Tracking preleukemic cells in vivo to reveal the sequence of molecular events in radiation leukemogenesis

Epidemiological studies have demonstrated an increased leukemia incidence following ionizing radiation exposure, but to date, the target cells and underlying mechanisms of radiation leukemogenesis remain largely unidentified. We engineered a mouse model carrying a different fluorescent marker on each chromosome 2, located inside the minimum deleted region occurring after radiation exposure and recognized as the first leukemogenic event. Using this tailored model, we report that following radiation exposure, more than half of asymptomatic CBA Sfpi1GFP/mCh mice presented with expanding clones of preleukemic hematopoietic cells harboring a hemizygous interstitial deletion of chromosome 2. Moreover, following isolation of preleukemic hematopoietic stem and progenitor cells irradiated in their native microenvironment, we identified the presence of Sfpi1 point mutations within a subpopulation of these preleukemic cells expanding rapidly (increasing from 6% to 55% in 21 days in peripheral blood in one case), hence identifying for the first time the presence of such cells within a living animal. Importantly, we also report a previously undescribed gender difference in the phenotype of the preleukemic cells and leukemia, suggesting a gender imbalance in the radiation-induced leukemic target cell. In conclusion, we provide novel insights into the sequence of molecular events occurring during the (radiation-induced) leukemic clonal evolution.

All data generated previously used fixed leukemic cells to study Del2 and Sfpi1 mutations, thus limiting further characterization of leukemogenesis. Here, we crossed CBA Sfpi1 GFP/GFP mice [16] with a newly generated CBA Sfpi1 mCh/mCh transgenic model to create an F1 CBA Sfpi1 mCh/GFP mouse expressing mCherry from a Rosa26 promotor construct located in the chromosome 2 MDR, and GFP being expressed from the other allele under the Sfpi1 promoter. Monthly blood sampling post-irradiation was used to monitor Del2, and preleukemic clonal expansion, by flow cytometry.
We report that more than half of mice presented with preleukemic cells harboring Del2. Moreover, we identified for the first time the presence of Sfpi1 point mutations within subpopulations of these preleukemic cells, within a living animal. We also provide evidence of a gender difference in the (pre)leukemic phenotype, suggesting a difference in the leukemic target cell between male and female mice.

Methods
Mice, rAML induction, and tissue preparation CBA Sfpi1 mCh/mCh mice were generated as previously described [17], and mated to CBA Sfpi1 GFP/GFP mice [16] to generate F1 CBA Sfpi1 mCh/GFP mice. Mice were given single 3Gy whole-body X-irradiation at 10-12 weeks of age (70 males and 50 females). Sham-irradiation of agematched mice (n = 20; assigned at random) was performed by placing the mice into the irradiator box for the appropriate time without X-rays being produced. rAMLs were diagnosed as described previously [16], using the criteria of the 'Bethesda proposals for classification of nonlymphoid neoplasms in mice' [18]. Spleen tissue was stored at −70°C in RNAlater® (Ambion, Austin, US) for nucleic acid extraction, in 4% formaldehyde for histopathological analysis, or disaggregated for FACS analysis, as described previously [16]. All animal procedures conformed to the UK Animals (Scientific Procedures) Act, 1986, Amendment Regulations 2012, and animal experimental protocols were reviewed and approved by the local Ethics Committee and the Home Office.

Pyrosequencing analysis
DNA was extracted from blood (10 µl) using DNeasy® kit (Qiagen). Ten ng of DNA was used to amplify the target sequence of Sfpi1 exon 5 by PCR with primer biotinylated forward 5′-GCATCCAGAAGGGCAACC-3′ and reverse 5′-TCGCCTGTCTTGCCGTAGT-3′ primers generating a 79 bp PCR product. Primers, DNA and PyroMark PCR master mix (Qiagen) were combined in a total volume of 25 µl, and amplified: 15 min at 95°C, then 45 cycles of (30 s at 95°C, 30 s at 60°C and 30 s at 72°C). Ten µl of the biotinylated PCR product was used to detect mutations on the first base of the 235th codon (CGC to TGC) with the following sequencing primer: 5′-CCTGTCTTGCCGTAGT-3′ using PyroMark48 (Qiagen).

