GPR56 contributes to the development of acute myeloid leukemia in mice

Abstract

The G protein-coupled receptor 56 (GPR56) was identified as part of the molecular signature of functionally validated leukemic stem cells isolated from patients with acute myeloid leukemia (AML). This report now demonstrates particularly high expression of GPR56 in patients with mutant NPM1 and FLT3-length mutation and association of high GPR56 expression with inferior prognosis in a large patient cohort treated in two independent multicenter phase III trials. Functional relevance of GPR56 expression was validated in mice, in which co-expression of Gpr56 significantly accelerated HOXA9-induced leukemogenesis and vice versa knockdown of Gpr56 delayed onset of HOXA9/MEIS1-induced AML. Overexpression of Gpr56 grossly changed the molecular phenotype of Hoxa9-transduced cells affecting pathways involved in G protein-coupled receptors (GPRCs) and associated intracellular signaling. Blockage of surface GPR56 by an anti-GPR56 antibody successfully impaired engraftment of primary human AML cells. In summary, these data demonstrate that high expression of GPR56 is able to contribute to AML development and characterize the GPR56 as a potential novel target for antibody-mediated antileukemic strategies.

Introduction

Acute myeloid leukemia (AML) is still a disease with dismal outcome despite therapeutic approaches, such as dose intense polychemotherapy and allogeneic transplantation. It is well accepted that a major cause for treatment failure lies within the comparably small leukemic subpopulation called AML stem cells, known to be responsible for the propagation of the disease and thought to be the major reservoir of chemoresistant and relapsing leukemic cells. These leukemic stem cells (LSCs) reside at the top of the malignant cellular hierarchy in AML patients, forming its bulk progeny, which lacks leukemia-propagating capacity.^{1, 2} One of the major goals in AML research is to identify differences between LSCs and their normal counterparts, the hematopoietic stem cells (HSCs). Reports from many groups have identified differential expression of membrane, cytoplasmic or nuclear proteins cell surface antigens between LSCs and HSCs, among them drug transporters, cytokine receptors, members of signal pathways such as members of the WNT pathway or transcription factors such as homeobox genes (reviewed in Jordan,³ Felipe Rico et al.⁴ and Misaghian et al.⁵) Another concept to identify proteins, which are essential for LSC biology is to search for markers, which are preferentially expressed on AML LSCs but not on its downstream progenies, which have lost LSC properties. Following this concept, Eppert et al.⁶ analyzed gene expression in functionally validated primary AML LSCs in comparison to their CD34-negative progeny lacking engraftment potential in immunocompromised mice, by this defining a gene signature consisting of 42 genes associated with LSC properties and inferior treatment outcome in patients with AML. Among these genes, G protein-coupled receptor 56 (GPR56) was highly expressed in the LSC-enriched fraction compared with the non-engrafting leukemic bulk with a 25% increase in the expression in CD34⁺ LSC compared with the CD34⁻ non-LSC fraction.⁶ GPR56 itself is a G protein-coupled adhesion molecule able to interact with the niche by binding to collagen III,⁷ indicating a possible role for GPR56 in the crosstalk between LSCs and their microenvironment niche, shown to be essential for LSC properties.^{8, 9, 10, 11} Based on its expression in AML LSCs, its potential link to the LSC niche and also its accessibility for antibody-mediated blocking as a surface protein, we hypothesized that high expression of GPR56 contributes to AML development and that this surface protein might be a potentially novel target for antibody-based anti-LSC therapies.

Materials and methods

Patients

Mononuclear cells isolated from diagnostic bone marrow (BM) or peripheral blood (PB) with AML leukemias from 43 adult patients were analyzed (PML-RARa n=7, inv (16) n=8, NPM-WT/FLT3-ITD n=10, NPMC⁺ n=9 and complex karyotype n=9). CD34⁺ from BM mononuclear cells (Lonza, Cologne, Germany) (n=3) from healthy individuals were taken as controls. Cytochemistry and cytogenetics were performed in all cases as described. Cases were classified according to the French–American–British criteria and the World Health Organization classification.^{12, 13} The study was approved by the ethics committees of all participating institutions, and informed consent was obtained from all patients before they entered the study in accordance with the Declaration of Helsinki (http://www.wma.net/en/30publications/10policies/b3/index.html). The Kaplan–Meier survival plots for the patient microarray data (n=423) was generated using the survival package with R statistical computing environment.^{14, 15} Samples were assigned into high and low GPR56 expression groups based on their expression values above or below the overall median of the samples under analysis. Statistical significance between treatment groups was analyzed using log-rank (Mantel–Cox) test.

