The influence of mutational status and biological characteristics of acute myeloid leukemia on xenotransplantation outcomes in NOD SCID gamma mice

Abstract

Purpose

This study aimed at analyzing the association of gene mutations and other acute myeloid leukemia (AML) characteristics with engraftment outcomes in immunodeficient mice and to select the engraftment outcomes that best reflect patient survival.

Methods

Mutations in 19 genes as well as leukemia- and patient-related characteristics were analyzed for a group of 47 de novo AML samples with respect to three engraftment outcomes: engraftment ability, engraftment intensity (percentage of hCD45⁺ cells) and engraftment latency. Leukemia-related characteristics were additionally analyzed in an extended group of 68 samples that included the 47 de novo samples, and additional 21 samples from refractory and relapsed cases. Engraftment outcomes were compared with overall and event-free survival of the patients.

Results

For the 47 de novo samples, no single mutation influenced engraftment, whereas the NPM1^mut/DNMT3A^mut co-mutation was associated with higher engraftment ability. NPM1^mut/FLT3-ITD^neg had lower engraftment intensity. Among leukemia-related characteristics, a complex karyotype was associated with higher engraftment intensity. Among patient-related characteristics, higher cytogenetic risk was associated with higher engraftment intensity, and failure to achieve clinical remission was associated with shorter engraftment latency. In the extended group of 68 samples, white blood count was associated with higher engraftment ability, and the presence of a complex karyotype was associated with higher engraftment intensity. Association with patient overall survival was seen only for engraftment intensity.

Conclusions

The engraftment of AML was influenced by mutation-interactions and other AML characteristics, rather than by single mutated genes, and engraftment intensity best reflected clinical penetrance of AML.

Electronic supplementary material

The online version of this article (10.1007/s00432-018-2652-2) contains supplementary material, which is available to authorized users.

Introduction

Acute myeloid leukemia (AML) is a heterogeneous malignancy that can be classified based on morphological features, represented by French–American–British classification, and/or together with disease origin and genetic abnormalities, represented by the World Health Organization (WHO) classification (Arber et al. 2016). The broad biological heterogeneity of AML is reflected by marked differences in treatment response and prognosis. Approximately 50% of AML patients carry chromosomal aberrations and can be stratified according to cytogenetic analysis (Gregory et al. 2009). The other group of patients lacking chromosomal abnormalities that have cytogenetically normal AML (CN AML) can be analyzed by molecular-biology methods, although the list of known AML-associated gene mutations was rather limited until recently. During the past several years, next-generation sequencing studies identified numerous mutational events that are involved in AML pathogenesis and can be used to further and better classify AML patients, mainly the subgroup of CN AMLs (Cancer Genome Atlas Research 2013; Metzeler et al. 2016; Papaemmanuil et al. 2016). However, only a few of these molecular lesions are considered for prognostic stratification of AML (i.e., prognostically relevant), whereas the prognostic value of the others remains either unclear or insignificant (Dohner et al. 2017).

Xenotransplantation experiments using immunodeficient mice represent a model of choice for functional studies of AML. A higher/faster engraftment of AML samples in these mice was generally shown to reflect a worse prognosis in patients (Kennedy et al. 2013; Pearce et al. 2006). Thus, AML samples from patients carrying mutations associated with poorer prognosis are generally assumed to show enhanced engraftment following xenotransplantation. To our knowledge, the role of mutated genes in the engraftment of AML samples has so far been investigated only for mutations in NPM1 or FLT3 genes. No impact of mutated NPM1 was observed in multiple studies, and mutated FLT3 was associated with higher engraftment ability only in some studies (Kennedy et al. 2013; Malaise et al. 2011; Pearce et al. 2006; Rombouts et al. 2000; Sanchez et al. 2009). Side by side comparison of engraftment results from different studies is often complicated by different definitions of engraftment outcomes that are usually based either on a simple categorization of engrafting vs non-engrafting samples or only on direct quantification of AML cells in the xenografts. To date, a direct comparison of these different outcomes, particularly in association with a complex set of AML characteristics that include mutational status, has not been performed.

In this study, we, therefore, analyzed how AML engraftment is affected by the mutational status of 19 genes that are typically mutated in myeloid malignancies as well as by other relevant leukemia- and patient-related characteristics. Three specific engraftment outcomes were assessed: ability to engraft (engrafting vs non-engrafting samples), resulting engraftment intensity (percentage of human leukocytes in the xenograft) and engraftment latency (time required to produce engraftment). The relevance of each of these engraftment outcomes to the actual outcome in patients was compared based on patient overall survival (OS) and event-free survival (EFS).

