A truncation mutant of Csf3r cooperates with PML-RARα to induce acute myeloid leukemia in mice

Abstract

Severe congenital neutropenia (SCN) is associated with a marked propensity to develop myelodysplasia (MDS) or acute myeloid leukemia (AML). Truncation mutations of CSF3R, encoding the G-CSF receptor (G-CSFR), are associated with the development of MDS/AML in SCN. However, a causal relationship between CSF3R mutations and leukemic transformation has not been established. Herein, we show that truncated G-CSFR cooperates with the PML-RARα oncogene to induce AML in mice. Expression of truncated G-CSFR significantly shortens the latency of AML in a G-CSF dependent fashion, and it is associated with a distinct AML presentation characterized by higher blast counts and more severe myelosuppression. Basal and G-CSF-induced STAT3, STAT5, and Erk1/2 phosphorylation were highly variable but similar in leukemic blasts expressing wild type and truncated G-CSFR. These data provide new evidence suggesting a causative role for CSF3R mutations in human AML.

INTRODUCTION

A shared feature of many congenital and acquired bone marrow failure syndromes is a marked propensity to develop acute myeloid leukemia (AML) or a myelodysplastic syndrome (MDS). The cumulative risk of developing AML or MDS in patients with chronic bone marrow failure ranges from 2% to greater than 20%. With a few exceptions, the molecular basis for the increased risk of leukemic transformation is unknown. Severe congenital neutropenia (SCN) is a congenital bone marrow failure syndrome characterized by severe neutropenia present from birth and an arrest of myeloid differentiation at the promyelocyte/myelocyte stage. The cumulative risk of MDS/AML in patients with SCN after 10 years is 11–21%.^1,2 Treatment with granulocyte colony-stimulating factor (G-CSF) is effective in increasing neutrophil counts and preventing infections³, and most patients now are treated lifelong with G-CSF. However, the quantity and duration of G-CSF treatment are correlated with an increased risk of MDS/AML.^1,2 This has sparked considerable debate on whether G-CSF treatment contributes to leukemic transformation in patients with SCN or other conditions requiring prolonged G-CSF treatment.

Nonsense mutations of CSF3R that truncate the distal cytoplasmic portion of the G-CSF receptor (G-CSFR) are present in approximately 40% of patients with SCN.⁴ These mutations are acquired, heterozygous, and strongly associated with the development of AML/MDS.⁴ Expression of mutant G-CSFR in myeloid cell lines results in enhanced proliferative and survival signals.⁵ Transgenic mice carrying truncating mutations of Csf3r display a hyperproliferative response to G-CSF.^6,7 Moreover, a recent study showed that expression of mutant Csf3r in mice confers a strong clonal advantage at the hematopoietic stem cell (HSC) level that is dependent upon exogenous G-CSF.⁸ However, Csf3r mutant mice do not develop MDS or AML; thus, a causal relationship between CSF3R mutations and leukemia has not been established.

The hyperproliferative response of the truncated G-CSFR is mediated, in least in part, by accentuated STAT5 activation. Compared with wild type G-CSFR, the amplitude and duration of STAT5 activation is increased in primary murine hematopoietic cells expressing truncated G-CSFR.^8–10 In contrast, STAT3 activation by the truncated G-CSF is modestly reduced.^8–10 STAT5 activation is a common feature of leukemias associated with activating mutations of tyrosine kinase genes, including BCR-ABL,¹¹ JAK2,¹² and FLT3.^13,14 STAT5 has been implicated in hematopoietic stem self-renewal.¹⁵ Consistent with this observation, we recently showed that the clonal advantage of hematopoietic stem cells expressing truncated G-CSFR is dependent on STAT5 activation.⁸ The importance of STAT5 activation in leukemic progression is further supported by studies showing that expression of constitutively active STAT5 results in a fatal myeloproliferative disease in mice.^16–18