Plasmid constructs and viral production

MSCV-based bicistronic retroviral vectors were used to express target genes and SFLV lentiviral vectors were used to express short hairpin RNA (shRNA) target sequences. Human HOXA9 was cloned into MSCV-based vector expressing yellow fluorescent protein (YFP) as described previously.¹⁶ The MEIS1 construct was kindly provided by Keith R Humphries (Terry Fox Laboratory, Vancouver, BC, Canada). Mouse full-length Gpr56 cDNA was generated from RNA isolated by TRIzol (Invitrogen, Carlsbad, CA, USA) from a C57BL/6J mouse (University of Ulm, Ulm, Germany) using the PrimScript RT Reagent Kit (Takara Bio, Saint-Germain-en-Laye, France) according to the manufacturer's instructions. Murine Gpr56 was amplified from C57BL/6J cDNA (Forward: 5′-ATGGCTGTCCAGGTGCTGC-3′; Reverse: 5′-TTAGATGCGGCTGGAGGAGGTG-3′) subcloned into pGEM-T-Easy (Promega, Mannheim, Germany) and cloned into the NotI site of MIEG3 upstream of the internal ribosome entry site (IRES) and enhanced green fluorescent protein (eGFP). The miR30-based GPR56 targeting sequences (Supplementary Table S1) were PCR-cloned into the XhoI and EcoRI sites of SFLV-BFP (blue fluorescent protein) vector (kindly provided by Lenhard Rudolph, Jena, Germany).

Phoenix-ECO 293 T cells (ATCC, Wesel, Germany) and Lenti-X 293 T cells were cultured in Dulbecco's modified Eagle's medium (Gibco, Life Technologies, Darmstadt, Germany) containing 10% fetal bovine serum, 50 μg/ml penicillin (Life Technologies, Life Technologies GmbH, Darmstadt, Germany) and 50 μg/ml streptomycin. Plasmids were transfected into cells using the calcium phosphate transfection method. To generate lentivirus, the transfection was supplemented with psPAX2 and pMD2.G (Addgene, Teddington, Middlesex, UK). Median shRNA-mediated knockdown efficacy was 72.33±18.12 (n=5) for sh1 and 35.29±11.16 (n=4) for sh2. Primers are given in Supplementary Table S1.

BM transplantation assay

All animal experiments were approved by ethical review permission. Animals were not randomized between the experimental groups. All mice in the experiments were used for analysis. Low-density BM cells were harvested from mice treated 4 days previously with 150 mg/kg body weight 5-fluorouracil (Medac GmbH, Wedel, Germany) and isolated by Ficoll-1083 (Sigma, Taufkirchen, Germany). Cells were cultured for 48 h in Dulbecco's modified Eagle's medium supplemented with 20% fetal bovine serum, penicillin/streptomycin (Life Technologies, GmbH), 100 ng/ml recombinant mouse stem cell factor, 10 ng/ml mouse interleukin 6 (IL-6) and 6 ng/ml mouse IL-3 (Immunotools, Friesoythe, Germany). BM cells were transduced with the MSCV-IRES-HOXA9/YFP, empty control vector or a combination of MSCV-IRES-HOXA9/YFP, MSCV-IRES-MEIS1/GFP, and corresponding empty control vectors (Supplementary Figure S1) on recombinant fibronectin, CH296 (Retronectin; Takara Bio) as previously described.¹⁷ Successfully transduced cells were sorted by YFP or YFP/GFP expression and co-transduced with the MSCV-IRES-Gpr56/GFP vector (Supplementary Figure S1) after 3 weeks. Double transduced cells were propagated for a further 3 weeks and then highly purified by fluorescence-activated cell sorter (FACS) before transplantation into animals. For the knockdown experiments with the SFLV shRNA/BFP construct, HOXA9/MEIS1-transduced cells were propagated in vitro for 3 weeks and then infected with the lentiviral shRNA constructs. Successfully transduced cells were sorted for BFP expression and injected highly purified in recipient mice 48 h after the end of shRNA transduction. Before transplantation, viability was assured by trypan-blue exclusion. In all, 4 × 10⁴ HOXA9/YFP cells double transduced with GPR56/GFP or vector control or 2 × 10⁴ triple transduced HOXA9/YFP, MEIS1/GFP and shRNA/BFP was combined with 4 × 10⁵ C3HxBoyJ BM helper cells. Chimerism of PB was determined in recipient mice every 3–4 weeks after transplant.

