Developmental differences in IFN signaling affect GATA1s-induced megakaryocyte hyperproliferation (original) (raw)
Prospective isolation of FL-MkPs and BM-MkPs. Since culturing of MkPs from FL or BM could potentially alter important gene expression differences, we performed our analysis directly on fluorescence-activated cell sorted (FACS) MkPs. E13.5 was chosen as a gestational time point to assess FL-MkPs, since the hyperproliferative phenotype of GATA1s Mks is apparent at this stage (21). We began our studies with WT mice since the developmental stage–specific hyperproliferation of FL-MkPs from GATA1s mice might confound the initial gene expression analysis. Pronk et al. reported that the immunophenotype Lin–Sca-1–c-KIT+CD150+CD41+ greatly enriches for committed MkPs from mouse BM (22). We used this set of cell surface markers to isolate MkPs from WT E13.5 mouse FL and adult BM (Figure 1, A and B). There were no significant morphologic differences based on May-Grünwald-Giemsa staining between cells sorted from FL versus BM (Figure 1, A and B). Culturing of the sorted cell populations in semisolid media containing cytokines supporting multilineage growth showed that greater than 95% of sorted cells derived from both sources gave rise to pure Mk colonies (Figure 1C). The unsorted starting population gave rise to multiple colony types, as expected. There were subtle morphological differences between Mk colonies derived from FL-MkPs versus BM-MkPs, with the former appearing somewhat larger and more light refractive than the latter (Figure 1D), although the significance of this remains uncertain. These findings indicate that the immunophenotype Lin–Sca-1–c-KIT+CD150+CD41+ markedly enriches for FL-MkPs similar to that reported for BM-MkPs and that there was minimal contamination with myeloid progenitor cells in our sorted samples.
Flow cytometric sorting of MkPs from adult BM and E13.5 FL. (A and B) FACS plots and gates used for cell sorting. Cells were stained with antibodies against lineage markers, c-KIT, Sca-1, CD41, and CD150. Dead cells were identified by staining with DAPI and were excluded. May-Grünwald-Giemsa stains of cytospun freshly sorted BM and FL MkPs are also shown (original magnification, ×1,000). (A) Percent BM-MkPs gated relative to the starting population was as follows: P3, 17%; P4, 16.5%; P5, 1.9%. (B) Percent FL-MkPs gated relative to the starting population was as follows: P3, 38.2%; P4, 24.4%; P5, 1.2%. (C) Colony-forming assays. Percentage of colony type from sorted or unsorted BM and FL cells cultured in semisolid medium (after red blood cell lysis) containing TPO, erythropoietin, stem cell factor, IL-3, IL-11, and GM-CSF. Colonies were enumerated after 8 days. Cell accumulations of 3 or more cells were considered a colony. (D) Representative Mk colonies (unstained) from BM (day 7 of culture) or FL (day 8 of culture). Original magnification, ×100.
Upregulation of IFN-α/β–inducible genes in FL-MkPs and BM-MkPs. Global gene expression of the FACS-sorted populations was then examined by cDNA microarray analysis. Total RNA was extracted from 20,000 freshly sorted FL-MkPs and BM-MkPs, amplified, reverse transcribed, and hybridized to Affymetrix 430 2.0 oligonucleotide microarrays. These gene chips contain about 45,000 probe sets, allowing for interrogation of more than 34,000 well-characterized genes. The cDNA microarray analysis was performed in triplicate using 3 independent cell harvests and sorts.
We first examined the expression of several key Mk transcription factors and cytokine receptor genes, including Gata1, Gata2, Fog1, Fli1, Gabpa, Runx1, and Mpl (Supplemental Figure 1; supplemental material available online with this article; doi:10.1172/JCI40609DS1). Transcript levels for all of these genes were considerably above background, confirming our selection of Mk-committed cells by the FACS sorting procedure. The expression differences between FL-MkPs and BM-MkPs were all relatively small, the largest being about a 1.8-fold increase of Gata2 mRNA and a 2-fold decrease of Gata1 mRNA in BM-MkPs compared with FL-MkPs.
