Retinoid-Regulated FGF8f Secretion by Osteoblasts Bypasses Retinoid Stimuli to Mediate Granulocytic Differentiation of Myeloid Leukemia Cells

Retinoic acid (RA) therapy for acute promyelocytic leukemia (APL) represents the most successful example of differentiation–induction therapy in current clinical oncology (1–3); however, this success has not extended to the remaining 85% of myeloid leukemia subtypes, largely because the mechanisms of RA-induced myeloid differentiation remain unclear. RA signaling is elicited through both the classical genomic pathway and rapid non-genomic signal transduction. The role of the classical RA-induced genomic pathway (4) in mediating cancer cell differentiation (5, 6) is well recognized (7). On the other hand, RA can exert rapid non-genomic effects independently of retinoic acid receptor (RAR)-mediated gene transcription (8) to induce cell differentiation or apoptosis (9, 10). Such non-genomic effects are mediated through cytoplasmic signaling kinases, for example, mitogen-activated protein kinase (MAPK) pathways related to phosphatidyl inositide-3 kinase (PI3K) or extracellular signal–related kinase (ERK) signal transduction (10, 11). Moreover, recent insights into the role of osteoblast-formed hematopoietic stem cell (HSC) niche cues in modulating HSC development (12, 13) and leukemogenesis (13, 14) have gained attention on RA-mediated epigenetic regulation of terminal differentiation of myeloid leukemia cells.

It is known that RA-activated RARα regulates myeloid differentiation by transcriptional regulation of differentiation target gene (4, 6, 7). RARα, a phosphoprotein and transcription factor, is a substrate of CAK complex (6, 15). Human CAK is an enzyme consisting of cyclin-dependent kinase 7 (CDK7; ref. 16), cyclin H (17), and MAT1 (18). Serine 77 (Ser-77) in the ligand-independent AF-1 domain of RARα (RARαS77) is the main residue phosphorylated by CAK (6, 15). CAK exists in cells either as free CAK or as part of the general transcription factor IIH (TFIIH) complex. Both free CAK and TFIIH-containing CAK phosphorylate Ser-77 of RARα (6, 15). Free CAK controls G₁ exit, a stage in which cells commit to proliferation or to differentiation (19–21). We have found that either RA-decreased CAK phosphorylation of RARα or expression of the phosphorylation-defective RARαS77 mutant, RARαS77A, not only mediates granulocytic differentiation of both malignant and normal hematopoietic precursors (20–22) but also regulates osteoblastic differentiation and inhibits osteosarcoma formation (23). These studies show that RA-induced CAK-RARα signaling is involved in regulating differentiation of both osteoblastic and hematopoietic precursors.

Osteoblastic cells represent a regulatory component of the bone marrow microenvironment that mediate myelopoiesis, B-lymphocyte commitment, and HSC plasticity (12, 13, 24, 25), whereas disruption of this niche signal induces leukemogenesis (13, 26). Osteoblasts derived from human mesenchymal stem cells constitute the HSC niche (13, 24), and RA induces osteoblastic differentiation of both mesenchymal stem cells and osteosarcoma cells (23, 27). Striking success in epigenetic reversion of the genetic malignant phenotypes, as exemplified by RA treatment of APL harboring a PML-RARα fusion gene (2), provides proof of concept that RA-mediated HSC niche signaling can effect changes in the differentiation state of myeloid leukemia cells, even as the genome remains malignant and unstable (28). Previous studies strongly indicated the existence of a reciprocal relationship between osteoblasts and hematopoietic cells (12, 29), but the dimension of these interactions is yet to be defined. Because RA-induced loss of CAK phosphorylation of RARα or phosphorylation-defective RARαS77A mediates osteoblastic differentiation pathway through induction of FGF8f (23), we sought to investigate whether this osteoblast-derived FGF8f mediates granulocytic differentiation in a paracrine manner. Our studies show that osteoblast-secretion of FGF8f induced by either RA or RARαS77A regulates terminal granulocytic differentiation of myeloid leukemic cells, revealing a novel CAK-RARα signaling induced by RA to coordinate granulocytic differentiation at the paracrine level.