For competitive transplants, 2–4-month-old C3HxBoyJ mice were lethally irradiated (10.25 Gy) and subsequently transplanted. Freshly transduced and sorted cells were transplanted into mice in a 1:2 ratio (1 × 10⁵/2 × 10⁵) of GPR56/GFP, the empty control vector or double-transduced HOXA9/YFP with either GPR56 or vector control in combination with C3HxBoyJ BM helper cells.

Analysis of mice

The lineage distribution was determined by multi-color FACS analysis as described previously.¹⁸ Gr-1, Sca-1, Ter-119, CD3e, Mac-1, c-Kit and B220 antibodies were used for analysis (all eBioscience, Frankfurt am Main, Germany). For histological analyses, sections of selected organs were prepared and hematoxylin/eosin-stained by using standard protocols as described previously.¹⁶

Xenograft transplants

Xenograft studies using human cell lines and patient cells were conducted in 8–12-week-old female NSG mice (NOD.Cg-Prkdc^Scid Il2rg^tm1Wjl/SzJ) (Jackson Laboratory, Bar Harbor, ME, USA). Prior to transplant of primary AML samples, BM cells were thawed, stained for CD3 and CD19 and sorted for negative CD3/19 expression. NSG mice were sublethally irradiated and injected intraperitoneally with 20 mg anti-IVIG antibody (Privigen, CSL Behring GmbH, Marburg, Germany) 1 day prior to transplant. Sick mice or those >12 weeks were analyzed for engraftment by flow cytometry using anti-human CD45/CD33 (BD Bioscience, Heidelberg, Germany). Primary AML patient samples were sorted for stem and progenitor cell populations. Lymphoid-primed multipotent progenitors (LMPPs) (CD34⁺/CD38⁻CD90⁻/CD45RA⁺), granulocyte-macrophage progenitors (GMPs) (CD34⁺/CD38⁺/CD123^+/low/CD110⁻/CD45RA⁺) and CD34⁻ sorted AML cells were adjusted to identical numbers of cells within the different subpopulations. Between 2.9 × 10⁵ and 1 × 10⁶ sorted BM or PB cells were transplanted for the LMPP, GMP and CD34⁻ subpopulations. AML engraftment was defined by >1% leukemic engraftment, characterized by the individual leukemia-associated phenotype and the absence of CD3- and CD19-positive cells determined at 12 weeks posttransplant. For blocking studies, the MV4-11 cell line (DMSZ) was incubated with anti-Gpr56 antibody (clone H11, purified mouse monoclonal IgG1κ) in concentrations of 10 or 50 μg per 1 × 10⁶ cells or 50 μg isotype control (cat. no. MABN310, Millipore, Darmstadt, Germany) per 1 × 10⁶ cells for 2 h at 4 °C.¹⁹ In all, 1 × 10⁶ treated cells were subsequently injected intravenously into sublethally irradiated (3.25 Gy) NSG mice. FLT3-ITD⁺ NPM1 wild-type AML patient samples were blocked with 50 μg of the anti-Gpr56 antibody (clone H11) or 50 μg isotype control (Millipore) per 1 × 10⁶ cells for 2 h at 4 °C before transplant.

Western blottings

Western blottings were performed according to standard protocols. Briefly, total protein was isolated with RIPA buffer supplemented with phosphatase and protease inhibitors, and protein concentration was determined by the Bradford method. Lysates were separated on a 10% sodium dodecyl sulfate/polyacrylamide gel electrophoresis and transferred to a polyvinylidene difluoride membrane. The blots were probed with anti-GPR56 (clone H11, Millipore) or β-actin (Santa Cruz Biotechnology, Heidelberg, Germany) and corresponding horseradish peroxidase-conjugated antibody (Santa Cruz Biotechnology). Equal amounts of total protein loaded onto the gel were confirmed by the 42-kDa β-actin signal.

