Ptch2 loss drives myeloproliferation and myeloproliferative neoplasm progression

In many malignancies, HH signaling is activated at the level of ligand secretion and not via point mutations within members of the canonical HH signaling pathway. In PB leukocytes of JAK2V617F⁺ MPNs and in BCR-ABL⁺ CML, we observed strong overexpression of Desert HH (DHH) transcript levels (but not SHH or IHH), compared with leukocytes from normal donors or patients with acute myeloid leukemia (AML; Fig. 1 a; patient information in Table S1). DHH overexpression in MPN/CML positively correlated with elevated PTCH1 transcript levels (Spearman correlation coefficient < 0.0001; Fig. 1 b). PTCH1 is a direct target gene of active canonical HH signaling and suggests cell-intrinsic HH pathway activation within MPN/CML cells, but not in AML blasts.

Excess HH ligand secretion was shown to activate classical canonical HH signaling (SMO-dependent GLI1 activation), and also induces ERK phosphorylation mediated by the intracellular signaling domain of the PTCH1 receptor (noncanonical HH signaling). Therefore, HH-stimulation of PTCH1-depleted cells can only activate canonical HH signaling, but lacks activation of the PTCH1-dependent noncanonical ERK pathway. In contrast to the PTCH1 receptor, the PTCH2 receptor lacks an intracellular signaling domain. We therefore hypothesized that depletion of the PTCH2 receptor might enhance ligand binding to the PTCH1 receptor and might therefore activate both SMO-dependent canonical GLI1 activation and PTCH1-dependent noncanonical ERK activation. To identify signaling differences in between the PTCH1 and PTCH2 receptors, we analyzed canonical and noncanonical pathway activation in BM stromal cultures from Ptch1^+/− mice, Ptch2^+/− mice, and Ptch2^−/− mice (genotypes; Fig. 1 c). Ptch1^+/− BM cultures showed activated canonical HH signaling as measured by elevated Gli1 transcription (Fig. 1 d), but as expected did not show enhanced Erk activation (Fig. 1 e). In contrast, BM stromal cells from Ptch2^−/− mice show increased canonical Hh signaling (enhanced Gli1 transcript levels; Fig. 1 d), but also possess strongly increased Erk phosphorylation levels (Ptch1-dependent noncanonical HH signaling; Fig. 1 e). In vitro experiments confirmed these results and show that depletion of PTCH1 by siRNA in HH ligand–producing 293T cells causes dephosphorylation of ERK, whereas depletion of PTCH2 actually induces ERK phosphorylation (Fig. 1 f). The Erk activation caused in Ptch2^−/− cells was strictly dependent on the presence of HH ligands and was completely abrogated by the addition of the HH ligand–blocking antibody Shh5E1 (Fig. 1 g), confirming our hypothesis that overstimulation of the Ptch1 receptor by HH ligands in cells devoid of the competitive Ptch2 receptor causes constitutive Erk phosphorylation. In concordance, overexpression of the huPTCH1 receptor but not the huPTCH2 receptor induced ERK phosphorylation in HH ligand–producing 293T cells (Fig. 1, h and i). From our experiments, we can conclude that depletion of either receptor on its own can activate GLI1 transcription, but only the depletion of PTCH2 induces combined canonical and noncanonical HH signaling, inducing ERK phosphorylation by directing the ligand toward the PTCH1 receptor.

Interestingly, a very similar situation was observed in many MPN and CML patients. Here, we found elevated DHH and PTCH1 transcript levels in nearly all patients and, additionally, we saw a loss of PTCH2 in 70–75% of MPN/CML cases investigated (Fig. 1 j), which could further enhance the effect of the HH ligands on the malignant cells.

Besides differences in signal transduction, the expression distribution of the Ptch1 and Ptch2 receptors might also cause differences in the hematologic phenotype. Therefore, we sorted different hematopoietic cell subsets and CD45⁻ stromal cells from the BM of C57BL/6 mice and detected the expression of Ptch1 and Ptch2 via qPCR. Both receptors were expressed in all cell subsets, including LT- and ST-HSCs and nonhematopoietic stromal cells, but they showed strong differences in their expression patterns. Ptch1 was most highly expressed in lymphoid cells and proliferating erythroid cells, but only at very low levels in the granulocytic/monocytic lineages. In contrast, the Ptch2 receptor exhibited the highest transcript levels (≤80-fold) in granulocytes, monocytes, and granulocytic-monocytic progenitors (GMPs), indicating that these myeloid cells are its major effector cells and can therefore be most affected by loss of Ptch2 (Fig. 1 k).

To investigate the effect of hyperactive canonical and noncanonical HH signaling on hematopoietic development, we examined the hematologic phenotype of Ptch2^−/− mice. Blood analysis at 3, 6, and 12 mo of age showed leukocytosis in Ptch2^−/− mice (Fig. 2 a), which was mainly driven by relative and absolute (3.5-fold at 12 mo) expansion of CD11b⁺ granulocytes and monocytes (Fig. 2, b–d). At 12 mo of age, the PB of Ptch2^−/− mice consisted of >60% CD11b⁺ myeloid cells, whereas control mice were <20% (Fig. 2, b and c). Besides the expansion of mature myeloid cells, we also detected a massive increase of immature myeloid progenitors (LK cells, fourfold increase) and LKS cells (sixfold increase) mobilized into the PB of Ptch2^−/− mice (Fig. 2 e). In contrast, T cells (both CD4⁺ and CD8⁺) and B cells (B220⁺) were significantly reduced (relative and total levels; Fig. 2, b and d), and the mice developed anemia and reduced thrombocyte counts (Fig. 2 a).

