Ubiquitous dysregulation of the Wnt pathway in immune thrombocytopenia genetic susceptibility

Key Points

• •ITP-associated Wnt pathway genes are dysregulated in multiple cell types, including lymphocytes and platelet/megakaryocyte lineage.
• •Some of these genes are upregulated, whereas others are downregulated in ITP, depending on the gene and cell type.

Visual Abstract

Abstract

The pathogenesis of immune thrombocytopenia (ITP) is complex and incompletely understood. Multiple cell types have been implicated, and their respective contribution is unclear. A recent genome-wide association study identified single-nucleotide polymorphisms (SNPs) associated with pediatric ITP within or near 5 genes in the canonical Wnt signaling pathway. To investigate whether this pathway was dysregulated in ITP and identify which cell types were involved, we leveraged extensive functional genomics and single-cell RNA sequencing (scRNA-seq) data. By linking the identified SNPs to likely regulated genes, we showed an enrichment in the Wnt pathway and identified 2 additional genes in this pathway involved in ITP. The SNPs affected regulation of the Wnt pathway genes in multiple cell types. Indeed, scRNA-seq showed some of these genes were expressed in lymphocytes, whereas others were expressed in platelets or megakaryocytes. By comparing the cell-specific expression of these genes between individuals with ITP and healthy controls, we demonstrated that several genes in the Wnt pathway were differentially expressed between these 2 groups. Some genes were upregulated, whereas other were downregulated in ITP. In sum, the Wnt signaling pathway is broadly dysregulated in ITP, with a complex pattern that varies across genes and cell types.

Introduction

The pathophysiology of immune thrombocytopenia (ITP) is complex and remains incompletely understood. Multiple immune cells have been implicated, and platelets themselves exhibit several functional alterations.¹ The respective contribution of each cell type remains unclear. Recently, to our knowledge, the largest genome-wide association study (GWAS) conducted in ITP, including 591 children, identified single-nucleotide polymorphisms (SNPs) located inside or near genes in the canonical Wnt signaling pathway: NKD1, NAV2, CCDC85A, GRIP1, and ABCC4.² The Wnt/β-catenin pathway plays a role in the function of many immune cells,³ including dendritic cells and T cells,^4,5 as well as in megakaryocyte development and platelet function.^6,7 Thus, a genetically mediated predisposition to ITP could involve several cell types. Confirming the contribution of the Wnt pathway and identifying the affected cell types would provide important insight into ITP pathophysiology. To that end, we leveraged extensive functional genomic data to further explore the role of the SNPs discovered in the GWAS.

Study design

SNP analysis

Using the SNPs identified in the GWAS from Kim et al (hereafter called original GWAS),² we gathered the genome-wide significant SNPs (n = 20) and suggestive SNPs (n = 6) in a case-control comparison, as well as a statistically suggestive SNP (n = 1) identified in a case-case comparison of patients with self-resolving ITP and those with chronic ITP (supplemental Table 1). To these 27 SNPs (sentinel SNPs), we added 221 additional SNPs (proxies) that were in linkage disequilibrium in the Trans-Omics for Precision Medicine (TOPMed) data set (supplemental Table 2; supplemental Figure 1).⁸

As adult ITP replication cohorts, we used summary statistics from the UK Biobank (cases, n = 814; controls, n = 357 646) and BioBank Japan (cases, n = 168; controls, n = 178 558).^9,10

Variant annotation, epigenomic data, and pathway enrichment

We used the Functional Annotation of Variant Online Resources (FAVOR) to annotate the SNPs using various resources including GeneCode, GeneHancer, Genotype-Tissue Expression (GTEx), and Activity-by-Contact (ABC) Encyclopedia of DNA Elements (ENCODE)_rE2G predictions.¹¹ We used EnrichR to perform a gene set enrichment analysis in different databases.¹²

scRNA-seq

We used Tabula Sapiens, a human single-cell RNA sequencing (scRNA-seq) atlas, to assess the cell-specific expression of the genes.¹³ We also reanalyzed published scRNA-seq data from bone marrow samples from 4 patients with ITP (n = 28 507 cells) and 4 healthy controls (HCs; n = 27 805 cells).¹⁴ We used the Seurat library to perform normalization, differential expression analysis, and gene set enrichment analysis–based computation of the Wnt pathway score.¹⁵ We annotated cell type using ANN-CAST (https://github.com/lavalleelab/ANN-CAST).¹⁶

Results and discussion

SNP replication in adult cohorts

We first aimed to replicate the 27 SNPs, but no other GWAS of pediatric ITP currently exists.¹⁷ Thus, we analyzed whether these SNPs were also associated with adult ITP. Four SNPs were nominally associated in the UK Biobank and 2 in BioBank Japan (supplemental Table 3). By examining the local region around the lead SNP rs4785426, we found that, despite this SNP not being replicated in the 2 adult cohorts, several proximal SNPs were nominally associated (supplemental Figure 2).

Thus, further genetic studies are required to replicate the findings of the original GWAS and clarify the association of these SNPs in adult ITP.

