Primary AML cells

Fresh BM or PB samples from adult AML patients were obtained from the Department of Internal Medicine III, Ludwig-Maximilians-Universität, Munich, Germany, during the years 2012 and 2013. Specimens were collected for diagnostic purposes before start of treatment. Written informed consent was obtained from all patients or legal guardians in the cases where patients were minors. The study was performed in accordance with the ethical standards of the responsible committee on human experimentation (written approval by Ethikkommission des Klinikums der Ludwig-Maximilians-Universität, Munich, number 068-08) and with the Helsinki Declaration of 1975, as revised in 2000. Pediatric AML samples were described previously [19].

Mice

NOD.Cg-Prkdc^scid IL2rg^tm1Wjl/SzJ (NSG; The Jackson Laboratory, Bar Harbour, ME, USA) were maintained under specific pathogen-free conditions in the research animal facility of the Helmholtz Zentrum München, Munich, Germany. Animals had free access to food and water, and were housed with a 12-hour light–dark cycle and constant temperature. All animal trials were performed in accordance with the current ethical standards of the official committee on animal experimentation (written approval by Regierung von Oberbayern, number 55.2-1-54-2532-95-10). When clinical signs of illness became apparent (more than 60% leukemic cells within PB, rough fur, hunchback, or reduced motility), mice were sacrificed equally in all passages. If leukemia became not apparent, mice were killed and analyzed 25 weeks after cell injection by latest.

Recovering PDX cells from mice and serial transplantation

From first generation mice harboring primary AML cells (see S1 Methods), PDX cells were reisolated out of femurs, tibiae and spleen by mincing the tissues and filtration through a cell strainer, followed by Ficoll gradient centrifugation in case of splenic cells. PDX AML cells were identified by staining for human CD45, CD33, CD3 and CD19 and flow cytometry analysis (see S1 Methods). Without further enrichment or manipulation, 1x10⁶–1x10⁷ total BM cells were reinjected into next recipient NSG mice for reexpansion (secondary transplantation).

Targeted next generation deep sequencing

Genetic characterization of patient specimens and PDX cells was performed using a targeted, multiplexed amplicon resequencing approach (Haloplex, Agilent, Boeblingen, Germany). Mutational hotspots or the entire coding sequences, as appropriate, of 43 genes known to be recurrently mutated in myeloid malignancies (S1 Table) were amplified from 250 ng of genomic DNA. The resulting libraries were sequenced on an Illumina MiSeq instrument (Illumina, San Diego, USA) using 2x250 bp paired-end reads. Sequence data were aligned to the human reference genome (version hg19) using BWA [32]. Single nucleotide variants and short insertions or deletions were called using the VarScan 2 [33] and Pindel algorithms [34], respectively, using a custom data analysis pipeline. Known and putative leukemia-associated mutations were identified through a review of published data and public databases including COSMIC and dbSNP, and novel variants were assessed based on their functional consequences on the amino acid level (i.e., nonsense and frameshift mutations were considered pathogenic). A variant allele frequency (VAF) of 2% was used as the threshold for mutation calling.

Statistical analysis

Mean and standard deviation of the mean were calculated using the Microsoft Excel 2010 software (Microsoft, Redmont, WA, USA). LIC frequencies were calculated according to Poisson statistics using the ELDA software application (http://bioinf.wehi.edu.au/software/elda).

Additional Methods

Please refer to S1 Methods for description on first engraftment of primary AML cells in NSG mice; monitoring engraftment of PDX cells by flow cytometry analysis of PB; surface antigen characterization of PDX AML cells by flow cytometry; cloning; lentiviral transduction; enrichment of transgene-expressing cells; in vivo bioluminescence imaging (BLI); quantification of BLI pictures; and Limiting dilution transplantation assay (LDTA).

High consistency of AML-specific mutations between primary and PDX AML cells

According to published protocols [6,16,17,19,35], we transplanted cells from 29 adult patients with AML. Patients' clinical characteristics are depicted in Table 1. In line with published data, 17/29 (59%) samples engrafted in NSG mice, defined by more than 0.1% hCD45+ hCD33+ cells within BM within 20 weeks after transplantation [17,35–38], although with broad differences regarding in vivo growth characteristics (Fig. 1; Table 1; see S1 Results for details).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Engraftment of primary AML cells in NSG mice predicts reengraftment capacity.

