Use of Genome Engineering to Create Patient Specific MLL Translocations in Primary Human Hematopoietic Stem and Progenitor Cells

One of the challenging questions in cancer biology is how a normal cell transforms into a cancer cell. There is strong evidence that specific chromosomal translocations are a key element in this transformation process. Our studies focus on understanding the developmental mechanism by which a normal stem or progenitor cell transforms into leukemia. Here we used engineered nucleases to induce simultaneous specific double strand breaks in the MLL gene and two different known translocation partners (AF4 and AF9), which resulted in specific chromosomal translocations in K562 cells as well as primary hematopoietic stem and progenitor cells (HSPCs). The initiation of a specific MLL translocation in a small number of HSPCs likely mimics the leukemia-initiating event that occurs in patients. In our studies, the creation of specific MLL translocations in CD34+ cells was not sufficient to transform cells in vitro. Rather, a variety of fates was observed for translocation positive cells including cell loss over time, a transient proliferative advantage followed by loss of the clone, or a persistent proliferative advantage. These studies highlight the application of genome engineering tools in primary human HSPCs to induce and prospectively study the consequences of initiating translocation events in leukemia pathogenesis.


Introduction
Leukemia is the most common form of childhood cancer, affecting approximately 5.2 per 100,000 children per year [1]. Over the past 60 years, scientific inquiry and advancements in treatment through clinical trials have taken what used to be a uniformly fatal disease and transformed it into a disease in which more than 90% of patients are cured [2,3]. However, certain subtypes of pediatric leukemia remain difficult to treat and retain a poor prognosis.
Many of these, including the majority of cases of infant leukemia and many treatment-related leukemias, are characterized by a translocation within the Mixed Lineage Leukemia (MLL) gene [4].
Aberrations of the MLL gene can be found in both primary and treatment-related acute leukemia in children and adults. However, the highest frequency of MLL rearrangements is seen in infants with acute leukemia [5]. For infants diagnosed with acute lymphoblastic leukemia (ALL), approximately 60-80% have an MLL rearrangement, which has been identified as a molecular feature associated with a very poor prognosis, with overall survival less than 50% [5,6]. For infants diagnosed with AML (acute myeloid leukemia), approximately 40% are found to have an MLL rearrangement [5]. While over 60 different translocation partners have been identified, the MLL-AF4 and MLL-AF9 translocations account for over half of the MLL rearrangements seen in infant leukemia [5,6]. Interestingly, the MLL-AF4 translocation is seen almost exclusively in ALL, while the MLL-AF9 translocation is more commonly seen in AML, but can also occur in ALL [7]. Translocations of the MLL gene appear to be a driving force in the pathogenesis of leukemia in these cases, with the resulting fusion protein sustaining aberrant expression of developmental genes critical in hematopoiesis [8]. Many attempts to model this process have involved forced expression of an MLL fusion protein in cells using a retroviral vector [8]. While these models have advanced our understanding of the MLL gene and fusion proteins, they have not fully recapitulated the clinical course seen in pediatric patients [9]. We believe that a system that more accurately models the initiating events that occur in nature will provide insight into the pathogenesis and possible treatments for this disease.
Chromosomal translocations, which are a hallmark of cancer cells, have been shown to result from mis-repair of simultaneous double-strand breaks (DSBs) on two different chromosomes [10][11][12]. The free end of one chromosome is ligated to a portion of a different chromosome either through classic or alternative non-homologous end-joining [13]. The evidence that DSBs on two different chromosomes can cause translocations came from studies in which recognition sites for specific nucleases were introduced into two different chromosomes and then translocations between the two artificial sites measured [10]. In the last ten years, several different platforms for engineering nucleases to induce double strand breaks at specific genome target sites have been developed giving rise to the field of genome editing. These platforms include zinc finger nucleases (ZFNs), TAL effector nucleases (TALENs) and RNA-guided endonucleases of the CRISPR/Cas9 family (RGENs). These new nuclease platforms have been used to engineer translocations and chromosomal rearrangements found in Ewing sarcoma, anaplastic large cell lymphoma, and lung cancer [12,14].
