Hoxb8 Requires Its Pbx-Interaction Motif to Block Di€Erentiation Of

HoxB8 requires its Pbx-interaction motif to block dierentiation of primary myeloid progenitors and of most cell line models of myeloid dierentiation

Paul S Knoep¯er2,3, David B Sykes1,3, Martina Pasillas1 and Mark P Kamps*,1

1Department of Pathology, University of California, San Diego, School of Medicine, 9500 Gilman Drive, La Jolla, California, CA 92093, USA; 2Department of Basic Sciences, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave N., Seattle, Washington, WA 98109, USA

HoxB8 was the ®rst homeobox gene identi®ed as a cause Introduction of leukemia. In murine WEHI3B acute myeloid leukemia (AML) cells, proviral integration leads to the expression Homeodomain (HD) proteins of the Hox, Pbx, and of both HoxB8 and Interleukin (IL-3). Enforced Meis families regulate gene expression in normal expression of HoxB8 blocks dierentiation of factor- development, and mutations that alter their expres- dependent myeloid progenitors, while IL-3 co-expression sion or function can lead to cancer. The mammalian induces autocrine proliferation and overt leukemogeni- Class I Hox genes are homologs of the Drosophila city. Previously, we demonstrated that HoxB8 binds homeobox genes located in the homeotic complex, DNA cooperatively with members of the Pbx family of and encode transcription factors that orchestrate transcription factors, and that HoxB8 makes contact anterior-posterior pattern formation (McGinnis and with the Pbx homeodomain through a hexameric Krumlauf, 1992), determine segmental identity during sequence designated the Pbx-interaction motif (PIM). embryogenesis (Krumlauf, 1994; Lewis, 1978) and E2a-Pbx1, an oncogenic derivative of Pbx1, both retains contribute to lineage-speci®c proliferation and/or its ability to heterodimerize with Hox proteins and arrest dierentiation of hematopoietic progenitors (Law- myeloid dierentiation. This observation prompts the rence et al., 1996). Class I Hox proteins can in¯uence question of whether E2a-Pbx1 and Hox oncoproteins use transcription through their binding to DNA as endogenous Hox and Pbx proteins, respectively, to monomers, and most are also capable of hetero- target a common set of cellular genes. Here, we use dimerizing on more complex sites with members of four dierent models of neutrophil and macrophage the three amino acid loop extension (TALE) family dierentiation to determine whether HoxB8 needs to of transcription factors, which include the products of bind DNA or Pbx cofactors in order to arrest myeloid the Pbx and Meis genes (Chang et al., 1996, 1997; dierentiation. The ability of HoxB8 to bind DNA or to Knoep¯er et al., 1996, 1997). Pbx1 is involved in the bind Pbx was essential (1) to block dierentiation of t(1;19) chromosomal translocation of pediatric pre-B factor-dependent myeloid progenitors from primary cell leukemia, which fuses the transcriptional activa- marrow; (2) to block IL-6-induced monocytic dierentiation domains of E2a to the HD of Pbx, producing a tion of M1-AML cells; and (3) to block granulocytic chimeric transcriptional activator that transforms dierentiation of GM-CSF-dependent ECoM-G cells. NIH3T3 ®broblasts (Kamps et al., 1991), prevents However, while DNA-binding was required, the HoxB8 dierentiation of myeloid progenitors (Kamps and Pbx-interaction motif was unnecessary for preventing Wright, 1994), and causes both AML (Kamps and macrophage dierentiation of ECoM-M cells. We Baltimore, 1993) and T-cell ALL in mice (Dedera et conclude that HoxB8 prevents dierentiation by directly al., 1993). Biochemical evidence (Lu and Kamps, in¯uencing cellular gene expression, and that the genetic 1996), later con®rmed by X-ray crystallography context within a cell dictates whether the eect of (Passner et al., 1999; Piper et al., 1999), revealed HoxB8 is dependent on a physical interaction with Pbx that a conserved tryptophan motif positioned N- proteins. Oncogene (2001) 20, 5440 ± 5448. terminal to the HD of Class I Hox proteins binds the Pbx1 HD. When the Hox tryptophan consensus Keywords: HoxB8; Pbx; myeloid; dierentiation motif (IYPWMR or ANWL; designated the Pbx- interaction motif or PIM) binds the Pbx HD, it alters the conformation of the Pbx HD so as to stabilize DNA-binding by the Pbx HD (Knoep¯er and Kamps, 1995; Lu and Kamps, 1996). Heterodimer- ization of Hox proteins with Pbx or E2a-Pbx1 on TGATTNAT (Pbx-Hox consensus motifs) also alters *Correspondence: MP Kamps; E-mail: [email protected] 3These authors contributed equally to this manuscript the conformation of the Hox HD, shifting the Received 1 March 2001; revised 1 June 2001; accepted 8 June 2001 binding speci®city for its 3' nucleotide core from The Pbx-interaction motif is required for myeloid immortalization by HoxB8 PS Knoepfler et al 5441 TAAT to TGAT or TTAT (Chang et al., 1996; sequences bind Pbx, and maintaining the possibility Knoep¯er et al., 1996; Lu and Kamps, 1997). that interaction with Pbx is essential for oncogenesis Subsets of Hox, Pbx,andMeis genes are expressed (Calvo et al., 2000). In contrast, mutations in the Pbx- during dierentiation of all hematopoietic lineages that interaction motif of Nup98-HoxA9 clearly abolish have been examined to date, raising the question of transcriptional cooperativity with Pbx and also prevent whether their function as oncogenes requires the the formation of foci on NIH3T3 monolayers (Kasper activity of endogenous heterodimer partners. One et al., 1999), providing direct evidence that Pbx oncogenic phenotype elicited by both Hox and E2a- proteins are important for transformation by Nup98- Pbx1 oncoproteins is blocking hematopoietic dieren- HoxA9. tiation: enforced expression of Hox11 (Hatano et al., HoxB8 was the ®rst Hox protein found to be 1991; Hawley et al., 1994), HoxA9 (Calvo et al., 2000), transcriptionally activated in AML (in WEHI3B mouse HoxA10 (Thorsteinsdottir et al., 1997), HoxB3 (Sau- myeloid leukemia cells). Persistent expression of vageau et al., 1997), HoxB4 (Sauvageau et al., 1995), HoxB8 prevents dierentiation of factor-dependent HoxB7 (Care et al., 1999), or HoxB8 (Blatt et al., 1988; myeloid progenitors (Blatt et al., 1988), and cooperates Perkins et al., 1990) prevents the dierentiation of a with mutations that establish autocrine proliferation to variety of myeloid and lymphoid progenitors. Among generate overt AML (Perkins et al., 1990). Here we these Hox proteins, the importance of interaction with examine whether the biochemical abilities of HoxB8 to Pbx and Meis partners for oncogenic function has been bind DNA and to bind Pbx proteins through its addressed for HoxB3, HoxB4, HoxB7, and HoxA9, as conserved Pbx-interaction motif are required for its well as Nup98-HoxA9, a transcriptionally activated cellular functions of blocking dierentiation of primary version of HoxA9 formed by the t(7;11) translocation marrow progenitors, as well as of three distinct cell line that fuses N-terminal sequences of Nucleoporin 98 models of neutrophil or macrophage dierentiation. (Nup98) to sequences just upstream of the HoxA9 Pbx- We demonstrate that DNA-binding is essential for all interaction motif and HD (Borrow et al., 1996; models of dierentiation arrest and that an intact Pbx- Nakamura et al., 1996). In the case of HoxB3 and interaction motif of HoxB8 is required to block HoxB4, the magnitude of their oncogenic function in myeloid dierentiation of all but one cell line. These Rat-1 ®broblast (number and size of colonies in soft data suggest that HoxB8 prevents myeloid dierentia- agar) is augmented by co-expression of Pbx1 (Krosl et tion in most cell types by a mechanism that requires al., 1998), providing indirect evidence that hetero- interactions with endogenous Pbx proteins on target dimerization with endogenous Pbx may be essential for gene promoters. their transforming properties. In the case of HoxB7, its ability to prevent G-CSF-induced granulocytic dierentiation of 32D cells requires its hexapeptide Results Pbx-interaction motif (Yaron et al., 2001). The question of whether transformation by HoxA9 The ability of HoxB8 to bind both Pbx and DNA is and NUP98-HoxA9 requires interaction with Pbx required to prevent differentiation of marrow-derived proteins has also been addressed by mutation of their myelomonocytic progenitors Pbx-interaction motifs. HoxA9 is required for normal myelopoiesis and lymphopoiesis as evidenced in HoxA9 A form of HoxB8 containing the conservative sub- knockout mice which are de®cient in their production stitution of phenylalanine for tryptophan in the Pbx- of mature granulocytes and lymphocytes (Lawrence et interaction motif (LFPWMR), designated al., 1997) and show both increased apoptosis and HoxB8W137F, was previously shown to be defective impaired T-cell dierentiation (Izon et al., 1998). in heterodimerization on DNA with Pbx proteins Enforced expression of HoxA9 blocks the dierentia- (Knoep¯er and Kamps, 1995). The ability of tion of myeloid progenitors capable of biphenotypic HoxB8W137F to bind DNA monomerically is equiva- dierentiation to either neutrophils in response to G- lent to that of wild-type HoxB8 (Figure 1, lane 5 vs 4) CSF or macrophages in response to M-CSF (Calvo et while its ability to heterodimerize with Pbx proteins al., 2000). Coactivation of HoxA9 and Meis1 tran- (Pbx1 or Pbx3) is reduced more than 20-fold (lanes 8 scription causes rapid AML in mice (Nakamura et al., and 11 vs 7 and 10). A form of HoxB8 predicted to be 1996; Kroon et al., 1998). HoxA9 is normally expressed incapable of binding DNA was produced by substitu- in CD34+ progenitors, downregulated during dier- tion of serine for asparagine at position 51 of the entiation, and expressed in all human AML (with the homeodomain (designated HoxB8N196S). Asp51 forms exception of promyelocytic leukemia), prompting an essential hydrogen bond with the third adenine speculation that persistent HoxA9 expression may residue in the core TAAT DNA element bound by Hox contribute to the dierentiation arrest in human proteins, while serine is incapable of forming this AML (Golub et al., 1999). Versions of HoxA9 that hydrogen bond. HoxB8N196S failed to bind DNA carry mutations in the Pbx-interaction motif are still (lane 6 vs 5) or to heterodimerize signi®cantly with capable of blocking myeloid dierentiation; however, Pbx1 (lane 9 vs 7) or Pbx3 (lane 12 vs 10). these same mutations do not preclude HoxA9 from The ability of both HoxB8W137F and HoxB8N196S cooperating with a transcriptionally activated versions to cooperate with Pbx proteins in transcription was of Pbx (VP16-Pbx1), suggesting that additional HoxA9 measured by cotransfection of expression vectors

