape of multiple sclerosis pathogenesis with bioinformatics.

Methods: In our study, gene matrix were retrieceed from gene expression database NCBI GEO. We then used bioinformatics to standardize the samples and obtain differential gene expressions (DEGs). We constructed a protein-protein interaction network with DEGs on the STRING website, and used Cytohubba plug-in and MCODE plug-in to to mine hub genes. Then we had a functional enrichment hub genes and DEGs. Meanwhile, we use CIBERSORTX algorithm to explore the characteristics of immune cells infiltration in brain tissues of multiple sclerosis, and did a Spearman correlation analysis between genes and immune cells. We also analyzed the correlation between genes and types of brain tissues with WGCNA method.

Results: We included a total of 90 samples from 2 datasets in the study, extracted 882 differential genes and 10 hub genes closely related to multiple sclerosis. Functional enrichment analysis suggested roles of immune response in multiple sclerosis. Besides, with CIBERSORTX algorithm we found brain tissues of MS contain a variety of infiltrating immune cells. Correlation analysis suggested that hub genes are highly relevant to chronic active white matter lesions and certain hub genes in our study may play a role in the activation of immune cells such as macrophages and natural killer cells.

Conclusions: Our study provides guidance for the study of genetic basis and immune infiltration mechanism of multiple sclerosis.

Translational Medicine

Multiple sclerosis

Genes

Signal transduction

Immunity

Computational biology

Multiple sclerosis (MS) is an immune-mediated demyelinating disease of the central nervous system¹. It is predominantly occurred in Caucasian and can cause severe neurological dysfunctions in advanced stages^2–3. The incidence of multiple sclerosis is closely related to environmental and genetic factors. Although genes play an important role in the pathogenesis of MS, the concordant rate among identical twins was 30%, and a meta-analysis estimated that genetic factors accounted for only 50% of the risk of MS^4–5, suggesting that non-genetic factors also greatly influence the susceptibility to the disease. Environmental risk factors including vitamin D deficiency, smoking, Epstein-Barr virus (EB virus) infection, etc⁶ can interact with genetic variations of MS and lead to immune dysregulation. However, the interactions between environment and genes are still under investigations.

Peripheral immune cells, especially B cells and T cells, play an important role in the pathogenesis of multiple sclerosis. In the early stages of multiple sclerosis, the innate immune system, dominated by macrophages, promotes the pro-inflammatory response of T and B cells, leading to tissue damage. Early microglial activation may also be one of the initial events in the development of multiple sclerosis. Activated microglia contribute to disease progression through the release of proinflammatory cytokines, free radicals and glutamate. In the progressive stage, pathological changes such as axonal degeneration, focal to diffuse white matter injury, microglial cell activation, diffuse lymphocyte and monocyte infiltration, and cortical involvement occur^7–8. On the one hand, immune cells are involved in promoting inflammatory response, promoting demyelination, axon damage and the formation of disease plaques; on the other hand, some immune cells also play the role of anti-inflammation, inhibiting disease progression by promoting tissue repair. The immune system alterations in multiple sclerosis patients is complex, and the contributions of immune cells to the pathology of MS has not been fully clarified⁸.

Studies of subtypes of infiltrating immune cells in brain of multiple sclerosis or experimental autoimmune encephalomyelitis (EAE) revealed that both Th1 and Th17 CD4 + T lymphocytes play an essential role in disease initiation after reactivation in situ. However, details of immune microenvironment of intracranial lesions in patients with multiple sclerosis has not been thoroughly explored. More and more studies focused on the role of certain immune cell types and their plasticity in interactions with cytokines, chemokines and brain compartments⁹.

As early as 20 years ago, sequencing techniques has arisen, and gene chips contain a large number of biological information¹⁰. Databases of gene expression, gene polymorphism, protein structure, and genetic map, etc were increasing. The large amounts of gene data brought up the need of bioinformatics¹¹. And we can extract, process and analyze gene chips data more effectively with bioinformatic tools. With emerging technique CIBERSORT, we can analyze the immune cells compositions of tissues based on gene chips¹².

In our study, we combined gene chips data extracted from existing multiple sclerosis samples, we used R language, STRING and Cytoscape to calculate hub genes, annotate and perform gene enrichment analysis. We use CIBERSORT algorithm to infer the proportions of peripheral immune cells in the brain lesions. And we also calculate correlations between hub genes and immune cells, and used WGCNA to further compare the internal gene differences in the dataset and analyze the correlations between hub genes and lesion sites. The purpose of our study was to explore susceptible genes and immune cells infiltration in brain lesion of multiple sclerosis, and to provide novel insights of disease prevention and therapy.

1. Differentially Expressed Genes (DEGs) Screening

Gene Expression Omnibus¹³ (GEO; https://www.ncbi.nlm. Sal) is an open genomics database. We downloaded human expression profile datasets of multiple sclerosis GSE108000 and GSE135511 from GEO database. We then divided the datasets into control group and MS group. And we used normalizeBetweenArrays function of the R language "limma" package to normalize the datasets and eliminate intra-group differences. We then combined GSE108000 and GSE135511 with "SVA" R package ¹⁴ to remove batches. We considered DEGs with p < 0.05 and |log2FC| > 0.5 to be statistically significant. With DEGs we got, we used R package "ggplot2" ¹⁵ to draw volcano plot to view the differences between up-regulated and down-regulated genes, and used pheatmap to view differences of genes between disease group and control group.

2. Construction of protein-protein interaction networks

2.1 STRING (https://string-db.org/) is a database for the study of the interaction networks between different proteins, with data derived from published experimental data and genes from computer mining, prediction and fusion. In our study, we used STRING 11.0 to analyze the DEGs obtained. We set the confidence score as 0.9. In view of the possibility that regions with more interconnections in the protein-protein interaction network (PPI) has a higher probability to participate in the biological regulation, and dissociated gene nodes seldom play a key role in the entire network, we chose to eliminate the dissociated gene nodes in our study. After that we downloaded the network file for subsequent analysis.

2.2 Cytoscape is a software for network analysis, which includes plug-ins such as Cytohubba, MCODE (full name is Molecular Complex Detection) and Cluego, etc. Cytohubba can rank network nodes via various topological analysis methods. Among all the topological methods, MCC method has the most accurate prediction value. MCODE can detect closely related regions in PPI network, which may infer different molecular complexes. With Cytohubba plug-in MCC method, we ranked network nodes and selected the top ten genes for further analysis. With MCODE plug-in, we divided the network into different sub-networks, and analyzed the first five sub-networks.

