Niche WNT5A regulates the actin cytoskeleton during regeneration of hematopoietic stem cells

Hematopoietic stem cells (HSCs) are defined by their ability to repopulate the blood cell system in steady-state and transplantation settings. The interplay of intrinsic (donor-derived) and extrinsic (host-derived) factors in successful engraftment is not entirely understood. Several intrinsic subprocesses enable HSCs to engraft: the ability of HSCs to find their niche, migrate into it, lodge there, proliferate, self-renew, and differentiate into mature hematopoietic cells without exhaustion or leukemic transformation. The extrinsic factors are orchestrated by these cellular events and attract HSCs, as well as secreted factors, to the niche, restrict unbridled HSC activation (Renström et al., 2009; Istvanffy et al., 2011; Ruf et al., 2016), prevent HSC senescence, promote survival, and guide proliferation and differentiation.

Interference with any of the engraftment subprocesses delays regeneration of the hematopoietic system and reestablishment of the pool of quiescent HSCs. Wnt signaling maintains the balance between HSC quiescence and HSC activation, through coupling of Frizzled receptors to G-proteins and Disheveled to the β- and γ-catenin–dependent canonical, as well as calcium-dependent or small GTPase-dependent, noncanonical pathways (Dijksterhuis et al., 2014). Catenin-dependent Wnt signaling is required for differentiation and HSC self-renewal through precise regulation of its amplitude: low levels of catenin activation promotes self-renewal, whereas higher levels drive differentiation (Luis et al., 2011). The further downstream mechanisms are likely to involve different degrees of oxidative stress and the ability to resolve DNA double-strand breaks (Lento et al., 2014). In leukemia models, activated β-catenin is required for survival and proliferation in MLL-AF9– (Wang et al., 2010) and BCR-ABL-driven leukemia (Zhao et al., 2007). Although it is not clear how catenin is up-regulated in leukemia, progression of BCR-ABL–driven leukemia is accelerated by activation of catenin, which may, at least in part, be caused by mis-splicing of GSK3β (Abrahamsson et al., 2009).

WNT5A is both glycosylated and palmitoylated, and is the main stimulator of noncanonical Wnt signaling. WNT5A is induced in inflammatory niche responses (Rauner et al., 2012), but its role for normal and malignant hematopoiesis is poorly understood. The expression of Wnt5a is strong in stromal cells, but is also found in B220⁺ lymphocytes (Liang et al., 2003). We previously found that Wnt5a is particularly strongly expressed in stromal cells, which maintain HSCs under noncontact conditions (Buckley et al., 2011). In addition, WNT5A promotes HSC maintenance in the absence of stromal cells through inhibition of canonical signaling (Murdoch et al., 2003; Nemeth et al., 2007). These reports are apparently conflicting with our finding that WNT5A expression is up-regulated during aging in HSCs and, as such, impaired HSC regenerative capacity through CDC42-mediated inhibition of actin polarization (Florian et al., 2013).

To shed more light on how WNT5A produced by the microenvironment regulates HSCs, we studied the role of Wnt5a in the regenerative response of normal hematopoiesis and in BCR-ABL–driven leukemogenesis. Because Wnt5a^−/− mice die just before birth as a result of multiple tissue developmental defects, we used adult Wnt5a-haploinsufficient mice (Yamaguchi et al., 1999). We find that HSCs regenerated in a Wnt5a^+/− environment, show severe defects in HSC engraftment. Gene expression analysis coupled to functional assays show that the HSCs regenerated in the Wnt5a-haploinsufficient niche display up-regulated small GTPase-mediated noncanonical signaling-impaired F-actin polarization. These alterations results in decreased adhesion, migration, and homing. In addition, we observed similar effects in BCR-ABL^p185 expressing cells that failed to develop leukemia in secondary recipients. These results indicate that the Wnt5a-haploinsufficient niche may prevent engraftment of leukemia-initiating cells.

To study the role of Wnt5a in HSC regeneration, we used heterozygous Wnt5a knockout (Wnt5a^+/−) mice, because homozygous knockout mice die just before birth (Yamaguchi et al., 1999; Liang et al., 2003). Wnt5a^+/− mice are viable and fertile, and adult Wnt5a^+/− mice show a 1.7-fold increase in Lin⁻IL-7R⁺ common lymphoid progenitors (CLPs), unchanged myeloid progenitors (MPs), and an unchanged clonogenic activity of progenitors (Fig. 1, A and B; and Fig. S1 A). Furthermore, we observed a 1.4-fold decrease in the Lin⁻ SCA-1⁺ KIT⁺ (LSK) population, which is mainly a result of a 1.7-fold decrease in CD34⁺ LSKs, because the CD34⁻ CD150⁺ LSKs are unchanged (Fig. 1 A).