Activating mutations of receptor tyrosine kinases (e.g., FLT3 and KIT), while present in approximately 30% of de novo AML, are rare in AML arising in the setting of SCN.¹⁹ Since receptor tyrosine kinases and the G-CSFR activate similar signaling pathways, including STAT5, we hypothesized that G-CSFR truncation mutations may contribute to leukemic transformation by providing an ‘activated receptor tyrosine kinase signal’. SCN-associated AML is characterized by monosomy 7 and deletions of chromosome 5.^19,20 Chromosomal translocations or point mutations in transcription factors are uncommon.^19,20 Thus, in contrast to de novo AML where the initiating mutation (e.g., PML-RARα) is known in some cases, the initiating mutations in SCN-AML are largely unknown. Consequently, to test this hypothesis, we asked whether expression of truncated G-CSFR is able to cooperate with the PML-RARα oncogene, a well characterized initiating mutation for human promyelocytic leukemia, to induce AML in mice. This experimental approach has been used to demonstrate the transforming potential of activating mutations of FLT3 and NRAS, which by themselves, are not sufficient to induce AML in mice.^21,22 We show that expression of truncated G-CSFR shortens the latency and alters the phenotype of AML in PML-RARα transgenic mice. Basal and G-CSF induced STAT5, STAT3, and Erk1/2 activation in leukemic blasts is highly variable but not affected by mutant G-CSFR expression. These data support a causative role for CSF3R mutations in leukemic progression in patients with SCN.

METHODS

Mice

d715 Csf3r mice backcrossed 10 generations onto a C57BL/6 background were generated as described previously.⁷ Transgenic mice in which the PML-RARα oncogene was knocked-in to the Cathepsin G (Ctsg) locus were generated as described previously;²³ these mice also were backcrossed at least 10 generation onto a C57BL/6 background. Mice were housed in a specific pathogen-free environment. All experiments were approved by the Washington University Animal Studies Committee.

G-CSF and pegylated-G-CSF administration

Human G-CSF was either given as a single subcutaneous dose (250 μg/kg) or twice daily for 7 days at 125 μg/kg. Pegylated G-CSF (1 mg/kg) was given every 4–5 days for 6 months. G-CSF and pegylated G-CSF were generous gifts from Amgen (Thousand Oaks, CA).

Bone marrow transplantation

A total of 5 million leukocytes harvested from the spleens of leukemic mice were injected into the tail vein of lethally irradiated wild type mice. Recipient mice were conditioned with 300 cGy from a ¹³⁷Cesium source at a rate of approximately 95 cGy/minute 24 hours prior to transplantation. Prophylactic antibiotics (Trimethoprim-Sulfamethoxazole, Alpharma, Baltimore, MD) were given during the initial two weeks after transplantation.

Assessment of STAT5, STAT3, and Erk1/2 phosphorylation

Spleen cells from leukemic mice were incubated with fluorescein isothiocyanate (FITC)-conjugated Gr-1 and allophycocyanin (APC)-conjugated c-Kit; as a control, bone marrow cells from healthy wild type mice were incubated with a cocktail of FITC-conjugated lineage markers (Gr-1, B220, Ter119, and CD3) and APC-conjugated c-Kit. In each case, cells were then incubated in BD Phosphflow Lyse/fix buffer (BD Biosciences) for 10 minutes and permeabilized in 100% methanol at 4° for 20 minutes cells. Cells were then incubated with phycoerythrin (PE)-conjugated anti-pSTAT5 (pY694), anti-pSTAT3 (pY705), or anti-pErk1/2 (pT202/pY204) at room temperature for 30 minutes. In some cases, cells from leukemic spleens were suspended in Opti-MEM (Invitrogen, Carlsbad, CA) with 10% fetal bovine serum and stimulated with G-CSF (100 ng/ml) for 30 minutes at 37 °C and then processed for STAT5, STAT3, and Erk1/2 phosphorylation. Cells were analyzed on a 2-laser, 5 color FACScan flow cytometer. All antibodies were purchased from BD Biosciences, unless otherwise indicated.

