Insulin-Like Growth Factor 1 Is a Direct HOXA9 Target Important for Hematopoietic Transformation

Home , Cdx, FRAT2, HOXA6, HOXA9, ParaHox

ORIGINAL ARTICLE Insulin-like growth factor 1 is a direct HOXA9 target important for hematopoietic transformation

J Steger, E Füller, M-P Garcia-Cuellar, K Hetzner and RK Slany

HOX homeobox proteins are key oncogenic drivers in hematopoietic malignancies. Here we demonstrate that HOXA1, HOXA6 and predominantly HOXA9 are able to induce the production of insulin-like growth factor 1 (Igf1). In chromatin immunoprecipitations, HOXA9 bound directly to the putative promoter and a DNase-hypersensitive region in the ﬁrst intron of the Igf1 gene. Transcription rates of the Igf1 gene paralleled HOXA9 activity. Primary cells transformed by HOXA9 expressed functional Igf1 receptors and activated the protein kinase Akt in response to Igf1 stimulation, suggesting the existence of an autocrine signaling loop. Genomic deletion of the Igf1 gene by Cre-mediated recombination increased sensitivity toward apoptosis after serum starvation. In addition, the leukemogenic potential of Igf1-negative, HOXA9-transformed cells was impaired, leading to a signiﬁcant delay in disease development on transplantation into recipient animals.

Leukemia (2015) 29, 901–908; doi:10.1038/leu.2014.287

INTRODUCTION factor FGF2 (fibroblast growth factor) are regulated by HOX 14,15 HOX homeobox genes are increasingly acknowledged as impor- proteins. Apart from these examples and despite the tant drivers of malignant transformation in the hematopoietic discovery of numerous HOXA9-binding sites across the genome 16 system. In line with their major regulatory role in hematopoietic by chromatin immunoprecipitation-sequencing ChIP-Seq, func- stem and precursor cell populations, HOX transcription needs to tionally characterized HOX targets are scarce. be extinguished for proper differentiation to occur (for detailed Here we identify Igf1 (insulin-like growth factor 1) as direct reviews of this topic see Alharbi et al.,1 Argiropoulos and HOXA9 subordinate gene and we show that Igf1 signaling Humphries2 and Eklund3). Consequently, inappropriate HOX contributes functionally to HOXA9-induced leukemia. The IGF activation has been detected in many hematologic malignancies. family has been widely implicated in the etiology of neoplastic High-level expression mainly of HOXA and HOXB family members disease. Insulin, IGF1 and IGF2 bind to a shared group of insulin/ is a hallmark of different leukemia subtypes like those harboring IGF receptor tyrosine kinases leading to the activation of PI3K MLL (mixed-lineage leukemia) or CALM (clathrin assembly (phosphatidylinositol-4,5-bisphosphate 3-kinase) and subse- lymphoid myeloid leukemia) fusion proteins, in cases with quently AKT (v-akt murine thymoma viral oncogene homolog). elevated CDX (caudal homeobox) expression or as a consequence In non-gluconeogenic tissues, this triggers multiple downstream 4–8 signaling cascades with preferential impact on proliferation and of a mutation in NPM1 (nucleophosmin). Besides, chromosomal 17 loci of the clustered HOX genes are themselves targets of survival (reviewed in Pollak ). In addition, the insulin/IGF system is chromosomal translocations9 and HOX expression can also be intricately involved in general aspects of physiology that show an association with the incidence of malignancies, like body mass endogenously upregulated, particularly in acute myeloid leuke- index and birth weight. Multiple clinical trials investigating mia. Completing the picture of HOX proteins as cancer hot spots, strategies to block insulin/IGF-triggered signaling are under way. in-frame fusions of members of the nucleopore complexes and On the basis of the studies reported here, it is possible that certain HOX sequences have been identified.10 Apart from these clinical leukemia types may also benefit from new therapeutics tested in findings, HOX factors are popular model oncoproteins that cause these clinical experiments. leukemia in experimental animals on co-overexpression with their known cooperating factors Meis (murine ecotropic integration site 1) and/or PBX (pre-B-cell leukemia homeobox). MATERIALS AND METHODS In contrast to the importance of HOX in normal and malignant hematopoiesis,1–3 disproportionately little is known about HOX Animals, cell lines, nucleic acids and data deposition downstream targets in hematopoietic stem and precursor cells. Conditional Igf1 knockout animals with a floxed exon 4 (B6.129(FVB)-Igf1- One of the most well-characterized genes under HOX control tm1Dlr-/J)18 and Mx-Cre (C57BL/6 J-Tg(Mx1-cre)1Cgn/J) mice19 have been codes for the proto-oncoprotein MYB (myeloblastosis) that can described previously. Both strains were obtained from Jackson Labora- function itself as hematopoietic oncogene.11 Recently, also LMO2 tories (Bar Harbor, MA, USA) through Charles River (Sulzfeld, Germany). C57BL/6 mice were bred in house or purchased from Janvier (LeGenest, (lim-only domain 2), a crucial hematopoietic regulator involved in France). Animal procedures have been approved by local and institutional T-ALL, and the guanine nucleotide exchange factor VAV2 animal welfare review boards under license number 54-2532.1-41/11. 12,13 have been identified as HOXA9 targets. In addition, the Primary cells transformed by inducible HOX constructs and cells for anti-apoptotic protein BCL2 (B-cell lymphoma) and the growth transplantation were obtained by retroviral transduction of a c-kit-selected