To determine whether the reduction in LSKs reflects a reduced HSC pool, we sorted HSC-enriched CD34⁻ CD150⁺ LSKs from the Wnt5a^+/+ (WT) and Wnt5a^+/− mice and transplanted these in equal numbers into congenic WT recipients (Fig. 1 C). We found that, despite the known expression of Wnt5a in CD34⁻ Flk2⁻ LSK cells (Florian et al., 2013) and B lymphocytes (Liang et al., 2003), lymphoid and myeloid engraftment of Wnt5a^+/− LT-LSKs was similar in primary, secondary, and tertiary recipients (Fig. 1, D–H; and Fig. S1, B and C). Also, engraftment of WT and Wnt5a^+/− LT-LSKs in the BM (Fig. 1, I–K) was unchanged. Additionally, amplification of the donor pool of WT and Wnt5a^+/− LT-LSKs after each round of transplantations was not significantly different (Fig. 1, L and M).

To test the effect of extrinsic niche WNT5A on HSCs regeneration, we transplanted WT HSCs into WT and Wnt5a^+/− recipient mice (Fig. 2 A). In primary recipients, WT HSCs regenerated similarly in peripheral blood (PB), SPLs, and BM with comparable percentage of lymphoid and myeloid populations in PB (Fig. 2, B and C). Moreover, the absolute numbers of engrafted Lin⁻, MP, LSKs and CD34⁻ CD150⁺ LSKs was similar 16 wk after transplantation (Fig. 2, D and E). To test the self-renewing activity of HSCs engrafted in WT (LSK-WT) and Wnt5a^+/− (LSK-5a) recipients, we serially transplanted equal numbers of donor LSKs. Surprisingly, in three independent experiments, we did not detect engraftment of >1% lymphoid or myeloid donor cells in the PB of secondary recipients at 5, 10, and 16 wk after transplantation from LSK-5a cells (Fig. 2, F–H). In addition, the LSK compartment was not regenerated by LSK-5a cells (Fig. 2, I and J). Transplanting higher numbers of LSK-5a cells showed an estimated 17-fold lower frequency of repopulating HSCs (Fig. 2 G). Hence, we find that regeneration of the HSC compartment in the Wnt5a-haploinsufficient environment is severely compromised.

To find out whether aberrant HSC regeneration is caused by altered niche composition, we first confirmed decreased WNT5A content in the cultured endosteal cells of Wnt5a^+/− mice compared with their WT littermates (Fig. 3 A). We identified that WNT5A expression in freshly isolated collagenase-treated BM cells was restricted to multipotent stromal cells (MSCs; Fig. 3 B). Additionally, we did not observe any gross abnormalities in the niche architecture as detected by laminin- (endothelial cells; ECs) and SCA-1–stained (ECs and MSCs) bone sections (Fig. 3 C).

In the transplanted primary recipients, the number of ECs, osteoblastic cells (OBCs), and MSCs was identical in WT and Wnt5a^+/− recipient mice (Fig. 3, D–F). To find out whether these cells from WT and Wnt5a^+/− mice show differences in gene expression, we generated RNA-Seq libraries from 3 independent pools of sorted MSCs (2.5–4 × 10³ cells) and OBCs (2–3 × 10³ cells) from recipient mice 16 wk after transplantation (Fig. 3 G). Samples were grouped according to the cell identity and genotype in the principal component analysis (PCA; Fig. 3 G). Unsurprisingly, comparison of MSCs and OBCs show a large amount of differentially expressed genes (DEGs; Fig. 3 H). Comparisons of reported niche and pericyte genes in WT and Wnt5a^+/− MSCs and OBCs were very similar, other than decreased Spp1 in OBCs (Fig. 3, I–K; and Table S3). We found only 79 significantly DEGs between MSCs from WT and Wnt5a^+/− recipients (Fig. 3 L). The analysis of specific pathways or biological processes was not informative (Table S4).