3. Enrichment analysis

Common enrichment analysis methods mainly include GO (Gene Ontology) (including CC (cellular component), BP (biological process). MF (Molecular Function)) and KEGG (Kyoto Encyclopedia of Genes and Genomes), and can reveal the functional tendency of gene sets. The commonly used enrichment tools include David database, Kobas database, key R package "Clusterprofiler" package ¹⁶, and plug-ins include Funrich, ClueGo, Cluepedia, Metascape, etc. In our study, We used Cytoscape plug-in GlueGo (version 2.5.7) and Cluepedia (version 1.5.7) to do gene annotation analysis on the genes sets obtained from MCODE and Cytohubba methods from step 2, and used "Clusterprofiler" R package to analyze the DEGs from step 1, and filtered pathways with P <0.05 and Q <0.05. The top 30 GO and KEGG enrichment results were selected and shown as bubble plots. Up-regulated and down-regulated genes of the KEGG pathways of interest were visualized using "Pathview" R package ¹⁷.

4. Evaluation of immune infiltration of brain tissues

Flow cytometry is currently the key way for evaluation of immune cells infiltration. However, it may take time to excavate the potential immune cells and determine the functional phenotypes of tissues in large samples by using flow cytometry only. CIBERSORT (cell-type identification by estimating relative subsets of RNA transcript) algorithm can simplify the process through deconvolving peripheral immune cells subtypes based on linear support vector regression. And CIBERSORT is one of the current methods for immune cell subtypes calculation based on gene expression profiles. In our study, we used CIBERSORT to analyze the characteristics of immune cells infiltration in brain tissues of multiple sclerosis. We input merge dataset into CIBERSORTX website (https://CIBERSORTx.stanford.edu/), and we set simulation calculation times as 100 times to obtain the proportion of 22 immune cells. Then we filtered out data with P<0.05 as data with significant differences for further analysis. And we used stacked bar chart to visualize the composition of different immune cells. We also calculated the correlation coefficient between different immune cell, and presented as a correlation heatmap. The differences between multiple sclerosis and control group were compared by Wilcoxon rank sum test. The differences of immune cell infiltration between the disease group and control group were visualized by violin map with the "ggplot2" R package. We also draw a two-dimensional PCA cluster plot to visualize the differences of immune cells infiltration between multiple sclerosis group and control group with "gggplot2" R package.

5. Correlation test between hub genes and immune cells

We obtained the proportion of 22 immune cells in the brain tissues of multiple sclerosis by CIBERSORT deconvolution method in Step 4. And we also did a spearman correlation analysis of the first 5 hub genes of Cytohubba and the infiltrating immune cells in the samples and presented as correlation maps using "ggplot2" R package.

6. Construction of weighted co-expression network

Weighted gene co-expression networks (WGCNA) ¹⁸ is an open source R package, which can be used to construct gene co-expression networks based on the following two points: One is that genes with similar expression patterns may be functionally related or has co-regulatory networks; the other is that the connections between genes conform to scale-free distribution¹⁸. In our study, we used step-by-step method to analyze the combined gene set. First step is to calculate the best soft threshold β, then we construct a hierarchical clustering tree based on the genetic correlation coefficient, and divided genes into different modules according to genes expression patterns, and we set a appropriate minimum quantity of gene modules and shear height threshold to further merge similar genes. The absolute value of Pearson correlation was used to measure the connectivity of genes in modules. And genes with high connectivity within the module were considered as the hub genes of the module. The hub genes within the module often had a strong correlation with specific traits. After we found the gene module of specific trait we aimed to analyze, we used "ggplot2" R package and acquire intersections of gene sets including MCODE Cluster 1 gene set, Cytohubba Top 10 gene set and specific gene module from WGCNA to observe the relationships between different genes clusters.

1. DEGs Screening

We included GSE108000 and GSE135511 into analysis. Descriptions of the datasets were shown in Table 1. GSE108000 dataset included gene expression profiles of pathological specimens of brain motor cortex of 20 patients with multiple sclerosis (specimens with and without meningeal infiltration were collected) and motor cortex specimens from 10 patients without neurological diseases. GSE135511 dataset included gene expression profiles of 30 white matter lesions from 7 patients with chronic active lesion multiple sclerosis and 8 patients with inactive lesion, and 10 white matter samples from control group. All specimens are non-living specimens.

We downloaded series matrix from GSE108000 and GSE135511. GSE108000 contains gene expression profiles of white matter from10 control group and 30 MS specimens, and GSE135511 dataset contains gene expression data of grey matter of 10 control group and 40 MS specimens. We combined and normalized the two datasets, removed batch effect between groups and got the merge dataset. We set the threshold of significant difference of DEGs as p < 0.05 and | log2FC | > 0.5 in order to get more DEGs, and we finally get 882 DEGs between multiple sclerosis and control group, including 399 up-regulated genes and 483 down-regulated genes (Fig. 1A, Fig. 1B).

2. Construction of protein-protein interaction network

We input 882 DEGs into STRING websites, set the confidence score as 0.9 and removed isolated gene. And we obtained a PPI network containing 872 nodes and 1265 edges. We downloaded network information and used Cytohubba plug-in MCC method to get the top 10 hub genes, as shown in Table 2 Top 10 Hub Gene column. The genes were displayed in different color levels according to the critical degree of genes in the network, as shown in Figure 2F. The PPI network of 872 nodes was analyzed with MCODE plug-in, and we got 24 subnetworks. We then selected the first 5 subnetworks (as shown in Table 2) for further genes enrichment analysis.

