Only Hematopoietic Stem and Progenitor Cells from Cord Blood Are Susceptible to Malignant Transformation by MLL-AF4 Translocations

Mixed lineage leukemia (MLL) (KMT2A) rearrangements (KMT2Ar) play a crucial role in leukemogenesis. Dependent on age, major differences exist regarding disease frequency, main fusion partners and prognosis. In infants, up to 80% of acute lymphoid leukemia (ALL) bear a MLL translocation and half of them are t(4;11), resulting in a poor prognosis. In contrast, in adults only 10% of acute myeloid leukemia (AML) bear t(9;11) with an intermediate prognosis. The reasons for these differences are poorly understood. Recently, we established an efficient CRISPR/Cas9-based KMT2Ar model in hematopoietic stem and progenitor cells (HSPCs) derived from human cord blood (huCB) and faithfully mimicked the underlying biology of the disease. Here, we applied this model to HSPCs from adult bone marrow (huBM) to investigate the impact of the cell of origin and fusion partner on disease development. Both genome-edited infant and adult KMT2Ar cells showed monoclonal outgrowth with an immature morphology, myelomonocytic phenotype and elevated KMT2Ar target gene expression comparable to patient cells. Strikingly, all KMT2Ar cells presented with indefinite growth potential except for MLL-AF4 huBM cells ceasing proliferation after 80 days. We uncovered FFAR2, an epigenetic tumor suppressor, as potentially responsible for the inability of MLL-AF4 to immortalize adult cells under myeloid conditions.


Introduction
The chromosomal translocations of the mixed lineage leukemia (MLL, KMT2A) gene are found in both acute myeloid leukemia (AML) and acute lymphoid leukemia (ALL) [1]. However, the frequency and the prognosis of KMT2A rearranged (KMT2Ar) leukemia differ with age and the reasons for that are only partially understood. In infants, KMT2Ar leukemia accounts for up to 80% of ALL with AF4 as the main fusion partner and is associated with a particularly poor prognosis [2]. The responsible genetic alterations occur in utero in a cell of fetal origin [3]. Whereas in adults, the majority of KMT2Ar leukemia is AML with AF9 as the main fusion partner being associated with an intermediate prognosis.
The translocation occurs most likely in hematopoietic stem or progenitor cells (HSPCs) derived from human bone marrow (huBM) possessing an inherent self-renewal capacity [1]. This is supported by the observation, that KMT2Ar leukemias often co-express myeloid and lymphoid markers suggesting that the cell of origin has to be a HSPC qualified to express both lineage markers [4]. Further evidence is supplied by the clinical observation that KMT2Ar ALL following CD19-targeted therapy can represent as AML in relapse [5]. Similarly, only HSPCs live long enough that additional genetic alterations, that may be required for KMT2Ar leukemogenesis, can accumulate in this cell type and its progeny. Whereas adult KMT2Ar leukemia often occurs in pretreated patients as secondary acute leukemia and additional cooperating mutations are mostly required for full leukemia transformation, infant KMT2Ar leukemias develop as de novo leukemia bearing only very few additional mutations [6][7][8]. Thus, the consequences of genetic changes in HSPCs derived from infant and adults leading to leukemia development are distinct in both cell types. In humanized mouse models, Horton et al. could show that human cord blood (huCB) cells were more susceptible for a retroviral MLL-AF9 immortalization, whereas retrovirally transduced huBM cells failed to immortalize in vitro and did not develop leukemia in vivo [9]. Similarly, by using mouse cells and an MLL-ENL fusion transcript Okeyo-Owuor et al. could demonstrate that the efficiency of MLL-ENL-driven AML changes with age with a peak shortly after birth [10]. In addition, the respective fusion partner has potential influence on leukemia development, as we and others could demonstrate its important role on the resultant phenotype: ENL exclusively led to ALL, whereas AF9 presented as AML, ALL or mixed phenotype in mouse xenograft models [11,12]. Moreover, the microenvironment plays a dominant role in instructing lineage fate [13]. In summary, these observations imply that crucial differences in leukemia development exist dependent on the fusion partner, the microenvironment and finally the cell of origin, in which the mutation develops. However, until now the performed studies were mainly based on artificial systems solely utilizing mouse cells or retroviral transduced oncogenes with unknown effects for the resultant human leukemias. In this study, we used CRISPR/Cas9 to introduce translocations of the MLL and AF4 or AF9 genes under physiologic promotors in both huCB and huBM cells, faithfully mimicking the patient nature of the disease.