Human myeloid leukemic HL60, HL60R (RA-resistant), NB4 (APL), and human osteosarcoma U2OS cells were cultured as described previously (20, 21, 23). Cells within 5 to 15 passages of HL60 and U2OS cell lines, expanded immediately after receiving the cells from the American Type Culture Collection, were used for less than 5 months. HL60R (20) and NB4 cells (21) were tested to be Mycoplasma free by PCR methods after receiving cells from our collaborators, and each of those 5 to 15 passages of HL60R and NB4 cells was used for less than 5 months. The cancer cells were authenticated by their ability to form cancers in non-obese diabetic/severe combined immunodeficient, and/or nude mice. Normal human primitive hematopoietic CD34⁺ cells were obtained from AllCells and maintained with myeloid medium as described (22). The myeloid medium adapted for inducing granulopoiesis (MM-G) is supplemented with hydrocortisone for blocking the growth of lymphoid cells, which eliminates erythropoietin to inhibit the growth of erythroid cells (22). CD34 cells, certified to be HIV- and Mycoplasma-free by AllCells, were cultured for maxi mum 12 days without passaging after their initial expansion by following the manufacturer's instructions. ATRA (RA) was from Sigma. RA concentration of 1 μM was used in the experiments. Recombinant human FGF8f was from R&D Systems.

Granulocytic differentiation, as judged by morphologic nuclear segmentation, has been described previously (20). Briefly, cells were cytocentrifuged for 5 minutes at 400 rpm in a Cytospin, fixed by using methanol, and stained with Wright–Giemsa stain (Sigma). The morphologic indicators of differentiation (nuclear/cytoplasmic ratio, nuclear shape, and degree of nuclear segmentation) were evaluated under a Zeiss Axioplan microscope. Images were color balanced in Adobe Photoshop.

U2OS cells treated with RA or transduced with lentiviral pCCL-FGF8f or vector (Supplementary Fig. S1) were grown in 24-well plates. After reaching 70% to 80% confluence, the cells were washed and cultured for 21 days with bone differentiation medium [culture medium supplemented with 10 nmol/L dexamethasone (Sigma, # D2915), 20 mmol/L β-glycerolphosphate (Sigma, # G9891), 50 μmol/L L-Ascorbic acid 2-phosphate (Sigma, # A8960)]. Cells were then fixed with 10% buffered formalin, and bone differentiation was judged by matrix mineralization as described previously (30) using Alizarin Red S (ARS; Sigma) staining.

Cell duplication was determined by cell count as described previously (31).

Transduction of U2OS cells with lentiviral human RARαS77A or FGF8f (23) has been described previously (22).

Western blotting was done as described previously (20). Antibodies for FGF8, P-p38-MAPK (Thr 180/Tyr 182), P-p42/44-ERK (Thr-202/Tyr-204), p38-MAPK, p42/44-ERK, and β-actin were from Santa Cruz Biotechnology.

The retained RA in the RA-osteoblast-conditioned medium (OCM) was monitored by using F9 reporter cells containing a LacZ reporter gene as described previously (32). X-gal was from Promega.

cDNAs were generated from DNase I–treated RNA (iScript; Bio-Rad), and quantitative real-time PCR (qRT-PCR) was carried out with QuantiTect SYBR Green PCR kits (Qiagen) by following the manufacturer's instructions, using 384-well optical plates on the 7900HT Fast qRT-PCR System (Applied Biosystems). The housekeeping gene GAPDH was used as an internal control. Two negative controls were carried in parallel through all steps of the experiments. Standard curves for cDNA were composed of three 10-fold dilutions of control cDNA. The primers sequences and amplification conditions are given as Supplementary Table S1.

To assay secreted FGF8f, media conditioned by RA treatment of U2OS cells and FGF8f (or RARαS77A) overexpression in U2OS cells were harvested. These different OCM were concentrated using Pierce microconcentrators (Pierce Biotechnology, Inc.), and the protein quantity was measured by the improved Lowry assay using Bio-Rad DC Protein Assay (33). The presence of FGF8f in the OCM was confirmed by Western blot analysis.

Antibody neutralization of FGF8f has been described previously (23). Blocking peptide against FGF8 antibody (sc-27144P; Santa Cruz Biotechnology) was used as described previously (23).

Flow cytometry using a direct immunofluorescence staining was carried out as described (20) for analysis of myeloid differentiation marker CD11b. CD11b-APC and CD34-PE antibodies were from BD Biosciences. Corresponding isotypes conjugated to irrelevant antibodies, isotype APC and isotype PE (BD Biosciences), were used as controls.

Student unpaired 2-tailed t test was used when appropriate.

