Transcriptome profiling of megakaryocytes and platelets: Application to GP9‐ and IKZF5‐related thrombocytopenia

Abstract

Platelets are anucleate cells produced in the bone marrow and derived from large progenitor cells called megakaryocytes (MKs). Platelets receive RNA transcripts from their progenitorial MKs during thrombopoiesis. However, the correspondence between platelet and MK transcriptomes is poorly understood, particularly in the context of germline mutations that cause platelet formation defects or thrombocytopenia. We have studied the effects of two such mutations on MK and platelet transcriptomes. We generated immortalized MK cell line (imMKCL)‐based models of Bernard–Soulier syndrome and IKZF5‐related thrombocytopenia. MKs derived from imMKCLs with either a homozygous deletion of GP9 (GP9 ^−/−) or a heterozygous Y121F variant in IKZF5 (IKZF5 ^WT/Y121F) exhibited reduced proplatelet formation (reductions of 96% and 57%, respectively). Platelets from patients with either GP9 ^−/− or IKZF5 ^WT/Y121F genotypes had broad transcriptomic dysregulation, suggesting that (pro)platelet formation defects due to mutations in glycoprotein receptor and transcription factor genes such as GP9 and IKZF5 already affect the MK transcriptome. RNA‐seq data from MKs at four stages of differentiation revealed widespread but distinct changes in expression over time between the GP9 ^−/− and the IKZF5 ^WT/Y121F genotypes. Dysregulated genes in GP9 ^−/− MKs were enriched for RNA metabolism and actin/tubulin folding pathways, whereas those in IKZF5 ^WT/Y121F MKs were enriched for cell cycle pathways. Most of these genes were also dysregulated in the platelets of patients with the corresponding diseases. Our results suggest that patients with inherited forms of thrombocytopenia present with specific transcriptomic changes during platelet formation.

INTRODUCTION

Platelets play a critical role in maintaining hemostasis. They are released into the bloodstream from their precursor cells, megakaryocytes (MKs), which reside in the bone marrow. ¹ , ² In humans, approximately 10¹¹ platelets are released into circulation each day, where they remain for 7–10 days. Although MKs account for only 1 out of every 10,000 cells in the bone marrow, each MK can form up to 3000 platelets. ³ Platelets are anucleate, but contain RNA transcripts from their progenitorial MKs, which are transported during thrombopoiesis. ¹ The mean quantity of RNA per platelet in a typical blood sample is only 2.2 fg, and young platelets contain much more RNA than old platelets. ⁴ A recent transcriptomic study of platelets in healthy controls reported a mean of 7162 expressed coding genes (defined as having a mean FPKM > 0.3). ⁵ Other studies have reported that approximately 10,000 coding genes are typically expressed in platelets based on observations in tens of healthy individuals (defined as having a mean RPKM > 0.3). ⁶ , ⁷ , ⁸ However, the number of genes declared expressed in any given platelet sample is influenced by bioinformatic and experimental factors. For example, the choice of platelet isolation method is key, because some protocols are prone to including messenger RNAs (mRNAs) from small numbers of leukocytes, which contain 1000 times more RNA than platelets. ⁹ Transcriptome studies of primary human MKs require the isolation of MKs using bone marrow biopsies, which are invasive and especially challenging in thrombocytopenia patients who might have an elevated risk of bleeding. Due to the limited quantity of RNA in platelets and the ethical and technical difficulties of harvesting native MKs from human bone marrow, it has been challenging to study the transcriptomes of platelets and MKs jointly using single‐cell or bulk RNA‐seq. ¹ Consequently, the degree to which the platelet transcriptome corresponds with that of MKs in humans has remained poorly understood. Such a comparison has only been conducted in a recent study of 185 participants, which used bulk RNA‐seq of native platelets and inducible pluripotent stem cell (iPSC)‐derived MKs from the same healthy donors. ¹⁰ In that study, only 5034 genes were deemed expressed in platelets (having a median FPKM > 0.3), 91.3% of which were also deemed expressed in MKs. The use of stringent quality control filtering may explain why such a small number of genes were declared to be expressed in platelets compared to prior analyses. Importantly, the iPSC‐derived MKs used for RNA‐seq comprised 60% ± 15% CD41⁺/CD42⁺ MKs, which are known to have a more primitive hematopoiesis signature than CD34⁺ hematopoietic stem cell‐derived MKs. ¹¹ , ¹²

It is plausible that patients with genetic platelet formation defects, such as those with inherited thrombocytopenia, have platelet transcriptomes that are dysregulated directly due to the underlying genetic defect and also indirectly due to impaired platelet formation and a disturbance in RNA transport. Several studies have analyzed the transcriptomes of patient‐derived platelets to determine which pathways are disrupted in certain monogenic diseases, including thrombocytopenia. ¹ , ¹³ , ¹⁴ However, a nonrandom subset of RNAs may be transported from the MKs of patients into their platelets, where they form a highly unstable transcriptome over a platelet's life span, and this might obscure the true effects of pathogenic variants on transcription. This is particularly likely to happen in studies of inherited thrombocytopenia because many of the underlying genetic defects affect MK differentiation or proplatelet formation. ¹⁴

