ARTICLE

https://doi.org/10.1038/s41467-019-09273-z OPEN Gfi1b regulates the level of Wnt/β-catenin signaling in hematopoietic stem cells and megakaryocytes

Peiman Shooshtarizadeh1, Anne Helness1, Charles Vadnais1, Nelleke Brouwer1, Hugues Beauchemin1, Riyan Chen1, Halil Bagci1,2, Frank J.T. Staal 3, Jean-François Coté 1,2,4,5 & Tarik Möröy1,6,7

Gfi1b is a transcriptional repressor expressed in hematopoietic stem cells (HSCs) and megakaryocytes (MKs). Gfi1b deficiency leads to expansion of both cell types and abrogates

1234567890():,; the ability of MKs to respond to integrin. Here we show that Gfi1b forms complexes with β-catenin, its co-factors Pontin52, CHD8, TLE3 and CtBP1 and regulates Wnt/β-catenin- dependent gene expression. In reporter assays, Gfi1b can activate TCF-dependent tran- scription and Wnt3a treatment enhances this activation. This requires interaction between Gfi1b and LSD1 and suggests that a tripartite β-catenin/Gfi1b/LSD1 complex exists, which regulates Wnt/β-catenin target genes. Consistently, numerous canonical Wnt/β-catenin target genes, co-occupied by Gfi1b, β-catenin and LSD1, have their expression deregulated in Gfi1b-deficient cells. When Gfi1b-deficient cells are treated with Wnt3a, their normal cellularity is restored and Gfi1b-deficient MKs regained their ability to spread on integrin substrates. This indicates that Gfi1b controls both the cellularity and functional integrity of HSCs and MKs by regulating Wnt/β-catenin signaling pathway.

1 Institut de recherches cliniques de Montréal, Montreal H2W 1R7 QC, Canada. 2 Department of Anatomy and Cell Biology, McGill University, Montreal,́ QC H3A 0C7, Canada. 3 Department of Immunohematology and Bloodtransfusion, Leiden University Medical Center, Leiden 2333 ZA, The Netherlands. 4 Programmes de Biologie Moleculaire,́ Departement́ de Medecine,́ Universitéde Montreal,́ Montreal,́ QC H3T 1J4, Canada. 5 Departement́ de Biochimie, Universitéde Montreal,́ Montreal,́ QC H3C 3J7, Canada. 6 Département de microbiologie, infectiologie et immunologie, Université de Montréal, C.P. 6128, succ. Centre-ville Montreal H3C 3J7 QC, Canada. 7 Division of Experimental Medicine, McGill University, Montreal H4A 3J1 QC, Canada. These authors contributed equally: Peiman Shooshtarizadeh, Anne Helness, Charles Vadnais. Correspondence and requests for materials should be addressed to T.M. (email: [email protected])

NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z

rowth factor independence 1b (Gfi1b) and its paralogue Results G Gfi1 are transcription factors that are expressed in a Gfi1b deficiency leads to expansion of HSCs and MKs.To complementary and partially overlapping manner in generate Gfi1b-deficient (KO) mice we introduced a Rosa-CreER hematopoietic stem cells (HSCs) and precursors for several transgene into Gfi1bfl/fl mice9,11. Floxed Gfi1b alleles were deleted lineages1,2.Gfi1b is expressed in HSCs, myeloid/erythroid by tamoxifen injections (Fig. 1a) and confirmed in both MKs and precursors (MEPs), megakaryocytes (MKs) and to varying HSCs by the absence of floxed exons 2–4 expression (Supple- levels during erythrocyte maturation2.BothGfi1andGfi1b mentary Fig. 1a). MKs and their progenitors were defined as lin−, have an N-terminal Snail/Gfi1(SNAG)domain,whichenables CD4high, CD9high cells, and HSCs were defined either as lin−, transcriptional repression through the recruitment of cofac- Sca-1+, cKit+ (LSK), CD48−, CD150+, CD41high/CD9high or tors Lysine (K)-specific demethylase 1A (LSD1/KDM1A) CD41low/CD9low to eliminate potential contamination by MK and CoREST/Rcor13–5.Interestingly,LSD1seemstoplayan precursors (Supplementary Fig. 1b), and also because we had essential structural rather than enzymatic role as part of the shown previously that the expression of integrin molecules, such Gfi1 repressive complex6.LSD1’slossdisruptsGfi1’sasso- as Itga2b (CD41) and Itgb3 (CD61), are up-regulated in Gfi1b- ciation with and repression of target loci. Gfi1andGfi1b also deficient HSCs9. Following the ablation of Gfi1b, we observed both interact with co-repressors, such as histone lysine increased numbers of MKs and CD41low/CD9low HSCs in the methyltransferase 2 (EHMT2/G9a) and histone deacetylases bone marrow and blood, (Fig. 1b, c) confirming previous (HDAC1/2)4,7,8. While germline deletion of Gfi1b in mice results11. Moreover, Gfi1b-deficient MKs and HSCs showed the causes lethality at around day 14.5 of embryonic development, same expansion in vitro after Tamoxifen injection (Fig. 1d, e) conditional knockout mice have been generated and show supporting that Gfi1b controls the cellularity of both HSCs and that Gfi1b controls HSC and MK expansion9,10.WhileGfi1b- MKs in bone marrow and in peripheral blood in a cell autono- deficient HSCs remain functional and give rise to all hema- mous manner9. topoietic lineages upon transplantation, MKs that lack Gfi1b We recently showed that Gfi1b loss leads to impaired MK cannot produce platelets and are unable to respond with spreading and motility on fibronectin and other matrix spreading and membrane ruffling to integrin receptor stimu- substrates11. On such substrates, Gfi1b KO MKs are smaller, lation due to defects in cytoskeletal organization11. stay completely round, and do not form protrusions like WT cells Wnt/β-catenin signaling also plays a crucial role in early or form platelets (Fig. 1f)11. A comparison of the roundness hematopoiesis, notably in HSCs. Loss- and gain-of-function index (RI) of cultured cells clearly showed that Gfi1b-deficient studies demonstrated that tight control of Wnt signaling and MKs maintained significantly low RIs, indicating that they are β-catenin activity is necessary for proper function and cellularity unable to spread and form protrusions, confirming a poor control of hematopoietic cells including HSCs and MKs12–15. integrin ligands response (Fig. 1g). Overactive Wnt/β-catenin signaling leads to exhaustion of HSCs, but insufficient activation is equally detrimental16,17. β-catenin GFI1B is associated with regulators of Wnt pathway.To acts as a transcriptional co-activator in complexes with tran- understand how Gfi1b deficiency leads to expansion of HSCs and scription factors, such as the T-cell factor/lymphoid enhancer MKs and loss of MK spreading, a Flag-tagged version of GFI1B factor (TCF/LEF) family members to regulate gene expression. was expressed in HEK293 cells to identify interacting proteins by The canonical Wnt signaling is under negative regulation at immune-precipitation and mass spectrometry. In addition to various levels. For instance, GRG/TLE (Groucho/transducin-like proteins already known to bind to GFI1B, such as LSD1, HDACs, enhancer) proteins associate with TCF molecules in the nucleus CoRest factors and G9a4, we found β-catenin and several to switch off expression of Wnt target genes in the absence of other proteins with regulatory roles in the canonical Wnt path- nuclear β-catenin18. CtBP1 and HDACs are other negative reg- way as new potential GFI1B binding partners (Fig. 2a, Supple- ulators of canonical Wnt signaling. Multiple non-canonical mentary Fig 1c, Supplementary Table 2). These included the Wnt signaling pathways also exist and although these pathways DNA helicase and chromatin remodeling factor CHD8, which all function in a β-catenin independent manner, crosstalk exists silences β-catenin mediated transcription23, Pontin52 (also between canonical and non-canonical signaling pathways in known as TIP49), a c-Myc interacting protein involved in chro- various contexts19,20. Several studies have shown that non- matin remodeling and transcription24, APC, a tumor suppressor canonical Wnt signaling antagonizes the canonical Wnt pathway and antagonist of the Wnt signaling pathway18, PP2A, a phos- through different mechanisms21,22; one example being NFAT5, phatase that regulates Wnt signaling at several levels25, and which is a transcription factor downstream of the non-canonical CtBP1 (C-terminal binding protein1), a repressor of Wnt sig- Wnt/Ca2+ pathway that inhibits canonical Wnt signaling via naling that associates with LSD126 (Fig. 2a, Supplementary inhibition of β-catenin acetylation21. Fig 1c). Gene Ontology (GO) enrichment analysis confirmed that Here we present evidence that Gfi1b controls HSC and MK numerous GFI1B associated proteins were annotated under the cellularity and MK spreading in response to integrin substrates by biological process positive regulation of Wnt signaling pathway regulating Wnt/β-catenin signaling. Our results show that Gfi1b (Supplementary Fig 1d). interacts with β-catenin as well as regulators of Wnt/β-catenin Similarly, a GFI1B-BirA* BioID approach showed that signaling pathway and that loss of Gfi1b affects the expression of regulators of the Wnt pathway, such as KDM2A and KDM2B, Wnt target genes in both MKs and HSCs. We also reveal a tri- which regulate the stability of nuclear β-catenin via demethyla- partite Gfi1b/LSD1/β-catenin complex that co-occupies key Wnt/ tion27, TLE1 and TLE3 which are both β-catenin inhibitors28 β-catenin signaling target regions like the Axin2 promoter. We and CREBBP (also known as CBP) a histone acetyltransferase show that Gfi1b can enhance transcription of TCF/LEF depen- that binds to β-catenin29 (Fig. 2b, Supplementary Fig 1e) can dent promoters and reporter genes in vitro and in vivo and we interact with GFI1B. In addition, we also found Spindlin1, a present evidence that Gfi1b does this by recruiting LSD1 via its chromatin reader that prompts Wnt signaling30, UBR5, an E3 SNAG domain to β-catenin containing complexes. In agreement ubiquitin-protein ligase that inactivates TLE31, CDC73, which is with this, we show that Gfi1b-deficient HSCs and MKs have a β-catenin interacting tumor suppressor32, CCAR2, which decreased levels of canonical Wnt signaling in vivo, which can be enhances LEF1-β-catenin complex formation33 and BRG1, a reversed when Wnt/β-catenin signaling is stimulated externally chromatin remodeler involved in transactivation of Wnt target by Wnt3A treatment. genes (Fig. 2b)34.

2 NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z ARTICLE

a b c Gfi1b WT

Gfi1b KO Tamoxifen Analysis 1.0 0.8 80 *** 0.08 0.008 7 ** * 70 6 d0 d1 d10 0.8 ** * 0.6 60 0.06 0.006 5 0.6 50 4 0.4 40 0.04 0.004 3 0.4 30 0.2 20 0.02 0.002 2

% lin- BM cells 0.2 1 10 % lin- blood cells 0.0 0.0 0 0.00 0.000 0 Rosa-cre tg HSCs HSCs MK HSCs HSCs MK Gfi1bWT/flox or Gfi1Bflox/flox CD41+/CD9+ CD41low/CD9low CD41+/CD9+ CD41low/CD9low

d e 15 4 Tamoxifen FACS Analysis 80 ** *** d0 d1 d2 d7 3 3 60 Lin depleted 10 * BM cuture 2 40 5 1 20 Absolute number of Rosa-cre tg cells in culture ×10 Gfi1bWT/flox or Gfi1Bflox/flox 0 0 0 HSCs HSCs MK CD41+/CD9+ CD41low/CD9low

f g PF4-cre tg PF4-cre tg 3.5 Gfi1bWT/flox Gfi1bflox/flox 3.0 2.5

2.0 *** 1.5

MK roundness index 1.0

50 μm 50 μm

DAPI CD41 βtubulin PF4-cre tg PF4-cre tg

Gfi1bWT/flox Gfi1bflox/flox

Fig. 1 Loss of Gfi1b in MKs and HCSs causes their expansion and mobilization into the blood. a Schematic of Tamoxifen injections; mice were analyzed by FACS at day 10 following the first injection. b, c Quantification of Gfi1b WT/KO MKs and HSCs in lineage negative depleted BM and blood (n = 3 mice per group). d Schematic of in vitro analysis of Gfi1b WT/KO HSCs and MKs expansion. e MKs and HSCs (CD41low/CD9low) were taken into culture and were quantified by FACS at day six (day seven post Tamoxifen injection, n = 3 mice per group). f, g Loss of Gfi1b leads to impaired MKs spreading on Fibronectin coated matrix as quantified by calculating their roundness index (n = 70 Gfi1b wt/flox cells and 220 Gfi1b flox/flox cells). (*p < 0.05, **p < 0.001, ***p < 0.0001 on a Welch corrected t-test, error bars show s.d)