Because the HSC niche formed by osteoblastic cells mediates HSC development (13, 24, 25) through secreting its regulatory signaling molecules (24, 25, 34), while RA stimuli or expression of a phosphorylation-defective RARαS77 mutant, RARαS77A, in osteosarcoma U2OS cells upregulates FGF8f expression (23), we investigated whether such upregulated FGF8f can be sufficiently secreted by U2OS cells and in turn directly mediates osteoblastic differentiation of U2OS cells. In a series of experiments using U2OS cells treated with RA or transduced with lentiviral RARαS77A or FGF8f (Supplementary Fig. S1), we found that overexpression of RARαS77A or FGF8f mimicked the effect of RA or RARαS77A on inducing cellular morphologic change (Fig. 1A), upregulating FGF8f secretion (Fig. 1B), inhibiting proliferation, and enhancing expression of osteoblastic differentiation regulator osteopontin (Fig. 1C). Moreover, bone differentiation analysis showed that, similar to RA stimuli, overexpression of FGF8f induces bone differentiation (Fig. 1D, top). Furthermore, to verify the effect of FGF8f on inducing bone differentiation, we decided to carry out antibody neutralization of FGF8f. Because of the special genetic structure of FGF8f shared with its alternatively spliced products FGF8e, FGF8b, and FGF8a (23, 35), an antibody specifically against FGF8f cannot be produced. However, in U2OS cells, FGF8f is the main isoform induced by RA or RARαS77A, and FGF8 antibodies only recognize a major 30-kDa protein, corresponding to the molecular mass of FGF8f (23). We, therefore, used FGF8 antibodies to neutralize the effect of FGF8f, as described previously (23), on inducing osteoblastic differentiation. We found that removal of FGF8f activities by FGF8 antibody neutralization significantly reduced bone matrix deposition of U2OS cells, as judged by ARS staining (Fig. 1D, bottom). These data collectively show that RA-induced FGF8f is sufficiently secreted by U2OS cells, which exhibits a direct regulatory effect on mediating osteoblastic differentiation at the autocrine level.

Osteoblast-formed HSC niche secretes signaling molecules to cross-mediate myelopoiesis (24, 25, 34), whereas FGF8f secreted by U2OS cells treated with RA or transduced with lentiviral FGF8f or RARαS77A (Fig. 1; ref. 23) is functionally active, as reflected by mediating osteoblastic differentiation in an autocrine manner (Fig. 1D). We, therefore, tested whether such osteoblast-secreted FGF8f can function in the absence of RA at the paracrine level to inhibit proliferation of myeloid leukemia cells, an indispensable biologic process required for induction of granulocytic differentiation. By coculturing HL60 cells with U2OS cells overexpressing FGF8f or RARαS77A, we found that the proliferation of both RA-sensitive HL60 and RA-resistant HL60R cells was inhibited (Fig. 2A). Because HL60R cells harbor a truncated ligand-dependent AF-2 domain of RARα (36, 37), the results show that osteoblast-derived paracrine FGF8f can bypass the upstream defect occurring at the levels of RA:RARα interaction to inhibit proliferation of RA-resistant leukemic myeloblasts, similar to the effect of RA on inhibiting proliferation of RA-sensitive cells. Moreover, by culturing NB4 cells, which harbor a PML-RARα fusion gene (2), as well as HL60 cells in the presence of either RA or OCM collected from U2OS cells transduced with RARαS77A (S77A-OCM) or FGF8f (FGF8f-OCM), we found that cell proliferation in both HL60 and NB4 cells was inhibited (Fig. 2B). These results indicate that, in the absence of RA, osteoblast-secreted FGF8f, a downstream target of RA signaling, is capable of inhibiting proliferation of both RA-sensitive and RA-resistant leukemic myeloblasts at the paracrine level.

Because the effect of RA signaling is identified with the induction of granulocytic differentiation (1, 3) and because differentiation can only be induced in cells where proliferation is arrested, we wondered whether RA-induced inhibitory effect mediated by paracrine FGF8f on cell proliferation (Fig. 2A and B) could lead to granulocytic differentiation of myeloid leukemic cells. To address this issue, HL60 cells were treated with RA or cultured with S77A-OCM or vector-OCM. Granulocytic morphology analysis indicates that FGF8f secreted into S77A-OCM (Fig. 1B, middle) induced HL60 cell differentiation (Fig. 2C). Correspondingly, the levels of CD11b, a myeloid differentiation marker expressed on myelocytes and more mature granulocytes, were induced by either RA or FGF8f-OCM in the absence of CD34⁺ expression, compared to blank or vector controls [Fig. 2D; (iii) and (iv) vs. (i) and (ii)]. Therefore, these data indicate that RA-mediated induction of paracrine FGF8f by osteoblasts instigates HSC niche signaling, in the absence of RA, to inhibit proliferation and induce differentiation of leukemic myeloblasts.