The hematologic phenotype within the PB could be recapitulated in other hematopoietic organs as assessed by IHC and flow cytometry (gating strategies in Figs. S1 and S2). Both neutrophils and monocytes were expanded not only in the PB but also in BM (IHC for CD11b⁺ cells) and spleens of Ptch2^−/− mice, where total numbers of granulocytes and monocytes were increased three- to fourfold compared with WT mice (flow cytometry; Fig. 2, f and g).

In contrast to the monocytic and granulocytic lineage, we detected a loss of all other hematopoietic lineages within the BM. Megakaryocytes (Fig. 2, h and i), as well as megakaryocytic progenitors (MegP, Lin⁻cKit⁺Sca1⁻CD41⁺CD16/32⁻; Fig. 2 j), the erythroid proliferation pool (CD44/Ter119 staining; Fig. 2 k), erythroid progenitors (EP, Lin⁻cKit⁺Sca1⁻CD41⁻CD16/32⁻; Fig. 2 j), and total megakaryocyte-erythroid progenitors (MEPs; Lin⁻cKit⁺Sca1⁻CD34⁻CD16/32⁻; Fig. 2 l) were strongly reduced within the BM compartment, explaining the anemia and thrombocytopenia detected in those mice. Reduced BM hematopoiesis in Ptch2^−/− mice was partially compensated for by enhanced extramedullary hematopoiesis with increased spleen weight and increased spleen cellularity (Fig. 2, m–o). Cells delocalized into the spleen were immature LKS cells, LK cells consisting mainly of MEPs and GMPs (Fig. 2 l) and mature myeloid cells (Ter119⁺ erythroid cells, granulocytes, and monocytes, Fig. 2, f and k). FACS data could be confirmed by colony assays showing a strong increase of erythroid colonies (CFU-Es) and granulocyte/monocyte colonies (CFU-GMs) within the spleen (Fig. 2 p). Similar to MPN patients, immature myeloid cells and colony forming cells (CFU-GMs and CFU-Es) were strongly increased in the PB of Ptch2^−/− mice (Fig. 2 p).

In conclusion, the Ptch2^−/− hematopoietic phenotype shows many features of human MPN patients with expansion of the granulocyte/monocyte compartment, enhanced extramedullary hematopoiesis with splenomegaly, and increased progenitor/LKS mobilization into the peripheral blood.

In contrast to granulocytes and monocytes, both lymphoid cell populations were progressively lost in Ptch2^−/− mice. T cells were consistently reduced in the PB, the spleen (involving both CD4⁺ and CD8⁺ T cells) and in the developing thymus of 3-mo-old mice and at all later time points (Fig. 3, a and b). Reduced thymocytes (46 × 10⁶ [WT] to 32 × 10⁶ [KO]) involved all major thymic T cell subpopulations including CD4⁻CD8⁻, CD4⁺CD8⁺, and both single-positive cell types, whereas relation in between the different T cell subsets was not altered (Fig. 3, c and d). All T cell subsets within the thymus showed a significant and strong increase in spontaneous apoptosis rates, pointing toward alterations within the selection process caused by depletion of Ptch2 (Fig. 3, b, e, and f).

B cells were normal at 3 mo, but strongly reduced in the PB, the BM, and the spleen at 12 mo (Fig. 2, b-d and Fig. 3, g and h). Impairment of B cell development already started at the earliest step of B/T cell–specific development, the CLP population (Fig. 3 g). Most additional steps of B cell development within the BM (pro–B cells, pre–B cells, and immature B cells) and the spleen (T1, T2, and T3 B cells) were reduced, pointing toward the lack of early progenitors being causative for reduction of B cell progenitors starting at the CLP stage and an additional loss at the transition from prepro–B cells to pro–B cells (Fig. 3 g).

We further examined the effect of Ptch2^−/− on the HSC compartment. Gating strategies are described in Figs. S1 and S2. At 3 mo of age, we found a relative and total increase in LKS cells within the BM and within spleens of Ptch2^−/− mice (Fig. 4 a). Cell cycle analysis using EDU click assay and FxCycle violet stain revealed increased cells in S/G2/M phase within all LKS subsets (MPPs, ST-HSCs, LT-HSCs) and also within the myeloid progenitor compartment (Fig. 4, b and c), explaining the drive toward increased myelopoiesis and leukocytosis seen in those mice. In addition, Ptch2^−/− mice show a reduction of quiescent HSCs in G₀ (Ki67/FxCycle violet stain; Fig. 4 d). These cell cycle changes resulted in a significant loss and exhaustion of HSCs over time, resulting in a reduction of LT-HSCs by 50% at 12 mo of age as assessed by flow cytometry using three different staining strategies of LKS cell subpopulations (Fig. 4, a, e, and f; and Fig. S1). Not only were LT-HSCs reduced, but so were all subsequent differentiation steps, such as ST-HSCs and MPPs (Fig. 4 f). Loss of repopulating LT-HSCs was also confirmed by performing a limiting dilution assay that showed a reduced CRU frequency of 1:33.361 in Ptch2^−/− mice versus 1:17.246 in WT mice (Fig. 4, g and h).