We noticed a well-conserved six amino-acid motif in the N- immune-precipitations in the presence of GFI1B or the GFI1B terminal sequence of GFI1B protein that we named WRD for ΔSNAG mutant that lacks the 20 N-terminal amino acids that Wnt regulatory domain, since a similar motif has previously been are required for LSD1 binding (Fig. 3b). Similarly, immune- shown in other proteins to mediate binding of the β-catenin complexes collected with GST-E-Cadherin that binds to β-catenin inhibitors TLE1 and TLE328 (Fig. 2c). Immune-precipitation of from U2OS cells engineered to express doxycycline-inducible Flag-tagged WT GFI1B or different mutants, in which either the GFI1B only contained more LSD1 when GFI1B expression SNAG, WRD or intermediate domain were disrupted, confirmed was induced (Fig. 3c), suggesting that GFI1B recruits LSD1 to that GFI1B interacts with β-catenin, CHD8, Pontin52, and CtBP1 β-catenin or β-catenin-containing protein complexes. To char- and to some extent with both TLE3 and TLE1 (Fig. 2d). The acterize these complexes further, we performed β-catenin-BirA* SNAG domain was required for an interaction between GFI1B BioID in the absence or presence of WT GFI1B or two GFI1B and CtBP1 and, as previously shown, LSD14, whereas the WRD mutants (Fig. 3d, e). All three GFI1B forms were biotinylated domain was necessary for GFI1B interaction with CHD8, TLE3, by β-catenin-BirA* further indicating an interaction or close Pontin52, and to some extent TLE1 (Fig. 2d). Moreover, GFI1B’s proximity between GFI1B and β-catenin (Fig. 3d, e) and zinc finger domain was required for its interaction with β-catenin confirming both our mass spectrometric and co-IP experiment (Fig. 2e). Endogenous immune-precipitation with extracts from findings that also suggest that GFI1B binds to β-catenin (Fig. 2a, non-transfected K562 and HEL cells demonstrated an enrichment d, and e). of complexes between GFI1B, β-catenin, and Pontin52 (Fig. 2f–h and Supplementary Fig. 2a). These results indicate that GFI1B GFI1B enhances transcription of TCF/LEF dependent reporter. interacts with β-catenin and several Wnt/β-catenin signaling We found that GFI1B enhanced transcription of a TCF promoter regulators. reporter both at basal level and in the presence of Wnt3A, a β-catenin co-precipitated with GFI1B and LSD1 and increased canonical Wnt signaling ligand, following over-expression of levels of LSD1 protein were collected when exogenous WT GFI1B Wnt3A or the active form of β-catenin or upon treatment with was present (Fig. 3a). This was not true when the GFI1B P2A LiCl, a GSK3β inhibitor that stabilizes β-catenin (Fig. 4a–e), or mutant was present instead, which is unable to bind LSD15 CHIR99021, a specificGSK3β inhibitor (Fig. 4f, Supplementary (Fig. 3a). Comparable results were obtained with anti-β-catenin Fig. 2b, c) indicating that GFI1B can activate β-catenin/TCF

NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z

a b c Histone modifying enzymes Wnt regulatory FSXXXL GFI1B-BIRA domain (WRD) IP flag Coverage % Exclusive unique spectrum count 106 111 KDM1A* nd nd kD –+GFI1B-flag KDM1A* CHD8 Human SFSWDTL RCOR1* 44 1 200 TLE1 RCOR2 6.4 2 KDM2A Mouse SFSWDTL RCOR3* 47 11 KDM2B TLE3 116 HDAC1* 45 29 X.Laevis NFSWDAL HDAC1* 97 MS HDAC2* 8.4 5 KDM3B EHMT2* 3.2 3 KDM4B KDM1A* GFI1B muWRD SASWDTA 66 HDAC1* KDM2A Wnt pathway regulators KDM2B Ig β HDAC2* -catenin 17 7 GFI1B-Flag CHD8 0.4 1 EHMT1 CREBBP 45 MS Pontin52 37 14 EHMT2* Spindlin1 GFI1B P2A-Flag APC 1.6 3

GFI1B CDC73 Wnt pathway regulators PP2A 4.9 1 RCOR1* Histone modifying enzymes GFI1B muWRD-Flag CCAR2 CtBP1 45 8 RCOR2 MS Δ Spindlin1 20 5 BRG1 GFI1B Zinc-Flag 31 UBR5 2 2 RCOR3* 0 50 GFI1B ΔInterm-Flag Ig CDC73 20 7 CCAR2 18 9 AvgSpec MS BRG1 8 5 Relative abundance 22 SNAG domain WRD domain HDAC1* 45 29 BFDR Zinc finger Mutation Flag * Known GFI1B binding proteins ≤ 0.01 ≤ 0.05 > 0.05

d –+– –GFI1B-Flag e –+– –GFI1B-Flag f IP g ––+ –GFI1B P2A-Flag – – + – GFI1B ΔInterm-Flag IgG GFI1B IP kDa Δ β-catenin ––– +GFI1B muWRD-Flag ––– +GFI1B Zinc-Flag 90 IgG GFI1B IP flag IP flag (low exposure) kDa Pontin 52 β-catenin kDa 90 50 (low exposure) kDa (high exposure) β-catenin 100 90 β-catenin (Ser675P) Pontin 52 100 LSD1 50 (high exposure)

80 37 TLE3 37 GFI1B 100 LSD1 60 GFI1B (N-terminal)

Input 20 50 Pontin 52

130 TLE1 Input 50 Pontin 52 100 LSD1 90 β-catenin Input 50 Pontin 52 37 h IP Input GFI1B IgG GFI1B IgG 300 CHD8 GFI1B-Flag kDa 20 100 β-catenin

50 CtBP1

37 GFI1B

100 LSD1 100 LSD1 Input –+––++ CHIR99021

37 GFI1B-Flag

Fig. 2 GFI1B interacts with β-catenin and several regulators of the Wnt/β-catenin signaling pathway. a Immune precipitation (IP) with anti Flag antibody from 293T cells overexpressing GFI1B-Flag followed by mass spectrometric analysis. b Dot Plot showing BioID interactions of GFI1B-BirA*-Flag with the indicated Histone modifying enzymes and Wnt pathway regulators in Flp-In T-REx HEK293 cells. Node color represents the average spectral counts. The edge color depicts the confidence score of the interaction (BFDR ≤ 1% as high confidence, 1% < BFDR ≤ 5% as medium confidence or 5% < BFDR as low confidence score). The relative abundance of prey across the bait is represented by the size of the circle. c A previously described Wnt regulatory domain (WRD) is present in GFI1B28. The indicated Flag-tagged and mutated forms used for IP. d, e Immune precipitation with anti Flag antibodies from 293T cells overexpressing WT or the indicated mutated forms of GFI1B followed by western blot. f–h Immune precipitation with an anti GFI1B antibody from K562 (f), HEL (g), and CHIR99021 treated K562 (h) cells dependent transcription. This GFI1B activity was reduced when its β-catenin regulators that interact with β-catenin through these SNAG (GFI1BP2A or GFI1BΔSNAG), WRD (GFI1BΔWRD), domains. In agreement with this, GFI1B was more active in intermediate or zinc-finger domains were disrupted, either in the combination with LSD1 than alone (Fig. 4g), suggesting that the presence of LiCl or CHIR99021 (Fig. 4c, f). We observed the same GFI1B/LSD1 complex can be an activator of β-catenin/TCF reduction in GFI1B activity using single (GFI1BF106A or GFI1- mediated transcription in this system. The repressive GFI1B BL111A) or double mutations in the WRD sequence (GFI1B- activity on its own promoter was not affected by WRD sequence muWRD, depicted in Fig. 2c) (Fig. 4d, e). This indicates that the deletion or CHIR99021 treatment (Supplementary Fig. 2b). Fur- observed effect of GFI1B on β-catenin/TCF mediated transcription thermore, we observed that GFI1B was able to partially reverse the depends on its ability to bind LSD1 or CtBP1 and/or other CtBP1 and TLE1 inhibitory effect on β-catenin/TCF-dependent

4 NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z ARTICLE

a b c GST-Ecad − + − GFI1B − + − GFI1B pull down − − + GFI1B P2A − − + GFI1B ΔSNAG U2OS-TetOn-GFI1B IP β-catenin IP β-catenin −+Dox kDa β-catenin kDa kDa 90 37 GFI1B 37 GFI1B 25 25 37 GFI1B

LSD1 100 100 LSD1

100 LSD1 Input Input

β β-catenin Input 90 -catenin 90

37 GFI1B 37 GFI1B 37 GFI1B

25 25 100 LSD1

90 β-catenin

de1 + EV Biological process Molecular function 2 + GFI1B 1234 1234 3 + GFI1B ΔSNAG Cellular localization Catalytic activity, acting on RNA 4 + GFI1B ΔWRD Intracellular receptor signaling pathway Enzyme regulator activity β-catenin-BIRA Tricarboxylic acid metabolic process Damaged DNA binding 1234 mRNA metabolic process ATPase activity, coupled GFI1B Regulation of Wnt signaling pathway RNA binding GSK3B RNA transport Structural constituent of cytoskeleton BCL9L Negative regulation of ubiquitin-protein transferase activity Helicase activity CTNNBIP1 Regulation of canonical Wnt signaling pathway Transcription coactivator activity CREBBP Protein polyubiquitination Translation initiation factor activity EP300 Nucleobase-containing compound metabolic process ADP binding DVL2 Macromolecule localization Ligand-dependent nuclear receptor binding TP53 Wnt pathway regulator Positive regulation of canonical Wnt signaling pathway Nuclear hormone receptor binding Protein sumoylation Ubiquitin protein ligase binding 050 AvgSpec Wnt signaling pathway, planar cell polarity pathway SUMO activating enzyme activity Relative abundance Stress-activated MAPK cascade Ligase activity, forming carbon-nitrogen bonds BFDR 0 25 040 ≤ 0.01 ≤ 0.05 > 0.05 –log10 corrected P value

Fig. 3 GFI1B recruits LSD1 to β-Catenin containing complexes. a, b Western blot analysis of indicated proteins after immune precipitation of endogenous β-catenin from 293T cells transfected with the indicated GFI1B constructs. c GST-E-Cadherin pull down of endogenous β-catenin in U2OS cells expressing inducible GFI1B (Tet-ON system) followed by Western blot analysis. d Dot Plot showing BioID interactions of β-catenin-BirA*-Flag coexpressing an Empty Vector (1), the wild-type form of GFI1B (2), GFI1B lacking the SNAG domain (3) or GFI1B lacking the WRD domain (4) with the indicated Wnt pathway regulators in Flp-In T-REx HEK293 cells. e Heat map illustrating biological processes and molecular functions of associated GO terms. The enrichment score of each GO term is shown as the −log10 of corrected P values, indicated by different color intensities transcription (Fig. 4h). Similarly, inducible GFI1B expression which abolishes DNA-binding of GFI1B, did not show any under doxycycline in U2OS cells reversed the effect of TLE1 in the inhibition in this modified TOP reporter assay but rather TOP/FOP reporter assay (Fig. 4i). This suggests that GFI1B can activation similar to WT GFI1B (Fig. 4j). GFI1B N290S mutant also activate β-catenin/TCF transcription by counteracting the in the unmodified TOP assay could still associate with β-catenin effect of β-catenin inhibitors. and activate TCF/LEF mediated transcription to the same extent Neither TOP nor FOP constructs contain a known GFI1B as WT GFI1B (Fig. 4k, l), indicating that the ability of GFI1B to consensus-binding motif (AATC). This suggested that GFI1B activate TCF-dependent transcription can be independent of activates TCF/LEF-dependent transcription without binding its ability to bind DNA and by inference also its function as a DNA directly. To test this, we generated TOP reporter genes transcriptional repressor. with GFI1B binding motifs (AAATCTCTGCA) before and after the TCF/LEF binding sites and observed that GFI1B was now able Gfi1b KO deregulates expression of Wnt target genes. To test to inhibit transcription of this modified TOP reporter (Fig. 4j). whether this GFI1B activity can also be observed in vivo, we used This inhibition was dependent on GFI1B DNA-binding capacity a well-established Axin2LacZ knockin reporter mouse, which since a N290S mutation of GFI1B’s fifth zinc-finger domain, enables monitoring of TCF/LEF dependent transcription by

NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z

TOP 7 TCFs LUC

FOP 6 muted TCFs LUC abcLiCl *** *** *** 300 *** 300 *** 500 Wnt3A treated *** Untreated 200 200 400

