Sulfatase modifying factor 1–mediated fibroblast growth factor signaling primes hematopoietic multilineage development

HSC differentiation and self-renewal are modulated by different mitotic stimuli. Thus, we analyzed whether the Wnt–β-catenin and FGF signaling pathways are altered in HSPCs from Sumf1^−/− mice. The HSPCs were purified from the BM of 3-wk-old Sumf1^−/− mice, and from age-matched WT mice, by lineage-negative selection. Sulf1 and Sulf2 activities were fully impaired in homogenates of Sumf1^−/− HSPCs (Fig. 1 A). Sulf1 and Sulf2 inactivation leads to an increase in the signaling of FGF, which is caused by stabilization of the HSPG–FGFR–FGF trimeric complex (Wang et al., 2004; Lamanna et al., 2008). Indeed, there was also increased phosphorylation of ERK, a downstream effector of FGF signaling, both in freshly isolated Sumf1^−/− HSPCs (Fig. 1 B) and in Sumf1^−/− HSPCs cultured without growth-factor stimulation (Fig. 1 C, right). As expected, no differences were seen when the cells were cultured in an enriched medium, because the added growth factors can themselves modulate several signaling pathways (Fig. 1 C, left). The FGF signaling pathway was specifically activated in the Sumf1^−/− HSPCs because there was increased phosphorylation of the downstream effector FSR2 (Kouhara et al., 1997) and increased transcription of GPC3, which positively modulates the pathway (Lai et al., 2008; Fig. S1, A and B).

Surprisingly, accumulation of β-catenin was seen in both freshly isolated and cultured Sumf1^−/− HSPCs (Fig. 1, B and C, right). This was not expected; indeed, inactivation of Sulf1 and Sulf2 should result in the inactivation of the canonical Wnt pathway, with the consequent β-catenin degradation (Ai et al., 2003). The accumulation of β-catenin and the increase in ERK phosphorylation were specifically caused by inactivation of both Sulf1 and Sulf2 in the Sumf1^−/− HSPCs, as confirmed by the analysis with HSPCs isolated from the BM of Sulf1^−/− Sulf2^−/−double-KO mice (Fig. 1 D, left). In contrast, there were no differences in the levels of p-ERK and in β-catenin accumulation in retrieved HSPCs from the BM of MPSIIIA (deficient in SGSH sulfatase) and MPSII (deficient in IDS sulfatase) mice, as compared with WT mice (Fig. 1, E and F), thus excluding a role of the other sulfatases.

To demonstrate the constitutive activation of FGF and Wnt signaling in HSCs, the c-kit⁺, lineage⁻, and Sca1⁺ (KLS) cell fraction of HSPCs was purified from WT and Sumf1^−/− mice. These KLS cells were then transduced with lentiviruses (LVs) carrying 4X β-catenin or an ERK-responsive element upstream of a minimal CMV promoter controlling enhanced GFP (eGFP) expression. eGFP expression was quantified as percentages of eGFP⁺ cells within the KLS cells (Fig. 2 A), LT-HSCs (Kit⁺, Lin⁻, Sca1⁺, and CD150⁺; Fig. 2 B) and ST-HSCs (Kit⁺, Lin⁻, Sca1⁺, and CD150⁻; Fig. 2 C). The FGF and Wnt/β-catenin pathways were constitutively activated in all of the Sumf1^−/− cell populations analyzed, as demonstrated by the increased frequency of GFP⁺ cells in these populations, as compared with their WT counterparts.

Finally, we tested the HSPC clonogenic potential by counting CFUs of WT and Sumf1^−/− HSPCs obtained with the colony-forming cell assay. Here, 30% less total CFUs were originated by Sumf1^−/− HSPCs, as compared with WT HSPCs. There were no differences in the morphologies and amounts of erythroid CFUs (CFU-E), but a significant reduction in myeloid CFUs (CFU-GM; Fig. 2 D), although the ratios of both CFU-E and CFU-GM with respect to the total number of colonies was comparable (Fig. 2 E).

These data demonstrate that the FGF and Wnt/β-catenin pathways are constitutively active in HSPCs from Sumf1^−/− mice, and also it might be possible that SUMF1 modulates these pathways through Sulf1 and Sulf2 activities.

We next asked whether accumulation of β-catenin in Sumf1^−/− HSPCs is caused by increased phosphorylation of ERK. Inactivation of GSK-3β, the kinase that is part of the β-catenin destruction complex, occurs via its phosphorylation. Although this can be mediated by activation of the canonical Wnt pathway, it can be mediated also by p-ERK (Ding et al., 2005). We tested for the latter possibility by inhibiting p-ERK with the MEK1/2 inhibitor UO126 (Fig. 3 A). Sumf1^−/− HSPCs were cultured in the presence of UO126, and Western blotting was then performed on the protein homogenates. Phosphorylation of ERK was completely abolished in these UO126-treated cells, and GSK3-β phosphorylation was also blocked. Remarkably, β-catenin did not accumulate in these cells, and remained at levels comparable to WT HSPCs (Fig. 3 A, lanes 1–3).