To directly compare the functionality of LT-HSCs from Ptch2^−/− mice versus WT mice, we performed a HSC competition assay and sorted 150 LT-HSCs (Lin⁻Kit⁺Sca⁺CD150⁺CD48⁻) from Ptch2^−/− versus WT BM (both CD45.2⁺) and diluted them with 1 × 10⁶ supporter BM cells (CD45.1⁺) and transplanted those into mixed CD45.1/CD45.2-expressing mice (Fig. 4, g and i). Chimerism in the PB was first assessed at 3 wk after transplantation and showed a slightly enhanced early engraftment for the faster cycling Ptch2^−/− LT-HSCs, which turned into a reduced, but stable long-term engraftment after 6–14 wk, suggesting a functional defect of the Ptch2^−/− LT-HSCs (Fig. 4 i). Interestingly, the frequency of Ptch2^−/− CD45.2⁺ cells did not decrease further, but slightly increased over time, giving us the first evidence that Ptch2^−/− HSCs in a normal WT niche do not suffer losses, and might even regain self-renewal properties.

Enhanced cell cycle within LT-HSCs increases the probability of differentiation and can cause exhaustion of the LT-HSC pool over time, but can be of advantage regarding stress hematopoiesis induced by low-dose 5-FU (150 mg/kg bw; Campaner et al., 2013). We therefore challenged the regeneration capacity of 3-mo-old mice by injecting them with 5-FU, which induces apoptosis in actively cycling cells and causes severe aplasia. As expected, 5-FU treatment more efficiently depleted the BM, reduced hemoglobin levels, and decreased WBC counts in the more actively cycling Ptch2^−/− mice compared with WT mice, 5 d after 5-FU injection (Fig. 4 j). At 7 d after 5-FU treatment, WBCs and hemoglobin levels were equivalently reduced in both groups (Fig. 4j). But in contrast to the previous measurement, total BM cells were already increased in Ptch2^−/− mice compared with WT mice, indicating faster BM regeneration in those mice (Fig. 4 j, bottom). Furthermore, LSK^med cells, representing the regenerating and proliferating LKS population (Campaner et al., 2013), were strongly increased in the regenerating Ptch2^−/− BM compared with WT BM, resulting in increased myeloid progenitors and also a faster increase in LKS cells (Fig. 4, k and l). Cell cycle analysis revealed that, even during BM regeneration, the cell cycle was increased in Ptch2^−/− mice compared with WT mice (Fig. 4, m and n). The accelerated regeneration capacity of Ptch2^−/− mice also translated into a significantly better survival after weekly 5-FU injections (Fig. 4 o), supporting the advantage of hyperactive canonical and noncanonical HH signaling for 5-FU–induced stress hematopoiesis.

In conclusion, the increased cell cycle and reduced quiescence of HSCs in Ptch2^−/− mice results in a reduction of HSCs over time, but enhances the regeneration capacity of the bone marrow after 5-FU–induced aplasia in young mice.

We next evaluated whether the observed phenotype in Ptch2^−/− mice is a result of hyperactive canonical and noncanonical HH signaling in the niche or within the hematopoietic system. Therefore, we performed the niche change transplantation system shown in Fig. 5 a: black, control (CD45.2⁺ WT BM was transplanted into CD45.1⁺ WT mice); blue, hematopoietic Ptch2^−/− (CD45.2⁺ Ptch2^−/− BM was transplanted into CD45.1⁺ WT mice); red, niche Ptch2^−/− (CD45.1⁺ WT BM was transplanted into CD45.2⁺ Ptch2^−/− mice). Chimerism was between 97 and 100% 3 mo after transplantation (unpublished data). Ptch2^−/− mice with WT BM (red) developed a pancytopenia with reduced leukocytes, mainly caused by decreased B and T cells, reduced thrombocyte counts, and reduced hemoglobin levels (Fig. 5, b and c), whereas granulocyte/monocyte levels were normal. BM analysis revealed that the Ptch2^−/− niche caused reduction of megakaryocytes, erythroid progenitors, MEPs, colony forming cells, LK cells, and LKS cells (Fig. 5, d–h). Within the HSC compartment, all HSC subsets were reduced, including LT-HSCs (Fig. 5 h). Loss of LT-HSCs could be confirmed via a limiting dilution assay of BM cells from mice 4 mo after the first niche change experiment, demonstrating a strongly reduced CRU frequency of 1:90.091 in mice with the Ptch2^−/− niche compared with 1:26.909 in the control group (Fig. 5 i). Long-term follow-up of mice with WT BM in a Ptch2^−/− niche (red) showed a shortened survival as a result of increasing cytopenias caused by the reduced regeneration pool (Fig. 5 j). Also, the reduction of thymic cells and the increased apoptosis rates within the thymic T cell subpopulations was dependent on the cell-extrinsic Ptch2^−/− (Fig. 5 k). Furthermore, loss of B cells starting at the CLP level was dependent on the niche Ptch2^−/− (unpublished data). In addition, we found that the mobilization of colony-forming cells, including progenitor and LKS cells, into the PB was also dependent on the cell-extrinsic hyperactivation of HH signaling in the niche (Fig. 5 l).

In contrast, the hematologic Ptch2^−/− phenotype (Ptch2^−/− BM in a WT mouse, blue) was characterized by leukocytosis resulting from an expansion of all cell subsets, including granulocytes, T cells, and B cells (Fig. 5, b and c). Hemoglobin levels and thrombocytes increased to normal levels within 4 mo after transplantation. Ptch2^−/− BM in a WT niche had no defect in LKS frequency and function (Fig. 5, h–j). In contrast, the in vitro replating of either 1% or of total cell numbers of 1,000 or 5,000 cells showed that the hematopoietic Ptch2^−/− significantly increased replating capacity at replating rounds 3–5 (Fig. 5, m and n). Further evaluation of the replated cells by flow cytometry showed that the Ptch2^−/− significantly enhanced the frequency of LKS cells from replating round to replating round (Fig. 5 o). In addition, homing experiments showed that all transplantation groups had similar homing capacities into the BM, but that the hematopoietic Ptch2^−/− increased the homing of hematopoietic cells into the spleen (Fig. 5 p).