10⁷ fresh primary AML cells were injected and successfully engrafted in NSG mice; shown are characteristics of the first engraftment regarding passaging time (time period from cell injection until clinical signs of leukemia or latest between 20 and 25 weeks) (A); percentage of cells positive for both hCD45 and hCD33 at time of sacrifice within mouse PB (B) and within BM (black cubes) or spleen (grey circles) (C). Each mark visualizes data obtained from a single mouse. Open cubes indicate 0% human cells. Dotted line discriminates samples that reengrafted in secondary recipients from samples that did not. Please refer to S1A Fig. for exemplary FACS plots.

https://doi.org/10.1371/journal.pone.0120925.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Patients’ clinical characteristics.

https://doi.org/10.1371/journal.pone.0120925.t001

AML is a genetically heterogeneous group of diseases, and various gene mutations have been shown to be associated with treatment response and patient outcome [39]. Furthermore, AML patients may harbor multiple, genetically related disease subclones [40,41]. The benefit of using patient-derived cells in preclinical in vivo models depends on their ability to faithfully recapitulate the genotypic heterogeneity of the disease. Thus, we performed comprehensive analyses as quality controls, as proposed recently [41]. Using a targeted resequencing approach that covers 43 genes with known roles in AML pathogenesis (S1 Table), we studied six primary patient samples and matched xenografts. Per sample, we identified two to six mutations (Fig. 2, Fig. 3F; S2 Table). Each patient sample contained mutations with a variant allele frequency (VAF) of close to 50%, indicating that these variants were present in most cells, assuming heterozygosity. All of these mutations, which mark the founding clone of the AML cell population, were preserved in the PDX cells recovered from mice after amplification, consistent with data published previously [41]. In each primary sample, we detected additional mutations with a significantly lower VAF, suggesting the presence of subclones within the AML cell population [41,42]. Most of these mutations were also detected within the respective PDX cells. In two cases, we observed an increase in the VAF of mutations in PDX compared to primary cells, indicating that the subclone carrying the mutation had an engraftment or growth advantage (AML-361 and AML-393; Fig. 2 and Fig. 3F). Four samples showed evidence of polyclonal engraftment. In these cases, variants found in the patient at a low VAF were detected in PDX cells with similar, low allelic frequencies, that were highly consistent between multiple mice transplanted in parallel (AML-372, AML-373, AML-407, and AML-412; Fig. 2). In two samples, small subclones (VAFs of the subclone-specific mutations below 10%) were detected in the patient sample but were undetectable in mice (AML-373 and AML-412). These subclones may have had a relative engraftment or proliferation disadvantage in our model. In one sample from a patient with a FLT3-ITD (60bp length), we detected a second, additional FLT3-ITD (96 bp length) in PDX cells obtained from three mice injected in parallel (VAF 2–4%). This second mutation was not detectable in the primary specimen (AML-373). Since the same, additional FLT3-ITD emerged in all three mice that were inoculated with the patient’s cells in parallel, we conclude that a small subclone carrying this variant was present in the patient sample, albeit below our limit of detection. After xenotransplantation, the subclone carrying the 96bp FLT3-ITD expanded relative to the 60bp FLT3-ITD-mutated clone, and thus had an engraftment or growth advantage in our model. Our data are in line with published results describing the challenge of reliably mimicking rare subclones upon xenotransplantation [41].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Engraftment and retransplantation of AML cells in NSG mice conserves genetic alterations of the primary sample.

Primary AML patient samples and matched PDX cells, reisolated out of the BM (CD45 chimerism 80–99%) after first passage in NSG mice (PDX-0) or after 1 or 2 re-transplantation cycles (PDX-1/-2), were characterized by targeted resequencing of 43 AML-related genes (S1 Table). Plots depict variant allele frequencies for each driver gene mutation found within the sample. a/b/c/d/f: PDX cells of three to five mice injected in parallel were analyzed. *: primary cells were frozen and thawed before injection. BCOR (BCL-6 corepressor); CEBPA (CCAAT/enhancer binding protein alpha); DNMT3A (DNA (cytosine-5)-methyltransferase 3 alpha); FLT3 (Fms-related tyrosine kinase 3); ITD (internal tandem duplication); KRAS (Kirsten rat sarcoma viral oncogene homolog); NPM1 (nucleophosmin-1); NRAS (neuroblastoma RAS viral oncogene homolog); SRSF2 (serine/arginine-rich splicing factor 2); TET2 (tet methylcytosine dioxygenase 2); TP53 (tumor protein p53). Raw data is depicted in S2 Table.

https://doi.org/10.1371/journal.pone.0120925.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. PDX AML cells allow genetic engineering without altering molecular sample characteristics.