Department of Genetics, University of Erlangen, Erlangen, Germany. Correspondence: Professor RK Slany, Department of Genetics, University of Erlangen, Erwin Rommel Strasse 3, 91058 Erlangen, Germany. E-mail: [email protected] Received 20 June 2014; revised 12 September 2014; accepted 22 September 2014; accepted article preview online 25 September 2014; advance online publication, 21 October 2014 Igf1 in HOX-induced leukemia J Steger et al 902 bone marrow fraction as described.20 Cells were cultivated in RPMI, 10% fetal calf serum and recombinant murine cytokines: 50 ng/ml stem cell

factor and 5 ng/ml each of IL3 (interleukin 3), IL6 and granulocyte- TAM) macrophage colony-stimulating factor (eBioscience, SanDiego, CA, USA). − For activation of HOX constructs, 100 nM 4-hydroxytamoxifen (Sigma, Taufkirchen, Germany) was added. The structure and cloning of HOXA1- estrogen receptor (ER), HOXA6-ER and HOXA9-ER have been published.21 For technical reasons, murine Hoxa9 was used, whereas HOXA1 and

HOXA6 were of human origin. Because human and mouse HOXA9/Hoxa9 = replicates B are identical in 267/272 amino acids and to avoid orthographical , confusion, we are using the human nomenclature with capital letters A (HOXA9) throughout this manuscript. Primers used for quantitative PCR (qPCR) were as follows: Igf1qPCR: fw-5ʹ-CTGGACCAGAGACCCTTTGCG-3ʹ; ʹ ʹ ʹ rev-5 -GGACGGGGACTTCTGAGTCTTG-3 , Igf1rqPCR: fw-5 -CTGTGGGGGC ABA B TCGTGTTTCTC-3ʹ; rev-5ʹ-GATCACCGTGCAGTTTTCCAGG-3ʹ, ChIP: 1fw-5ʹ-G TGCCTCCCATACTGCTTCCTTG-3ʹ;1rev-5ʹ-CTAGATCGAAAGGCAGCTCTCAG-3ʹ, 2fw-5ʹ-GAGCCAAGAATCGGGAATTCTTTG-3ʹ; 2rev-5ʹ-CGCGGTGAGTCTAAGAG CAGAG-3ʹ,3fw-5ʹ-AGAGAATAAGTCAGAGTGGCTGC-3ʹ;3rev-5ʹ-CTCTGGCCAG CTCCTCTACTG-3ʹ, Xfw-5ʹ-AGGGTTTGCTTCCACCCACTCAC-3ʹ;Xrev-5ʹ-CTGGAC CACCCAGAGCTAAACCA-3ʹ, genotyping: ES1-5ʹ-GTTAAAAGCCTCTCAACT On vs off AAGACAATA-3ʹ, IA6-5ʹ-AAACCACACTGCTCGACATTG-3ʹ,ID3-5ʹ-CACTAAGG AGTCTGTATTTGGACC-3ʹ,normfw:5ʹ-GTGCCTCCCATACTGCTTCCTTG-3ʹ; normrev: 5ʹ-CTAGATCGAAAGGCAGCTCTCAG-3ʹ. Array data have been deposited at ArrayExpress (www.ebi.ac.uk) and can be accessed under E-MEXP-3648 and E-MTAB-2603. TAM) Log2 change Hox on (+TAM) Hox off ( −

Isolation of nascent RNA To determine RNA synthesis rates, newly transcribed RNA was labeled by the addition of 100 μM 4-thiouridine (Sigma) directly to the cell culture

medium. After 1 h, total RNA was isolated by RNeasy spin column = replicates HOXA1 B puriﬁcation according to the instructions of the manufacturer (Qiagen, , Hilden, Germany). RNA (100 μg) was biotinylated with 200 μg/ml EZ-Link A HPDP Biotin (Pierce-Thermo, Rockford, IL, USA) in 1 ml of 10 mM Tris-HCl, pH7.4, 1 mM EDTA. Biotinylated nucleic acid was bound to μMACS ABA B

streptavidin magnetic beads, washed as recommended by the manufac- ND ND ND ND 0.88 42 797 40 530 17 005 28 228 turer (Miltenyi, Bergisch-Gladbach, Germany) and separated by magnetic force. Elution of retained material was done with 100 mM dithiothreitol. After a ﬁnal puriﬁcation on RNeasy columns, nascent RNA was subjected to

standard reverse transcription and qPCR. a

Enzyme-linked immunosorbent assay and antibodies On vs off Enzyme-linked immunosorbent assay (ELISA) reagents speciﬁc for mouse Igf1 were obtained from AssayPro (St Charles, MO, USA) and used exactly according to the instructions of the manufacturer. Assay supernatants were 6 TAM) Log2 change Hox on (+TAM) Hox off (

generated by incubating 2 × 10 cells for 18 h in 1 ml of fresh medium. − Mouse monoclonal anti-Igf1 antibody (MAB791) was purchased from R&D Systems (Minneapolis, MN, USA). Anti-Igf1rβ (111A9), anti-Igf1rβ (Tyr1150- /1151P)(19H7) and anti-Akt(Ser473)(D9E) were monoclonal rabbit antibodies, anti-panAkt (4DD4) was of murine origin. These antibodies were developed by Cell Signaling Technologies (Danvers, MA, USA). = replicates HOXA6 B , Igf1r stimulation and serum starvation A Starvation experiments were done in serum-free RPMI1640 supplemented with 0.1% buffered bovine serum albumin (Life Technologies, Darmstadt,