We next examined the dataset more globally. After filtering for genes having at least a 4-fold change in expression level and P < 0.05 (see Methods), there were 200 upregulated and 122 downregulated genes in BM-MkPs versus FL-MkPs (Figure 2A and Supplemental Tables 1 and 2). Gene Ontology (GO) analysis showed enrichment for genes associated with immune function in BM-MkPs and mitochondrial/metabolism function in FL-MkPs (Supplemental Figure 2, A and B). Gene set enrichment analysis (GSEA) using all available curated gene sets revealed striking enrichment for IFN-α–responsive genes in BM-MkPs versus FL-MkPs (Figure 2B and Supplemental Figure 2E). There was also significant enrichment for genes induced by the other type I IFN, IFN-β (Supplemental Figure 2F), which signals through the same receptor as IFN-α. Gene sets involved in metabolism and mitochondrial function were enriched in FL-MkPs versus BM-MkPs (Figure 2C). Given the known potent antiproliferative effects of IFN-α on Mk growth (23) and the clinical response of myeloproliferative disorders to IFN-α treatment (24–26), we chose to focus the remainder of the current study on the type I IFN signaling pathway.
Gene expression analysis of FL-MkPs versus BM-MkPs. (A) Total number of genes whose expression changed >4-fold with P < 0.05 among the 3 biologic replicates and were represented by probes on the array comparing FL-MkPs versus BM-MkPs for male WT and GATA1s mice. (B and C) Analysis of gene sets enriched in BM-MkPs relative to FL-MkPs and vice versa from WT mice. Asterisks and bold type denote IFN-α–responsive gene sets. (D and F) GSEA for the combined set of 92 IFN-α–induced genes (Supplemental Table 3 and refs. 44, 45) compared with the differentially expressed genes in BM-MkPs versus FL-MkPs in WT (D) and GATA1s (F) mice. (E and G) Probe set intensities corresponding to each of the 92 genes, compared with intensities of 164 randomly selected probe sets, arranged from highest to lowest from the FL-MkP and BM-MkP datasets in WT (E) and GATA1s (G) mice. (H and I) Validation of differences in IFN-α–responsive gene expression in FL-MkPs versus BM-MkPs by conventional PCR (H) and qRT-PCR (I). Some lanes in H were run on different gels or noncontiguous lanes of the same gel.
After combining the 4 different publicly available IFN-α–induced gene sets shown in Figure 2B and eliminating duplicate genes, we generated a more comprehensive list of 92 IFN-α–induced genes (Supplemental Table 3). GSEA using the 92-gene list demonstrated significant enrichment in BM-MkPs compared with FL-MkPs (false discovery rate [FDR], 0.10; nominal P < 0.01; Figure 2D). A plot of the normalized signal intensities of the 164 probes corresponding to these 92 genes compared with the average intensity of 164 randomly selected probes is shown in Figure 2E.
We validated the differences in selected gene expression by quantitative real-time RT-PCR (qRT-PCR) from independently harvested/sorted cells. However, the signals from the FL-MkPs were so low for many of the genes that meaningful numbers could not be generated. Figure 2H shows an ethidium bromide–stained gel of the PCR products for a more qualitative analysis. Quantitative analysis of IFN-α/β receptor 1 (IFNAR1), 1 of the 2 IFNAR subunits, showed about 1.5-fold higher expression in BM-MkPs than FL-MkPs (data not shown).
Developmental stage–specific differences in type I IFN–inducible gene expression in GATA1s mice. In order to determine whether the differences we observed in the WT mice are preserved in the setting of exclusive GATA1s production, MkP isolation and gene expression analysis were repeated using GATA1s mice (21). Numerous genes were differentially expressed between BM-MkPs and FL-MkPs of GATA1s mice, with 577 genes upregulated and 151 genes downregulated greater than 4-fold (P < 0.05; Figure 2A). As expected, GO analysis of highly expressed genes in FL-MkPs versus BM-MkPs showed significant enrichment for mitosis-related genes (Supplemental Figure 2D). However, significant enrichment for IFN-α– and IFN-β–responsive gene sets was still apparent in BM-MkPs versus FL-MkPs, similar to the results obtained from WT mice (Figure 2F and Supplemental Figure 2, C, E, and G). Although the GATA1s FL-MkPs expressed these genes at relatively higher levels than WT FL-MkPs, there was still a significant increase comparing GATA1s BM-MkPs and FL-MkPs (FDR, 0.09; nominal P < 0.01; Figure 2, G–I).