To characterize the transcriptomic effects of inherited platelet formation defects, we examined platelet and MK transcriptomes corresponding to two types of inherited thrombocytopenia (Figure 1). One is caused by a defect in one of the three components of glycoprotein (GP)Ib/IX/V, and would a priori not be expected to alter the expression of many genes. The other is caused by a defective transcription factor, which, due to the role of transcription factors in regulating gene expression, would a priori be expected to alter the expression of many genes. Biallelic defects in any of the three genes that encode the GPIb/IX/V complex (GP1BA, GP1BB, and GP9) cause Bernard–Soulier syndrome (BSS), which is characterized by severe macrothrombocytopenia and bleeding. ¹⁵ MKs derived from cultures of CD45⁺ cells obtained from the blood of BSS patients present with normal differentiation but almost no proplatelet formation. ¹⁶ Pathogenic heterozygous missense variants in the transcription factor gene IKZF5 cause mild thrombocytopenia. MKs derived from cultures of CD34⁺ cells obtained from the blood of IKZF5 patients show normal MK differentiation and a mild reduction in proplatelet formation. ¹⁷ We have studied the platelet transcriptomes of two related BSS patients with a homozygous GP9 deletion and four unrelated patients with pathogenic missense variants in IKZF5, including Y121F. We have opted not to study the transcriptomes of MK cultures from blood‐derived CD34⁺ or CD45⁺ cells because such cells are very heterogeneous and produce proplatelets in very low numbers. This would necessitate using a technique—single‐cell RNA‐seq—that has not yet been optimized for platelets, complicating comparisons between MKs and platelets. Instead, we introduced pathogenic variants into the immortalized MK cell line (imMKCL). ¹⁸ We and others have previously used the imMKCL to characterize genes that are etiological for thrombocytopenia and platelet function disorders. ¹⁹ , ²⁰ , ²¹ Although imMKCL‐derived MKs differentiate to mature, polyploid, and proplatelet‐forming MKs, it is still unclear how similar their transcriptomes are to those of native MKs or MKs derived from CD34⁺ hematopoietic stem cells (HSCs). Here, we characterize the function of MKs derived from the imMKCL harboring each of a homozygous GP9 deletion (GP9 ^{− / −}) and a heterozygous Y121F variant in IKZF5 (IKZF5 ^WT/Y121F). We performed bulk RNA‐seq at four differentiation stages to study expression changes over time for the two genotypes and wild‐type (WT). We also performed bulk RNA‐seq of platelets from BSS and IKZF5 patients. We then modeled gene expression in MKs and in patient‐derived platelets jointly to understand the relationships between genotypes, MKs (over differentiation stages), and platelets.

Figure 1.

Schematic of the cell types used in the transcriptomic study of a glycoprotein receptor knockout (GP9 ^{− / −}) and a transcription factor defect (IKZF5 ^WT/Y121F). Wild‐type (WT) and mutant megakaryocytes (MKs) were harvested at Days 0, 1, 4, and 5 of differentiation and compared to platelets derived from peripheral blood from study participants. imMKCL, immortalized megakaryocyte cell line.

MATERIALS AND METHODS

Genetic modification and MK differentiation of imMKCLs

We generated three GP9 ^{− / −} imMKCLs using CRISPR/Cas9 technology as previously described, ¹⁹ resulting in a 23‐bp deletion (KO1 and KO3) and a 2‐bp insertion (KO2) in GP9 (Supporting Information S1: Figure 1). We generated two heterozygous Y121F IKZF5 knock‐in imMKCLs by introducing the patient variant (chr10:122 994 678 T>A, Supporting Information S1: Figure 2). The WT imMKCL was unedited. Additional details about the genetic modifications are provided in Supporting Information S1: Supplementary Methods. We characterized the differentiation of imMKCLs to proplatelet‐forming MKs on differentiation days ranging from 0 to 6 using flow cytometry, bright field microscopy, and immunoblot analysis. We analyzed propidium iodide (PI)‐stained MKs as previously described. ¹⁹ , ²² We differentiated WT and mutant imMKCLs independently three times into proplatelet‐forming MKs and counted proplatelets on Day 6 using the Cytation5 imager by blinded assessment, as previously described. ¹⁹

Flow cytometry analysis

We analyzed patient platelets with a BD FACS Canto cytometer and imMKCL cells with a BD Symphony cytometer. We used the FCS Express software (De Novo Software) to analyze the flow cytometry data. The antibodies used are listed in Supporting Information S1: Supplementary Methods.

Immunoblot analysis

We lysed imMKCL cells to obtain protein samples as previously described. ¹⁹ , ²² We imaged blots using a ChemiDoc XRS+ imager and quantified them using Image Lab software (BioRad). The antibodies used for imaging are listed in Supporting Information S1: Supplementary Methods.

