Clonal selection in xenografted TAM recapitulates the evolutionary process of myeloid leukemia in Down syndrome

Neonates with Down syndrome (DS) are at high risk of developing a unique hematologic disorder referred to as transient abnormal myelopoiesis (TAM), transient myeloproliferative disorder, or transient leukemia. In most cases, TAM resolves spontaneously within 3 months.^1,2  However, after spontaneous remission, 20% of TAM patients develop myelodysplastic syndrome and acute megakaryocytic leukemia referred to as myeloid leukemia of DS (ML-DS) within 4 years.^3,4  Blast cells in most patients with TAM and ML-DS have mutations in exon 2 of the gene coding for the transcription factor GATA1,^5-8  which is essential for the normal development of erythroid and megakaryocytic cells.^9,10  Although blast cells in most TAM and ML-DS patients share the identical GATA1 mutation, recurrent additional cytogenetic abnormalities are commonly observed during disease progression.^2,5,11,12  In fact, a ML-DS case derived from a minor clone with a distinct GATA1 mutation in the TAM phase was previously reported by our group.¹³  These clinical findings suggest that although most TAM cells disappear in the early neonatal phase, a few clones persist during apparent remission to develop ML-DS later. Because only one fifth of TAM cases progress to ML-DS, additional genetic events besides GATA1 mutation are likely to be involved in the progression of TAM to ML-DS.¹⁴  As mentioned above, the development of ML-DS is significantly correlated with karyotypic abnormalities such as duplication (dup)(1q), deletion (del)(6q), del(7p), dup(7q), +8, +11, and del(16q),^2,11,12  which are rarely observed in the TAM phase. These clinical findings have led many physicians to consider TAM as preleukemia and the progression of TAM to ML-DS as an attractive model to investigate multistep leukemogenesis.

Animal models have contributed to our understanding of the pathogenesis of TAM/ML-DS and other leukemias.^15-21  Mice models in which primary human leukemic cells were transplanted into immunodeficient hosts provided significant clues to advance our understanding of the pathogenesis of human leukemia.^19-22  However, xenograft models of primary patient samples from the preleukemic phase have been rarely reported, and the TAM xenograft model would be an attractive method to investigate leukemogenesis.

We previously described the development of novel immunodeficient NOD/Shi-scid, interleukin (IL)-2Rγ^null (NOG) mice with a superior capacity for the engraftment of human hematopoietic and neoplastic cells.^23-26  In contrast to a previous study in which TAM cells showed a limited ability to expand in immunodeficient mice,²⁷  we established a xenograft model where TAM cells were transplanted into NOG mice to recapitulate the pathophysiology of TAM/ML-DS. This xenograft model in combination with high-throughput genomic technology was used to show that genetically heterogeneous minor subclones with leukemia-initiating potential already exist in the neonatal TAM phase and could serve as initiating clones evolving to ML-DS in a patient. Our TAM xenograft model may be of value to gain insight into the evolutionary process of leukemia.

Peripheral blood (PB) samples were obtained from patients diagnosed with TAM associated with DS in acute and complete remission phases. Mononuclear cells were separated by Ficoll-Hypaque (Pharmacia, Uppsala, Sweden) density gradient centrifugation, as previously described.²³  Informed consent was obtained from the patients’ parents in accordance with the Declaration of Helsinki, and the research was approved by the institutional ethics committee of Kyoto University Hospital.

NOG mice were developed at the Central Institute of Experimental Animals (Kawasaki, Japan) as previously described²⁸  and were maintained in our pathogen-free facility and cared for in accordance with the institutional guidelines for animal welfare.

Xenotransplantation and analysis of TAM cells were performed using a previously reported method with some modifications.²⁶  In brief, PB mononuclear cells (PBMCs) obtained from TAM patients (1–3 × 10⁶ cells) were injected into 2.4 Gy–irradiated 8- to 12-week-old NOG mice through the tail vein. To screen for the proliferation of TAM-derived cells, bone marrow (BM) cells were aspirated from the tibia every 4 weeks. Engraftment was defined as >1% of cells staining positive for human CD7 (hCD7), hCD33, hCD41a, hCD45, and hCD117 at 12 weeks after transplantation. For serial transplantation, recipient BM cells were collected 12 to 18 weeks after transplantation; the equivalent of 1 × 10⁶ hCD45⁺ cells was intravenously transplanted into new mice. For a detailed determination of chromosomal and genetic alterations in TAM-derived cells, serial transplantation experiments using preserved PBMC samples were performed.