Here we designed TALENs, which consist of a fusion of a sequence specific TAL effector DNA binding domain to the nuclease domain from FokI, to specifically engineer chromosomal translocations involving the MLL gene in both K562 cells and primary hematopoietic stem and progenitor cells (HSPCs). We found that the frequency of translocations is higher in K562 cells than in HSPCs. Interestingly, the creation of MLL translocations in HSPCs was not sufficient to fully transform the cells in vitro into leukemia. Instead we found that there is a heterogeneous response to the creation of an MLL translocation whereby some cells develop a clear proliferative advantage, others develop a clear proliferative disadvantage, while still others develop a transient proliferative advantage that then disappears. These studies, which model how leukemia might occur in humans through the generation of patient specific translocations involving the endogenous genes in a small number of cells, provide the genetic foundation for studying leukemogenesis.

TALEN construction and validation
MLL, AF4, and AF9 cleavage sites were designed based on patient specific translocation breakpoints available through GenBank using the TAL Effector Nucleotide Targeter 2.0 [15]. Three pairs of TALENs were created for each cleavage site using the Golden Gate TALEN Assembly Method [16] and inserted into the MR015 vector. Each pair was then tested for cutting efficiency using the surveyor assay [17]. Briefly, K562 cells were nucleofected with a combination of TALEN pairs. Genomic DNA was isolated using the Qiagen DNeasy DNA isolation system and the region of interest (MLL, AF4, or AF9 cut sites) was amplified by PCR with the following primers: MLL AF4 5'GGTTTGACCAATTGTCCCAATAAT3', 5'TCTGGTTTGTCCTTTCCATTTGTA3'; AF4 5'TTTGGGAGACACTGGGGTAACAAT3', 5'CCCACCTGAGGAATTTCACCTTCT3'; MLL AF9 : 5'AGCAATCTCACAGGGTTCCT3', 5'TGGGACAATTGGTCAAACCT3'; AF9 5'ATCTTGTTCCAGTAGAAGGCTGTTTC3', 5'AACTCATGAATGATAAGGAAGCAAAA3'. Following isolation by gel extraction, the PCR products were heated to 95°C to denature DNA and then cooled slowly to allow the strands to re-anneal. The samples were incubated for 1h at 37°C with the T7 endonuclease, which cleaves at sites of DNA mismatches. Colony-forming cell (CFC) assays CD34+ cells were nucleofected with MLL and AF4 or AF9 TALENs or control (GFP) and were seeded in triplicate (10,000cells/dish) at days 25 and 39 of extended liquid culture in Methocult H4230 methylcellulose medium (StemCell Technologies Inc., Vancouver) supplemented with the same cytokines used in liquid cultures. Cultures were incubated at 37°C, 5% CO2, and colonies were scored after 12-14 days. Secondary CFC assays were performed by harvesting all cells from the primary cultures after colony enumeration, and replating them into new assays under identical conditions. Morphology of the colonies was determined as previously described [19]. All statistical analyses were performed with the Student's t test. P<0.05 was considered statistically significant.