Platelet RNA sequencing and data analysis

We performed RNA‐seq of samples from study participants with approval from the UZ Leuven Ethical Committee (S63666). We obtained informed consent from all participants. We performed RNA extraction of CD45⁺ leukocyte‐depleted platelets as previously described. ¹⁷ We performed RNA‐seq of platelets from 19 healthy controls, two GP9 ^{− / −} patients, and one patient with the Y121F variant in IKZF5. The RNA‐seq data from the IKZF5 patient with the Y121F variant were analyzed jointly with previously generated data on three additional IKZF5 patients with different missense variants and 70 additional healthy controls ¹⁷ (Figure 1). We obtained gene expression estimates using MMSEQ. ²³ We performed gene set enrichment analysis (GSEA) using Reactome (v90). ²⁴ Additional details on the expression analysis of platelets are provided in Supporting Information S1: Supplementary Methods.

ImMKCL RNA sequencing and data analysis

We extracted RNA from imMKCL‐derived MKs at Days 0, 1, 4, and 5 of differentiation (Figure 1), as previously described, ¹⁷ and performed RNA‐seq. We obtained gene expression estimates using MMSEQ. ²³ We performed polytomous model selection using MMDIFF ²⁵ to compare the statistical models depicted in Supporting Information S1: Figure 3. We performed GSEA using Reactome (v90). ²⁴ Additional details on the expression analysis of imMKCL samples are provided in Supporting Information S1: Supplementary Methods.

RESULTS

Clinical and genetic description of thrombocytopenia patients

Three thrombocytopenia patients were diagnosed using a multigene panel test. ²⁶ Patients A‐II.1 and A‐II.2 are siblings from consanguineous parents with BSS due to a homozygous deletion spanning GP9 (Supporting Information S1: Figure 4). The GP9 deletion was confirmed by quantitative polymerase chain reaction (qPCR; Supporting Information S1: Figure 5). The patients present with bleeding, macrothrombocytopenia with the presence of giant platelets on blood smear analysis, and absent GPIX (CD42a) and GPIbα (CD42b) expression on platelets (Supporting Information S2: Table 1). Patient B‐II.1 carries a de novo Y121F variant in IKZF5 and presents with mild thrombocytopenia and bleeding (Supporting Information S1: Figure 6, Supporting Information S2: Table 1). HEK293 cells transfection studies showed impaired chromatin binding of IKZF5 Y121F (Supporting Information S1: Figure 7), which is consistent with observations in patients carrying other pathogenic missense variants in IKZF5. ¹⁷

Functional characterization of megakaryopoiesis in GP9 ^{− / −} and IKZF5 ^WT/Y121F MKs

Immunoblot analysis confirmed absent expression of GPIbɑ and GPIX on Day 4 differentiated MKs from the three GP9 ^{− / −} imMKCLs. As expected, Day 0 non‐differentiated MKs did not express GPIX (Figure 2A). Flow cytometry of Day 4 MKs also demonstrated complete absence of GPIX (CD42a) expression and mildly reduced ITGA2B (CD41a) expression (Figure 2B,C). The ploidy distribution was not significantly different between GP9 ^{− / −} and WT MKs (Figure 2D,E). Bright field live‐cell microscopy revealed a substantial and significant reduction (96% on average) in proplatelet formation in Day 6 GP9 ^{− / −} MK (Figure 2F,G), which was consistent with prior observations in MKs derived from the stem cells of BSS patients. ¹⁶ In the two heterozygous IKZF5 ^WT/Y121F knock‐in imMKCLs, flow cytometry analysis of Day 4 MKs did not reveal significant alterations in CD41a or CD42a expression levels compared to WT (Figure 3A,B). The distribution of MK ploidy was not significantly different in IKZF5 ^WT/Y121F MKs compared to WT MKs (Figure 3C,D). However, a mild but significant reduction (57% on average) in proplatelet formation was observed in Day 6 IKZF5 ^WT/Y121F MKs compared to WT MKs (Figure 3E,F), which is consistent with previous observations in MKs derived from the stem cells of IKZF5 thrombocytopenia patients. ¹⁷

Figure 2.

Functional characterization of GP9 ^{− / −} and wild‐type (WT) immortalized megakaryocyte cell lines (imMKCLs). (A) Western blot of WT, GP9KO1, and GP9KO2 imMKCL lysates on Day 0 and Day 4 of differentiation using antibodies for GPIbɑ (145 kDa), GPIX (22 kDa), and glyceraldehyde‐3‐phosphate‐dehydrogenase (GAPDH) (36 kDa) as a loading control. Dead or apoptotic cells were excluded using 4′,6‐diamidino‐2‐phenylindole (DAPI) staining. (B) Representative flow cytometry measurements of WT and GP9 ^{− /−} megakaryocytes (MKs) using markers for CD41a and CD42a. (C) Box plots showing the fraction of positive cells for markers CD41a and CD42a in the two GP9 ^{− / −} Day 4 differentiated MKs and in Day 4 differentiated WT MKs. (D) Histogram of fluorescent intensity measured by flow cytometry of propidium iodide (PI) stained Day 5 differentiated WT and GP9 ^{− / −} MKs and quantification of peaks corresponding to different levels of ploidy. Dead or apoptotic cells were excluded using DAPI staining. (E) Relative proportions of ploidy states in WT, GP9KO1, and GP9KO2 MKs on Day 5 of differentiation measured in duplicate. The vectors of proportions did not differ significantly between the cell lines (all P > 0.05, pairwise comparisons of Dirichlet regression models using likelihood ratio tests [LRTs]). (F) Representative images of proplatelet‐forming WT and GP9 ^{− / −} MKs on Day 6 of differentiation. (G) Fraction of proplatelet‐forming WT and GP9 ^{− / −} MKs obtained from three wells and three separate differentiations on Day 6 of differentiation.