For analysis of TAM-derived cells in murine BM, mice were euthanized, and the BM was removed and mechanically dispersed. Mononuclear cells were purified from the BM and stained with antibodies. Dead cells were excluded according to 4′,6-diamidino-2-phenylindole staining. Blast cells were identified by classical CD45/SSC blast gating.²⁹  See supplemental Methods on the Blood Web site for details.

Human cell isolation was performed according to a previously described method with some modifications.^23,24  See supplemental Methods for details.

Leukemic colony formation was assessed according to a previously described method with some modifications.³⁰  See supplemental Methods for details.

The GATA1 gene was amplified using polymerase chain reaction (PCR) as previously described⁸  and sequenced by an ABI 3130xl Genetic analyzer (Applied Biosystems, Foster City, CA).

DNA copy number analysis was performed using GeneChip Human Mapping 250K Nsp arrays (Affymetrix, Inc., Santa Clara, CA) according to the manufacturer's standard protocols. Genomic copy numbers including allele-specific copy numbers were calculated using CNAG/AsCNAR software (http://www.genome.umin.jp), and genomic DNA obtained from PB of patients in the remission phase was used as a control. Copy number abnormalities and other allelic imbalances were detected using a hidden Marcov model–based algorithm.

Data are presented as the mean ± standard deviation. The 2-sided P value was determined by testing the null hypothesis that the 2 population medians are equal. P values <0.05 were considered to be significant.

To determine whether NOG mice provide a TAM xenograft model, Ficoll-purified PB samples from 11 TAM patients were transplanted into irradiated NOG mice. Patient characteristics are shown in Table 1. Patients’ ages at sample collection, percentage of blast cells, number of cells injected, and number of engrafted recipients for each PB sample are shown in supplemental Table 1. Of 11 patient samples, 3 (patients 1, 2, and 9) were engrafted successfully in the recipient mice. Engraftment was maintained ≥12 weeks in all cases (Figure 1A). The spleen and liver of the recipients were also infiltrated with hCD45⁺ blast cells (data not shown). These TAM-derived cells were morphologically similar to the primary TAM cells obtained from the patients (Figure 1B). Flow cytometric analysis of surface antigens detected the expression of CD117, CD34, CD33, and CD41a on hCD45⁺ cells, which was consistent with the pattern observed in primary cells of TAM patients (Figure 1C). The presence of the same GATA1 mutation was confirmed in the primary TAM cells and the engrafted cells in NOG mice (Figure 1D; supplemental Table 1). Chromosomal analysis of engrafted cells showed no abnormalities other than trisomy 21 (Figure 1E). These TAM-derived cells were detectable in the recipient’s BM for >24 weeks (data not shown).

To examine the self-renewal capacity of TAM-derived cells, we performed serial transplantation of engrafted cells in the BM of recipient mice. Only the TAM-derived cells from patient 1 were successfully engrafted into the secondary (2°) and tertiary (3°) recipients. The morphology and surface antigen expression of these engrafted cells remained unchanged throughout the serial transplantation (Figure 2A-B). Interestingly, the TAM-derived cells expanded rapidly in the 3° recipients (Figure 2C). The colony-forming ability of the engrafted cells also increased in subsequent generations (Figure 2D). These cells could be grown by serial transplantation for >1 year and ≥8° recipients, indicating that some TAM clones had long-term self-renewal capacity, a characteristic of leukemia. Indeed, patient 1 developed ML-DS at the age of 1 year, whereas the other patients did not (Table 1).