Table 1 summarizes the findings from the niche change experiment. We can conclude that hyperactivation of canonical and noncanonical HH signaling in the niche causes depletion of hematopoiesis starting with the LT-HSC compartment, causes progenitor mobilization, and also affects selection processes in the thymus and B cell lymphopoiesis. In contrast, the cell-intrinsic hyperactivation of canonical and noncanonical HH signaling causes leukocytosis and can enhance self-renewal properties, as well as homing into the spleen.

Because the major hematopoietic changes apparent in our mice, particularly the loss of LT-HSCs and the mobilization of progenitors into the PB, are caused by cell-extrinsic effects, we aimed to identify alterations of the hematopoietic niche in Ptch2^−/− mice. The BM niche consists of various cell types, such as osteoblasts, endothelial cells, CaR cells, and MSPCs. These provide factors essential for stem cell maintenance, stem cell differentiation, homing, and retention in the BM. Bone sections revealed a decreased frequency of endosteal osteoblasts and slightly reduced bone substance in Ptch2^−/− mice (Fig. 6, a and b). Furthermore, in vitro BM stromal cultures from Ptch2^−/− versus WT BM for 12 d showed reduced expression of several important HSC regulators (Fig. 6 c): Cxcl12, which regulates HSC retention in the BM and HSC quiescence, as well as multilineage reconstitution (Sugiyama et al., 2006; Greenbaum et al., 2013); Jagged1 (Jag1), which was shown to be required for quiescence and self-renewal of HSCs (Poulos et al., 2013; Mendelson and Frenette, 2014); and angiopoietin-1. Next, we sorted osteoblasts, CaR cells, endosteal cells, and MSPCs fromBM and from digested bones of Ptch2^−/− versus WT mice (gating strategy in Fig. S1). RNA from sorted niche cells was amplified, and we performed qPCR for Smo, Ptch1, Ptch2,Dhh, and for HSC regulators, in addition to microarray gene expression analysis. Smo was preferentially expressed in MSPCs and osteoblasts, indicating that these niche cell subsets are the most sensitive to alterations within the canonical HH signaling pathway (Fig. 6 d). In agreement with this, the canonical HH signaling pathway (represented in the KEGG pathway analysis as HH and basal cell carcinoma pathway) was only activated in MSPCs, but not in other niche cells (Fig. 6 e and Fig. S3). Ptch1 and Ptch2 were expressed in all niche cells, with Ptch1 being most highly expressed in endothelial cells and Ptch2 in MSPCs. Interestingly, the Ptch1 receptor-dependent noncanonical Erk pathway represented as Ras signature pathway (oncogenic signature and Kegg gene sets in Fig. S3 and Table S2) was activated in all four niche cell subsets, indicating that noncanonical HH signaling is considerably more involved than the canonical pathway. Dhh was exclusively expressed in endothelial cells, pointing to these cells as being the natural regulators of HH signaling within the niche (Fig. 6 d). qPCR for essential HSC regulators shows that all niche cells are involved in deregulation of niche homeostasis in Ptch2^−/− mice and confirms the data from the BM stromal cultures (Fig. 6, c and f). Scf is down-regulated in all Scf-expressing niche cells, especially endothelial cells and CaR cells. Cxcl12, the major retention factor in the niche, is down-regulated in CaR cells, but also in osteoblasts. Osteopontin is most highly expressed and strongly down-regulated in osteoblasts. Angiopoietin-1 is reduced in CaR cells and osteoblasts, but maintained at high levels in MSPCs. Jagged1 is down-regulated in endothelial cells, where it is normally highly expressed. Interestingly, thrombopoietin (Thpo), a strong regulator of HSC quiescence and proliferation, is strongly up-regulated in osteoblasts and MSPCs. Although low levels of Thpo restrain HSCs in a quiescent state, high Thpo levels drive HSC proliferation and could therefore result in the increase in cell cycling detected in the HSC compartment of Ptch2^−/− mice.

From these experiments, we can conclude that the niche is strongly affected by alterations in canonical, and even more by the noncanonical, HH signaling pathway, and that essential factors for stem cell maintenance, such as Scf and Jagged1, stem cell proliferation as in Thpo, or retention factors such as Cxcl12 are all strongly affected and could be involved in reduced HSC numbers and enhanced HSC mobilization seen in Ptch2^−/− mice.

Our initial experiments show that the majority of patients with chronic myeloproliferative diseases have an overexpression of DHH ligands within hematopoietic cells, which may autostimulate hematopoietic cells or stimulate HH-responsive niche cells. To distinguish the influence of niche-dependent HH activation versus cell-intrinsic HH activation on disease development, we performed the niche change experiment again, but in this experiment we infected the BM cells with retroviral particles containing either the BCR-ABL/GFP oncogene or the JAK2V617F/GFP oncogene.