Identification of additional genes of the Wnt pathway dysregulated in ITP

We then aimed to analyze the pathways affected by the SNPs identified in the original GWAS.

We identified 154 genes predicted to be regulated by at least 1 of the 248 SNPs (27 sentinels + 221 proxies), including 14 long noncoding RNAs and 94 protein-coding genes (the others being micro RNAs, small nuclear RNAs, or novel genes). We found enrichment in Wnt/β-catenin signaling (MSigDBHallmark 2020; odds ratio [OR], 9.45; P = .02) and in Wnt signaling pathways (BioCarta 2016; OR, 21.22; P = .0005). These enrichments were driven by 3 genes: NKD1, MAP3K7, and DKK1. The last 2 have not been previously associated with ITP.^2,17 We also found the enrichment of the NRF2 pathway (WikiPathways 2024 human; OR, 7.12; P = .001), identified due to GSTM1, ABCC4, GSTM4, GSTM2, and SLC6A16.

Thus, we confirmed that the SNPs of the original GWAS mainly affected the Wnt pathway and identified 2 additional genes in the Wnt pathway that are likely dysregulated: MAP3K7 and DKK1.

The identified SNPs regulate Wnt pathway genes in multiple cell types

We next aimed to identify in which cell types the identified SNPs might regulate Wnt pathway signaling.

For 4 genes (NAV2, CCDC85A, ABCC4, and GRIP1), the identified SNPs were located within intronic regions. For NAV2, CCDC85A, and GRIP1, the SNPs (sentinel and proxies) did not colocalize with a predicted regulatory region, expression quantitative trait loci (QTLs) or splice QTLs (Figure 1). For GRIP1, some of the proxies appeared to affect HMGA2, either because they lie in intronic regions or predicted regulatory regions of this gene. HMGA2 is a transcription factor associated with several blood traits,¹⁸ but it is also a downstream effector of the Wnt pathway.^19,20 The sentinel at ABCC4 locus was in a predicted regulatory region in megakaryocytes (supplemental Figure 3).

Figure 1.

Genomic landscape of the NAV2, NKD1, GRIP1, and NKD1 loci with regulatory and functional annotations. This figure presents a multilayered visualization of genomic features at the NAV2 (top left), NKD1 (top right), and GRIP1 (bottom) loci, integrating data from various regulatory and functional data sets. Dashed vertical lines mark variant genomic positions, providing a reference for overlapping regulatory and functional elements. (A) Variants and CADD score; displays the genetic sentinel variant (triangle) and its proxies (dots) mapped to the region, along with their Combined Annotation Dependent Depletion (CADD) scores indicating potential functional impact. (B) RefSeq genes; annotated gene models from the RefSeq database. (C) Predicted regulatory regions; regulatory elements predicted in relevant cell types using the ABC_rE2G model (score ≥ 0.5), which links regulatory regions to target genes.¹¹ (D) Chromatin immunoprecipitation-Atlas–Assay for Transposable-Accessible Chromatin (ATAC)-seq tracks; open chromatin regions identified across multiple blood-derived cell types, highlighting accessible regulatory regions. (E) GTEx expression²¹; expression QTLs (eQTLs) from the GTEx database across all available tissues and cell types (q ≤ 0.05), in which colors represent the direction and magnitude of the effect (red, positive; blue, negative). (F) Kammers et al²²; eQTLs identified in megakaryocytes and platelets, linking genetic variants to gene expression changes.

Two of the top GWAS sentinels (rs4785426 and rs4785216) were proximal to NKD1.² They were located 89 and 2.9 kilobase pairs (kb) upstream of the transcription start site, respectively. Some of their proxies were located close to the predicted NKD1 regulatory region in T cells and natural killer cells, as well as in open chromatin regions in B cells and bone marrow (Figure 1; supplemental Table 2). Two proxies were located at an open chromatin region in all hematopoietic-derived cells. None of the sentinels or proxies were known as expression QTLs or splice QTLs for NKD1.

Among the SNPs not associated with any gene in the original article, rs79328664 is in a predicted regulatory region of DKK1 in type B pancreatic cells and is relatively close to the gene (29 kbp). DKK1 is a member of the Dickkopf family of proteins and inhibits β-catenin–dependent Wnt signaling.²³ Finally, although the sentinel rs78144867 and its proxies were not associated with any gene in the epigenomic data, they are located ∼300 kb from the MAP3K7 gene (supplemental Figure 4).

Thus, the SNPs identified in the original GWAS regulate the Wnt pathway genes in multiples cell types.

ITP-associated genes of the Wnt pathway are expressed in multiples cell types

We then assessed whether the Wnt pathway genes identified were expressed in platelets, megakaryocytes, and/or immune cells (supplemental Table 4).

We first compared the expression between platelets/megakaryocytes and lymphocytes in HCs. Among the 7 genes associated with Wnt pathway, 2 had higher expression in platelets/megakaryocytes (ABCC4 and DKK1) and 3 in lymphocytes (NKD1, NAV2, and MAP3K7; Figure 2A).