Methods

Patient and sample characteristics

Samples were collected from patients treated at the University Hospital, Brno. All patients provided written informed consent and the study was approved by local ethical committee at the University Hospital, Brno. A total of 47 samples were collected at the time of initial diagnosis from the same number of AML patients (referred to hereafter as “diagnostic” samples). Furthermore, 68 AML samples, referred as the “extended sample group”, were obtained from 52 patients. This group included the 47 diagnostic samples plus seven samples collected from patients that failed to achieve clinical remission after induction therapy (“refractory” samples) and 14 samples from hematological relapse (“relapse” samples) cases. Samples were selected with a preference for higher leukocyte counts and previously confirmed DNMT3A, NPM1, or FLT3 gene mutations. Patient characteristics are provided in Online Resource 1. The median age at diagnosis was 52 years (range 19–67 years). A majority of patients (50/52, 96%) were treated with curative intent (single or double induction chemotherapy with 45–90 mg/m² daunorubicin for 3 days, and 100–200 mg/m² cytosine arabinoside for 7 days; post-remission therapy consisted of 2–3 consolidation chemotherapies with 1.5–3.0 g/m² cytosine arabinoside BID on days 1, 3 and 5). Allogeneic hematopoietic stem cell transplantation (HSCT) was performed when indicated based on European Leukemia Net HSCT criteria and if donor was available (Dohner et al. 2010). In total, 26/47 (55%) of the patients from the diagnostic group underwent allogeneic HSCT (Online Resource 1). Meanwhile, 2/47 patients were only treated symptomatically and were not included in treatment outcome analyses. Cytogenetic and flow-cytometric analyses were performed according to standard diagnostic procedures. Patient cytogenetic risk was defined according to the Medical Research Council (MRC) classification (Grimwade et al. 1998). The OS and EFS were analyzed only for patients in the diagnostic sample group. OS was measured from initial diagnosis to patient’s death; patients alive were censored at last follow-up. EFS was measured from initial diagnosis to the date of refraction, relapse (hematological or molecular) or death; patients in stable clinical remission and alive were censored at last follow-up.

Xenotransplantation experiments

Fresh or cryopreserved bone marrow (BM) leukocytes (1–2 × 10⁶), obtained by red blood cell lysis, were injected via the lateral tail vein into 8–12 week-old non-irradiated NOD SCID gamma mice (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ; NSG), with 1–5 mice per AML sample. The injected cells were pre-treated with anti-CD3 antibody OKT3, 1 µg/1 × 10⁶ leukocytes approximately 30–60 min prior to xenotransplantation, to eliminate expansion of T-cells in the xenograft, according to a previously reported protocol (Wunderlich et al. 2014). Animals that developed wasting disease syndromes were sacrificed immediately. Murine BM was obtained from crushed femoral and iliac bones. Animals were maintained according to standard procedures.

Flow-cytometry

Phenotypization of the samples pre- and post-xenotransplantation was performed using fluorescent antibodies against human CD45, CD33, CD34, CD14, CD3, CD19 (Sony Biotechnology, Biolegend), on FACSCanto II and FACSVerse (BD Biosciences). 7-aminoactinomycin D was used to discriminate live cells (Thermo Fisher Scientific).

Mutational screening

Detection of mutations was performed in diagnostic peripheral blood (PB) or BM AML samples. In 40/47 samples, mutational screening of 19 genes (Fig. 2) was performed by targeted amplicon sequencing using the ClearSeq AML Haloplex (Agilent) panel on MiSeq and NextSeq instruments (Illumina). MPL gene was not analyzed, due to poor coverage. To achieve high sensitivity and specificity of variant detection, three variant callers (Pindel, VarDict, VarScan) were used. The mutation detection threshold was set to 2% variant allele frequency (VAF, mutated/[mutated + wild type] × 100). All FLT3-ITDs were confirmed by fragment length analysis, with a detection threshold of 1%. CEBPA mutations were confirmed by Sanger sequencing, due to low coverage. In 7/47 samples that were not analyzed by next-generation sequencing, DNMT3A, NPM1 and FLT3-ITD mutations were analyzed in 6/7 samples, and in 1/47 samples only FLT3-ITD was analyzed. These analyses were performed by conventional Sanger Sequencing (DNMT3A) or fragment length analysis (NPM1, FLT3-ITD).

Fig. 2.

Number of mutation-positive engrafting (black bars) and non-engrafting (gray) diagnostic AML samples. Open bars represent the corresponding number of samples lacking the given gene mutation. Mutational status of all genes was analyzed in all 22 engrafted samples. Mutational status of NPM1 and DNMT3A was analyzed in 24/25 non-engrafting samples and FLT3-ITD in 25/25 non-engrafting samples. The mutational status of the rest of the genes was analyzed in 18/25 non-engrafting samples

Statistical analyses

In univariate analysis, categorical data were analyzed by Fisher’s exact test, continuous data by Pearson’s r test, and combination of categorical vs continuous data by Mann–Whitney test.