In addition, we examined the accumulation of two transcription factors that regulate erythroid and myeloid differentiation: C/EBPα and GATA-1 (Laiosa et al., 2006b; Tsai et al., 1989). Semiquantitative PCR analysis and Western blotting showed that both C/EBPα and GATA-1 accumulate in Sumf1^−/− HSPCs (Fig. S2 and Fig. 3 A, lane 2). Their levels were substantially diminished in UO126-treated Sumf1^−/− HSPCs, and were comparable to those of WT HSPCs (Fig. 3 A, lanes 1–3). Thus, C/EBPα was up-regulated here by increased levels of p-ERK, as has been previously shown (Prusty et al., 2002). Interestingly, we also saw a marked p-ERK–dependent increase in GATA-1 levels. Conversely, there were similar levels of β-catenin, p-ERK, and c/EBPα in Sumf1^−/− HSPCs that was not dependent on whether they were treated with Dkk1, a Wnt pathway inhibitor that acts on the Frizzled receptor (Bafico et al., 2001; Fig. 3 A, lanes 2 and 4). As expected, bFGF treatment of HSPCs increased p-ERK and p-GSK3-β even further, in respect to untreated Sumf1^−/− cells; in turn, c/EBPα, GATA-1, and β-catenin accumulated accordingly (Fig. 3 A, lanes 2 and 5). In addition, Axin2 and Notch1, which are β-catenin target genes, and Hes-1, a Notch1 target gene, were up-regulated in Sumf1^−/− HSPCs, in respect to WT HSPCs. These levels were diminished by treating the cells with UO126, but not with Dkk1, whereas FGF treatment resulted in an increase in the expression levels in respect to the untreated Sumf1^−/− HSPCs (Fig. 3, B–D). These data clearly show that SUMF1 inactivation results in an increase in ERK phosphorylation, which in turn leads to β-catenin, c/EBPα, and GATA-1 accumulation. Sulf1/2^−/− double-KO mice also showed c/EBPα and GATA-1 accumulation (unpublished data).

HSCs are regulated by extrinsic cues, such as a variety of cytokines, growth factors, and ligands, and multiple developmental signaling pathways. The regulation of Wnt, BMP, Notch, Hedgehog, and FGF signaling pathways can contribute to the modulation of HSC-niche activity and can influence the quiescence state and mobilization of HSCs, and their recruitment to the vascular niche (Blank et al., 2008). Thus, we investigated the activation of the Wnt/β-catenin and FGF pathways in mesenchymal stromal cells (MSCs), which are nonhematopoietic cells that support HSC activities and functions in the niche. The MSCs isolated from the total BM of WT and Sumf1^−/− littermates at 3 wk of age showed increased p-ERK, accumulation of β-catenin, and a consequent increased transcription of Notch1, Deltex-1, Hes-1, and Axin2 (Fig. 3, E and F). Inhibition of ERK phosphorylation by UO126 in Sumf1^−/− MSCs led to reduced accumulation of β-catenin and down-regulation of target genes (Fig. 3, E and F).

These data indicate that SUMF1 controls the FGF signaling pathway, which in turn also regulates the amount of β-catenin accumulation, also in the stromal component of the HSC niche.

Next, we investigated whether altered Wnt and FGF signaling pathways and the accumulation of c/EBPα and GATA-1 transcription factors can result in differentiation impairment of HSCs in Sumf1^−/− mice.

We first quantified the total number of cells in the BM, spleen, and thymus of Sumf1^−/− mice, as compared with age-matched WT littermates, with organ cellularity adjusted according to body or organ weights (caused by the very low body weights of the mutant mice). There was low cellularity in the hematopoietic organs of Sumf1^−/− mice, as compared with WT controls (Fig. S3, tables I and II). Similar differences in cellularity were also seen in the spleen and thymus of Sulf2^−/− mice, although these were not seen in the Sulf1^−/−, Ids^−/−, and Sgsh^−/− mice (Fig. S3, tables III and IV, and not depicted).

We then studied the phenotype of the BM of Sumf1^−/− mice. A cohort of ≥14 Sumf1^−/− mice and age- and sex-matched WT mice (4 wk old) were sacrificed, and their BM cells were flushed from their long bones and characterized by multichannel FACS analysis, for the expression of lineage/differentiation-stage markers.

Erythroid differentiation normally progresses from CD71⁺Ter119^low/− pro-erythroblasts through CD71⁺Ter119⁺ basophilic erythroblasts, CD71^low Ter119⁺ chromatophilic erythroblasts, and CD71⁻Ter119^+/high orthochrotomaphilic erythroblasts. In the Sumf1^−/− BM, we saw accumulation of CD71⁺Ter119^low/− pro-erythroblasts and a strong reduction in CD71^low Ter119⁺ chromatophilic erythroblasts and mature CD71⁻Ter119^high orthochrotomaphilic erythroblasts, indicating a block in erythroid differentiation (Fig. 4 A). The accumulation of erythroblasts in the spleen of Sumf1^−/− mice might reflect a compensatory response to ineffective erythropoiesis, or a peripheral accumulation of immature cells (Fig. 4 B). An increase in CD71⁺Ter119^low/− pro-erythroblasts was also seen in Sulf2^−/− mice, but not in Sulf1^−/−, Ids^−/−, or Sgsh^−/− mice (Fig. S4, A and B, and not depicted). No differences were detected in the percentages of BM CD71^lowTer119⁺, of BM CD71^–Ter119^high, and of spleen Ter119⁺ cells from these animal models (Fig. S4 C). In the BM of Sumf1^−/− mice, there was also a marked reduction in monocytes (CD11b⁺ and CD115⁺) and a slight decrease in granulocytes (CD11b⁺ and CD115⁻), as compared with WT mice, with no differences in their percentages in the spleen (Gr-1⁺; Fig. 4, C and D). In contrast, myeloid development progressed normally in the Sulf1^−/−, Sulf2^−/−, ids^−/−, and sgsh^−/− mice (unpublished data).