The expression of the JAK2V617F oncogene in C57BL/6 mice causes a myeloproliferative disease characterized by increased hemoglobin/hematocrit in the initial phase of the disease, followed by leukocytosis (granulocytes/monocytes) in the PB. In addition, the mice develop increased extramedullary hematopoiesis with splenomegaly and low-grade myelofibrosis, but they do not die from this chronic disease. Interestingly, the hyperactivation of canonical and noncanonical HH signaling either within the niche or within the hematopoietic system or both drastically enhanced the JAK2V617F disease phenotype and turned the chronic myeloproliferative disease into a lethal leukemic disease (Fig. 7 a). The combination of JAK2V617F and the hematopoietic Ptch2^−/− caused a massively increased leukocytosis in the PB consisting of mature CD11b⁺ myeloid cells (Fig. 7 b) and also CD11b⁺Kit⁺ immature blasts accounting for 15% of total PB leukocytes (Fig. 7 b). Immature cells were also increased in BM and spleens, and the accelerated disease resulted in the death of 6/10 mice. JAK2V617F plus the Ptch2^−/− niche did not enhance leukocytosis or splenomegaly. In contrast, the mice developed a more severe thrombocytopenia and anemia (HGB < 6 g/dl). Furthermore, they showed an increased mobilization of cKit⁺ cells into the PB, accounting for 20% of leukocytes and fulfilling the criteria for an acute myeloid leukemia (Fig. 7 b). Combined depletion of Ptch2 in the niche and the BM showed the most severe phenotype, and all of the mice died from the JAK2V617F-driven disease within 1 yr of transplantation (Fig. 7 a). At the 3-mo time point, the majority of mice had >25% blasts in the PB, had developed thrombocytopenia (Fig. 7 b), and had developed a significantly increased splenomegaly (Fig. 7, d and e). BM aspiration 100 d after treatment showed that the polycythemia vera phenotype with massively increased Ter119⁺ erythroid progenitors had already turned into a granulopoiesis-driven phenotype with increased Gr-1⁺/CD11b⁺ cells and increased cKit⁺ immature progenitors within the BM and the PB in Ptch2^−/− mice (Fig. 7 c). Investigation of JAK2V617F⁺ LKS and LK cells at the time of death revealed that the addition of Ptch2 within hematopoiesis and the niche favors the expansion of the immature malignant cells primarily in the spleen, but also in the BM, creating a leukemia-promoting environment (Fig. 7 f). Despite the presence of the JAK2V617F oncogene, we found similar alterations in cytokines expressed by the Ptch2^-/- niche cells, especially down-regulation of Cxcl12 and Scf (Fig. 7 g).

In contrast to JAK2V617F, the BCR-ABL oncogene always causes a lethal myeloproliferative disease in C57BL/6 mice. To prolong their survival, mice were transplanted with BM containing only 5% BCR-ABL⁺ cells. In this context, only the cell-intrinsic deletion of Ptch2 enhanced disease development, resulting in an earlier death of the mice by further increasing leukocytosis, whereas the Ptch2^−/− niche did not alter disease development (Fig. 7, h and i).

Our findings show that hyperactivation of combined canonical and noncanonical HH signaling in myeloproliferative diseases can accelerate disease development, and in the case of JAK2V617F, can turn a nonlethal myeloproliferative disease into a lethal acute leukemia. Although cell-intrinsic HH activation enhances leukocytosis and splenomegaly in JAK2V617F and BCR-ABL-pos. mice, the niche Ptch2^−/− mobilizes leukemic blasts into the PB and enhances thrombocytopenias and anemias, resulting in a premature death of those mice.

Aberrant activation of HH signaling in malignancies caused by point mutations in the canonical SMO-dependent HH pathway (PTCH1/SMO mutations) drives specific cancer entities, such as basal cell carcinomas, medulloblastomas, or rhabdomyosarcomas, which in turn are highly responsive to treatment with SMO inhibitors (vismodegib and erismodegib). In contrast, the majority of cancer types show activated HH signaling within the malignant cells or within the surrounding stromal cells induced by excess ligand secretion. Previous clinical trials with SMO inhibitors for these entities have failed to show treatment responses, perhaps because HH ligands stimulate a much broader signaling event, as follows: SMO-dependent canonical GLI1/2 activation; retention of activated CYCLIN B1 in the cytoplasma by binding to PTCH receptors (Barnes et al., 2001); and constitutive ERK activation directly downstream of PTCH1 (Chang et al., 2010). ERK signaling is a major player in the oncogenic RAS signaling pathway cascade, and is involved in driving proliferation, migration, and stemness, through various stages of cancer development. In our experiments, we show that the PTCH1 receptor is the exclusive mediator of ERK signaling induced by HH ligands. The depletion of PTCH1 blocked ERK phosphorylation, whereas depletion of its ligand competitor PTCH2 actually induced ERK phosphorylation, by redirecting the entire ligand toward the PTCH1 receptor. In this regard, Ptch2^−/− mice constitute an ideal model to simulate the situation of excess ligand stimulation of effector cells, allowing for the evaluation of its influence on both malignant and niche cells.