300 100 100

TOP/FOP n.s. TOP/FOP 200

10 10 TOP/FOP *** 5 5 *** 100

0 0 0

WRD pCMV GFI1B ΔSNAG pCMV GFI1B ΔSNAG pCMV GFI1B Wnt3A pCMV GFI1B Δ β-catenin β-catenin GFI1B P2A GFI1B P2A GFI1B P2A GFI1B GFI1B GFI1B/Wnt3A GFI1B GFI1B/

d e LiCl f n.s. *** CHIR99021 * * 8 * 60 *** *** 200

6 150 40

4 100 TOP/FOP TOP/FOP TOP/FOP 20 2 50

0 0 0

ΔZinc Interm pCMV GFI1B pCMV GFI1B pCMV GFI1B ΔSNAG Δ GFI1B P2A GFI1B P2A GFI1B GFI1B L111AGFI1B F106A GFI1B L111AGFI1B F106A GFI1B muWRD GFI1B muWRD GFI1B GFI1B

g LiCl h i *** LiCl LiCl *** ** 100 *** 200 *** 5 *

80 4 150 *** n.s. *** 60 3 100 40 2 TOP/FOP TOP/FOP TOP/FOP 50 20 1

0 0 0 Dox 0 μg/ml Dox 5 μg/ml LSD1 TLE1 pCMV GFI1B pCMV GFI1B CtBP1 TLE1 – + – +

GFI1B/LSD1 GFI1B – – + + GFI1B/TLE1GFI1B/CtBP1

jklTOP with GFi1B binding motifs TOP –+– – GFI1B-Flag GFI1B 7 TCFs GFI1B LUC 7 TCFs LUC ––+ – GFI1B N290S-Flag Δ LiCl LiCl ––– + GFI1B Zinc-Flag IP flag *** *** kDa 10 n.s. 3 *** 100 β-catenin 8 5 5 2 6 100 LSD1 4 RLU × 10 RLU × 10 1 2 Input

0 0 37

pCMV GFI1B pCMV GFI1B GFI1B-Flag

GFI1B N290S GFI1B N290S

20

100 β-catenin

measuring β-galactosidase (LacZ) activity in mouse primary three cell subsets and was statistically significant in MKs (Fig. 5c). cells (Fig. 5a, Supplementary Fig. 2d)35. Compared to controls, To validate this, we used a second in vivo reporter system, the β-galactosidase (LacZ) activity was reduced in HSCs and MKs TCF/LEF:H2B-GFP transgene reporter mice, in which six copies with Gfi1b deletion (Fig. 5b). Quantification of mean fluorescence of a TCF/LEF responsive element are placed directly 5′ of a his- intensities (MFI) of LacZ in these cells showed the effect of Gfi1b tone H2B-GFP fusion gene36. Similar to the Axin2LacZ reporter deficiency on the Axin2 promoter activity was measurable in all system, a reduction of GFP intensities was observed in HSCs and

6 NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z ARTICLE

Fig. 4 GFI1B enhances transcription of a TCF dependent promoter/reporter system. a, b TOP/FOP flash reporter assay in 293T cells. Cells were transfected with the indicated plasmids together with TOP or FOP reporter constructs 36 h before luciferase measurement. GFI1B enhances the TOP/FOP ratio at both basal levels and following activation with a 10% Wnt3A-conditioned media or b co-transfection with active form of β-catenin and/or Wnt3A expressing vectors. c–e TOP/FOP flash reporter assay in 293T cells transfected with WT Gfi1b and the indicated mutated forms. Cells were treated with 25 mM LiCl (to activate canonical Wnt signaling in c and e) for 5 h before luciferase measurement. f TOP/FOP flash reporter assay in 293T cells treated with CHIR99021, a specific GSK3β inhibitor for 5 h before luciferase measurement. g, h TOP/FOP flash reporter assay in 293T cells and GFI1B-TetON-U2OS cells transfected with the indicated vectors. i TOP/FOP flash reporter assay in U2OS cells stably expressing a doxycycline inducible GFI1B (Tet-ON system) transfected or not with a construct for TLE in the presence or absence of doxycycline. j, k TOP and modified TOP (flanked by GFI1B binding motifs) reporter assay in 293T cells. l Immune precipitation with anti Flag antibodies from 293T cells overexpressing WT or the indicated mutated forms of GFI1B followed by western blot. (*p < 0.05, **p < 0.005, ***p < 0.0001 on a Welch corrected t-test, error bars show s.d, n = 3 biologically independent samples for each data point)

a b Rosa-cre tg Rosa-cre tg c Tamoxifen Analysis Gfi1bwt/flox Gfi1bflox/flox d0 d1 d10 Gfi1b WT Axin+/LacZ 3000 Axin+/+ Gfi1b KO 2000 HSCs + + HSCs CD41 /CD9 + + Rosa-cre tg 1000 CD41 /CD9 Gfi1bWT/flox or Gfi1bflox/flox +/LacZ Axin II 0

d Tamoxifen Analysis 2000 d0 d1 d10 1500 HSCs HSCs low low CD41 /CD9 1000 CD41low/CD9low

Rosa-cre tg 500 Gfi1bWT/flox or Gfi1bflox/flox TCF-H2B-GFP 0

Gfi1b WT 100 5000 e 250 Gfi1b KO 80 4000 p=0.03 200 p=0.02 60 3000 MK 150 40 2000 MK 20

100 Normalized to mode 1000 0 50 0 3 3 4 5 3 3 4 5 FDG (Lac-Z) MFI (corrected to background)normalized mode

TCF-H2B GFP (MFI) –10 0 10 10 10 –10 0 10 10 10 0 FDG (LacZ) HSCs HSCs MK CD41+/CD9+ CD41low/CD9low

Fig. 5 Gfi1b is required for the expression of Axin II- or TCF-dependent reporter alleles in HSCs and MKs in vivo (a–c) Axin2+/LacZ Wnt-reporter mice were used. Littermates not carrying the reporter transgene (Axin2+/+) were used to correct for endogenous β-galactosidase activity. a Mice were analysed by FACS at day 10 following the first injection of tamoxifen. b FDG (LacZ staining) histograms of indicated FACS gated cells. A representative result from three independent experiments is shown. c Quantification of the mean fluorescence intensity (MFI) measured in the FDG channel (LacZ) normalized to endogenous β-galactosidase activity. Paired t-test was used for the analysis. (n = 5 mice per group) (d, e) In vivo measurement of canonical Wnt signaling activity in Gfi1b WT/KO MKs and HSCs by FACS using TCF-H2B-GFP transgene reporter mice. A Mann–Whitney U non-parametric test was used for the analysis. Boxplot Centre line shows median and bounds of box and whiskers show interquartile ranges of GFP MFI, n = 5 GFI1b WT and five GFI1B KO mice

MKs following Gfi1b deletion, again most significantly in MKs (Fig. 6d, Supplementary Table 2, Supplementary Fig. 3). To (Fig. 5d, e). These data indicate that Gfi1b absence leads to a obtain this information, we interrogated published ChIP-seq transcriptional down-regulation of β-catenin driven target genes datasets from the mouse embryonic stem cell (ESC)-derived in vivo. hematopoietic precursor cell line HPC-7 for Gfi1b38,mouseESCs RNA-seq analysis comparing flow-sorted Gfi1b WT/KO CD41 expressing a tagged-β-catenin39 and with wt ESCs for LSD140. +/CD9+ HSCs, MKs, and CD41low/CD9lowHSCs showed that We found 523 genes co-occupied by Gfi1b and β-catenin and many genes, which were differentially expressed between WT 133 genes co-occupied by all three factors at promoters (Supple- and Gfi1b-deficient cells, belong to previously published sets mentary Fig. 4a, Supplementary Table 3). Among them were known of genes25 involved in Wnt/β-catenin signaling (Fig. 6a, b). A canonical Wnt genes that were down-regulated in Gfi1b-null set of genes that were described to be up-regulated following cells such as Ccnd1 (regulator of cell cycle progression)41, Egr1 activation of canonical Wnt signaling37 was enriched in Gfi1b (transcription factor)42 and Ptgs2 (enzyme in prostaglandin WT CD41+/CD9+ HSCs and CD41low/CD9lowHSCs and a set of biosynthesis)43 (Supplementary Figs 3 and 4b, Fig. 6d). However, genes down-regulated upon activation of canonical Wnt signal- key non-canonical Wnt genes such as Ptk7 (Wnt co-receptor)22, ing, was enriched in Gfi1b KO MKs (Fig. 6c). Nfat521, Zfp467 (regulator of the canonical Wnt inhibitor SOST)44, Next, we identified genes that were expressed over 1.5 fold-up and Rock2 (serine/threonine kinase downstream of PTK7)45 were or 0.6 fold-down in MKs, HSCs and CD41−/CD9− HSCs and up-regulated in Gfi1b-deficient MKs (Supplementary Table 2, were also co-occupied by Gfi1b and β-catenin or also by LSD1 Supplementary Figs. 3 and 4b).

NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z

ab Gene setTotal Differentially expressed Tamoxifen

d0 d1 NUSSE TARGETS UP 92 29

LABBE_WNT3A_TARGETS_UP 112 58

10 days LABBE_WNT3A_TARGETS_DN 97 31 BM lineage depletion WILLERT_WNT_SIGNALING 24 9

Rosa-cre tg KOLLIGS TARGETS 64 28 Gfi1bWT/flox (n=10/sorting) FACS sorting GO:0060070, CANONICAL WNT SIGNALING PATHWAY 89 30 or Gfi1bflox/flox (n=5/sorting) GO:0016055, WNT SIGNALING PATHWAY 223 72 KEGG_WNT_SIGNALING_PATHWAY 151 44 LK CD41+/CD9+ LSK CD150+/CD48– (MK) (HSCs) Canonical Wnt target genes 157 94 Non-canonical Wnt target genes 162 132 HSCs CD41low/CD9low

d 20

RNA-Seq Zfp467

c 15 Snora16a Enrichment plot: LABBE_WNT3A_TARGETS_UP Enrichment plot: LABBE_WNT3A_TARGETS_UP Ptk7 Nrep 0.00 0.1 –0.05 –0.10 0.0 HSCs HSCs + + –0.15 low low CD41 /CD9 –0.20 CD41 /CD9 –0.1 –0.25 –0.2 –0.30 10 Irf2bpl –0.35 –0.3 –0.40 Glul Enrichment score (ES) Enrichment score (ES) –0.45 –0.4 –0.50 Tgif1 Lpp Chst15 Rbks 7.5 Ubl3 10.0 ‘KO’ (positively correlated) ‘CD9hiCD41low_HSC_KO’ (positively correlated) 5 7.5 5.0 Ccng2 B3gnt1 Zfp13 5.0 2.5 Dhrs7 Irf2bpl Bbip1 2.5 0.0 Nfat5 Kazald1 Zfp13 0.0 Zero cross at 7007 Zero cross at 6758 Qser1 –2.5 Zfp503 –2.5 NES= –1.60 –5.0 NES= –1.45 –5.0 –7.5 p Value= 0.026 –7.5 p Value= 0.006 ‘WT’ (negatively correlated) ‘CD9hiCD41low_HSC_WT’ (negatively correlated) 1.5 0 2500 5000 7500 10000 12,500 0 2500 5000 7500 10000 12,500 0.6 Rank in ordered dataset list metric (Diff_of_Classes) Ranked Rank in ordered dataset Ranked list metric (Diff_of_Classes) Ranked Enrichment profile Hits Ranking metric scores Enrichment profile Hits Ranking metric scores

Enrichment plot: LABBE_WNT3A_TARGETS_DN 0.5

0.4 0.3 MK Expression fold (Gfi1b KO/WT) Expression fold Pdss1 0.2 0.4 Celsr3 Nme1 Ier2 Kazald1 0.1 Hmga1 Dnph1

Enrichment score (ES) 0.0 Hmga1-rs1 Rnu11 –0.1 Egr1 Dusp6 Cd244

10.0 Bhlhe40 7.5 ‘ROSA_MK_KO’ (positively correlated) Ccnd1 5.0 0.2 Lphn1 2.5 0.0 Zero cross at 6490 Rnu11 –2.5 –5.0 NES= –1.43 –7.5 p Value= 0.048 ‘ROSA_MK_WT’ (negatively correlated) Atf3 –10.5 Ptgs2 0 2500 5000 7500 10000 12,500

Ranked list metric (Diff_of_Classes) Ranked Rank in ordered dataset Enrichment profile Hits Ranking metric scores

