CXCR4 Expression, CXCR4 Activation, and Wild Type NPM1 Are Independently Associated with Unfavorable Prognosis in Patients with Acute Myeloid Leukemia

Abstract

BACKGROUND

CXC chemokine receptor 4 (CXCR4) is activated by phosphorylation and is essential for migration of hematopoietic precursors to bone marrow. CXCR4 overexpression predicts unfavorable prognosis in patients with acute myeloid leukemia (AML). Nucleophosmin (NPM1) mutation is the most frequent genetic abnormality in AML patients and predicts a favorable prognosis. In vitro studies have suggested that mutant NPM decreases CXCR4-mediated chemotaxis by downregulating CXCR4, thereby linking the NPM and CXCR4 pathways.

PATIENTS AND METHODS

In a group of 117 untreated adults with AML we used immunohistochemistry to assess bone marrow specimens for CXCR4 and phosphorylated (p) CXCR4 (pCXCR4) expression. All cases were also analyzed for NPM1 mutations using PCR-based methods.

RESULTS

CXCR4 expression was detected in 75 (64%) and pCXCR4 expression was detected in 31 (26%) patients. NPM1 mutations were detected in 63 (54%) patients. NPM1 mutations did not correlate with CXCR4 (p = 0.212) or pCXCR4 (p = 0.355) expression. The median 5-year overall survival was 27% (95% CI: 19-36%), with a median follow-up of 8 months (95%CI: 6-15). In a multivariate Cox proportional hazards model, reduced overall and progression-free survival rates were associated with a history of antecedent hematological disorder, failure to achieve complete remission, thrombocytopenia, unfavorable cytogenetics, CXCR4 expression, and wild type NPM1. pCXCR4 expression was independently associated with shorter progression-free survival.

CONCLUSION

There is no correlation between NPM1 mutations and CXCR4 or phosphorylated CXCR4 expression suggesting that the CXCR4 and NPM pathways act independently in adult AML.

INTRODUCTION

Acute myeloid leukemia (AML) is estimated to affect 14,590 individuals in the United States in 2013.¹ Despite improvements in therapy over the past decade, 71% of these patients are expected to die of disease.¹ Therefore, biomarkers that predict and refine prognosis for AML patients are needed to improve risk-adapted therapy.

CXCR4 is a 352 amino acid rhodopsin-like G protein coupled receptor with no intrinsic kinase activity that selectively binds CXC chemokine stromal cell-derived factor 1 alpha (SDF-1α), also known as CXCL12.^{2, 3} The CXCR4/ SDF-1α interaction is essential for hematopoiesis as well as for general development, organogenesis, and vascularization.^4-9 CXCR4 expression is tightly controlled at several levels. At the transcriptional level, a number of signaling molecules can either increase or attenuate CXCR4 expression.^10-17 At the post-translational level, phosphorylation is a key mechanism of CXCR4 regulation. SDF-1α -promoted tyrosine phosphorylation activates the JAK/ STAT pathway,^{18, 19} whereas cytokine-induced tyrosine phosphorylation promotes ligand-independent internalization of CXCR4.¹³ CXCR4 is phosphorylated in response to ligand binding in a G-protein-coupled receptor kinase 2–dependent fashion.^20-22 Receptor phosphorylation stimulates the interaction of β-arrestin with the carboxy terminus.^{21, 23} This interaction terminates CXCR4-mediated activation of Gαi but promotes dynamin-dependent, clathrin-mediated receptor endocytosis and enhances CXCR4-RAF-dependent signaling.^{20, 23, 24} Phosphorylation of CXCR4 also can occur in response to the activation of other receptors and involves additional kinases, such as protein kinase C^20-22 or tyrosine kinases.^{13, 25}

CXCR4 expression has been detected in 23 cancers of various origins²⁶ and is the most common chemokine receptor expressed by cancer cells.²⁷ In the neoplastic setting, a number of other factors also can enhance CXCR4 expression, such as vascular endothelial growth factor,²⁸ activation of nuclear factor kappa B,²⁹ PAX3-FKHR,^{30, 31} and RET/PTC.³² Ubiquitination of CXCR4 is another mechanism of posttranslational regulation of CXCR4 expression.^{33, 34}

Nucleophosmin, encoded by the nucleophosmin (NPM1) gene on chromosome 5q35, is ubiquitously expressed in human tissues.^{35, 36} NPM1 mutations occur in 25-35% of all AML patients and 45-64% of AML cases with a normal karyotype.^{37, 38} In AML patients, NPM1 mutations are associated with a favorable prognosis in the absence of fms-related tyrosine kinase 3 gene (FLT3) internal tandem duplications (ITD).³⁹ In two retrospective studies of patients with FLT3 wild type AML, patients with NPM1-mutation did not benefit from allogeneic stem cell transplantation, in contrast with patients with wild type NPM1.^{39, 40} These data have led others to propose that AML patients with a normal karyotype should be tested routinely for FLT3-ITD and NPM1 mutations thereby facilitating risk-adapted therapy.