CRISPR/Cas9
Demonstrates High Cutting Efficiencies and Induces t(9;11) and t(4;11) Chromosomal Translocations in Human HSPCs Derived from huBM Previously, we were able to introduce MLL-AF4 and -AF9 chromosomal translocations based on patient sequences in HSPCs (CD34 + ) derived from huCB in high frequency [1,14]. To translate our results in an adult system, we used HSPCs derived from huBM in comparison to huCB to evaluate whether the cell of origin and/or the fusion partner influence leukemia initiation. By nucleofecting plasmid-and virus-free single guide (sg) RNAs for the genes MLL, AF4 or AF9 with Cas9 protein in K562 cell line as proof-of-principle and in HSPCs derived from huBM, respectively, we demonstrated successful cutting efficiencies in both cell types ( Figure 1A). To induce t(9;11) and t(4;11) translocations in adult HSPCs, we isolated CD34 + cells and nucleofected them using Cas9 protein and sgRNAs targeting MLL and AF4 or AF9, respectively. Cas9 alone was used as control. Following nucleofection, the cells were maintained in liquid culture supplemented with cytokines and chemokines optimized for growth of KMT2Ar cells [15]. PCR analyses of genomic DNA revealed signals of MLL-AF4 translocations in three out of 10 and MLL-AF9 translocations in four out of eight performed experiments with different donors, demonstrating an easy translation of our previously used CRSIPR/Cas9-system to adult cells ( Figure 1B). Sanger sequencing revealed specific fusion sequences comparable to our huCB approach ( Figure 1C,D) [14]. These results demonstrate that we were able to induce MLL translocations with high frequency in HSPCs derived from huBM by using genome engineering.

Engineered Adult KMT2Ar Cells Are Characterized by KMT2Ar-Typical Gene Expression, Phenotype and Morphology
To characterize the engineered KMT2Ar huBM cells, we performed RT-PCR to determine the functional expression on RNA level revealing both MLL-AF4 and -AF9 fusion transcripts ( Figure 2A). Furthermore, we assessed the expression of common KMT2Ar-specific target genes like MEIS1 and HOXA9 in KMT2Ar huBM cells that were comparably high to the respective expression levels of patient-derived KMT2Ar cells ( Figure 2B).  MLL-AF4 and MLL-AF9 huBM cells were normalized to control cells (CD34 + huBM cells nucleofected with Cas9 alone) and compared to patient cells harboring t(4;11)(q21;q23) or t(9;11)(p22;q23), respectively. Experiment was performed in biological duplicates (n = 2) and horizontal bars represent the mean. Student's t test was used: * p < 0.05. Error bars indicate standard deviation (SD). (C) Representative contour plots of flow cytometry analyses of KMT2Ar huBM and control cells regarding myelomonocytic markers CD15, CD33 and CD64 as well as the expression of CD34 and CD38 after reaching purity are shown. (D) KMT2Ar huBM cells (black line) present with higher expression levels of immaturity marker CD117, lower expression levels of differentiation marker CD14 and higher expression of known KMT2Ar surface marker CD9 compared to control cells (gray shading) [15,16]. (E) Representative morphologies of KMT2Ar huBM and control cells are shown to display less cell differentiation and cell death of KMT2Ar cells. Scale bars define 20 µm.
Blast cells from KMT2Ar leukemia patients typically display a myelomonocytic phenotype and no expression of CD34 in contrast to the most non-KMT2Ar leukemia patients [16,17]. Likewise, the genome-engineered KMT2Ar huBM cells in our model expressed myelomonocytic markers like CD64, CD33 and CD15 and lacked CD34 expression on their surface ( Figure 2C). Interestingly, despite prolonged culture time of several weeks the KMT2Ar huBM cells expressed almost no CD14 as marker of differentiation, but CD117 as marker of immaturity, whereas control cells differentiated upon time in culture ( Figure 2D). Importantly, as further validation of our adult KMT2Ar model, the genome engineered KMT2Ar huBM cells expressed CD9 as typical KMT2Ar leukemic surface marker ( Figure 2D) [15,16].
To further characterize the KMT2Ar huBM cells, we performed cytospins followed by May-Gruenwald-Giemsa staining that revealed an immature morphology of the KMT2Ar huBM cells while the control cells presented a macrophage-like morphology and karyopyknosis indicating ongoing apoptosis ( Figure 2E). These results demonstrate that using CRISPR/Cas9 to induce MLL translocations in HSPCs derived from huBM leads to expression of the fusion transcript, upregulation of KMT2Ar-specific target genes, a myelomonocytic phenotype and immature morphology hereby authentically mimicking KMT2Ar leukemia; therefore, this model can be used as a reliable patient-derived in vitro model.