Colony-forming cell (CFC) assay

CFC assays were performed using fresh transduced BM cells (1 × 10³ cells) sorted for YFP/GFP. Cells were plated in 1 ml of methylcellulose medium (M3434, Stem Cell Technologies, Köln, Germany) or M3234 supplemented with 100 ng/ml mouse stem cell factor, 10 ng/ml IL-6, 6 ng/ml IL-3 (all cytokines from Immunotools) or in serum-free methylcellulose (M3236, Stem Cell Technologies). Cultures were plated in duplicate and incubated for 7 days at 5% CO₂, 37 °C. On day 7, colonies containing at least 50 cells were scored. CFCs from day 7 cultures were harvested and replated at 500 cells/well in 1 ml of methylcellulose medium, and secondary colonies were counted 7 days after replating.

Both cell cycle and apoptosis assays were performed after an 18-h serum starvation and subsequent 24-h serum activation. Cell cycle analysis was performed using the APC BrdU Flow Kit (BD Pharmingen, Heidelberg, Germany: cat. no. 51-900009AC), and apoptosis assay was performed by using the APC Annexin V Apoptosis Detection Kit (BD Pharmingen: cat. no. 556547).

Quantitative real-time PCR (qRT-PCR)

Total RNA was isolated using the TRIzol method and treated with DNase I. RNA was reversed transcribed into cDNA using the Prime Script RT Reagent Kit according to the manufacturer's protocols for use of random hexamer. Expression levels of human GPR56 and HOXA9 assayed by TaqMan qRT-PCR Taqman probes (Life Technologies GmbH) were used for human samples: GPR56:Hs00173754_m1, HOXA9:Hs00365956_m1, Human TBP Endogenous Control, and Human ActB Endogenous Control. The relative expression of each gene to the house keeping gene (TBP) was determined by calculating fold change (2^−ΔCt). Gpr56 expression in murine samples, long-term culture and sorted mouse subpopulations was determined via SYBR Green strategy (Life Technologies GmbH). Gene expression was quantified relative to a calibrator standard curve (C57Bl/6 total BM-derived cDNA) and normalized to an endogenous control. Primers used for quantification by SYBR Green: Gpr56 5′-GTGACTCTGCAGTGCGTGTTC-3′, 5′-TGACACCATCAGCACTGCAA-3′; and HPRT1 5′-CAGTCCCAGCGTCGTGATTA-3′, 5′-ATGACATCTCGAGCAAGTCTTTCA-3′.

RNA-Seq

RNA-Seq of HOXA9 and HOXA9/Gpr56 was performed in triplicates from two biologically independent samples. Libraries were prepared by using the TruSeq RNA Sample Preparation Kit (Illumina, San Diego, CA, USA). The samples were loaded on the flowcell using the cBot from Illumina and run on a Illumina HiSeq2000 as a paired-end 50 base pair read. After trimming Illumina sequencing adapters using trimm galore, high-quality raw Fastq files (phred score of 20 or higher) were aligned to the mouse genome version mm10 RefSeq using tophat, and differential expression analysis was performed using Cufflinks^{20, 21} and R packages.²²