Interestingly, in MPNs, but not in AML, we can detect high levels of DHH and PTCH1 in the malignant cells, accounting for activated canonical HH signaling. In addition, we find loss of PTCH2 in ∼70% of the cases. This is despite the fact that granulocytes and monocytes, which are the main pool of leukocytes within the PB in MPNs, usually have the highest PTCH2 levels within the whole hematopoietic system (Fig. 1 l). In embryonic development, it was shown that PTCH2 expression is necessary to abrogate canonical HH signals and limits HH activity to the floor plate (Holtz et al., 2013), thus implicating that lack of PTCH2 in myeloid cells/GMPs further amplifies HH ligand stimulatory effects. Indeed, Ptch2^−/− mice show massive expansion of granulocytes/monocytes starting at the GMP level, resulting in leukocytosis and splenomegaly, even though the regenerative HSC pool is concomitantly reduced. Interestingly, Dhh^−/− mice, which have reduced canonical and noncanonical HH signaling, show the exact opposite phenotype, with a reduction in GMPs and granulocytes, whereas MEPs are increased (Lau et al., 2012). In Ptch2^−/− mice, differentiation from CMPs to MEPs is blocked as a result of cell-extrinsic mechanisms. This results in reduced numbers of megakaryocytes and erythroid progenitors in the BM. The effect on erythropoiesis is partially compensated for by the relocation of erythropoiesis into the spleen, further increasing splenomegaly. These myeloproliferative features, with increased granulocytes/monocytes and enhanced erythropoiesis in the spleen and splenomegaly, strongly resemble human MPNs.

In contrast to myeloid cells, Ptch2 is nearly absent in B and T cells, whereas Ptch1 is most strongly expressed. In agreement with these results, hematopoietic Ptch2^−/− did not affect B and T cell development, suggesting that the Ptch1 receptor functions as the regulator of cell-intrinsic HH ligand responses in T cell development and T cell activation (Rowbotham et al., 2009; Furmanski et al., 2012; de la Roche et al., 2013). Nevertheless, the niche Ptch2^−/− strongly enhances apoptosis rates and reduces the frequencies of all T cell subsets within the thymus, probably as a result of increased selection pressures within the thymic niche.

In addition, lack of Ptch2 in the niche resulted in loss of B cells at 12 mo of age. This is partially caused by the reduction in the HSC pool, which occurs by that time, followed by a reduction in differentiating CLPs. In addition, B cell maturation in the BM is guided by direct interaction of cKit⁺ pro–B cells with Scf-expressing stromal cells. Because Scf levels are drastically reduced in Ptch2^−/− BM niche cells, the development of prepro–B cells is also affected.

The effect of the complete Ptch2^−/− LT-HSCs is primarily characterized by an increase in cell cycling, followed by HSC exhaustion after 12 mo, loss of HSC quiescence, and HSC mobilization into the PB and the spleen. Most of the effects on the HSC compartment can be explained by niche cell alterations in Ptch2^−/− mice. Cxcl12 is strongly reduced in Ptch2^−/− CaR cells. CXCL12 inhibitors are potent HSC mobilizers used in the clinic, and a lack of Cxcl12 reduces HSC retention in the BM, causing HSC mobilization into the PB. Loss of Scf and other self-renewal factors, such as Jagged1, are most likely the cause of the reduction in stem cell maintenance, whereas the loss of megakaryocytes results in strongly increased Thpo levels, driving the cell cycle in HSCs. The influence of the HH signaling pathway on the BM niche is mainly caused by the noncanonical pathway, as Cxcl12-expressing CaR cells, osteoblasts (osteopontin), and endothelial cells only show noncanonical Erk–Ras pathway activation, but not canonical HH pathway activation. The only niche cell type that can activate both the canonical and noncanonical HH signaling pathway is MSPCs. Our findings implicate that excess HH ligand stimulation of the niche cells, or otherwise induced constitutive activation of canonical and noncanonical HH signaling, can strongly alter the production of cytokines important for the maintenance of HSCs, therefore HH ligand–producing malignant cells possess the ability to drastically alter their niche.

Although the Ptch2^−/− niche causes loss of HSCs, the intrinsic Ptch2^−/− enhances replating capacity and LKS maintenance in vitro. Furthermore, the hematopoietic Ptch2^−/− causes leukocytosis by increasing the frequency of nearly all cell types (myeloid cells and B and T lymphocytes). These cell-intrinsic effects are also strongly supported by the experiments using myeloproliferative oncogenes such as BCR-ABL or JAK2V617F. BCR-ABL causes a lethal myeloproliferative disease in mice, and death of the mice was strongly accelerated by hematopoietic Ptch2^−/−, resulting in faster and more pronounced leukocytosis, whereas cell-extrinsic depletion of Ptch2 was not relevant to disease development. This confirms our previous findings that depletion of hematopoietic canonical HH-signaling (Smo) is relevant for disease development and stem cell function in BCR-ABL–driven disease (Dierks et al., 2008).

The JAK2V617F oncogene causes a myeloproliferative disease in humans, which is stable for years and then progresses to AML in ∼4–12% of the cases. The mechanisms driving this acceleration are currently not understood. In C57BL/6 mice, the JAK2V617F oncogene also causes nonlethal MPN, which is characterized by initial erythrocytosis, followed by leukocytosis and a mild myelofibrosis at later stages. Also in this context, the hematopoietic Ptch2^−/− drastically increased leukocytosis, accelerated the infiltration of cells into the spleen, and increased the numbers of Kit⁺ immature blasts in the PB (Fig. 7,b–e). The constitutive activation of canonical and noncanonical HH signaling in niche cells caused mobilization of immature Kit⁺ blasts into the PB, approaching the criteria for a myeloid leukemia, and also increased the severity of anemias and thrombocytopenias, causing death in those mice. These findings are in agreement with previous publications showing that tight control of the niche is necessary to guide HSC maturation and genetic stability, and that loosening this control results in the development of leukemias (Kode et al., 2014).