Germany) and 25 μg/ml iron-saturated bovine transferrin (Sigma). After 4 h, ABA B 50 ng/ml recombinant murine Igf1 (eBioscience) was added for 15 min to activate Igf1r and downstream pathways. For apoptosis assays, serum-free but cytokine-supplemented medium was used for overnight incubation. Probe was not present on array, values indicate arbitrary normalized expression units. a

ChIP, apoptosis detection and transplantation experiments HOXA9

Apoptosis was recorded by annexinV and propidium iodide staining Log2 change Hox on (+TAM) Hox off ( according to standard protocols. ChIP was done exactly as described22 using antibodies against the estrogen receptor ligand-binding region anti-ER (TE111.5D11). A secondary amplifying antibody, rabbit anti-mouse IgG/IgM (31198) was added in

equimolar amounts to enhance the ChIP signal. Both antibodies were from ve probe sets with strongest expression in HOXA9-transformed cells ﬁ ThermoScientiﬁc (Pierce, Rockford, IL, USA).

Transplantation studies were done with C57BL/6 mice as recipients. Top Three days before transplantation, animals were treated with 1.1 g/l neomycin and 1 × 106 U/l polymyxin as addition to drinking water. Twenty- four hours after myeloablation by total body irradiation (8 Gy), 0.5 × 106 Accession Gene On vs off NM_001198914NM_010512 mybNM_184052M16449 Igf1 1.00 Igf1 3.01 48 myb 493 3.12 39 724 37 386 0.96 18 530 30 067 12 059 25 642 9468 25 3784 907 21 2686 546 0.69 2347 10 031 1871 14 26 380 746 3.40 26 446 1.21 ND 17 046 2891 15 996 952 2297 1.00 717 278 348 216 66 306 64 081 365 25 228 2.22 39 974 2.23 1120 619 984 10 548 14 363 312 NM_177603 Frat2 1.43 25 588 9664 7241 5868 1.44 3951 4378 1465 1613 1.23 18 197 24 427 6778 11 434 transformed cells and 1 × 106 total bone marrow cells for short-term rescue Table 1. Abbreviation: ND, not determined.

Leukemia (2015) 901 – 908 © 2015 Macmillan Publishers Limited Igf1 in HOX-induced leukemia J Steger et al 903 were transplanted by retroorbital injection. Animals exceeding predefined conjunction with HOX-mediated transformation. The shortlist also termination criteria (hunched posture, ruffled fur, labored breathing, loss of included Frat2, a guanine nucleotide exchange factor activating body weight, pale and/or cold extremities) were killed. Spleen weight was RAC in MLL-induced leukemia.24 Whereas Myb and Frat2 expression recorded as an objective surrogate marker for leukemia burden. levels were not proportional to the in vivo leukemogenic potential of the respective HOX genes, Igf1 clearly complied with this RESULTS prerequisite (Figure 1a). HOXA9 cells contained an order of magnitude more Igf1 RNA than HOXA1 and HOXA6 transformed Igf1 expression is controlled by multiple Hox genes according to populations. To confirm these results, freshly transformed cell lines their leukemogenic potential were generated by retroviral transduction with inducible HOX-ER We have shown previously21 that several members of the HOXA constructs. RNA was harvested in the presence and 72 h after cluster can provoke hematopoietic malignancies in experimental withdrawal of tamoxifen as inducer of HOX activity. Real-time qPCR animals. Each HOX gene tested induced disease with a character- for Igf1 (Figure 1b) supported the array results. Igf1 was under istic latency and penetrance. HOXA9 caused the most aggressive control of each HOX protein tested, as inactivation of the ER fusions leukemia, whereas HOXA1 or HOXA6 elicited leukemia either with lead to a rapid downregulation of Igf1 RNA in all samples. Yet, even long latency or myeloproliferative disease, respectively. These in fully activated state, HOXA6 and, particularly, HOXA1- properties were solely dependent on the nature of the particular transformed cells contained significantly less Igf1 RNA than HOXA9 homeodomain.23 Swapping the homeobox DNA-binding region samples. This was also corroborated at the protein level by a murine between two HOX proteins not only exchanged disease character- Igf1-specific ELISA (Figure 1c). HOXA9-transformed cultures pro- 6 istics reciprocally but also switched the respective target gene duced 2.4 ng/ml Igf1 per 2 × 10 cells in 18 h. Under similar spectrum. Therefore, we surmised that there may be HOX targets conditions, neither HOXA6 nor HOXA1 cells yielded Igf1 above whose expression should be correlated with the respective threshold levels. In addition, Igf1 in extracts from HOXA9 but not transformation potency. When scouring our previous array from HOXA6 cells could be visualized by western blot (Figure 1d). results23 for genes that are commonly regulated by HOXA1, HOXA9 and HOXA6, the five most highly expressed probe sets in HOXA9 Igf1 is a direct target gene of HOXA9 cells represented three transcripts (Table 1). Topmost in expression HOXA9 was the predominant upstream regulator of Igf1 was the well-characterized HOXA9 target Myb, followed by two transcription in our array experiments and Hoxa9 binding at a probe sets for Igf1 that hitherto has not been described in putative enhancer 90 kb upstream of the Igf1 transcription start

