Bone marrow–specific loss of ABI1 induces myeloproliferative neoplasm with features resembling human myelofibrosis

The phenotype of primary myelofibrosis (PMF) is characterized by progressive bone marrow fibrosis, organomegaly, extramedullary hematopoiesis, thromboembolism, and ultimately, marrow failure or transformation to acute myeloid leukemia (AML).^1-3  Median survival in PMF varies between 1 and 15 years, depending on risk factors, and treatment options are limited.^3,4  Identification of JAK2-activating mutations as major drivers in myeloproliferative neoplasms (MPNs) prompted clinical development of JAK2 inhibitors.^5,6  Ruxolitinib, an ATP-mimetic JAK1/2 inhibitor, induces symptomatic improvement in PMF, but exacerbates associated cytopenias and does not have curative potential, and responses occur regardless of presence of JAK2 mutations.^7-11  Therefore, a major need remains to identify other targetable mechanisms contributing to the pathogenesis of PMF and related MPNs, polycythemia vera (PV), and essential thrombocythemia (ET).

Abelson Interactor 1 (Abi-1) is a negative regulator of Abl1 kinase,^12-15  involved in regulation of cell proliferation.^16,17  By forming a complex with Wiskott-Aldrich syndrome protein family member 2 (WAVE2),^18,19  Wiskott-Aldrich Syndrome protein (WASP), or Diaphanous (Dia) formin,^16,18-23  Abi-1 affects actin remodeling, cell adhesion, and migration. Abi-1 also interacts with integrin α4 and is involved in integrin β1 signaling.^24-26  Abi-1-deficient mice uniformly die in utero with lethal defects of the heart and placenta.^19,24  The role of Abi-1 in carcinogenesis is controversial, as both loss or overexpression were implicated in cancer.^27-30  Its involvement in malignant hematopoiesis, although reported by us and others, remains unclear.^31-35 

Here, we present evidence for direct involvement of Abi-1 in homeostasis of hematopoietic system. We found that conditional deletion of Abi1 in murine bone marrow results in impairment of hematopoietic stem cell self-renewal, progressive anemia, megakaryocytosis and myeloid hyperplasia, with resulting PMF-like phenotype characterized by marrow fibrosis and splenomegaly. Furthermore, Abi-1 protein and mRNA levels are decreased in hematopoietic progenitors from patients with PMF, but not from those with ET or PV. Mechanistically, loss of Abi-1 leads to upregulation of Src family kinases (SFKs), STAT3, and NF-κB signaling, suggesting that the Abi-1/SFKs/STAT3/NF-κB axis may represent a new regulatory module involved in the pathophysiology of MPNs.

CD34⁺ cells were isolated from the bone marrow of patients with PMF or from healthy marrow purchased from AllCells (Alameda, CA). Granulocytes were isolated from peripheral blood (PB) of patients with ET, PV, primary or secondary (PV- or ET-derived) myelofibrosis, and healthy donors (supplemental Table 1, available on the Blood Web site). CD34 MicroBeads (Miltenyi Biotec, San Diego, CA) and gradient centrifugation were used for CD34⁺ and granulocyte isolation, respectively. Human subject participation was conducted with informed consent and approved by local ethics committees.

Conditional Abi1(fl/fl) mice¹⁹  were crossed with B6.Cg-Tg(Mx1-cre+1)Cgn/J strain (#003556, JAX, Bar Harbor, ME)³⁶  to generate Abi1(fl/fl);Tg(Mx1-cre[^+/−]) mice. These animals were back-crossed to B6.SJL-Ptprc^APepc^B/BoyJ(#002014, JAX)(CD45.1) background. Abi1(fl/fl);Tg(Mx1-cre[⁻]), Abi1(fl/wt);Tg(Mx1-cre[⁺]), or Abi1(fl/fl);Tg(Mx1-cre[⁺]) mice were subjected to polyinosinic:polycytidylic acid [poly(I:C)]-induced (Invivo Gen, San Diego, CA) activation of the Cre recombinase under the control of Mx1 promoter to obtain animals with an Abi1(fl/fl);Tg(Mx1-cre[⁻]), Abi1(−/wt);Tg(Mx1-cre[⁻]), or Abi1(^−/−);Tg(Mx1-cre[⁺]) genotype, designated as Abi-1^WT, Abi-1^HET, or Abi-1^KO, respectively. Recombination of Abi1^floxed allele was confirmed by polymerase chain reaction (PCR). We evaluated 76 Abi-1^WT, 41 Abi-1^HET, and 85 Abi-1^KO animals. Animal experiments were approved by the Institutional Animal Care and Use Committee.

