The insertion/deletion polymorphism rs201494641 at ITGA9 influences blood CD34⁺ cell levels by altering ZNF384 binding

Stem cell transplantation is a cornerstone in the treatment of blood malignancies. The most common method for harvesting stem cells for transplantation is by leukapheresis. This requires the mobilization of CD34⁺ hematopoietic stem and progenitor cells (HSPC) from the bone marrow into the blood. Understanding the genetic factors that influence blood CD34⁺ cell levels could reveal previously unappreciated genes and mechanisms that control HSPC behavior in humans, as well as potential new drug targets for HSPC mobilization.

Recently, we reported the first large-scale genome-wide association study on blood CD34⁺ cell levels.¹ Across 13,167 individuals, we identified 11 independent genetic associations. One of the most significant associations maps to the ITGA9 locus at chromosome 3p22 (lead variant rs201494641:TTT>T; P=4.7×10-¹¹; |3=-0.123). The ITGA9 protein is the a subunit of the α9|31 integrin receptor, which has been reported to modulate HSPC growth, differentiation, and retention within the bone marrow by interacting with various ligands, like fibronectin,² tenascin-C,³ and osteopontin.⁴ The 3p22 association with blood CD34⁺ cell levels is represented by 46 variants in high linkage disequilibrium (LD) (r²>0.8) with the lead variant. All of these variants are located in ITGA9 intron 3 and 4 (Figure 1A). However, the causal variants and their mechanisms of action remain unknown. We therefore sought to dissect functionally the association at 3p22 between ITGA9 and blood CD34⁺ cell levels.

To identify causal variants, we integrated ATAC-sequencing (ATAC-seq) and mRNA-sequencing data for sorted blood cell types.^5,6 We found a strong positive correlation between ITGA9 expression and chromatin accessibility in an approximately 1,300 bp-long segment in ITGA9 intron 3 (Figure 1A).¹ Notably, this segment is selectively accessible in CD34⁺ blood cell populations, including hematopoietic stem cells (HSC), multi-potent progenitors, common myeloid progenitors, and megakaryocyte-erythroid progenitors (Figure 1B). The segment encompasses four variants within the ITGA9 LD block: three single nucleotide polymorphisms (rs73053290, rs17227369 and rs17227404) and one insertion-deletion polymorphism (rs201494641; Figure 1B). Using promoter capture Hi-C data for primary CD34⁺ cells,⁷ we detected a chromatin looping interaction between these four variants and the ITGA9 promoter (Figure 1A).¹ Consistent with a gene-regulatory effect, analysis of mRNA-sequencing data for CD34⁺ cells from 155 blood donors showed association between rs17227369 and ITGA9 mRNA levels in blood CD34⁺ cells, with the minor allele yielding lower expression (linear regression P=2.0×10-¹¹; Figure 1C).¹ To investigate if this effect translates to the protein level, we quantified ITGA9 surface expression on CD34⁺ cells in 458 blood donors by flow cytometry, observing a significant association in the same direction (linear regression P=8.1×10-¹⁵; Figure 1D; Online Supplementary Figure S1A, B).

To confirm the regulatory role of the identified segment on ITGA9 expression, we used dual single-guide RNA (sgRNA) CRISPR/Cas9 genome editing⁸ to delete a 486-bp region harboring the four candidate variants (Online Supplementary Table S1) in the human erythroleukemia HEL cells, which show an HSPC-like transcriptional profile and are homozygous for the major alleles of the four variants of interest.⁹ This led to the downregulation of ITGA9, further supporting a regulatory role (Figure 2A). To assess the transcriptional activity of each of the four candidate variants, we carried out luciferase experiments with constructs representing their reference and alternative alleles in HEL cells (Online Supplementary Table S2). We observed higher transcriptional activity with rs201494641-TTT construct than with rs201494641-T construct (one-sided Student’s t test P=2.8x10-³; Figure 2B), consistent with the direction of the effects on ITGA9 transcript and protein levels (Figure 1C, D). Similarly, we detected allele-dependent accessibility at rs201494641 (14 vs. 7 reads containing the TTT and T alleles, respectively; Binomial test P=2x10-²) but not at the other three variants in ATAC-seq data for the acute myeloid leukemia cell line MUTZ-3, which is heterozygous for all four variants of interest. We also noted a DNAase I footprint at rs201494641 in CD34⁺ cells (Figure 2C).¹⁰ Collectively, these data identify rs201494641 as a likely causal regulatory variant underlying the 3p22 association with blood CD34⁺ cell levels.

Further, we searched for differentially binding transcription factors using the FABIAN tool.¹¹ The strongest differential binding score was seen for the zinc finger protein ZNF384, which binds the rs201494641-harboring region (Figure 2C).¹² FABIAN predicted higher binding affinity for the minor (T) compared to the major (TTT) allele (Figure 2D). The ZNF384 core binding motif is a poly-A/poly-T sequence (Figure 2E),¹³ whose length is affected by rs201494641. Small interfering RNA (siRNA)-mediated knockdown of ZNF384 yielded up-regulation of ITGA9 in HEL cells (one-sided Student’s t test P=1.0×10-²; Figure 2F). Additionally, we observed reduced DNAase I accessibility across the repetitive poly-A/poly-T sequence as well as the flanking regions (Figure 2G). Collectively, these observations are consistent with ZNF384 acting as a transcriptional repressor, preferentially binding the minor ITGA9-low-expressing allele rs201494641-T.