Overall, these data indicate that SUMF1 contributes to the regulation of the development of the erythroid and myeloid lineages, and that SUMF1 loss of function results in a significant loss of erythroid and myeloid terminally differentiated cells.

B lymphocytes differentiate in the BM from a common lymphoid progenitor (CLP) through a series of well-defined intermediates. We analyzed the pro–B cell stage (CD127⁺, B220^low, IgM⁻) in the BM of Sumf1^−/− mice, in comparison with WT littermates: there was an accumulation of pro–B cells, a reduction in immature B cells (CD127⁻, B220^low, IgM⁺), and apparently normal percentages of mature B cells (B220^high, IgM⁺; Fig. 5 A). However, a strong decrease in terminally differentiated B lymphocytes (B220⁺) was seen in the spleen of mutant mice (Fig. 5 B). These data led us to conclude that there was a block in maturation of B lymphocytes at the pro–B stage. Absolute B cell counts in the spleen confirmed this finding (unpublished data), thus weakening the alternative hypothesis, that lack of terminally differentiated B cells is a secondary effect caused by the erythroid expansion occurring in the spleen of the Sumf1^−/− mice. No differences were detectable in pro–B and mature B cells in any of the other mice models analyzed, with the exception of Sulf2^−/− mice, in which a slight decrease in immature and mature B lymphocytes was seen in the BM, although this was in the absence of a defect in mature cells in the spleen. (Fig. S5, A and B).

Next, we analyzed T cell development. As with B lymphocytes, T cells originate from a CLP, but during differentiation, the pro–T cell (c-kit⁺, CD127⁺, CD25⁻, CD3⁻, CD4⁻, and CD8⁻) BM-resident cells migrate into the thymus to give rise to mature T lymphocytes. Sumf1^−/− and WT mice had similar pro–T cell percentages in the BM (Fig. 5 C). However, a reduction in immature double-positive (DP) CD4⁺ CD8⁺ thymocytes and an increase in single-positive (SP) CD4⁺ or SP CD8⁺ thymocytes were seen in the thymus of Sumf1^−/− mice (Fig. 5 D). The CD4 and CD8 antigens are expressed only after CD3 (a TCR accessory protein; Clevers et al., 1988; Teh et al., 1988). By looking at the CD3⁺ population, no differences in the frequencies of SP CD3⁺ CD4⁺ and SP CD3⁺ CD8⁺ cells were seen in Sumf1^−/− mice, as compared with WT mice, clearly showing the existence of aberrant Sumf1^−/− populations of CD4⁺ and CD8⁺ thymocytes that did not express CD3 (Fig. 5 E). This might be caused by accumulation of Notch1 and β-catenin in the Sumf1^−/− thymocytes (Fig. 5 F). Indeed, Notch1 has been shown to substitute for CD3 expression in promoting development of thymocytes from the DP to SP stages (Deftos et al., 2000; Michie et al., 2007). In the Sulf1^−/−, Ids^−/−, and Sgsh^−/− mice, pro–T cells, total DP and SP thymocytes, and DP and SP cells gated on CD3⁺ were comparable in frequency to those of WT mice (Fig. S5, C–H, and not depicted). However, a slight decrease in DP thymocytes and a parallel increase in SP CD4⁺ cells was seen in Sulf2^−/− mice when gated both on the total population and on the CD3⁺ cells (Fig. S5, F–H). Interestingly, development of invariant natural killer T cells, a highly specialized subset of thymocytes, was also severly impaired in the Sumf1^−/− mice (Plati et al., 2009).

The HSC-enriched KLS compartment, defined as c-kit⁺, Lin⁻, Sca1⁺, includes LT-HSCs, ST-HSCs, and multipotent progenitor (MPP) cells (Spangrude et al., 1988; Laiosa et al., 2006a). The analysis of the lineage-negative HSPC population of WT and Sumf1^−/− mice revealed a similar content of KLS cells in the BM (Fig. 6 A). In addition, to identify LT-HSCs (c-kit⁺, Lin⁻, Sca1⁺, and CD150⁺), ST-HSCs, and MPP cells (ST-HSC/MPP: c-kit⁺, Lin⁻, Sca1⁺, and CD150⁻), we analyzed the CD150 marker in KLS cells. Comparable proportions of LT-HSCs and ST-HSC/MPP cells were present in the BM of Sumf1^−/− and WT mice (Fig. 6 B).

The HSPC fraction also includes early committed progenitor cells, such as CLPs and common myeloid progenitors (CMPs). The CLP population differentiates subsequently into pro–T and pro–B cells, whereas the CMP population generates megakaryocyte–erythrocyte progenitors (MEPs) and granulocyte–monocyte progenitors (GMPs; Morrison and Weissman, 1994; Morrison et al., 1995; Kondo et al., 1997; Akashi et al., 2000; Laiosa et al., 2006a). To evaluate the relative frequencies of these progenitor cells, Sumf1^−/− and WT HSPCs were stained for Sca1⁺, c-kit⁺, CD127, CD16, and CD34. Flow cytometry analysis revealed no significant differences in the contents of CMPs (Lin⁻, c-kit⁺, Sca1⁻, CD127⁻, CD34^high, and CD16^low), MEPs (Lin⁻, c-kit⁺, Sca1⁻, CD127⁻, CD34⁻, and CD16^low), GMPs (Lin⁻, c-kit⁺, Sca1⁻, CD127⁻, CD34^high, and CD16^high), and CLPs (Lin⁻, c-kit⁺, Sca1⁺, and CD127⁺) in Sumf1^−/− HSPCs, in respect to WT cells; however, the progenitor cells showed a tendency to increase in Sumf1^−/− HSPCs, in respect to the WT cells (Fig. 6 C). No major differences in HSPC subfractions were seen in Ids^−/−, Sulf1^−/−, and Sulf2^−/− mice (unpublished data).