Results

GPR56 is highly expressed in human AML and correlates with treatment outcome

As GPR56 was identified as part of an LSC-associated signature in human AML, we analyzed the expression levels in different human AML genotypes based on the TCGA data set:⁶ transcript levels varied between different AML genotypes and showed a characteristic expression pattern with the highest expression in patients with NPMc⁺/FLT3-ITD⁺ AML and the lowest expression in CBF leukemias. The difference between these two different groups were significant (median 19.6-fold; P<0.0003; Figure 1a). To extend these observations in the human AML bulk population, we analyzed 43 clinically and molecularly annotated AML samples of different genotypes by qRT-PCR, confirming the highest expression in NPM1 mutant/FLT3-ITD-positive AML in line with the TCGA data and lower expression in the CBF INV16 AML subtype with a 4.7-fold difference between the two genotypes. Of note, the expression of GPR56 was highest in normal CD34⁺ hematopoietic stem and progenitor cells, indicating that there is substantial, but not aberrantly high, GPR56 expression in human AML cells (Figure 1b). The high expression of GPR56 in normal HSCs for mouse and man was confirmed in published data sets²³ as well as in our own RT-PCR analyses on normal mouse BM (Supplementary Figures S2A–C). In addition, GPR56 expression was tested on functionally validated LSCs: for this, CD34⁺ leukemic cells were separated into fractions with LMPP- and GMP-like phenotype according to their immunophenotype and transplanted into NSG mice. These fractions that were able to induce leukemic engraftment in mice were defined as functionally validated LSCs. Of note, GPR56 was substantially higher expressed in the CD34⁺ LSC subpopulations with LMPP and GMP phenotype compared with their CD34⁻ leukemic progeny without LSC activity (Figure 1c).

Figure 1

Expression of GPR56 in AML. (a) Expression of GPR56 in primary patient samples (n=74) with different genotypes from the AML data set available in the TCGA database.³⁹ Significance refers to expression levels of FLT3-ITD⁺ NPM1C⁺ patients. Significance was determined by Mann–Whitney ***P<0.0005, **P<0.005, *P<0.05. (b) Expression of GPR56 was determined in primary AML patient samples by qRT-PCR. Data are represented as ΔCT (2^−ΔCT; (n=46). (c) Expression levels of GPR56 determined by RNA-Seq in highly purified functionally validated AML stem cells with LMPP and GMP phenotype and their CD34⁻ leukemic progeny lacking LSC activity in NSG mice (LMPP n=6, GMP n=8, CD34⁻ n=7). Survival (d) and event-free survival (e) of the total cohort of AML patients and survival (f) and event-free survival (g) of the subgroup of patients with normal karyotype treated in the randomized prospective clinical trials HD98A and HD98B according to the expression level of GPR56 determined by cDNA arrays. Median expression levels were defined as cutoff between low and high expression of GPR56. Patient characteristics are described in Frohling et al.,⁴⁰ Schlenk et al.⁴¹ and Buchner et al.⁴²

To test whether the expression levels of GPR56 correlated with treatment outcome, microarray-based GPR56 expression of 423 clinically and molecularly annotated newly diagnosed patients treated in two independent prospective clinical trials of the Austrian–German Study group (AMLSG) was correlated with event-free and overall survival. All the patients included into this analysis received standard treatment with two induction cycles containing standard-dose cytarabine and anthracyclines, followed by one consolidation cycle comprising intermediate-dose cytarabine. When the median expression level of GPR56 was taken as cutoff, high GPR56 expression was associated with inferior event-free and overall survival in the total cohort of patients (n=423) as well as in the patients with normal karyotype (n=184) (Figures 1d–g).

Taken together, these data suggested functional relevance of GPR56 expression in patients with AML.