Figure 3.

Functional characterization of IKZF5 ^WT/Y121F and wild‐type (WT) immortalized megakaryocyte cell lines (imMKCLs). (A) Representative flow cytometry measurements of WT and Y121FHet megakaryocytes (MKs) using markers for CD41a and CD42a. Dead or apoptotic cells were excluded using 4′,6‐diamidino‐2‐phenylindole (DAPI) staining. (B) Box plots showing the fraction of positive cells for markers CD41a and CD42a in the two Y121FHet Day 4 differentiated MKs and in Day 4 differentiated WT MKs. The mean fractions did not differ by cell line (all P > 0.05, Wilcoxon rank sum test between pairs of vectors). (C) Histogram of fluorescent intensity measured by flow cytometry of propidium iodide (PI) stained Day 5 differentiated WT and IKZF5 ^WT/Y121F MKs and quantification of peaks corresponding to different levels of ploidy. Dead or apoptotic cells were excluded using DAPI staining. (D) Relative proportions of ploidy states in WT, Y121FHet1, and Y121FHet2 MKs on Day 5 of differentiation measured in duplicate. The vectors of proportions did not differ significantly between the cell lines (all P > 0.05, pairwise comparisons of Dirichlet regression models using likelihood ratio tests [LRTs]). (E) Representative images of proplatelet‐forming WT and Y121FHet MKs on Day 6 of differentiation. (F) Fraction of proplatelet‐forming WT and Y121FHet MKs obtained from three wells and three separate differentiations on Day 6 of differentiation.

GP9 ^{− / −} and IKZF5 variants have broad transcriptomic effects in platelets

RNA‐seq and differential gene expression analysis of platelets comparing two GP9 ^{− / −} patients (A‐II.1 and A‐II.2) with 19 healthy unrelated controls revealed 1540 differentially expressed genes (DEGs), of which 325 were significantly overexpressed and 1215 were significantly underexpressed in GP9 patients (posterior probability > 0.95, |log_e fold change| > 0.5) (Supporting Information S1: Figures 8 and 9). Given that only one study participant harbors the IKZF5 variant Y121F, we supplemented newly generated RNA‐seq data on platelets from that patient (B‐II.1) and the 19 controls with previously published platelet RNA‐seq data from three IKZF5 thrombocytopenia patients (C‐II.1, D‐II.1, and E‐III.3) and 70 controls. ¹⁷ We declared 3541 DEGs, of which 1379 were significantly overexpressed and 2162 significantly underexpressed in IKZF5 patients (posterior probability > 0.95, |log_e fold change| > 0.5) (Supporting Information S1: Figure 9). Broad transcriptome alterations were expected in the context of a platelet defect caused by a defective transcription factor, but not in the context of an absent GP reception. Our results suggest the profound proplatelet‐formation defect present in GP9 ^{− / −} MK may affect the transfer of transcripts into platelets. GSEA using Reactome revealed multiple, partially overlapping pathways to be significantly associated with up‐ and downregulated genes in each of the conditions (Supporting Information S1: Figure 10). The pathway “Platelet activation, signaling and aggregation” was the most significantly enriched pathway amongst each of the sets of downregulated DEGs identified in GP9 and IKZF5 patients. However, given the possibility of defective transcript transfer in these conditions, enriched pathways in platelets may not reflect the etiological defects in the corresponding MKs. Though RNA transfer from MKs to proplatelets has never been studied in detail, we detected total RNA in the nucleus, cytoplasm but also in the proplatelets of Day 5 WT MKs (Supporting Information S1: Figure 11).