Additional chromosomal alterations are frequently observed in ML-DS in comparison with TAM, suggesting that these alterations in genomic structure could be related to the evolution of ML-DS from TAM.^2,11,12  Therefore, we first investigated whether the serially engrafted TAM-derived cells (from patient 1) had DNA copy number alterations (CNAs) using Affymetrix GeneChip Mapping 250K arrays. Primary samples from patient 1 had no CNAs other than the gain of chromosome 21. However, the TAM-derived cells in the 1° recipients showed heterozygous deletion of 16q22 and 16q24 (Figure 3). To determine whether these deletions were present in the same cell, we calculated the signal intensities of each deletion using array data. Nearly 100% of TAM-derived cells harbored each deletion, indicating that these 2 deletions exist in a single TAM-derived cell. Although 2° recipients showed the same CNAs, 3° recipients showed additional CNAs, namely the gain of the entire chromosome 1q (Figure 3; supplemental Figure 1A). Interestingly, the 1q gain was not detected in the 4° to 7° recipients, whereas deletions of 16q22 and 16q24 were present (Figure 3). In this series of transplantations, the original GATA1 mutation found in the primary patient sample (patient 1) remained unchanged (supplemental Figure 1B).

Gain of 1q and deletions in 16q are recurrent chromosomal abnormalities in ML-DS.^11,12,31  The result of G-band karyotyping of TAM-derived cells in 3° recipients was 47,XX,+1, der (1;15)(q10;q10),+21 in 20/20 metaphase cells (supplemental Figure 1C), confirming genomic structural change, which is a hallmark of ML-DS. These data suggest that leukemic evolution of TAM-derived cells was observed in our NOG mouse model.

To examine the kinetics of the leukemic evolution of TAM cells, another 2 sets of serial transplantations were performed using the preserved patient 1 sample (Figure 4A). Four of 5 mice in the second group (m2-1–m2-5) and 5 of 11 mice in the third group (m3-1–m3-11) harbored TAM cells from the patient. Of the total of 9 engrafted mice, 2 had the same CNAs detected in the first series of serial transplantations: deletion of 16q22 and 16q24 (m3-5 and m3-8; Figure 3). Moreover, 2 combinations of new CNAs were detected in the 1° recipients: deletion of 9q22 +12p12 (m3-4 and m3-7) and gain of 1q25.2-1q44 (m3-11). No CNAs other than the gain of chromosome 21 were detected in the other recipients (m2-1, m2-2, m2-4, and m2-5).

Each 1° engrafted mouse was subjected to 2° transplantation, and 5 of 9 series (m2-5, m3-4, m3-7, m3-8, and m3-11) successfully gave rise to the xenografts in the 2° recipients. It is noteworthy that the TAM-derived cells of the 2° recipients in 2 of the 3 analyzed series (m2-5 and m3-4) acquired additional CNAs, whereas the CNAs in 2 descendent 2° recipients of m3-8 remained unchanged. The additional CNA of gain of 1q was detected in the 2° recipients of m3-4, similar to that observed in the 3° recipient in Figure 3. Although gain of 1q was recurrently observed in this series, the duplicated regions were diverse: 1q25.2-1q44 (1°, m3-11), 1q21.3-1q44 (2°, m3-4), 1q31.2-1q44 (2°, m3-4), and the whole arm of chromosome 1q (3° in Figures 3 and 4A; supplemental Figure 2). In m2-5, a deletion of 3q24 appeared in the 2° and 3° recipients. These results demonstrated that TAM cells derived from patient 1 acquired various CNAs and showed divergent repopulating capacity in our xenograft model.

ML-DS can arise from a minor TAM clone with a GATA1 mutation that is distinct from that of the major TAM clone in a patient.¹³  To determine whether the GATA1 mutation in the primary patient’s TAM cells was preserved in engrafted TAM-derived cells, GATA1 mutation analysis was performed. TAM-derived cells in the series m3-4, m3-5, m3-7, and m3-8 had the same GATA1 mutation (c.38_39delAG) as that of patient 1 (Figure 4A). Surprisingly, this mutation was not detected in TAM-derived cells in m2-1, m2-2, m2-5, and m3-11; instead, these samples showed a distinct GATA1 mutation (c.1A>G) that was not detectable in the primary patient sample by direct sequencing. One of the 1° recipients (m2-4) showed both GATA1 mutations. These results suggested that a minor clone with a distinct GATA1 mutation (c.1A>G) was present in the primary patient sample and that this minor clone coexisted with, or predominated over, other clones in some 1° recipients. Therefore, a mutation-specific restriction enzyme digestion assay was performed using the primary sample from patient 1, which confirmed the presence of cells with the GATA1 mutation (c.1A>G) as a minor clone (Figure 4B). Moreover, this minor clone propagated and acquired CNAs in NOG mice independently of the major clones (Figure 4A), further demonstrating the genetic heterogeneity of TAM cells. Interestingly, the major clone in the original patient 1 sample with a c.38_39delAG GATA1 mutation and no CNAs did not become dominant in any of the recipients.