Clonal expansion of Del2 hematopoietic cells following radiation exposure
Mice received 3 Gy whole-body X-irradiation and underwent monthly tail vein bleeding for lifespan to identify leukocytes carrying Del2, indicated by fluorescence loss (Supplementary Figure S1A). In sham-irradiated mice, all leukocytes expressed mCherry. GFP expression is controlled by PU.1 promoter, hence lymphocytes did not express GFP (i.e., mCherry+GFP−). Myeloid cells (monocytes and granulocytes) require PU.1 for terminal differentiation and maturation, and expressed GFP throughout lifespan (i.e., mCherry+GFP+). Either PU.1 copy (mCherry or GFP) can be deleted during leukemic transformation. However, mCherry loss was used as the 'lead' for detection of Del2, allowing GFP to be used as a proxy for PU.1 expression from the remaining copy.
To determine PU.1 expression changes on the remaining chr2 after exposure, blood of male mice with clonal expansion of mCherry− leukocytes was analyzed for mCherry/GFP expression. When comparing GFP expression between mCherry− and mCherry+granulocytes, mCherry− granulocytes had a markedly higher GFP expression (Fig. 2a). Similarly, GFP expression was higher in mCherry−Lin− than in mCherry+Lin− cells (Fig. 2b).

Exon 5 Sfpi1 point mutation is solely observed in Del2 hematopoietic cells
There are data suggesting that the Sfpi1 point mutation most probably leads to complete abolition of PU.1 activity [21,22]. To identify and quantify the presence of the 'second' driving mutation, Lin− cells were sorted, based on their mCherry expression, from irradiated male mice with mCherry− blood leukocytes. mCherry allele loss occurred specifically in the sorted mCherry− population (Fig. 3a). Overall, 16% of mCherry−Lin− cells in mouse 2 had the characteristic murine rAML C to T substitution, indicating that these mutations are restricted to Del2 Lin− cells (Fig. 3b). Point mutations were also analyzed in terminal blood samples. Although not detectable in sham or irradiated CBA Sfpi1 mCh/GFP mice with no mCherry− leukocytes, 83% of irradiated mice harboring mCherry− clonal expansion showed detectable levels of C to T substitution, linking point mutation occurrence to a prior presence of Del2 (data not shown). Previously, we reported that murine CBA rAML cases without Sfpi1/PU.1 involvement are rare (<10%; either as chromosome 2 deletion or as Sfpi1/PU.1 exon 5 mutation), and that within this subset of rAMLs, 50% of cases has internal tandem duplications (ITDs) within FMS-like tyrosine kinase 3 (Flt3) [23]. We have now screened a panel of 134 murine rAML samples and found Flt3-ITDs in 3% of the cases, none of them carrying deletions or mutations of Sfpi1/PU.1. Interestingly, 2.2% of cases carried a KRAS G12 mutation (manuscript in preparation).

Simultaneous expansion of competing preleukemic clones
At 15 months, one male mouse presented with an unusual clonal expansion of both mCherry−/GFP+ and mCherry +/GFP− leukocytes, with the point mutation detected in blood leukocytes (6%; Fig. 4a). Three weeks later, the point mutation had increased to 55% in blood (58% in spleen; Fig. 4a were sorted. Despite being asymptomatic at sacrifice, this mouse had leukemic pathological features, demonstrating that appearance of the point mutation in Del2 blood cells was indicative of a preleukemic mouse at least 1 month prior to AML presentation. Hence, the point mutation occurs late in the leukemogenic process, but presents a rapid increase. Furthermore, both mCherry−/GFP+ and mCherry +/GFP− leukocytes were detected. A schematic representation of simultaneous expansion of two preleukemic clones (one clone with GFP loss and one with mCherry loss) is presented for this leukemic case (Fig. 4b). Sfpi1 mutation was analyzed in mCherry+Lin− and mCherry −Lin− cells, and most mCherry− and mCherry+ cells carried the mutation (63% and 71%, respectively); loss of the GFP allele in mCherry+Lin− cells was also confirmed (Fig. 4b).