3. Functional enrichment analysis of differential expressed genes, hub genes and subnetwork genes

GO and KEGG enrichment analysis of the first 10 hub genes was conducted with Cytoscape plug-ins ClueGo and the Cluepedia. As shown in Figure 3.1, results with the highest GO enrichment intensity suggested that the molecular functions of the 10 hub genes were closely related to the activity of MHC class II receptors, polypeptide antigens and polysaccharides binding, ER-phagosome pathway and cytokine signal transduction in the immune system. While the top 5 KEGG enrichment intensity were allograft rejection, graft-versus-host disease, type 1 diabetes, asthma, and autoimmune thyroid disease, all of which were autoimmune diseases. The enrichment results of Cluster 1 obtained from MCODE were similar to the enrichment results of the first 10 genes of Cytohubba. GO analysis suggested that the gene set was related to MHC II, MHC I, glycation, growth factor binding and amyloidosis. Cluster 2 was associated with ubiquitin mediated proteolysis pathway. Cluster 3 was related to spliceosome and mRNA surveillance pathway, and Cluster 4 was related to the production of ROS, RNS and collagen degradation by phagocytes. Cluster 5 was associated with the regulation and signal transduction of TRP channels by inflammatory mediators.

GO (BP, CC, MF) enrichment was performed on the DEGs obtained in step 1 by using "Clusterprofiler" in R package. Results of the most significant pathways were displayed, as shown in figure 3.2. The most significant GO BP showed negative immunoregulation, myeloid differentiation, antigen and positive regulation of cytokines production, control of protein catabolism, glial cells activity and response to interferon gamma pathway, etc. The cellular components are mainly related to collagen fibers, extracellular matrix, adhesion plaques, endocytic vesicles, endocytic vesicle membrane, secretory granular membrane, MHC protein complex, etc. Molecular function is related to amide binding, protein serine/threonine kinase activity, polypeptide binding, carboxylic acid binding, organic acid binding, cytokine binding, immune receptor activity, cell adhesion mediator activity and MHC class II receptor activity. KEGG enrichment analysis results are shown in Figure 3.2. Enrichment pathways can be classified into virus infection (such as the human papilloma virus, swine flu, cytomegalovirus, EB virus, etc.), bacterial infection (salmonella, staphylococcus aureus, etc.), autoimmune diseases (include inflammatory bowel disease, type 1 diabetes, autoimmune thyropathy, asthma, etc.) and others (include atherosclerosis, endogenous ligand and other related pathways).

GO and KEGG enrichment results of the first 10 hub genes obtained by Cytoscape Cytohubba plug-in, the first subgene set Cluster 1 obtained by MCODE plug-in are consistent with enrichment results of DEGs, which mainly focus on immune regulatory pathways and immune-related disease pathways. However, the enrichment results of DEGs are more comprehensive and suggest that in addition to immune response, the response of glial cells of brain tissues to antigens also plays an indispensible role in the pathological mechanisms.

In our study, we also selected Epstein-Barr virus pathway to study how the specific environment factor interacts with susceptible genes of multiple sclerosis. Pathways were visualized by using "PathView" R package. As shown in Figure 3.3, Epstein Barr virus mainly acts on B cells through B cell receptors such as MHC-II and TLR2. After entering B cells, Epstein Barr virus further affects the immune activation pathway and up-regulates genes related to antigen presentation of MHC-I, which further activates cytotoxic T cells. In addition, Epstein-Barr virus may indirectly activate the PI3K pathway and affect cell proliferation and cell cycle.

4. Characteristics of immune infiltration in brain tissues of multiple sclerosis patients

In order to explore the characteristics of immune infiltration in brain tissues of multiple sclerosis from the cellular level, we uploaded the merge dataset through the CIBERSORTX website, calculated 100 times by CIBERSORT deconvolution method and finally we obtained the proportion of 22 immune cells of different samples. A total of 70 samples with a relatively stable proportion of immune cells (including 12 control samples and 58 multiple sclerosis samples) were screened by P<0.05, as shown in Figure 4.1. According to the results of analysis, the proportion of resting mast cells, macrophage M2, neutrophils, plasma cells, T cells CD8 and macrophage M0 were relatively high in multiple sclerosis brain tissues.

The correlation heatmap of immune cells between multiple sclerosis and control was drawn as shown in 4.2A. The correlation coefficient with the highest absolute value was the correlation between follicular helper T cells and eosinophils (R=0.5), suggesting that there was a certain positive correlation between follicular helper T cells and eosinophils. Correlation coefficient of macrophage M0 and neutrophils (R=-0.41) was followed, suggesting there is a negative correlation between macrophage M0 and neutrophils. Overall, the correlations between immune cells in the specimens ere weak to moderate. Differentiation of immune cells in the MS and control samples was analyzed and shown as the violin diagram. As shown in Figure 4.2B, CD4+T cells naive were significantly reduced in the MS group compared with the healthy control group (P=0.013) and resting synaptic cells were significantly increased (P=0.003). The PCA cluster diagram showed differences in the level of immune cell infiltration between MS and healthy control samples (4.2C).

5. Construction of weighted gene co-expression network

In order to further observe the differences in gene expression in different brain regions, we also performed weighted gene co-expression analysis (WGCNA) on the combined dataset to observe the differences in gene expression in white matter lesions, normal white matter tissue, gray matter lesions and normal gray matter tissue of multiple sclerosis patients. The variance of all genes was taken and the top 25% of the variance was extracted to obtain the expression profile of 3709 genes as the input dataset of WGCNA. The optimal soft threshold β value was calculated to be 4 based on R2, slope and mean k parameters. In our study, R2=0.95, which means the connectivity degree of the scaling free network construction is optimal (as shown in Figure 5.1). The network was constructed by step-by-step method, and the minimum number of genes in the module was set as 30. Then we further merged modules of those module similarity above 0.75, and finally we obtained 11 different module module (figure 5.2). We further draw the correlation heatmap between gene modules and tissue characteristics (figure 5.3). Among the modules obtained, the ones with the most genes and the highest correlation with traits were turquoise modules (946 genes). Correlation analysis between gene modules and tissue traits showed that there were significant differences in gene expression between white matter of multiple sclerosis group and control group, and between gray matter of disease group and control group. In addition to distinguishing control group and disease group, the module with the greatest correlation with traits (turquoise module) may well distinguish control group and disease group. There was also a certain degree of differentiation between the chronic active white matter lesions and the chronic active lesion adjacent tissues in the disease group, but no effective gene module that could clearly distinguish the gray matter lesions from the normal gray matter tissue. In our study, we further compared the turquoise gene module with the hub genes of Step 2, and found that all the hub genes were in the turquoise module (as shown in Figure 5.4), suggesting that there was a strong correlation between the hub genes and chronic active white matter lesions. There was no significant difference between normal gray matter and pathological gray matter in multiple sclerosis, but there were significant differences between the two gray matter and the control group. However, in the correlation diagram between gene modules and traits, the modules with the greatest difference correlation fell in the gray module, which belonged to the unclassified gene module. In conclusion, compared with the control population without central disease, patients with multiple sclerosis generally have abnormal gene expression in both gray matter and white matter of brain tissue. The turquose module gene set obtained in our study can further distinguish chronic active lesions and chronic active adjacent tissues in patients with multiple sclerosis.