To verify the inhibitory effect of paracrine FGF8f on leukemic cell proliferation, we chose to use FGF8 antibody neutralization of paracrine FGF8f effect, as described above (Fig. 1D). We cultured both HL60 and HL60R cells with FGF8f-OCM in the presence or absence of FGF8 antibodies as described previously (23). The results showed that whereas FGF8f-OCM inhibited cell proliferation in either HL60 or HL60R cells, this inhibitory effect of paracrine FGF8f was neutralized by FGF8 antibodies (Fig. 3A). Moreover, because FGF8f is a downstream target of RARαS77A (Fig. 1B, middle; ref. 23); we next verified whether the inhibitory effect of both RARαS77A- and FGF8f-OCM on cell proliferation is FGF8f-specific. We found that, in the presence of negative (blank, vector) and positive (RA-OCM) controls, FGF8 antibodies neutralized the inhibitory effect of either RARαS77A-OCM or FGF8f-OCM-containing FGF8f on cell proliferation, while the blocking peptides of FGF8 antibodies overrode such counteraction of FGF8 antibodies (Fig. 3B). In parallel, the RA levels in RA-OCM were monitored to not exceed physiologic levels (10⁻⁹ mol/L) by using F9 RA-reporter cells (Supplementary Fig. S2). Therefore, the above-described data show that OCM-containing paracrine FGF8f is required for inhibiting proliferation of both RA-sensitive and RA-resistant leukemic myeloblasts.

To show that the observed effects of paracrine FGF8f are not due to unknown cytokines cosecreted and present in the conditioned medium, we first cultured HL60 cells in the presence of human recombinant FGF8f (rFGF8f). The result showed that rFGF8f significantly inhibited HL60 cell proliferation in the absence of RA (Fig. 3C). To examine that paracrine FGF8f specifically mimicked RA induction of granulocytic differentiation, we cultured HL60 and NB4 cells in the presence of RA or rFGF8f. The results showed that similar to RA stimuli, rFGF8f actually induces granulocytic morphology differentiation of HL60 and NB4 cells, an essential phenotype of terminal granulocytic differentiation (Fig. 3D, top; sections 3 vs. 2, respectively). Furthermore, using the antibody neutralization approach to define that paracrine FGF8f in OCM is a key factor in inducing granulocytic differentiation, HL60 cells were cultured with RA-OCM in parallel to different controls, including vector-OCM, RA stimuli, and blank. Distinctively, while FGF8 antibodies reversed the effect of RA-OCM containing FGF8f on inducing granulocytic morphologic differentiation of HL60 cells (Fig. 3D, bottom; sections 5 vs. 6), vector-OCM with FGF8 antibodies showed no effect on cellular status, similar to blank and vector-OCM groups (Fig. 3D, bottom; sections 4 vs. 1 and 3). Taken together, these data showed that paracrine FGF8f secreted by osteoblastic niche cells specifically cross-regulates granulocytic differentiation of different subtypes of leukemic myeloblasts.

Hematopoietic development and granulopoiesis have been well characterized (38, 39). Although other groups and we have determined that RA mediates granulocytic differentiation of both normal and malignant hematopoietic precursors (1, 6, 20–22, 40), it remains unknown whether signaling instigated by RA from the HSC niche cross-regulates granulopoiesis. Given that paracrine FGF8f bypasses RA stimuli to induce terminal granulocytic differentiation of leukemic myeloblasts (Figs. 2 and 3), and that many pathways associated with cancer normally regulate stem cell development (41), we investigated whether paracrine FGF8f has similar effect on mediating terminal granulocytic differentiation of normal primitive hematopoietic precursors. CD34⁺ cells were treated with RA or cultured with RA-OCM or S77A-OCM or FGF8f-OCM, in parallel to different controls. The results showed that both RA- and OCM-containing FGF8f inhibited proliferation of CD34⁺ cells (Fig. 4A and B). Because change in morphology is an essential phenotype for assessment of terminal granulocytic differentiation, we further analyzed granulocytic morphologic differentiation of CD34⁺ cells treated with RA or RA-OCM. The results showed that RA-OCM containing FGF8f mimics the effect of RA on inducing granulocytic differentiation (Fig. 4C). To verify that the observed response in CD34⁺ cells is not due to unknown cytokines cosecreted in the RA-OCM, we treated CD34⁺ cells with RA-OCM in the presence and absence of FGF8 antibodies. The results showed that, while RA-OCM–containing FGF8f mimics the RA effect on inducing terminal granulocytic morphology differentiation of CD34⁺ cells (Fig. 4D; sections 5 and 11 vs. 2 and 8), FGF8 antibodies spatiotemporally neutralized such effect of RA-OCM–containing FGF8f (Fig. 4D; sections 5 and 11 vs. 6 and 12). These data provide evidence that RA-mediated secretion of FGF8f by osteoblasts can bypass RA stimuli to inhibit proliferation and induce granulocytic differentiation of normal hematopoietic precursors.