Results
Generation of TALENs that create specific double strand breaks within the MLL, AF4, and AF9 genes Previous studies have identified hot spots within the MLL, AF4, and AF9 genes, known as breakpoint cluster regions, that are frequently the sites of chromosomal translocations in patients with leukemia [7]. To design TALENs for the MLL and AF4 genes, we chose a specific patient translocation sequence (UPN010) that is available through the GenBank database (AJ408902.1) and is representative of the most common site of translocations in patients with infant leukemia [20]. The Golden Gate assembly method [16] was used to create pairs of TALENs that induce a specific double strand break within the breakpoint cluster regions of MLL and AF4 corresponding to the UPN010 translocation ( Fig 1A). Following nucleofection with the MLL and AF4 TALENs, respectively, resultant double strand breaks within the MLL and AF4 genes were detected using the surveyor assay, which results in cleavage of PCR products at sites of DNA mismatch that result from the creation of indels at sites of double strand breaks (Fig 1B) [17]. A second set of TALENs was designed to create double strand breaks in the MLL and AF9 genes within the breakpoint cluster regions typically seen in patients with MLL-AF9 leukemia (Fig 1C). This design was based on the specific translocation sequence found in a child with AML (P1) that was characterized by Langer et al. [21]. The surveyor assay confirmed activity of both the MLL and AF9 TALENs (Fig 1D).

Induction of double strand breaks within the breakpoint cluster region of MLL and its fusion partner is sufficient to generate MLL translocations
To test whether the simultaneous induction of two double strand breaks is sufficient to induce a specific chromosomal translocation, we nucleofected K562 cells with either the MLL and AF4 TALENs, or the MLL and AF9 TALENs. Genomic DNA isolated on day 3 post-nucleofection demonstrated both the primary MLL translocation (MLL-AF4; MLL-AF9) and the reciprocal translocation (AF4-MLL; AF9-MLL) (Fig 2A and 2B). Translocations were not detected when the MLL TALENs were expressed alone. TALEN transfected cells appeared to have a survival disadvantage in the K562 cultures as the relative intensity of the translocation product within the cell populations was significantly decreased after 14 days in culture. Resultant PCR products were isolated, and genomic rearrangements confirmed by Sanger sequencing (Fig 2C and  2D). A sub-culturing strategy was used to quantify the MLL-AF4 translocation frequency in K562 cells, which was approximately 1 in 1100 cells (8.9 x 10 −4 ). These studies demonstrate that TALENs specifically targeted to the respective breakpoint cluster regions are capable of generating MLL chromosomal translocations.  TALENs targeted to breakpoint cluster regions are sufficient to generate MLL translocations in human primary hematopoietic stem and progenitor cells Since K562 cells have significant chromosomal abnormalities at baseline and likely do not recapitulate what would be observed in primary cells, we explored the feasibility of inducing translocations in human primary HSPCs. After isolation of human primary CD34+ cells from fresh umbilical cord blood, the indicated TALENs were introduced by nucleofection. Cells were maintained in culture and monitored over time for the appearance of an MLL translocation. After 1-2 weeks following nucleofection, MLL-AF4 or MLL-AF9 translocations and the respective reciprocal translocations (AF4-MLL; AF9-MLL) were detected (Fig 3A and 3B). The resultant PCR products were isolated and the translocations confirmed by Sanger sequencing (Fig  3C and 3D). In order to assess the frequency of induction of MLL-AF4 translocations by TALENs, human CD34+ cells were subdivided into a 96 well plate immediately following nucleofection with MLL and AF4 TALENs (50,000 cells per well). Parallel cultures were maintained over time and monitored by PCR for the presence of the MLL-AF4 as well as the reciprocal AF4-MLL translocation. The estimated MLL-AF4 translocation frequency following nucleofection with TALENs was 2.6-4.0 x 10 −5 . The reciprocal AF4-MLL translocation was found to have a lower frequency of 1.3-2 x 10 −5 . Taken together, these data demonstrate the induction of MLL chromosomal translocations by TALEN-mediated genome engineering of primary human HSPCs.