Statistical Analysis

Statistical significance was determined either by a two-way ANOVA analysis or by a two-sided Student t test. Survival differences were assessed using the Log-rank (Mantel-Cox) test. All data represent the mean ± SEM, unless otherwise stated.

RESULTS

The d715 Csf3r can cooperate with PML-RARa to induce AML in mice in a G-CSF dependent fashion

Previous studies showed that activating mutations of FLT3 or KRAS were able to cooperate with PML-RARα to induce leukemia in mice.^21,22 As a proof-of-principle experiment, we asked whether Csf3r truncation mutations also could cooperate with PML-RARα to induce leukemia. We previously described transgenic mice carrying a targeted mutation of their Csf3r gene (termed d715) that reproduces the CSF3R mutation found in a patient with SCN.⁷ These mice (Csf3r^d715/d715) were crossed with transgenic mice expressing PML-RARα²³ and a tumor watch established (Table 1); all mice were backcrossed at least 10 generations onto a C57Bl/6 background. To simulate the high systemic level of G-CSF present in patients with SCN (either through increased endogenous production or exogenous administration), some of the mice were treated with G-CSF (pegylated human G-CSF, 1 mg/kg every 4–5 days) for 6 months starting at 6–10 weeks of age. Of note, the neutrophil response to chronic G-CSF treatment was similar in all cohorts (data not shown). All mice were followed for at least 18 months.

Table 1.

Tumor watch

PML-RARα	Csf3R	G-CSF Rx*	N
No	WT	No	19
No	d715	No	45
No	WT	Yes	12
No	d715	Yes	47
Yes	WT	No	57
Yes	WT/d715	No	46
Yes	d715	No	67
Yes	WT	Yes	54
Yes	WT/d715	Yes	42
Yes	d715	Yes	50

WT/d175 refers to Csf3r ^WT/d715 mice

d715 refers to Csf3rd715^/d715 mice

^*Pegylated human G-CSF (1 mg/kg) was given subcutaneously every 4–5 days for 6 months

As expected, none of the mice not containing the PML-RARα transgene developed AML (data not shown), confirming a previous study showing that expression of the d715 G-CSFR is not sufficient to induce AML in mice.⁷ The cumulative incidence of AML in PML-RARα mice (with wild type G-CSFR) was approximately 40%, with a median survival of 541 days (Figure 1A). In the absence of G-CSF treatment, the presence of heterozygous or homozygous Csf3r d715 mutation had no significant impact on leukemia penetrance or latency (Figure 1A). However, in mice treated with G-CSF, the presence of the d715 G-CSFR significantly shortened the latency to develop AML (Figure 1B and Table 2). Moreover, a non-significant trend to increase AML penetrance was observed (Table 2). These data demonstrate that the d715 G-CSFR is able to cooperate with PML-RARα to induce AML in mice in a G-CSF dependent fashion.

Figure 1. The d715 Csf3r cooperates with PML-RARα to induce AML in mice in a G-CSF dependent fashion.

(A). Kaplan Meier survival curves are shown for mice expressing PML-RARα (PR) and wild type Csf3r, heterozygous d715 Csf3r, or homozygous d715 Csf3r. A separate cohort of mice (B) was treated with pegylated human G-CSF (1 mg/kg every 4–5 days) for 6 months. There were 42–67 mice in each cohort. Significance was assessed using the Mantel-Cox test.

Table 2.