Differential protein levels of type I IFN–responsive genes in BM versus FL Mks. In situ immunohistochemistry was performed next to confirm the differences in protein expression of type I IFN–responsive genes in Mks. As shown in Figure 3A and Supplemental Figure 3, we observed marked staining of Mks in WT adult BM for both IRF1 and IFI205. In contrast, only background staining was apparent in morphologically recognizable Mks from E13.5 FL. Similar differences in IRF1 and IRF8 protein levels were observed in GATA1s mouse BM versus FL Mks (Figure 3B and Supplemental Figure 4). Staining for the Mk marker protein vWF was strongly positive in both BM and FL samples. We conclude that murine BM-MkPs/BM Mks have substantial upregulation of type I IFN–inducible genes compared with E13.5 FL-MkPs/FL Mks and that this also occurs in a GATA1s genetic background.
Increased protein levels of type I IFN–responsive genes in BM versus FL Mks. (A) In situ immunohistochemical staining for IRF1 and IFI205 in adult femur BM and E13.5 FL of WT mice. Staining for vWF served as a positive control for Mks. Positive staining appears brown; counterstain is blue. Arrows indicate Mks. See also Supplemental Figure 3. (B) In situ immunohistochemical staining for IRF1 and IRF8 in GATA1s male mice. Staining for vWF is indicated as a positive control. See also Supplemental Figure 4. Scale bars: 5 μm.
Exogenous IFN-α inhibits hyperproliferation of GATA1s-containing Mks. If increases in type I IFN–responsive gene expression were an important determinant of the stage-specific effects of GATA1s on Mk hyperproliferation, then culturing of early GATA1s FL-MkPs with a type I IFN would be expected to counteract their hyperproliferative phenotype. To test this, single-cell suspensions were prepared from E13.5 FL from GATA1s or control WT mice and cultured in collagen-based semisolid media containing thrombopoietin (TPO) and increasing concentrations of IFN-α. As previously described (21), GATA1s Mks markedly hyperproliferated compared with WT Mks in the presence of TPO alone; however, their hyperproliferation was markedly attenuated in a dose-dependent manner when IFN-α was included (Figure 4, A, C, and D). IFN-α also limited the proliferation of WT E13.5 FL Mks, as expected (23), but the effects were not as pronounced as for GATA1s Mks (Figure 4B).
Antiproliferative activity of IFN-α on GATA1s-containing E13.5 FL Mks. (A and B) Representative CFU-Mk colony morphologies from male GATA1s or WT E13.5 FL cultured with 5 ng/ml TPO and 0 or 1,000 U/ml recombinant murine IFN-α. Colonies were fixed and stained for AChE activity after 7 (GATA1s) or 8 (WT) days. Original magnification, ×40. (C) AChE stained CFU-Mk colonies generated from male GATA1s E13.5 FL cultured with 5 ng/ml TPO and 0, 100, 500, or 1,000 U/ml recombinant IFN-α. Original magnification, ×40. Images of the entire slides are shown at left. (D) Effect of IFN-α treatment on CFU-Mk colony size from GATA1s E13.5 FL cultured in the presence of 5 ng/ml TPO and the indicated concentrations of IFN-α. Colony size is given as the number of pixels encompassing the entire colony after photographing the culture dishes (see Methods). Horizontal bars denote means. (E) Relative lineage-selective effects of IFN-α on Mk growth. Shown are representative BM AChE-stained CFU-Mk colonies from WT or Ifnar1–/– mice cultured in the presence of 5 ng/ml TPO and the indicated IFN-α concentrations. Original magnification, ×40. Images of the entire slides are shown at left. (F) Quantitation of BFU-E, CFU-GM, and CFU-GEMM of separate cultures grown in MethoCult (Stem Cell Technologies) semisolid medium containing SCF (50 ng/ml), IL-3 (10 ng/ml), IL-6 (10 ng/ml), EPO (3 IU/ml), insulin (10 μg/ml), transferrin (200 μg/ml), and the indicated IFN-α concentrations. See also Supplemental Figure 5.