Oncogene The Pbx-interaction motif is required for myeloid immortalization by HoxB8 PS Knoepfler et al 5442 cassette encoding neomycin phosphotransferase. The cells were selected in G418 and analysed by immuno- blotting using antibodies to HoxB8. Both HoxB8W137F and HoxB8N196S were expressed at levels at least as great as that of wild-type HoxB8 (Figure 2a, lanes 2 ± 4), suggesting that neither mutation signi®cantly altered overall protein stability. Equal titers of retrovirus were used to infect marrow progenitors cultured in medium containing GM-CSF. In each of eight cultures, expression of wild-type HoxB8 resulted in the outgrowth of factor-dependent populations of myeloid progenitors within 14 days (Figure 2b). Each immortalized cell population expressed HoxB8, as detected by immunoblot analysis of cells derived from day 20 of the assay (Figure 2c). These cell populations were clearly immortal, having now been cultured continuously for over 2 years. In contrast to cultures infected by wild-type HoxB8 virus, none of eight parallel cultures infected with retrovirus encoding either HoxB8W137F or HoxB8N196S be- came immortalized; rather, all eight proliferated and Figure 1 DNA-binding and heterodimerization by wild-type and mutant forms of HoxB8. The recombinant proteins indicated dierentiated identically to cultures infected with above each lane were synthesized by coupled transcription/ empty vector control retrovirus, ultimately dieren- translation, were mixed with radiolabeled probe containing the tiating into neutrophils and macrophages within 6 ± 15 TGATTTAT Pbx-Hox consensus sequence, and were resolved by days (Figure 2b). We conclude that HoxB8 requires its electrophoresis in 6% non-denaturing polyacrylamide genes. Mutant `HoxB8W137F' contains a phenylalanine substitution ability to bind DNA and interact with Pbx via its for the tryptophan residue in the Pbx-interaction motif, while tryptophan motif in order to prevent dierentiation of `HoxB8N196S' contains a serine substitution for the asparagine primary GM-CSF-dependent myelomonocytic progeni- residue in the Hox HD, disrupting a critical contact with the tors. adenosine residue at position 3 of the TAAT core element