MLL-AF9 Can Immortalize Neonatal and Adult Cells, Whereas MLL-AF4 Only Immortalizes Neonatal Cells
Following nucleofection with the respective sgRNAs targeting MLL and AF4/AF9 and Cas9 protein, the huBM cells were kept in culture and monitored over time by PCR to detect the fusion gene. Both MLL-AF4 and -AF9 PCR products were detected with increasing signal intensity over time ( Figure Figure 3F). Collectively, these data indicate that MLL-AF9 can immortalize CD34 + cells of both neonatal as well as adult origin in vitro, whereas MLL-AF4 can solely immortalize cells of neonatal origin comparable to KMT2Ar patient leukemia, as the portion of MLL-AF4 leukemia in adults is very rare. These results indicate an important role of the cell of origin dependent on the expression of the respective fusion transcript for leukemia development.

Identification of Common KMT2Ar Target Genes and Uncovering of FFAR2 as Possible Intrinsic Factor Responsible for Cell Transformation
We performed RNA sequencing (RNA-seq) of the MLL-AF4/-AF9 huCB and huBM cells and control cells to shed light on the transformation potential of MLL-AF4 and -AF9 translocations in the different cell types. Genes changed specifically in adult huBM cells upon MLL-AF4 expression. This may provide insight into the mechanisms responsible for the observed phenomenon that solely huCB cells but not huBM cells were immortalized by the expression of the MLL-AF4 fusion transcript. We identified 335 and 502 differentially expressed genes (DEGs) in huBM cells upon MLL-AF4 and MLL-AF9 expression, respectively. Regarding huCB, we detected 654 DEGs upon MLL-AF4 and 939 DEGs upon MLL-AF9 expression, respectively ( Figure 4A). Of these genes, 73 were concordantly changed in all groups following MLL translocation and therefore represent a common KMT2Ar gene signature that was irrespective of tissue source ( Figure 4B). Interestingly, this signature comprised the upregulation of classical KMT2Ar target genes like HOXA9, 10, 10-AS and MEIS1 ( Figure 4C). These data confirm that both our CRISPR/Cas9-KMT2Ar models are physiologically relevant and that the inability of MLL-AF4 to immortalize adult cells is not explained by absence of KMT2Ar-induced expression of these common target genes. deviation (SD). (C) Representative contour plots of flow cytometry analyses of KMT2Ar huBM and control cells regarding myelomonocytic markers CD15, CD33 and CD64 as well as the expression of CD34 and CD38 after reaching purity are shown. (D) KMT2Ar huBM cells (black line) present with higher expression levels of immaturity marker CD117, lower expression levels of differentiation marker CD14 and higher expression of known KMT2Ar surface marker CD9 compared to control cells (gray shading) [15,16]. (E) Representative morphologies of KMT2Ar huBM and control cells are shown to display less cell differentiation and cell death of KMT2Ar cells. Scale bars define 20 µm.  To gain more insight into the hampered transformation potential driven by MLL-AF4 in huBM cells, we compared the expression profile from MLL-AF4 huBM cells against all others and revealed 45 differentially expressed genes ( Figure 5A). Strikingly, within these 45 genes, we uncovered FFAR2, also known as GPR43 or FFA2, as the most downregulated in all other KMT2Ar cells although it was less in MLL-AF4 huBM cells indicating a possible important role in the KMT2Ar transformation potential ( Figure 5B,C). To further confirm the RNA-seq results, we performed qPCR with the samples submitted to RNA-seq and further CRISPR/Cas9-KMT2Ar cells, KMT2Ar patient (UPN1) and non-KMT2Ar patient (UPN2) samples and revealed in all cases a significant downregulation of FFAR2 in contrast to healthy controls but again a specifically less downregulation in MLL-AF4 rearranged cells ( Figure 5D). To elucidate the impact of FFAR2 in cancer in general, we mined the literature and compared the expression level in different cancer entities. Strikingly, we discovered the lowest levels of FFAR2 expression in breast cancer, prostate cancer and hematological diseases especially in leukemia ( Figure 5E) [18]. This could be further confirmed by analyzing leukemia patient data of different entities, again showing very low levels of FFAR2 in contrast to healthy controls ( Figure 5F) [19]. Notably, after 62 days of cell culture, treatment with the FFAR2 antagonist GLPG0974 resulted in increased proliferation of MLL-AF4 huBM cells whereas MLL-AF4 huCB cells were not affected ( Figure 5G) demonstrating the mechanistic role of FFAR2 in transforming MLL-AF4 huBM cells [20].     Taken together, our data show that our CRISPR/Cas9-KMT2Ar models are authentic by demonstrating common KMT2Ar target gene expression that is not responsible for the immortalization. Moreover, we uncovered downregulation of FFAR2 as a potential key player in the KMT2Ar leukemogenesis.