Transcriptome‐wide analysis of WT, GP9 ^{− /−}, and IKZF5 ^WT/Y121F MKs and platelets

We analyzed the transcriptional profiles of WT and mutant MKs by performing RNA‐seq at differentiation Days 0 (non‐differentiated imMKCLs), 1 (early differentiation stage), and 4 and 5 (both late differentiation stages). We refer to Day 0 and Day 1 MKs as early MKs and Day 4 and Day 5 MKs as late MKs. Volcano plots of differential gene expression on each day are shown in Supporting Information S1: Figure 9. Principal component analysis (PCA) of normalized gene expression estimates of the 500 most variable genes among WT and mutant MKs showed that the first two principal components discriminate between different stages of differentiation and between mutations (Figure 4A). Projection of WT platelet transcriptomes onto the plane formed by the two first principal components placed the platelets in a region flanked by Day 1 and Day 4 MKs, suggesting that platelet transcriptomes from healthy individuals resemble Day 2–3 MKs. Platelets from thrombocytopenia patients, on the other hand, clustered more closely with early MKs, suggesting they tend to receive a less mature transcriptome from MKs. Of the 500 most variable genes, the subset with a positive loading for the first principal component (i.e., those contributing to the the left‐to‐right repositioning towards maturity in Figure 4A) were enriched for pathways related to MK development (e.g., “Hemostasis”) and platelet function (e.g., “Platelet activation, signaling and aggregation,” “Response to elevated platelet cytosolic Ca²⁺,” “Platelet degranulation”) (Figure 4B). Furthermore, 26 of the 27 genes that had previously been reported to be positive or negative markers of MK maturity in an MK differentiation study of blood‐derived CD34⁺ stem cells exhibited changes in expression between early and late MKs that were consistent with those identified in that study ²⁷ (Supporting Information S1: Figure 12). This suggests that MKs derived from stem cells and from imMKCLs share some common gene expression signatures.

Figure 4.

Principal component analysis (PCA) of megakaryocyte (MK) and platelet transcriptomes. (A) Scatter plot of the first two principal components obtained by performing PCA on the 500 genes with the greatest variability of expression in wild‐type (WT) and mutant MKs. Control and patient platelet transcriptomes were projected onto the plane. (B) Gene set enrichment analysis of the genes contributing to the first principal component with a positive loading. The P‐values corresponding to the 20 most significantly enriched pathways are shown as bars. The number of genes overlapping each pathway is shown in brackets. ECM, extra cellular matrix; imMK, immortalized megakaryocyte; VWF, von Willebrand factor.

GP9 ^{− / −} and IKZF5 ^WT/Y121F have distinct maturity‐dependent effects on MK transcriptomes

For each of the two mutant genotypes, GP9 ^{− / −} and IKZF5 ^WT/Y121F, we used a Bayesian regression model selection ²³ , ²⁵ approach to classify each gene into one of five general patterns of expression in MKs over differentiation time (early vs. late) and genotype (WT vs. mutant) (Supporting Information S1: Supplementary Methods, Supporting Information S1: Figure 3). The baseline model, Model 0, contained an intercept column corresponding to a global mean expression parameter. In addition to the intercept column, Model 1 contained a covariate corresponding to a parameter representing the difference in expression between early and late MKs; Model 2 included a covariate corresponding to a parameter representing the difference in expression between WT and mutant MKs; Model 3 contained both of the additional covariates in Models 1 and 2; and Model 4 contained the additional covariate in Model 2 and two covariates each corresponding to a parameter representing the difference in expression between early and late stages in either WT or mutant MKs. Genes that were initially classified into Models 1–4 but had effect sizes below biologically meaningful thresholds (see 2) were excluded from downstream analyses.

Independent of the mutant cell line, over 80% of ≈57,000 genes in our downstream analyses were assigned to Model 0 (i.e., they were not declared differentially expressed). These groups included genes with very low mean expression (Supporting Information S1: Figure 13). Approximately 8% of genes (4652 and 5017 in GP9 ^{− / −} and IKZF5 ^WT/Y121F analyses, respectively) were assigned to Model 1 (i.e., expression changed over differentiation time independently of genotype) (Figures 5 and 6). Overall, the GP9 ^{− / −} genotype affected the expression of 2429 genes (3.9% of the total) (Models 2–4, Figure 5, Supporting Information S2: Table 2): 119 had constant levels of expression throughout MK differentiation that differed between genotypes (Model 2); 469 had similar changes throughout MK differentiation but at different levels of expression between genotypes (Model 3); and most genes (1841) exhibited different expression changes throughout MK differentiation between the two genotypes (Model 4). In contrast, the IKZF5 ^WT/Y121F genotype affected the expression of 1053 genes (1.8% of the total) (Models 2–4, Figure 6, Supporting Information S2: Table 3): only 187 were assigned to Model 2; 752 were assigned to Model 3; and only 114 were assigned to Model 4. Our results suggest that mutation of a GP (in this case, GPIX) can have a greater impact on the MK transcriptome than mutation of a transcription factor such as IKZF5. This might explain why GP9 ^{− / −} MKs have a more severe proplatelet‐formation defect than IKZF5 ^WT/Y121F MKs (Figures 2F,G and 3E,G). Furthermore, the defect in IKZF5, which is a constitutively active transcription factor, tends to cause expression changes that are stable over differentiation time relative to WT (consistent with Model 3). In contrast, the defect in GP9 tends to cause differentiation time‐dependent changes in expression (consistent with Model 4), possibly because expression of the GPIb‐IX‐V receptor complex increases as MKs mature.

Figure 5.