A biotin-conjugated antibody cocktail containing anti-TER119, CD127, CD8a, Ly-6G, CD11b, CD4, and CD45R was used to stain lineage-committed cells. BUV395-Streptavidin, anti-CD34-FITC, CD117/c-Kit-APC, Ly-6A/E/Sca-1-PE-Cy7 (or -BV605), CD135/Flt3-PE, and CD16/CD32-PE were used to stain hematopoietic stem/progenitors. Cells were sorted using Legacy MoFlo High-Speed cell sorter.

For noncompetitive bone marrow transplantation, 5 × 10⁶ marrow cells isolated by flushing from poly(I:C)-uninduced Abi1(fl/fl);Tg(Mx1-cre[⁺]) or Abi1(fl/fl);Tg(Mx1-cre[⁻]; CD45.1) mice were injected via tail vein into lethally irradiated (2x475cGy) C57BL/6 wild-type mice recipients (CD45.2; #000664, JAX). Four weeks posttransplant, Abi1 inactivation was performed by poly(I:C) induction. Donor chimerism was evaluated in PB every 4 weeks for 24 weeks. After 24 weeks, marrow was harvested from primary recipients and 5 × 10⁶ cells were transplanted into CD45.2 secondary recipients conditioned as earlier. Donor chimerism in PB was evaluated as earlier. For competitive repopulation assays, marrow cells were isolated via flushing from poly(I:C)-induced Abi-1^KO or Abi-1^WT (CD45.1) mice. After confirming Abi1 inactivation, donor cells (1 × 10⁶ Abi-1^KO or Abi-1^WT) were mixed with competitor cells (1 × 10⁶; 1:1, CD45.1:CD45.2) and injected via tail vein into lethally irradiated CD45.2 recipients. Donor chimerism in PB was evaluated as earlier.

Levels of interleukin (IL)-1α, IL-1β, IL-6, IL-10, IL-12p70, IL-17A, IL-23, IL-27, monocyte chemoattractant protein-1 (MCP-1/CCL2), interferon (IFN)-β, IFN-γ, tumor necrosis factor α (TNF α), and granulocyte-macrophage colony-stimulating factor were detected in plasma of Abi-1^WT (n = 11) or Abi-1^KO (n = 11) 14-week-old animals, using LEGENDplex Mouse Inflammation Panel (Biolegend, San Diego, CA).

Healthy CD34⁺ cells (n = 3) were expanded for 48 hours in the presence of thrombopoietin, Flt3/Flk-2 (Fms-like tyrosine kinase 3/fetal liver kinase-2) ligand, stem cell factor, IL-3, IL-6, and granulocyte-macrophage colony-stimulating factor (StemCell Technologies, Cambridge, MA) and incubated for additional 48 hours with 15 μM ABI1-silencing antisense 2′-deoxy-2′fluoro-β-d-arabinonucleotides (FANA-ABI1-ASO; AUM BioTech, Philadelphia, PA). After 48 hours culture with FANA-ABI1-ASO, Abi-1 protein levels were evaluated by immunoblotting, and cell cycle status was determined by 5-ethynyl-2′-deoxyuridine (EdU) incorporation assay.

Bone marrow was isolated from 20-week-old Abi-1^WT (n = 3) or Abi-1^KO (n = 3) animals by flushing. Cell pellets were lysed, and 100 µg protein per sample was subjected to tryptic digestion. Tryptic peptides were subjected to liquid chromatography-tandem mass spectrometry, using an automated proteomic technology platform.^37,38 

Gene array analysis of lineage⁻, Sca-1⁺, c-Kit⁺ (LSK)-enriched cells from 14-week-old Abi-1^KO (n = 4) or Abi-1^WT (n = 5) mice was done using the Affymetrix-WT Pico Expression Kit (Affymetrix, Louisville, KY) and Affymetrix-3000 7G gene scanner. Partek Genomics Suite v6.6 was used for quality control and data analysis. Microarray data are deposited in GEO as GSE83288.