For multivariate analysis, we used the AML sample characteristics as predictor variables and the resulting engraftment outcomes as predicted variables. All predictor variables were checked for their normality. Two sample Hotelling’s T-Squared tests were used to estimate two state predicted variables, engraftment ability and engraftment latency. Generalized linear model (GLM) was used to predict the continuous variable, engraftment intensity. To identify the best multivariate combination, we used additive step analysis wherein the criterion for model quality was the resulting p-value for Hotelling’s T-Squared test and Akaike information criterion (AIC) for GLM. Bartlett’s test was used with each Hotelling’s T-Squared test to filter predictor combinations with non-equal variance–covariance matrices. Only predictors with statistically significant coefficients were selected from the best-scoring GLMs.

Results

Successful engraftment was defined as > 1% of hCD45⁺ cells detected in BM from at least one transplanted mouse per AML sample. The analysis was performed at 12 weeks if peripheral blood check showed > 1% of hCD45⁺ cells; otherwise, the analysis was postponed to 16 weeks. In one case, the mice were sacrificed already at 10 weeks due to development of wasting disease symptoms, and the sample was included in the group engrafting at 12 weeks. Successful engraftment was observed for 22/47 (47%) diagnostic, 1/7 (14%) refractory and 7/14 (50%) relapse samples. In all 30 engrafted cases, the human cells represented a phenotypically homogeneous blast/granulocyte population, with appropriate scatter characteristics (Fig. 1a–c). For 28/30 AML samples, 88–100% of the human cells stained for CD33 or CD34 (Fig. 1a). From the two remaining cases, one carried 65–80% of CD33⁺ or CD34⁺ cells in the human population (three mice), and the other case that was defined as “acute leukemia of ambiguous lineage” showed less than 2% of CD33 or CD34 positive cells in the hCD45⁺ fraction (four mice; Fig. 1b) (Arber et al. 2016). Both samples engrafted with very high intensity (> 90% hCD45⁺ cells in all mice), confirming the leukemic origin of xenografts. Besides leukemic engraftment, we observed an engraftment of a human lymphocyte population (SSC^lowhCD45^high) in only 1/30 samples, represented by CD3⁺ T-cells (2/3 mice; Fig. 1c). The lymphocytes were not included in the hCD45⁺ fraction calculated to assess the engraftment. None of the engrafted samples produced B lymphocyte (SSC^lowhCD45^highCD19⁺) engraftment that would suggest concurrent engraftment of normal hematopoietic stem cells (Risueno et al. 2011; Taussig et al. 2005). The leukemic engraftment in our cohort was, therefore, solely assessed based on the percentage of hCD45⁺ cells, with exclusion of lymphocytes.

Fig. 1.

Flow-cytometry analysis of engrafted samples. a Example of leukemic engraftment with high proportion of CD33⁺ cells (patient J-58, diagnostic sample). b Leukemic engraftment without CD33 or CD34 expression (patient S-86, diagnostic sample). c Leukemic engraftment and concurrent T-lymphocyte engraftment/expansion (patient K-51, diagnostic sample)

Three engraftment outcomes were defined in our study: (1) engraftment ability, or the general capability to produce engraftment. The two further outcomes were evaluated only for successfully engrafted samples: (2) engraftment intensity, defined as the percentage of hCD45⁺ cells in murine BM, ranging from 1 to 100%, and (3) engraftment latency, defined as engraftment at 12 or 16 weeks. The three engraftment outcomes were then correlated with characteristics of the AML samples, arbitrarily divided into: (1) mutational status (analyzed only in diagnostic samples); (2) leukemia-related characteristics (analyzed in all samples), including presence of non-CN AML without complex karyotype, presence of complex karyotype, white blood count (WBC), PB and BM blast percentages, expression of CD34 and CD33 on BM leukocytes; (3) patient-related characteristics (analyzed only in diagnostic samples), including cytogenetic risk stratification, achievement of clinical remission, occurrence of relapse. Furthermore, we have analyzed the effect of two characteristics related to sample processing/quality: (1) number of injected blasts, calculated as (percentage of viable cells/100) × (percentage of blasts/100) × (total number of leukocytes injected), and (2) sample cryopreservation status (freshly isolated vs cryopreserved). Finally, the engraftment outcomes were also compared with patient treatment outcomes, OS and EFS.

Association between engraftment ability and AML sample characteristics

Mutations were found in 15 of the 19 (79%) sequenced genes, in 43/47 diagnostic samples (91%; Fig. 2). Univariate analysis showed no relationship between the presence of any single mutation and the ability of samples to engraft (Table 1). We next analyzed the possible role of interactions of the three most frequently co-mutated genes in our cohort: DNMT3A, NPM1, FLT3-ITD. The NPM1 mutation, regardless of FLT3-ITD status, did not affect engraftment ability, whereas the NPM1^mut/DNMT3A^mut co-mutation was predictive of engraftment, as all 5 (100%) double-mutant samples were successfully engrafted (p = 0.02). The effect of high FLT3-ITD allelic burden could not be evaluated, as only two of the FLT3-ITD positive samples carried allelic ratio above 0.5 (calculated as mutated/wild type; corresponding to > 33% VAF). None of the analyzed leukemia- and patient-related characteristics influenced engraftment of the diagnostic samples (Table 1).