AML parameters

Parameter	PR Csf3rWT/WT	PR Csf3rd715/WT	PR Csf3rd715/d715	^ P value
Median Survival	437 days	406 days	317 days	0.013
* Cumulative incidence of AML	61.4%	65.1%	77.1%	0.089
# Blood
WBC(× 10−3/μl)	81.6 ± 59.2	88.7 ± 37.8	62.8 ± 45.9	0.808
RBC (× 10−6/μl)	6.7 ±1.3	4.8 ± 0.9	1.6 ± 0.7	0.013
Platelets (× 10−3/μI)	527± 133	535 ± 115	188 ± 62	0.052
% Blasts	5.16 ± 1.2	7.93 ± 2.9	22.6 ± 9.8%	0.195
# Bone marrow
Cells per femure (× 10−6)	18.2 ± 4.6	24.1 ± 4.1	11.9 ± 3.1	0.276
% Blasts	20.7 ± 5.3	26.1 ± 7.9	38.2 ± 16.7	0.381
# Spleen
Weight (grams)	1.32 ± 0.33	1.20 ± 0.17	0.94 ± 0.12	0.235
% Blasts	25.4 ± 4.7	20.2 ± 8.3	34.3 ± 7.3	0.37

All mice had a history of G-CSF treatment

^{^}P value (t-test) comparing PR Csf3r^WT/WT with PR Csf3rd^715/d715

^*Cumulative incidence of AML @ 560 days

^#Data based on the analysis of morbidly ill mice: n = 3–6 PR Csf3r^WT/WT, 4–8 PR Csf3r^dWT/d715, and 4–9 PR Csf3r^d715/d715 mice.

Expression of the d715 G-CSFR is associated with higher blast counts and more severe myelosuppression

We next compared the phenotype of AML arising in the setting of wild type or d715 Csf3r. We focused on mice with a history of G-CSF treatment, since changes in AML latency and penetrance were confined to these experimental cohorts. A similar degree of leukocytosis and splenomegaly were observed in all cohorts (Table 2). However, PML-RARα Csf3r^d715/d715 mice had more severe anemia and thrombocytopenia at presentation with a trend to decreased bone marrow cellularity. Moreover, a trend to an increase in the percentage of myeloblasts (defined as Gr-1+ Kit+ double positive cells)²³ in the bone marrow, blood and spleen was observed (Figure 2A and Table 2). Spleen cells from leukemic PML-RARα Csf3r^wt/wt and PML-RARα Csf3r^d715/d715 mice were able to induce a fatal leukemia upon transplantation into sub-lethally irradiated syngenic hosts with a similar latency (Figure 2B). Collectively, these data show that expression of the d715 G-CSFR is associated with a distinct AML presentation characterized by higher blast counts and more severe myelosuppression.

Figure 2. Phenotypic characterization of leukemic mice.

(A) Peripheral blood cells from leukemic mice of the indicated genotype were analyzed by flow cytometry for Gr-1 and c-Kit expression. Leukemic blasts were identified as Gr-1+ Kit+ cells. Data are representative of at least 3 independent experiments. (B) Spleen cells (1 × 10⁶) from leukemic mice of the indicated genotype were transplanted into syngenic irradiated mice. Kaplan Meier survival curves are shown.

STAT3, STAT5, and Erk1/2 are not constitutively activated in leukemic blasts expressing the d715 Csf3r

Constitutive activation of STAT3, STAT5, and Erk1/2 has been detected in a subset of AML.^18,24–26 Moreover, sustained activation of these signaling molecules has been reported in myeloid cells expressing the d715 Csf3r.^7,9,27 Based on these observations, we hypothesized that excessive activation of one or more of these proteins by the d715 G-CSFR may have contributed to leukemic transformation. To test this hypothesis, we first isolated splenic cells from leukemic PML-RARα Csf3r^wt/wt and PML-RARα Csf3r^d715/d715 mice (no mice were treated with G-CSF at the time these analyses were performed) and measured the basal level of phosphorylated STAT3, STAT5, and Erk1/2 in blasts (Gr-1+ Kit+) by flow cytometry (Figure 3). As a control, we also analyzed lineage- Kit+ cells from non-leukemic wild type mice. Considerable variability in the basal level of STAT3, STAT5, or Erk1/2 phosphorylation was observed. In PML-RARα Csf3r^wt/wt blasts, constitutive activation of STAT3, STAT5, and Erk1/2 (as defined by a signal greater then the 95% confidence interval for control cells) was observed in 23.4%, 11.8%, and 26.7% of cases, respectively. Similar data were observed with PML-RARα Csf3r^d715/d715 blasts; constitutive activation of STAT3, STAT5, and Erk1/2 was observed in 22.7%, 14.8%, and 25.0% of samples, respectively.