6. Correlation test between hub genes and immune cells

Spearman correlation analysis was performed between the first 5 hub genes of Cytohubba (HLA-DRA, HLA-DRB1, HLA-DRA5, HLA-DRA and HLA-DPA1) and the 22 immune cells in the samples obtained from CIBERSORT. It was found that the expression levels of 5 hub genes were all negatively correlated with the number of M0 in macrophages, and positively correlated with the number of M1 or M2 in macrophages and the number of gamma delta in T cells. In neuroinflammatory diseases, macrophages play a dual role in the process of tissue damage according to their activation state (M1 / M2). Macrophage M1 can damage neurons, while M2 macrophages are believed to contribute to the regeneration and repair of neurons, and M0 is the resting state [95]. The correlation diagrams of the analyzed gene correlations (P <0.05) was shown as scatter plots in Figure 6.

Little was found in the literature on the questions of differentially expressed genes in different brain regions and immune cells infiltration pattern of brain tissues in multiple sclerosis patients，an initial objective of the study was to identify hub genes of grey and white matter lesions and infiltrating immune cells in brain tissues. Gene matrix, an emerging technology with the advantage of efficient and large-scale collection of gene expression profile data of diseases, have been widely used in bioinformatics research¹⁹. With bioinformatic tools, we performed a comprehensive analysis of the differentially expressed gene between multiple sclerosis patients and subjects with no neurological diseases with data from Gene Expression Omnibus (GEO) database, and used the CIBERSORT algorithm to analyze the immune infiltration of multiple sclerosis brain tissues.

Most of the hub genes are located in HLA-DR loci in the MHC region, and the genes were found to be associated with immunity. Screening the functions of hub genes also confirms that the ten genes are susceptible genes for multiple sclerosis^20-30. HLA-DRA, a member of the HLA class II alpha chain, is expressed on the surface of various antigen-presenting cells and plays a central role in the immune response through presenting polypeptides, particularly of pathogenic origin, to T cells. HLA-DRB1 and HLA-DRB5 are associated with antigen binding and T-cell receptor (TCR) recognition²¹. HLA-DQB1 and HLA-DQB2 are associated with antigen binding and MHC class II receptor activity. Studies identified DRB1 and DQB1 as the strongest genetic determinants of multiple sclerosis and classified HLA genotypes into high risk types (including 2 susceptible gene haplotypes) and low-to-medium risk types (including 1 susceptible gene haplotype or no susceptible gene haplotypes) and found a significant decrease in brain volume in high-risk group compared to low-to-medium risk group²³. HLA-E, a non-classical MHC class Ib molecule, is a ligand for the NKG2A inhibitory receptor on natural killer cells and CD8+ cells and plays a dual role in immune system^24-25. HLA-C has an independent impact on MS susceptibility and related to antigen cross presentation and immune response including IFNα/β and IFNγ signaling pathways. CD44 encodes a protein of cell surface glycoprotein involved in intercellular interactions, cell adhesion and migration, and in a variety of cellular functions, including lymphocyte activation, circulation and homing and hematopoiesis; CD44 has multiple protective roles including influencing cytokine production, T cell differentiation, T cell-endothelial cell interactions, and blood-brain barrier integrity in EAE, CD44-deficient mice exhibit more severe lesions²⁸, and recent studies suggest that central nevous system (CNS) lesions following CD44 gene deletion may result from alterations of gut microbiota and short-chain fatty acids²⁹.

Our study also found that the ten genes are closely related to chronic active white matter lesions through WGCNA algorithm. We hypothesized that the genes obtained in our study were significantly expressed in chronic active white matter lesions, but can not rule out that the genes also influence on other sites. At present, there are relatively few studies on the relationships between genes and brain regions. Some studies evaluated the pathogenicity of NF-κB signaling pathway related genes in different brain tissues (including cerebellum, cortex, hippocampus, white matter, putamen, etc.) of multiple sclerosis patients. By using inverse variance weighted estimation and weighted median estimation, studies found that there were genes specifically expressed in different parts, and there was also mutual regulation between genes³¹. Comparison of miRNAs in different brain regions of multiple sclerosis patients³¹ found that active and chronic inactive white matter lesions share regulatory miRNAs, which mainly involved in astrocyte proliferation, microglial proliferation, demyelination and other processes while microglia activation and inflammatory infiltration are not obvious in chronic inactive white matter lesions. Most of the miRNAs overlap in chronic inactive white matter lesions and gray matter lesions were related to the response of astrocytes to inflammatory stimulation, maintenance of inflammatory cytokines, apoptosis and regeneration of glial cells. MiRNAs in chronic inactive white matter lesions and normal white matter tissues are negatively regulated, and the biggest differences lie in the degree of myelin loss and remyelination. Study of gene expression in different brain tissues is helpful to better uncover the pathological mechanism of pro-inflammatory and anti-inflammatory in brain tissues.
GO analyses indicate that hub genes significantly enrich in immune pathways including antigen presentation and antibodies production, T cell activation, proliferation and differentiation. GO analyses of DEGs also suggest glial cell proliferation together contribute to the pathogenesis. Overall, the pathogenesis of multiple sclerosis involves peripheral immune cells and a cascade of brain glial cell activation. Previous studies have shown that glial cells have bidirectional interactions with components of the immune system, not only mediating tissue damage and repairing immunity, but also directly participating in inflammatory processes, further aggravating inflammation and axon injury ³¹. Signals between inflammatory cells and the target tissues may also be bidirectional. A toxic microenvironment mediated by immune cells (peripheral or central nerve cell compartmentalization) and resident cells of the central nervous system (microglia and astrocytes) can produce inflammatory mediators, reactive oxygen species, and iron deposition, leading to chronic demyelination and axonal lesions. Structural damage to myelin oligodendrocytes and loss of nutritional support lead to increased energy requirements for nerve impulse transmission, while mitochondrial damage and disruption of ion homeostasis further impair normal electrical activity. The above processes jointly promote brain atrophy, and explain the pathological mechanism of brain atrophy in patients with advanced multiple sclerosis^35-36. KEGG enrichment pathways in our study suggests that multiple sclerosis is closely related to environmental risk factors including viral infection, bacteria infection and certain autoimmune diseases. Epstein-Barr virus is regarded as the leading environmental risk factor ³⁷. And the cross-reactivity of CD4+T cells partly explained the synergistic effect between EBV infection and genetic susceptibility to multiple sclerosis³⁸. However, the pathways of interactions between EB virus and DEGs of multiple sclerosis of our study still needs to be verified by further studies. KEGG pathway also suggests that multiple sclerosis is associated with other acquired autoimmune diseases (such as inflammatory bowel disease, type 1 diabetes, autoimmune thyroid disease, rheumatoid arthritis, etc.). MHC genes mutations contribute to the pathogenesis of most autoimmune diseases and affect the regulation of immune cells response, T cell differentiation, cytokine secretion and so on. Most of the diseases are characterized by recurrent attacks and chronic progression³⁹.