To further demonstrate that the constitutive FGF signaling that leads to β-catenin accumulation is responsible for the multilineage differentiation block seen in Sumf1^−/− mice (Figs. 4 and 5), we wanted to unlock the differentiation of Sumf1^−/− terminal lineages by inhibiting FGF1 signaling in vivo with SU5402, a specific FGF1 receptor antagonist. Thus, Sumf1^−/− mice were i.p. injected at postnatal day (P) 7 with a suspension of 500 µM SU5402 adsorbed onto beads (to allow slow SU5402 release). The mice were then analyzed at P21–P27 (Fig. 7 A). Upon SU5402 treatment, the frequency of intermediate precursors and terminally differentiated CD71⁻Ter119⁺ cells significantly increased in the BM of Sumf1^−/− mice, demonstrating a rescue of the block of erythroid lineage differentiation (Fig. 7 B). Similarly, the myeloid lineage was unlocked, and comparable frequencies of monocytes and granulocytes were seen, in respect to those in the BM of WT littermates (Fig. 7 C). Moreover, SU5402 also increased the mature B lymphocytes in treated mutants, as compared with untreated mice (Fig. 7 D). Although SU5402 treatment did not affect the pro–T frequency in the BM (Fig. 7 E), thymocytes developed normally; comparable DP and SP T lympocyte frequencies were seen in the treated mutants, both for the total (Fig. 7 F) and for electronically gated CD3⁺ populations (not depicted), in respect to WT mice. We did not observe any correction/restoration of the erythrocyte and B lymphocyte proportions in the spleen of treated mice, which was probably because we analyzed the mice 2–3 wk after SU5402 administration, a time that is too early to see any rescue effect (Fig. 7 G).

Collectively, these findings demonstrate that SUMF1 controls hematopoietic lineage differentiation via modulation of the FGF pathway.

To demonstrate that differentiation defects of Sumf1^−/− HSPCs are cell autonomous and independent of the cell niche, we transplanted these into WT recipient mice. Purified WT HSPCs and Sumf1^−/− HSPCs were transduced with a LV vector carrying GFP cDNA, to follow their long-term fate after transplantation (Biffi et al., 2004). LV-transduced HSPCs (7 ×10⁵) were transplanted into lethally irradiated, 2-mo-old WT recipient mice. 10 wk after noncompetitive transplantation, the hematopoietic organs were collected and analyzed by flow cytometry, as described above. The cells of donor origin were analyzed according to GFP expression. This analysis showed a considerable reduction in the total cell number in the BM, spleen, and thymus of mice transplanted with Sumf1^−/− HSPCs, with respect to mice transplanted with WT cells (Fig. 8 A), as also seen in the donor mice. Interestingly, the overall engraftment of GFP⁺ Sumf1^−/− and WT HSPCs were comparable (Fig. 8 B). All further analyses were performed on the GFP⁺ population to avoid confounding effects from residual autologous cells. When we analyzed terminally differentiated cells in the BM, spleen, and thymus of the transplanted mice, a block in erythroid lineage differentiation of the Sumf1^−/− HSPCs transplanted mice was clearly seen, as strong reductions in the frequencies of CD71^lowTer119⁺ chromatophilic erythroblasts and of mature CD71⁻Ter119^high orthochrotomaphilic erythroblasts (Fig. 8 C). Notably, and probably because of the limited time of observation (10 wk after the transplant), in these mice we did not detect any accumulation of CD71⁺Ter119⁺ cells in the BM, although the same compensatory increase in Ter119⁺ cells was seen in the spleens of the donors (Fig. 8 D).

Impaired B cell development and aberrant T cell development were seen in the Sumf1^−/− HSPCs transplanted mice, which showed a decrease in pre–/pro–B cells, low percentages of terminally differentiated B cells in the spleen, a reduction in DP CD4⁺ CD8⁺ cells in the thymus, and an increase in SP CD4⁺ and SP CD8⁺ thymocytes (Fig. 8, E–G). In addition, as for the donor mice, there was aberrant CD3⁻ SP CD4⁺ and SP CD8⁺, without expression of the TCR accessory protein (unpublished data).

Furthermore, although the number of granulocytes and monocytes in the BM were similar in both of these groups of transplanted mice, there was a slight decrease in Gr-1⁺ cells in the spleen of the Sumf1^−/− HSPCs transplanted mice, leading us to envision a compensatory mechanism in the BM, with mutated cells that might unlock GMP progression into differentiation (unpublished data).