Table 1.

Results of univariate statistical analysis

Sample type	Characteristics	Engraftment ability (Y/N)			Engraftment intensity (hCD45+ %)			Engraftment latency (12/16 weeks)
		n patients	Statistical test	Statistical significance	n patients	Statistical test	Statistical significance	n patients	Statistical test	Statistical significance
Diagnostic samples only	Mutations
	NPM1	19 of 46	Fisher’s exact test	p = 0.4	11 of 22	Mann Whitney	p = 0.1	11 of 22	Fisher’s exact test	p = 0.4
	FLT3-ITD	11 of 47		p = 0.7	6 of 22		p = 0.7	6 of 22		p = 1.0
	DNMT3A	9 of 46		p = 0.3	6 of 22		p = 0.8	6 of 22		p = 1.0
	NRAS	9 of 40		p = 0.5	4 of 22		p = 0.5	4 of 22		p = 1.0
	CEBPA	8 of 40		p = 1.0	4 of 22		p = 0.8	4 of 22		p = 1.0
	CEBPA biallelic	4 of 40		p = 1.0	2 of 22		N/A	2 of 22		p = 1.0
	FLT3-TKD	7 of 40		p = 0.2	2 of 22		N/A	2 of 22		p = 1.0
	TET2	6 of 40		p = 0.7	4 of 22		p = 0.3	4 of 22		p = 1.0
	ASXL1	5 of 40		p = 0.7	3 of 22		p = 0.6	3 of 22		p = 1.0
	IDH1	4 of 40		p = 0.3	1 of 22		N/A	1 of 22		p = 1.0
	IDH2	4 of 40		p = 0.6	3 of 22		p = 0.09	3 of 22		p = 1.0
	U2AF1	3 of 40		p = 0.6	2 of 22		N/A	2 of 22		p = 0.09
	SRSF2	2 of 40		p = 1.0	1 of 22		N/A	1 of 22		p = 1.0
	TP53	1 of 40		p = 1.0	1 of 22		N/A	1 of 22		p = 1.0
	RUNX1	1 of 40		p = 1.0	0 of 22		N/A	0 of 22		N/A
	DNMT3Amut/NPM1mut	5 of 46		p = 0.02	5 of 22		p = 0.8	5 of 22		p = 1.0
	DNMT3Amut/FLT3-ITDmut	3 of 46		p = 0.6	2 of 22		p = 0.4	2 of 22		p = 1.0
	NPM1mut/FLT3-ITDmut	9 of 46		p = 0.7	5 of 22		p = 0.7	5 of 22		p = 1.0
	NPM1mut/FLT3-ITDneg	10 of 46		p = 0.5	6 of 22		p = 0.02	6 of 22		p = 0.1
	DNMT3Amut/NPM1mut/FLT3-ITDmut	2 of 46		p = 0.2	2 of 22		N/A	2 of 22		p = 1.0
	Patient-related
	Diagnostic vs relapse samples	47 vs 14	Fisher’s exact test	p = 1.0	22 vs 7	Mann Whitney	p = 0.03	22 vs 7	Fisher’s exact test	p = 0.4
	Diagnostic vs refractory samples	47 vs 7		p = 0.2	22 vs 1		N/A	22 vs 1		N/A
	Achievement of clinical remission	37 of 45		p = 0.7	18 of 21		p = 0.7	18 of 21		p = 0.03
	Occurrence of relapse	16 of 45		p = 0.8	8 of 21		p = 0.1	8 of 21		p = 0.7
	Intermediate vs poor cytogenetic risk	37 vs 10		p = 0.7	18 vs 4		p = 0.006	18 vs 4		p = 0.08
	Leukemia-related
	Non-CN AML	9 of 43	Fisher’s exact test	p = 0.5	5 of 19	Mann Whitney test	p = 0.4	5 of 19	Fisher’s exact test	p = 1.0
	Complex karyotype	6 of 47		p = 1.0	3 of 22		p = 0.01	3 of 22		p = 0.2
	WBC PB	47	Mann Whitney	p = 0.1	22	Pearson’s r	p = 0.4	22	Mann Whitney	p = 0.9
	PB blast %	47		p = 0.9	22		p = 0.2	22		p = 0.07
	BM blast %	47		p = 0.8	22		p = 0.1	22		p = 0.3
	CD34 expression on LEU	47		p = 0.2	22		p = 0.3	22		p = 0.3
	CD33 expression on LEU	47		p = 0.4	22		p = 0.08	22		p = 0.4
	Sample-related
	Total number of transplanted blasts	46	Mann Whitney	p = 0.1	22	Pearson’s r	p = 0.4	22	Mann Whitney	p = 0.7
	Fresh vs cryopreserved	11 vs 36	Fisher’s exact test	p = 0.7	6 vs 16	Mann Whitney	p = 0.2	6 vs 16	Fisher’s exact test	p = 0.6
Diagnostic + refractory + relapse samples	Leukemia-related
	Non-CN AML	15 of 57	Fisher’s exact test	p = 0.4	8 of 24	Mann Whitney test	p = 0.2	8 of 24	Fisher’s exact test	p = 1.0
	Complex karyotype	10 of 67		p = 0.3	6 of 30		p = 0.01	6 of 30		p = 0.7
	WBC PB	68	Mann Whitney	p = 0.01	30	Pearson’s r	p = 0.2	30	Mann Whitney	p = 0.6
	PB blast %	68		p = 0.09	30		p = 0.2	30		p = 0.06
	BM blast %	68		p = 0.1	30		p = 0.06	30		p = 0.1
	CD34 expression on LEU	67		p = 0.3	30		p = 0.2	30		p = 0.09
	CD33 expression on LEU	65		p = 0.09	29		p = 0.3	29		p = 0.3
	Sample-related
	Total number of transplanted blasts	67	Mann Whitney	p = 0.02	30	Pearson’s r	p = 0.5	30	Mann Whitney	p = 0.9
	Fresh vs cryopreserved	13 vs 55	Fisher’s exact test	p = 0.5	7 vs 23	Mann Whitney	p = 0.2	7 vs 23	Fisher’s exact test	p = 0.2