Figure 2.

scRNA-seq data in Tabula Sapiens database for the genes of interest. (A) Volcano plot of differentially expressed genes. We compared platelet and megakaryocyte clusters (n = 58 662 cells) with lymphocytes (n = 11 854 143 cells) using Tabula Sapiens differential expression tool (t test). Among 55 512 genes, 34 291 are significantly differentially expressed (adjusted P ≤ .001). We show the genes that are overexpressed in platelets/megakaryocytes in red (positive effect) and the genes that are overexpressed in lymphocytes in blue (negative effect). We show the name of the significantly differentially expressed genes associated with the sentinel and/or the proxies in at least 1 annotation tool used. (B) Expression of each gene of interest in different cell clusters in blood or bone marrow. Genes are organized by locus. We selected clusters with >500 cells, and in which at least one of our genes shows a scaled expression >0.2 in >10% of cells. Color represents the scaled expression, and the size of the dots represents the proportion of expressing cells. We highlighted platelets and megakaryocytes in blue box. We also highlighted with a green circle the clusters in which expression is >2 in at least 500 cells.

However, these genes were not tissue-specific. ABCC4 and DKK1 also had low levels of expression in several immune cells. NKD1 is expressed at a low level in a small proportion of cell lines of interest, including platelets and megakaryocytes. Although NAV2 had higher expression in lymphocytes and is not expressed in megakaryocytes, it is part of the platelet signature genes in Web-based Cell-type Specific Enrichment Analysis of Genes (WebCSEA).²⁴ MAP3K7 may also be expressed in platelets and megakaryocytes. Finally, GRIP1 is mainly expressed in a subset of lymphocytes. We found low expression in platelets, although others did not.²⁵

We then sought to replicate our scRNA-seq findings in the bone marrow of patients with ITP (Figure 3A). We confirmed that ABCC4 and DKK1 were more highly expressed in platelets/megakaryocytes than lymphocytes (fold change, 28.4 and 12.7; P < 1 × 10^–217 and 8 × 10^–134, respectively; Figure 3B). MAP3K7 was mainly expressed in platelet and B-cell precursors. The low expression of NKD1 limited the interpretation, although it was expressed in megakaryocytes and plasma cells (supplemental Figure 5). The expression of NAV2 was too low to provide a confident assessment.

Figure 3.

scRNA-seq in ITP and HCs. (A) UMAP of the scRNA-seq of the 4 bone marrow samples of patients with ITP. We show the annotation of each cell type. (B) Expression in patients with ITP of 3 genes of the Wnt pathway. Color gradients show the relative expression. (C) Differential expression of the genes of the Wnt pathway between patients with ITP and HCs in all cells (left) or in the cell type they are expressed (right). cDC, conventional dendritic dell; CDP, common dentritic cell progenitor; HSC, hematopoietic stem cell; MEP, megakaryocyte-erythroid progenitor; MHC, major histocompatibility complex; NK, natural killer; pDC, plasmacytoid dendritic cell; UMAP, Uniform manifold approximation and projection.

Thus, the Wnt genes affected by the SNPs were expressed in several immune cells and in the platelet lineage.

Wnt genes are differentially expressed in ITP

We next investigated whether the Wnt genes identified were differentially expressed in patients with ITP compared to HCs. Considering all cell types, we found that 7 of the 8 genes had significant differential expression (Figure 3C). Three (ABCC4, GRIP1, and NKD1) were upregulated, whereas 4 (MAP3K7, DKK1, NAV2, and CCDC85A) were downregulated in patients with ITP. We then compared the expression of Wnt genes with known expression levels in platelets/megakaryocytes or lymphocytes. We found the same direction of effect in these cell types as when examining expression in all cell types (Figure 3C). Two of the 5 genes (DKK1 in platelets/megakaryocytes and MAP3K7 in lymphocytes) also exhibited significant differential expression.

Finally, we compared a Wnt pathway score that considers the expression of all genes of the pathway in patients with ITP vs HCs. We found significant differences in several cell types with both directions of effect. The score was higher in patients with ITP for some cells (eg, pre-B cells) and lower in others (eg, plasma cells; supplemental Figure 6).

In sum, ITP is associated with altered expression of several Wnt pathway genes. This alteration is complex because it varies across genes and cell types.

Conclusions

Our results confirm and expand the alterations of the Wnt pathway suggested by the original GWAS.² Although we could not replicate the SNPs in adult GWAS, scRNA-seq from adult samples suggest the Wnt pathway is also dysregulated in adult ITP. The dysregulation of the Wnt pathway involves multiple genes and cell types, including the platelet lineage. Future studies should consider this complexity while deciphering the contribution of the Wnt pathway in ITP.

Conflict-of-interest disclosure: T.P. received research funding from Biossil Inc, unrelated to immune cytopenia. T.O.K. received honoraria from Sobi. A.B.G. received research funding from Novartis and consultancy fees from Sanofi. The remaining authors declare no competing financial interests.