The ability of HoxB8 to bind both Pbx and DNA is required to prevent IL-6-induced monocytic encoding E2a-Pbx1, HoxB8, and a luciferase reporter differentiation of M1-AML cells gene driven by six TGATTTAT Pbx-Hox heterodimer binding sites. The ability of E2a-Pbx1 to activate The ability of HoxB8W137F and HoxB8N196S to transcription from this vector is absolutely dependent prevent myeloid dierentiation was next tested in M1- on coexpressed Hox proteins, and has been shown AML cells, factor-independent myeloid progenitors previously to be activated cooperatively by E2a-Pbx1 that exhibit monocytic dierentiation following treat- plus HoxC8 or HoxA5 (Lu and Kamps, 1997). Unlike ment with IL-6. This system has the advantage of HoxC8 or HoxA5, coexpression of HoxB8 with E2a- permitting one to con®rm expression of each mutant Pbx1 in Nalm-6 pre-B cells failed to activate transcrip- oncoprotein in the progenitor populations prior to tion of this reporter (data not shown). HoxB8 also determining their impact on IL-6-induced dierentia- exhibited broad transcriptional inhibition, preventing tion. M1-AML cells transduced with retrovirus encod- both Hox-independent transcriptional activation by ing mutant and wild-type HoxB8 proteins were E2a-Pbx1 through TGATTGAT motifs and also pre- subjected to immunoblot analysis. As in NIH3T3 venting transcription activated by the Rous Sarcoma ®broblasts, the abundance of HoxB8W137F and Virus (RSV) LTR. Transcriptional inhibition was inde- HoxB8N196S in M1-AML cells was at least as great pendent of the ability of HoxB8 to bind DNA or to as that of wild-type HoxB8 (Figure 3a). In contrast to interact with the Pbx1 HD, as both HoxB8W137F and wild-type HoxB8, which prevented morphologic dier- HoxB8N196S inhibited transcription as actively as entiation of M1-AML cells in response to treatment wild-type HoxB8 (data not shown). Therefore, an in with IL-6 (Figure 4f ± h) and strongly suppressed the vivo experiment could not be performed to demonstrate growth inhibitory eect of IL-6 (Figure 3b), neither that disruption of HoxB8 binding to Pbx or to DNA HoxB8W137F nor HoxB8N196S inhibited morpholo- eliminates its ability to in¯uence transcription coopera- gic dierentiation (Figure 4i and j, respectively) or tively with Pbx proteins in vivo. suppressed growth inhibition by IL-6 (Figure 3b). We The ability of HoxB8W137F and HoxB8N196S to conclude that the ability of HoxB8 to bind both Pbx prevent dierentiation of GM-CSF-responsive marrow and DNA are essential for its ability to prevent IL-6- progenitors was compared with that of HoxB8. To induced dierentiation in the M1-AML cell clone. examine the stability of each HoxB8 mutant in vivo, The presence of endogenous Pbx proteins expressed NIH3T3 ®broblasts were infected with retrovirus in M1-AML cells was determined in order to support expressing each HoxB8 cDNA in tandem with a the hypothesis that HoxB8W137F fails to prevent

Oncogene The Pbx-interaction motif is required for myeloid immortalization by HoxB8 PS Knoepfler et al 5443

Figure 2 The ability of HoxB8 to bind Pbx and DNA are essential for the prevention of dierentiation of GM-CSF-dependent marrow progenitors. (a) Western blot analysis demonstrates that the HoxB8 mutants are as stable as wild-type HoxB8 in NIH3T3 ®broblasts. The mobility of HoxB8 is designated at left. (b) Growth kinetics of primary mouse marrow cells infected with retrovirus encoding no insert (triangles), wild-type HoxB8 (solid circles), HoxB8N196S (open squares), and HoxB8W137F (®lled squares). (c) Western blot analysis using anti-HoxB8 sera to identify HoxB8 proteins in eight immortalized cell populations independently derived by infection with HoxB8 retrovirus (lanes 2 ± 9). Control myeloblasts, immortalized by E2a-Pbx1, are represented in lane 1

cells, both Pbx1 and Pbx2 cofractionated with nuclei, and were extracted eciently by 400 mM NaCl (Figure 5b). Their fractionation pattern was not altered by expression of HoxB8 or HoxB8W137F. These data indicate that Pbx1 and Pbx2 reside in the nuclei of M1- AML cells, and that expression of HoxB8 proteins does not alter their nuclear location (Figure 5b). In contrast to wild-type HoxB8, which remained associated with the salt-insensitive nuclear fraction, ap- Figure 3 The ability of HoxB8 to bind Pbx and DNA are proximately 50% of both HoxB8W137F and essential for the prevention of IL-6-induced monocytic dierentia- HoxB8N196S were localized in the cytosolic fraction tion of M1-AML cells. (a) Anti-HoxB8 Western blot of M1-AML (Figure 5c). This result suggests that the nuclear cells transduced with control retrovirus (lane 1) and retrovirus anity of HoxB8 is stabilized by its ability to bind encoding wild-type HoxB8 (lane 2), HoxB8W137F (lane 3), and HoxB8N196S (lane 4). (b) Cells remaining after 7 and 8 days of both Pbx and DNA, implicating that the majority of culturing the populations assayed in a in the absence (®lled bars) HoxB8 is bound to DNA in the form of Pbx-HoxB8 or presence (cross-hatched bars) of 10 ng/ml IL-6. The identity of complexes. each population is indicated below the bar graph