(A) Scheme of the process of generating transgenic PDX (t-PDX) AML cells. PDX cells were transduced after first or second retransplantation cycle. (B) Scheme of the vector constructs. (C) Transduction rate in t-PDX AML cells after lentiviral transduction and cell amplification in mice was measured by FACS analysis of fluorochrome or NGFR expression. Each mark visualizes data obtained from a single transduction. Open mark: no transgenic cells were detectable. (D) Enrichment of transgenic cells using flow cytometry was measured using mCherry expression after cell amplification in mice. (E) Genetic engineering does not alter immunophenotype; primary cells, untransduced PDX cells after fourth retransplantation and enriched transgenic t-PDX cells were analyzed by multicolor flow cytometry; specific fluorescence intensity is depicted. See also S3C Fig. for exemplary FACS plots of AML-372. Raw data is depicted in S3 Table. (F) Genetic engineering does not markedly alter AML-specific mutations; genomic DNA was isolated out of primary cells, untransduced PDX cells and enriched transgenic t-PDX cells; VAF of mutations was profiled by targeted resequencing. BCOR (BCL-6 corepressor); KRAS (Kirsten rat sarcoma viral oncogene homolog); NRAS (neuroblastoma RAS viral oncogene homolog); TP53 (tumor protein p53). Raw data is depicted in S2 Table.

https://doi.org/10.1371/journal.pone.0120925.g003

These results illustrate several important characteristics of our model: (i) driver mutations present in the founding clone of primary AML specimens are preserved in xenografts; (ii) certain subclones present within the patients’ primary cell population may have an engraftment or growth advantage or disadvantage in mice, as indicated by the increase or decrease, respectively, of the VAF in PDX cells compared to the patient sample; (iii) in the three samples that were xeno-transplanted in multiple mice in parallel, engraftment characteristics of subclones are reproducible; (iv) PDX cells do not gain mutations in the studied AML driver genes, if they were unmutated in the patient's specimen.

Serial transplantation of PDX AML cells

Next, capacity of initially engrafted cells to reengraft in next generation recipient mice was analyzed. 10/17 (59%) of established PDX samples, corresponding to 10/29 (34%) of all transplanted samples, reengrafted after first retransplantation. As positive predictive markers for successful repassaging, we identified the following growth characteristics of primary cells in the first transplantation: (i) overt leukemia in mice within 20 weeks after cell injection, accompanied by (ii) a substantial percentage of human cells in mouse BM and (iii) more than 5% leukemic cells in the blood at time of sacrifice (Fig. 1, S1B Fig.). A threshold of 0.1 to 1% human cells within BM to define positive engraftment, as proposed by others [10,17,35–38], was not sufficient to reengraft these cells in our hands, and might be explained by the absence of long-term LICs.

Furthermore, serial transplantation of PDX cells was evaluated, to enable repetitive and reproducible experiments. Samples were retransplanted in up to six consecutive generations of mice. Until now, 6 samples could be analyzed for capacity to reengraft serially for more than three passages. 5/6 (83%) samples engrafted serially in all retransplantation experiments performed (S2 Fig.). Only one specimen (AML-373) did not allow retransplantation for more than one cycle. These results are in line with published data where successful secondary and tertiary engraftment was used as prove for the presence of long-term LICs which were not present in all AML samples [9].

Serial reengraftment of PDX AML samples was further characterized. Similar between all samples, murine BM consisted mainly of human leukemic cells at time of overt illness (minimum 80%, maximum 99.8%). In contrast, other characteristics varied widely between samples, including passaging time (minimum 4 weeks, maximum 15 weeks), percentage of leukemic cells in PB (minimum 1%, maximum 99.8%), or amount of leukemic cells in the spleen (minimum 1x10⁶, maximum 7x10⁸) (S2 Fig.). Splenic enlargement was observed in one sample (AML-393), while all other samples showed only minor infiltration of the spleen. This is in contrast to PDX ALL cells, where splenomegaly represents a common feature between all samples and 1x10⁸ human cells can regularly be reisolated out of the spleen [31], but consistent with other studies analyzing AML engraftment in NSG mice [17].

Serial transplantation might affect clonal composition of PDX cells. Targeted resequencing of serially transplanted cells revealed that they still contained the same leukemia-associated mutations as primary patient specimens and PDX cells after initial engraftment (Fig. 2; S2 Table). In addition, analysis of six AML associated antigens revealed that the immunophenotype of PDX cells after four retransplantations was similar to the primary specimens for most antigens tested (Fig. 3E, S3 Fig.; S3 Table). Thus, serial expansion did not severely alter molecular sample characteristics of AML cells, indicating that PDX cells closely resemble the disease they originate from.