Figure 1. HOX-dependent regulation of Igf1 RNA and protein. (a) Heatmap analysis of array data. Log2-transformed expression values for two Igf1-specific probe sets (Igfa and Igfb, corresponding to lines) are plotted for two independent samples (A and B, columns) of cells transformed by inducible versions of HOXA9, HOXA6 and HOXA1 as indicated. RNA was isolated in the presence of tamoxifen (+TAM) and 72 h after HOX activity was switched off by withdrawal of inductor (− TAM). (b) Real-time qPCR analysis of Igf1 expression. Primary cells transformed by conditional HOX derivatives were generated by retroviral transduction. RNA was isolated as in a and Igf1 RNA abundance was measured by qPCR normalized to β-actin. Average and s.d. of a technical triplicate are shown. n.d., not determined. (c) Igf1 production in cell culture supernatants. The supernatant of 2 ×106 HOXA9-transformed cells cultivated in 1 ml of tamoxifen-containing medium for 18 h was analyzed by a murine Igf1-specific ELISA in triplicate. The bar depicts average and s.d. For comparison, the calibration curve is shown. (d) Western blot. Cell extracts of HOXA9, as well as HOXA6 cells grown in the presence and absence of tamoxifen were probed by immunoblotting with an Igf1-specific antibody. Recombinant Igf1 served as positive control.

© 2015 Macmillan Publishers Limited Leukemia (2015) 901 – 908 Igf1 in HOX-induced leukemia J Steger et al 904 site could be detected in high-throughput data obtained in a hypersensitive region. Together, these data argue for a direct previous ChIP-Seq study16 (Figure 2a). Therefore, further ChIP regulation of Igf1 transcription by HOXA9. experiments were performed in HOXA9 cells to scrutinize whether To record the direct influence of HOXA9 on Igf1 transcription binding also occurs around the putative Igf1 promoter (Figure 2b). rates, nascent RNA was isolated from cells transformed by Because the human IGF1 gene contains a DNase-hypersensitive conditional HOXA9-ER. Samples were prepared from cultures region (http://genome.ucsc.edu/ENCODE/) just downstream of grown in the presence of tamoxifen and during a time course after exon 1 in a region conserved in mouse, this potential regulatory inactivation of HOXA9-ER (Figures 2d and e). Newly transcribed feature was included into the analysis. Precipitation with an ER- RNA was labeled by incubation of cells with 4-thiouridine (4S-U) specific antibody yielded a strong 10-fold enrichment of for 1 h. Total RNA was extracted and molecules incorporating 4S-U sequences at the 5ʹ end of the DNase-sensitive area compared were specifically purified by thiol-mediated biotinylation followed with a control site located within an X-based microsatellite. by streptavidin affinity chromatography. Tamoxifen-responsive Weaker HOXA9 binding was detected 500 bp upstream of the estrogen receptor fusions25 are post-transcriptionally regulated. In transcription start site and immediately downstream of the the absence of inducer, these proteins are inactivated by

Figure 2. HOXA9 binds Igf1 regulatory sequences. (a) HOXA9 binds to a putative Igf1 enhancer site on chromosome 10. Genome-wide ChIP- Seq data from a previous study16 were scoured for HOXA9 binding in the vicinity of the Igf1 gene. The arrow indicates a ChIP-Seq peak independently found in two replicates. The genomic structure of Igf1 is drawn to scale. Visualization of data was done with Integrated Genomics Viewer.38 (b) Location of primers used for ChIP. Precipitation of crosslinked material was tested with primers located ~ 500 bp upstream of the transcription start site (TSS, pr1), at the 5ʹ end of a DNase-hypersensitive region (HS, pr2) in the first intron and 3ʹ of this sequence (pr3). As control, a primer amplifying a X-chromosomal satellite sequence (prX) was used. (c) ChIP analysis of HOXA9 binding to putative Igf1 regulatory sequences. Enrichment was calculated relative to background values obtained for primerX. Given are averages and s.d. of triplicates. (d) Igf1 RNA synthesis rates are correlated to HOXA9 activity. HOXA9-ER-transformed cells in the presence of tamoxifen (0 h) or at various time points after washing out the inductor as indicated were labeled for 1 h with 4-thiouridine. RNA synthesized during this period was specifically isolated by thio-selective biotinylation and streptavidin column purification followed by complementary DNA synthesis and qPCR analysis. Newly synthesized Igf1 RNA was quantified by real-time PCR and normalized to β-actin. Averages and s.d. of a technical triplicate are depicted. (e) Schematic depiction of labeling scheme for isolation of nascent RNA.