GPR56 accelerates myeloid leukemogenesis in collaboration with HOXA9

To test the functional relevance of GPR56 in AML, first BM progenitor cells retrovirally engineered to express HOXA9/YFP were transduced with the MSCV-IRES-Gpr56/GFP or the empty MSCV/GFP vector (control). Overexpression was confirmed by western blotting at the protein level and on the level of surface expression by FACS (Supplementary Figures S3A and B). Strikingly, co-expression of Gpr56 with HOXA9 increased primary colony formation in vitro (n=3) (Figure 2a). To determine whether the increase in the number of CFC was due to an increase of Gpr56 stimulation by extracellular matrix proteins such as collagen III present in serum,⁷ the CFC assay was performed in serum-free methylcellulose. Of note, Gpr56 enhanced colony formation also under serum-free conditions (P<0.004), indicating that the effect of the protein was independent of extrinsic serum factors (Figure 2b). In contrast, Gpr56 did not increase the replating capacity of HOXA9-transduced cells (data not shown). To determine whether Gpr56 expression forced more cells into cycle, cells were serum starved for 18 h to synchronize cell in G₀ phase and then released back into cycle upon the addition of cytokines and fetal bovine serum. The percentage of cells in S phase 24 h after release was determined by a bromodeoxyuridine incorporation assay. Although Gpr56-transduced HOXA9 cells tended to have more cells in S phase and less in G₀/G₁, this trend was not significant. There was also no significant difference in apoptosis, measured by Annexin staining (data not shown). Importantly, co-expression of Gpr56 induced leukemia in transplanted mice after a median time of 148 days after transplant (range 93–264), in contrast to the HOXA9 control, which did not develop any signs of disease in the observation period of up to 380 days after transplant (Figure 2c). Leukemic mice were pale and lethargic and partly showed leukocytosis (Table 1). FACS analysis of the PB, BM and spleen showed an AML with high expression of Gr-1 and CD11b with negativity for lymphoid marker such as B220 (Figure 2d). Immunohistochemistry documented multi-organ infiltration by leukemic blasts and positivity for CAE, in line with the diagnosis of AML without differentiation following the Bethesda criteria for non-lymphoid malignancies in mice (Figure 2e).²⁴ Leukemias were transplantable with short latency (median latency 103 days). Importantly, the leukemogenic potential of Gpr56 was dependent on co-expression of HOXA9 as retrovirally induced constitutive expression of Gpr56 did not impact colony formation in vitro and hematopoietic differentiation in vivo (Supplementary Figures S4A–D; Supplementary Table S2). As we had seen that overexpression of Gpr56 accelerates HOXA9-associated leukemogenesis, we hypothesized that knockdown of endogenous Gpr56 would delay leukemia development. To test this, Gpr56 was depleted by shRNA-induced knockdown in HOXA9/Meis1-transduced cells before transplantation into mice. shRNAs had substantial different knockdown efficiencies, allowing to correlate readouts with the extent of knockdown (Supplementary Figures S5A and B). Onset of leukemia was significantly delayed by shRNA1 construct with nearly 100% knockdown (P<0.05) in contrast to shRNA2, which achieved <20% Gpr56 depletion compared with the shRNA control (Table 2).

Figure 2

Impact of Gpr56 overexpression on the clonogenicity of HOXA9-transduced hematopoietic cells determined in the CFC assay in (a) the presence of serum (n=3) (NS) or under (b) serum-free conditions (n=4). Successfully transduced cells were highly purified by FACS before plating into methylcellulose. Significance was determined by Mann–Whitney NS=P>0.05. (c) Kaplan–Meier survival curve for mice transplanted with HOXA9/Gpr56-transduced cells or HOXA9/GFP-transduced cells. Gpr56 accelerated disease onset (P=0.0001). Significance was determined by Mantel–Cox test. (d) Immunophenotype of leukemias in diseased mice. Shown are flow cytometric profiles of BM, spleen and PB from a representative HOXA9/Gpr56 leukemic mouse #4. Gating for GFP⁺ and YFP⁺GFP⁺ cells is indicated in the top panel. (e) Histological analysis of a representative HOXA9/Gpr56 diseased mouse #1. Hematoxylin and eosin stain (H&E), myeloperoxidase (MPO) and chloroacetate esterase stain (CAE) demonstrate cell infiltration in the liver and the white pulp of the spleen (positive for CAE and MPO and negative for the lymphoid markers CD3 and B220). Wright–Giemsa staining of cytospins and PB smears denotes the presence of blasts.

Table 1 Characteristics of mice overexpressing HOXA9/Gpr56 or HOXA9 alone

Table 2 Characteristics of mice depleted for Gpr56

Because increased Gpr56 expression accelerated myeloid leukemogenesis in vivo, we sought to determine whether Gpr56 affected the adhesion and homing rate of the transplanted BM cells. Adhesion to stroma was performed by modifying the CAFC (cobble-stone area-forming cell assay) adhesion assay as previously published.²⁵ Homing was determined by injecting HOXA9-GFP/GPR56-YFP- versus HOXA9/YFP-transduced cells intravenously into lethally irradiated mice and determining the percentage of GFP/YFP-positive cells in the BM of recipient mice 18 h after transplantation by FACS. However, co-expression of Gpr56 with HOXA9 had no impact on homing in vivo or on adhesion in vitro, despite its significant effect on leukemogenesis (data not shown).

In line with its effect on colony formation in serum-free conditions, this suggests that cell intrinsic effects mediated by Gpr56 overexpression have a major role in accelerating HOXA9-induced leukemia.