MK HSCs HSCs CD41low/CD9low

e 50 f Gfi1b 0.05 CCND1 0.025 NFAT5 0.4 BIRC5 37 ** 0.04 ** 0.020 0.3 0.03 0.015 0.2 * * 0.02 * Tubulin CT G6PD 0.010 * 50 ΔΔ ** ** 0.1 0.01 ** *** 0.005 *** 0.00 0.000 0.0 C1 C5 Sh11 Sh23 C1 C5 Sh11 Sh23 C1 C5 Sh11 Sh23 C1 C5 Sh11 Sh23 g 0.025 PTK7 0.10 ROCK2 * CCND1 NFAT5 *** 15 50 ** * * * 0.020 ** 0.08 *** 40 * * 10 0.015 0.06 30

CT G6PD 0.010 0.04 20 * 5 * * ΔΔ * * * 0.005 0.02 10 Fold enrichment Fold 0.000 0.00 0 0 C1 C5 Sh11 Sh23 C1 C5 Sh11 Sh23 C1 C5 Sh11 Sh23 C1 C5 Sh11 Sh23

Fig. 6 Deregulation of Wnt/β-catenin target gene expression in Gfi1b-deficient MKs and HSCs. a Gene expression analysis was done on sorted Gfi1b WT/ KO MKs, HSCs and HSCs CD41low/CD9low as indicated. b, c Gene set enrichment analysis (GSEA) of RNA-seq data from the sorted cells using published Wnt related gene sets (see Supplementary Table 2 for gene set references). d RNA-seq expression analysis of Gfi1b/β-catenin co-occupied genes (based on published ChIP-seq data) in the three Gfi1b WT/KO populations. Genes indicated in red also have peaks for LSD1 at their promoter. e GFI1B expression in two independent K562-derived cell lines stably expressing a scrambled control shRNA (C1, C5) and two lines expressing a GFI1B specific shRNA (Sh11, Sh23). f Expression analysis of Wnt target genes by RT-PCR in GFI1B shRNA KD and scramble shRNA stable K562 clones. g Enrichment of acetylated H3K9 at GFI1B/β-catenin target gene promoters by ChIP-qPCR in GFI1B shRNA KD and scramble shRNA stable K562 clones. (*P < 0.05, **P < 0.001, ***P < 0.0001 on a Welch corrected t-test compared to C1 (in blue) or C5 (in red), error bars show s.d, n = 3 technical replicates)

8 NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z ARTICLE

To validate our findings further, we generated a GFI1B did not alter the majority of GFI1B and LSD1 targeted promoters knockdown using shRNA in the human erythroleukemia cell and enhancers (Supplementary Fig. 7a, b). β-catenin bound most line K562 (Fig. 6e). We confirmed reduced GFI1B levels frequently to promoters and enhancers where GFI1B and LSD1 decreased the expression of CCND1 and BIRC5 (canonical Wnt are already present prior treatment (Supplementary Fig. 7a, b). genes) and increased the expression of NFAT5, ROCK2 and PTK7 To assess whether recruitment of β-catenin is GFI1B genes (non-canonical Wnt genes) (Fig. 6f), similar to that dependent, we carried out both GFI1B and β-catenin ChIP- observed in cells from Gfi1b-deficient mice (RNA-seq data qPCRs on CHIR99021 treated K562 cells with WT (K562-C5) Fig. 6d, Supplementary Fig 3, Supplementary Table 2). Further- and KD (K562-S23) levels of GFI1B. Levels of GFI1B enrichment more, the regulation of CCND1 or NFAT5 expression in GFI1B confirmed the decrease in GFI1B protein at target genes within knockdown K562 cells correlated with lower or higher H3K9 GFI1B KD K562 cells (Fig. 8d). Importantly, depletion of GFI1B acetylation at their promoters, respectively (Fig. 6g). This suggests lead to a significant decrease in β-catenin enrichment at target that the expression changes induced by GFI1B correlate with the loci, such as SLC38A8 and YAF2, indicating that β-catenin appropriate H3K9 acetylation changes at promoters of target recruitment is GFI1B-dependent at these promoters (Fig. 8d). genes. Other loci, like AXIN2 and SP5, did not show significant changes Next, we performed RNA-seq analysis on two sets of Gfi1b WT in β-catenin levels with GFI1B depletion, indicating that β- and KO MKs obtained by cultivating lin-BM cells in vitro with catenin may also bind to specific loci independently of GFI1B in TPO and SCF for 7 days (Fig. 7a). As with the previous RNA-Seq, K562 cells. many genes that were differentially expressed between WT and Since LSD1 is a histone demethylase, specifically for histone 3, KO samples were part of previously published genes sets involved lysine 4, and 9 (H3K4 and H3K9), we assessed H3K4me1 and in Wnt/β-catenin signaling (Fig. 7b, c)25. Especially, two H3K9me2 levels by ChIP-seq in K562 cells before and after canonical Wnt gene sets were significantly down-regulated in Wnt3a treatment. The majority of genes have high H3K4me1 Gfi1b KO MK samples (Nusse Targets up and Labbe Targets up, enrichment if targeted by one or more of β-catenin, GFI1B and Fig. 7d). Interestingly, unsupervised clustering of canonical or LSD1 upon treatment (60–90%). Only when none of the three non-canonical Wnt target genes clearly separated Gfi1b KO and factors were detected at promoters did similar proportions of WT MKs in both biological replicates (Fig. 7e). These findings genes experience increases or decreases in H3K4me1 levels confirmed that Gfi1b loss in HSCs or MKs can lead to both up- (Supplementary Fig. 7c). Furthermore, promoters targeted by one and down-regulation of Wnt target genes. or more of the three factors mostly had a distinct decrease in H3K9me2 enrichment (60–100%). This indicates that upon Wnt stimulation chromatin priming may occur with a general increase β-catenin co-occupies regions targeted by Gfi1b and LSD1. The in H3K4me1 and decrease in H3K9me2 at promoter targets of analysis of published ChIP-seq datasets from HPC7 cells for one or more of the three factors and not due to LSD1 alone. Gfi1b and knock-in ES cells for β-catenin38–40,46 showed that Gfi1b and β-catenin co-occupy several loci within 1 kb of a Genomic distribution of β-catenin, GFI1B and LSD1 peaks. promoter (Supplementary Fig. 5a and b). We next examined the GFI1B associates mostly at promoter regions (~48%), while LSD1 genomic features found closest to a binding site for one factor as a was located primarily within intergenic (~37%) or intronic function of its proximity to a binding site of the other factor using (~41%) regions (Supplementary Fig. 7d). This genomic dis- these datasets and a CTCF ChIP-seq dataset46 as a control, which tribution did not significantly change upon Wnt3A stimulation. showed no relation with features associated with Gfi1b and only a β-catenin was bound within intergenic regions mostly (~61.5%), limited relation for β-catenin. When Gfi1b peaks overlap a β- however, upon treatment, the most notable shift appears as a catenin peak, they are more frequently present at transcription 55.7% decrease in the proportion of β-catenin peaks at promoters. start sites (TSS) and conversely, β-catenin peaks are more likely The genomic distribution for one factor as a function of its found at a TSS when they overlap a Gfi1b peak, as compared to proximity to a binding site of another factor showed that β- peaks with more than 1 kb distance to the other factor (Supple- catenin peaks overlapping with a GFI1B or LSD1 peak shift from mentary Fig. 5a–c). A comparable pattern was found for both predominantly promoter (61 to 24.5%) to intergenic regions Gfi1b and β-catenin peaks relative to their distance to LSD1 (26.5–50%) upon Wnt3A activation (Fig. 8e). Analysis of estab- binding sites (Supplementary Fig. 5b and c). lished enhancer regions showed there was a strong increase in the To confirm a co-occupation of GFI1B, β-catenin, and LSD1 at number of enhancers co-bound by β-catenin, GFI1B, and LSD1 specific genomic sites at endogenous expression levels in the same following Wnt activation (Fig. 8f–h and Supplementary Table 5). cells, we performed ChIP-seq analysis with K562 cells treated or Similarly to promoters, a shift was seen mostly with β-catenin not with Wnt3A for 4 h to obtain the highest levels of active, localization rather than GFI1B or LSD1 at enhancers with Wnt3a nuclear β-catenin (Supplementary Fig. 6). β-catenin peaks treatment (Supplementary Fig. 7b). showed a statistically significant overlap with GFI1B and LSD1 peaks at promoters (Fig. 8a, Supplementary Fig. 7a and Supplementary Table 4) and also that many more genes are Increased Wnt/βcatenin signaling rescues Gfi1b KO pheno- occupied by the tripartite complex (GFI1B/β-catenin/LSD1) than type. To test whether Gfi1b null phenotypes can be rescued by by LSD1/β-catenin or by GFI1B/β-catenin (Fig. 8a). AXIN2 and reactivating Wnt/β-catenin signaling pathway in primary HSCs YAF2 are examples of gene promoters specifically targeted by β- and MKs, we explanted bone marrow cells from Rosa Cre-ER, catenin, GFI1B, and LSD1 upon Wnt3a treatment (Fig. 8b). Gfi1bfl/fl mice or controls after injection of Tamoxifen to induce Motif analysis revealed, as expected, that sites bound by GFI1B Gfi1b deletion, followed by a lineage depletion. The cells were put were significantly enriched for the GFI1B binding motif as well as in culture and treated with recombinant Wnt3A (a canonical Wnt GATA binding motifs genome-wide (Fig. 8c). However, ligand) (Fig. 9a). After 7 days in culture, non-treated HSC and sequences at sites co-occupied by β-catenin and GFI1B were MK samples from Gfi1b KO mice expanded as expected several found to be prominently enriched for the LEF1 motif as well as fold over WT controls (Fig. 9b). However, treatment with Wnt3A GATA factor motifs, but not for GFI1B’s own binding motif inhibited this expansion in a dose dependent manner, most sig- (Fig. 8c). This suggests that GFI1B has a different DNA binding nificantly in MKs (Fig. 9b). To validate this hypothesis further, we activity specifically at β-catenin sites. Notably, Wnt stimulation infected isolated CD45.2+ Gfi1b-deficient lin-BM cells with

NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z

ab<–2.0 fold >2.0 fold GFI1b/β-CATENIN/LSD1

Tamoxifen GFI1b/β-CATENIN

d0 d1 GFI1b/LSD1 MK in vitro culture

40 CD33 10 days BM Lineage depletion

Rosa-cre tg Cass4 Gfi1bWT/flox or Gfi1bflox/flox Culture with TPO & SCF value)

7 days P 20 RNA-Seq BSA gradient separation Adcy9 – log 10 (

Adk Zwint Pttg1ip Tia1 Crtc1 Arhgef37 Il4i1 Axin2 Snx16 Fbxo31 Sdr39u1 Ccdc92 P = 0.05 c 0

Gene setTotal Differentially expressed –4 –2 024 Log 2 ( WT/KO fold change) NUSSE TARGETS UP 92 42

LABBE_WNT3A_TARGETS_UP 112 89 d Gfi1b WT/KO MKs in culture Enrichment plot: NUSSE targets up Enrichment plot: LABBE_WNT3A_targets_up

0.10 LABBE_WNT3A_TARGETS_DN 97 54 0.1 0.05 0.0 0.00 –0.05 WILLERT_WNT_SIGNALING 24 16 –0.1 –0.10 –0.2 –0.15 –0.20 KOLLIGS TARGETS 64 49 –0.3 –0.25 –0.30

Enrichment score (ES) –0.4 Enrichment score (ES) –0.35 –0.5 –0.40 GO:0060070, CANONICAL WNT SIGNALING PATHWAY 89 53 –0.45

GO:0016055, WNT SIGNALING PATHWAY 223 151

3 'KO' (positively correlated) 3 'KO' (positively correlated) KEGG_WNT_SIGNALING_PATHWAY 2 2 151 101 1 1

0 Zero cross at 7889 0 Zero cross at 7889 –1 –1 Canonical Wnt target genes 157 106 NES= –1.53 NES= –1.38 –2 p Value= 0.019 –2 p Value= 0.028 –3 'WT' (negatively correlated) –3 'WT' (negatively correlated) 0 2500 5000 7500 10,000 12,500 15,000 0 2500 5000 7500 10,000 12,500 15,000 Ranked list metric (Diff_of_Classes) Ranked Non-canonical Wnt target genes 162 128 list metric (Diff_of_Classes) Ranked Rank in ordered dataset Rank in ordered dataset

Enrichment profile Hits Ranking metric scores Enrichment profile Hits Ranking metric scores e