TAM-derived cells in multiple 1° recipients derived from patient 1 had various CNAs including deletions of 16q22 and 16q24 (Figure 4A). To determine whether these subclones were present at low levels in the primary sample of patient 1, specific PCR for the 16q22 deletion was performed using primer pairs designed to bookend the deletion site. CNA analysis and genome sequencing data in these deletion sites (16q22 and 16q24) revealed the presence of genomic breakage and inversion (Figure 5A; supplemental Figures 3 and 4; see supplemental Methods for details). A primer set was designed to detect the deduced breakpoint and used to perform PCR on TAM-derived cells from patient 1 in the recipients with 16q22 and 16q24 deletions. PCR using genomic DNA from TAM-derived cells in the 1° to 8° recipients of the first series of transplantations (Figure 3) produced a uniformly bright DNA fragment of the same size, consistent with the results of CNA profiling (Figure 5B). A faint fragment was detected by applying this PCR method to genomic DNA from the primary patient sample (patient 1), which was confirmed to contain the deletion breakpoint in 16q22 by Sanger sequencing. These results demonstrated that TAM cells with the 16q22 and 16q24 deletions already resided as a minor population in the original sample of patient 1. The frequency of the mutant cells was estimated to be ∼1.0% to 0.2% of the patient’s PBMCs by a serial dilution assay (supplemental Figure 5).

The same method was used to detect a subclone with a 3q24 deletion in the primary patient sample (m2-5; Figure 4A; supplemental Figure 6). At the site of the deletion, genomic breakage was confirmed, and the ends were bound by insertion of a G-nucleotide (Figure 5C; supplemental Figure 7). Consistent with the results of CNA profiling, PCR using DNA from engrafted cells in the 2° and 3° mice (m2-5; Figure 4A) produced a bright DNA fragment, which was confirmed to contain the deletion breakpoint in 3q24 by Sanger sequencing (Figure 5D). Engrafted cells from the BM of the 1° recipient produced a faint DNA fragment, although CNAs were not detected in these cells by array-based methods. We could not detect the corresponding DNA fragment in the primary sample of patient 1. These results suggest the subclone with the 3q24 deletion arose in the 1° recipient mouse as a minor population, emerged as a major population in the 2° recipient, and subsequently engrafted into the 3° recipients. However, because the sensitivity of the specific PCR targeting of the 3q24 deletion was ∼0.1% as determined by the dilution assay (data not shown), it is also possible that this minor clone already existed in the primary patient sample at a frequency below the sensitivity limit. Collectively, our results provide evidence that subclones with additional genetic alterations already exist in the TAM phase and suggest that clonal selection occurs continuously in this xenograft model.

To assess whether TAM cells derived from the patients who did not develop ML-DS had similar self-renewal capacity and genetic instability to those from patient 1, CNA analysis of TAM-derived cells was performed by transplanting the preserved PBMC samples of patients 2 and 9. In patient 2, 4 1° transplantation attempts resulted in successful engraftment. The primary sample of patient 2 had no CNAs (Figure 6A). However, TAM-derived cells in 1 of the 1° recipients (m2-2) showed 7p and 7q deletions, suggesting that a subclone with these CNAs may exist in the primary patient sample. The other 2 1° recipients had no additional CNAs (m2-1 and m3-6). In patient 9, engraftment succeeded in 5 1° recipients, and no additional CNAs were detected in either primary patient sample or engrafted TAM-derived cells (Figure 6B). The engrafted cells in all of the recipient mice harbored the same GATA1 mutation as that of the primary samples of patients 2 and 9. In these 2 cases, our xenograft assay did not detect potent TAM clones with self-renewal capacity in serial transplantation assays (Figure 6A-B; supplemental Table 1), which may reflect the favorable clinical outcome of these patients.