MLL translocations in primary CD34+ cells confer a survival advantage in extended culture and induce clonal expansion in colony-forming cell assays
To assess whether de novo MLL translocations result in a survival advantage in primary cells, human CD34+ cells were nucleofected with MLL and AF4 or AF9 TALENs or GFP and maintained in culture. Cell viability was determined by flow cytometry over time (S1 Fig). Whereas the control sample died out after 2 months in culture, the MLL-AF4 and MLL-AF9 samples continued to proliferate until approximately day 100-120. We did not see a difference in the proliferation capacity between the MLL-AF4 and MLL-AF9 samples. Further analysis of the subcultures revealed a variable pattern on a clonal level. We used both semi-quantitative PCR (MLL-AF4 and AF4-MLL; Fig 4A) and qPCR (MLL-AF4; Fig 4B) to track the kinetics of a given clone over time. Seventeen out of the 20 subcultures had a detectable MLL-AF4 translocation one week following nucleofection. Many of these clones persisted for 14-21 days in culture, but disappeared between days 21-41. However, a number of subcultures contained a clone with an MLL-AF4 translocation that appeared to have a significant survival advantage as demonstrated by the increased prominence of the translocation product over time. While some of these clones displayed a rapid early expansion and subsequently disappeared after day 21, four clones remained prominent thru days 41-57. FISH analysis was performed on subcultures 9 and 20 at day 41 to better quantify the percentage of cells with an MLL translocation (Fig 5). Using an MLL break apart probe, an MLL translocation was detected in 49.5% (99/200) of cells in subculture 9, 17.5% (35/200) of cells in subculture 20, and 0% (0/200) of cells in the GFP control culture. Interestingly, the four subcultures that demonstrated the most significant expansion and prolonged survival in extended culture also had a detectable reciprocal AF4-MLL translocation by PCR (S2 Fig) and the reciprocal translocation was detected in subculture 9 by FISH analysis. Additionally, expression of the MLL-AF4 fusion product transcript was confirmed by RT-PCR in subcultures 3,9,16,and 20 (S3 Fig). In order to further understand the clonal dynamics of the MLL translocations in these subcultures,  We next performed colony-forming assays to further assess the clonal expansion and transformative potential of the translocated cells. For this purpose, we plated 10,000 cells of the MLL translocated samples (both AF4 and AF9) or control samples each on day 25 and 39 of the extended culture in semisolid medium supplemented with the same cytokines that were used during the extended cultures. All three cultures generated colonies after 12-14 days, but the MLL-AF4 and MLL-AF9 samples displayed significantly higher clonogenic potential after the second round of replating as compared to the control samples ( Fig 6A). Remarkably, approximately 20% of the colonies in the experimental samples containing the translocated cells, but not the control sample, demonstrated a compact morphology consistent with more immature cells [19] (Fig 6B). However, the colony forming assays confirmed that the translocated cells were not fully transformed since further replating resulted in decreasing cell numbers comparable to what was observed in our extended in vitro cultures.

Discussion
Using tools developed for genome editing/engineering, we induced patient specific MLL chromosomal translocations in primary human hematopoietic stem and progenitor cells (freshly isolated CD34+ cells), and prospectively investigated what happened to the cells as a result of the initiating event. We believe that this is the first example in which TALEN nucleases have induced the relevant translocation in primary human HSPCs, a critically important milestone in using engineered nucleases to study the ontogeny of translocation associated leukemia.
MLL-rearranged leukemia presents a unique opportunity to explore leukemia pathogenesis in a prospective manner. We know from infant studies that MLL rearrangements are an initiating event that occurs in utero in infants who develop MLL-rearranged leukemia [22,23]. Other chromosomal translocations such as TEL-AML1 and BCR-ABL have been shown to occur in utero, but demonstrate more twin discordance, longer latency, and are thought to require subsequent mutations for leukemia development [24,25]. Epidemiologic studies suggest that an MLL translocation alone may be sufficient for leukemogenesis given the high penetrance and short latency seen in twin studies [26]. MLL translocations also occur in therapy related leukemias in which a secondary leukemia usually occurs 1-3 years after exposure to topoisomerase inhibitors such as etoposide and teniposide [27][28][29]. Thus, the latency in infant leukemia and therapy related leukemia both highlight the importance of understanding the developmental period between the occurrence of the translocation and the appearance of overt leukemia.