To determine how regenerated LSKs respond to the two environments, we generated RNA-Seq libraries from sorted 1,500 to 3,000 LSK-WT and LSK-5a (Fig. 4 A) and mapped in a mean of 12.3 million sequence reads per library. The mapped samples were grouped according to the recipient genotype in the principle component analysis (Fig. 4 B). From the 14,029 genes tested, 656 were differently expressed between LSK-5a and LSK-WT cells (FDR < 0.05; Fig. 4 C and Table S5). These DEGs were enriched in the pathways with protein processing, inositol phosphate metabolism, and cell cycle and actin cytoskeleton regulation among the 10 most enriched pathways (Fig. 4 D and Table S6). Surprisingly, no enrichment for DEGs annotated in the Wnt signaling pathway (Kyoto Encyclopedia of Genes and Genomes [KEGG] ID, mmu04310) or other ontology categories related to the Wnt signaling was found (unpublished data). Indeed, using manual curation, we found that LSK-WT and LSK-5a cells express similar levels of Fzd receptors, as well as mediators of canonical or calcium-dependent noncanonical pathways (Table S5). Interestingly, regulation of actin cytoskeleton (KEGG ID mmu04810) is the pathway with the most genes up-regulated in LSK-5a cells (Fig. 4, D and E). We confirmed these results on the protein level (Fig. 4 F). Interestingly, alterations in protein levels of DOCK2, CDC42, WAVE2, and F-Actin was correlated with a decrease in polarized localization (Fig. 4 G).

A closer look at donor HSC-enriched subpopulations showed that F-actin is also distributed in an apolar manner in CD34⁻ LSK-5a cells (Fig. 4 H). Furthermore, in accordance with our finding that expression of canonical Wnt-related mediators is not altered in LSK-5a cells, we also found no differences in protein levels of the main mediators β-catenin and MYC (Fig. 4 I). These results are reminiscent of our previous finding in aged HSCs, which show reduced F-actin polarization caused by CDC42 activation (Florian et al., 2013). Indeed, a pull-down assay showed that the relative quantity of GTP-binding activated CDC42 is higher in sorted donor Lin⁻ cells from primary Wnt5a^+/− recipients (Fig. 4 J).

To interrogate which transcription factors may contribute to the differential gene expression in LSK-5a cells, the Integrated System for Motif Activity Response Analysis (ISMARA) prediction tool showed enrichment of DEGs in ZEB1, AIRE, FOXQ1, NKX2-2,8, and NFATC1.3 motifs (Fig. 4 K). Importantly, genes in the actin regulatory pathway are enriched in ZEB1 motifs (Fig. 4 L). Indeed, in Zeb1-KO MEFs, Arhgef7, Pak4, and Wasf2 were found to be directly regulated by ZEB1 (Fig. 4 M). Given these findings, we considered actin regulation as a promising mechanism that could explain why WT LSKs regenerated in the Wnt5a-haploinsufficient niche lose their stemness.

The actin-driven cellular polarization is at the basis of migratory processes required for HSC engraftment. To interrogate these processes, we compared the general mechanical properties, adhesion, chemotaxis, and homing properties with LSK-WT and -5a cells (Fig. 5 A). We did not observe differences in the diameter and the cell stiffness of LSK-5a as detected with apparent Young’s modulus (Fig. 5, B and C). CXCL12 promotes F-actin polarization (Voermans et al., 2001), but did not rescue the apolar distribution of F-actin in LSK-5a cells (Fig. 5, D and E).

To determine functional consequences of reduced actin assembly, in a first experiment, we found that sorted LSK-5a cells only poorly adhered to VCAM1-Ig–coated slides (Fig. 5 F). In a second experiment, we found that although the starting number and composition of Lineage-depleted donor cells from WT and Wnt5a^+/− recipients was similar, migration of the LSK-5a cells and Lin⁻ and MPs from Wnt5a^+/− recipients toward CXCL12 was profoundly less efficient than their WT counterparts in a Boyden chamber assay (Fig. 5 G).

To find out whether in vivo responses of LSK-5a were affected in a manner similar to the in vitro adhesion and migration, we injected lineage-depleted donor cells recovered from WT and Wnt5a^+/− recipients into irradiated WT recipient mice. In these experiments, we found strongly reduced homing of Lin⁻ Kit⁺, MP, and LSK donor subpopulations from Wnt5a^+/− recipients 16 h after injection (Fig. 5, H and I). This confirmed the migratory defects of hematopoietic cells regenerated in Wnt5a^+/− recipients in vivo.

To determine whether the remaining LSK-5a cells that do home to the marrow fail to engraft as a result of decreased cell division or reduced survival, we set up single-cell cultures (Wohrer et al., 2014; Istvánffy et al., 2015). These showed that cell division and survival of CD34^- LSK-5a cells were similar to CD34⁻ LSK-WT cells (Fig. 5, J and K). This experiment further showed that addition of rWNT5a does not enhance or reduce the clone size or the number of CD34⁻ LSK-5a cells dividing (Fig. 5, J and K). Interestingly, however, addition of rWNT5A increases differentiation of CD34⁻ LSK-5a cells, with almost complete loss of LSKs, whereas their WT counterparts preserved LSK cells (Fig. 5, L and M).