The ability of HoxB8 to bind both Pbx and DNA is required to prevent granulocytic differentiation of dierentiation because it fails to interact with endo- ECoM-G cells, and to prevent transcription of genous Pbx proteins on essential target gene promo- granulocytic differentiation genes ters. Immunoblots using Pbx-speci®c antisera detected signi®cant expression of Pbx1 and Pbx2, but not of ECoM-G cells are GM-CSF-dependent neutrophil Pbx3 (Figure 5a). Cell fractionation was used to progenitors whose intrinsic dierentiation program is determine whether expression of HoxB8 altered the regulated by a conditional form of E2a-Pbx1 in an intracellular location of Pbx proteins and to determine estrogen-responsive manner (Sykes, 2001). The condi- whether the anity of HoxB8 for nuclei was stabilized tional form of E2a/Pbx1, designated EPD578ER, by interaction with Pbx. Cells were swollen in contains the estrogen-binding domain of the estrogen hypotonic buer, lysed by Dounce homogenization, receptor inserted between the E2a and Pbx1 domains and separated into cytosolic and nuclear fractions by of a derivative of E2a-Pbx1 containing an internal low speed centrifugation. The nuclei were then Pbx1 deletion (E2A-PBX1D487-578, Calvo et al., extracted with high salt buer (as described in 1999). In this system, EPD578ER is active in the Materials and methods), and nuclear extract was presence of estrogen, permitting myeloid dierentiation separated from the remaining nuclei by centrifugation. arrest, while removal of estrogen inactivates EP- While crude, this method yielded clean fractionation of D578ER, permitting growth arrest and normal granu- numerous background bands (Figure 5c) and of wild- locytic dierentiation. The conditional dierentiation type Pbx and HoxB8 proteins. In parental M1-AML of ECoM-G cells can be rearrested by heterologous

Oncogene The Pbx-interaction motif is required for myeloid immortalization by HoxB8 PS Knoepfler et al 5444

Figure 4 The ability of HoxB8 to bind Pbx and DNA are essential for the prevention of morphologic dierentiation of M1-AML cells to monocytes and morphologic dierentiation of ECoM-G cells to neutrophils. Wright-Giemsa analysis of ECoM-G cells (a ± e) and M1-AML cells (f ± j) expressing empty vector (a, b, f, g), wild-type HoxB8 (c, h), HoxB8W137F (d, i), or HoxB8N196S (e, j). ECoM-G cells were cultured in the presence of estrogen (1 mM b-estradiol) (a) or in the absence of estrogen to permit granulocytic dierentiation (b ± e). M1-AML cells were cultured in the absence of IL-6 (f) or in 10 ng/ml IL-6 to induce monocytic dierentiation (g ± j)

human myeloid leukemia oncoproteins (HoxA9, Aml1- which occurred within 5 days of withdrawing estrogen Eto, Pml-RAR) by mechanisms characteristic of the from ECoM-G cells during their dierentiation to secondary oncoprotein (e.g. parental ECoM-G cells are neutrophils (Figure 6c). Furthermore, when the level unresponsive to retinoic acid while those cells whose of gene expression was reduced in cells expressing dierentiation is re-arrested by Pml-RAR exhibit HoxB8W137F, it was reduced at least as much in cells granulocytic dierentiation in response to retinoic expressing HoxB8N196S, which serve as a control to acid). Previously, we reported that dierentiation of measure the non-speci®c, DNA-binding-independent ECoM-G (previously referred to as CM1 puro) cells eect of HoxB8 on gene transcription. Therefore, the can be rearrested in the absence of estrogen by HoxA9 DNA-binding-dependent block in transcription by (Calvo et al., 2000). The ability of wild-type HoxB8 HoxB8, associated with preventing granulocytic dier- and HoxB8 mutant proteins to prevent dierentiation entiation of ECoM-G cells, was dependent upon an of ECoM-G cells was assayed. HoxB8 retrovirus was intact Pbx-interaction motif. The same result was used to infect ECoM-G cells, and clones expressing the obtained by analysis of c-Myb and c-Myc transcription, same amount of HoxB8 proteins were selected (Figure which is normally suppressed during granulocytic 6a). Expression of wild-type HoxB8 prevented both the dierentiation. Transcription of both genes was reduced morphologic dierentiation of ECoM-G cells to at least ®vefold in parental cells though remained neutrophils (Figure 4a ± c), as well as their exit from unchanged in cells expressing wild-type HoxB8. Cells the cell cycle (Figure 6b). In contrast, neither expressing either HoxB8W137F or HoxB8N196S were HoxB8W137F nor HoxB8N196S prevented morpholo- similar in that they did not exhibit the same magnitude gic dierentiation (Figure 4d,e) or exit from the cell of transcriptional suppression as that found in parental cycle (Figure 6b). cells. Again, this result indicated that the DNA-binding- The ability of HoxB8, compared with HoxB8W137F dependent ability of HoxB8 to prevent transcriptional and HoxB8N196S, to prevent dierentiation-speci®c repression of c-Myb and c-Myc was dependent upon the gene expression was next examined. This assay integrity of the Pbx-interaction motif. represents a transcriptional cognate of morphologic dierentiation arrest by HoxB8 in which downstream The ability of HoxB8 to bind DNA, but not to bind Pbx, myeloid dierentiation genes that fail to be expressed by is required to prevent monocytic differentiation of a Pbx-dependent mechanism represent the genes that ECoM-M myeloblasts are responding to the oncogenic property of HoxB8. Such genes may be either direct target genes repressed ECoM-M cells are GM-CSF-dependent macrophage by HoxB8 or indirect target genes that fail to become progenitors whose intrinsic dierentiation program is activated because Hoxb8 alters transcription of genes regulated by a second form of E2a-Pbx1, designated controlling their transcription. Wild-type HoxB8, but EPD623ER. This estrogen-responsive form of E2a- not HoxB8W137F or HoxB8N196S, prevented the Pbx1 contains the estrogen-binding domain of the transcriptional upregulation of genes encoding lacto- estrogen receptor inserted between E2a and Pbx1 ferrin, the gp91phox subunit of NADPH oxidase, GR-1 domains found in another internal Pbx1 deletion (Ly6G), neutrophil gelatinase B, c-Fos, and Egr-1, mutant (E2A-PBX1D487-623, Calvo et al., 1999). In

Oncogene The Pbx-interaction motif is required for myeloid immortalization by HoxB8 PS Knoepfler et al 5445