Nfat5 Stk17b Fscn1 Add1 Rarg Znrf2 Gfi1b Mpp7 Hspa13 Npnt Ncam1 Iqgap1 Nrcam Cant1 Rplp0 Dazap2 Notch2 Laptm4a Met Ralgapa2 Itgav WT Dlk1 Rnf2 Tbp Usp37 Fst Mzt1 1 Cdh1 Ppp1r14c Anp32e Tiam1 Slc38a7 Jag1 Trmt1 KO Klf5 Wdr55 Vegfa Abi2 Lamc2 Nipa1 Prkaa1 Mmp2 Slc11a2 Jun Ccdc69 Mycbp Tmem55a Nedd9 Adam22 0.5 Plaur Otud1 Hnrnpk Runx2 Atp6v1c1 Cebpd Myc Ptpra Myc Lamp2 Ccnd1 Tmem30a Ccnd2 Chsy1 Hadh Egr1 Ap4e1 Tcf7 Hs6st2 Cdon Ccnd1 Cep97 Ets2 Rcn2 0 Rps3 Cyr61 Pisd Mycn Anxa6 Birc5 Tmem62 Ahr Pfn2 Birc5 Tsnax Adam10 Rundc3b Bcl2l1 Dag1 Gbx2 Fech Six1 Srprb Actb Clptm1l Ddx41 −0.5 Mmp9 Tbrg1 Cldn1 Cdk14 Cdkn2a B9d1 Tgfb3 Enpp4 Inhba Ak3 Gapdh Slc25a42 Nt5c Ptgs2 Pros1 Mycbp2 Hectd3 Fn1 Tmem56 Ppard B4galt4 Hsd17b11 −1 Angptl4 Aip Sgk1 Ndrg2 Cdc25a Abhd2 B2m Agps Dnmt1 Fam129a Ddit3 Fosl1 Vapa Suz12 Rrs1 Itfg2 Tmco3 Lef1 Cfl2 Stra6l Xpr1 Snx13 Smc3 Lin7c Smo Fktn Gja1 Pbrm1 Myb Ints6 Dkkl1 Ccdc88a Slc44a1 Bcl2l2 Sra1 Fzd7 Gk5 Nos2 Hyal3 Lrp1 Cnnm2 Srsf3 Gapvd1 Lcor Mmp14 Usp47 Tcf4 Ppp6r3 Gusb Mapk14 Efnb1 Pcyt2 Ube4a Fgf1 Htatip2 Yap1 Lpgat1 Wisp1 Trove2 Irs1 Esd Tcf7l2 Rnf111 Nfe2l2 Fzd3 Rsbn1l Il6 Gng12 Axin2 Mapk3 Nrp1 Axin2 Tmbim6 Mitf Fam160b1 Sh2b3 Hes1 Chac1 Ptch1 Ncoa2 Tert Kdm6b Bmi1 Adam10 Lbh Rab11fip1 Plekhf2 Id2 Plod2 Antxr1 Aldh9a1 Dab2 Cpd Plau Kdm3b Lman2 Igf1 Hmox1 Pttg1 Klhl5 Cd44 Kctd3 Btrc Wnk1 Wnt10b Camk4 Galnt1 Ephb2 Usp2 Tle1 Cdca4 L1cam Tle1 Cdc37 Enc1 Ska2 Bambi Dnajc11 Mkl2 S100a6 Rad23b S100a4 Rars WT1 WT2KO2 KO1 KO2 KO1 WT1 WT2 Canonical Wnt target genes Non-canonical Wnt target genes

Fig. 7 Canonical and non-canonical Wnt target genes are deregulated in Gfi1b KO MK in vitro. a Gfi1b WT/KO mice were taken to obtain lineage depleted BM cells, n = 2 GFI1b WT and GFI1B KO mice. Cells were taken into culture in the presence of TPO and SCF for 7 days and MK were separated using BSA gradient followed by RNA-seq analysis. b Volcano plot of expression fold change (KO/WT) over the P value of differences between WT and KO. Vertical dashed lines show the threshold of two-fold change in expression in KO samples compared to WT. The color code shows the co-occupation of promotor by indicated factors as seen in ChIP-seq data from K562 cells. c, d Gene set enrichment analysis (GSEA) of RNA-seq data from the sorted cells using published Wnt related gene sets and two new sets of Wnt target genes pooled from published data. e Unsupervised clustering of RNA-seq data using indicated gene sets retroviral vectors directing the expression of the constitutively Wnt signaling pathway by retroviral expression of active β-cate- active form of β-catenin or GFP as a control. These cells were nin, inhibits in vivo expansion of Gfi1b KO HSCs. Finally, we transplanted into irradiated CD45.1 recipient mice in order to follow up on the findings that Gfi1b loss leads to impaired MKs study the impact of canonical Wnt signaling activation in Gfi1b spreading on fibronectin coated matrix and treat Gfi1b null MKs KO cells. Flow cytometric analysis, 4 months post transplanta- with Wnt3A in a dose dependent manner. The effects are tion, showed that Gfi1b-deficient β-catenin+ HSCs frequencies quantified by measuring the RI and find that Gfi1b null MKs were significantly lower than those of GFP controls (Supple- became less round, start spreading and resemble WT MKs when mentary Fig. 8a and b) suggesting that activating the canonical treated with Wnt3A (Fig. 9c, d). Conversely, treatment of Gfi1b

10 NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z ARTICLE

a 5 kb 5 kb Bound promoters Bound promoters b 5725 – Wnt3A + Wnt3A

LSD1 -CATENIN β 11,659 Gfi1b 5856 7378 5617

624 9029 1618 5920 GFi1B

108 202 + Wnt3a 27 35 14 1 LSD1 17 15 143 β-catenin P = 2.98e–192 P = 1.92e–99 122 172 H3K4me1

c 2.0 1.5

GFI1B 1.0 Bits H3K9me2 0.5 P =1.16e–130 AXIN2 YAF2 0.0 1234567891011 Gfi1b 2.0 GATA3 1.5 GATA3 d 8 Gfi1b 1.0 Gfi1b ChIP K562-C5 Bits IgG P = 1.96e–88 0.5 P = 3.62e–12 Gfi1b β-catenin 0.0 Gfi1b 12345678 K562-S23 6 IgG

2.0

GATA1 1.5 GATA1

1.0 P = 3.43e–88 Bits P = 1.48e–11 0.5 4

0.0 1234567891011

2.0

1.5 2

LEF1 Relative enrichment (input) 1.0 Bits

0.5 P = 1.35e–05

0.0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 SLC38A8 Ccrl2 Yaf2 Cdh1 Axin2 Sp5 Sp8 100 90 β-catenin ChIP β 80 -catenin e Distance from Gfi1b peaks Distance from LSD1 peaks K562-C5 70 IgG 0 kb >1 kb 0 kb >1 kb 60 β-catenin K562-S23 IgG

30 3′

–Wnt3A 5′ 20

Exon Relative enrichment (input) 10 Intergenic Total = 251Total = 331 Total = 211 Total = 353 Intron 0 Non-coding SLC38A8 Ccrl2 Yaf2 Cdh1 Axin2 Sp5 Sp8 Promoter-TSS TTS

g Bound enhancer regions +Wnt3A 823 886 LSD1 LSD1 5820 4068 Total = 332 Total = 446 Total = 376 Total = 352 Gfi1b Gfi1b 891 813 1038 850 β-catenin ChIP-seq 68 4996 148 3151

10 + Wnt3a 36 0 4 f Distance from Gfi1b peaks Distance from LSD1 peaks 1 31 0 kb >1 kb 0 kb >1 kb 12 19 10 40 β-catenin β-catenin 67 23 11 90

h β-catenin ChIP-seq peaks Overlap with other factors NoneGfi1b LSD1 Both –Wnt3A 26% Promoter

– Wnt3A Enhancer Total = 8509Total = 60,291 Total = 8461 Total = 769 58% 4% 12% Transcribed Untranscribed 0 20406080100%

11% Promoter +Wnt3A 11% Enhancer + Wnt3A 55% Transcribed Total = 8658Total = 45,138 Total = 8652 Total = 1824 23% Untranscribed LSD1 ChIP-seq Gfi1b ChIP-seq 0 20406080100%

WT MKs with Wnt3A led to a RI decrease and higher Wnt3A Discussion concentration treated WT MKs resemble untreated Gfi1b KO The concept that Gfi1b regulates Wnt/β-catenin target genes is MKs (Fig. 9d and Supplementary Fig. 8c). This data suggest that supported by several, independent lines of experimental evidence Gfi1b is required to maintain a level of Wnt/β-catenin activity from this study: Gfi1b is found in β-catenin protein complexes necessary to control HSC and MK cellularity and the response of found in immune-precipitations analyzed by mass spectrometry MKs to integrin. as well as BioID experiments and western blotting; Gfi1b, LSD1,

NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z

Fig. 8 GFi1b, β-catenin and LSD1 co-occupy sites at enhancer regions and gene promoters. a Venn diagram showing GFI1B, β-catenin and LSD1 ChIP-seq promoter binding prior to and after Wnt3A treatment of K562 cells. P indicates statistical significance of overlap between GFI1B and β-catenin calculated using Fisher’s exact test. b Example of target genes Axin2 and Yaf2 bound by GFI1B, β-catenin and LSD1 and H3K4me1 and H3K9me2 enrichment before and after Wnt3A treatment. Arrowhead indicates the presence of Gfi1b and TCF binding motifs in the Axin 2 promoter. c Motif enrichment analysis under GFI1B peaks alone (top) and Gfi1b/β-catenin overlapped peaks (bottom). d GFI1B (top) and β-catenin (bottom) ChIP-PCR analysis in K562 stable clones expressing GFI1B (S23) or scramble (C5) shRNA (error bars show s.d, n = 3 technical replicates). e Distribution of β-catenin peaks based on their distance to GFI1B peaks (left panel) and LSD1 peaks (right panel). f Distribution of LSD1 peaks based on their distance to GFI1B peaks (left panel) and reversely (right panel). g Venn diagram showing GFI1B, β-catenin and LSD1 binding at enhancer regions prior to and after Wnt3A treatment of K562 cells. h Distribution of β-catenin peaks (left) and their overlaps with GFI1B and/or LSD1 peaks based on the genomic feature of each bound region