The antiproliferative effects of IFN-α were relatively selective for the Mk lineage. Colony number and size for adult BM myeloid (CFU-G, CFU-M, and CFU-GM) and mixed myeloid (CFU-GEMM) were only slightly affected at the doses used, in contrast to the marked effect on CFU-Mk (Figure 4, E and F, and Supplemental Figure 5). BFU-Es were also modestly reduced (Figure 4F). As expected, CFU-Mks and BFU-Es from mice lacking IFNAR1 (27) were not affected by IFN-α treatment (Figure 4, E and F, and Supplemental Figure 5).
Delayed resolution of GATA1s Mk hyperproliferative phenotype in the absence of IFN-α/β signaling. Our model also predicted that loss of IFN-α/β signaling in the postnatal BM environment would abrogate or delay resolution of the developmental stage–specific hyperproliferation of GATA1s MkPs. To test this, GATA1s mice were bred to Ifnar1–/– mice. The resulting compound Ifnar1–/–::GATA1s mice were born at the expected Mendelian ratio. We first examined the peripheral blood counts of these mice at 3–4 weeks of age. Male Ifnar1–/– mice and hemizygous GATA1s mice (GATA1 is located on the X chromosome) had platelet counts similar to those of C57BL/6 WT controls. However, Ifnar1–/–::GATA1s male mice had significantly lower platelet counts (624 × 109 ± 20 × 109 platelets/l; n = 10) than male WT (838 × 109 ± 27 × 109 platelets/l; n = 10), Ifnar1–/– (812.7 × 109 ± 30 × 109 platelets/l; n = 10), or GATA1s (790 × 109 ± 18 × 109 platelets/l; n = 10) littermate controls (P < 0.05, 2-tailed Student’s t test; Figure 5A). They also had larger mean platelet volumes (Figure 5B). In contrast, there were no significant differences in red blood cell or total white blood cell counts (Figure 5, C and D). Thus, a genetic interaction exists between GATA1s and IFNAR1 with regard to thrombopoiesis.
Enhanced postnatal proliferation of GATA1s-containing Mks in an Ifnar1–/– genetic background. (A–D) Peripheral blood platelet count (A), mean platelet volume (B), red blood cell count (C), and white blood cell count (D) of WT (C57BL/6), Ifnar1–/–, GATA1s, and Ifnar1–/–::GATA1s male mice at 3–4 weeks of age. (E–H) Representative AChE-stained BM CFU-Mk colonies from each mouse genotype at 3–4 weeks of age. Original magnification, ×40. Images of the entire slides are shown at left. Red arrows in H indicate hyperplastic colonies. (I) Quantitation of colony size (mean number of pixels covered by colonies derived from photographs) from 20 randomly selected colonies. (J) Representative flow cytometry plots for BrdU and 7AAD (DNA content stain) of CD41+ gated cells obtained from the BM of 3- to 4-week-old mice of the indicated genotypes. (K) Percent CD41+BrdU+ cells from J (n = 3).
We next examined the proliferation status of BM-MkPs from 3- to 4-week-old mice. CFU-Mk colony assays showed significantly larger Mk colony size in male Ifnar1–/–::GATA1s mice compared with male WT, Ifnar1–/–, or GATA1s mice (Figure 5, E–I). As previously described (21), Mk colonies from GATA1s mice contained a large number of acetylcholinesterase-negative (AChE–) cells, with scattered AChE+ cells mixed in, suggesting an expansion of more immature Mks in these cultures. There was an exaggerated number of AChE– cells in the base of Ifnar1–/–::GATA1s colonies (Figure 5H), which suggests expansion of the more immature Mk cells. In vivo BrdU incorporation analysis confirmed the higher proliferative rate of CD41+ cells in Ifnar1–/–::GATA1s compared with WT, Ifnar1–/–, and GATA1s mice (Figure 5, J and K).