Dysregulated PU.1 expression upon leukemic progression
We demonstrated that following Del2, PU.1 expression on the remaining homolog is upregulated, indicative of a negative feedback mechanism; with upregulation of the intact chr2 PU.1 promotor being a counter-balancing mechanism. However, in several cases, this feedback mechanism became dysregulated, possibly when the point mutation occurs, resulting in leukemia development. At 6 months, GFP expression of mCherry− granulocytes slightly decreased compared with mCherry+granulocytes (Fig. 5a). By 9 months, GFP expression was no longer increased. One month later, mCherry− leukocytes percentage had increased further. Shortly thereafter, the mouse presented with outward physical signs of rAML. Although still expressing GFP, 93.5% of cells were mCherry−, indicating Rosa26-mCherry chr2 homolog loss. Spleen cell immunophenotyping showed that mCherry−GFP+ cells expressed immaturity cell surface markers (Fig. 5b). Isolated DNA from these cells revealed a TGC sequence. Blood smear showed nucleated cells with blastic appearance while leukemic spleen cells had lost Rosa26-mCherry construct (Fig. 5c).
Notably, most female mice had clonal mCherry− lymphocytes expansion. Blood analysis of a female mouse showed that myeloid leukocytes retained mCherry expression at 13 months but 1.9% of lymphocytes were mCherry −, increasing further over time. At 17 months, the mouse presented with AML. All mCherry− leukemic spleen cells had lost mCherry and GFP expression, indicative of a lymphoid origin. Immunophenotyping showed that cells only expressed CD31 and B220 (Supplementary Figure   Fig Figure S4B). Loss of Rosa26-mCherry construct was confirmed. While leukemic cells did not express GFP (i.e., PU.1 downregulation upon commitment towards the lymphoid lineage), the GFP construct was still detectable by PCR (Supplementary Figure S4C). Seven of the eight female mice were diagnosed with lymphocytic leukemia, while the remaining female was diagnosed with myeloid leukemia, with a much shorter latency compared with male AML cases (5 months and 11 months, respectively).