Taken together, the results show that only the TAM cells derived from patients who subsequently developed ML-DS had long-term self-renewal capacity with additional CNAs in our serial transplantation assay.

New genomic technologies have led to a better understanding of the complex clonal architecture of leukemia and have shown that disease progression occurs through clonal evolution.^20-22,32  However, most studies have been based on the retrospective analysis of frank leukemia samples, and data on the evolutionary process occurring in the preleukemic phase are limited because primary preleukemia samples are rarely available and are difficult to maintain in vitro or in vivo. TAM is a unique hematologic condition associated with DS that is mostly self-limited but leads to ML-DS in 20% of cases after spontaneous remission. Therefore, TAM has been considered a preleukemic state and is a suitable pathological condition to analyze the evolutionary process of leukemia.

Because mice models in which primary human leukemic cells were transplanted into immunodeficient hosts provided significant clues to advance our understanding of the pathogenesis of human leukemia,^19-22  we hypothesized that xenograft models of TAM cells would be an attractive method to investigate leukemogenesis. In this report, we demonstrated the long-term engraftment of primary TAM cells in NOG mice and showed that TAM cells from a patient that subsequently developed ML-DS had the potential to gain diverse additional genomic alterations and self-renewal capacity. Although we were unable to determine whether the clonal evolution of TAM cells observed in our model reflected the clinical phenotype of the original patient because of insufficient sample from the ML-DS phase, our model is likely to enable the prospective evaluation of leukemic evolution and can be a powerful tool to study the pathophysiology of leukemogenesis. Our model using NOG mice contrasts somewhat with the study by Chen et al,²⁷  who reported that TAM cells resided only in the BM after intra-BM infusion into NOD/SCID mice. We speculate that a severe and unique immunodeficient microenvironment may have contributed to the successful engraftment of TAM cells in NOG mice.

In the present study, primary TAM cell samples from 3 of 11 patients engrafted in NOG mice (Figure 1), but serial engraftment was successful only with cells obtained from the patient who developed ML-DS at the age of 1 year (Figures 2, 3, and 6). The results of extensive serial transplantation revealed the emergence of subclones with various additional CNAs characteristic of ML-DS (Figures 3 and 4). Furthermore, we showed that minor subclones with various CNAs and a distinct GATA1 mutation were already present in the ML-DS patient during the early TAM phase (Figures 4 and 5), as previously described for polyclonality of TAM.^2,33  These findings suggest that several preleukemic clones with high leukemia-initiating potential may already reside as minor clones in TAM cells of patients fated to develop ML-DS and show high repopulating capacity in the special microenvironment of NOG mice. Our findings support the hypothesis that ML-DS develops from a pool of heterogeneous minor clones through clonal selection, illustrating the early evolutionary process of leukemia.³⁴ 

Long-term engraftment of TAM-derived cells was observed for only a minority of TAM patients. This finding suggests that factors other than the properties of the TAM-derived cells, such as technical issues, affected engraftment efficiency. In this regard, increasing the number of transplanted cells resulted in a higher rate of engraftment in recipients of samples from patient 9 (supplemental Table 1). However, there was no clear association between the percentage of TAM blast cells in transplanted samples and successful engraftment (supplemental Table 1). Likewise, frozen samples from 3 patient samples (patients 1, 2, and 9) were as efficient for engraftment as fresh samples from these patients. Therefore, although the number of injected TAM cells and technical issues may affect engraftment, we speculate that engraftment efficiency is an intrinsic property of each TAM-derived cell population.

In addition to trisomy 21, somatic GATA1 mutation is considered an early essential event of TAM and ML-DS occurring in utero.^35,36  Interestingly, our TAM-NOG mice model enhanced the emergence of a minor clone with a distinct GATA1 mutation that was not detectable in the original patient sample by conventional sequencing methods. In our model, a minor GATA1 mutant clone expanded predominantly in some recipients and acquired CNAs independently of clones with the original GATA1 mutation, raising the possibility that leukemic evolution occurred from this minor clone, similar to the clinical observation in our previous report.¹³  In this scenario, a common founder clone of TAM/ML-DS may be established before the acquisition of the GATA1 mutation, or TAM clones with distinct GATA1 mutations may arise independently in the fetal period.