Leukemia (2015) 901 – 908 © 2015 Macmillan Publishers Limited Igf1 in HOX-induced leukemia J Steger et al 905 sequestration into macromolecular protein complexes.26 Conco- functional Igf1r pathway irrespective of HOXA9 activity, thus mitant with an immediate and direct effect of HOXA9 on Igf1 enabling a potential autocrine signaling loop. transcription, loss of HOXA9 activity was accompanied by a synchronous downregulation of Igf1 synthesis, supporting a direct Igf1 production of HOXA9 cells affects apoptosis and regulation of this transcriptional unit by HOXA9. leukemogenic activity To examine the physiological consequences of Igf1 signaling HOXA9-transformed cells possess a functional Igf1 receptor during HOXA9-induced transformation in a defined genetic enabling Igf1 signaling model, we obtained conditional Igf1 knockout (k.o.) mice that To examine whether the gene coding for the Igf1 receptor is also carry exon 4 of the Igf1 gene flanked by loxP recombination 18 under HOX control, Igf1r RNA was quantified in HOXA9-ER cells in sites. Exon 4 codes for amino acids that are essential for receptor 19 the presence and 72 h after withdrawal of tamoxifen (Figure 3a). In binding. These animals were crossed with Mx-Cre deleter mice contrast to Igf1, Igf1r was only marginally influenced by HOXA9 and subsequently the offspring was bred to homozygosity for the fl/fl indicating that this gene is constitutively active. Western blots floxed Igf1 allele (Igf × Mx-Cre). Hematopoietic progenitors from confirmed this result (Figure 3b, upper panel). At the protein level, these animals were retrovirally transduced with HOXA9 in ‘ ’ ‘ combination with Meis1 to allow for efficient leukemogenesis in even more Igf1r was present in HOX-off cells than in their HOX- fl/fl on’ counterpart. This indicates a post-transcriptional upregulation recipient animals. Igf1 mice without Cre transgene served as control. Transformed cell lines could be easily established from during differentiation after inactivation of HOXA9. The biochem- fi ical function of the Igf1r was investigated by starvation experi- both genotypes and there was no signi cant difference in ments, as the high levels of Igf1 in serum (25–50 ng/ml according proliferation, surface phenotype or colony-forming cell numbers in the two populations under normal growth conditions (not to Honegger27) caused unacceptable background. Cells were shown). In addition, co-transduction with Meis1 did not alter HOX incubated for 4 h in serum-free media followed by stimulation control of Igf1 transcription compared with HOXA9 acting alone with recombinant Igf1 because ‘HOX-on’ cells did not produce fi (Figure 4a). A PCR experiment checking the Igf1 excision status suf cient Igf1 themselves during this short incubation period (Figure 4b) showed that retroviral transduction had already under suboptimal starvation conditions to reach the detection activated the Mx promoter. Excision was nearly complete leading limit for autostimulation of Igf1r. Nevertheless, the receptor was to a 98–99% loss of the Igf1 exon 4 in the HOX/Meis cells as clearly functional, as a potent and rapid activation of the Igf1r was determined by qPCR. Incubation of the cells in medium containing evidenced by the appearance of Y1150/51-phosphorylated 1000 U/ml murine IFNβ had no additional effect. Because high receptor chains after addition of recombinant Igf1 (Figure 3b, levels of Igf1 in bovine serum obfuscates any differences between lower panel). A parallel result was obtained for the Igf1r Igf1 k.o. and wild-type (wt) cells, the importance of potential downstream signal component Akt (Figure 3c, upper and lower autocrine Igf1 signaling and pro-survival Akt activation was tested panels). The presence of sufficient Igf1 ligand translated into an again in starvation experiments (Figure 4c). HOX/Meis cells of Igf1 increased phosphorylation of Akt at serine 473. Interestingly, and k.o. and wt genotypes were cultivated overnight in the absence of in line with the observed increase of Igf1r protein, ‘HOX-off’ cells serum. Annexin/PI staining demonstrated that Igf1 wt cells were also contained significantly more Akt than the ‘HOX-on’ popula- significantly better protected from cell death than their knockout tion, suggesting a possible upregulation of the whole Igf1- counterparts. Under these conditions, apoptosis rates increased downstream signaling pathway as a consequence of differentia- from 30% to over 50% in the absence of endogenous Igf1 tion. In summary, these observations confirmed the presence of a production.

Figure 3. HOXA9 cells express a functional Igf1 receptor and respond to Igf1 stimulation independently of HOX activity. (a) Real-time PCR quantiﬁcation of Igf1 receptor RNA in HOXA9-ER-transformed cells in the presence of tamoxifen (HOX-on) and 72 h after withdrawal of inductor (HOX-off). (b) Detection of total and activated Igf1 receptor β-subunit in HOXA9-ER-transformed cells starved for 4 h before they were either stimulated with Igf1 (25 ng) or vehicle only ( − ). Upper panel: Pan-Igf1rβ was detected by immunoblotting in extracts from cells cultured under HOX-activating conditions (+TAM) and 72 h after withdrawal of inductor ( − TAM) disabling HOX function. Equal amounts of total protein were loaded into each lane. Lower panel: the same extracts were analyzed for receptor activation by immunoblotting and detection of Tyr-phosphorylated Igf1rβ.(c) Stimulation of Akt by Igf1 treatment. Upper panel: cell extracts produced as in b were examined for total amounts of Akt protein kinase by an anti-pan Akt blot. Middle and lower panel: corresponding immunoblot performed with the same extracts and speciﬁcally probed for Akt Ser473P. Two different exposures are shown to allow visualization of Akt status also in HOX-on cells.