As an independent means to test our model, we acutely inhibited IFN-α/β signaling in 4- to 6-week-old mice by intraperitoneal injection of neutralizing IFN-α and IFN-β antibodies (Figure 6A). WT mice showed only a minimal increase in the percentage of BM CD41+forward scatterhi cells (i.e., Mks) 8 days after injection compared with control mice that received an equivalent amount of control IgG (19.6% ± 6.8%; n = 3; Figure 6, B and C). GATA1s mice receiving control IgG had a higher percentage of CD41+forward scatterhi cells at baseline than WT mice injected with IgG, consistent with low-level hyperproliferation of GATA1s Mks even in the postnatal period at this age. However, there was an additive increase of CD41+forward scatterhi BM cells in GATA1s mice after injection of the neutralizing IFN-α/β antibodies (65% ± 10.5% versus control IgG; n = 3; Figure 6C), indicative of their proliferative sensitivity to low IFN-α/β signaling. Enumeration of BM Mks in situ by vWF immunohistochemical staining showed similar results (Figure 6, D and E). We conclude that IFN-α/β signaling normally dampens the hyperproliferative phenotype of GATA1s Mks postnatally.
Expansion of BM Mks in GATA1s mice injected with neutralizing IFN-α/β antibodies. (A) Experimental scheme. 4- to 6-week-old GATA1s or age-matched WT mice were injected intraperitoneally with 1 × 104 neutralizing units of anti–IFN-α and anti–IFN-β antibodies each, or an equivalent amount of normal rabbit IgG. 8 days after injection, animals were euthanized, and BM was harvested. (B) Flow cytometric color-contour plots for CD41 staining and forward scatter from whole BM (after red blood cell lysis) from 1 representative experiment. (C) Change in CD41+ cell frequency in mice injected with neutralizing IFN-α/β antibodies versus control IgG (n = 3). Results of 3 independent experiments are shown. (D) Representative vWF immunohistochemical stained sections of femur BM from mice injected 8 days earlier with neutralizing anti–IFN-α/β antibodies or equivalent amounts of control IgG. Scale bar: 1 mm. (E) Quantitation of the data in D, showing the mean number of Mks per 2-mm2 field in 10 randomly selected sections.
Role of IFN-α/β signaling in human Mk ontogeny. In order to determine whether the ontologic differences in Mk IFN-α/β signaling also apply to humans, we examined the expression of the IFN-α/β–responsive gene IRF1 by in situ immunohistochemistry in FL from aborted fetuses (12- to 22-week estimated gestational age) versus postnatal BM (>1 year of age). Similar to the mouse studies, this experiment showed a marked increase in Mk IRF1 protein levels in BM versus FL, with no significant difference in vWF staining (Figure 7A and Supplemental Figure 6). We also found cells staining for IFN-α and IFN-β in the BM, but not in FL samples (Figure 7B and Supplemental Figure 7). Thus, differences in IFN-α/β production in the BM versus FL microenvironment may account for the differences in type I IFN–induced gene expression and their effects on GATA1s-induced MkP hyperproliferation.
Developmental stage–specific difference in human Mk IFN-α–responsive gene expression. (A) In situ immunohistochemical staining for IRF1 and vWF in FL from 12- to 22-week estimated gestational age aborted human fetuses and from BM of >1-year-old individuals. Arrows indicate Mks. Original magnification, ×1,000. See also Supplemental Figure 6. (B) In situ immunohistochemical staining for IFN-α and IFN-β of human FL and BM as in A. Arrows indicate positive cells. Original magnification, ×1,000. See also Supplemental Figure 7. (C) Expression of IFN-α–responsive genes in human DS-TMD and DS-AMKL cells (based on data in ref. 6) compared with an equivalent number of randomly selected probes, shown in rank order. (D) Model for developmental stage–specific reduction in GATA1s-related Mk hyperproliferation with increases in IFN-α/β–responsive gene expression.
Type I IFN signaling in human DS-TMD and DS-AMKL. Finally, we analyzed previously reported gene expression profiles of DS-TMD and DS-AMKL primary cells (6) using the human equivalent of the 92 IFN-α–induced gene panel described above (Supplemental Table 3). These samples represent mixtures of postnatal peripheral blood and BM-derived specimens. FL samples were not available for comparison. Nonetheless, the DS-TMD samples expressed IFN-α–responsive genes at levels considerably higher than background (Figure 7C), similar to our mouse postnatal BM-MkP data. This is consistent with their anticipated proliferative decline as the DS-TMD phase resolves. The DS-AMKL cells also expressed IFN-α–responsive genes a levels higher than background (Figure 7C). We speculate that additional genetic events in DS-AMKL cells allow them to escape the antiproliferative effects of type I IFN signaling.