Genetic engineering in AML PDX cells

Until now, preclinical studies in AML were restricted to established AML cell lines [22–24], although they might inherit unphysiologic mutations acquired upon prolonged in vitro culture. The use of PDX AML cells will allow repetitive molecular studies on cells highly related to the patients, but these cells have never been used for genetic engineering so far.

Here, we established lentiviral transduction of PDX AML cells, which we and others had successfully used to genetically modify PDX ALL cells [28–31] or primary AML cells [9] (Fig. 3A), although lentiviral transduction holds the risk of additional untargeted genetic alterations [43]. Lentiviral constructs coding for, e.g., luciferase and the red fluorochrome mCherry or the antigene nerve growth factor receptor (NGFR), were cloned under the control of the EF1-alpha promoter; transgenes were connected by a T2A linker allowing equimolar expression (Fig. 3B).

Four PDX samples from adult AML patients as described above (AML-361, AML-372, AML-393, AML-412) were included into further studies together with two pediatric PDX samples (AML-346, AML-356) that had been engrafted and serially transplanted before [19] (S4 Table). After transduction and reamplification in mice, percentage of transgenic PDX (t-PDX) cells was determined by flow cytometry. For one sample (AML-361), no transgenic cells were detectable after six independent transduction trials. Within the other five samples, transduction efficiency ranged from 1.1% to 90% (Fig. 3C). Although defined factors determining lentiviral transduction efficiencies could not be identified, specific characteristics of samples, constructs and viruses might influence the process. t-PDX cells were enriched by flow cytometry (Fig. 3D) or by magnetic-activated cell sorting (MACs), and were reinjected into another generation of recipient mice (Fig. 3A). To generate samples with a high percentage of transgenic cells (>90%), transgenic cells were enriched by two rounds of cell sorting. As some samples were transduced with a rather low transduction efficiency (Fig. 3C), cell enrichment inherits the threat of clonal selection, as we could not guarantee that all subclones of a sample have been transduced equally. Nevertheless, lentiviral transduction and cell enrichment did not markedly alter sample characteristics like growth behavior and immunophenotype in t-PDX compared to PDX AML cells or primary specimens, respectively (Fig. 3E, S4A Fig.; S3 Table). Furthermore, the mutational pattern within the defined set of AML driver mutations studied and within the resolution applied was not altered in the 2 samples analyzed, apart from a loss of a rare subclone carrying an NRAS mutation (VAF 5%) in t-PDX of AML-372 (Fig. 3F; S2 Table). Transgene expression level remained stable over numerous consecutive transplantations (S4 Fig.). In contrast, CMV promoters were silenced in our ALL model within a short period of time (own unpublished data).

Taken together, lentiviral transduction was feasible in 5/6 (83%) PDX samples; both lentiviral constructs allowed successful transduction indicating a more general applicability of lentiviral constructs. Cell enrichment allowed generation of t-PDX cells without markedly altering sample characteristics. Successful genetic engineering in PDX AML cells now opens a broad variety of possible molecular studies.

Highly sensitive repetitive BLI of t-PDX AML cells in single mice

We next wanted to evaluate transgene expression in PDX cells and used the expression of luciferase as both a convenient example system and an attractive target protein. Expression of luciferase enables in vivo bioluminescence imaging (BLI) for sensitive and repetitive disease monitoring in single living mice over time [44]. BLI would greatly improve analysis of preclinical treatment trials in the individualized mouse model of AML. Until now, monitoring of disease development and treatment efficiencies was hampered by the lack of sensitive and repeatable detection procedures.

A codon-optimized form of firefly luciferase (effluc) was used which enables intense light emission for sensitive monitoring [45]. BLI is frequently and successfully used in tumor cell lines to monitor growth or treatment response in vivo [46–50] or to monitor engraftment of human hematopoietic stem cells in mice [51]. We and others have recently established BLI to monitor PDX ALL cells growing in mice [28–31], while BLI could not be performed in PDX AML cells so far due to the lack of genetic engineering.

1x10⁵ t-PDX cells of AML-372 were injected into mice and repetitively monitored using BLI (Fig. 4A, S5A Fig.). One day after cell injection, positive BLI signals were visible in the liver and the lower extremities most probably emitted by the BM (S5A Fig.). BLI signal persisted and increased within the legs over time. The liver signal observed shortly after cell injection vanished after another day and might be due to AML blasts that had intermediately homed to the liver, but left the liver again. After ten days, BLI allowed detection of signal in the upper extremities and the sternum. Only at late time points close to animal death, BLI visualized t-PDX AML cells in the spleen (Fig. 4A), reproducing our experience that splenic involvement is difficult to visualize in BLI [31].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. BLI is highly sensitive and reliable in single mice.