Figure 3. STAT3, STAT5, and Erk1/2 phosphorylation in leukemic cells.

Spleen cells from leukemic mice of the indicated genotype were harvested and immediately analyzed for phosphorylated STAT3 (A), STAT5 (B), or Erk1/2 (C). Representative histograms showing staining for the indicated phosphorylated protein after gating on Gr-1+ Kit+ leukemic blasts (left 3 panels). As a control, bone marrow cells from healthy wild type mice were analyzed in the same fashion, gating on Lineage- Kit+ cells (Lin- Kit+). The right panel shows the median fluorescent intensity (MFI) after subtraction of the signal from the isotype control. Each data point represents a single mouse, and the mean ± SEM are shown. CTRL: lineage- Kit+ wildtype cells; WT: PR Csf3r^wt/wt blasts; d715: PR Csf3r^d715/d715 blasts.

We next stimulated leukemic cells with G-CSF in vitro and assessed STAT3, STAT5, and Erk1/2 activation using the phospho-flow assay. G-CSF stimulation induced STAT3 and STAT5 phosphorylation to a similar degree in PML-RARα Csf3r^wt/wt and PML-RARα Csf3r^d715/d715 leukemic blasts (Figure 4). G-CSF did not consistently induce Erk1/2 phosphorylation in either cell type. Together, these data suggest that, at least in fully transformed leukemic blasts, basal and G-CSF induced activation of STAT3, STAT5, and Erk1/2 is similar in cells expressing wild type or d715 G-CSFR.

Figure 4. G-CSF induced STAT3, STAT5, and Erk1/2 activation in leukemic cells.

Splenic cells from leukemic mice of the indicated genotype were stimulated in vitro with G-CSF (100 ng/ml) for 15 minutes, and phosphorylation of Stat 3 (A), STAT5 (B), or Erk1/2 (C) was measured. Each point represents data from an individual mouse. Data are gated on Gr-1⁺ Kit⁺ leukemic blasts. The difference in signal between untreated and G-CSF treated samples (ΔMFI) is shown in the right panel. The mean ± SEM is shown. WT: PR Csf3r^wt/wt blasts; d715: PR Csf3r^d715/d715 blasts.

DISCUSSION

The contribution of CSF3R truncation mutations to transformation to MDS/AML in patients with SCN is controversial. CSF3R mutations are strongly associated with the development of MDS/AML. Germeshausen et al. reported that the incidence of CSF3R truncation mutations was 78% (18/23) in individuals with SCN and monosomy 7, MDS, or AML, compared with 34% (43/125) in patients without MDS or AML.⁴ In mice, expression of truncated G-CSFR confers a strong clonal advantage to HSCs that is dependent on G-CSF.⁸ This is consistent with the observation, that in most (but not all) patients with SCN, the frequency of circulating cells containing mutant CSF3R increases over time.^4,28 Moreover, a recent study showed that CSF3R mutations are present in all hematopoietic lineages (including lymphocytes), suggesting that the clonal expansion in patients with SCN occurs at the HSC level.²⁹ On the other hand, CSF3R mutations are not sufficient to induce MDS/AML in humans or mice. The time course from detection of CSF3R mutations to leukemic transformation is quite variable and some patients with CSF3R mutations remain transformation free for years.^4,28 Moreover, transgenic mice expressing mutant Csf3r do not develop leukemia, despite prolonged treatment with G-CSF.^7,9