The table shows association of AML characteristics with three engraftment outcomes—engraftment ability (categorical), engraftment intensity (continuous) and engraftment latency (categorical), for the group of 47 diagnostic samples (upper part) and the extended group of all 68 samples from patients at different disease stages (lower part). For the 68-sample group, only the effect of leukemia-related characteristics was examined

BM Bone marrow, CN AML AML with cytogenetic aberrations other than complex karyotype, LEU leukocytes, n count, N negative, N/A not analyzed, PB peripheral blood, WBC white blood count, Y positive

Meanwhile, multivariate analysis of all mutations combined with leukemia- and patient-related characteristics identified no set of characteristics that predicted engraftment ability (Table 2).

Table 2.

Results of multivariate analysis

	i. Analysis of diagnostic sample group		ii. Analysis of extended sample group
	Samples—diagnostic		Samples—diagnostic, refractory, relapse
	Predictors—gene mutations, leukemia-related and patient-related characteristics		Predictors—leukemia-related characteristics
Engraftment ability (categorical, engraftment Y/N)	Engrafted n = 21		Engrafted n = 26
	Non-engrafted n = 18		Non-engrafted n = 25
	Hotelling’s T-Squared test		Hotelling’s T-Squared test
	No predictors identified	–	no predictors identified	–
Engraftment intensity (continuous, percentage of hCD45+ cells)	Engrafted n = 21		Engrafted n = 26
	General linear model		General linear model
	Cytogenetic risk	p < 0.001	Complex karyotype	p = 0.008
	DNMT3A	p = 0.002	BM blast %	p = 0.006
	TET2	p = 0.001	CD33+ %	p = 0.04
	NRAS (negative predictor)	p = 0.004
	BM blast %	p = 0.01
Engraftment latency (categorical, engraftment at 12 or 16 weeks)	Engrafted n = 21		Engrafted n = 26
	Hotelling’s T-Squared test		Hotelling’s T-Squared test
	PB WBC + CD33+ %	p = 0.02	PB blast % + CD33+ %	p = 0.01
	PB WBC + FLT3-TKD	p < 0.05	PB blast % + CD33+ % + BM blast %	p = 0.02
	PB WBC + CEBPA double mutated	p < 0.05
	PB WBC + TET2	p < 0.05
	PB WBC + SRSF2	p < 0.05
	PB WBC + CD33+ % + Cytogenetic risk	p = 0.03

Association of AML sample characteristics (predictors) with three engraftment outcomes—engraftment ability (categorical), engraftment intensity (continuous), engraftment latency (categorical), was assessed. Hotelling’s T-Squared test was applied for categorical outcomes; we list all statistically significant combinations of two predictors plus the best fitting combination of three predictors. General linear model (GLM) was used to select the best-fit model for prediction of the continuous outcome, engraftment intensity; we list the prediction model with the highest AIC score and only statistically significant coefficients. The analyses were performed for the group of 47 diagnostic samples (on the left) and the extended group of all 68 samples from different disease stages (on the right). For the extended group, only the effect of leukemia-related characteristics was examined

BM Bone marrow, non-CN AML AML with cytogenetic aberrations other than complex karyotype, LEU leukocytes, n count, N negative, PB peripheral blood, WBC white blood count, Y positive

We found no statistically significant effect of sample-related characteristics (number of injected blasts and cryopreservation status) on engraftment ability of diagnostic samples.