Figure 6 The ability of HoxB8 to interact with Pbx and DNA is essential for the prevention of changes in gene expression associated with granulocytic dierentiation. (a) Immunoblot analysis of wild-type and mutant HoxB8 proteins expressed in ECoM-G cells. The identity of each HoxB8 protein is indicated above each lane. `B8 Marrow' denotes primary marrow-derived myeloid progenitors immortalized by wild-type HoxB8. (b) Proliferation of ECoM-G cells cultured in the absence of estrogen and expressing wild-type HoxB8 (open squares), HoxB8W137F (open triangles), HoxB8N196S (®lled squares) and no HoxB8 protein (®lled triangles). (c) Northern blot analysis of gene expression in ECoM-G cells cultured in the presence (+) or absence (7) of estrogen and expressing no HoxB8 (parent), wild- type HoxB8, HoxB8W137F, or HoxB8N196S

expressing HoxB8N196S, dierentiated normally into macrophages in the absence of estrogen, those Figure 5 Expression of HoxB8 does not change the location of expressing both HoxB8 and HoxB8W137F remained endogenous Pbx1 or Pbx2, while the ability of HoxB8 to bind Pbx fully blocked in dierentiation and continued to divide and DNA are essential to maintain its quantitative nuclear association. (a) Immunoblot analysis of endogenous Pbx proteins exponentially (data not shown). Therefore, the ability in parental M1-AML cells (lanes 4, 8 and 12) or M1-AML cells of HoxB8 to prevent dierentiation of ECoM-M cells, expressing HoxB8 (lanes 5, 9 and 13), using antisera speci®c for while dependent on DNA-binding, is independent of its Pbx1 (lanes 4 and 5), Pbx2 (lanes 8 and 9), and Pbx3 (lanes 12 ability to interact tightly with Pbx proteins. and 13). Recombinant Pbx1a (lane 1), Pbx1b (lane 2), Pbx2 (lane 6), and Pbx3 (lane 10) were produced by coupled transcription/ translation. (b) Immunoblot analysis of Pbx1 and Pbx2 in subcellular fractions of parental M1-AML cells and M1-AML Discussion cells expressing each HoxB8 mutant. M1-AML cells expressing no HoxB8 protein (lanes 1 ± 3 and 10 ± 12), expressing wild-type Our data demonstrate that in most genetic contexts HoxB8 (lanes 4 ± 6 and 13 ± 15), or expressing HoxB8W137F (lanes 7 ± 9 and 16 ± 18), were subjected to subcellular fractiona- HoxB8 requires both its ability to bind Pbx cofactors tion. `C' denotes cytosol, `E' denotes nuclear extract, and `N' and DNA in order to prevent morphologic dierentia- denotes proteins remaining in the extracted nuclei. Samples tion and the accompanying cell cycle downregulation. derived from the same number of cells were loaded in each lane. These properties support a model in which the ability of The major proteins migrating above and below the Pbx2 band serve as internal controls for the eciency of fractionation, the HoxB8 to heterodimerizes with endogenous Pbx upper band cofractionated with cytosol, and the lower band with proteins on DNA alters gene transcription in a manner extracted nuclei. (c) Immunoblot analysis of subcellular fractiona- that prevents progression through an intrinsic genetic tion of HoxB8 proteins in M1-AML cells. Cellular fractions are dierentiation program. This interpretation is consistent denoted as described above, and samples derived from the same with the facts that Pbx genes are expressed in these number of cells were loaded in each lane. Nonspeci®c bands representing proteins that reside in the cytoplasm (cyto), nuclear progenitors (Pbx1 and Pbx2 in M1-AML, Pbx2 in extract (extract), and post-extract nuclei (nuclear) are indicated at ECoM-G). It was formally possible that expression of left, and serve as controls for the eciency of fractionation HoxB8 blocks myeloid dierentiation by altering the location of endogenous Pbx cofactors, especially con- sidering the fact that changing the location of Exd (the ECoM-M cells, EPD623ER prevents dierentiation in Drosophila ortholog of Pbx) from nuclear to cytoplasmic the presence of estrogen, and removal of estrogen induces a profound developmental defect, causing results in growth arrest and monocytic dierentiation antenna progenitors to develop into leg; however, our (Sykes and Kamps, 2001). Analysis of the abilities of data do not support this type of mechanism. HoxB8, HoxB8W137F, and HoxB8N196S to prevent The DNA-binding and Pbx-dependent mechanism dierentiation of ECoM-M cells yielded a surprising by which overexpressed HoxB8 alters transcription of result: while cells infected with vector alone, or cells relevant myeloid target genes in vivo is not yet clear. In