Polytomous model selection among five statistical models capturing different patterns of gene expression in GP9 ^{− / −} and wild‐type (WT) megakaryocytes (MKs) (as described in Supporting Information S1: Figure 3 ). For each gene, a model is selected based on posterior probability and estimated effect size (|log_e fold change|) thresholds, which were set to 0.7 for time and 0.5 for genotype. The first two columns depict weighted log₂ fold changes of effect time and genotype. The third column shows the number of genes attributed to this model. The fourth column shows the log expression parameter “mu” in the early and late stages of differentiation and the estimated regression line for WT and GP9 ^{− / −} MKs.

Figure 6.

Polytomous model selection among five statistical models capturing different patterns of gene expression in IKZF5 ^WT/Y121F and wild‐type (WT) megakaryocytes (MKs) (as described in Supporting Information S1: Figure 3 ). For each gene, a model is selected based on posterior probability and estimated effect size (|log_e fold change|) thresholds, which were set to 0.7 for time and 0.5 for genotype. The first two columns depict weighted log₂ fold changes of effect time and genotype. The third column shows the number of genes assigned to this model. The fourth column shows the log expression parameter “mu” in the early and late stages of differentiation and the estimated regression line for WT and IKZF5 ^WT/Y121F MKs.

For each of the two genotypes, GP9 ^{− / −} and IKZF5 ^WT/Y121F, we performed GSEA on each of the 14 gene sets identified by assigning genes to Models 2–4 and partitioning each gene according to its pattern of expression in its assigned model (combination of signs for the regression coefficients of the fitted model). Ten of the gene sets were significantly enriched in at least one pathway (false discovery rate [FDR] < 0.05) with which the overlap included at least three genes (Figure 7A–C, Supporting Information S1: Figures 14 and 15). Of these gene lists, three exhibited especially strong pathway enrichments (FDR < 1e−4) (Figure 7A–C). Genes that were downregulated specifically in early or late megakaryopoiesis due to GP9 ^{− / −} (Model 4) were strongly and significantly enriched for the RNA metabolism and the tubulin and actin folding pathways (Figure 7A,B). Genes that had the same fold change in early and late megakaryopoiesis due to IKZF5 ^WT/Y121F (Model 3) were strongly and significantly enriched for the cell cycle pathway (Figure 7C). We confirmed the downregulation of three genes involved in tubulin and actin folding and the cell cycle pathway by quantitative reverse transcription polymerase chain reaction (RT‐qPCR) of MK RNA from three independent differentiation experiments (Supporting Information S1: Figure 16). The cell cycle defect was validated in IKZF5 ^WT/Y121F MKs by quantifying cells at different stages of the cell cycle, which showed a strong reduction in the relative number of cells in the G2 and S phases (Figure 7D,E).

Figure 7.

(A) Gene set enrichment analysis for downregulated genes in early stage maturation of GP9^{− / −} megakaryocytes (MKs). Pathways with false discovery rate (FDR) < 0.05 and a minimum of three genes overlapping with those in the pathways (indicated between brackets) are shown. Scatter plot on the right panel shows corresponding log₂ fold changes in early‐stage MKs (mutant vs. wild‐type) compared with log₂ fold changes in platelets (patients vs. controls) of genes involved in “metabolism of RNA” pathways (including ribosomal RNA [rRNA] processing), marked with an asterisk. (B) Gene set enrichment analysis for downregulated genes in late‐stage maturation of GP9 ^{− / −} MKs. Pathways with FDR < 0.05 and a minimum of three genes overlapping with those in the pathways (indicated between brackets) are shown. Scatter plot on the right panel shows corresponding log₂ fold changes in late‐stage MKs (mutant vs. wild‐type) compared with log₂ fold changes in platelets (patients vs. controls) of genes involved in “metabolism of RNA” and actin/tubulin folding pathways, marked with an asterisk. (C) Gene set enrichment analysis for genes downregulated in time and by genotype in IKZF5 ^WT/Y121F MKs. The top 20 pathways with FDR < 0.05 and a minimum of three genes overlapping with those in the pathways (indicated between brackets) are shown. Scatter plot on the right panel shows the corresponding log₂ fold changes in MKs (mutant vs. wild‐type) compared with log₂ fold changes in platelets (patients vs. controls) of genes involved in cell cycle pathways, marked with an asterisk. (D) Histogram of fluorescent intensity measured by flow cytometry of propidium iodide (PI) stained Day 1 differentiated WT and IKZF5 ^WT/Y121F MKs and quantification of peaks corresponding to different levels of ploidy. Dead or apoptotic cells were excluded using 4′,6‐diamidino‐2‐phenylindole (DAPI) staining. (E) The percentage of MKs in G1, S, and G2 cell cycle stages quantified through flow cytometry and PI staining on Day 1 differentiated MKs (pre‐endoreplication stage). This experiment was performed in three independent differentiation replicates. P‐values are shown for pairwise t‐tests between WT and mutant conditions. iMK, immortalized megakaryocyte; tRNA, transfer RNA.