Finally, we analyzed the HSPC fraction in both of these groups of transplanted mice. Interestingly, there was a slight increase (although not statistically significant) in the frequency of KLS cells in the mice transplanted with Sumf1^−/− GFP⁺ HSPCs (Fig. 9 A). Within the KLS cells, we saw a decrease in the frequency of CD150⁺ LT-HSCs and a consequent increase in the ST-HSC/MPP fraction in the group of Sumf1^−/− HSPC transplanted mice (Fig. 9 B), suggesting that the SUMF1 defect might render primitive HSC more prone to differentiate into ST-HSCs and MPPs. Furthermore, there were relevant increases in CLPs and CMPs in the Sumf1^−/− HSPC transplanted mice (although, not statistically significant; Fig. 9 C).

In conclusion, transplantation of Sumf1^−/− HSPCs reproduces the phenotype of Sumf1^−/− mice in a WT host, leading us to conclude that SUMF1 has an important role in different steps of the lineage differentiation process of HSPCs.

HSCs represent a rare population of adult stem cells that can give rise to all blood cell lineages in a hierarchical process that occurs daily in adult mammals (Morrison and Weissman, 1994; Morrison et al., 1995; Kondo et al., 1997; Akashi et al., 2000; Laiosa et al., 2006a). The more the cells proceed in the developmental progression, the more restricted they become in their differentiation potential, losing their self-renewal ability. Cytokines, hormones, and activation of different signaling pathways control these well-tuned developmental processes that couple and integrate extrinsic and intrinsic fate determinants. Developmentally conserved signaling pathways have emerged as important controlling devices for HSC fate, including Wnt, FGF, Notch, sonic hedgehog (Shh), and bone morphogenetic protein (BMP; Esko and Lindahl, 2001).

Several studies have reported on the role of the Wnt/β-catenin pathway in controlling self-renewal and differentiation of HSCs in vivo and ex vivo, with apparent discrepancies seen. Some studies claim that stabilization of β-catenin is positive for self-renewal and differentiation of HSCs (Reya et al., 2003; Willert et al., 2003; Baba et al., 2005), whereas others assert that β-catenin stabilization impairs HSC repopulation and engraftment in a recipient (Kirstetter et al., 2006; Scheller et al., 2006). We believe that fixed β-catenin thresholds at specific developmental stages can provide different phenotypes. This is evident in different biological contexts, such as during cardiac differentiation: differentiating embryonic stem cells treated early with Wnt-3a undergo cardiac differentiation; conversely, late activation of the Wnt pathway reduces their cardiac differentiation (Ueno et al., 2007). Periodic β-catenin accumulation in embryonic stem cells at a specific threshold can lead to reprogramming of somatic cells after fusion. In contrast, high or low levels of β-catenin can affect the ability of embryonic stem cells to reprogram somatic cells (Lluis et al., 2008).

This scenario is even further complicated, as many signaling pathways show cross-talk, and their interactions will determine the different developmental fates in blood lineages. An interaction between the Shh and BMP-4 pathways is evident in the finding that Shh can induce proliferation of primitive hematopoietic cells via BMP-4 (Bhardwaj et al., 2001). However, BMP-4 can maintain HSCs in culture, but cannot control their expansion (Utsugisawa et al., 2006). FGFs and Wnts have also been shown to interact in a variety of developmental systems, including brain, kidney, and tooth in some vertebrates (Moon et al., 1997), and FGF signaling controls proliferation and subsequent lineage commitment of neural stem cells through a concerted action with β-catenin (Israsena et al., 2004). The role of the FGF signaling pathway in the development of blood lineages is poorly understood, with FGF1 and FGF2 used for ex vivo culture of LT-HSCs, together with some additional growth factors and cytokines (Yeoh et al., 2006). In contrast, TGF-β is generally accepted to be a potent inhibitor of HSC expansion in vitro (Sitnicka et al., 1996; Batard et al., 2000).

So, overall, very little is known about the cross-talk between these signaling pathways. It is certain that the influences they have on each other will finally regulate the complex hierarchical development of different blood lineages. If one signaling pathway is imbalanced or dysregulated, it might lead to cells of one lineage being reprogrammed into another lineage, thus resulting in a change in cell fate. This can arise from aberrant expression of key transcription factors, such as GATA-1, GATA-3, C/EBPα, and PU.1 (Orkin and Zon, 2008). Indeed, it has been shown that myeloid transdifferentiation of pro–B cells is caused by ectopic expression of c/EBPα (Heavey et al., 2003). This might also account for the compensation of the granulocytes that we see in the Sumf1^−/− mice.

Here, we have shown direct interconnections between at least three different pathways: FGF, Wnt/β-catenin, and Notch signaling. These are regulated by the activity of SUMF1, probably via the Sulf1 and Sulf2 sulfatases. We have shown that activation of the FGF signaling pathway leads to increases in ERK phosphorylation that, upon GSK-3 inactivation, trigger β-catenin accumulation. In thymocytes, we observed FGF-dependent accumulation of β-catenin and Notch activation. Interestingly, SUMF1 controls autophagy, proliferation, and differentiation of chondrocytes by limiting FGF signaling (Settembre et al., 2008). The constitutive activation of FGF and Wnt/β-catenin leads to an increase in ST-HSCs, as clearly seen in the recipient transplanted mice. As a result, a lack of terminally differentiated cells occurs because of developmental failure in the three lineages; in addition, aberrant thymocytes also develop. The concerted activities of accumulated β-catenin and Notch1 might, in part, stall the transition of thymocytes from the double-negative to the DP stage, and increase the levels of SP CD4⁺ or CD8⁺ thymocytes; indeed, it has been shown that β-catenin can, in part, transcribe the CD4 gene (Huang et al., 2006) and increase the timing and maturation rate of CD8⁺ cells (Yu and Sen, 2007; Staal and Sen, 2008). The rescue of erythroid lineage differentiation and the normal development of T lymphocytes after FGF1 signaling inhibition in Sumf1^−/− mice further reinforce the observation that FGF activation controls Wnt and Notch signaling activation in the differentiation of HSPCs.