MSCV-MPL^WT-IRES-GFP and MSCV-MPL^W515L-IRES-GFP retroviral vectors were generously provided by Ross Levine (Memorial Sloan Kettering Cancer Center). Marrow was isolated from 14-week-old Abi-1^WT or Abi-1^HET animals, and Abi1 recombination was verified by PCR. Retroviral transduction and murine marrow transplant assay were performed as previously described.^39,40  Abi-1^WT/MPL^WT, Abi-1^HET/MPL^WT, Abi-1^WT/MPL^W515L, and Abi-1^HET/MPL^W515L transplantation groups were used.

Two-tailed unpaired t-tests or log-rank tests were used for between-group comparisons, using Bonferroni correction where appropriate. P < .05 was considered statistically significant. Throughout, *P < .05, **P < .01, and ***P < .001. Two-tailed unpaired t-test and q-values for multiple hypothesis tests, using the R package QVALUE,⁴¹  were used to select peptides with significant change in paired analyses.

Additional materials and methods are described in supplemental Information.

ABI1 transcript levels in CD34⁺ cells isolated from bone marrow were decreased by approximately 40% in PMF (n = 5) compared with controls (n = 5; Figure 1A). Gene expression profiles (GEO/GSE53482)⁴²  of CD34⁺ cells isolated from PB also showed significant downregulation of ABI1 in PMF (n = 42) with mutations in JAK2 or CALR relative to controls (n = 31; supplemental Figure 1A-C). Furthermore, granulocytes from patients with PMF (n = 36), or secondary myelofibrosis post-PV (n = 9), showed a 40% to 60% decrease in ABI1 mRNA relative to control patients (n = 16; Figure 1B). Downregulation of ABI1 in PMF granulocytes was observed regardless of JAK2 or CALR mutation status (Figure 1C). Notably, no significant changes in granulocyte ABI1 transcript levels were noted in ET (n = 15), PV (n = 20), or post-ET myelofibrosis (n = 4; Figure 1B). Consistent with these findings, examination of gene expression profiles of CD34⁺ cells from bone marrow of ET (n = 24) and PV (n = 26) patients (GEO/GSE103176)⁴³  showed no downregulation of ABI1 (supplemental Figure 1D), indicating that reduced ABI1 expression may be specific to PMF.

To determine the role of Abi-1 in the homeostasis of the hematopoietic system, we phenotyped transgenic Abi-1^KO mice carrying bone marrow-selective knockout of Abi1. We first confirmed both inducible inactivation of the Abi1^flox allele and loss of Abi-1 protein in the marrow (Figure 2A-C; supplemental Figure 2A-B). Abi-1 levels were also reduced in spleen and liver, but not in heart, lung, or kidney (supplemental Figure 2C). Abi-1^KO mice showed femur pallor and splenomegaly that progressed with age, but no change in liver size (Figure 2D-E). Blood count analysis at 14, 35, and 56 weeks indicated progressive leukocytosis, mild anemia, and thrombocytosis in the Abi-1^KO, which were absent in Abi-1^WT animals (Figure 2F; supplemental Table 2). Survival of Abi-1^KO mice decreased 20 weeks post-poly(I:C) injection, with significant weight loss after 60 weeks and death in all animals by 66 weeks of age (Figure 2G; supplemental Figure 2D), likely because of progressive disease. We observed no changes in gross pathology, blood counts, or survival in 56-week-old Abi-1^HET animals. Therefore, in subsequent analyses, we focused on characterization of Abi-1^KO mice. Our findings suggest that marrow-specific inactivation of Abi1 disrupts normal hematopoiesis.