Mirroring between GP9 ^{− / −} and IKZF5 ^WT/Y121F MKs and platelets from patients

We analyzed the platelet expression of genes assigned to the most highly represented models in MKs that were also enriched for specific pathways (Figure 7A–C, Supporting Information S2: Tables 4 and 5). For example, 33 of the 318 genes downregulated in GP9 ^{− / −} MKs early in megakaryopoiesis are involved in RNA metabolism. Of these 33 downregulated genes, 30 are downregulated in the platelets of Bernard–Soulier patients (Figure 7A). Similarly, 52 of the 289 genes downregulated in GP9 ^{− / −} MKs late in megakaryopoiesis are involved in RNA metabolism, and 50 are downregulated in patient platelets. Additionally, five of the 289 genes are involved in tubulin and actin folding, of which all but one are downregulated in patient platelets (Figure 7B). The MK defect in Bernard–Soulier platelets therefore appears broadly reflected in platelet transcriptomes. In contrast, of the 235 genes declared downregulated over differentiation time that also exhibit time‐independent downregulation in IKZF5 ^WT/Y121F MKs, 64 are involved in the cell cycle. However, only 40 of these genes are downregulated in the platelets of patients with IKZF5‐related thrombocytopenia (Figure 7C). Here, higher expression of some genes in immature MKs (the transcriptomes of which resemble those of mutant platelets, Figure 4A) relative to mature MKs (the transcriptomes of which resemble those of WT platelets) compensates for the time‐independent downregulation caused by mutation of IKZF5 (Supporting Information S1: Figure 17).

DISCUSSION

Platelet RNAs are produced through transcription of nuclear DNA in MKs, which can contain up to 10,000 different protein‐coding transcripts. ⁶ However, due to ethical concerns, there exist no comparisons between the transcriptomes of platelets and native bone marrow‐derived MKs. Our study compared transcriptomes of platelets and MKs throughout megakaryopoiesis under healthy conditions and under genetic alterations that cause thrombocytopenia. To overcome the difficulties of obtaining bone marrow‐derived MKs, we used genetically engineered MKs derived from the imMKCL.

Transcriptomic analysis of imMKCL‐derived MKs recapitulated the expression changes of differentiation‐specific marker genes previously identified in blood stem cell‐derived MKs, validating the imMKCL as a model system (Supporting Information S1: Figure 12). ²⁷ MKs derived from GP9 ^{− / −} and IKZF5 ^WT/Y121F imMKCLs mimicked the proplatelet‐formation defect previously reported in blood stem cell‐derived MKs from thrombocytopenia patients with defects in these genes. ¹⁶ , ¹⁷ Our PCA analysis of MK and platelet transcriptomes (Figure 4A,B) showed that the makeup of transcripts in platelets resembles that of MKs of intermediate maturity rather than that of mature proplatelet‐forming MKs. Platelets from thrombocytopenia patients appear to have transcriptomes that resemble those of immature MKs.

Only 10% of circulating platelets (the youngest subpopulation) contain considerable quantities of RNA. ⁴ Inhibition of thrombopoiesis by a genetic defect may impair MK transcriptomes, which would explain the surprisingly broad transcriptomic alterations we observed in GP9‐ and IKZF5‐related thrombocytopenia patient platelets. Therefore, transcriptomic studies of platelet disorders that only include platelets might fail to distinguish alterations due to impaired megakaryopoiesis from those that are due to impaired MK maturation or even the transfer of transcripts. The latter hypothesis should be studied in more detail in further studies. We performed a longitudinal analysis of imMKCLs with WT, GP9 ^{− / −}, and IKZF5 ^WT/Y121F genotypes to investigate the effects of genetic changes over differentiation time. We categorized genes based on their patterns of expression in MKs. Although many dysregulated genes in the most biologically relevant enriched pathways were similarly dysregulated in platelet transcriptomes with corresponding genetic defects, the pathways could not be identifiable as enriched using platelet transcriptomics alone (Supporting Information S1: Figure 10). The largest set of classified genes in GP9 ^{− / −} MKs corresponded to expression changes in the late differentiation stage. Amongst these genes, those that were downregulated in late MKs were enriched for pathways related to RNA metabolism and the folding of actin and tubulin by the CCT/TriC complex (Figure 7B). Several genetic defects (in ACTN1, FLNA, DIAPH1, and TUBB1) affect the actin and tubulin MK cytoskeleton and have been implicated in the etiologies of macrothrombocytopenia. ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² The GPIb‐IX‐V receptor complex is thought to determine the structure of the submembranous actin network through binding of amino acids 556–577 in the central region of the cytoplasmic domain of GPIbα and Ig‐like repeat number 17 in the C‐terminal half of filamin A. ³³ A defect in this receptor complex has previously been reported to cause macrothrombocytopenia. ³⁴ Eukaryotic group II chaperonin TRiC/CCT has been shown to regulate the folding of key cytoskeletal proteins actin and tubulin ³⁵ , ³⁶ but also the mitotic regulator CDC20 ³⁷ and other cytosolic proteins. The role of TRiC/CCT has not been studied in platelets or MKs but Cct3 (the ortholog of one of the genes identified in our analysis, Supporting Information S2: Table 4) depleted mice have an increased mean platelet volume (MGI:104708).