In the Sulf2^−/− mice, there was a slight increase in pro-erythroblasts, a small decrease in BM B mature cells, a decrease in DP thymocytes, and an increase in SP CD4⁺ cells, and we would expect lineage defects to be even more severe in the Sulf1^−/− Sulf2^−/− double-KO mice because of the redundancy in the activities of these two sulfatases (Ratzka et al., 2008). This double-KO phenotype is very severe and the mice die very early (Holst et al., 2007); indeed, we only obtained a sufficient number of these Sulf1^−/− Sulf2^−/− double-KO mice for the biochemical analysis here. Sulf1^−/− Sulf2^−/− HSPCs showed constitutive activation of FGF and Wnt signalings. Unfortunately, because of the severe phenotype and the lack of Sulf1^−/− Sulf2^−/− double-KO mice, we could not carry out the BM phenotype analysis. In contrast, Ids^−/− and Sgsh^−/− mice did not show any of the defects seen in the Sumf1^−/− strain, ruling out any involvement of these other sulfatases in the hematopoietic lineage phenotype.

Mesenchymal stromal cells (MSCs) reside in the niche that supports stem cell activity in vivo. MSCs from Sumf1^−/− mice show constitutive activation of the FGF and Wnt signaling pathways. These might induce irreversible conditioned functions in the HSCs. Osteoblasts overexpressing the Wnt inhibitor Dkk1 have irreversible effects on HSC function, reducing their ability for long-term repopulation in serial transplant experiments (Fleming et al., 2008). It is possible that concerted dysregulation of the FGF, Wnt/β-catenin, and Notch signaling pathways in both the MSC niche component and in the HSCs themselves forces the HSC pool to enter into the cell cycle, and thus to switch them toward a short-term reconstituting phenotype, with a consequent decrease in terminally mature cell development.

Moreover, we have also show that these altered interacting signaling pathways induce lineage defects in a mouse model of MSD. Patients affected by MSD show a lack of activity of their entire family of sulfatases, and the consequent accumulation of undegraded sulfatase substrates in their lysosomes. In addition, these patients have developmental problems that can be ascribed to a lack of activity of Sulf1 and Sulf2, the only sulfatases that can modulate signaling pathways. The lysosomal storage in murine models of sulfatase deficiencies does not cause defects in lineage development. In further support of these data, we recently showed that invariant natural killer T cells develop aberrantly only in the MSD mouse model, and not in lysosomal disease models (Plati et al., 2009), ruling out a role for lysosomal engulfment in lineage development.

Sulf1^−/− Sulf2^−/− double-KO mice and Sumf1^−/− mice show very high perinatal mortality (Holst et al., 2007), and our data lead us to believe that blood lineage failure might contribute to this mortality. Alternatively, the MSD disease is very heterogeneous with respect to the developmental features, which is possibly caused by the escape of β-catenin stabilization in patients carrying less-detrimental SUMF1 mutations. Interestingly, we observed that even in mixed background strains, the BM phenotype of Sumf1^−/− mice is relatively heterogeneous, with some mice showing less severe defects (unpublished data).

All in all, our data highlight that regulation through the sulfation state of the HSPGs, which provide a repository of growth factors such as Wnt and FGF, should be finely tuned. This is also modulated by the Sulf activities (Dhoot et al., 2001; Ai et al., 2003; Viviano et al., 2004; Wang et al., 2004). Thus, we hypothesize that only a small fraction of Sulf proteins have the FGly residue and can cleave the sulfate of the heparan sulfate in HSPGs; the majority of the Sulfs should be activated by SUMF1. This represents a fine regulation mechanism, and SUMF1 appears to be the master controlling factor here.

The strains used in this study were as follows: Sumf1^−/− (Settembre et al., 2007), Sulf1^−/−, Sulf2^−/− (Ai et al., 2003), Ids^−/− (Cardone et al., 2006), and Sgsh^−/− (Fraldi et al., 2007). All the procedures were approved by the Animal care and use Committee of Telethon Institute of Genetics and Medicine and were communicated to the Ministry of Health and local authorities, according to Italian law.

A FACSCanto instrument (BD) was used for the FACS analysis. Phycoerythrin-conjugated anti–mouse mAbs to c-kit, CD34, CD71, CD3, CD4, IgM, and Gr-1; allophycocyanin-conjugated anti–mouse mAbs to Ter119, CD11b, CD16, CD25, and CD8; and phycoerythrin-conjugated-cycrome5 anti–mouse mAbs to B220, CD3, and 7-amino-actinomycin D were purchased from BD. Phycoerythrin-conjugated-cycrome7 anti–mouse mAbs to Sca1 and c-kit; allophycocyanin-conjugated-Alexa Fluor 750 anti–mouse mAbs to c-kit and CD8; and allophycocyanin-conjugated CD150 were purchased from eBioscience. The FITC-conjugated anti–mouse mAb to CD127 was purchased from Abcam, and the phycoerythrin-conjugated anti–mouse mAb to CD115 was purchased from Serotec. Isolated BM, spleen, and thymic cells were prepared for flow cytometry by blocking the aspecific sites using anti-Fc blocking reagent from BD, and then incubating the cells with the appropriate antibodies or control isotype-matched antibodies, followed by PBS washes.