Overexpression of Gpr56 induces gross changes in the molecular phenotype of HOXA9 leukemic cells

To characterize the cell intrinsic impact of constitutive expression of Gpr56 in the context of HOXA9 co-expression in more detail, RNA-Seq was performed. Comparison of the molecular signature of BM cells expressing HOXA9/GFP versus HOXA9/Gpr56 revealed 867 differentially expressed genes (P<0.05, false-discovery rate q<0.05). Of the 867 genes, 166 could be mapped to specific pathways using Panther pathway analysis.²⁶ In the category ‘molecular function’, the subcategories ‘catalytic activity’, ‘binding’ and ‘receptor activity’ scored most (Figure 3a). Within the ‘receptor activity’, nearly half of the genes belonged to ‘G protein-coupled receptor activity’ (Figure 3b). Among the top pathways affected were the ‘Wnt pathway’, the ‘Inflammation mediated chemokine and cytokine signaling pathway’ and the ‘Integrin’ pathway (Supplementary Table S3, Supplementary Figure S6). At the level of individual genes, Xlr3b belonged to the top 10 upregulated genes, originally described for its expression in plasmacytoma cell lines.²⁷ Another significantly upregulated gene was Epcam, associated with ‘stemness’ and Evi-associated transformation.²⁸ Among the top downregulated genes, several genes are associated with cell adhesion processes, such as Vcam-1, the claudin gene Cldn11, involved in tight junction formation and function and belonging to a gene family often deregulated in human cancers²⁹ and the homeobox gene Barx2³⁰ (Supplementary Table S4).

Figure 3

Panther analysis for pathways differentially expressed between BM cells expressing HOXA9/Gpr56 compared with HOXA9 alone. (a) Histogram presenting Panther molecular function of genes that are differentially expressed. (b) Dissection of the three top differentially expressed ‘Panther molecular functions’ into subcategories according to differential gene expression in BM cells expressing HOXA9/Gpr56 compared with HOXA9 alone.

Antibody-mediated blockage of GPR56 impairs human AML engraftment

As GPR56 is part of the LSC signature in human AML associated with inferior prognosis and as we had shown that GPR56 in collaboration with the oncogene HOXA9 contributes to myeloid leukemogenesis, we sought to determine whether the surface receptor GPR56 can be used as a target to impair leukemic engraftment in the NSG mouse model. MV4-11 cells are known to overexpress HOXA9 and were chosen as an AML cell line model: blockage of GPR56 with a monoclonal naked antibody significantly impaired engraftment into NSG mice (Figure 4a). Importantly, also FLT3-ITD3⁺ NPM1c⁻ primary patient samples showed a decrease in BM engraftment at 12 weeks when treated with the anti-GPR56 antibody when compared with isotype control with a 68% and 56% reduction in engraftment for patients 1 and 2, respectively (P<0.04; Figure 4b).

Figure 4

(a) Kaplan–Meier survival curve for mice transplanted with MV4-11 cells following antibody-mediated blockage of GPR56 compared with isotype control. Significance was determined by Mantel–Cox. (b) Percentage of CD45⁺ cells in the BM of NSG mice 12 weeks after transplant following antibody blockage of GPR56 compared with the isotype control. The P-value between the two groups for the total cohort of mice was <0.04.