As Abi-1^KO animals aged, the fraction of myeloid (CD11b⁺/Gr-1⁺) cells in their PB increased at the expense of lymphoid B220 cells, whereas the CD3⁺ lymphoid fraction remained unchanged (Figure 3A). Abi-1^KO erythrocytes demonstrated polychromasia, anisopoikilocytosis, and teardrop-shaped forms (Figure 3B; supplemental Figure 3A). Giant platelets were observed in 2/10 14-week-old, 4/10 36-week-old, and 5/10 56-week-old or older Abi-1^KO mice. Bone marrow showed hypercellularity with myeloid hyperplasia, erythroid hypoplasia, and megakaryocytosis (Figure 3C-E; supplemental Figure 3B-C). Bone marrow cellularity at the distal aspect of femur in the medial and lateral condyles was 95% for Abi-1^WT and 100% for 56-week-old Abi-1^KO mice (supplemental Figure 3D). Spleen histology showed expansion of red pulp, with numerous islands of megakaryocytic, myeloid, and erythroid infiltration (Figure 3C-D; supplemental Figure 3B-C). Clusters of megakaryocytes were commonly observed in marrow and spleen of 56-week-old Abi-1^KO mice (supplemental Figure 3C). Abi-1^KO marrow samples and spleens showed progressive deficits of stainable iron on Prussian blue staining (Figure 3D and 3F; supplemental Figure 4A-B), possibly linked to observed anemia (supplemental Table 2; Figure 2F). Silver stain and semiquantitative grading⁴⁴  showed progressive increase in reticulin fibrosis in Abi-1^KO marrow, from grade 1 (out of 3) at 14 weeks to grade 2 to 3 at 56 weeks (Figure 3G; supplemental Figure 4C). Increased thickness and density of reticulin fibers was present in Abi-1^KO spleen, from grade 1 at 14 weeks to grade 2 at 56 weeks (Figure 3G; supplemental Figure 4C). Progressive splenomegaly, megakaryocytosis, and marrow and spleen fibrosis in Abi-1^KO animals meet Mouse Models of Human Cancers Consortium criteria for MPN.⁴⁵ 

We assessed the frequency of hematopoietic stem and progenitor cells in the marrow of 14-week-old Abi-1^WT or Abi-1^KO mice (supplemental Figure 5A). Deletion of Abi1 resulted in no change in Lin- fractions. However, there were increased frequencies of LSKs (from 2.3% [±0.4%] to 4.6% [±0.8%]) and LK progenitors (from 4.9% [±0.9%] to 17.6% [±1.1%]; Figure 4A-B), a 2.6-fold increase in LT-HSCs, no change in ST-HSCs, a 47% reduction in common myeloid progenitors (CMPs), and a 25% increase in granulocyte-monocyte progenitors (GMPs) (Figure 4C; supplemental Figure 5B). Evaluation of the absolute number of stem/progenitor cells confirmed these trends with the exception of CMP, which showed an increased number in Abi-1^KO mice (supplemental Figure 5C).

Given this evidence for expansion of HSCs, we evaluated the cell cycle activity of LSK and LT-HSC cells. Abi-1^KO mice at 14 weeks showed a higher proportion of LSK cells in S-phase, and 46% more LT-HSCs in the S/G2/M phases relative to Abi-1^WT animals (Figure 4D-E; supplemental Figure 5D-E). Transcript expression analysis of genes regulating cell cycle progression in LT-HSCs confirmed these observations (supplemental Figure 5F).

To determine the effect of Abi-1 loss on hematopoietic progenitors, we performed colony-forming unit assays. Results showed a significant decrease in burst forming unit-erythroid and increase in GMP colonies derived from Abi-1^KO marrow isolated from 14-week-old mice, consistent with anemia and leukocytosis (supplemental Figure 5G). We further observed a 3-fold increase in megakaryocyte colonies, consistent with the observed megakaryocytosis in Abi-1^KO mice (supplemental Figure 5H). We found no indication of growth factor independence or hypersensitivity.

To examine the effect of Abi-1 loss on long-term engraftment and self-renewal of HSCs, we transplanted bone marrow cells isolated from Abi-1^WT or Abi-1^KO mice (before poly[I:C] injection), into lethally irradiated recipient C57BL/6 wild-type mice in the absence of competitor cells. Four weeks after the transplantation, Abi-1 loss was induced by poly(I:C)-induction. Whereas initial engraftment remained comparable (∼95%) for primary recipients of Abi-1^WT or Abi-1^KO transplants, 12 weeks postpoly(I:C)-induction, Abi-1^KO-transplanted mice showed progressive loss of chimerism (Figure 4F). These mice also exhibited significant weight loss at 24 weeks posttransplant, with an average weight of 22.2 g (±1.8 g) compared with 28.6 g (±3.1 g) for Abi-1^WT marrow recipients (P = .04). Bone marrow cells were harvested from primary recipients 24 weeks posttransplant and transplanted into secondary recipients. A further decrease in donor chimerism and significantly reduced initial engraftment of Abi-1^KO bone marrow cells were seen in secondary recipients (Figure 4G).