Oncogene The Pbx-interaction motif is required for myeloid immortalization by HoxB8 PS Knoepfler et al 5446 conjunction with Pbx, HoxB8 could alter transcription proteins that is further stabilized on target promoters of Pbx target genes by direct or indirect mechanisms. by other factors unique to ECoM-M cells. Such a By direct mechanisms, HoxB8 could displace endogen- possibility is supported by the fact that mutations in ous Hox proteins in Pbx-Hox heterodimers formed on the Pbx-interaction motif of HoxA9 disrupt hetero- TGATTTAT-like elements (similar to those that dimerization with Pbx proteins on TGATTTAT in regulate transactivation of the HoxB2 promoter EMSA, but do not prevent its ability to activate (Jacobs et al., 1999), the HoxB1 promoter (Di Rocco transcription cooperatively with VP16-Pbx1 through et al., 1997; Popperl et al., 1995), or that activate TGATTTAT motifs in vivo (Calvo et al., 2000), transcription of the EphA2 gene (Chen and Ruley; suggesting that other surfaces of HoxA9 contact Pbx. 1998)) or on TAAT-like Hox monomer sites adjacent Therefore, we can not eliminate the possibility that to Pbx-Meis/Prep heterodimer elements, as occurs in essential target genes in ECoM-M cells are also the somatostatin enhancer (Goudet et al., 1999; Peers regulated by Pbx-HoxB8W137F complexes. In this et al., 1995). This would introduce a uniform regard, the ability of HoxB8W137F to prevent transcriptional function at Pbx-Hox elements by dierentiation of a GM-CSF-dependent monocytic disrupting selective targeting of endogenous Hox progenitor may be somewhat similar to our former proteins. In the context of HoxB8, this uniform observation that HoxA9 proteins containing the function could represent the strong activity in HoxB8 homologous WF mutation retain a potent and selective that prevents transcription in a DNA-binding indepen- ability to immortalize a progenitor exhibiting mono- dent manner. Through an indirect mechanism, over- cytic rather than one capable of both granulocytic and expressed HoxB8 could also sequester Pbx away from monocytic dierentiation. Lastly, dierentiation in other regulatory elements (Pbx-Meis/Prep or Pbx-Hox ECoM-M could be modulated by a subset of genes motifs) to unoccupied Pbx-Hox elements. This mechan- responsive to HoxB8 in a fully Pbx-independent ism would disrupt the ability of Pbx complexes to manner. In this light it is important to note that promote or repress transcription of normal genetic ECoM-M cells dierentiate into macrophages under targets even though those genes are not intrinsically the in¯uence of GM-CSF, a process distinct from controlled by HoxB8. Identifying such transcriptional either the IL-6-induced macrophage dierentiation mechanisms must await the identi®cation of target exhibited by M1-AML cells, or the granulocytic genes whose transcription is directly controlled by dierentiation exhibited by ECoM-G cells. The fact overexpressed HoxB8. This type of non-speci®c that the importance of functional motifs in Hox disruption of developmental fate by inappropriate proteins varies with respect to the cell system used Hox protein expression, which is also observed for for analysis indicates that prudence should be taken HoxB7 in terms of its Pbx-dependent mechanism of not to over-interpret results based on a single model of preventing G-CSF-induced neutrophil dierentiation of cellular dierentiation. 32D cells (Yaron et al., 2001) is similar to the How can the genetic basis of myeloid dierentiation observation in Drosophila that expression of Abd-A arrest by oncogenic Hox proteins be delineated, and Abd-B, which play speci®c roles in de®ning especially in light of the fact that Hox proteins may thoracic identity, transforms wing into haltere without alter transcription of a large fraction of cellular genes, producing any qualities of the thorax. many of which neither directly nor indirectly prevent The observations that Pbx and Hox proteins interact dierentiation? We propose that the most reasonable physically, and that Pbx1 can be converted into an method to identify relevant target genes is not simply oncoprotein, suggests that E2a-Pbx1 may also use to identify all direct targets, but rather to focus on endogenous Hox proteins as cofactors as part of a those targets that are responsible for preventing genetic mechanism that prevents myeloid dierentia- upregulation of normal downstream myeloid dier- tion. In this regard, oncogenesis by Pbx and Hox entiation genes, such as those encoding lactoferrin, oncoproteins may be similar to that of mutant forms of gp91phox, and GR-1. In this light, ECoM-G cells can Fos and Jun or of AML1 and CBFb, examples of be used to identify the transcriptional mechanism transcription factors that interact as heterodimers and responsible for normal upregulation of dierentiation are likely to eect oncogenesis by targeting the same or genes, and subsequently, the mechanism by which overlapping sets of cellular genes (Downing 1999; Vogt HoxB8 prevents upregulation of such genes can be and Bos, 1990). identi®ed. Such a mechanism can then be compared Why isn't the Pbx-interaction motif of HoxB8 with that executed by other Hox oncoproteins and by essential for blocking progression through dierentia- E2a-Pbx1 to distinguish between common or distinct tion in all models of myeloid dierentiation? Three pathways of dierentiation arrest. possible explanations may account for the fact that in some progenitors, the Pbx-interaction motif of HoxB8 is dispensable for preventing myeloid dierentiation. First, these cells might express other transcription Materials and methods factors that stabilize HoxB8 in a Pbx-independent manner on the same critical enhancers targeted by the Construction of recombinant plasmids and retroviral constructs Pbx-HoxB8 complexes. Secondly, HoxB8W137F may cDNA's encoding wild-type or mutant versions of murine retain a weak but signi®cant interaction with Pbx HoxB8 were subcloned into the polylinker of the proviral