To further explore the role of immune cells in the brain tissues of multiple sclerosis, our study used CIBERSORTX algorithm to analyze brain tissues samples of multiple sclerosis patients. In general, there are more CD4+T cells in multiple sclerosis than in the control group. Our study suggests a reduction in initial CD4+T cells, which suggests that the proportions of CD4+T cells may be various in brain tissues of multiple sclerosis patients according to cell subtypes⁴⁰. CD8+T lymphocytes serve as resident cells of the brain and spinal cord tissues, and transmit neuroinflammation locally when the cells encounter homologous antigens⁴¹. Although it has long been thought that multiple sclerosis is a T-cell-mediated disease, in recent years much more studies focus on other important components include B cells, endothelial cells, complements, autoantibodies, cytokines, and chemokines⁴². B cells and myeloid cells are equally important in the inflammatory response of the central nervous system. At present, anti-CD20 monoclonal antibodies (such as rituximab) are used clinically to deplete B cells, which has certain effects on reducing recurrence and slowing the progression of the disease, further verifying the important role of B cells and antibodies in the progression of brain injury in multiple sclerosis⁴³. Macrophages and microglia have similar inflammatory effects, and the infiltration of macrophages is strongly correlated with the model of advanced experimental autoimmune encephalomyelitis⁴⁴, suggesting the role of macrophages in the late stage of the disease. EAE studies have found that the inactivation of macrophages can improve the severity of the disease⁴⁵. Permanent myeloid cells can be found in all the tissues and organs, and in the case of infection or tissue injury, monocytes or granulocytes (especially neutrophils) can be recruited from the circulation to participate in the promotion of inflammatory response at the beginning, and play roles in tissue repair and regeneration at the later stage⁴⁶. Previous studies on immune cell patterns of gene expression suggest that the development of multiple sclerosis is caused by a wide subset of cells⁴². Pathogenic cells include autologous reactive CD4, CD8, and B cells, while regulatory cells include regulatory T cells, regulatory B cells, and mononuclear phagocytes and NK cells that regulate all cells. Patients may have dysregulatory proportions of immune cells, which results in differences in treatment responses. Susceptible genes for major cell subsets are responsible for transcription factors, chemokines, receptors, intracellular enzymes, and signal transduction factors that control cell proportion, cell differentiation, and cell status⁴². The identification of target antigens recognized by tissue-hosting CD8+ T cells and B cells, as well as the molecular properties and corresponding gene targets of soluble inflammatory mediators that may cause tissue damage, can help guide the treatment of multiple sclerosis⁴¹. Some hub genes were negatively correlated with the resting state immune cells and positively correlated with the active state, suggesting that certain hub genes may be related to the activation of peripheral immune cells such as macrophages. However, the role of genes in the regulation of immune cells in multiple sclerosis patients remains to be further investigated.

The 2019 International Multiple Sclerosis Genetics Consortium (IMSGC) explored the expression of multiple sclerosis genome-wide immune cells and immune organs⁸. IMSGC annotated genome-wide gene fragments of interest using gene expression profiles and epigenetic signature information, and found that multiple sclerosis susceptibility loci were significantly enriched in many different immune cell types and tissues, including natural killer cells, dendritic cells, and the thymus. The difference between our study and the IMSGC study is that the subjects of our study are brain tissues, which can intuitually show the changes of pathological microenvironment of brain tissues. And the traditional methods for evaluating tissue immune infiltration include immunohistochemistry, flow cytometry, etc., which are mostly limited by the size of samples, making it difficult to analyze the composition of infiltrating immune cells more comprehensively.

Multiple sclerosis is closely related to intracranial lesion immune infiltration. In recent years, studies on the pathological mechanism of multiple sclerosis has changed from animal and cell experiments to the genomic analysis of human cells and tissues samples, which together reflect the genetic and immune characteristics of multiple sclerosis more objectively. In order to better understand the genetic composition, cell diversity and pathway characteristics of infiltrating immune cells, our study screened the DEGs using the "Limma" package of R language, and analyzed genes networks through Cytoscape software plug-in. With the help of Cibersort algorithm, we quantified different types of infiltrating immune cells in the samples. We avoided tedious surface markers and cells separation methods, and used scientific algorithm to identify the proportions of 22 common immune cell types of multiple sclerosis, and explore the relationships between immune cells and DEGs, and further obtain a more comprehensive landscape of immune infiltration in multiple sclerosis brain tissues, and provide guidance for future research on genetic and immune patterns of multiple sclerosis.

Similar to most bioinformatic studies, our studies have following limitations. Firstly, sample size of the study and patient information are limitied by original researches, and we can not eliminate the bias from gender, comorbidity, disease stage, and medications. Secondly, although merge data increase the sample size, it cannot eliminate the system error caused by different studies methods and experiments conditions. Thirdly, the samples are from cadavers, and gene expression of the samples may be affected by the sample storage time and processing methods, which are somewhat different from living tissues. Currently, the development of brain biopsy is restricted, and the number of well-preserved pathological samples for rare diseases is limited. Although most of the results obtained in our study are supported by previous studies, more tissue samples and samples information are needed for further research.