Discussion
In this study we used CRISPR/Cas9 to generate t(9;11) and t(4;11) chromosomal translocations in primary human HSPCs derived from both huCB and huBM to unravel the differences on KMT2Ar leukemia based on the fusion partner and the cell of origin, in which leukemia initiation occurs. Our models are based on patient-specific sequences and share not only genetical but also morphological, phenotypical and transcriptomic attributes of KMT2Ar leukemias and hereby closely parallel the nature of the disease. Previously, we could demonstrate that by using CRISPR/Cas9 in huCB the creation of an authentic infant KMT2Ar model based on endogenous oncogene activation was possible to unravel the pathogenesis of KMT2Ar leukemogenesis in vitro [14]. In this study, we successfully transferred our genetic tool to HSPCs derived from huBM and could again demonstrate the feasibility to generate both MLL-AF4 and MLL-AF9 translocations with high efficiency in an adult system. To our knowledge, this is the first demonstration that both the induction of MLL translocation and the outgrowth of pure KMT2Ar cells in HSPCs derived from huBM is feasible. Interestingly, all generated cells except for MLL-AF4 huBM cells showed unlimited growth potential in in vitro cultures. Although MLL-AF4 huBM cells were similarly able to reach purity within 60 days and expressed KMT2Ar target genes as a hallmark of KMT2Ar leukemia, they succumbed to apoptosis in all performed experiments within around 80 days under myeloid culture conditions. This situation closely parallels human KMT2Ar patients, as the portion of MLL-AF4 myeloid leukemias in adults is negligible, whereas MLL-AF4 leukemia is the most occurring lymphoid leukemia in infants [1]. One could argue that MLL-AF4 cells become preleukemic in adults but full transformation to myeloid leukemia is not possible due to the cell of origin. Strikingly, prior to apoptosis our genome-engineered MLL-AF4 huBM cells lost their CD9 surface expression as typical KMT2Ar marker in in vitro cultures [16]. This indicates that CD9 loss is a potential early marker for detection of upcoming cell death upon targeted therapy. Further studies in KMT2Ar patients during anti-cancer treatment are necessary to confirm this phenomenon. Our RNA-seq data of the CRISPR/Cas9-KMT2Ar cells derived from huCB and huBM revealed a common KMT2Ar signature including the overexpression of known common target genes like MEIS1 and HOXA9 that was irrespective of the fusion partner or cell of origin. This confirmed that our genome-engineered models based on patient sequences are authentic and that the observed differences are independent from the KMT2Ar-typical target gene expression. Moreover, we could identify FFAR2 as the most downregulated DEG in MLL-AF9 huCB and huBM as well as MLL-AF4 huCB cells. In contrast, MLL-AF4 huBM cells failed to efficiently downregulate this gene indicating a major role for disease development. In addition, the fact that the FFAR2 antagonist GLPG0974 specifically favored the proliferation of MLL-AF4 huBM cells underpins the essential role of FFAR2 in KMT2Ar leukemogenesis [20]. Until now, the impact of FFAR2 is only poorly understood in leukemia. It has been demonstrated that downregulation of FFAR2 is necessary for leukemia survival in vitro and in vivo [21]. By re-analyzing publicly available datasets, we were able to assign FFAR2 an important role in leukemia by demonstrating a significant downregulation in leukemic patients in contrast to different other tumor entities and healthy controls [18,19]. We could confirm this observation by performing qPCRs with our CRISPR/Cas9-KMT2Ar models and patient samples. Interestingly, in colon cancer FFAR2 acts as an epigenetic tumor suppressor since loss of FFAR2 leads to high-level of H3K4me3 promoting colon carcinogenesis [22]. Importantly, leukemic stem cells (LSC) in KMT2Ar leukemia display high levels of H3K4me3 and low levels of H3K79me2 thereby playing a crucial role in determining LSC fate [23]. Recently, a FFAR2 agonist with favorable pharmacokinetic properties has been developed allowing the targeting of FFAR2 as a new therapeutic strategy in KMT2Ar leukemia [20].
In summary, our study highlights the feasibility of engineering chromosomal translocations at their endogenous loci in primary human HSPCs derived from both huCB and huBM to generate pure KMT2Ar cells in a short period of time serving as innovative and authentic human leukemia model. Further, we provide robust data that MLL-AF4 was unable to transform HSPCs derived from huBM cells under myeloid conditions indicating an important role of the fusion partner and the cell of origin. Downregulation of FFAR2 seems to be mandatory for leukemia development and therefore could serve as therapeutic target in the treatment of poor prognosis KMT2Ar leukemia.