In competitive repopulation assays, C57BL/6 wild-type mice transplanted with the whole bone marrow isolated from Abi-1^KO mice 4 weeks post-poly(I:C)-induction (after PCR confirmed recombination) showed progressive decrease in donor chimerism relative to Abi-1^WT marrow recipients, reaching significance 4 months after the transplant (supplemental Figure 5I). Of note, initial engraftment at 4 weeks posttransplant was comparable between recipients of Abi-1^KO or Abi-1^WT marrow (46% and 45% chimerism, respectively).

To confirm a direct link between Abi-1 loss and cell cycle activity in hematopoietic stem/progenitor cells, we performed EdU incorporation assays using CD34⁺ cells isolated from the bone marrow of healthy donors (n = 3) and exposed to ABI1-silencing FANA antisense oligonucleotides (FANA-ABI1-ASO). More than 50% efficiency of ABI1 silencing after 48 hours incubation with FANA-ABI1-ASO was noted for CD34⁺ cells (Figure 4H; supplemental Figure 6A). Cellular penetrance of FANA oligos was confirmed by confocal microscopy and fluorescence-activated cell sorting (supplemental Figure 6B-C). EdU incorporation assays showed that 48-hour exposure to FANA-ABI1-ASO resulted in a nearly 2-fold increase in CD34⁺ cells in S-phase (Figure 4I; supplemental Figure 6D). Overall, our data suggest that deletion of Abi-1 leads to abnormal cell cycle activity, impaired HSC self-renewal, and defective reconstitution of hematopoiesis.

To determine factors contributing to the abnormal expansion of hematopoietic stem/progenitor cells in Abi-1-deficient animals, we performed a genome-wide expression analysis of LSK cells from 14-week-old Abi-1^KO and Abi-1^WT mice. We found significant overexpression of genes regulated by or involved in regulation of the NF-κB pathway (Figure 4J; supplemental Table 3). MetaCore Pathway Analysis indicated that the NF-κB pathway was first of the top 10 dysregulated pathways in the Abi-1^KO LSK cells (supplemental Figure 6E). Ingenuity Pathway Analysis identified 3 overlapping networks in which NF-κB activation was predicted on the basis of increased expression of known NF-κB targets (Cd69, Cxcl10, Cxcl2, Hp, and Nfkbia) and regulators (Myd88, Chuk, and Ikbkb; supplemental Figure 6F).^46,47  Of note, gene expression profiling of CD34⁺ cells from patients with PMF (GEO/GSE53482) showed upregulation of the same NF-κB human target genes CXCL10, CD69, and HP (supplemental Figure 7A), as well as genes involved in leukocyte migration and recruitment (FCER1A, FCER1G, FCGR2A, TPSAB1, and TPSB2; supplemental Figure 7B).⁴² 

We also assessed levels of 13 pro-inflammatory cytokines, finding a nearly 2-fold increase in IL-1B, IL-12, IL-17, IL-23, IL-27, and MCP-1/CCL2; a roughly 10-fold increase in IFN-γ; and no significant changes in IL-1A, IL-6, IL-10, TNF α, or IFN-β in Abi-1^KO plasma (Figure 4K). Overall, these data link upregulation of NF-κB pathway and inflammatory signaling to Abi-1 deficiency in hematopoietic system.

On the basis of our gene expression data indicating active NF-κB signaling in Abi-1^KO marrow, we evaluated major components of this pathway by immunoblotting. We found increased levels of IκB accompanied by increased phosphorylation of NF-κB (Figure 5A; supplemental Figure 8A). Surprisingly, Abi-1^KO bone marrow showed only modest increase in JAK2 and STAT5 activities, whereas the phosphorylation of STAT3, Akt, and Erk1/2 increased significantly (Figure 5B; supplemental Figure 8B).