It has long been considered that the linear sequential acquisition of genetic alterations induces disease progression in TAM/ML-DS.³⁷  By contrast, recent studies using high-throughput genomic technology indicate that evolutionary trajectories are more complex and branching in other cancers and leukemias, as previously proposed by Nowell.³⁸  In this theory, genomic instability in founder cells gives rise to heterogeneous mutant subclones, and under selective pressure, some subclones expand to result in disease progression, whereas others become extinct or remain dormant. Thus, leukemic clones may evolve and emerge through the complex interaction of selectively advantageous “driver” mutations, additional advantageous “cooperating” mutations, neutral “passenger” mutations, and deleterious mutations.^32,38  It is clinically true that genomic alterations are more frequently observed in ML-DS than in TAM.^2,11,39  In this paper, we showed that diverse subclones with various CNAs can be generated in TAM, and these events occurred preferentially in a patient who later developed ML-DS. These findings suggest the presence of leukemic driver mutations in the early phase of TAM in this patient, which may have induced genomic instability. We were unable to find any candidate tumor-associated genes on the deletion sites (3q24, 9q22, 12p12, 16q22, and 16q24) of TAM-derived cells using the The National Center for Biotechnology Information database, suggesting that other genetic mutations and epigenetic events may contribute to the progression to ML-DS, including a few candidate mutations identified previously.^40-42  It is also noteworthy that subclones in each recipient mouse showed different repopulating capacities in this study. The dominant clones in each recipient were not always identical in the 1° generations, and the dominant clone in a certain recipient did not always propagate dominantly in the next generation recipients (Figure 4A). Differences between the recipient mice or technical problems may have caused variations in engraftment outcome, which is a potential weakness of this xenograft model; however, it is more likely that cooperating genetic event(s) important for leukemogenesis led to the cells of a specific TAM clone becoming the dominant population in each recipient. Such cooperating event(s) could have a considerable impact on a preleukemic TAM clone, and clonal selection might occur in a somewhat random manner. Thus, leukemic evolution may depend on random chance to an extent. Our TAM xenograft model may help demonstrate the branching architecture of clonal evolution in a preleukemic phase, which contrasts with a linear and deterministic pattern of evolution.^34,38  Further genomewide analysis is needed to elucidate the true driver or cooperating mutation(s) and unravel the evolutionary process of leukemia.

In conclusion, we established a xenograft model of TAM using highly immunodeficient NOG mice. Our model enabled the observation of clonal selection and expansion of minor mutant TAM clones and is likely to mimic the early phase of the leukemic evolutionary process, demonstrating the striking genetic heterogeneity and the propagating potential of minor clones in a preleukemic phase. Our xenograft model could be valuable tool for gaining insight into the leukemogenesis of ML-DS and for evaluating the prognosis of TAM patients.

The authors thank all the TAM patients and their families for their participation. The authors thank Drs Akira Niwa, Masashi Sanada, Hironao Numabe, Tomoki Kawai, Takahiro Yasumi, and Ryuta Nishikomori for technical advice.

This work was supported by grants from the Japanese Ministry of Education, Culture, Sports, Science, and Technology, and from the Japanese Ministry of Health, Labor and Welfare.

Contribution: S.S., I.K., T.M., H.F., and K.U. performed sample collection and processing; S.S., K.T., K.Y., and R.W. performed experiments; A.S.-O. performed microarray analysis (accession number GSE44739); Y.S. and S.M. provided expert statistical analysis; S.S., Y.O., and T.T. analyzed results and made the figures; M.I. and T.N. generated NOG mice; and S.S., K.W., H.H., S.A., E.I., S.O., and T.H. designed the research and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Toshio Heike, Department of Pediatrics, Graduate School of Medicine, Kyoto University, 54 Kawahara-cho, Shogoin, Sakyo-ku, Kyoto, 606-8507, Japan; e-mail: heike@kuhp.kyoto-u.ac.jp.