2-mo-old C57Bl6 mice were lethally irradiated for 4 h before transplantation (10 Gy, divided into two doses). Next, 7 × 10⁵ WT or Sumf1^−/−-transduced HSPCs were injected into the tail veins of the recipient mice. The transplanted mice were sacrificed 10 wk after transplantation, and their BM, spleen, and thymus were analyzed.

Analyses were performed by one-way and two-way ANOVA, using Bonferroni correction as post-test, or Student’s t test/Mann Whitney tests, and the data are expressed as means ± SD; p-values < 0.05 were considered statistically significant.

Total BM was flushed out from femurs and tibia of 3-wk-old and age-matched WT and Sumf1^−/− mice. For in vitro experiments, the lineage-fraction was purified from total BM cells by untouched isolation on a magnetic sorter, using the Lineage Depletion kit (Miltenyi Biotec), according to the manufacturer protocol. The HSPCs were cultured for two days in Stem Span Sfem medium (STEMCELL Technologies) at a concentration of 10⁶ cells/ml in the presence of 100 ng/ml recombinant mouse SCF (R&D Systems), 20 ng/ml mouse IL3 (Peprotech), 20 ng/ml human IL6 (Peprotech), and 10 ng/ml Flt3 ligand (Peprotech). They were then harvested for all of the analyses. HSPCs for the in vivo studies were purified from total BM cells by untouched isolation using the Murine Progenitor Enrichment Cocktail (STEMCELL Technologies).

Total BM was flushed out from femurs and tibia of 3-wk-old and age-matched WT and Sumf1^−/− mice. The total cells were cultured for 3 d in DME with 10% FBS. The medium was then changed to remove nonadherent cells. The mesenchymal stromal cells (MSCs) were grown until confluence, and then harvested for analysis.

Splenocytes were obtained by mechanic disaggregation of the spleen in RPMI with 10% FBS, 100 IU/ml penicillin, 100 µg/ml streptomycin, 2 mM l-glutamine, supplemented with 28.4 mM 2-mercaptoethanol. The thymocytes were obtained by mechanic disaggregation of the thymus in IMDM with 20% FBS, 100 IU/ml penicillin, 100 µg/ml streptomycin, and 2 mM l-glutamine. After centrifugation at 1,500 rpm for 5 min at room temperature, 2.5 ×10³ cells were resuspended in 50 µl blocking solution.

Freshly isolated HSPCs from WT and Sumf1^−/− mice were transduced with PGK.GFP.LV or 4xTCF/LEF.mCMV.eGFP or 4xERK.mCMV.eGFP, at a multiplicity of infection (MOI) of 100 in Stem Span Sfem medium (Stem Cell Technologies), at a concentration of 10⁶ cells/ml in the presence of 100 ng/ml recombinant mouse SCF (R&D Systems), 20 ng/ml mouse IL-3 (Peprotech), 20 ng/ml human IL-6 (Peprotech), and 10 ng/ml Flt3 ligand (Peprotech) for 12–16 h. The transduction efficiency was evaluated at the clonal level on colonies from the colony-forming cell assay or by FACS analysis.

Total homogenates from WT and Sumf1^−/− HSPCs were obtained using SIE buffer (250 mM saccharose, 3 mM imidazole, pH 7.4, 1% ethanol, and 1% NP-40). 10 µg total cell homogenates were incubated with 10 µM estrone 3-O-sulfamate at 37°C for 1 h, to inhibit steroid sulfatases, before addition of 10 mM 4-MUS substrate in 10 mM lead acetate buffer, for a total of 200 µl final reaction mixture. The reaction was stopped after 24 h by adding 1 ml of 0.5 M Na₂CO₃/NaHCO₃, pH 10.7, and read at 460 nm using a VersaFluor Fluorometer (Bio-Rad Laboratories). The activities were expressed as nanometers/hour/milligram.

Total cell lysates were obtained in RIPA buffer (50 mM Tris, 250 mM NaCl, 0.1% SDS, 0.5 Na-deoxycholate, and 1% NP-40). Total protein (50 µg) from each lysate was subjected to SDS-PAGE under reducing conditions. After electrophoresis, the protein was transferred onto nitrocellulose filter membranes (Immobilon-P; Millipore) using a Trans-Blot Cell (Bio-Rad Laboratories) and transfer buffer containing 25 mM Tris, 192 mM glycine, and 20% methanol. The membranes were then placed in 5% nonfat milk in TBS plus 0.5% Tween 20 (TBST) at 4°C for 2 h, to block nonspecific binding sites. The filters were incubated with specific antibodies before being washed three times in TBST and then incubated with a peroxidase-conjugated secondary antibody (GE Healthcare). After further washing in TBST, peroxidase activity was detected using the ECL system (Thermo Fisher Scientific).