The above-described data show that osteoblast-secretion of FGF8f mediated by RA signaling coordinates osteoblastic differentiation (Fig. 1) with granulocytic differentiation of hematopoietic precursors (Figs. 2–4). These findings prompted us to ask a question: Which pathway is possibly involved in coupling FGF8f effect at both the autocrine and paracrine levels? Because FGF signaling is largely transduced through MAPK pathway (42) and because, in parallel to FGF8f induction, RA upregulates expression of MAPK signaling molecules in osteoblasts (23), we investigated whether MAPK signaling lies downstream of RA-originated FGF8f regulation in those osteoblastic and hematopoietic precursors. Western blot analysis using specific antibodies against different phosphorylated MAP kinases showed that, in FGF8f-mediated osteoblastic differentiation of U2OS cells (Fig. 1), p38-MAPK phosphorylation was inhibited (Fig. 5A, top) but p44-ERK phosphorylation was upregulated in association with decreased p42-ERK phosphorylation (Fig. 5A, bottom), in contrast to their even levels of total proteins. Interestingly, a similar dynamic pattern was also detected in HL60 cells cultured with RA-OCM, as shown by upregulated p44-ERK phosphorylation (Fig. 5B, bottom), accompanied by downregulated phosphorylation of both p38-MAPK (Fig. 5B, top) and p42-ERK (Fig. 5B, bottom). Furthermore, in CD34⁺ cells cultured with RA-OCM, the phosphorylation of p42/44-ERK was periodically sustained (Fig. 5C, bottom) whereas the p38-MAPK phosphorylation was inhibited in general (Fig. 5C, top) during granulocytic differentiation induced by RA-OCM (Fig. 4). Altogether, these results support the notion that possibly through modulation of MAPK signaling at both the autocrine and paracrine levels, RA-induced FGF8f coordinates differentiation of osteoblastic and hematopoietic precursors.

The difficulty of integrating signals instigated by the genetic programs of hematopoietic precursors with the developmental cues originating from osteoblast-constituted HSC niche, where myeloid leukemia cells appear to arise, is a major barrier to effectively manipulate RA therapy for diverse types of myeloid leukemia. Our studies show that RA-mediated FGF8f induction by osteoblasts coordinates osteoblastic maturation (Fig. 1) with granulocytic differentiation of both leukemic myeloblasts and normal hematopoietic precursors (Figs. 2–4). Moreover, such paracrine FGF8f is capable of inhibiting proliferation of myeloid leukemic cells by overcoming RA-resistance (Figs. 2A and 3A), which results from mutation deletion of ligand-dependent AF-2 domain of RARα (36, 37). This effect of paracrine FGF8f on cross-regulating granulocytic differentiation of hematopoietic precursors is possibly through modulation of MAPK signaling (Fig. 5) via FGF8f-mediated interaction with the FGF receptor (FGF-R; Fig. 6). Therefore, these findings provide evidence that, in the absence of RA, the downstream regulatory target of RA signaling, HSC niche-derived FGF8f, is able to execute the RA-induced signaling effect on inducing granulocytic differentiation of both malignant and normal hematopoietic precursors (Figs. 2–4). Success in identifying which FGF-R on hematopoietic precursors interacts with paracrine FGF8f is crucial to determine the mechanisms of HSC niche-mediated granulopoiesis through modulation of MAPK signaling.