Discussion
Using an engineered mouse model, we assessed Del2 occurrence after radiation exposure. A significant number of studies into the cytogenetics of rAML in various inbred mouse strains [24][25][26] revealed that structurally abnormal chr2-usually consisting of large hemizygous interstitial deletion (i.e., Del2)-occurs in~90% of rAML cases, and typically is detectable in 90-100% of the leukemic cells within any sample [27]. It represents by far the most common and consistent chromosomal aberration seen in rAML mouse models. Of the other cytogenetic abnormalities identified in these studies, change in chromosome number is the most frequent. In particular, loss or gain in Y chromosome. More recently, we performed array comparative genomic hybridization (aCGH) at unprecedentedly high Leukemic progression following Sfpi1 dysregulation. a At 6 months, GFP expression in mCherry− granulocytes of a male mouse had decreased slightly compared with mCherry− granulocytes at 4 months. GFP upregulation in mCherry− granulocytes decreased further, and at 9 months, GFP expression was no longer different. b mCherry−GFP+ spleen cells expressed immaturity cell surface markers. c Exon 5 Sfpi1 DNA sequence of the leukemic mouse zoomed in on the CGC codon known to exhibit the point mutation (arrow indicates the point mutation). Blood smear showed nucleated cells twofold larger than RBCs with a blastic appearance. Confirmation of loss of mCherry allele: control DNA was obtained from a wild-type mouse and 2 sham-irradiated CBA Sfpi1 mCh/GFP mice [Rosa26-mCherry allele (510 bp) and wild-type allele (383 bp)] resolution on a unique panel of 79 CBA rAMLs [27]. Besides the characteristic Del2, small deletions were observed on chrs 3, 4, 5, 6, 11, and 16 in individual cases, but importantly no consistent event was identified.
At 9 months after radiation exposure, 25% of males presented with mCherry− leukocytes, in line with previously reported percentages in irradiated mice with BM Del2 clones, detected with cytogenetic methods [14]. All these male mice had mCherry loss in both myeloid and lymphoid lineages, indicative of a primitive hematopoietic cell type (HSC or MPP1). Using SLAM markers, we identified mCherry− HSCs (LSK CD48−CD150+) [28] in mice with mixed lineage mCherry− clonal expansion in blood. These mice had mCherry− myeloid and lymphoid progenitors, based on differential CD127 expression in these cells [29,30], indicating that Del2 HSCs gave rise to mCherry− daughter HSCs as well as mCherry− lineagecommitted progeny. At 15 months, 8 out of 18 males presented with lymphoid lineage-only mCherry− cells, indicating that in these mice Del2 likely occurred in a hematopoietic cell-like a CLP giving rise to lymphoid progeny specifically.
We assessed PU.1 expression from the remaining chr2 homolog in mice with mCherry− leukocytes (GFP as reporter for PU.1 transcription) [27,31]. GFP expression in mCherry− myelocytes was 1.4-fold higher compared with mCherry+myelocytes. It can be directly attributed to an increased PU.1 promoter activity through PU.1 autoregulation to compensate the loss of the second allele. It was previously reported that both URE and Sfpi1 proximal promoter have binding sites for PU.1 itself [32][33][34]. Here we report that during leukemogenesis, autoregulation of PU.1 was observed in mCherry− Lin− cells. This autoregulation becomes dysregulated, resulting in GFP expression decreasing to the level of mCherry+leukocytes. In these cases, immunophenotyping showed immature phenotypes consistent with myeloid leukemia. We also confirmed that the remaining Sfpi1 copy carried a point mutation replacing arginine 235 with cysteine (the most common C to T substitution in murine rAMLs) [35] probably impairing transcriptional autoregulation. Our experimental observations are consistent with a two-hit model of murine radiation leukemogenesis in which the first irreversible mutational hit (deletion of one Sfpi1 copy) results in preleukemic cells with growth advantage. Subsequently, they acquire a second mutational hit (point mutation in the remaining Sfpi1), leading to full malignancy, clonal expansion and leukemia [13].
It could be hypothesized that the point mutation occurs when the GFP levels return to 'normal' levels (i.e., equal to mCherry+ leukocytes), linking point mutation occurrence with loss of PU.1 autoregulation. Data obtained from terminal blood samples of irradiated mice revealed that 83% of mice had mutated cells. In contrast, the point mutation couldn't be detected neither in sham-irradiated mice nor irradiated mice without mCherry− leukocytes. In addition, point mutations were exclusively detected in the BM mCherry−Lin− cell fraction of mice with mCherry− leukocytes. Most importantly, these observations link, for the first time, the occurrence of the point mutation to a prior presence of Del2. The leukemic case described with two competing preleukemic clones provides evidence of how rapid the leukemic progression occurs following acquisition of the second hit (6% to 55% in merely three weeks). To the best of our knowledge, only one study so far reported an alternative mutation occurring in a mutually exclusive way with Sfpi1/PU.1 [23]. RFLP and sequencing analysis revealed in a subset of rAMLs the presence of Flt3-ITDs (one of the most frequent mutations in human AMLs [36]). These ITDs are similar to those seen in human AML cases and the mutual exclusion with Sfpi1/PU.1 mutations suggests that Flt3 mutations are driver mutations in these rare rAML cases. We have now screened a panel of 134 CBA rAML samples and found Flt3-ITDs in 3% of the cases (unpublished data). Although this is clearly a minor pathway, it represents the only significant alternative pathway to rAML identified so far in this mouse model with a direct link to human AML. In addition, we recently found a few rAML with leukemic cells carrying a KRAS mutation concomitantly to Del2 and Sfpi1 mutations (data not shown), none of them with additional Flt3-ITDs, suggesting different leukemogenic pathways.
Overall male rAML incidence was 9% with an average latency of 11 months, consistent with previous work in male CBA mice [8,16,37]. One in three females had mCherry− lymphoid but not myeloid leukocytes at 12 months. The mean leukemic latency was similar to males (i.e., 13 and 11 months, respectively; P = 0.49). All but one male leukemia cases were categorized as myeloid leukemia. In contrast, 7 out of 8 female leukemia cases were categorized as lymphoid leukemia. Previous work reported AML incidences higher among irradiated males than females. [16,[38][39][40] Our data suggest that a gender-specific hematopoietic cell subpopulation leads to clonal expansion (HSC/ MPP1 vs CLP). Interestingly, PU.1-knockout in lymphoid progenitors does not alter B-cell-maturation or proliferation [41], possibly explaining the mature B-cell phenotype in our female AMLs (i.e., B220 expression). On the other hand, PU.1-knockout in myeloid progenitors inhibits their maturation but not their proliferation [41], consistent with the immature myeloid phenotype observed in our male AMLs. However, we assume that all lineages are prone to Del2 to the same extent, but that cell intrinsic or external factors promoting their expansion, may be gender-specific.
An intrinsic gender-specific radiosensitivity is highly unlikely, and albeit not yet well understood, sex hormones might have a pivotal role. For example, castration does not affect rAML incidence, whereas ovariectomy results in a twofold increase [42]. Interestingly, gonadectomy does not alter leukemia incidence in unexposed mice [42]. It was also shown that estrogen binding to its receptor enhanced HSC self-renewal by upregulation of cell-cycle genes [43]. Follow-up of atomic bomb survivors reported that female AML baseline rate is~40% of that for men who displayed a more rapid increased AML incidence rate with attained age [1]. In conclusion, we propose, based on our experimental data, a model with gender dependent leukemic pathways (Supplementary Figure S5). Our study provides novel insights into (radiation) leukemogenesis, and the model should enable further deciphering of this complex multistep process.
Acknowledgements The transgenic reporter gene model C57BL/6 Sfpi1 GFP/GFP was generously provided by Professor Stephen Nutt from the Walter and Eliza Hall Institute in Melbourne, Australia.

Compliance with ethical standards
Conflict of interest The authors declare that they have no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commonslicense, unless indicated otherwise in a credit line to the material. If material is not included in the article'sCreative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/