Association between engraftment intensity and AML sample characteristics

None of the single mutations were associated with the engraftment intensity of 22 successfully engrafted AML samples (Table 1). Only the NPM1^mut/FLT3-ITD^neg samples had lower engraftment intensity than the other samples (median 2.7 vs 40.1% of hCD45⁺ cells, p = 0.02), which corresponds to the favorable prognostic value of this mutation combination in AML patients.

None of the leukemia-related characteristics was associated with higher engraftment intensity of diagnostic samples (Table 1), except for complex karyotype (median 97.3 vs 8.7% of hCD45⁺ cells, p = 0.01), wherein 3/3 (100%) samples having a complex karyotype produced > 95% of hCD45⁺ cells.

Among patient-related characteristics, poor cytogenetic risk group samples (including the complex karyotype samples) engrafted with higher intensity than intermediate risk group samples (median 96.5 vs 8.2% of hCD45⁺ cells, p = 0.006; no favorable risk patients were present).

A multivariate analysis of mutational status, leukemia- and patient-related characteristics identified five statistically significant predictors of higher engraftment intensity: (1) cytogenetic risk group, (2) mutated DNMT3A, (3) mutated TET2, (4) mutated NRAS (negative predictor), and (5) percentage of BM blasts (Table 2).

The engraftment intensity was not affected by cryopreservation or the number of injected blasts.

Association between engraftment latency and AML sample characteristics

Mutational status and leukemia-related characteristics did not affect engraftment latency in univariate analysis. Of the patient-related characteristics, failure to achieve clinical remission was associated with shorter engraftment latency, as reflected by 3/3 (100%) samples engrafting at 12 weeks (Table 1).

Multivariate statistical analysis yielded several combinations of AML characteristics that predicted engraftment latency (Table 2). When considering only two parameters, the strongest association included the percentage of PB blasts and CD33 expression on leukocytes, and for consideration of three parameters, cytogenetic risk group was added to these two.

Again, the sample-related characteristics were not associated with engraftment latency.

Of note, when comparing the two engraftment outcomes–intensity and latency, samples with faster engraftment (12 weeks) engrafted with higher intensity than the samples with slower engraftment (16 weeks), both in the diagnostic group (median 69.6 vs 7.8% of hCD45⁺ cells, p = 0.03) and the extended group (median 94.5 vs 10.2% of hCD45⁺ cells, p = 0.004).

Engraftment of AML samples from different disease stages

The leukemia-related characteristics were specific to disease stage rather than to patient information. Therefore, in the extended group of all 68 xenotransplanted samples, we only analyzed the association between the leukemia-related parameters and the three engraftment outcomes, as discussed below.

In these 68 samples, higher engraftment ability was associated with higher PB WBC (median 25.3 vs 9.9 × 10⁹/L in engrafting vs non-engrafting samples, p = 0.01; Table 1). In multivariate analysis, none of the leukemia-related characteristics predicted the engraftment ability (Table 2).

Univariate analysis showed an association between higher engraftment intensity and complex karyotype (median 95.1 vs 25.3% of hCD45⁺ cells, in complex- vs non-complex karyotype samples, p = 0.01). In multivariate analysis, engraftment intensity was predicted by three statistically significant characteristics: (1) presence of complex karyotype; (2) BM blast percentage; and (3) CD33 expression on leukocytes.

Engraftment latency was not associated with any of the leukemia-related characteristics in the extended group of 68 samples, as assessed by univariate analysis. In multivariate analysis, the most statistically significant effect with two parameters was seen for PB blast percentage and CD33 expression on leukocytes, with the addition of BM blast percentage for a combination of three variables.

No difference was found between the freshly isolated vs cryopreserved samples in their engraftment ability, intensity or latency in the extended sample group (Table 1). However, the number of injected blasts was shown to be lower for the non-engrafting samples vs the engrafting samples (median 1.0 × 10⁶ vs 1.4 × 10⁶ injected blasts, p = 0.02), whereas engraftment intensity and latency were not affected.

We also performed a direct comparison of engraftment of diagnostic vs refractory and diagnostic vs relapse samples (Table 1). The analysis of diagnostic vs refractory samples was limited by the low number of refractory samples. Although only 1/7 (14%) refractory samples engrafted, the difference in engraftment ability was not statistically significant and the other two engraftment outcomes could not be compared. The relapse samples engrafted with higher intensity than diagnostic samples (median 92.9 vs 14.0% of hCD45⁺, p = 0.03), whereas engraftment ability or latency was not significantly different. A comparison of engraftment of diagnostic samples obtained from patients who later relapsed vs engraftment of diagnostic samples of non-relapsing patients showed no difference in any of the engraftment outcomes between these two groups (Table 1). This result suggests that the higher leukemia penetrance seen for relapse samples is acquired only at the time of relapse, but not present at diagnosis.