To assess whether WNT5A-mediated cross talk between Wnt5a-competent WT HSCs and whether the Wnt5a^+/− niche is responsible for the reduced engraftment in secondary transplants, we transplanted HSCs from Wnt5a^+/− donors into WT (HET/WT) or Wnt5a^+/− recipients (HET/HET; Fig. 6 A). This experiment showed that Wnt5a^+/− donor LSKs regenerated in Wnt5a^+/− recipients behaved essentially similar as LSK-5a cells in both primary and secondary transplants (Fig. 6, B and C), with a low detection of mostly myeloid engraftment in HET/HET mice 5 wk after secondary (2°) transplantation. Moreover, as in LSK-5a cells, the polar distribution of DOCK2, WAVE2, and F-actin was also decreased in Wnt5a^+/− CD34⁻ LSK cells regenerated in Wnt5a^+/− recipients (HET/HET; Fig. 6, D–F).

To find out whether Wnt5a deficiency affects leukemogenesis, we studied a model of BCR-ABL^p185–induced leukemia (Kelliher et al., 1991). Neonatal WT mice transplanted with BM cells transduced with an e1a2 BCR-ABL^p185 expression vector succumbed to lympho-myeloproliferative disease (LMPD) characterized by increasing cell numbers in the periphery and splenomegaly with a lag time of 58 d (Fig. 7, A–D). The juvenile LMPD mice mostly suffer from a mixed lymphoid-myeloid (42%), lymphoid (42%), and in remaining cases a myeloid leukemia (16%). In neonatal Wnt5a^+/− recipients, we observed a significant number of mice (12/28, 43%) which did not develop leukemia. These disease-free recipients showed essentially similar numbers of GFP⁺ blood cells and spleen weights as the MSCV-IRES-GFP backbone control vector-transduced donor cells (Fig. 7 D). Sixteen Wnt5a^+/− recipients transplanted with BCR-ABL^p185+ cells developed a lethal LMPD with a high GFP⁺ cell engraftment, and higher PB count and more pronounced splenomegaly than leukemic WT recipients (Fig. 7 D).

To assess leukemogenic potential of cells from the primary recipients, we transferred 1 × 10⁶ GFP⁺ spleen cells into secondary WT recipients (Li et al., 1999). These secondary recipients from BCR-ABL^p185 WT cells all caused secondary LMPD with a lag time of 38 d (Fig. 7 E). In contrast, only one of eleven secondary recipients of BCR-ABL^p185+ cells from leukemic Wnt5a^+/− recipients was able to transfer leukemia (Fig. 7 E) and failed to engraft a significant number of GFP⁺ cells (not depicted).

To determine whether BCR-ABL^p185+ cells from Wnt5a^+/− recipients show similar defects in actin regulation as LSK-5a cells, we isolated GFP⁺ from the BM of primary recipients with LMPD to study cellular signaling and their migratory properties. As we were unable to isolate sufficient GFP⁺ LSK cells from the BM for functional analyses (Fig. S3), we studied GFP⁺ B cells from the BM of LMPD mice instead. Immunofluorescent analysis showed increased expression of CDC42, WAVE2, and F-actin, as well as loss of CDC42, WAVE2, and F-actin polarization in BCR-ABL⁺ cells from Wnt5a^+/− recipients (Fig. 7 F). Unlike normal LSK-5a cells, the polarized expression of DOCK2, was not affected (Fig. 7 G). In contrast to normal LSK-5a cells, and in accord with the more severe LMPD in Wnt5a^+/− recipients (Fig. 7 D), in leukemic cells we find an increased expression of the canonical Wnt mediators DVL2 and β-catenin (Fig. 7 H).

To find out whether the changes in polarized localization of F-actin, CDC42, and WAVE2 have similar functional consequences as in LSK-5a cells, we set up experiments to study actin-dependent cellular responses on leukemic B cells. These experiments demonstrated that leukemic GFP⁺ cells show reduced CXCL12-stimulated F-actin polarization (Fig. 7 I), but that adhesion to VCAM-1 is only marginally reduced (Fig. 7 J). Yet, GFP⁺ BCR-ABL^p185+ B cells show reduced chemotaxis toward CXCL12 (Fig. 7 K). Thus, our results show that similar to LSK-5a cells, leukemic BCR-ABL^p185+ B cells formed in Wnt5a^+/− recipients demonstrate defective actin regulation, and fail to transfer BCR-ABL^p185+ LMPD.