Despite extensive investigations over the past 20 years, our understanding of the role of MLL translocations in the pathogenesis of acute leukemia is still not complete. Many animal models have been developed and helped to define MLL fusion proteins as key transcriptional and epigenetic regulators that lead to leukemia pathogenesis [7,9,[30][31][32][33][34]. However, current animal models have failed to fully recapitulate the role of MLL fusion proteins, especially in the case of the MLL-AF4 translocation. The MLL-AF4 translocation has been difficult to model in mice due both to the size of the MLL-AF4 translocation product and its chromosomal orientation [8]. Additionally, mouse models that develop leukemia by forced expression of an MLL-AF4 fusion product have not fully recapitulated the clinical phenotype seen in infants and are complicated by potential effects related to retroviral insertion. There is ongoing debate as to whether the MLL-AF4 fusion protein alone is sufficient to induce leukemogenesis or whether subsequent mutations are required [35]. Previous studies have demonstrated Engineering MLL Translocations in HSPCs MLL-AF4 leukemias to have fewer additional mutations as compared to other leukemias, but have identified N-Ras, K-Ras, and Flt-3 as a source of potential cooperating mutations [36][37][38][39][40]. Additionally, it remains unclear whether there is a role for the reciprocal AF4-MLL translocation product, as only a fraction of patients have a detectable reciprocal translocation; however one report has demonstrated leukemia development in a mouse model by enforced expression of the reciprocal AF4-MLL fusion product alone [41]. The genome engineering approach developed in our studies provides a novel approach to more fully explore the role of various initiating events in leukemia pathogenesis in a prospective, physiologic manner using primary HSPCs.
In vitro, we observed that many cells with an induced MLL translocation do not persist in culture. This is likely a result of an intact apoptotic response in primary cells. However, a subset of cells that acquired an MLL translocation persisted over time, appeared to have a selective advantage in extended culture, and demonstrated a higher clonogenic potential in colony forming assays. While this effect was not as robust as that seen in previous studies when the fusion product has been introduced with retroviral vectors, this is likely related to lower expression of the fusion product under the control of the endogenous promoter. Interestingly, the clones that demonstrated the most robust survival advantage either in subculture analysis or colony forming assay contained the reciprocal translocation product. This further suggests that the reciprocal translocation may play an active role in leukemia pathogenesis. While the MLLrearranged cells demonstrated a selective advantage, which may represent a pre-leukemic clone, our studies would suggest that they were not fully transformed as these cells eventually died out in culture as well as with serial replating in colony forming assays. This is consistent with other studies showing that enforced expression of an MLL-AF4 transgene in primary HSPCs conveys a proliferative advantage but is insufficient to generate leukemia [9,38]. An alternative explanation for the enhanced proliferation followed by disappearance is that the culture conditions do not support immortalized growth of transformed cells since it is widely known that primary ALL, including those with MLL translocations, are extremely difficult to grow in culture. Finally, it could also be that the clones were pre-leukemic but also died out because they retained some of the characteristics of their cell of origin (an early progenitor) and had a defined but limited replicative lifespan. Further studies are required to address these hypotheses. Nevertheless, this model system presents a unique opportunity, as we typically do not have access to pre-leukemic clones in order to explore the steps necessary to transform into leukemia. We are currently using this system both in vitro and in vivo to explore the immunophenotypic, gene expression and epigenetic changes that occur in cells over time as a result of the MLL translocation. Additionally, we are using this system to generate MLL translocations in CD34+ cells freshly isolated from umbilical cord blood followed by transplantation into immunocompromised (NSG) mice in order to develop a novel mouse model of MLL-rearranged leukemia. We believe this system will allow us to more fully and accurately answer whether or not MLL translocations are sufficient for leukemogenesis or whether they require cooperating events. It will also allow exploration of the relationship between the MLL translocation partner as well as cell of origin in relation to disease phenotype.