Comparison of engraftment outcomes

Finally, we compared the three engraftment outcomes based on their relevance to patient prognosis, as defined by OS and EFS. We identified no difference in treatment outcomes for patients with engrafting vs non-engrafting diagnostic samples (median OS of 17.0 vs 31.0 months, p = 0.4; median EFS of 10.0 vs 10.5 months, p = 1.0; Fig. 3a, b). In contrast, analysis based on engraftment intensity showed that patients with higher than the median percentage of engrafted hCD45⁺ cells showed a significantly shorter OS and a trend towards shorter EFS (median OS of 9.0 months vs not reached, p = 0.04; median EFS 7.0 months vs not reached, p = 0.1; Fig. 3c, d). No difference in OS and EFS was observed between patients whose samples engrafted at 12 weeks vs those that engrafted at 16 weeks (engraftment latency; median OS of 10.0 vs 17.0 months, p = 0.2; median EFS of 8.0 vs 10.0 months, p = 0.3; Fig. 3e, f).

Fig. 3.

Survival of curatively treated patients stratified according to engraftment outcomes of their diagnostic samples that were xenotransplanted into NSG mice. a, b OS and EFS of patients with engrafting or non-engrafting samples; c, d OS and EFS for patients stratified according to the percentage of hCD45⁺ cells present in xenografts; e, f OS and EFS for patients stratified according to time required for engraftment. p values were calculated from Mantel-Cox test. n Number, w week

Discussion

In the present study, we analyzed the association between the engraftment of AML samples and their characteristics including mutational status. Whereas most previous studies used one evaluation criterion, we performed a “three-parameter” assessment of engraftment outcomes in terms of engraftment ability, intensity and latency. These outcomes were also compared to select the best strategy for engraftment assessment.

The investigation of mutational status showed that only interacting mutations, rather than single mutations, were associated with engraftment outcomes. This result is consistent with previous studies demonstrating that single NPM1 mutations were not associated with engraftment outcome (Kennedy et al. 2013; Malaise et al. 2011; Pearce et al. 2006). NPM1^mut/FLT-ITD^neg samples engrafted with lower intensity than other samples, although the engraftment ability was similar to that of other samples. In contrast to our results and that of several other groups, some studies reported higher engraftment ability or intensity of FLT3-ITD positive samples (Kennedy et al. 2013; Lumkul et al. 2002; Malaise et al. 2011; Pearce et al. 2006; Rombouts et al. 2000; Sanchez et al. 2009). In our study, the co-mutation NPM1^mut/DNMT3A^mut was associated with higher engraftment ability, which indicates the importance of mutational context in which the NPM1 appears. As mentioned above, NPM1 in the absence of a proliferative FLT3-ITD mutation was associated with lower engraftment intensity. On the other hand, co-occurrence of NPM1 with DNMT3A mutations improved engraftment ability in our study, which could suggest a lower dependence on extrinsic signals or microenvironment. DNMT3A occurs in AML as a pre-leukemic event with a landscaping role and requires an additional leukemia-initiating event, such as NPM1 mutation, to drive malignant transformation (Corces-Zimmerman et al. 2014; Shlush et al. 2014; Thol et al. 2017). However, mutated DNMT3A alone showed an adverse prognostic impact in two large AML patient cohorts (Metzeler et al. 2016; Papaemmanuil et al. 2016). To date, the effect of other mutations beyond NPM1 and FLT3-ITD that could have potential for prognostic stratification has not been investigated in terms of engraftment (Dohner et al. 2017). Of these mutations, the biallelic CEBPA mutation and ASXL1 mutation showed no role in engraftment in our study, and TP53 and RUNX1 were mutated in very few samples, which limited our analysis of their effects.

In most previous studies, higher cytogenetic risk was associated with better engraftment outcomes, although this effect was dominantly due to poor engraftment of favorable risk samples (Kennedy et al. 2013; Lumkul et al. 2002; Pearce et al. 2006). Our cohort included no favorable risk patients, but we did observe increased engraftment intensity of diagnostic samples in a comparison of poor and intermediate risk patients that was driven by the high engraftment intensity of samples of poor-risk patients with complex karyotypes. A higher PB WBC was previously shown to correlate with better engraftment outcomes, and to date the influence of blast percentage has been evaluated in only one study that showed no effect of BM blast load on engraftment (Kennedy et al. 2013; Rombouts et al. 2000). In our cohort, PB WBC was associated with higher engraftment ability in univariate analysis, for the extended sample group. The percentage of blasts in PB or BM was not associated with any of the engraftment outcomes in univariate analysis, although both parameters scored as predictors in multivariate analysis. Previous studies investigated CD34 expression in relation to both leukemogenic potential of different cell populations and patient prognosis (Quek et al. 2016; Zeijlemaker et al. 2015). In our cohort, CD34 expression on BM leukocytes ranged between 0 and 90% and was not related to any engraftment outcome. Moreover, three samples with < 1% CD34⁺ leukocytes produced intense engraftment with > 80% hCD45⁺ cells, indicating that leukemia initiating capacity is not restricted to cells having a specific phenotype or state of differentiation arrest, which is consistent with earlier fiindings (Quek et al. 2016; Taussig et al. 2010). The CD33 expression has been previously reported to differ between various AML subgroups, with high expression found in AML patients with NPM1, FLT3-ITD and CEBPA mutations and low expression found in patients with complex karyotypes or core-binding factor AMLs (Ehninger et al. 2014; Khan et al. 2017). In our study, CD33 expression was not associated with engraftment as an independent factor in univariate analysis, but it scored as one of the predictors for engraftment latency and also intensity in multivariate analysis.