On the basis of our previous work showing high-affinity binding between Abi-1 and the Src homology domain 2 of SFKs, we assessed the activity of SFKs.^48,49  We found increased phosphorylation of the Src auto-inhibitory Tyr-527 site. However, pan-phospho-Src family antibody showed increased phosphorylation on Tyr-416, consistent with activation of other SFKs (Figure 5C; supplemental Figure 8C). We did not detect phosphorylation of c-Abl on Tyr-412 in Abi-1^KO bone marrow cells (supplemental Figure 8D). Immunoblotting showed an Abi-1-loss-associated decrease in the stability of WAVE2 complex components (WAVE2, Nap1, and Sra-1; Figure 5D; supplemental Figure 9A), consistent with our previous findings.¹⁹ 

To obtain unbiased insight into the mechanistic effects of Abi-1 loss, we performed label-free, intensity-based quantitative proteomic analysis of bone marrow from 20-week-old Abi-1^KO and Abi-1^WT mice. The analysis yielded 12 103 peptides derived from 2464 unique proteins. When quantified, there were significant changes in the abundance of 318 peptides representing 226 unique proteins in the Abi-1^KO samples (supplemental Table 4; supplemental Figure 9B). A volcano plot showed that among peptides with a more than 2-fold difference, 25.1% showed lower and 74.9% higher abundance (Figure 5E). Among peptides showing a greater than 2-fold increase in abundance in Abi-1^KO samples, we detected peptides derived from calreticulin, CD177, CD97, haptoglobin, Mac-1, myeloperoxidase, STAT1, STAT3, and SFKs Hck and Fgr. We interpreted these data as confirming not only activation of SFKs and STAT3 signaling but also an increase in proteins previously associated with MPN. Within the group of peptides showing a greater than 2-fold decrease in abundance, we detected peptides derived from WAVE2, Sra-1, and Nap1 (Figure 5F).

Finally, immunoblotting analysis performed on CD34⁺ or CD34⁻ cells isolated from PMF patients showed decreases of 80% and 50% in Abi-1 protein levels, respectively. Similar to Abi-1-depleted murine marrow, we also found significant upregulation of SFKs and NF-κB activities (Figure 5G; supplemental Figure 9C-D). These data indicate that SFKs/STAT3/NF-κB signaling operating in Abi-1-deficient marrow may be involved in pathogenesis of the observed PMF-like phenotype. A model of Abi-1/SFKs/STAT3/NF-κB cross talk is presented in Figure 5H.

To further explore the role of Abi-1 in the pathogenesis of MPNs, we assessed involvement of Abi-1 in 1 of the established models of the disease. Such models are based on expression of mutated JAK2, CALR, or MPL. Because of less pronounced ABI1 loss in granulocytes bearing CALR mutation, and no significant overactivity of JAK2-STAT5 in Abi-1^KO marrow, we decided to use the transplant model of MPL^W515L-mediated MPN.³⁹  As lethality and penetrance of the MPL^W515L model is variable in C57/BL6 wild-type mice recipients (in contrast to full penetrance in tumor growth-permissive Balb/C background),^39,50  and homozygous loss of Abi1 allele is sufficient to produce MPN phenotype, we decided to use Abi-1^HET animals to achieve controllable manifestation of the disease in double-mutant (Abi-1^HET/MPL^W515L) animals.

Bone marrow isolated from Abi-1^WT (n = 3) and Abi-1^HET (n = 3) animals was transduced by retrovirus encoding MPL^WT or MPL^W515L. Transduction efficiency was higher than 60%. Abi-1^WT marrow cells expressing MPL^WT or MPL^W515L (Abi-1^WT/MPL^WT or Abi-1^WT/MPL^W515L), or Abi-1^HET marrow cells expressing MPL^WT or MPL^W515L (Abi-1^HET/MPL^WT or Abi-1^HET/MPL^W515L) were transplanted into lethally irradiated C57/BL6 mice (n = 6 per group) (Figure 6A). Analyses performed 60 days posttransplant showed no femur pallor or significant changes in spleen or liver sizes in the Abi-1^WT/MPL^WT, Abi-1^WT/MPL^W515L, or Abi-1^HET/MPL^WT groups. Conversely, Abi-1^HET/MPL^W515L animals showed 3-fold and 1.7-fold increases in spleen and liver size, respectively, as well as femur pallor (Figure 6B-C). Furthermore, we noted pronounced leukocytosis, thrombocytosis, and polycythemia in Abi-1^HET/MPL^W515L animals compared with Abi-1^WT/MPL^W515L (Figure 6D). We also observed a significant increase in Mac-1⁺/Gr-1⁺, CD41, TER119⁺/CD71⁺, TER119⁺/CD71⁻, and B220⁺ cells in both marrow and blood, and a decrease in CD3⁺ cells in the marrow of the Abi-1^HET/MPL^W515L animals (Figure 6E; supplemental Figure 10A-B). Staining for reticulin fibers in the Abi-1^HET/MPL^W515L mice showed significant (3+) thickening in both marrow and spleen (Figure 6F). Overall, these data suggest that loss of 1 copy of Abi1 markedly accelerates the MPL^W515L-mediated MPN in mice.