In conclusion, we functionally dissected the genetic association between ITGA9 and blood CD34⁺ cell levels. We show that the association maps to a regulatory region in ITGA9 intron 3, and identify rs201494641:TTT>T as a likely causal variant. Our data are consistent with rs201494641:TTT>T increasing the affinity of the zinc finger protein ZNF384, which represses ITGA9 transcription. Previously, ZNF384 has been reported to undergo somatic rearrangements in B-cell precursor acute lymphoblastic leukemia, including gene fusions with more than ten distinct partner genes, including TCF3, EP300, TAF15, and CREBBP¹⁴. However, its precise role in hematopoiesis remains unexplored. In summary, our findings provide new insight into the genetic factors that influence blood CD34⁺ cell levels and implicate ITGA9 as a regulator of circulating HSPC levels in humans.

Figure 1.

Functional fine-mapping of the 3p22/ITGA9 locus.¹ (A) Top: close-up of the 3p22 signal. Middle: chromatin looping interactions in CD34⁺ cells with standard and internal promoter (blue arches; y-axis indicates PCHi-C P-score⁷). Bottom: co-accessibility plot showing positive correlation (blue peak) between ITGA9 expression across sorted blood cell populations and ATAC-sequencing signal (100-bp sliding window) (y-axis indicates false discovery rate for Pearson correlation).¹⁵ (B) Four credible set variants map to the identified segment, which is accessible in hematopoietic stem cells (HSC), multipotent progenitors (MPP), lympho-myeloid primed progenitors (LMPP), common myeloid progenitors (CMP), granulocyte-monocyte progenitors (GMP) and megakaryocyte-erythroid progenitors (MEP) (y-axis indicates ATAC-sequencing signal). (C) Correlation between rs17227369 genotype (credible set proxy for rs201494641 used for TaqMan genotyping) and ITGA9 mRNA levels (residual FPKM).¹ (D) Correlation between rs17227369 genotype and ITGA9 protein levels (median fluorescent intensity [MFI], quantified using phycoerythrin (PE)-conjugated monoclonal antibody towards ITGA9. Statistics are for Pearson correlation, ***P≤0.001.

Figure 2.

ZNF384 preferential binding to rs201494641-T repressing ITGA9. (A) ITGA9 expression in HEL cells subjected to CRISPR/Cas9 editing with a non-targeting single-guide RNA (sgRNA) pair control (Ctrl, 60-bp cut, 11 biological replicates), or an sgRNA pair designed to delete the region harboring the 4 putative causal variants at 3p22 (CRISPR, 486-bp cut) (Online Supplementary Table S2); one-sided t test, ***P≤0.001. (B) Luciferase activities of the 4 candidate causal variants in HEL cells (4 biological replicates). Data normalized to empty vector control; one-sided t test, ***P≤0.001, *P≤0.05, NS: not significant. (C) Upper panel: chromatin accessibility (ATAC-sequencing signal intensity) across different blood cell types (colors as in Figure 1B) in the approximately 1,300-bp wide region in ITGA9 intron 3; lower panel: DNAse I footprint in primary CD34⁺ cells and chromatin immunoprecipitation sequencing (CHIP-seq) signals in K562 cells from ReMap in the rs201494641-harboring region. (D) ZNF384 gain of binding to rs201494641-T from the Fabian-variant database. (E) JASPAR detailed Transcription Factor Flexible Model TFFM0157.1 showing the ZNF384 consensus motif. (F) ZNF384 and ITGA9 expression in HEL cells transfected with Ctrl or ZNF384 small interfering RNA (siRNA) (4 biological replicates); one-sided t test, ***P≤0.001, *P≤0.05. (G) ENCODE DNAse I consensus footprint data for the ZNF384 motif in primary CD34⁺ cells, showing mean cleavage ratios in a 200-bp window centered on the core poly-T motif (dotted lines), for 2 different repeat lengths. In both cases the DNA binding footprint (solid lines) extends beyond the core poly-T stretch, suggesting that ZNF384 directly binds flanking sequences on either side of the repeat sequence. a.u.: arbitrary units; mRNA: messenger RNA; HSC: hematopoietic stem cells.

Funding Statement

Funding: This work was supported by grants from the European Research Council (CoG-770992), the Swedish Research Council (2017-02023 and 2018-00424), the Swedish Cancer Society (23 2851 Pj), the Swedish Children’s Cancer Fund (PR2023-0067 and PR2020-0056), Inga-Britt and Arne Lundberg’s Foundation (2017-0055).

Data-sharing statement

ATAC-seq raw data deposited in Sequence Read Archive, accession number PRJNA1040035.