Figure 4. Genetic Igf1 ablation increases apoptosis and impairs leukemogenesis of HOXA9-transformed cells. (a) Regulation of Igf1 RNA by HOXA9 in HOXA9-ER/Meis1-cotransduced cells. Analysis was done by real-time qPCR. (b) Targeting strategy and proof of excision. Primer pairs and genomic location of recombination sites are depicted schematically. The gel shows end point PCR products amplified from HOXA9/Meis1- transformed cells derived from either Igf1fl/fl × Cre (Mx-Cre)orIgf fl/fl control mice as indicated. Genomic DNA was isolated from cell populations grown under standard conditions or after 3 days incubation with 1000 U/ml recombinant murine interferon β (IFNβ). The bar chart represents a quantification of the excision reaction, as measured by qPCR with primers IA6+IA3 and normalized to a normfw/normrev amplicon. Averages and s.d. of a triplicate measurement are shown. (c) Annexin/propidium iodide (PI) staining of serum-starved cells. Cells with Igf1fl/fl or Igf1fl/fl × Cre genotype were incubated in triplicate for 18 h in normal or serum-free media as indicated. Proportions of cells negative for annexin V/PI (live), annexin V positive (annex) and double-positive apoptotic cells (annex/PI) are given as average and s.d. (d) Transplantation experiments. Left panel: Kaplan–Meier survival curve of animals receiving a syngenic graft of HOXA9/Meis1-transformed cells as shown in a. One animal receiving cells of Igf fl/fl × Cre genotype did not survive myeloablative conditioning and was censored from the experiment. Right panel: recorded spleen weights after killing. n.s., not significant. (e) Quantification of Igf1 excision in cell samples re-derived from spleen of leukemic animals. q-PCR analysis was done as in a. Control DNA from an Igf1fl/fl leukemia was used.

The leukemogenic potential of Igf1 k.o. vs wt cells was were re-grown from spleens of leukemic recipients contained examined in syngenic transplantation experiments (Figure 4d). more than 10% of cells with Igf1 wt configuration (Figure 4e). This In these experiments, cells with a deletion of Igf1 took significantly is a clear increase from the ~ 1% non-recombined cells present in longer to elicit fatal leukemia than wt cells. Importantly, spleen the original graft. This is best reconciled with a growth weights of wt and Igf1 k.o. cohorts did not differ, indicating that and/or survival advantage of an occasional cell that escaped the disease scoring system was functional and that both groups Cre-mediated excision before transplantation. were killed at the same stage of disease. This suggested that Igf1 k.o. cells needed more time to accumulate to numbers that were deleterious for the health of the animals. The lower fitness of Igf1 DISCUSSION k.o. cells compared with that of their wt counterparts was Igf1 is a key molecule in the regulatory circuit that controls fetal underlined by the fact that three out of five leukemia samples that and juvenile growth as well as body size. As such, Igf1 affects the