One of the major goals in stem cell biology is to understand the factors involved in successful engraftment of stem cells and how these cells regenerate tissues. This process requires for the stem cells to find their anatomical niche, lodge there, and start regeneration of all cell types of the tissue. In transplantation settings, single HSCs efficiently lodge in the BM, and can repopulate the entire blood cell system. Our data show that the microenvironment significantly regulates the quality of the HSCs that are regenerated. We demonstrate that although HSCs efficiently engraft Wnt5a-haploinsufficient recipients, they are conditioned to show alterations in their transcriptome. In particular, we show dysregulated expression of several genes involved in the small GTPase-dependent actin polymerization pathway. These changes in gene expression result in reduced ability to form polarized F-actin bundles, leading to defects in adhesion, migratory behavior, and homing to the BM. Our results further suggest that the dysregulated gene expression in the Wnt5a-haploinsufficient environment is caused at least in part by up-regulated expression of targets of the transcription factor ZEB1.

Our study shows that the noncanonical Wnt5a not only signals through small GTPases but also regulates the amplitude of their response by up-regulating the expression of mediators of the small GTPase-dependent actin-regulatory pathway. Our results support the view that in LSK-5a this pathway is overactive, as we observed that F-actin polarization is defective as a result of CDC42 activation (Florian et al., 2012). Our results further show that this activation cannot be overcome by CXCL12 stimulation, which normally increases F-actin polarization (Voermans et al., 2001). The experiments reported here underscore the importance of the small GTPase pathway for different subprocesses of HSC engraftment: adhesion, migration, and homing. It has previously been shown that RAC1 (Cancelas et al., 2005) and CDC42 (Yang et al., 2007a) regulate HSC engraftment through acting on cell shape, localization and retention, and mobilization. Not only are the small GTPases themselves important, but upstream factors like DOCK2 (Kikuchi et al., 2008), ARHGAP5 (p190-B; Xu et al., 2009), ARHGEF7 (β-PIX; Reddy et al., 2016), or PTPRS (Quarmyne et al., 2015), or downstream factors such as PAK2 (Dorrance et al., 2013) and WASF2 (WAVE2; Ogaeri et al., 2009) also affect engraftment. Our study confirms these studies and adds that the niche regulates the generation of functional HSCs by modulating the activity of this pathway in HSCs not only through secretion of ligands such as the chemokine CXCL12 but also through modifications in the expression level of critical mediators of the pathway, such as β-PIX, DOCK2, and WAVE2.

We further show that the effects of altered gene expression of regenerated cells in a Wnt5a-haploinsufficient environment also affect BCR-ABL^p185+ cells, which, in 42% of the studied mice, fail to generate leukemia, and, in the remaining cases, fail to transfer leukemia to secondary mice. Similarly to the normal cells, BCR-ABL^p185+ cells from Wnt5a-haploinsufficient recipients show defects in the actin-regulation pathway and F-actin polarization, resulting in defective migration and secondary engraftment. The GTPase-dependent actin regulatory pathway has been described mostly to affect BCR-ABL^p210-driven leukemogenesis, which, in contrast to BCR-ABL^p185, comprises the Bcr guanine exchange domain (Tala et al., 2013). It has been shown that RAC1 deficiency depletes BCR-ABL^p210 progenitors (Sengupta et al., 2010), and that the RAC1 and CDC42 exchange factor VAV drives leukemogenic proliferation and survival (Chang et al., 2012), as well as motility (Daubon et al., 2008). These studies and our findings suggest that coupling of small GTPase signals to actin polarization, which is required for directed migration may be a target to reduce leukemogenesis or to prevent reengraftment of leukemia after therapy, not only as has been suggested for BCR-ABL^p210 (Thomas et al., 2007) but also for BCR-ABL^p185.

Our study expands on current concepts of how the niche regulates the engraftment potential of HSCs, by showing that several different signaling mechanisms are simultaneously regulated. We and others have previously shown that secreted niche factors regulate engraftment through their effect on HSC cell cycle progression and self-renewal (Schreck et al., 2014). We here add to the concepts of niche regulation and demonstrate that the niche has a major extrinsic impact on the gene expression during the regeneration of the HSCs pool. In particular, reduced microenvironmental WNT5A causes defects in actin polymerization and cellular processes dependent on this signaling pathway. These findings with HSCs regenerated in a Wnt5a-haploinsufficient are reminiscent of aged HSCs, in particular with regard to reduction of short-term homing and LT-HSC frequency (Dykstra et al., 2011), as well as depolarization of F-actin localization and CDC42 activation (Florian et al., 2013). Our results, which reduced ability of HSCs to produce WNT5A does not rescue their failure to engraft secondary recipients, strongly suggest that although regeneration of hematopoiesis in a Wnt5a-haploinsufficient environment shows similarities with aged mice, in regeneration, the niche appears to play a dominant role. Indeed, in previous studies comparing transcriptomes of young and old mice, the actin-regulatory pathway was only a minor dysregulated pathway (Sun et al., 2014), thus revealing the clear differences in regenerative and aging processes. A common denominator in both is CDC42 activation. However, the mechanisms through which this small GTPase is activated may differ and defining the different small GTPase activation pathways may offer new ways to improve tissue regeneration.