Activating mutations of receptor tyrosine kinase genes (FLT3 and KIT) are present in approximately 30% of cases of de novo AML but are rare in SCN-associated AML.¹⁹ Similar to mutant Csf3r, these mutations are not sufficient to induce AML in mice,^22,30 demonstrating that they are not leukemia-initiating mutations. However, these mutations are able to cooperate with initiating mutations, such as PML-RARα, to induce AML. There is strong evidence that CSF3R truncation mutations also are activating. The truncated G-CSFR transmits a hyperproliferative signal that is secondary, in part, to sustained STAT5 activation that, in turn, is related to impaired receptor internalization and the failure of the truncated G-CSFR to recruit and activate SOCS3 and SHP2.^31,32 Since the G-CSFR and receptor tyrosine kinases share many signaling pathways, these observations suggest that the mutant G-CSFR may provide the “activated tyrosine kinase signal” that is present in many cases of de novo AML. A previous study showed that activating mutations of FLT3 were able to cooperate with PML-RARα to induce leukemia in mice.²² Likewise, in the present study, we show that expression of truncated G-CSFR cooperates with PML-RARα to induce AML in mice. Though to the best of our knowledge the t(15;17) translocation producing PML-RARα has not been reported in patients with SCN-associated MDS/AML, we suggest this “proof-of-principle” experiment provides strong evidence that CSF3R truncation mutations are leukemogenic. Interestingly, this cooperativity was dependent upon G-CSF administration. This is relevant to patients with SCN, since G-CSF levels are chronically elevated, either secondary to exogenous administration or through increased endogenous production (due to the severe neutropenia). In fact, the G-CSF dependence of the truncated G-CSFR provides an explanation for why CSF3R mutations are common in MDS/AML arising in the setting of SCN but are rare in de novo AML.

We observed considerable heterogeneity in baseline and G-CSF-induced STAT3, STAT5, and Erk1/2 phosphorylation in leukemic blasts. Constitutive STAT3 and STAT5 activation were observed in approximately 23% and 13% of cases, respectively, irrespective of G-CSFR mutation status. Though highly variable, similar data have been reported in primary human AML blasts. Constitutive activation of STAT3 has been observed in 20–44% of cases of AML and STAT5 activation in approximately 20% of cases.^{24,26,33–35} Relevant to this study, STAT3 activation has been observed in the majority (7 of 9 cases, 78%) of cases of M3-AML.^34,35 However, Spiekerman et al showed that expression of PML-RARα was not sufficient to induce STAT3 or STAT5 activation.³³ Contrary to expectations, enhanced STAT5 activation after G-CSF stimulation was not observed in leukemic blasts carrying mutant Csf3r, likely reflecting altered G-CSFR expression or signal transduction. In this regard, it is interesting to note that some cases of AML with activating mutations of FLT3 also do not display constitutive STAT3 or STAT5 activation.³³ It is important to note that our analysis was confined to fully transformed leukemic blasts. It is possible that enhanced STAT5 and/or STAT3 activation contributed to leukemic transformation in preleukemic progenitors. It also is possible that mutant G-CSFR contributes to leukemic transformation in a STAT3 and STAT5 independent fashion. For example, similar to activating mutations of FLT3,²² the mutant G-CSFR is associated with increased reactive oxygen species (ROS) production,³⁶ which has been implicated in oxidative DNA damage.³⁷

In summary, our data show that Csf3r truncation mutations, similar to other mutations that activate receptor tyrosine kinase signaling pathways, cooperate with the PML-RARα oncogene to induce AML in mice. These data provide strong new evidence that truncating mutations of the CSF3R are leukemogenic.

Nonstandard abbreviations

G-CSF

granulocyte colony-stimulating factor

G-CSFR

G-CSF receptor

HSC

hematopoietic stem cells

HSPC

hematopoietic stem/progenitor cells

AML

acute myeloid leukemia

MDS

myelodysplastic syndrome

LSK cells

lineage- Sca+ Kit+ cells