Finally, the combination of hematopoietic and niche-dependent induction of canonical and noncanonical HH signaling simulates excess HH ligand stimulation in both compartments, transforming the nonlethal myeloproliferative disease into a full-blown acute leukemia, with >30% blasts in the PB and in which all mice succumbed to the disease. These findings suggest that excess HH ligand secretion is also a potential driver of the transformation process of MPNs into acute leukemias in humans. Furthermore, it implicates HH ligand inhibitors instead of SMO inhibitors in fully targeting the canonical and noncanonical HH pathway within hematopoiesis, and the niche as a possible therapeutic option to prevent the transformation of MPNs into acute leukemias.

Ptch1^+/− (Ptch1^tm1Mps/Ptch1⁺; The Jackson Laboratory), Ptch2^−/− (Ptch2^tm1Dgen/Ptch2^tm1Dgen; Deltagen), C57BL/6 (Ly5.2), Ly5.1 (B6.SJL-Ptprca Pepcb/BoyJ; The Jackson Laboratory), and Ly5.1/Ly5.2 (Ly5.1 x Ly5.2 crossbreed, both from The Jackson Laboratory) mice were maintained and genotyped as previously described. Ptch1 and Ptch2 mice were backcrossed 10 times with C57BL/6. All mice used for experiments (inclusive control groups) were obtained from our inbred strains. All animal experiments were conducted in compliance with the US Department of Health and Human Service Guide for Care and Use of Laboratory Animals, and in concordance with German law. Experiments were approved by the Regierungspräsidium Freiburg.

For niche change transplantations (Fig. 5 a), BM cells were isolated from donor mice (Ly5.1, Ptch2^−/−, or C57BL/6) and red blood cells were lysed using Gay solution. 5 × 10⁶ BM cells were transplanted via retroorbital injection into lethally irradiated (10 Gy) hosts (Ly5.1, Ptch2^−/−, or C57BL/6). To monitor hematopoietic recovery, blood was analyzed using an ADVIA120 (Siemens). Donor cell engraftment was determined using Ly5.1/Ly5.2 discrimination. For the limiting dilution competitive transplantation assay, we transplanted 1 × 10⁶ Ly5.1 BM with 3 × 10³, 15 × 10³, 3 × 10⁴, and 1 × 10⁶ competitor Ptch2^−/− or C57BL/6 BM cells in lethally irradiated Ly5.1/Ly5.2 recipient mice (Fig. 4 g). For the competitive LT-HSC repopulation assay, 1 × 10⁶ Ly5.1 BM cells were mixed with 150 SLAM⁺ competitive Ptch2^−/− or WT LT-HSCs and transplanted into lethally irradiated Ly5.1/Ly5.2 mice. Engraftment was monitored by Ly5.1/Ly5.2 staining at 4, 8, 12, and 16 wk after transplantation. Engraftment was assessed as positive if chimerism was >0.3% including all lineages.

For BM transplantation experiments, including retroviral JAK2V617F or BCR-ABL insertion, donor mice (Ptch2^−/− or C57BL/6) were injected i.p. with 5-FU (150 mg/kg) and mice were sacrificed 4 d later. RBC-depleted BM cells were cultivated in DMEM containing 10% fetal bovine serum, IL-3, IL-6, and SCF. Cells were infected twice with a pMSCV/JAK2V617F/IRES-GFP or MSCV/BCR-ABL/IRES-GFP or pMSCV/IRES-GFP (control) retrovirus. 5 × 10⁵ mononuclear cells were transplanted retroorbitally into lethally irradiated Ptch2^−/− or C57BL/6 mice. Disease development was monitored by inspection for general disease symptoms, twice weekly weight measurements, and blood cell counts (Advia120; Siemens), followed by flow cytometric analysis. BM was aspirated after 100 d, and then analyzed by FACS. Mice were sacrificed when moribund organs of interest (spleen, femur, and liver) were extracted for weight measurements, flow cytometry analysis, or fixed for IHC staining.

7 × 10⁶ Ly5.1, WT, or Ptch2^−/− BM cells were injected i.v. into lethally irradiated Ly5.1, WT, or Ptch2^−/− recipient mice. After 4 h, the mice were sacrificed and BM and spleen cells were analyzed using CD45.1/CD45.2 and LIVE/DEAD Fixable Aqua Dead Cell stain(Thermo Fisher Scientific).

Antibodies used are as follows (all from BioLegend): CD3, CD11b, Gr1, CD19, Ter119, Il7R (lineage markers), CD31, CD45, CD45.1, CD45.2, CD51, CD4, CD8, B220, CD34, FLT3, CD48, CD150, CD41, CD44, CD16/32, CD90.1 (Thy1.1), CD140 (VEGFR), CD166 (Alcam), cKit, Sca1, and Ki67. For cell cycle analysis, cells were fixed with 2% formalin, permeabilized with chilled 90% methanol, and stained with FxCycle Violet stain (Invitrogen), Ki67, or with Click-iT reagent to detect incorporated BrdU according to the manufacturer's instructions (Invitrogen). BrdU (100 µl/20 g body weight, 10 mM EdU) was injected i.p. 15 h before BM harvest. For intracellular phospho-Erk staining, cells were fixed and permeabilized, and then incubated with a phospho-p42/44 MAPK mouse mAb (Cell Signaling Technology), followed by the secondary anti–mouse IgG Alexa Fluor 488–conjugated Ab (Cell Signaling Technology). Cells were analyzed using a CyanADP flow cytometer. Annexin V/7AAD staining (BD) was performed according to the manufacturer’s instructions.