Our results show that reduction of WNT5A from the niche conditions HSCs during regeneration that this is not affected by cross-talk of HSCs which may up-regulate Wnt5a, as Wnt5a-haploinsufficient HSCs behave similar to WT HSCs. Interestingly though, in culture, the conditioned LT-LSK-5a are not rescued by addition of WNT5A. Instead, these cells almost completely differentiate into mature cell types, rapidly depleting LSK cells. Because WNT5A receptors and downstream mediators show no alterations in their expression, we hypothesize that the observed differentiation may be linked to actin regulation. Indeed, CDC42 is required for B cell differentiation (Guo et al., 2009) and regulates the balance between myelo- and erythropoiesis (Yang et al., 2007b). The related RAC1 GTPase is crucial for T-lymphopoiesis (Guo et al., 2008). Also, in Wiskott-Aldrich syndrome (WAS), a recessive disorder of WAS-family proteins, such as WAVE2, terminal differentiation of multiple blood cell lineages is reduced (Kajiwara et al., 1999). These findings suggest that study of the actin-regulatory pathway may lead to targets that can alter terminal differentiation outcomes.

Dysregulated actin polarization is not restricted to normal HSCs but also affects mature B cells and BCR-ABL^p185+ leukemic cells. Indeed, despite the fact that BCR-ABL is a strong oncogene, leukemic cells generated in the Wnt5a-haploinsufficient environment do not transfer leukemia, indicating that the actin-regulatory pathway is a possible therapeutic target to prevent reengraftment of hosts after initial therapy. We further present evidence suggesting that alterations in HSC gene expression by environmental Wnt5a may be caused by increased use of sequence motifs bound by the transcription factor ZEB1. In summary, our results reveal a novel mechanism by which the niche preserves HSC behavior required for successful engraftment. Extrinsic conditioning of gene expression patterns aids HSCs in their ability to adhere, migrate, and find their niches in times of stress. Our results further suggest that modulation of up- or downstream members of the actin-regulatory pathway could improve the regeneration of a functional stem cell pool.

129S2 mice were obtained from Charles River and C57BL/6.J (B6, CD45.2) and B6.SJL-Ptprca.Pep3b/BoyJ (CD45.1) were obtained from Taconic. Wnt5a^+/− (Yamaguchi et al., 1999) mice obtained from The Jackson Laboratory were backcrossed to the (129S2 x C57BL/6.J)F1 (129B6, CD45.2) background for at least six generations. Age- and gender-matched Wnt5a^+/+ (WT) littermates were used as controls in all experiments. In transplantation experiments, first-generation crosses of 129S2 and B6.SJL-Ptprca Pepcb/BoyJ (129Ly5.1, CD45.1xCD45.2) were used as recipients in intrinsic transplantation assays and as donors in extrinsic transplantations. All animal experiments were approved by the section Consumer Protection, Veterinary Services, and Food Hygiene of the Government of Upper Bavaria (Regierung Oberbayern, Munich, Germany). All animals were housed for at least a week before experimental use in microisolators under specific pathogen–free conditions, according to the Federation of Laboratory Animal Science Associations and institutional recommendations.

Lin⁻ SCA-1⁺ Kit⁺ (LSK) cells and further CD34⁻ and CD150⁺ subpopulations were isolated as described previously (Renström et al., 2009; Istvanffy et al., 2011). Surface antigens were stained with antibodies from eBioscience (Natutec), except for PE-Cy5.5-streptavidin conjugate, which was obtained from Invitrogen (Tables S1 and S2). FACS analyses were performed on a CyAn ADP Lx P8 (Coulter-Cytomation). Data were analyzed with FlowJo software (Tree Star). Sorting of cell populations was done with a MoFlo High Speed cell sorter (Beckman Coulter).

The intracellular staining was performed on endosteal cell fraction isolated from the four long bones, as previously described (Ruf et al., 2016). Antibodies used are listed in the Table S1. Stained cells were analyzed on CyAn ADPLx P8 (Beckman Coulter) flow cytometer.

8-µm bone sections of frozen femurs were prepared using the Kawamoto tape method (Kawamoto, 2003). In brief, bones were excised from mice, fixed, decalcified, and placed in sucrose before embedding in OCT and freezing. 8-µm sections of frozen femurs were then cut using the Leica tape system, stained with indicated antibodies overnight at 4°C, and images were acquired using an LSM710 microscope. Images were prepared using ImageJ (National Institutes of Health) software.