Extensive evidence implicates dysregulated JAK/STAT signaling in the pathophysiology of MPNs. Nonetheless, JAK2 inhibitors do not have salutary disease-modifying effects at the stem cell level.^7-10  The cross talk between JAK/STAT, MAPK, and Akt pathways, and the potential of therapeutic targeting of these pathways, has received much less attention.^39,51-53  Inflammation signaling and cytokine involvement in MPNs, particularly PMF, is widely acknowledged; however, the specific role of NF-κB signaling in MPNs was highlighted only recently.⁵⁴ 

Abi-1 is an adapter protein involved in regulation of Abl1, PI3K, and Ras signaling, with a previously unknown role in hematopoiesis.^12,17,49  To address this gap, we generated bone marrow-specific murine knockout of Abi1. Loss of Abi-1 proved sufficient to induce a MPN-like phenotype that replicates many features of human myelofibrosis. Myeloid hyperplasia, megakaryocytosis, progressive fibrosis, splenomegaly, and extramedullary hematopoiesis observed in Abi-1^KO mice resemble other murine models of PMF.^6,39,55-58  Notably, we observed no transformation to acute leukemia, yet Abi-1^KO animals had shorter lifespan and evidence of cytokine dysregulation. Although there was no significant phenotypic effect of heterozygous Abi-1 loss, given that PMF typically develops in patients older than 55 years (equivalent to 18 months in mice), we continue observation of the effect of heterozygous Abi1 loss in mice beyond 72 weeks of age.

Molecular analyses of the MPN-like phenotype in Abi-1^KO mice showed unexpectedly only modest activation of JAK2-STAT5 signaling, with hyperactivation of SFKs, STAT3, and NF-κB. Upregulation of Mpo, integrin αMβ2, Runx1, Hck, Fgr, STAT3, and NF-κB–haptoglobin, seen in proteomic analysis, are consistent with increase in myeloid cells and upregulated SFKs, STAT3, and NF-κB signaling in Abi-1^KO mice. Further work will be needed to elucidate how overactive SFKs/STAT3/NF-κB signaling relates to established MPN models. JAK2, MPL, or CALR mutations are frequent events in MPNs,^5,6,39,59  and murine models incorporate these variants.^{39,51,56,57,60-62}  In contrast to JAK2^V617F models that progress to secondary myelofibrosis via PV-like stage,^51,60,61,63  our model resembles the thrombopoietin -driven models expressing MPL^W515L or CALR^del52, in which fibrosis develops after a period of primary thrombocytosis.^39,56,57,62  To explore the association between Abi-1 deficiency and established pathogenetic mechanisms in MPNs, we assessed how heterozygous loss of Abi-1 affected the MPL^W515L-mediated mouse model. Results showed accelerated MPN-like disease in Abi-1^HET/MPL^W515L animals. These data support a contributory role of Abi-1 to the pathophysiology of MPNs. In addition to these findings, proteomics and immunoblotting (supplemental Figure 9E) indicated upregulation of calreticulin in Abi-1^KO. Calreticulin has an established role in MPN pathophysiology, and somatic CALR mutations occur in 30% of PMF.⁵⁹  At this point, the contribution of increased calreticulin to the Abi-1 loss-dependent phenotype remains unclear. Strikingly, Abi-1 deficiency in hematopoietic stem/progenitor cells results in upregulation of genes controlled, directly or indirectly, by NF-κB. Genome-wide expression profiling of Abi-1^KO LSK cells highlighted overexpression of pro-inflammatory pathways associated with NF-κB activation. Upregulation of pro-inflammatory cytokines in patients with PMF,⁶⁴  as well as TLR activation,⁶⁵  may contribute to marrow fibrosis. Our data support a mechanism in which Abi-1 loss results in upregulation of SFKs, which in turn activate STAT3, ultimately modulating the NF-κB pathway. Notably, STATs can be activated by SFKs,^66,67  and STAT3 and NF-κB pathways are linked.^68,69  Interestingly, Ser-32-phosphorylated IκB is not degraded and remains highly expressed in Abi-1^KO marrow. IκB phosphorylation on Tyr allows NF-κB activation without IκB degradation.⁷⁰  Elucidation of how Abi-1 loss might lead to increased NF-κB signaling and the role of IκB phosphorylation in this process will require further research.