Total RNA was extracted from WT and Sumf1^−/− HSPCs using the Agilent Total RNA Isolation Micro kit 50 (Agilent Technologies), according to the manufacturer protocol. The eluted RNA was reverse-transcribed with SuperScript III (Invitrogen) according to the manufacturer protocol. Quantitative real-time PCR reactions were set up in duplicate using Platinum SYBR Green qPCR SuperMix-UDG with ROX (Invitrogen), and amplification was performed using a 7,000 ABI Real-Time PCR machine. Real-time PCR was performed using 12.5 µl SYBR green Master Mix, 2.5 µl of each primer (10 µM each), and DNase- and RNase-free water for a final reaction volume of 25 µl, with the following parameters: 95°C for 3 min, followed by 40 cycles of 95°C for 10 s, 60°C for 20 s, and 72°C for 10 s. Primers were designed using PrimerExpress software (Applied Biosystems). Data analyses were performed using the Applied Biosystems SDS software, version 1.2.3. All experiments were performed in triplicate, and differences in cDNA input were “compensated” for by normalization to expression of GAPDH. The primers used were as follows: Axin2 primers mix (SuperArray Bioscience Corporation); GATA-1 primers mix (SuperArray Bioscience Corporation); GPC3 primers mix (SuperArray Bioscience Corporation); Hes-1 (forward) 5′-AAAGCCTATCATGGAGAAGAGGCG-3′ and (reverse) 5′-GGAATGCCGGGAGCTATCTTTCTT-3′; Notch1, (forward) 5′-GCAGCCACAGAACTTACCACTCCAG-3′ and (reverse) 5′-TAAATGCCTCTGGAATGTGGGTGAT-3′; Deltex-1, (forward) 5′-GCCATGTACTCCAATGGCAACAAG-3′ and (reverse) 5′-CGGGATGAGGTGAAACTCCATCTT-3′; C/EBPα, (forward) 5′-GGCGGCATCTGCGAGCACGA-3′ and (reverse) 5′-GGCTGTGCTGGAAGAGGTCG-3′; GAPDH, (forward) 5′-ACTCCCACTCTTCCACCTTC-3′ and (reverse) 5′-TCTTGCTCAGTGTCCTTGC-3′.

50 µl of packed AG1-2X resin (Bio-Rad Laboratories) was incubated in the presence of 500 µM SU5402 (Calbiochem) for 1 h at room temperature. The suspension was than centrifuged to remove the excess SU5402, washed twice in PBS, and resuspended in 100 µl PBS, for i.p. injection.

Gating strategies for stem cells and progenitors were applied according to the Weissman Laboratory protocol (Irving L. Weissman, Institute of Stem Cell Biology and Regenerative Medicine, Stanford Cancer Center, Stanford University School of Medicine, Stanford, CA). In brief, the following populations were identified: KLS, staining was performed on lineage-selected cells, with KLS cells defined as c-kit⁺, Sca1⁺; CLPs, staining was performed on lineage-selected cells, with CLP cells defined as c-kit⁺, Sca1⁺, CD127⁺; CMPs, staining was performed on lineage-selected cells, with CMP cells defined as c-kit⁺, Sca1⁻, CD127⁻, Cd34⁺, CD16^low; MEPs, staining was performed on lineage-selected cells, with MEPs defined as c-kit⁺, Sca1⁻, CD127⁻, CD34⁻, CD16^+/low; GMPs, staining was performed on lineage-selected cells, with GMPs defined as c-kit⁺, Sca1⁻, CD127⁻, Cd34⁺, CD16^high; pro–B cells, staining was performed on total BM, with pro–B cells defined as B220^+/low, IgM⁻, CD127⁺; immature B cells, staining was performed on total BM, with immature B cells defined as B220^+/low, IgM⁺; mature B cells, staining was performed on total BM, with mature B cells defined as B220^bright, IgM^+/−; pro–T cells, staining was performed on total BM, with pro–T cells defined as CD127⁺, CD3, CD4, CD8^-/low, CD25⁻, c-kit⁺.

Colony-forming assays were performed by plating 5 × 10³ untransduced or transduced, WT, and Sumf1^−/− HSPCs in methylcellulose medium (mouse MethoCult; Stem Cell Technologies) supplemented with cytokines. After 1 wk, the colonies were scored under light or fluorescence microscopy, to evaluate the percentages of GFP⁺ cells.

Fig. S1 shows the Western blot analysis of phospho-FSR2 and (Glypican 3) GPC3 mRNA in WT and Sumf1^−/− HSPCs. Fig. S2 shows GATA-1 and c/EBPα mRNA expression in WT and Sumf1^−/− HSPCs. Fig. S3 shows the analysis of BM, spleen, and thymus cellularity in WT, Sumf1^−/−, Ids^−/−, Sulf1^−/−, and Sulf2^−/− mice. Fig. S4 shows analysis of the erythroid lineage in Ids^−/−, Sulf1^−/−, and Sulf2^−/− mice. Fig. S5 shows analysis of lymphoid lineage in Ids^−/−, Sulf1^−/−, and Sulf2^−/− models.

The authors thank A. Baldini, A. Ballabio, G. Diez-Roux and C. Settembre for critical reading of the manuscript, X. Ai for comments on the manuscript and for providing the Sulf1^−/− Sulf2^−/− mice, C. Procaccini and G. Matarese for useful suggestions on the experiments, B. Gentner for precious suggestions and help in the cytofluorimetric analysis, and L. Mele for technical support.

We are grateful for financial support from the Telethon Foundation (M.P. Cosma and A. Biffi) and from Associazione Italiana Ricerca sul Cancro (M.P. Cosma).

The authors declare that they have not competing financial interests.