RNA-sequencing data from donor-derived LSK cells was generated using the Smart-seq2 protocol and niche cell data were generated using UMI-seq, as described previously (Parekh et al., 2016). To remove noise from lowly expressed genes, count datasets were subjected to data-driven gene filtering using the HTSFilter R package (Rau et al., 2013; Tables S3 and S5).

Differential expression (DE) analysis was done by applying the DESeq2 package (Love et al., 2014) using the likelihood ratio to test for differential expression in 14,029 genes that fulfilled the test assumptions. Using an FDR <5% (Benjamini, 1995) as cut-off, 656 genes are differently expressed (Table S5). Overrepresentation of significantly DEGs in KEGG pathways (Table S6) was tested by a fixed network enrichment analysis implemented in the neaGUI R package (Alexeyenko et al., 2012). We used the topGO package (Alexa et al., 2006) for Gene Ontology (GO) enrichment analysis. Variance stabilized counts of all DEGs were plotted as heat map normalized to the mean value of LSK-WT cells. ISMARA analysis (Balwierz et al., 2014) was done using default parameters, but without miRNA motifs because miRNAs had not been measured in our samples. Enrichment analysis (Fig. S3 G) was done using a Fisher’s exact test categorizing the ISMARA annotated promoters by whether they are part of the actin regulation KEGG pathway and whether they are regulated by the tested motif.

For Wnt5a expression analysis, Western blotting was performed with lysates from BM primary stroma of Wnt5a^+/+ and Wnt5a^+/− mice containing 10% glycerol, 25 mM sodium fluoride, 1 mM sodium orthovanadate, and a protease inhibitor cocktail (Roche). After blotting the proteins on polyvinylidene difluoride membranes (EMD Millipore), 2% BSA in TBS with 0.1% Tween-20 was used for blocking and dilution of anti-Wnt5a antibody. Primary antibody was detected with an anti–rabbit horseradish peroxidase-conjugated secondary antibody and developed with the use of Super Signal chemiluminescent substrates.

Relative levels of GTP-bound CDC42 were determined by an effector pull-down assay. In brief, lineage-depleted BM cells (10⁶) were lysed in a Mg²⁺ lysis/wash buffer (EMD Millipore), including protease inhibitors as in the previous paragraph. Samples were incubated with PAK-1–binding domain/agarose beads (EMD Millipore), and bound (activated) and unbound (nonactivated) CDC42 fractions were probed by immunoblotting with an anti-CDC42 antibody (rabbit polyclonal; EMD Millipore). Activated protein was normalized to total protein (Stain Free System; Bio-Rad Laboratories) and the relative amount was quantified by densitometry (ChemiDoc Imaging System; Bio-Rad Laboratories).

As described previously (Renström et al., 2009; Istvanffy et al., 2011; Istvánffy et al., 2015), 500–1 × 10³ sorted LSK-WT and LSK-5a cells or donor GFP⁺ B-cells were spotted on poly-l-lysine-coated slides. In some experiments, 400–1,000 CD34⁺ or CD34⁻ LSK from different transplantation experiments were used. Cells were fixed with 4% paraformaldehyde in PBS, blocked with 10% FCS/0.1% Triton-X in PBS, and stained with primary and secondary antibodies listed in Tables S1 and S2. All stains were counterstained with SlowFade Gold Antifade Reagent with DAPI (Invitrogen). Staining was assessed using on a DM RBE fluorescent microscope (Leica). Fluorescence intensities of stained cells were quantified in total pixels using ImageJ. Each stain included a negative Ig control, the detected pixels of which were deducted from the total pictures as background.

The number of colony-forming cells was determined with culture in growth factor–supplemented methylcellulose medium as described by the manufacturer (MethoCult GF M3434; StemCell Technologies).

LSK cells adhering to a rhVCAM-coated surface were probed using an AFM cantilever equipped with a 2.5-µm-radius bead. Experiments were conducted in absence ("unstimulated") or presence ("+rmCXCL12") of 150 ng/ml murine rCXCL12 (R&D Systems). Apparent Young’s Moduli for individual cells (from fits to the Hertz model, averaged for repeated measurements) are presented as dot plots. The number of probed cells is indicated. Mean apparent Young’s Moduli of the probed cell population are denoted by horizontal lines. Cell diameters were determined from phase contrast images and presented as boxplots with 25th, 50th, and 75th percentiles denoted as horizontal lines, and 10th and 90th percentiles as whiskers.