Abi-1 loss induced significant impairment of the hemostasis of HSCs. Increased cell cycle activity of Abi-1^KO hematopoietic stem/progenitor cells is possibly leading to their expansion, displacement from the marrow, and progressive exhaustion, and is reflected in their impaired self-renewal capacities. Similarly, increased cycling and impaired self-renewal of HSCs was noted in animal models of hematopoietic tissue-specific overactivation of Akt or NF-κB.^71,72  Given the activity of Mx1 promotor-driven Cre in mesenchymal stromal cells, it is reasonable to expect that changes in homeostasis of Abi-1^KO HSCs may prove to be both intrinsic and nonautonomous in future experiments.

Analyses of ABI1 transcripts in MPNs indicated that its loss may be specific to PMF. Mechanisms responsible for loss of ABI1 transcripts in PMF are however not clear. Notably, single nucleotide polymorphism arrays performed on 151 MPN samples showed microdeletions in the 10p12 region encompassing ABI1 locus.⁷³  Sequencing of ABI1 is warranted to provide mutational profile of the gene in MPNs.

In aggregate, our data provide the first evidence for a role of Abi-1 in the pathogenesis of PMF, and a link to upregulation of the SFKs/STAT3/NF-κB signaling axis. To our knowledge, this is the first model of MPN mechanistically uncoupled from JAK2-STAT5 hyperactivity. Full elucidation of the cross-communication between JAK/SFKs/STAT3/NF-κB signaling may provide insight into the association between proliferative and pro-inflammatory features that characterize MPNs, particularly PMF. Targeting Abi-1 and associated pathways may offer a new therapeutic strategy for human PMF.

The authors thank Mandy Pereira, Laura Goldberg, Mark Dooner, Elaine Pappa, Michael DelTatto, Joan Boylan, Aleksander F. Sikorski, Paola Guglielmelli, and Loren Fast for helpful suggestions and technical assistance, and Eric Milner for his editorial help. The authors thank Matthew Adler and Veenu Aishwarya from AUM BioTech, LLC and Chris Gould from Biolegend for technical assistance and helpful suggestions. The authors also thank John Crispino and Qiang Jeremy Wen (Northwestern University) for critical comments, discussion, and technical assistance. The authors thank Ross Levine (MSKCC) for MSCV-MPL WT-IRES-GFP and MSCV-MPL W515L-IRES-GFP retroviral vectors.

This work was supported by Brown Biomed Division Dean's Award grant (P.M.D., E.O.), University of Medicine Foundation Integration Funding (P.M.D.), Rhode Island Foundation grant (P.M.D.), a pilot grant from the Lifespan Center of Biomedical Research Excellence for Cancer Research Development (National Institutes of Health/National Institute of General Medical Sciences 1P30GM110759 to P.M.D. and 1P20GM119943 to P.J.Q. and P.M.D.) and Savit Foundation (P.J.Q., P.M.D.). Work in Florence was supported by a grant from Associazione Italiana per la Ricerca sul Cancro (AIRC, Milan, Italy), Special Program Molecular Clinical Oncology 5x1000 to AIRC-Gruppo Italiano Malattie Myeloproliferative project #1005. A.H.-J. and L.K. were supported by National Institutes of Health/National Cancer Institute R01CA161018 (L.K.). A.J.O. is supported by the American Society of Hematology Research Scholar award.

Contribution: A.C., J.M., N.A., M.P., N.K., Y.C., K.L., C.S., X.Y., R.Z., A.P., A.T., J.C., A.H.-J. and O.L. performed experiments, acquired and analyzed data, and contributed to manuscript writing; N.A., D.O.T., A.J.O., T.C.Z., E.O., J.L.R., L.K., P.J.Q., P.A.G., R.M., and A.M.V. analyzed data and contributed to manuscript writing; and P.M.D. designed research, analyzed data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Patrycja M. Dubielecka, Warren Alpert Medical School of Brown University, Signal Transduction Laboratory, Rhode Island Hospital, 1 Hoppin St, Coro West, Suite 5.01, Providence, RI 02903; e-mail: patrycja_dubielecka-szczerba@brown.edu.