and β-catenin co-target promoters and enhancers at endogenous repressor. For instance, the canonical Wnt/β-catenin target gene levels in the erythroleukemic K562 cell line; β-catenin binds Ccnd1 promoter is occupied by Gfi1b, β-catenin, and LSD1 in significantly less at specific target regions in K562 cells upon mouse ESCs and its expression and H3K9 acetylation levels are GFI1B depletion; GFI1B activates TCF-dependent transcription decreased in Gfi1b-deficient cells upon Gfi1b knockdown. in a TOP/FOP reporter assay and Gfi1b deficiency leads to lower Additionally, our findings that LSD1 and GFI1B together enhance Wnt/β-catenin signaling in MK and HSCs in vivo when measured TCF mediated transcription even more than GFI1B alone support by two different Wnt/β-catenin dependent reporter mice. Also, a model of a Gfi1b/LSD/β-catenin activator complex. Alter- Gfi1b- deficient murine HSCs and MKs show deregulated natively, we show that GFI1B interacts with a number of β- expression of Wnt target genes and knockdown of GFI1B in K562 catenin inhibitors, such as CtBP1 or TLE1. It is therefore con- cells reproduces this deregulation of Wnt target genes seen in cells ceivable that GFI1B can also sequester these inhibitors, thereby from Gfi1b knockout mice. And lastly, cellular defects associated liberating β-catenin from their inhibitory effect, leading to with Gfi1b deficiency can be in part restored by overexpressing β- increased β-catenin target gene expression. The results from our catenin or by stimulating the Wnt/β-catenin signaling pathway. TOP/FOP reporter assay with GFI1B in the presence of CtBP1 or Our biochemical analyses suggest that a tripartite Gfi1b/LSD1/ TLE1 also support this possibility. β-catenin complex exists occupying promoter and enhancer In contrast, the non-canonical Wnt genes NFAT5, PTK7 and regions. This is supported by published ChIP-seqs from trans- ROCK2 show increased expression in Gfi1b-deficient cells and the fected cells over-expressing these factors and our own ChIP-seqs NFAT5 promoter shows increased H3K9 acetylation. This sug- on K562 cells in which this complex exists endogenously at target gests that non-canonical Wnt targets respond differently to sites. One important line of evidence suggesting the importance GFI1B than canonical Wnt/β-catenin targets. Also, regulation of of GFI1B for β-catenin recruitment to DNA is the fact that almost Wnt/β-catenin target genes by Gfi1b is evident in primary flow all β-catenin binding sites significantly overlap with GFI1B sorted hematopoietic cells and MKs obtained after prolonged binding sites (see Fig. 8a). Notably, we discovered that GFI1B is culturing with growth factors and cytokines. However, their Gfi1b essential for β-catenin to associate with specific target regions, like dependent regulation of expression is not the same, suggesting YAF2 and CCRL2, upon Wnt signaling activation. Other loci, like that the regulatory link between Gfi1b and β-catenin and their AXIN2 and SP5, did not show a significant change in β-catenin gene expression effects is context dependent. levels upon GFI1B depletion, despite the tripartite complex being Analysis of consensus motifs at GFI1B and β-catenin binding detected at these loci in K562 cells. This indicates that β-catenin sites also supports a regulatory interaction between the two may also bind to some loci independently of GFI1B. proteins but do not reveal the mechanism of this interaction. Interestingly, the majority of β-catenin targets are co-occupied While there are promoters where GFI1B and β-catenin bind only by GFI1B and LSD1, while β-catenin only co-occupies a very small after Wnt treatment (AXIN2 and YAF2 are examples), Wnt sti- fraction of GFI1b/LSD1 targets (108 of 5727 in Wnt treated cells, mulation does not alter the majority of GFI1B and LSD1 binding see Fig. 8a). This may indicate that besides GFI1B and LSD1’s and β-catenin generally binds to promoters where GFI1B and influential role in Wnt/β-catenin biology, both are involved in LSD1 are already present. The fact that GFI1B binding motifs are multiple other processes, whereas many β-catenin targets are less frequent at such sites compared to sites bound by GFI1B but affected by GFI1B and LSD1. In addition, it is notable that while not β-catenin suggests that GFI1B plays a role at these sites which the overall number of β-catenin peaks increases following Wnt3A is specific to the regulation of β-catenin targets. It should be noted treatment (from 603 to 815), a considerable number of promoters that consensus motifs are not strictly required for binding to were bound by β-catenin in untreated cells (247), suggesting the DNA. It is possible for transcription factors to bind in the absence existence of a baseline level of activity of β-catenin. of their established motif, and to have their nucleotide sequence The tripartite Gfi1b/LSD1/β-catenin complex also acts in such a preference altered by DNA shape, genomic context, DNA mod- way that GFI1B recruits LSD1 to β-catenin or at least enhances the ifications and coding and noncoding (genetic) variation51. Our interaction between LSD1 and β-catenin. This is supported by data suggest that GFI1B recruitment to TCF/LEF sites can be reporter gene assays and immune-precipitation experiments with independent of its consensus element. An alternative possibility is mutant forms of GFI1B that lack or disrupt the SNAG domain and that GFI1B can be recruited in a complex with β-catenin to TCF no longer bind to LSD1. Our data corroborate findings in other at sites where TCF binds DNA. In this situation GFI1B does not systems in particular the observation that β-catenin can recruit necessarily have to contact DNA directly and thus would not LSD1 to regulate the expression of the tumor suppressor Lefty1 in require a DNA binding motif to be present at this site. Inde- mouse embryonic stem cells47. Similarly, another study identified pendently of this, we also suggest an additional potential LSD1 as a component of β-catenin complexes that also contained mechanism whereby GFI1B could activate TCF-dependent tran- DNMT1 in human colorectal cancer cell line HCT11648. It has been scription by sequestering negative regulators without binding shown as well that LSD1 promotes Wnt/β-catenin pathway acti- DNA or occupying a specific site via another factor. This vation in liver cancer initiating cells and HCT 116 cell line49,50. mechanism is also independent of DNA binding. At the moment, We have obtained data supporting that a tripartite Gfi1b/ we have evidence supporting these models without excluding one LSD1/β-catenin complex can act as either an activator or or the other. It is also possible that GFI1B is present at sites prior

12 NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z ARTICLE

a b 3 80 Wnt3A ng/ml 15 0 ** * 6 Tamoxifen FACS analysis 100 * * d0 d1 d2 d7 60 500 10 * Lin depleted 4 *** BM cuture + Wnt3A 40

5 2 Rosa-cre tg 20 Gfi1b WT/flox or Gfi1bflox/flox

0 0 0

Absolute number of cells in culture ×10 WT MK KO MK WT HSCs KO HSCs WT HSCs KO HSCs CD41+/CD9+ CD41+/CD9+ CD41low/CD9low CD41low/CD9low

c d

PF4-cre tg PF4-cre tg PF4-cre tg WT/flox n.s Gfi1b Gfi1bWT/flox Gfi1bflox/flox 3.5 n.s 3.0 *** n.s n.s 2.5

2.0 ***

1.5 MK roundness index 50 μm +Wnt3A 1.0

Wnt3A (μg/ml) 00 0.9 1.3 1.9 2.5 3.1

PF4-cre tg flox/flox PF4-cre tg Gfi1b WT/flox 3.5 Gfi1b n.s n.s 3.0 2.5 * 2.0 * ***

1.5 MK roundness index 50 μm +Wnt3A 1.0 DAPI CD41 βtubulin Wnt3A (μg/ml) 0 0.9 1.3 1.9 2.5 3.1 e Gfi1b Canonical Non-canonical Wnt Wnt HSC β-catenin Gfi1b WT MK MK Fibronectin

Gfi1b Non-canonical Wnt Gfi1b KO MK β Canonical -catenin Wnt Fibronectin

Fig. 9 Up-regulation of canonical Wnt signaling rescues Gfi1b-deficient phenotypes in HSCs and MKs. a, b Lineage depleted Gfi1b WT/KO BM cells were taken into culture under the indicated concentrations of Wnt3A. MKs and HSCs CD41low/CD9low were quantified by FACS at day seven post Tamoxifen injection, n = 5 GFI1b WT and five GFI1B KO mice. c Gfi1b KO MKs spreading on fibronectin-coated matrices is rescued by Wnt3A treatment. d MK roundness index plotted as bean plot. Upper part: spreading of Gfi1b KO MKs on Fibronectin coated matrices is rescued by Wnt3A treatment at increasing concentrations. Lower part: spreading of WT MKs on Fibronectin coated matrices is abrogated by Wnt3A treatment at increasing concentrations. (*p < 0.05, **p < 0.001, ***p < 0.0001 on a Mann–Whitney U-test, colored horizontal line shows median, n values for upper part are, in order from left to right: 39, 46, 95, 83, 91, 111, and 87. n values for lower part are, in order from left to right: 39, 120, 89, 108, 86, and 81). Each horizontal short line represents one individual cell. e Proposed model: Gfi1b interacts with β-catenin and regulates Wnt signaling in HSCs and MKs. Deletion of Gfi1b alters the balance between canonical and non-canonical Wnt signaling leading to expansion of HSCs and MKs and impaired MKs spreading on Fibronectin coated matrix to the recruitment of β-catenin as a form of priming, and that While Gfi1b deletion leads to both HSC and MKs expansion and, once β-catenin is recruited to these sites, it forms a new complex upon Wnt3a treatment, a significant decrease in expansion in both with GFI1B that is responsible for the regulation of the gene. cell populations, there are differences in the expression of Wnt target The fact that recruitment of β-catenin to many of these sites genes between these cell populations (Fig. 6d). This may be due to is reduced in the context of GFI1B knockdown supports this different transcription factors and other gene regulator within the two hypothesis. different cell populations. For example, self-replicating and

NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z pluripotent HSCs distinctly contain Gfi1andPU.1,whilethemore Immunofluorescence of Gfi1b WT/KO MKs. Lineage depleted bone marrow from differentiated MKs contain Gata1 and Fli1 which are not prominent PF4-Cre Gfi1b-flox/flox and Gfi1b-wt/flox mice were cultured in Stemspan/2.6% FBS in HSCs52,53. These transcription factors could distinctly influence supplemented with 1% L-Glutamine and SCF (20 ng per ml). TPO (50 ng per ml) fi ’ was added to fresh medium at day 2 and cells were cultured for 4 more days. MKs G 1b s genomic target regions within these different cells populations were enriched on a BSA gradient and plated on fibronectin (500 µg per ml; Life and lead to expressional differences of Wnt target genes, further Technology) coated 12-µ-chamber slide (ibidi). Cells were incubated at 37 °C, 5% – highlighting the need for future research. CO2 for 3 6 h with or without Wnt3A (R&D systems) allowing megakaryocytes to fi Our finding that treatment of Gfi1b-deficient HSCs and MKs attach and spread. After incubation cells were xed with 4% formalin, permeabilized with 0.1% Triton-X100 in PBS, and blocked with FcBlock (1:500; BD Biosciences). with Wnt3A treatment can rescue both the expansion of HSCs Cells were then labeled with FITC-CD41 (BD Biosciences), AF555-β-tubulin (Cell and MKs, as well as the spreading of Gfi1b-deficient MKs Signaling) and covered with Vector Shield containing DAPI (Vector Laboratories). strongly supports a role of Gfi1b as a regulator of the Wnt/β- Immunofluorescence imaging of the slides was done using a DCX-950P DP72 catenin pathway. This may be possible through β-catenin camera (Sony) mounted on a Leitz DMRB microscope (Leica). Images were analyzed recruitment to target loci that do not depend on Gfi1b, such as using ImageJ v1.46r (NIH). AXIN2 and SP5, but which nevertheless are targeted by the tri- partite complex. Other transcription factors like, Gata1, Fli1, BioID-MS data analysis. The BioID-MS data were analyzed using the ProHits fl β software54,55. The Proteowizard4 tool was used to convert RAW files to.mzXML TCF1 or LEF1 could potentially be in uencing factors of - files. Peptide search and identification were performed by using Human RefSeq catenin recruitment in these cases. Further evidence comes from version 57 and the iProphet tool integrated in ProHits56. our experiments using retroviral transduction of active β-catenin fi into bone marrow cells of Gf1b- de cient mice, which led to a GO term enrichment analysis. Functional annotation of GO terms was analyzed significant reduction of HSCs and MKs cell numbers in mice by using the g:Profiler tool57. Biological process or molecular function of prey transplanted with transduced BM cells. proteins identified in GFI1B-BirA*-Flag BioID/MS or β-catenin-BirA*-Flag Our findings describe an unexpected function for GFI1B in BioID/MS with the indicated constructs are shown in heat map analyses. Reviewed β UniProtKB entries of the prey proteins analyzed in Significance Analysis of modulating Wnt/ -catenin signaling, which differs from its tra- INTeractome (SAINT) results generated in ProHits were entered in the Query field ditional activity as a transcriptional repressor, since mutants that on g:Profiler for GO term analysis54,55,58. Contaminant proteins identified in lack direct DNA binding still activate TCF dependent transcrip- GFI1B-Flag AP/MS were filtered by using the Contaminant Repository for Affinity fi Purification (CRAPome)59 repository prior to the GO term analysis on g:Profiler. tion and the genomic sequences at sites of co-occupation of G 1b/ − β-catenin and LSD1 lack the classical Gfi1/Gfi1b DNA consensus The enrichment score of GO terms is shown as the log10 of corrected P values. motif, suggesting that GFI1B may act as part of yet undescribed fi β regulatory complexes. This new function of Gfi1b as a regulator of Dot plot analysis. SAINT output les of GFI1B-BirA*-Flag BioID/MS or -cate- β nin-BirA*-Flag BioID/MS data analyzed in ProHits were submitted in ProHits-viz the Wnt/ -catenin pathway appears essential for HSCs and MKs to perform dot plot analyses60. to maintain their cellularity and for MKs to respond to integrin ligands properly (Fig. 9e). Although further investigation into the TOP/FOP Flash reporter assays. 293T Cells were transfected with TOP/FOP precise mechanism of this novel role is required, it is plausible that Flash reporters, β-galactosidase, and effector plasmids using Lipofectamine® the level of Gfi1b expression is critical not only for the fine-tuning (ThermoFisher) and luciferase activity was normalized by β-galactosidase activity61. of the expression of Wnt/β-catenin target genes, but also for the balance between canonical and non-canonical Wnt signaling in Quantification of intracellular β-galactosidase activity. Lineage negative BM these cells to maintain their functionality. cells were loaded with 2 mM Fluorescein di-β-D-galactopyranoside (FDG) sub- strate (AAT Bioquest) by hypotonic shock at 37 °C for 5 min, prior to cell surface antibody staining. β-galactosidase reaction was stopped with 1 mM Phenylethyl Methods β-D-thiogalactopyranoside (PETG, Sigma). Mice.Gfi1bfl/fl mice9 were crossed with Rosa-CreERT2, and Axin2+/LacZ (Jackson, B6.129P2-Axin2tm1Wbm/J) or TCF/Lef:H2B-GFP (Jackson, 61Hadj/J) transgene fi reporter line. PF4-cre mice were used to delete Gfi1b in MKs for immuno- Gene expression pro ling by RNA-seq analysis. Bone marrow from 2 tibiae, fi fl/fl fi wt/fl fluorescence microscopy analysis. Age and sex-matched mice were used in all 2 femora and 2 humeri (from 5 Rosa-Cre G 1b and 10 Rosa-Cre G 1b mice experiments. For BM transplantation C57B6 CD45.1 mice were used. All mice were treated with tamoxifen 2 weeks prior the experiment) was harvested in PBS/2.5% housed under SPF conditions and the IRCM Institutional Review Board approved FBS and pooled prior to lineage negative depletion using autoMACS Pro separator all animal protocols and experimental procedures were performed in compliance (Miltenyi Biotec). Cells were incubated with a lineage antibody cocktail (B220, with IRCM and CCAC (Canadian Council of Animal Care) guidelines. CD3, CD4, CD8, Gr1, CD11b, NK1.1, Il7R, CD19) and then labeled with PE/Cy5- streptavidin, PE-anti-CD41, AF700- anti-CD9, APC- anti-CD150, BV421-anti- cKit, BV510-anti-CD48 and PE/Cy7-anti-Sca1 antibodies. MKs (Lin− cKit+ CD41 Treatment. Suboptimal doses of tamoxifen were injected at 100 mg per kg at day 0 + CD9+), HSCs (Lin− cKit+ Sca1+ Cd48− CD150+), and HSCs CD41low CD9 low followed by 50 mg per kg the day after, and mice were analyzed at day 10. were sorted on FACSAria II sorter (BD Biosciences). RNA was extracted using MagMax-96 Total RNA Isolation kit (Ambion) and quality-checked with RNA Flow cytometry analysis, sorting of HSCs and MKs. Hematopoietic cells were 6000 Pico kit (Agilent). RNA-seq libraries were prepared from the RNA extracts fl analyzed with LSR, or LSR Fortessa ow cytometers (BD Biosciences, Mountain using the Illumina TruSeq Stranded mRNA Kit according to the manufacturer’s View, CA) and analyzed using BD FACS Diva software (BD Biosciences) or FlowJo instructions, and sequenced using the TruSeq PE Clusterkit v3-cBot-HS on an (for histogram overlays; Tree Star). For cell sorting, lineage negative BM cells were Illumina HiSEq 2000 system. Sequencing reads were aligned to the mm10 genome fi rst depleted using mouse lineage cell depletion kit (Miltenyi Biotec) then applied using Tophat v2.0.1062. Reads were processed with Samtools63 and then mapped to fi to ve-laser FACSAria II sorter (BD Biosciences). Ensembl transcripts using HTSeq64. Differential expression was tested using the DESeq R package65(R Core Team 2015, http://www.r-project.org/). A genome Cell culture. K562 (ATCC CRL-3344), HEL (ATCC TIB-180) cells were main- coverage file was generated and scaled to RPM using Bedtools66. RNA-seq data tained in RPMI media (Multicell) supplemented with 10% Bovine Growth Serum are available under accession number GSE71310 (MKs) and GSE85737 (HSCs). (RMBIO Fetalgro) and 100 IU Penicillin and 100 μg/ml Streptomycin (Multicell). Total read numbers and aligned read numbers for each experiment is shown in HEK-293 (ATCC CRL-1573) and U2OS (ATCC HTB-96) cells were maintained Supplementary Data 1. DMEM media (Multicell) with above mentioned supplements. We verified that none of the cell lines used in this study were found in the Register of Misidentified Cell Lines maintained by the International Cell Line Authentication Committee. Functional analysis. The enrichment of selected biological functions of interest (Supplementary Table 1) was also analyzed using the GSEA tool67. Normalized read counts for Ensembl genes from HTSeq were used and enrichment calculated Antibodies. The following antibodies were used for western blots analysis at β using 1000 Gene Set permutations. Unsupervised clustering analysis was done 1/1000 dilution: -catenin (cell signaling 9566 and 4176), TLE1 (sc-9121), TLE3 using web tool ClustVis (https://biit.cs.ut.ee/clustvis/). (sc-13374), Pontin52 (ab133513), CHD8 (cell signaling 7656), LSD1 (sc-67272 and cell signaling 4064), Gfi1b (ARP30094 and sc-28356), Tubulin (cell signaling 9099), G6PD (ab993), HDAC1 (ab702A). Uncropped and unprocessed scans of the Consensus motif analysis. Motif scanning was performed using the AME tool western blots are provided in Supplementary Data 3. from the MEME Suite68 using the JASPAR CORE 2016 database.

