The variables that are important for initiation of leukemia in xenograft models are not well defined. Numerous studies have shown that 30–50 percent of primary human AML samples do not engraft immunodeficient mice.8-11
We have noticed a similar trend using the MA9 model of transduced human CD34+
cells. As described in our recent paper, the long-term cultured MA9 myeloid cell lines induce AML and the tumor cells are typically transferable to secondary mice.4
describes the data obtained using two such cell lines, clones 3 and 6, which reproducibly induce AML upon injection into mice, with disease latency and tumor burden a function primarily of mouse strain (i.e., the presence of the three human cytokines in the NOD-SCID-SGM3) irrespective of length of time in culture or route of injection. However, we have generated two other myeloid cells lines (clones 1 and 4) with virtually identical phenotypic and morphologic characteristics that do not reproducibly induce AML when injected into mice (). While clone 1 displays impressive killing ability initially, this ability is subsequently diminished after continued in vitro culture. Clone 4 is fairly ineffective initially and became completely unable to engraft and kill mice when injected at later timepoints of in vitro culture ( and data not shown). Based on these data, it seems that in vitro immortalization and in vivo transformation are separable phenotypes. One possible explanation for these results is that the clone which predominates in vitro is the most fit under cell culture conditions, but the characteristics that are necessary for in vivo expansion are not essential components in vitro and may not be an obligate phenotype of every immortal line. These characteristics could include homing efficiency, adherence to bone marrow stroma, metastatic capability, and propensity to respond to the microenvironmental signals in the bone marrow environment rather than the Petri dish. It is also possible that the genetic background of the donor cell influences these different parameters. Additional mutations may also be required for leukemogenesis above and beyond those required for immortalization, and some in vitro clones may be more or less likely to possess or acquire these genetic alterations. Obviously each cell line or tumor that is established is derived from a unique, outbred human sample, unlike the inbred mouse models that are established using similar approaches with murine cells. Given the variability that is seen in human leukemia patients with regard to disease presentation, treatment response, relapse and overall survival, we would expect that genetic factors influence some of the parameters in our model system as well. This is in fact one of the strengths of the current model in our opinion and an advantage over the use of inbred mouse strains for leukemia modeling.
Long-term engraftment and induction of acute myeloid leukemia in mice using cultured MA9 cell lines
As we have previously demonstrated, the sub-strain of immunodeficient mouse that is used has a significant impact on the latency of disease.4
The NOD-SCID-SGM3 mouse, that contains the three human cytokine transgenes encoding SCF, GM-CSF and IL-3, promotes a significantly faster myeloid disease than does the NOD-SCID mouse. This is evident in the data presented for clone 6 in , and has been seen repeatedly using multiple different samples. These three cytokines have been identified as particularly important for promoting myeloid cell proliferation, and the lack of cross-reactivity of the murine cytokines for human receptors is likely to be a critical factor in the xenograft model for the cytokine dependent MA9 cell lines as well as for primary AML patient samples. We have also noticed a significant improvement in penetrance and shortened latency of disease upon direct intrafemoral injection compared to intravenous injection (). Whether this is due to a homing deficiency in the MA9 cells or is due to an increased initial “load” of tumor cells in the proper niche remains to be determined.
One interesting result that has emerged from our use of two different MLL-AF9 fusion genes is the finding that these genes reproducibly generate immortal cell lines that differ in a number of ways. As we have previously shown, the surface phenotype of these myeloid cell lines show consistent differences for some common cell surface markers, including CD13, CD117 (c-kit) and CD135 (Flt3).4
The importance of these differences is not clear, and the high variability of expression of these molecules, as well as the fact that even those myeloid lines that show essentially no Flt3 surface expression are still dependent on Flt3L for growth and survival, raises questions as to whether these are technical artifacts rather than real differences. Importantly, both constructs induce immortalization with similar success rates (). However, one highly reproducible difference between these two groups of cell lines is the in vitro clonogenic efficiency. Those myeloid cell lines that were generated using the REW-MA9 retroviral construct have an average clonogenic rate of about 40% while the cell lines generated from the second retroviral construct show a ten-fold lower frequency of approximately 5% (). These in vitro differences do not correlate with in vivo aggressiveness; in fact, the two cell lines that show the highest clonogenic potential, MA9.1 and MA9.4, are the two that do not efficiently initiate tumorigenesis in mice (). There are two differences between these retroviral constructs. The chromosomal breakpoints differ in the MLL portion of the fusion genes, resulting in additional MLL coding sequence in the REW construct.4
In addition, for the REW construct, we engineered it so that EGFP is expressed as a fusion protein with MLL-AF9. This fusion protein (EGFP-2A-MA9) is proteolytically cleaved after translation to generate individual proteins, as a result of the presence of the Foot and Mouth Disease Virus 2A peptide.4,12
Which of these factors is responsible for the clonogenic differences remains to be determined. However, this data clearly highlights the need for further characterization of this model system and the variables that impact on the different readouts.
Figure 1 The clonogenic frequency differs among UCB-derived myeloid MA9 cell lines based on the retroviral construct used. Cell lines established using the REW-MA9 retroviral construct (MA9.1, MA9.3, MA9.4 and MA9.6) show relatively high clonogenic frequency compared (more ...)
In a pilot series of experiments using a single MA9 myeloid cell line, we performed limiting dilution clonal analyses and expanded some of these single cell clones for leukemogenicity testing. Interestingly, for four purified clonal lines that were generated, all four were effective in leukemia induction, with similar kinetics to the parental myeloid cell line (). This data indicates that the high clonogenic rate that is observed in the MA9 cell lines (approximately 30% for the MA9.3 clone used in this experiment, see ) can be considered as a measure of the LSC frequency in these cultures. This is surprising data, given the low frequency of LSC that has been observed in experiments using AML patient samples.13
However, it appears likely that MLL leukemia is a unique disease with characteristics that are not common to other subtypes of leukemia, including the ability to transform committed murine progenitor cells and to impart self-renewal properties on a high fraction of cells within the murine tumor.14-16
The uniqueness of this class of oncogenes extends to human cells as well, given that this is the only group of oncogenes that has thus far been successful in transforming primary human HSPC. The signaling pathways downstream of MLL oncogenes that are initiating such efficient self-renewal signals remain to be identified.
Figure 2 Myeloid cell lines generated from single cell clones retain in vivo leukemogenic potential. Single cells were isolated as described in the legend to . Cells were expanded and one million cells were injected intravenously into mice. Mice were sacrificed (more ...)