The largest set of classified genes in IKZF5 ^WT/Y121F MKs corresponded to time‐independent expression changes due to the mutation in genes with expression levels that change over differentiation time. Amongst these genes, those downregulated over differentiation time and also due to the mutation were enriched for cell cycle pathways (Figure 7C). This defect was confirmed through quantification of WT and IKZF5 ^WT/Y121F MKs at each of the different phases of the cell cycle (Supporting Information S1: Figure 17). The role of the cell cycle in megakaryopoiesis and platelet production is not well understood, although it is known that MKs undergo polyploidization by endomitosis to increase their size. ³⁸ IKZF5 ^WT/Y121F MKs, however, have normal polyploidy. Defective cell cycle regulation might play a role through mechanisms such as in MASTL (microtubule‐associated serine/threonine protein kinase–like) related thrombocytopenia. ³⁹ MASTL is a Ser/Thr kinase that inhibits PP2A‐B55 complexes during mitosis ⁴⁰ but the missense variant associated with thrombocytopenia has a “gain‐of‐function” effect by increased phosphorylation of Cdk and PP2A substrates and overstabilization of actin cytoskeleton. The IKZF5 transcription factor binding motif is “GNNTGTNG,” ¹⁷ which bears resemblance to those of KLF9 and RUNX1 (P‐values of 1.00 × 10^{− 3} and 1.89 × 10^{− 3}, respectively, for observed similarity being due to chance ⁴¹ ). Dominant RUNX1 variants are known to cause thrombocytopenia, and out of 170 genes that were reported as being most downregulated in RUNX1‐deficient platelets, ⁴² 138 were also downregulated in platelets from the IKZF5 patients with a posterior probability > 0.5 (data not shown). The downregulation of the cell cycle pathway as observed in IKZF5 ^WT/Y121F megakaryopoiesis, is also altered in RUNX1‐deficient patients' platelets. RUNX1 was shown to regulate the cell cycle by shortening the G1/S phase in hematopoietic cells through the binding and induction of cyclin D promoters. ⁴³ , ⁴⁴ , ⁴⁵ There is also phenotypic similarity between both thrombocytopenia types with patients exhibiting a mild bleeding phenotype and dense granule deficiency but IKZF5 deficiency is not known to result in hematological malignancy.

Our study includes a limited number of patients and MK models, with only a single causal variant for each condition. Future work should validate the transcriptome changes in platelet samples from additional patients and use other MK models, including imMKCL‐derived MKs with different IKZF5 and GP9 variants but also using CD34⁺ HSC‐derived MKs obtained from the blood of patients. In addition, our work has used the imMKCL to create inherited thrombocytopenia models, but it is still not clear how well they represent the defective megakaryopoiesis that occurs in the bone marrow of patients. Nevertheless, our study highlights the merit of having this transcriptomic model for the validation of variants and elucidating disease mechanics in thrombocytopenia during megakaryopoiesis and thrombopoiesis.

AUTHOR CONTRIBUTIONS

Koenraad De Wispelaere: Conceptualization; investigation; writing—original draft; writing—review and editing; visualization; validation; software; formal analysis; data curation; methodology; funding acquisition; project administration. Fabienne Ver Donck: Writing—review and editing; investigation. Kato Ramaekers: Writing—review and editing; investigation. Chantal Thys: Writing—review and editing; investigation; validation. Koji Eto: Writing—review and editing; resources. Veerle Labarque: Writing—review and editing; funding acquisition; resources; project administration. Ernest Turro: Writing—review and editing; writing—original draft; funding acquisition; resources; software; methodology; conceptualization; project administration; supervision. Kathleen Freson: Writing—review and editing; writing—original draft; funding acquisition; resources; conceptualization; methodology; project administration; supervision.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICS STATEMENT

We performed RNA‐seq of samples from study participants with approval from the UZ Leuven Ethical Committee (S63666). We obtained informed consent from all participants.

FUNDING

K.F. is supported by KU Leuven BOF grant C14/23/121, FWO grant G072921N, and an unrestricted research grant from Swedish Orphan Biovitrum AB (SOBI). K.F. and E.T. are supported by NIH grant R01 Hl161365. K.D.W. received a fellowship from the Belgian American Education Foundation (BAEF). University Hospitals Leuven are member of the European Reference Network on Rare Haematological Diseases (ERN‐EuroBloodNet)‐Project ID No 101157011. ERN‐EuroBloodNet is funded by the European Union within the framework of the Fourth EU Health Program.

Supporting information

Supporting Information.

Supporting Information.

DATA AVAILABILITY STATEMENT

The data that support the findings of this study is openly available in the European Genome Archive (EGA). EGA data accession codes for the fastq files and additional analysis files are accessible at https://github.com/kdewispelaere/10.1002-hem3.70217.