Human CRISPR/Cas9-KMT2Ar Model and Patient Samples
CD34 + HSPCs were isolated (human CD34 MicroBead Kit UltraPure, Miltenyi, Bergisch Gladbach, Germany) either from fresh huCB obtained from the Department of Gynecology (IRB approval 751/2015BO2) or from huBM obtained from the Department of Hematology and Oncology of the University Hospital Tuebingen (IRB approval 309/2018BO2) and maintained in culture as previously described [14]. CRISPR/Cas9 was used to target patient-specific MLL-AF4 and -AF9 breakpoints for KMT2Ar model induction and T7 endonuclease I assay (NEB, Ipswich, MA, USA) was performed to evaluate cutting efficiencies [14]. Rearrangements were identified via PCR (AccuPrime Pfx DNA Polymerase, Thermo Fisher Scientific, Waltham, MA, USA), reverse transcriptase (RT)-PCR, Fluorescence in situ hybridization (FISH, Cytocell MLL (KMT2A) Breakapart Probe, Cambridge, UK), karyotyping and Sanger sequencing (Seqlab, Goettingen, Germany) as previously described [14]. For compound treatment the FFAR2 antagonist GLPG0974 (Tocris, Bristol, UK) was prepared in a stock solution with DMSO [24][25][26][27]. After 62 days of culture KMT2Ar huCB and huBM cells were subjected to treatment as indicated. Cells were retreated and reseeded at original density every second day. Cell counts were determined by staining with Trypan blue (Gibco, Thermo Fisher Scientific) using the Neubauer counting chamber.

Quantitative PCR (qPCR)
Total RNA was isolated (NucleoSpin RNA Kit, Macherey Nagel, Dueren, Germany), cDNA was generated (Thermo Fisher Scientific) and qPCR was performed for detection of MEIS1, HOXA9 and FFAR2 using 18S rRNA as housekeeper and employing the ddCT method as previously described [14]. The results were normalized on 18S rRNA and respective control cells were used as calibrator. Primers for FFAR2 detection were: huFFAR2 FOR 5 CCCTCACGAGTTTTGGCTTC and huFFAR2 REV 5 GGAGCCACGTGCTGCAGTA.

May-Gruenwald-Giemsa Cytospin Staining
Cytospins were performed as previously described [14]. Images were collected using a Zeiss Primovert microscope with an ×40 objective and the Axiocam 105 color camera using ZEN 3.0 blue edition software (all Carl Zeiss AG, Oberkochen, Germany) at a resolution of 2560 × 1920 pixels.

Cell Proliferation Analysis
Increasing amount of translocation positive cells was identified on DNA level via semi-quantitative PCR (AccuPrime Pfx DNA Polymerase, Thermo Fisher Scientific) of 100 ng genomic DNA with primers as previously described [14]. Proliferation of polyclonal cultures was determined by staining with Trypan blue (Thermo Fisher Scientific) and total cell count was calculated over a period of 120 days. Cell viability was determined by flow cytometry using Fixable Viability Dye eFluor 506 (eBioscience).

RNA Sequencing and Gene Expression Analyses
RNA was isolated (NucleoSpin RNA Kit, Macherey Nagel) and quality assessment was carried out by NanoDrop (Thermo Fisher Scientific) and Bioanalyzer measurements (Agilent, Santa Clara, CA, USA).

Statistical Analyses
To summarize pooled data of independent experiments the mean was calculated and standard deviation (SD) was used to describe the variability. Student's t test was used for statistical analysis and p < 0.05 was considered statistically significant.

Data Sharing Statement
For original data, please contact corina.schneidawind@med.uni-tuebingen.de. Raw sequencing files and count data are available through Gene Expression Omnibus (GEO) under accession number GSE148714.

Conclusions
In this study, we used CRISPR/Cas9 to introduce translocations of the MLL (KMT2A) and AF4 or AF9 genes under physiologic promotors in both huCB and huBM cells. All genome-engineered cells faithfully mimic the genuine nature of the disease by sharing morphological, phenotypical and transcriptomic attributes of KMT2Ar leukemias therefore constituting an innovative and authentic human model of KMT2Ar leukemia. However, the oncogene MLL-AF4 only transformed cells derived from infant cells demonstrating a major role of the fusion partner and the cell of origin in leukemia development. Finally, we uncovered intrinsic properties like an absent downregulation of free fatty acid receptor 2 (FFAR2), an epigenetic regulator, possibly responsible for this phenomenon.