Leukemia (2015) 901 – 908 © 2015 Macmillan Publishers Limited Igf1 in HOX-induced leukemia J Steger et al 907 proliferation of many different cell types explaining its universal Certainly, IGF1 is not the only factor impacting on the effects on cell proliferation and apoptosis. Here we present aggressiveness of HOX-induced leukemia. Nevertheless, all avail- evidence that Igf1 is induced by several leukemogenic HOXA able evidence points to a significant contribution of this central proteins. This goes well with the function of HOX transcription signaling molecule toward leukemia development. The relatively factors during embryonic development and it fits also with their well-known PI3K/AKT pathway triggered by IGF1 in combination purpose to control the size of the transiently proliferating with a cornucopia of different inhibitors in well-advanced clinical precursor pool in steady state hematopoiesis. The correlation of studies suggests that an anti-IGF1 medication may be well advised Igf1 induction and the severity of the leukemic phenotype caused at least as adjunct therapy during leukemia treatment. by the respective HOX protein suggests that Igf1 is at least one of the contributing factors that modulates disease outcome. In a clinical setting, IGF1 and the related insulin signal pathway CONFLICT OF INTEREST has been long recognized as a risk factor for the development of The authors declare no conflict of interest. malignant disease. This is spectacularly demonstrated by patients suffering from either congenital IGF1 deficiency or from defects in ACKNOWLEDGEMENTS growth hormone or its receptor (Laron syndrome), thus losing the physiological rheostat that adjusts the production of IGF1.28,29 We thank Renate Zimmermann for technical assistance and Constanze Breitinger for Compared with controls, this patient cohort enjoys a clear help in early stages of this study. This work was supported by research funding from protection from the development of malignant disease. In the Deutsche Forschungsgemeinschaft (SL27/8-1) to RKS. contrast, growth hormone and, therefore, also IGF1 overproduc- tion in acromegaly, is associated with an increased risk for acute REFERENCES myeloid leukemia and other malignancies.30 1 Alharbi RA, Pettengell R, Pandha HS, Morgan R. The role of HOX genes in normal Investigations assessing the role of IGF1 signaling in leukemia hematopoiesis and acute leukemia. Leukemia 2013; 27: 1000–1008. focused mainly on the IGF1 receptor demonstrating that an intact 2 Argiropoulos B, Humphries RK. Hox genes in hematopoiesis and leukemogenesis. 31 IGF1R is an absolute prerequisite for the development of T-ALL. Oncogene 2007; 26: 6766–6776. Activation of IGF1R also stimulated proliferation of B-ALL samples 3 Eklund EA. The role of HOX genes in malignant myeloid disease. Curr Opin and B-ALL cell lines.32 Moreover, blocking IGF1R tyrosine kinase Hematol 2007; 14:85–89. activity by knockdown or the inhibitor sorafenib affected viability 4 Caudell D, Zhang Z, Chung YJ, Aplan PD. Expression of a CALM-AF10 fusion gene of B-cell chronic lymphocytic leukemia.33 Inappropriate IGF1R leads to Hoxa cluster overexpression and acute leukemia in transgenic mice. 67 – activity is also the major reason for the constitutive PI3K and AKT Cancer Res 2007; : 8022 8031. 5 Mullighan CG, Kennedy A, Zhou X, Radtke I, Phillips LA, Shurtleff SA et al. Pediatric activation that can be frequently detected in myeloid leukemia.34 35 acute myeloid leukemia with NPM1 mutations is characterized by a gene A recent report examining the consequences of genetic Igf1r expression profile with dysregulated HOX gene expression distinct from ablation for MLL-AF9-mediated leukemogenesis in mice supports MLL-rearranged leukemias. Leukemia 2007; 21: 2000–2009. our present studies. It is well established that MLL-AF9 induces 6 Okada Y, Jiang Q, Lemieux M, Jeannotte L, Su L, Zhang Y. Leukaemic Hox gene expression and a significant part of the transforming transformation by CALM-AF10 involves upregulation of Hoxa5 by hDOT1L. activity of MLL fusion proteins can be explained by ectopic Hox Nat Cell Biol 2006; 8: 1017–1024. activation. Attenuation of Igf1r expression by a hypomorphic allele 7 Rawat VP, Humphries RK, Buske C. Beyond Hox: the role of ParaHox genes in 120 – caused a cell autonomous defect reducing MLL-AF9-transforming normal and malignant hematopoiesis. Blood 2012; :519 527. capability in clonogenic assays, underlining the importance of the 8 Zeisig BB, Milne T, Garcia-Cuellar MP, Schreiner S, Martin ME, Fuchs U et al. Hoxa9 and Meis1 are key targets for MLL-ENL-mediated cellular immortalization. Mol Cell Igf1 signal for in vitro transformation. However, once established Biol 2004; 24: 617–628. in recipient animals, MLL-AF9 cells became independent on Igf1r 9 Speleman F, Cauwelier B, Dastugue N, Cools J, Verhasselt B, Poppe B et al. Anew activity. In our experiments, a possible impact of Igf1 ablation on recurrent inversion, inv(7)(p15q34), leads to transcriptional activation of HOXA10 clonogenic activity (colony-forming cell assays) was obscured and HOXA11 in a subset of T-cell acute lymphoblastic leukemias. Leukemia 2005; because the fetal calf serum necessary for cell viability supplies 19: 358–366. copious amounts of this growth factor. Yet, during in vivo 10 Gough SM, Slape CI, Aplan PD. NUP98 gene fusions and hematopoietic leukemogenesis, we observed effects of Igf1 deficiency that malignancies: common themes and new biologic insights. Blood 2011; 118: 6247–6257. would not necessarily appear in an Igf1 receptor knockout 11 Hess JL, Bittner CB, Zeisig DT, Bach C, Fuchs U, Borkhardt A et al. c-Myb is an 36 essential downstream target for homeobox-mediated transformation of because Igf1 also signals through the insulin receptor. Igf1r hematopoietic cells. Blood 2006; 108:297–304. itself is not under HOX regulation and Igf1r protein was even 12 Breitinger C, Maethner E, Garcia-Cuellar MP, Schambony A, Fischer KD, Schilling K upregulated in cells after shutdown of HOXA9 activity. Most likely, et al. HOX genes regulate Rac1 activity in hematopoietic cells through control of Igf1 production by leukemic cells in addition to the naturally Vav2 expression. Leukemia 2013; 27:236–238. occurring plasma levels creates higher local concentrations that 13 Calero-Nieto FJ, Joshi A, Bonadies N, Kinston S, Chan WI, Gudgin E et al. HOX- make the microenvironment more favorable for engraftment, mediated LMO2 expression in embryonic mesoderm is recapitulated in acute survival and proliferation. This may endow cells that synthesize leukaemias. Oncogene 2013; 32: 5471–5480. Igf1 with a competitive advantage compared with their knockout 14 Brumatti G, Salmanidis M, Kok CH, Bilardi RA, Sandow JJ, Silke N et al. HoxA9 regulated Bcl-2 expression mediates survival of myeloid progenitors and the counterparts. Indeed, a selective expansion of residual cells that severity of HoxA9-dependent leukemia. Oncotarget 2013; 4:1933–1947. failed to recombine the loxP sites was seen in our transplantation 15 Shah CA, Bei L, Wang H, Platanias LC, Eklund EA. The leukemia-associated experiments. Mll-Ell oncoprotein induces fibroblast growth factor 2 (Fgf2)-dependent Another aspect of our findings impacts on the long-standing cytokine hypersensitivity in myeloid progenitor cells. J Biol Chem 2013; 288: observation of an association between higher birth weight and 32490–32505. the risk of childhood leukemia.37 Fetal growth is under direct 16 Huang Y, Sitwala K, Bronstein J, Sanders D, Dandekar M, Collins C et al. Identifi- control of circulating IGF1. In this respect, it is interesting to note cation and characterization of Hoxa9 binding sites in hematopoietic cells. Blood that the majority of infant leukemia that manifests at birth harbor 2012; 119: 388–398. MLL rearrangements and therefore they bear high-level HOX 17 Pollak M. The insulin and insulin-like growth factor receptor family in neoplasia: an update. Nat Rev Cancer 2012; 12:159–169. expression. It seems conceivable that not only fetal IGF1 18 Liu JL, Grinberg A, Westphal H, Sauer B, Accili D, Karas M et al. Insulin-like exacerbates tumor burden, but also that tumor-derived IGF1 growth factor-I affects perinatal lethality and postnatal development in a gene supports fetal growth, thus establishing a mutual dependency dosage-dependent manner: manipulation using the Cre/loxP system in that may explain at least part of the epidemiological evidence. transgenic mice. Mol Endocrinol 1998; 12: 1452–1462.