Besides the AML-characteristics, we have also analyzed the influence of sample-related parameters on engraftment. We did not see any difference between engraftment outcomes of cryopreserved or fresh samples. However, lower number of injected blasts was associated with lower engraftment ability, although this was only seen in the extended group of 68 samples. To evaluate if this parameter might have biased our analysis, we performed additional analysis for the engraftment ability that included only a group of samples above a specific threshold of injected blasts. This threshold was set to 0.6 × 10⁶ injected blasts, as none of the samples with lower blast number provided engraftment (Online Resource 1). However, the results of the additional analysis did not differ from the original analysis on all samples (results not shown).

OS and EFS represent the main outcome parameters in clinical studies that characterize the prognosis of different AML patient groups or the success of treatment approaches. Here we used OS and EFS to compare engraftment outcomes with AML patient prognosis. Only engraftment intensity, and not engraftment ability or latency, was associated with patient survival and thus best reflected the actual aggressiveness of AML in patients. In contrast, some studies showed that engraftment ability was also associated with OS and/or EFS in patients (Kennedy et al. 2013; Pearce et al.2006). We did not observe such an association, which could be due to the use of different mouse strains and/or different intervals for engraftment analysis (6–10 weeks compared to 12–16 weeks post-transplant in our analyses). Experiments using humanized mouse models demonstrated that the engraftment ability of AML is linked with microenvironment (in)dependency, as favorable-risk AMLs showed no/poor engraftment in the standard mouse model, but readily engrafted in mice carrying humanized scaffolds or expressing human cytokines (Antonelli et al.2016; Ellegast et al.2016). In contrast, another study showed that favorable risk AMLs engrafted even in standard mice, but only when the engraftment period was extended to up to 1 year (Paczulla et al.2017). In our study the samples that were analyzed only at 16 weeks would probably be falsely classified as non-engrafters if the engraftment period was only 12 weeks. Together, these findings indicate that determination of engraftment ability is highly dependent on the time of analysis. We further observed that samples with shorter engraftment latency (12 weeks) also engrafted with higher intensity than the slow engrafters (16 weeks). This clearly demonstrates that faster engraftment is also connected with a higher proliferation rate. Based on the association between engraftment latency and intensity, as well as the dependency of engraftment ability on the engraftment period, we propose that engraftment could be more accurately described by only a single outcome, defined as “engraftment rate”, which can be expressed as the percentage of hCD45⁺ cells determined at one predefined time point for all samples. The samples could be than categorized into: (1) non/slow-engrafting (close to 0% hCD45⁺), defining samples with long engraftment latency requiring prolonged analysis or humanized conditions to detect robust engraftment, usually representing AML from the favorable cytogenetic risk group; (2) moderately-engrafting (~ 5–90% hCD45⁺ cells), defining samples with good engraftment; and (3) rapidly-engrafting (reaching ~ 100% hCD45⁺ cells or inducing terminal disease progression before the given analysis time point), defining the most aggressive samples, usually representing AMLs with poor prognosis.

In summary, our study provides a comprehensive investigation of the effect of a complex set of AML characteristics, including the effect of recurrent mutations on engraftment of AML samples defined by three separate outcomes. We showed that the analysis of individual mutated genes was insufficient to predict engraftment outcomes of AML samples in our cohort, and an association with engraftment was only found when considering interactions of mutated genes. We also showed that the survival of AML patients was best reflected by engraftment intensity (i.e., percentage of hCD45⁺ cells generated in the xenograft). Furthermore, for future studies, we propose use of a new engraftment outcome, the engraftment rate, which defines: (1) the ability of a sample to produce engraftment with (2) a certain percentage of hCD45⁺ cells within (3) a specific time. This combines the three engraftment outcomes analyzed in our study.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Funding

This study was supported by funds from the Faculty of Medicine MU to junior researcher Martin Culen. Supported by Ministry of Health of the Czech Republic, Grant No. 15-25809A. All rights reserved. Supported by MUNI/A/0968/2017.

Compliance with ethical standards

Conflict of interest

All authors declare that they have no conflicts of interest.

Ethical approval

Samples were collected from AML patients treated at the University Hospital, Brno. Informed consent was obtained from all individual participants included in the study. The study was approved by the Institutional Review Board of the University Hospital, Brno. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. All animal experiments were approved by institutional review board of the Masaryk University, Brno, and performed in accordance with all European guidelines for the protection of laboratory animals.