Oncogene The Pbx-interaction motif is required for myeloid immortalization by HoxB8 PS Knoepfler et al 5447 vector, MSCV2.1. Helper-free virus for the infection of separation through 17% polyacrylamide gels run in 0.56TBE primary murine marrow was produced by cotransfection of (27 mM Tris; 27 mM boric acid; 0.6 mM EDTA). Coupled 293T cells with MSCV vectors and a packaging de®cient transcription/translation was performed in vitro using the MLV provirus (Kamps and Wright, 1994). cDNAs encoding Promega TNT Coupled Reticulocyte Lysate System in all proteins used in EMSA were cloned into pGEM3zf- accordance with manufacturer's protocol using SP6 polymer- (Promega) under the SP6 promoter. ase. For EMSA, 40 000 c.p.m. of probe was incubated with 3±6 mlofin vitro translated proteins in the presence of 1 mg of poly (dIdC) in a buer containing 10 m Tris (pH 7.5), Marrow infections M 1mM DTT, 0.1% NP40, and 5% glycerol for 30 min at room Primary murine progenitors were isolated from the femurs temperature. Complexes were separated by electrophoresis in and tibias of female Balb/c mice and separated from mature 6% polyacrylamide gels and run in 0.56TBE (27 mM Tris; monocytic and lymphocytic cells on a Ficoll-Paque gradient 27 mM boric acid; 0.6 mM EDTA) for 2 h at 160 V. Protein- (Pharmacia). These progenitors were infected with helper-free DNA complexes were visualized by autoradiography. Equal retrovirus and cultured in RPMI containing 10% fetal bovine molar amounts of wild-type and mutant HoxB8 proteins were sera and GM-CSF (Kamps and Wright, 1994). Brie¯y, 16106 added as assayed by parallel transcription/translation reac- progenitors were transferred to each well of a 6-well tissue tions performed in the presence of 35S-methionine followed by culture plate, incubated for 1 h with 1 ml of retrovirus quantitation of proteins resolved by SDS-gel electrophoresis. containing 56105 G418-resistance units, and 1 ml of Marrow Culture Medium (MCM; 500 ml RPMI1640, 10% FBS, Cooperative transactivation assays 16antibiotics (Penicillin, Streptomycin), 16glutamine, 16 units/ml GM-CSF) was added. Polybrene was added to a Nalm-6 cells were grown to log phase in RPMI, 10% fetal ®nal 8 mg/ml. After 1 h, 6 ml of MC media was added to bovine serum, 1% penicillin/streptomycin and electroporated dilute the polybrene to 2 mg/ml. After 3 days, 4 ml of in 200 ml at a concentration of 76106 cells with 6 mg of each medium was removed and replaced with 4 ml of fresh MCM. appropriate expression plasmid (e.g. HoxB8+mutants, E2A- Non-adherent cells were transferred every 7 days to new Pbx1) and Luciferase and Renilla reporter plasmids. The plates. Luciferase reporter plasmid contains 66binding sites (TGATTTAT) upstream of a minimal Fos promoter 5' of the luciferase gene in the pGL3 basic Luciferase vector Antiserum and immunoblots (Promega). Cells were harvested 60 h post electroporation Antibodies to HoxB8 were purchased from Santa Cruz and the luciferase assay was performed using the Dual- Biotechnology. Proteins in primary myeloblast populations Luciferase Reporter Assay System (Promega) according to were dissolved in Laemmli sample buer and boiled for manufacturer's speci®cations. Luciferase production was 5 min. Protein from 2.56105 cells was resolved by electro- normalized to internal Renilla controls. phoresis through 12.5% SDS polyacrylamide gels, transferred to a PVDF-Plus membrane, and detected immunologically, M1-AML differentiation assays using the chemiluminescent Phototope- HRP Detection Kit (New England Biolabs). M1-AML cells were cultured in RPMI 1640, 10% FBS, 1% Pen/Strep. M1-AML were infected with MSCV retrovirus lacking HoxB8 cDNA sequences or encoding wild-type or Northern blots mutant HoxB8 proteins, cultured for 48 h, and selected for Total cellular RNA was puri®ed from 56108 myeloid retroviral expression in 1 mg/ml G418. Populations of G418- progenitors. Fifteen micrograms of total cellular RNA was resistant cells were split for immunoblot analysis with resolved by electrophoresis through 1% denaturing alkaline- antibodies to HoxB8 and for macrophage dierentiation in agarose gels. RNA was transferred to GeneScreen Plus response to IL-6. IL-6 was added to a ®nal concentration of (NEN) membranes and subjected to hybridization with 32P- 100 ng/ml in RPMI 1640, 10% FBS, 1% Pen/Strep, and the labeled DNA probes using Ultrahyb (Ambion) following the phenotype of cells was analysed after 7 and 8 days. manufacturer's protocols. Differentiation assays for ECoM-G and ECoM-M Mutagenesis ECoM-G and ECoM-M were produced by the transduction of Site directed mutations in the Pbx-interaction motif and in hematopoietic progenitors with MSCV-based retrovirus en- the HD were created using the QuikChangeTM Site-Directed coding conditional versions of E2a-Pbx1 created by fusion of Mutagenesis Kit (Stratagene) according to the manufacturers E2a-Pbx1 with the estrogen-binding domain of the estrogen protocol. DNA sequencing to verify all mutations was receptor (ER). The complete characterization of these cell lines performed by the Molecular Pathology Shared Resource, as a tool for studying gene expression in myeloid dierentiation UCSD Cancer Center, which is funded in part by NCI as well as for studying how oncoproteins interfere with such Cancer Center Support Grant #5P0CA23100-16. genetic programs is currently underway (Sykes and Kamps, 2001). ECoM-G and ECoM-M were cultured in RPMI 1640, 10% FBS, 1% Pen/Strep/Gluatmine, and 1 m b-estradiol, Electrophoretic mobility shift assays M which is required to maintain the conditional E2a-Pbx1-ER Oligonucleotides containing a consensus Pbx-Hox binding proteins in their active state. Removal of b-estradiol from the site (tcacggTGATTTATgagcgactgctcgg) were synthesized culture medium inactivates E2a-Pbx1-ER, and permits normal (Genosys Biotechnologies, Inc., The Woodlands, TX, USA) dierentiation of cells to neutrophils (ECoM-G) or macro- and labeled with a-32P-ATP by phosphorylation of a reverse phages (ECoM-M), within 6 and 9 days, respectively. Both cell oligonucleotide that was annealed to the 3' portion of the lines are GM-CSF-dependent. The estrogen-binding domain single-stranded oligonucleotide and ®lled in using dNTPs and harbors the G400V mutation which makes it unresponsive to Klenow polymerase. The radioactive probe was isolated by the trace amounts of b-estradiol in the fetal bovine serum or to

Oncogene The Pbx-interaction motif is required for myeloid immortalization by HoxB8 PS Knoepfler et al 5448 the estrogenic eects of compounds such as phenol red (Tora et each well and the cells were cultured for 24 h prior to drug al., 1989). selection. Drug selection was carried out for 3 ± 4 days in Wild-type and mutant forms of HoxB8 were cloned into 1 mg/ml G418 or 1 mg/ml puromycin, respectively. Clones of the MSCV vector containing the neomycin or puromycin ECoM-G expressing equal levels of wild-type or mutant selection genes. Retrovirus derived from these vectors were HoxB8 proteins were selected for analysis of dierentiation introduced into ECoM-G and ECoM-M by spinoculation. properties and of RNA transcripts. Brie¯y, non-tissue culture treated 12-well plates were incubated with 10 mg/ml human ®bronectin (Sigma) in PBS for 1 h at 378C. The ®bronectin was aspirated and cells (250 000 cells in 250 ml of MCM) and viral supernatant Acknowledgment (1 ml) were centrifuged at 2500 g for 1.5 h. Following This work was supported by Public Health Services grant spinoculation, an additional 3 ml of MCM was added to CA56876.