Given that multiple sclerosis is not a simple genetic disease, and there are complex interactions among genetics, environment and immunity, it is difficult to realize the application of genes in the practical clinical diagnosis and treatment of multiple sclerosis. Currently, more and more studies of genes and environment, and immune mechanism behind, future research should consider to include more informations include clinical phenotypes, prognosis and treatment response, genetic (including epigenetic) and pathology characteristics, and expand sample sizes further explore the molecular mechanism of clinical phenotype and therapeutic prognosis heterogeneity, which will help further clarify the molecular classifications of disease, and guide disease prevention and medications development of multiple sclerosis.

In our study, 10 key genes of multiple sclerosis, HLA-DRA, HLA-DRB1, HLA-DRB5, HLA-DQB2, HLA-DPA1, HLA-DQB1, HLA-E, HLA-C, GBP2, CD44, were obtained by the algorithm. Enrichment pathways are associated with T cell activation, proliferation and differentiation, antigen presentation, autoantibody production, and glial cell proliferation, suggesting that the pathogenesis of multiple sclerosis involves the involvement of the peripheral immune system and a cascade of brain glial cell activation. The infiltrating immune cells in brain tissues are diverse, which are not only related to T cells, but also include B cells, macrophages, granulocytes and other peripheral immune cells. The key genes are correlated with some immune cells, which may indicate the role of susceptibility genes in promoting the activation of macrophages and natural killer cells. The gene expression profiles in brain tissues (including gray matter and white matter) of patients with multiple sclerosis were significantly different from those in the control group, with the most significant differences in chronic active white matter lesions, followed by tissues adjacent to chronic active white matter lesions. Ten key genes were closely related to chronic active white matter lesions. The analysis of the pathway of differentially expressed genes enrichment and peripheral immune cell infiltration in our study can provide more ideas for the etiological and pathological studies of multiple sclerosis.. Gene expressions and the underlying mechanisms of immune infiltration of multiple sclerosis and their correlations with environment and disease stage need further exploration.

Abbreviations	Full Names
BP	Biological Process
CC	Cellular Component
CNS	Central Nervous System
Contr	Control
DEGs	Differentially Expression Genes
EAE	Experimental Autoimmune Encephalomyelitis
EB virus	Epstein-Barr Virus
GBP2	Gtp Binding Protein 2
GEO	Gene Expression Omnibus Database
GO	Gene Ontology
GPL	GEO Platform
HLA	Human Leukocyte Antigen
IFN	Interferon
IL	Interleukin
KEGG	Kyoto Encyclopedia Of Genes And Genomes
MAPK	Mitogen-Activated Protein Kinase
MCODE	Molecular Complex Detection
MF	Molecular Function
MHC	Major Histocompatibility Complex
miRNAs	MicroRNAs
MS	Multiple Sclerosis
NCBI	National Center Of Biotechnology Information
NK cell	Natural Killer (Cell)
OR	Odd Ratio
P.adj	Adjust P-Values
PI3K	Phosphatidylinositide 3-Kinases
TCR	T Cell Receptor
TLR	Toll Like Receptor
WGCNA	Weighted Correlation Network Analysis

Authors’ contributions

All authors contributed to conceptualization and data curation. XZ wrote the first draft. LY reviewed the data and the final manuscript. All authors read and approved the final manuscript.

Acknowledgements

Not applicable.

Availability of data and materials

Not applicable.

Competing interests

The authors declare no competing interests.

Consent for publication

All authors have reviewed the final version of the manuscript and approved it for publication.

Ethics approval and consent to participate

Not applicable.

Funding

None.