14 NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z ARTICLE

Chromatin immuno-precipitation (ChIP). GFI1B, LSD1, β-catenin and histone 2. Vassen, L., Okayama, T. & Moroy, T. Gfi1b:green fluorescent protein knock-in modification ChIPs were performed on 1–20 × 106 K562 cells treated or untreated mice reveal a dynamic expression pattern of Gfi1b during hematopoiesis that in culture with 100 ng per ml rhWnt3a (5036-WN; R&D Systems) or CHIR99021 is largely complementary to Gfi1. Blood 109, 2356–2364 (2007). (SML1046;Sigma) for 4 or 8 h, respectively. The cells were cross-linked with 1.5 mM 3. Chowdhury, A. H. et al. Differential transcriptional regulation of meis1 by EGS for 20 min and 1% formaldehyde for 8 min before quenching with 125 mM Gfi1b and its co-factors LSD1 and CoREST. PLoS. ONE. 8, e53666 glycine. Cells were lysed in lysis buffer and sonicated using a Covaris E220 to (2013). generate 200–600 bp fragments69. Samples were immuno-precipitated with 2–5µg 4. Saleque, S., Kim, J., Rooke, H. M. & Orkin, S. H. Epigenetic regulation of of either anti- GFI1B (ARP30094_P050; Aviva Systems Biology), anti-LSD1 hematopoietic differentiation by Gfi-1 and Gfi-1b is mediated by the cofactors β – (ab17721;Abcam), anti- -catenin antibody (71 2700; ThermoFisher), anti- CoREST and LSD1. Mol. Cell 27, 562–572 (2007). H3K4me1 (ab8895, Abcam), anti-H3K4me2 (ab11946; Abcam), anti-H3K4me3 5. Upadhyay, G., Chowdhury, A. H., Vaidyanathan, B., Kim, D. & Saleque, S. (ab8580; Abcam) or anti-H3K9me2 (ab1220; Abcam). Libraries were generated Antagonistic actions of Rcor proteins regulate LSD1 activity and cellular ’ according to Illumina s instructions. Libraries were sequenced on the Illumina Hi- differentiation. Proc. Natl Acad. Sci. USA 111, 8071–8076 (2014). ’ seq 2000 following the manufacturer s protocols to obtain 50 bp paired end reads. 6. Maiques-Diaz, A. et al. Enhancer activation by pharmacologic displacement fi 38 46 ChIP-seq results for G 1b binding in HPC7 cells and CTCF in ESCs were of LSD1 from GFI1 induces differentiation in acute myeloid leukemia. obtained from the laboratory website of Dr. Goettgens (http://hscl.cimr.cam.ac.uk/). Cell Rep. 22, 3641–3659 (2018). Results for β-catenin39 and LSD140 binding in ESCs were obtained from GEO 7. Duan, Z., Zarebski, A., Montoya-Durango, D., Grimes, H. L. & Horwitz, M. accessions GSE43597 and GSE22557, respectively. External datasets were obtained Gfi1 coordinates epigenetic repression of p21Cip/WAF1 by recruitment of in the form of.bed files of peaks and.wig visualization tracks, aligned to the mm9 histone lysine methyltransferase G9a and histone deacetylase 1. Mol. Cell. Biol. build, with the exception of LSD1, which only included the.bed peak file. 25, 10338–10351 (2005). 8. Vassen, L., Fiolka, K. & Moroy, T. Gfi1b alters histone methylation at target Annotation databases used. For gene promoters, we used the Ensembl Genes 92 gene promoters and sites of gamma-satellite containing heterochromatin. database, dataset GRCh38.p12. (https://useast.ensembl.org/index.html) For EMBO J. 25, 2409–2419 (2006). enhancer regions, we used the Fantom5 human_permissive_enhancer- 9. Khandanpour, C. et al. Evidence that growth factor independence 1b regulates s_phase_1_and_2 enhancers (February 2015) dataset (http://fantom.gsc.riken.jp). dormancy and peripheral blood mobilization of hematopoietic stem cells. Blood 116, 5149–5161 (2010). Megakaryocytes in vitro culture and roundness index. Primary megakaryocytic 10. Vassen, L. et al. Growth factor independence 1b (gfi1b) is important for cultures were prepared by flushing bone marrow into DMEM/10% FBS and sus- the maturation of erythroid cells and the regulation of embryonic globin pended in 1x ACK red blood cell lysis buffer (0.15 M NH4Cl, 10 mM KHCO3,1 expression. PLoS. ONE. 9, e96636 (2014). mM EDTA). Cells were then labeled with a biotinylated lineage antibody cocktail 11. Beauchemin, H. et al. Gfi1b controls integrin signaling-dependent (B220, Gr1, CD11b, CD16/32) and anti-biotin magnetic beads and separated on cytoskeleton dynamics and organization in megakaryocytes. Haematologica an AutoMACS. The negative fraction was washed with DMEM/10% FBS and 102, 484–497 (2017). suspended in StemSpan SFEM (StemCell Technologies) that contained 2.6% FBS, 12. Ono, M., Matsubara, Y., Shibano, T., Ikeda, Y. & Murata, M. GSK-3beta 1% L-Glutamine and SCF (20 ng per ml). Cells were then cultured two days at negatively regulates megakaryocyte differentiation and platelet production 37 °C and 5% CO2. The media was then replaced with fresh StemSpan SFEM that from primary human bone marrow cells in vitro. Platelets 22, 196–203 (2011). contained 2.6% FBS, 1% L-Glutamine, SCF (20 ng per ml) and TPO (50 ng per ml) 13. Macaulay, I. C. et al. Canonical Wnt signaling in megakaryocytes regulates and cells were cultured for 4 more days. On the 7th day, mature megakaryocytes proplatelet formation. Blood 121, 188–196 (2013). were enriched on a BSA gradient and plated in fibronectin coated 12-well μ- 14. Luis, T. C. et al. Canonical wnt signaling regulates hematopoiesis in a dosage- Chamber glass slides (ibidi). Cells were then allowed to attach and to spread for dependent fashion. Cell. Stem. Cell. 9, 345–356 (2011). – 3 8 h at 37 °C under 5% CO2. To quantify the spreading of MKs on coated slides, 15. Fleming, H. E. et al. Wnt signaling in the niche enforces hematopoietic stem we compared the periphery (P) of each cell to the circumference of a perfect circle cell quiescence and is necessary to preserve self-renewal in vivo. Cell. Stem. π (2 r) having the same area as this cell and calculated a roundness index (RI) Cell. 2, 274–283 (2008). resulting in a numerical value that increases as the complexity of a spreading cell 16. Kirstetter, P., Anderson, K., Porse, B. T., Jacobsen, S. E. & Nerlov, C. = 11 increases, starting at RI 1 for a perfectly round cell . Activation of the canonical Wnt pathway leads to loss of hematopoietic stem cell repopulation and multilineage differentiation block. Nat. Immunol. 7, Statistical analysis. The paired Student t-test was chosen for analyzing the dif- 1048–1056 (2006). ferences in the intracellular β-galactosidase activity in MKs and HSCs. Welch 17. Luis, T. C. et al. Wnt3a deficiency irreversibly impairs hematopoietic stem corrected Student’s t tests were used to calculate the statistical significance of results cell self-renewal and leads to defects in progenitor cell differentiation. Blood from luciferase reporter assays. Overlap between binding sites was calculated using 113, 546–554 (2009). Fisher’s exact test. All p-values were calculated two-sided, and values of p < 0.05 18. Staal, F. J., Luis, T. C. & Tiemessen, M. M. WNT signalling in the immune were considered statistically significant. Statistical analysis was done with Graph- system: WNT is spreading its wings. Nat. Rev. Immunol. 8, 581–593 (2008). Pad Prism software (GraphPad software, La Jolla, CA, USA). The sample size of 19. Malhotra, S. et al. Contrasting responses of lymphoid progenitors to canonical data points for each assay is shown in Supplementary Data 2. and noncanonical Wnt signals. J. Immunol. 181, 3955–3964 (2008). 20. Kokolus, K. & Nemeth, M. J. Non-canonical Wnt signaling pathways in Reporting summary. Further information on experimental design is available in hematopoiesis. Immunol. Res. 46, 155–164 (2010). the Nature Research Reporting Summary linked to this article. 21. Wang, Q. et al. NFAT5 represses canonical Wnt signaling via inhibition of beta-catenin acetylation and participates in regulating intestinal cell differentiation. Cell Death Dis. 4, e671 (2013). Data availability 22. Peradziryi, H. et al. PTK7/Otk interacts with Wnts and inhibits canonical Wnt The raw proteomics data, which are presented in Figs. 2a, b, and 3d, is publicly available signalling. EMBO J. 30, 3729–3740 (2011). on massIVE [https://massive.ucsd.edu/ProteoSAFe] under the following accession 23. Nishiyama, M., Skoultchi, A. I. & Nakayama, K. I. Histone H1 recruitment by number: MSV000083125. The raw ChIP-seq and RNA-seq data, which is presented in CHD8 is essential for suppression of the Wnt-beta-catenin signaling pathway. Figs. 6–8 and Supplementary Figures 3 and 7, have been uploaded to the GEO Datasets Mol. Cell. Biol. 32, 501–512 (2012). repository [https://www.ncbi.nlm.nih.gov/gds] and is available under the following 24. Bellosta, P. et al. Myc interacts genetically with Tip48/Reptin and Tip49/ accession numbers: GSE71310, GSE85737 and GSE117944. Pontin to control growth and proliferation during Drosophila development. Previously published ChIP-seq data, which is presented in Supplementary Figs 4 and 5, is Proc. Natl Acad. Sci. USA 102, 11799–11804 (2005). available under the following accession numbers: GSE43597, GSE22557. 25. Yu, N. et al. HSP105 recruits protein phosphatase 2A to dephosphorylate beta- catenin. Mol. Cell. Biol. 35, 1390–1400 (2015). 26. Chinnadurai, G. The transcriptional corepressor CtBP: a foe of multiple tumor Received: 21 April 2017 Accepted: 25 February 2019 suppressors. Cancer Res. 69, 731–734 (2009). 27. Lu, L. et al. Kdm2a/b lysine demethylases regulate canonical Wnt signaling by modulating the stability of nuclear beta-catenin. Dev. Cell. 33, 660–674 (2015). 28. Arce, L., Pate, K. T. & Waterman, M. L. Groucho binds two conserved regions of LEF-1 for HDAC-dependent repression. Bmc. Cancer 9, 159 (2009). 29. Mosimann, C., Hausmann, G. & Basler, K. Beta-catenin hits chromatin: regulation References of Wnt target gene activation. Nat. Rev. Mol. Cell Biol. 10,276–286 (2009). 1. Fiolka, K. et al. Gfi1 and Gfi1b act equivalently in haematopoiesis, but have 30. Su, X. et al. Molecular basis underlying histone H3 lysine-arginine distinct, non-overlapping functions in inner ear development. EMBO Rep. 7, methylation pattern readout by Spin/Ssty repeats of Spindlin1. Genes & Dev. 326–333 (2006). 28, 622–636 (2014).

NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications 15 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-019-09273-z

31. Flack, J. E., Mieszczanek, J., Novcic, N. & Bienz, M. Wnt-Dependent 60. Knight, J. D. R. et al. ProHits-viz: a suite of web tools for visualizing Inactivation of the Groucho/TLE Co-repressor by the HECT E3 Ubiquitin interaction proteomics data. Nat. Methods 14, 645–646 (2017). Ligase Hyd/UBR5. Mol. Cell 67, 181–193 e185 (2017). 61. Vadnais, C. et al. Autocrine activation of the wnt/beta-catenin pathway by 32. Mosimann, C., Hausmann, G. & Basler, K. Parafibromin/Hyrax activates CUX1 and GLIS1 in breast cancers. Biol. Open 3, 937–946 (2014). Wnt/Wg target gene transcription by direct association with beta-catenin/ 62. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Armadillo. Cell 125, 327–341 (2006). Nat. Methods 9, 357–359 (2012). 33. Yu, E. J. et al. Positive regulation of beta-catenin-PROX1 signaling axis by 63. Li, H. et al. The sequence alignment/Map format and SAMtools. DBC1 in colon cancer progression. Oncogene 35, 3410–3418 (2016). Bioinformatics 25, 2078–2079 (2009). 34. Griffin, C. T., Curtis, C. D., Davis, R. B., Muthukumar, V. & Magnuson, T. 64. Anders, S., Pyl, P. T. & Huber, W. HTSeq–a Python framework to work The chromatin-remodeling enzyme BRG1 modulates vascular Wnt signaling with high-throughput sequencing data. Bioinformatics 31, 166–169 (2015). at two levels. Proc. Natl Acad. Sci. USA 108, 2282–2287 (2011). 65. Anders, S. & Huber, W. Differential expression analysis for sequence count 35. Jho, E. H. et al. Wnt/beta-catenin/Tcf signaling induces the transcription data. Genome Biol. 11, R106 (2010). of Axin2, a negative regulator of the signaling pathway. Mol. Cell. Biol. 22, 66. Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for 1172–1183 (2002). comparing genomic features. Bioinformatics 26, 841–842 (2010). 36. Ferrer-Vaquer, A. et al. A sensitive and bright single-cell resolution live 67. Subramanian, A. et al. Gene set enrichment analysis: a knowledge-based imaging reporter of Wnt/ss-catenin signaling in the mouse. Bmc. Dev. Biol. 10, approach for interpreting genome-wide expression profiles. Proc. Natl Acad. 121 (2010). Sci. USA 102, 15545–15550 (2005). 37. Labbe, E. et al. Transcriptional cooperation between the transforming growth 68. McLeay, R. C. & Bailey, T. L. Motif enrichment analysis: a unified framework factor-beta and Wnt pathways in mammary and intestinal tumorigenesis. and an evaluation on ChIP data. BMC Bioinforma. 11, 165 (2010). Cancer Res. 67,75–84 (2007). 69. Lee, T. I., Johnstone, S. E. & Young, R. A. Chromatin immunoprecipitation and 38. Hannah, R., Joshi, A., Wilson, N. K., Kinston, S. & Gottgens, B. A microarray-based analysis of protein location. Nat. Protoc. 1, 729–748 (2006). compendium of genome-wide hematopoietic transcription factor maps supports the identification of gene regulatory control mechanisms. Exp. Hematol. 39, 531–541 (2011). Acknowledgements 39. Zhang, X., Peterson, K. A., Liu, X. S., McMahon, A. P. & Ohba, S. Gene We are indebted to Mathieu Lapointe for technical assistance, Marie-Claude Lavallée and regulatory networks mediating canonical Wnt signal-directed control of Jo-Anny Bisson for excellent animal care, Eric Massicotte and Julie Lord for FACS and pluripotency and differentiation in embryo stem cells. Stem Cells 31, cell sorting, Virginie Calderon for assistance with bioinformatics analysis. We thank 2667–2679 (2013). Genome Quebec for performing HT sequencing. Tarik Möröy holds a Canada Research 40. Nair, V. D. et al. Involvement of histone demethylase LSD1 in short-time-scale Chair (Tier 1) and grants from the CIHR (MOP-84238, MOP-94846, FDN-148372) and gene expression changes during cell cycle progression in embryonic stem cells. from the Canadian Hemophilia Society. Charles Vadnais was supported by fellowships Mol. Cell. Biol. 32, 4861–4876 (2012). from the Fonds de recherche Quebec – Santé (FRQS) and from the CIHR. Halil Bagci 41. Shtutman, M. et al. The cyclin D1 gene is a target of the beta-catenin/LEF-1 was supported by a doctoral training award from FRQS (#33603) and NSERC Discovery pathway. Proc. Natl Acad. Sci. USA 96, 5522–5527 (1999). Grant (RGPIN-2016-04808 to Jean-François Coté). Jean-François Coté holds the 42. Yu, S. et al. Hematopoietic and leukemic stem cells have distinct dependence TRANSAT chair in breast cancer research. Tarik Möröy is supported by a Canada on Tcf1 and Lef1 transcription factors. J. Biol. Chem. 291, 11148–11160 (2016). Research Chair (Tier 1) and by grants from CIHR (MOP-84238, MOP-94846, FDN- 43. Nunez, F., Bravo, S., Cruzat, F., Montecino, M. & De Ferrari, G. V. Wnt/beta- 148372) and the Canadian Hemophilia Society. catenin signaling enhances cyclooxygenase-2 (COX2) transcriptional activity in gastric cancer cells. PLoS. ONE. 6, e18562 (2011). Author contributions 44. You, L., Chen, L., Pan, L., Gu, W. S. & Chen, J. Y. Zinc finger protein 467 P.S.: Concept and design, Collection of data, data analysis, manuscript writing and final regulates Wnt signaling by modulating the expression of sclerostin in adipose approval, A.H.: Concept and design, Collection of data, data analysis, manuscript writing. derived stem cells. Biochem. Biophys. Res. Commun. 456, 598–604 (2015). C.V.: Data analysis, manuscript writing, N.B.: Collection of data, data analysis, H.B.: 45. Andreeva, A. et al. PTK7-Src signaling at epithelial cell contacts mediates Collection of data, data analysis, R.C.: Concept and design, collection of data, H.B.: Data spatial organization of actomyosin and planar cell polarity. Dev. Cell. 29, analysis, F.J.T.S.: Concept and writing, reagents, J-F.C.: Concept and writing, reagents. T. 20–33 (2014). M.: Concept and design, data analysis and supervision, manuscript writing, final 46. Kagey, M. H. et al. Mediator and cohesin connect gene expression and approval, provision of funds. chromatin architecture. Nature 467, 430–435 (2010). 47. Yakulov, T., Raggioli, A., Franz, H. & Kemler, R. Wnt3a-dependent and Additional information -independent protein interaction networks of chromatin-bound beta-catenin in Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- – mouse embryonic stem cells. Mol. Cell. Proteom. 12, 1980 1994 (2013). 019-09273-z. 48. Song, J. et al. A protein interaction between beta-catenin and Dnmt1 regulates Wnt signaling and DNA methylation in colorectal cancer cells. Mol. Cancer Competing interests: The authors declare no competing interests. Res. 13, 969–981 (2015). fi 49. Lei, Z. J. et al. Lysine-speci c demethylase 1 promotes the stemness and Reprints and permission information is available online at http://npg.nature.com/ + chemoresistance of Lgr5( ) liver cancer initiating cells by suppressing reprintsandpermissions/ negative regulators of beta-catenin signaling. Oncogene 34, 3188–3198 (2015). fi 50. Huang, Z. et al. Lysine-speci c demethylase 1 (LSD1/KDM1A) contributes to Journal peer review information: Nature Communications thanks Shireen Saleque and colorectal tumorigenesis via activation of the Wnt/beta-catenin pathway by the other anonymous reviewer(s) for their contribution to the peer review of this work. down-regulating Dickkopf-1 (DKK1) [corrected]. PLoS. ONE. 8, e70077 (2013). Peer reviewer reports are available. 51. Inukai, S., Kock, K. H. & Bulyk, M. L. Transcription factor-DNA binding: – beyond binding site motifs. Curr. Opin. Genet. Dev. 43, 110 119 (2017). Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in 52. Dore, L. C. & Crispino, J. D. Transcription factor networks in erythroid cell published maps and institutional affiliations. and megakaryocyte development. Blood 118, 231–239 (2011). 53. Moroy, T., Vassen, L., Wilkes, B. & Khandanpour, C. From cytopenia to leukemia: the role of Gfi1 and Gfi1b in blood formation. Blood 126, 2561–2569 (2015). Open Access This article is licensed under a Creative Commons 54. Methot, S. P. et al. A licensing step links AID to transcription elongation for Attribution 4.0 International License, which permits use, sharing, mutagenesis in B cells. Nat. Commun. 9, 1248 (2018). adaptation, distribution and reproduction in any medium or format, as long as you give 55. Liu, G. et al. ProHits: integrated software for mass spectrometry-based appropriate credit to the original author(s) and the source, provide a link to the Creative interaction proteomics. Nat. Biotechnol. 28, 1015–1017 (2010). Commons license, and indicate if changes were made. The images or other third party 56. Shteynberg, D. et al. iProphet: multi-level integrative analysis of shotgun material in this article are included in the article’s Creative Commons license, unless proteomic data improves peptide and protein identification rates and error indicated otherwise in a credit line to the material. If material is not included in the estimates. Mol. Cell. Proteom. 10, M111 007690 (2011). article’s Creative Commons license and your intended use is not permitted by statutory 57. Reimand, J. et al. g:Profiler-a web server for functional interpretation of regulation or exceeds the permitted use, you will need to obtain permission directly from gene lists (2016 update). Nucleic Acids Res. 44, W83–W89 (2016). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ 58. Choi, H. et al. SAINT: probabilistic scoring of affinity purification-mass licenses/by/4.0/. spectrometry data. Nat. Methods 8,70–73 (2011). 59. Mellacheruvu, D. et al. The CRAPome: a contaminant repository for affinity purification-mass spectrometry data. Nat. Methods 10, 730–736 (2013). © The Author(s) 2019

16 NATURE COMMUNICATIONS | (2019) 10:1270 | https://doi.org/10.1038/s41467-019-09273-z | www.nature.com/naturecommunications