(A) 1x10⁵ t-PDX AML-372 cells were injected into two mice. At indicated days after cell injection, mice were monitored by BLI. Images of one representative mouse are shown. See S5A Fig. for further images. (B) BLI signals from the kinetic in A were quantified in both animals (diamonds); cells positive for both hCD45 and hCD33 in PB were analyzed in parallel (circles). (C) t-PDX AML-372 cells were injected into three mice per group at absolute numbers indicated; 1 and 8 days after cell injection, mice were monitored by BLI; images are shown of one representative mouse per group.

https://doi.org/10.1371/journal.pone.0120925.g004

When BLI signals were quantified, signal intensity increased exponentially over time implying logarithmic growth of PDX AML cells in mice, which is identical to results obtained for ALL [31]. Over time, BLI signal increased by more than four orders of magnitude, enabling a broad range of BLI signals for sensitively discriminating different AML tumor burdens. By using BLI, we were able to visualize AML growth in mice markedly earlier than by flow cytometry measurement in PB, where human cells were detectable not before six weeks after cell injection (Fig. 4B). Furthermore, BLI using effluc-t-PDX cells seems to be more sensitive and covers a broader range than imaging applying Alexa Fluor 680-labeled monoclonal antibodies [52].

To determine the sensitivity of BLI, different amounts of t-PDX AML cells were injected and mice were imaged as early as 18 hours after cell injection. Application of 1x10⁵ (3/3 mice) and 3x10⁴ (2/3 mice) t-PDX AML cells resulted in a positive BLI signal at both femurs, whereas 1x10⁴ cells yielded no detectable signal (Fig. 4C). Thus, PDX AML cells seem to home to the mouse BM within few hours. Assuming that (i) each mouse has about 3x10⁸ BM cells in total (see supplemental results for calculation of this value) and that (ii) within 18 hours from injection 100% of injected t-PDX cells had successfully homed to the BM, but (iii) cells had not yet started to proliferate, BLI was able to visualize less than 30000 leukemia cells per mouse, hence one human cell among 10000 murine BM cells (0.01%). As homing efficiency is unlikely to reach 100%, BLI probably is even more sensitive.

Thus, expression of luciferase in PDX AML cells enables repetitive BLI for highly sensitive visualization of t-PDX cells in mice at tumor burdens which range within the different definitions of minimal residual disease in AML patients [53], to analyze homing efficiency of PDX cells into the BM niche, and for convenient monitoring of leukemia development in vivo.

BLI for quantifying leukemia stem cell surrogates

AML represents a stem cell disease, and various scientific questions require quantification of leukemia initiating cells (LIC) as stem cell surrogates [20]. The gold standard method for determining LIC frequency is the limiting dilution transplantation assay (LDTA) [6,36,37,54–57]. LDTAs are usually quantified by post mortem analysis of numerous mice on the same day, e.g., 16 weeks after cell injection, and thus evaluating a single time point. We validated the use of BLI for monitoring LDTA over time.

Limiting numbers of t-PDX AML-372 cells were injected into groups of mice and repetitive BLI was performed in all animals (Fig. 5A). Reproducing results obtained before, 2/3 mice receiving 3x10⁴ cells were positive in BLI one day after cell injection (S5B Fig.). Seven days later, 2/3 mice of the next dilution (1x10⁴ cells) became positive in BLI; at day 29 after cell injection, all mice that would ever engraft were positive in BLI (Fig. 5A-B; S5 Table). Calculation of LIC frequency showed a clear time-dependent increase, which remained stable as early as four weeks after cell injection (Fig. 5C). Similar results were obtained in t-PDX AML-346 cells (S6 Fig.; S6 Table). LIC frequencies were 1 in 5.1x10³ (AML-372) and 1 in 2.8x10³ (AML-346), consistent with data obtained from sorted primary AML samples [56]. Thus, BLI enabled reliable determination of LIC frequencies in LDTAs within short periods of time.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. BLI facilitates quantifying leukemia stem cell surrogates.

After limiting dilution of t-PDX AML-372 (98% hCD45+ hCD33+ mCherry+), cells were injected into three mice per group (12 mice total) at absolute cell numbers indicated; leukemia development was monitored by BLI over time. (A) Images of one representative mouse per group. See S5B Fig. for further images. (B) BLI signals from the kinetic in A were quantified. Shown are mean values plus/minus standard deviation. (C) LIC frequency was calculated at each time point using ELDA software. Raw data is depicted in S5 Table.

https://doi.org/10.1371/journal.pone.0120925.g005