F. Rinaldi, M. Calabrese, P. Grossi, et al. Cortical lesions and cognitive impairment in multiple sclerosis. NEUROL SCI 2010;31(Suppl 2):S235-37. doi: 10.1007/s10072-010-0368-4 pmid:206351132010-11-01].
HMB Lunde, J. Assmus, K. M. Myhr, L. Bo, N. Grytten. Survival and cause of death in multiple sclerosis: a 60-year longitudinal population study. J Neurol Neurosurg Psychiatry 2017;88(8):621-25. doi: 10.1136/jnnp-2016-315238 pmid:28365589.
Global, regional, and national burden of multiple sclerosis 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. LANCET NEUROL 2019;18(3):269-85. doi: 10.1016/S1474-4422(18)30443-5 pmid:30679040.
H. Kuusisto, J. Kaprio, E. Kinnunen, et al. Concordance and heritability of multiple sclerosis in Finland: study on a nationwide series of twins. EUR J NEUROL 2008;15(10):1106-10. doi: 10.1111/j.1468-1331.2008.02262.x pmid:187276712008-10-01].
C. Fagnani, M. C. Neale, L. Nistico, et al. Twin studies in multiple sclerosis: A meta-estimation of heritability and environmentality. Mult Scler 2015;21(11):1404-13. doi: 10.1177/1352458514564492 pmid:25583848.
L. Belbasis, V. Bellou, E. Evangelou, J. P. Ioannidis, I. Tzoulaki. Environmental risk factors and multiple sclerosis: an umbrella review of systematic reviews and meta-analyses. LANCET NEUROL 2015;14(3):263-73. doi: 10.1016/S1474-4422(14)70267-4 pmid:25662901.
B. Hemmer, M. Kerschensteiner, T. Korn. Role of the innate and adaptive immune responses in the course of multiple sclerosis. LANCET NEUROL 2015;14(4):406-19. doi: 10.1016/S1474-4422(14)70305-9 pmid:25792099.
Multiple sclerosis genomic map implicates peripheral immune cells and microglia in susceptibility. SCIENCE 2019;365(6460) doi: 10.1126/science.aav7188 pmid:31604244.
W. N. Brandao, M. G. De Oliveira, R. T. Andreoni, et al. Neuroinflammation at single cell level: What is new? J Leukoc Biol 2020;108(4):1129-37. doi: 10.1002/JLB.3MR0620-035R pmid:32779279.
A. Butte. The use and analysis of microarray data. NAT REV DRUG DISCOV 2002;1(12):951-60. doi: 10.1038/nrd961 pmid:124615172002-12-01].
Isaac S. Kohane, Atul Butte. Bioinformatics in Functional Genomics. Encyclopedia of Biostatistics 2000 doi: https://doi.org/10.1002/0470011815.b2a050992000-01-01].
Y. J. Deng, E. H. Ren, W. H. Yuan, et al. GRB10 and E2F3 as Diagnostic Markers of Osteoarthritis and Their Correlation with Immune Infiltration. Diagnostics (Basel) 2020;10(3) doi: 10.3390/diagnostics10030171 pmid:322357472020-03-22].
T. Barrett, S. E. Wilhite, P. Ledoux, et al. NCBI GEO: archive for functional genomics data sets--update. NUCLEIC ACIDS RES 2013;41(Database issue):D991-95. doi: 10.1093/nar/gks1193 pmid:231932582013-01-01].
H. S. Parker, J. T. Leek, A. V. Favorov, et al. Preserving biological heterogeneity with a permuted surrogate variable analysis for genomics batch correction. BIOINFORMATICS 2014;30(19):2757-63. doi: 10.1093/bioinformatics/btu375 pmid:24907368.
Leland Wilkinson. ggplot2: Elegant Graphics for Data Analysis by WICKHAM, H. BIOMETRICS 2011;67(2):678-79. doi: https://doi.org/10.1111/j.1541-0420.2011.01616.x..
G. Yu, L. G. Wang, Y. Han, Q. Y. He. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 2012;16(5):284-87. doi: 10.1089/omi.2011.0118 pmid:224554632012-05-01].
W. Luo, C. Brouwer. Pathview: an R/Bioconductor package for pathway-based data integration and visualization. BIOINFORMATICS 2013;29(14):1830-31. doi: 10.1093/bioinformatics/btt285 pmid:237407502013-07-15].
P. Langfelder, S. Horvath. Fast R Functions for Robust Correlations and Hierarchical Clustering. J STAT SOFTW 2012;46(11) pmid:230502602012-03-01].
H. R. Ali, L. Chlon, P. D. Pharoah, F. Markowetz, C. Caldas. Patterns of Immune Infiltration in Breast Cancer and Their Clinical Implications: A Gene-Expression-Based Retrospective Study. PLOS MED 2016;13(12):e1002194. doi: 10.1371/journal.pmed.1002194 pmid:279599232016-12-01].
D. A. Hafler, A. Compston, S. Sawcer, et al. Risk alleles for multiple sclerosis identified by a genomewide study. N Engl J Med 2007;357(9):851-62. doi: 10.1056/NEJMoa073493 pmid:17660530.
H. L. Lang, H. Jacobsen, S. Ikemizu, et al. A functional and structural basis for TCR cross-reactivity in multiple sclerosis. NAT IMMUNOL 2002;3(10):940-43. doi: 10.1038/ni835 pmid:122443092002-10-01].
S. Saidha, P. A. Calabresi. Anti-interleukin-2 receptor alpha for multiple sclerosis? LANCET 2013;381(9884):2141-43. doi: 10.1016/S0140-6736(13)60065-3 pmid:235620082013-06-22].
L. Lorefice, G. Fenu, C. Sardu, et al. Multiple sclerosis and HLA genotypes: A possible influence on brain atrophy. Mult Scler 2019;25(1):23-30. doi: 10.1177/1352458517739989 pmid:291118832019-01-01].
S. A. Joosten, L. C. Sullivan, T. H. Ottenhoff. Characteristics of HLA-E Restricted T-Cell Responses and Their Role in Infectious Diseases. J IMMUNOL RES 2016;2016:2695396. doi: 10.1155/2016/2695396 pmid:276991812016-01-20].
P. F. Durrenberger, L. V. Webb, M. J. Sim, et al. Increased HLA-E expression in white matter lesions in multiple sclerosis. IMMUNOLOGY 2012;137(4):317-25. doi: 10.1111/imm.12012 pmid:23039207.
F. Ye, J. Liang, J. Li, H. Li, W. Sheng. Development and Validation of a Five-Gene Signature to Predict Relapse-Free Survival in Multiple Sclerosis. FRONT NEUROL 2020;11:579683. doi: 10.3389/fneur.2020.579683 pmid:33343487.
N. L. Fransen, C. C. Hsiao, M. van der Poel, et al. Tissue-resident memory T cells invade the brain parenchyma in multiple sclerosis white matter lesions. BRAIN 2020;143(6):1714-30. doi: 10.1093/brain/awaa117 pmid:32400866.
K. M. Flynn, M. Michaud, J. A. Madri. CD44 deficiency contributes to enhanced experimental autoimmune encephalomyelitis: a role in immune cells and vascular cells of the blood-brain barrier. AM J PATHOL 2013;182(4):1322-36. doi: 10.1016/j.ajpath.2013.01.003 pmid:23416161.
K. N. Chitrala, H. Guan, N. P. Singh, et al. CD44 deletion leading to attenuation of experimental autoimmune encephalomyelitis results from alterations in gut microbiome in mice. EUR J IMMUNOL 2017;47(7):1188-99. doi: 10.1002/eji.201646792 pmid:28543188.
J. A. Hollenbach, J. R. Oksenberg. The immunogenetics of multiple sclerosis: A comprehensive review. J AUTOIMMUN 2015;64:13-25. doi: 10.1016/j.jaut.2015.06.010 pmid:26142251.
T. Fazia, A. Nova, D. Gentilini, et al. Investigating the Causal Effect of Brain Expression of CCL2, NFKB1, MAPK14, TNFRSF1A, CXCL10 Genes on Multiple Sclerosis: A Two-Sample Mendelian Randomization Approach. Front Bioeng Biotechnol 2020;8:397. doi: 10.3389/fbioe.2020.00397 pmid:32432099.
L. M. Healy, M. Yaqubi, S. Ludwin, J. P. Antel. Species differences in immune-mediated CNS tissue injury and repair: A (neuro)inflammatory topic. GLIA 2020;68(4):811-29. doi: 10.1002/glia.23746 pmid:31724770.
J. Gruchot, V. Weyers, P. Gottle, et al. The Molecular Basis for Remyelination Failure in Multiple Sclerosis. CELLS-BASEL 2019;8(8) doi: 10.3390/cells8080825 pmid:313826202019-08-03].
A. Geladaris, D. Hausler, M. S. Weber. Microglia: The Missing Link to Decipher and Therapeutically Control MS Progression? INT J MOL SCI 2021;22(7) doi: 10.3390/ijms22073461 pmid:338016442021-03-27].
M. A. Friese, B. Schattling, L. Fugger. Mechanisms of neurodegeneration and axonal dysfunction in multiple sclerosis. NAT REV NEUROL 2014;10(4):225-38. doi: 10.1038/nrneurol.2014.37 pmid:246381382014-04-01].
D. H. Mahad, B. D. Trapp, H. Lassmann. Pathological mechanisms in progressive multiple sclerosis. LANCET NEUROL 2015;14(2):183-93. doi: 10.1016/S1474-4422(14)70256-X pmid:25772897.[Copyright (c) 2015 Elsevier Ltd. All rights reserved.:*2015-02-01].
T. Olsson. Epstein Barr virus infection and immune defense related to HLA-DR15: consequences for multiple sclerosis. EUR J IMMUNOL 2021;51(1):56-59. doi: 10.1002/eji.202049030 pmid:33350470.
H. Zdimerova, A. Murer, C. Engelmann, et al. Attenuated immune control of Epstein-Barr virus in humanized mice is associated with the multiple sclerosis risk factor HLA-DR15. EUR J IMMUNOL 2021;51(1):64-75. doi: 10.1002/eji.202048655 pmid:32949466.
R. Horton, L. Wilming, V. Rand, et al. Gene map of the extended human MHC. NAT REV GENET 2004;5(12):889-99. doi: 10.1038/nrg1489 pmid:155731212004-12-01].
B. Arneth. Activated CD4+ and CD8+ T Cell Proportions in Multiple Sclerosis Patients. INFLAMMATION 2016;39(6):2040-44. doi: 10.1007/s10753-016-0441-0 pmid:275960062016-12-01].
H. Lassmann. Pathogenic Mechanisms Associated With Different Clinical Courses of Multiple Sclerosis. FRONT IMMUNOL 2018;9:3116. doi: 10.3389/fimmu.2018.03116 pmid:306873212018-01-20].
G. P. Parnell, D. R. Booth. The Multiple Sclerosis (MS) Genetic Risk Factors Indicate both Acquired and Innate Immune Cell Subsets Contribute to MS Pathogenesis and Identify Novel Therapeutic Opportunities. FRONT IMMUNOL 2017;8:425. doi: 10.3389/fimmu.2017.00425 pmid:284586682017-01-20].
G. Comi, A. Bar-Or, H. Lassmann, et al. Role of B Cells in Multiple Sclerosis and Related Disorders. ANN NEUROL 2021;89(1):13-23. doi: 10.1002/ana.25927 pmid:33091175..
B. Ajami, J. L. Bennett, C. Krieger, K. M. McNagny, F. M. Rossi. Infiltrating monocytes trigger EAE progression, but do not contribute to the resident microglia pool. NAT NEUROSCI 2011;14(9):1142-49. doi: 10.1038/nn.2887 pmid:218045372011-07-31].
E. D. Ponomarev, T. Veremeyko, N. Barteneva, A. M. Krichevsky, H. L. Weiner. MicroRNA-124 promotes microglia quiescence and suppresses EAE by deactivating macrophages via the C/EBP-alpha-PU.1 pathway. NAT MED 2011;17(1):64-70. doi: 10.1038/nm.2266 pmid:211319572011-01-01].
T. Owens, A. Benmamar-Badel, A. Wlodarczyk, et al. Protective roles for myeloid cells in neuroinflammation. SCAND J IMMUNOL 2020;92(5):e12963. doi: 10.1111/sji.12963 pmid:32851668.