For adhesion assays, WT donor LSK cells regenerated in WT and Wnt5a^+/− primary recipients were sorted out after 16 wk, and 500 cells were spotted on slides coated with recombinant VCAM1 (R&D Systems) in PBS. The cells were stimulated 15 min after depositing with murine rCXCL12 (150 ng/ml: R&D Systems) and imaged on a DM RBE fluorescent microscope (Leica). Adhesion was calculated by measurement of light incidence with ImageJ.

A total of 600 µl IMDM supplemented with 2% FCS containing 150 ng/ml murine rCXCL12 was added to the lower chamber of a 24-well 5 µm transwell filter (Corning). Sorted Lin⁺ cells (2 × 10⁵) in 100 µl IMDM were loaded to the upper chamber and were allowed to migrate for 4 h at 37°C. Migrating cells were collected from the lower chamber, stained for mature hematopoietic markers CD4, CD8a, CD11b, B220, and Gr-1, and counted using a CyAn ADP Lx P8 flow cytometer.

For this assay, 1,000 donor LSKs together with 30,000 donor MP cells sorted from the BM of WT and Wnt5a^+/− recipients, 16 wk after primary transplantation, were injected i.v. into secondary WT (C57BL/6.J) recipients. 16 h after i.v. injection, the secondary recipient mice were sacrificed and the BM were harvested. The homing ability of the WT cells was measured by quantifying the content of CD45.1⁺ donor cells by flow cytometry.

Competitive repopulation was performed using transplantation of donor cells into lethally irradiated recipient mice, as described previously ( Istvanffy et al., 2011; Istvánffy et al., 2015). For secondary and tertiary transplants, 129Ly5.1 (CD45.1xCD45.2)⁺ Lin⁻ SCA-1⁺ KIT⁺ CD34⁻ CD150⁺ donor BM cells regenerated in primary recipients were used as donor cells, which were injected into lethally irradiated 129B6 (CD45.2) WT mice with freshly isolated 129B6 competitor cells (BM, 100,000; Spleen, 500,000).

Unless otherwise indicated, the functional biological and the biochemical data are presented as the mean the standard error associated with the mean. In these experiments, the two-tailed Student’s t test or the Mann-Whitney U test for independent samples (MWU test) with a level of significance set at 0.05 were performed for two-group comparisons of the differences between the samples under study. In the analyses of the RNA-seq data, we used the FDR to control for the expected proportion of false discoveries among the rejected hypotheses (Benjamini, 1995).

Sequence reads from the RNA sequencing can be accessed at the Gene Expression Omnibus under accession no. GSE84993.

Fig. S1 shows gating of BM staining of primary, secondary, and tertiary transplantations (related to Fig. 1). Fig. S2 show gating of single-cell culture of donor CD34⁻ LSK cells isolated from transplanted WT and Wnt5a^+/− recipients (related to Fig. 5). Fig. S3 shows flow cytometry of primary BCR-ABL^p185-leukemic mice (related to Fig. 7). Tables S1–S6 are available as Excel files. Table S1 lists primary antibodies used in this study. Table S2 lists secondary reagents and antibodies used in this study. Table S3 is a comparison of mapped RNA sequences from MSC-WT and MSC-5a. Table S4 lists GO enrichment analysis of MSC transcriptomes from transplanted recipients. Table S5 is a comparison of mapped RNA sequences from LSK-WT and LSK-5a. Table S6 lists KEGG pathway enrichment analysis of the genes differentially expressed between LSK-WT and LSK-5a.

We thank Prof. Justus Duyster (1st Department of Medicine, University Clinic Freiburg, Germany) for the generous gift of the MIG control and the BCR-ABL1^p185 expression vectors. This project was supported by the Deutsche Forschungsgemeinschaft (DFG; OO 8/5-2, OO 8/9-1, FOR 2033 [projects A1, A3, A4, and B3], and SFB 1243 [projects A09 and A14]), and the German José Carreras Leukemia Research Foundation (DJCLS R 11/12).

The authors declare no competing financial interests.

Author contributions: C. Schreck, R. Istvánffy, and C. Ziegenhain designed and performed experiments, collected and analyzed data; T. Sippenauer, F. Ruf, F. Gärtner, M.C. Florian, N. Mende, A. Taubenberger, A. Prendergast, S. Grziwok, C. Pagel, and L. Henkel performed the experiments; B. Vieth and W. Enard analyzed data; K.S. Götze, J. Guck, S. Massberg, M. Schiemann, M. Essers, C. Waskow, and W. Enard provided infrastructure and contributed critical reagents/materials/analysis tools; C. Peschel, W. Enard, and R.A.J. Oostendorp designed research; and C. Schreck, R. Istvánffy, and R.A.J. Oostendorp wrote the manuscript.