© 2015 Macmillan Publishers Limited Leukemia (2015) 901 – 908 Igf1 in HOX-induced leukemia J Steger et al 908 19 Kuhn R, Schwenk F, Aguet M, Rajewsky K. Inducible gene targeting in mice. 30 Ozeki K, Morishita Y, Saito S, Umemura K, Yamaguchi Y, Tatekawa S et al. Science 1995; 269: 1427–1429. Acute myeloid leukemia and colon carcinoma during the course of acromegaly. 20 Lavau C, Szilvassy SJ, Slany R, Cleary ML. Immortalization and leukemic transfor- Int J Hematol 2013; 98: 620–624. mation of a myelomonocytic precursor by retrovirally transduced HRX-ENL. EMBO 31 Medyouf H, Gusscott S, Wang H, Tseng JC, Wai C, Nemirovsky O et al. J 1997; 16: 4226–4237. High-level IGF1R expression is required for leukemia-initiating cell activity 21 Bach C, Buhl S, Mueller D, Garcia-Cuellar MP, Maethner E, Slany RK. Leukemogenic in T-ALL and is supported by Notch signaling. J Exp Med 2011; 208: transformation by HOXA cluster genes. Blood 2010; 115: 2910–2918. 1809–1822. 22 Milne TA, Zhao K, Hess JL. Chromatin immunoprecipitation (ChIP) for analysis of 32 Yamada H, Iijima K, Tomita O, Taguchi T, Miharu M, Kobayashi K et al. Effects of histone modifications and chromatin-associated proteins. Methods Mol Biol 2009; insulin-like growth factor-1 on B-cell precursor acute lymphoblastic leukemia. Int J 538: 409–423. Hematol 2013; 97:73–82. 23 Breitinger C, Maethner E, Garcia-Cuellar MP, Slany RK. The homeodomain region 33 Yaktapour N, Ubelhart R, Schuler J, Aumann K, Dierks C, Burger M et al. Insulin-like controls the phenotype of HOX-induced murine leukemia. Blood 2012; 120: growth factor-1 receptor (IGF1R) as a novel target in chronic lymphocytic 4018–4027. leukemia. Blood 2013; 122: 1621–1633. 24 Walf-Vorderwulbecke V, de Boer J, Horton SJ, van Amerongen R, Proost N, Berns A 34 Chapuis N, Tamburini J, Cornillet-Lefebvre P, Gillot L, Bardet V, Willems L et al. et al. Frat2 mediates the oncogenic activation of Rac by MLL fusions. Blood 2012; Autocrine IGF-1/IGF-1R signaling is responsible for constitutive PI3K/Akt activation 120: 4819–4828. in acute myeloid leukemia: therapeutic value of neutralizing anti-IGF-1R antibody. 25 Littlewood TD, Hancock DC, Danielian PS, Parker MG, Evan GI. A modified Haematologica 2010; 95:415–423. oestrogen receptor ligand-binding domain as an improved switch for the 35 Jenkins CR, Shevchuk OO, Giambra V, Lam SH, Carboni JM, Gottardis MM et al. regulation of heterologous proteins. Nucleic Acids Res 1995; 23: 1686–1690. IGF signaling contributes to malignant transformation of hematopoietic 26 Hager GL, Lim CS, Elbi C, Baumann CT. Trafficking of nuclear receptors in progenitors by the MLL-AF9 oncoprotein. Exp Hematol 2012; 40: living cells. J Steroid Biochem Mol Biol 2000; 74:249–254. 715–723 e716. 27 Honegger A, Humbel RE. Insulin-like growth factors I and II in fetal and adult 36 Belfiore A, Frasca F, Pandini G, Sciacca L, Vigneri R. Insulin receptor isoforms and bovine serum. Purification, primary structures, and immunological cross- insulin receptor/insulin-like growth factor receptor hybrids in physiology and reactivities. J Biol Chem 1986; 261:569–575. disease. Endocr Rev 2009; 30: 586–623. 28 Guevara-Aguirre J, Balasubramanian P, Guevara-Aguirre M, Wei M, Madia F, Cheng CW 37 Chokkalingam AP, Metayer C, Scelo G, Chang JS, Schiffman J, Urayama KY et al. et al. Growth hormone receptor deficiency is associated with a major reduction in Fetal growth and body size genes and risk of childhood acute lymphoblastic pro-aging signaling, cancer, and diabetes in humans. SciTranslMed2011; 3: 70ra13. leukemia. Cancer Causes Control 2012; 23: 1577–1585. 29 Steuerman R, Shevah O, Laron Z. Congenital IGF1 deficiency tends to confer 38 Thorvaldsdottir H, Robinson JT, Mesirov JP. Integrative Genomics Viewer (IGV): protection against post-natal development of malignancies. Eur J Endocrinol 2011; high-performance genomics data visualization and exploration. Brief Bioinform 164: 485–489. 2013; 14: 178–192.