Discussion

GPR56 belongs to the family of GPCRs, which is one of the largest families of proteins with a plethora of physiological functions. In this superfamily, GPR56 belongs to the so-called ‘adhesion family’, which is characterized by their large extracellular N-terminus, which shows homology to adhesion proteins.³¹ Gpr56 has been associated earlier with stem cell biology, first in the neurological system, showing high expression of Gpr56 in neural stem and progenitor cells, but downregulation on the cell surface in the more differentiated cells, such as astrocytes and βIII-tubulin neurons.³² In the hematopoietic system, Gpr56 was found to be important as early as day 10.5 in the mouse embryo.³³ RNA sequencing demonstrated that endothelial cells have low levels of Gpr56 expression, but the expression increased in HSCs and in embryonic cells transitioning from the endothelial to the hematopoietic system. Using a morpholino oligo knockdown approach in zebrafish, knockdown of Gpr56 showed a decrease in the number of HSCs emerging from the vasculature, thus suggesting that Gpr56 is required for the emergence of HSCs in the mouse embryo. In adult murine hematopoiesis, however, there are contradictory results with regard to the relevance of Gpr56 expression for HSCs: in a report by Saito et al.,³⁴ it was shown that Gpr56 is highly expressed in long-term (LT-HSC) and short-term (ST-HSC) HSCs but decreased in the more differentiated hematopoietic cells. The percentage of LT-HSCs and ST-HSCs was lower in the BM of Gpr56 knockout mice when compared with wild type, which was consistent with a decrease in the number of colony-forming units. Furthermore, HSCs from knockout Gpr56 mice showed an increase in migration toward an SDF-1α gradient in vitro but a decrease in adhesion to stroma cells or extracellular matrix proteins. In addition, the ability of Gpr56 knockout cells to reconstitute the BM was significantly decreased and BM analysis revealed fewer knockout cells located close to the periosteum when compared with wild type.³⁴ The stem cell-associated expression of Gpr56 was confirmed in a recent publication, but no impact of Gpr56 deficiency on maintenance and function of HSCs could be discovered.¹⁹ Importantly, GPR56 expression was part of a core signature of functionally validated LSCs. Of note, this 42 gene LSC signature was prognostic and patients showing high expression levels of LSC-genes had an inferior event-free and overall survival in retrospective analysis of cytogenetically normal AML.⁶ In a separate study, EVI-1 was shown to directly bind to the promoter region of GPR56. Knockdown of GPR56 in EVI-1-high AML cell lines decreased viability and rendered the cells more susceptible to chemotherapy drugs.³⁴ We now show that Gpr56 robustly contributes to the development of AML in mice, using a homeobox-driven AML model. Surprisingly, the effect of the surface receptor Gpr56 expression seemed not to depend on increased homing or adhesion. This was further underlined by increased colony formation in vitro under serum-free conditions after retrovirally engineered Gpr56 overexpression and large changes in gene expression of murine BM progenitor cells only 48 h after induction of retrovirally induced constitutive expression of the gene. This in line with findings that Gpr56 forms a complex with Gα_q/11 and G12/13 in neural progenitor cells.^{35, 36} This Gα_q/11-Gβ signaling pathway is part of both the ‘inflammation mediated chemokine and cytokine signaling pathway’ and the ‘CCKR pathway’, which belonged to the top 10 differentially expressed pathways after Gpr56 overexpression in murine progenitor cells in our experiments. This data set on changes in gene expression will provide a starting point for future analyses on the underlying mechanisms of GPR56-associated leukemogenesis and importantly to identify relevant downstream targets of GPR56. GPR56 has been linked to tumorigenesis before: GPR56 expression is upregulated in several cancers, such as that of lung, colon, glioblastoma and pancreas. Furthermore, it was shown to affect cell transformation and survival, metastasis and angiogenesis, partly interacting with the tumor microenvironment.^{37, 38} One of the major goals is to target GPRs involved in cancer. Our data could demonstrate, that GPR56 is highly expressed on AML bulk and AML stem cells and future studies will aim at understanding mechanisms determining GPR56 expression levels in human AML. As GPR56 is expressed on the cell surface, it is accessible by blocking antibodies. We could indeed show that the human AML cell line MV4-11 as well as primary AML patient samples are efficiently targeted by a blocking anti-GPR56 antibody, resulting in a major reduction of engraftment potential in NSG mice. These observations are encouraging, although future experiments will have to show whether normal human HSC engraftment in NSG mice depends on GPR to the same extent as human LSCs. The observation in mice that Gpr56 might be dispensable for normal HSCs¹⁹ would indicate a possible ‘therapeutic window’ between HSCs and LSCs, encouraging further investigations on a potential therapeutic role of anti-GPR56-directed antibodies in human AML. However, given the high expression of GPR56 on normal human CD34+ cells, antibody-mediated anti-GPR56 treatment might also affect normal hematopoiesis. Finally, the presented data demonstrate that the previously published molecular signature of AML LSCs⁶ includes genes that are functionally relevant and affect LSC behavior. This indicates that, beyond their potential role as prognostic score, LSC signatures are an important tool for the discovery of novel LSC genes.