Protein expression of PTCH1, PTCH2, ERK, and pERK was determined in 293T cells transfected with siRNA against PTCH1 or PTCH2 using standardized protocols with antibodies against the following proteins: PTCH1 (LS-Bio; LS-B3939; 1:1,000), PTCH2 (LS-Bio; LS-C80730; 1:1,000), ERK (Cell Signaling Technology; MAPK, Erk1/2, 9102; 1:1,000), pERK (Cell Signaling Technology; Phospho-p44/42 MAPK, Erk1/2, 9106; 1:1,000), and β-actin (Sigma-Aldrich; 1:20,000).

Paraffin-embedded spleen and liver sections, as well as paraffin-embedded and decalcified BM slides, were acquired from the Pathology Department of the University of Freiburg. Diaminobencidine immunoperoxidase staining (EnVision+System-HRP/DAB; DAKO) was performed for 10 min according to the manufacturer’s instructions. Primary Abs to CD11b (IgG rabbit polyclonal; LSBio) diluted 1:400 were incubated overnight, followed by a 30-min incubation with secondary Abs (EnVision+System-HRP/DAB; DAKO). Immunodetection was performed with 3,3 diaminobencidine. HE stain was done for 30–45 s. Images were taken using an Axioplan2 microscope (Carl Zeiss).

Femurs of WT or Ptch2^−/− mice were flushed and cells were when cultivated in Petri dishes for 12 d without the addition of cytokines. Floating cells were removed after 24 h, and then every second day, and only adherent cells were used for further analysis regarding HH gene expression, Erk phosphorylation, and the expression of cytokines.

RBC-depleted BM (2 × 10⁴), spleen (5 × 10⁵), and PB cells (5 × 10⁵) were plated in MethoCult GF M3434 (StemCell Technologies, Inc.) and colonies were counted after 7 d. Benzidine staining was performed to identify erythroid colonies.

RNA from human PBMCs from patient samples, enriched adherent BM cells, or sorted BM cells was isolated using the RNeasyKit (QIAGEN), reverse transcribed, and amplified for 50 cycles at an annealing temperature of 60°C. Murine and human primer/probe sets for TaqMan PCR were obtained from Applied Biosystems (Table S3). Results were quantified according to the ΔΔ-CT method, based on the relative expression of a target gene versus a reference gene (GAPDH), and normalized to the median of all samples.

WT and Ptch2^−/− niche cells were pooled from five mice and sorted as described in Fig. S1. RNA was extracted with the RNeasy Micro Kit (QIAGEN). RNA integrity was analyzed by capillary electrophoresis using a Fragment Analyzer (Advanced Analytical Technologies). RNA samples had an RNA quality number of >8.5 and were further processed with the NuGEN Ovation Pico WTA V2 Kit as described by the manufacturer. The amplified cDNA was further fragmented and labeled using the Affymetrix WT Terminal labeling and control kit. Labeled fragments were hybridized to GeneChip Mouse Gene ST 2.0 arrays for 16 h at 45°C with 60 rpm in an Affymetrix Hybridization oven 640. After washing and staining, the arrays were scanned with the Affymetrix GeneChip Scanner 3000 7 G. CEL files were produced from the raw data with Affymetrix GeneChip Command Console Software Version 4.0. We used Partek Genomics Suite software for further analysis.

Here, CEL files were imported, including control and interrogating probes. Prebackground adjustment was set to adjust for GC Content and probe sequence and RMA background correction was performed. Arrays were normalized using Quantile normalization, and probeset summarization was done using Median Polish. Probe values were log₂ transformed. To identify differentially expressed genes between the groups, we performed a one-way analysis of variance in Partek. Using Fisher's Least Significant Difference as a contrast method, WT subpopulations (MSPCs, endothelial cells, OBCs, and CaR cells) were compared with Ptch2^−/− subpopulations. Gene set enrichment analysis was performed by using the KEGG gene set database or the oncogenic signature gene set database. The threshold for significance was set to a nominal value of P < 0.05, normalized enrichment score >1.2 or <−1.2 and false discovery rate <0.25.

Data are represented as the mean ± SEM. For statistical comparisons between groups, two-tailed Student t test or Mann-Whitney test was performed when indicated. For all analyses, P < 0.05 was considered statistically significant: *, P < 0.05; **, P < 0.01; ***, P < 0.001. Correlations were assessed with the Spearman rank correlation coefficient. Analyses were performed using GraphPad Prism 5.03.

Gene expression data are available from ArrayExpress (accession no. E-MTAB-3361).

Fig. S1 shows gating strategies for PB, spleen/BM cells analysis, and niche cell purification. Fig. S2 shows flow cytometry analysis for B cell lymphopoiesis. Fig. S3 shows enrichment plots and heat maps for core enrichment genes. Table S1 lists patient characteristics. Table S2 lists Molecular Signature Database oncogenic signatures. Table S3 lists genes and primers.

This work was supported by the Emmy-Noether program of the DFG (DI 1664/1-1), FOR 2033 TP B1 (Nichem), and the Deutsche Krebshilfe (grant 110670). C. Dierks is supported by EXC 294, Centre for Biological Signaling Studies BIOSS.

The authors declare no competing financial interests.

Author contributions: C. Klein, A. Zwick, K. Aumann, S. Decker, T. Benkler, and S. Kissel performed the experiments. D. Pfeifer performed Affymetrix arrays. M. Follo performed the cell sorting. A.L. Illert, C. Klein, R.A.J. Oostendorp, C. Dierks, and J. Duyster designed research. C. Klein and C. Dierks wrote the manuscript.