Table 1. Basic information of data from GSE108000 and GSE135511

Dataset ID	Sample Source	Data Platform	Sample Grouping	Disease subtypes	Sample Number	Published Date
GSE108000	brain tissue	GPL13497 (Agilent)	multiple sclerosis and control	SPMS	40 (7M/18F)	Jan-18
GSE135511	brain tissue	GPL6883 (Illumina)	Multiple Sclerosis and control	No specific	50 (14F/16M)	Dec-19

Abbreviation: SPMS, secondary progressive multiple sclerosis; M, male; F, for female.

Table 3. The first 10 core genes obtained from Cytohubba plug-in and genes of 5 gene sets from MCODE plug-in

Gene Sets	Gene lists
Top 10 hub genes	HLA-DRA, HLA-DRB1, HLA-DRB5, HLA-DQB2, HLA-DPA1, HLA-DQB1, HLA-E, HLA-C, GBP2, CD44
Cluster 1	ACTR1B, HLA-DRB5, HLA-DQB2, HLA-DPA1, HLA-DMB, B2M, HLA-DRA, HLA-DMA, HLA-DRB1, HLA-DQB1, IRF8, KLC2, DYNLL2, DYNC1I1, GBP2, GBP1, DNAJC3, HLA-C, HLA-E, IFI30, CD44, IRF3, TMEM132A, LGALS1, PRKCSH, FSTL1, LTBP1, IGFBP7, CHGB, ADCY1, CALU, DRD4, HSP90B1, GNB4, HTR5A, CXCL16, CP, GNG3, CCR1, SPP1, C5AR1, ANXA1, GNG12, ADORA1, GNAI3, APOE, TIMP1
cluster 2	WWP1, FBXO30, KLHL22, FBXO44, FBXL4, UBE2S, DTX3L, MGRN1, RNF126
cluster 3	CD2BP2, PAPOLA, SYMPK, FUS, U2AF1L4, POLR2G, ELAVL2, PLRG1, PRPF31
cluster 4	OLR1, CD300A, TMEM63A, ITGAV, COL8A1, ATP6V0C, P4HA1, PLOD2, COL1A2, COL9A3, COL8A2, P4HA2, COL4A2, CYBA, COL4A1, CD33, SERPINH1, CD53
cluster 5	GPR65, GNRH1, P2RY2, CCKBR, TAC3, CCK, PIK3CA

Download PDF

Journal Publication

published 26 Jun, 2022

Read the published version in Computational Intelligence and Neuroscience →

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Integration of Genetic and Immune Infiltration Insights Into Data Mining of Multiple Sclerosis Pathogenesis

Status:

Journal Publication

Version 1

Abstract

Figures

Background:

Methods:

Results:

Discussion

Conclusion

Abbreviations

Declarations

References

Tables

Status:

Journal Publication

Version 1