How does RA mediate osteoblast induction of FGF8f to coordinate the regulation of osteoblastic differentiation with granulocytic differentiation? The important role of RARα is identified with myeloid development (3) as well as its mutation-resistance to myeloid differentiation (36, 37). It has been shown that RA-activated RARα regulates myeloid differentiation by transcription of differentiation target genes (1, 3, 4, 7), and that RA-suppressed CAK phosphorylation of RARα on RARαS77 of the AF-1 domain (6, 15) induces differentiation of both osteosarcoma cells (23) and leukemic myeloblasts (20, 21). Previous studies have shown that the human FGF8 promoter contains RARα-binding sites, and that RA-mediated interaction of RARα with these sites induces CAT activity (43). In addition, we showed that RA-suppressed CAK phosphorylation of RARα remodels RARα–chromatin interaction to induce transcription of differentiation target genes (6), including FGF8f that is induced by RA stimuli or overexpression of RARαS77A (23). Moreover, whereas FGF-Rs are expressed on nearly every cell of hematopoietic origin, FGFs are basically produced by bone marrow stromal cells, and cells from some mature peripheral blood lineages (13, 44). These findings indicate that RA-mediated FGF8f induction by osteoblasts (Fig. 1B; ref. 23) cross-regulates granulocytic differentiation (Figs. 2–4), possibly through its interaction with specific FGF-R expressed ubiquitously on hematopoietic cells (13, 44) in the bone marrow (Fig. 6). Therefore, determining the mechanism of RA-mediated transactivation of FGF8f in the HSC niche as well as the interaction of FGF8f:FGF-R on hematopoietic precursors are crucial to establish the role of RA-instigated epigenetic regulation of granulocytic differentiation.

FGF signals are generally transduced through the MAPK pathways (42). Some FGFs can act at both the autocrine and paracrine levels to mediate the biologic processes between blood- and bone-forming cells by interacting with distinct FGF-Rs (44, 45). Data from embryonic stem cells show that rapid retinoid induction of FGF8 and downstream ERK activity ensures loss of self-renewal and initiates differentiation (46). Furthermore, recent studies indicate that RA activates RARα and induces granulocytic differentiation of leukemic myeloblasts through the MAPK pathway (47, 48). These findings indicate a mechanistic link of RA-mediated granulocytic differentiation among retinoid stimuli, RARα activation, FGF8 induction, and MAPK signaling. Our studies show that RA-decreased CAK phosphorylation of RARα induces FGF8f (23). The induced FGF8f (Fig. 1B), in turn, mediates osteoblastic differentiation in an autocrine manner (Fig. 1) and couples granulocytic differentiation of hematopoietic precursors at the paracrine level (Figs. 2–4). Such bidirectional regulation of FGF8f is associated with a dynamic change in ERK phosphorylation in both osteoblastic and hematopoietic precursors (Fig. 5). These findings indicate that RA coordinates CAK-RARα regulation with FGF8f-ERK signaling to couple osteoblastic maturation with granulocytic differentiation (Fig. 6). It is therefore important, in future studies, to delineate the involvement of ERK signaling mediated by FGF8f-dependent epigenetic regulation, using combined loss-of-function strategies including antibody neutralization, short-hairpin RNA silencing, and inhibition of ERK pathways. However, because FGF8f is also engaged in cross-talk with Wnt8B or WISP3 (23) through GSK3 or catenin signals (49) to activate the FGF–PI3K–AKT pathway in mediating granulocytic differentiation, the above-mentioned loss-of-function approaches might encounter difficulty in modulating granulocytic differentiation through blocking FGF8f-ERK pathway. If so, it is necessary to alternatively test whether paracrine FGF8f mediates granulocytic differentiation through cross-talk with Wnt signaling to activate the PI3K–AKT pathway, which has been implicated in the granulocytic differentiation of myeloid leukemia cells (50). Success in determining the mechanisms of this orchestration mediated by paracrine FGF8f should provide new insights into the niche-derived myeloid leukemogenesis.

In conclusion, our studies discover that FGF8f, a downstream targeting molecule induced by RA signaling in osteoblastic niche cells, is capable of cross-mediating granulocytic differentiation of both malignant and normal hematopoietic precursors in the absence of RA. These findings show that, without RA stimuli, modulation of the downstream targeting molecule of RA signaling not only can efficiently implement the effect of RA on inducing granulocytic differentiation of myeloid leukemic cells but also can bypass the upstream defect occurring at the levels of RA:RARα interaction in RA-resistant cells. These findings thereby present a compelling molecular rationale for developing effective differentiation therapies against different types of myeloid leukemia by targeting paracrine FGF8f and its potential cofactors.

No potential conflicts of interest were disclosed.

This work was supported by grants from the National Institutes of Health (R01 CA120512 and ARRA-R01CA 120512 to L. Wu).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.