References

Blatt C, Aberdam D, Schwartz R and Sachs L. (1988). Knoep¯er PS and Kamps MP. (1995). Mol. Cell. Biol., 15, EMBO J., 7, 4283 ± 4290. 5811 ± 5819. Borrow J, Shearman AM, Stanton VP, Becher R, Collins T, Knoep¯er PS, Lu Q and Kamps MP. (1996). Nucleic Acids Williams AJ, Dube I, Katz F, Kwong YL, Morris C, Res., 24, 2288 ± 2294. Ohyashiki K, Toyama K, Rowley J and Housman DE. Kroon E, Krosl J, Thorsteinsdottir U, Baban S, Buchberg (1996). Nat. Genet., 12, 159 ± 167. AM and Sauvageau G. (1998). EMBO J., 17, 3714 ± 3725. Calvo KR, Knoep¯er P, McGrath S and Kamps MP. (1999). KroslJ,BabanS,KroslG,RozenfeldS,LargmanCand Oncogene, 18, 8033 ± 8043. Sauvageau G. (1998). Oncogene, 16, 3403 ± 3412. Calvo KR, Sykes D, Pasillas M and Kamps MP. (2000). Mol. Krumlauf R. (1994). Cell, 78, 191 ± 201. Cell. Biol., 20, 3274 ± 3285. Lawrence HJ, Helgason CD, Sauvageau G, Fong S, Izon DJ, Care A, Valtieri M, Mattia G, Meccia E, Masella B, Luchetti Humphries RK and Largman C. (1997). Blood, 89, 1922 ± L, Felicetti F, Colombo MP and Peschle C. (1999). 1930. Oncogene, 18, 1993 ± 2001. Lawrence HF, Sauvageau G, Humphries RK and Largman Chang CP, Brocchieri L, Shen WF, Largman C and Cleary C. (1996). Stem Cells, 14, 281 ± 291. ML. (1996). Mol. Cell. Biol., 16, 1734 ± 1745. Lewis EB. (1978). Nature, 276, 565 ± 570. ChangCP,JacobsY,NakamuraT,JenkinsNA,CopelandNG Lu Q and Kamps MP. (1996). Mol. Cell. Biol., 16, 1632 ± and Cleary ML. (1997). Mol. Cell. Biol., 17, 5679 ± 5687. 1640. Chen J and Ruley HE. (1998). J. Biol. Chem., 273, 24670 ± Lu Q and Kamps MP. (1997). Oncogene, 14, 75 ± 83. 24675. McGinnis W and Krumlauf R. (1992). Cell, 68, 283 ± 302. Dedera DA, Waller EK, Lebrun DP, Sen-Majumdar A, Nakamura T, Largaespada DA, Lee MP, Johnson LA, Stevens ME, Barsh GS and Cleary ML. (1993). Cell, 74, Ohyashiki K, Toyama K, Chen SJ, Willman CL, Chen IM, 833 ± 843. Feinberg AP, Jenkins NA, Copeland NG and Shaugh- Di Rocco G, Mavilio F and Zappavigna V. (1997). EMBO J., nessy JD. (1996). Nat. Genet., 12, 154 ± 158. 16, 3644 ± 3654. Nakamura T, Largaespada DA, Shaughnessy JD, Jenkins Downing JR. (1999). Br. J. Haematol., 106, 296 ± 308. NA and Copeland NG. (1996). Nat. Genet., 12, 149 ± 153. Golub TT, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Passner JM, Ryoo HD, Shen L, Mann RS and Aggarwal AK. Mesirov JP, Coller H, Loh ML, Downing JR, Caligiuri (1999). Nature, 397, 714 ± 719. MA, Bloom®eld CD and Lander ES. (1999). Science, 286, Peers B, Sharma S, Johnson T, Kamps MP and Montminy 531 ± 537. M. (1995). Mol. Cell. Biol., 15, 7091 ± 7097. Goudet G, Delhalle S, Biemar F, Martial JA and Peers B. Perkins A, Kongsuwan K, Visvader J, Adams JM and Cory (1999). J. Biol. Chem., 274, 4067 ± 4073. S. (1990). Proc. Natl. Acad. Sci. USA, 87, 8398 ± 8402. Hatano M, Roberts CW, Minden M, Crist WM and Piper DE, Batchelor AH, Chang C, Cleary ML and Korsmeyer SJ. (1991). Science, 253, 79 ± 82. Wolberger C. (1999). Cell, 96, 587 ± 597. Hawley RG, Fong AZ, Lu M and Hawley TS. (1994). Popperl H, Bienz M, Studer M, Chan SK, Aparicio S, Oncogene, 9, 1±12. Brenner S, Mann RS and Krumlauf R. (1995). Cell, 81, Izon DJ, Rozenfeld S, Fong ST, Komuves L, Largman C and 1031 ± 1042. Lawrence HJ. (1998). Blood, 92, 383 ± 393. Sauvageau G, Thorsteinsdottir U, Eaves CJ, Lawrence HF, Jacobs Y, Schnabel CA and Cleary ML. (1999). Mol. Cell. Largman C, Lansdorp PM and Humphries RK. (1995). Biol., 19, 5134 ± 5142. Genes Dev., 9, 1753 ± 1765. Kamps MP and Baltimore D. (1993). Mol. Cell. Biol., 13, Sauvageau G, Thorsteinsdottir U, Hough MR, Hugo P, 351 ± 357. Lawrence HJ, Largman C and Humphries RK. (1997). Kamps MP, Look T and Baltimore D. (1991). Genes Dev., 5, Immunity, 6, 13 ± 22. 358 ± 368. Sykes DB and Kamps MK. (2001). Blood, in press. Kamps MP and Wright DD. (1994). Oncogene, 9, 3159 ± Tora L, Mullick A, Metzger D, Ponglikitmongkol M, Park I 3166. and Chambon P. (1989). EMBO J., 8, 1981 ± 1986. Kasper LH, Brindle PK, Schnabel CA, Pritchard CE, Cleary Thorsteinsdottir U, Sauvageau G, Hough MR, Dragowska ML and Van Deursen JM. (1999). Mol. Cell. Biol., 19, W, Lansdorp PM, Lawrence HJ, Largmen C and 764 ± 776. Humphries RK. (1997). Mol. Cell. Biol., 17, 495 ± 505. Knoep¯er PS, Calvo KR, Chen H, Antonarakis SE and Vogt PK and Bos TJ. (1990). Adv. Cancer Res., 55, 1±35. Kamps MP. (1997). Proc. Natl. Acad. Sci. USA, 94, Yaron Y, McAdara JK, Lynch M, Hughes E and Gasson JC. 14553 ± 14558. (2001). J. Immunol., 166, 5058 ± 5067.

Oncogene