In 2007, Zhang and colleagues published data suggesting that the CXCR4-CXCL12 axis and NPM pathway are linked.⁴¹ These authors showed that CXCR4 forms complexes with NPM, and that overexpression of mutant NPM decreased CXCR4-mediated chemotaxis by downregulating CXCR4 expression.⁴¹ If relevant in vivo, NPM1 mutation theoretically could result in less protection of leukemic cells by the bone marrow microenvironment suggesting that the favorable prognosis of NPM1 mutation in AML patients could be related to negative regulation of CXCR4 expression and function by mutant NPM protein.

Recently, Chou and colleagues introduced a knock-in mouse model of NPM1 mutation inserting TCTG after nucleotide c.857. ⁴² and demonstrated that the expression of CXCR4/CXCL12-related genes was significantly suppressed in mutant myeloid precursors compared to myeloid precursors with wild type NPM1; similarly, suppression of CXCL12 and CXCR4 pathway signatures was detected by genome-wide expression microarray analysis on BM mononuclear cells in NPM1-mutated AML patients compared to NPM1-wild AML. ⁴²

We and others have shown that CXCR4 expression correlates with inferior prognosis in adult AML patients;^43-45 however, there are little data regarding pCXCR4 expression in AML. In addition, there are no studies analyzing the potential interaction of the CXCR4 and pCXCR4 pathways, and the NPM pathway in AML patients. Therefore, the goals of this study were to assess for CXCR4 and pCXCR4 expression as well as NPM1 mutations and correlate with prognosis in adult patients with AML.

PATIENTS AND METHODS

Patient Identification

After approval of the protocol by the Institutional Review Board, the files of the Department of Hematopathology from January 1, 2003, to July 31, 2008 were searched for cases of AML with available banked bone marrow aspirate material and fixed, paraffin-embeded bone marrow core biopsy specimens. Patients who did not receive treatment at our institution were excluded.. The diagnosis of AML was established based on the results of bone marrow (BM) aspiration and biopsy according to the current World Health Organization criteria. None of the patients received prior therapy. Available Wright-Giemsa stained BM aspirate smears and hematoxylin-eosin stained BM aspirate clot and trephine biopsy specimens were reviewed. Patient clinical information was obtained by review of medical records.

Morphologic, Cytochemical, and Immunophenotypic Analysis of Blasts

Bone marrow aspirate smears were assessed via cytochemical analysis for myeloperoxidase and alpha-naphthyl butyrate esterase using methods reported previously.⁴⁶ Flow cytometry immunophenotypic analysis was performed on BM aspirate specimens using a four-color FACScalibur cytometer (Becton Dickinson) and analyzed using the CellQuest software package (Becton Dickinson) as has been described.⁴⁷ Antibodies specific for the following antigens were used: CD3, CD7, CD10, CD13, CD19, CD20, CD33, CD34, CD45, CD56, CD64, CD117; HLA-DR, myeloperoxidase, and terminal deoxynucleotidyl transferase. All antibodies were obtained from Becton-Dickinson. Blasts were gated for analysis using CD45 expression and light side-scatter characteristics. Blasts were considered positive for an antigen based on an arbitrary cutoff level of at least 20% blasts that expressed the antigen compared with an isotype control.

Conventional Cytogenetic Analysis

Conventional cytogenetic analysis of BM aspirate specimens was performed using standard Giemsa trypsin G-banding procedures as described previously.⁴⁷ The results of cytogenetic studies were categorized according to a classification system proposed by Haferlach et al.⁴⁸ Briefly, the favorable risk category included AML cases associated with t(8;21), t(15;17), inv(16), and t(16;16), and the unfavorable risk AML category included cases with −5/5q-, −7/7q-, inv(3), t(3;3), 11q23 abnormalities, 17p abnormalities, and a complex aberrant karyotype (three or more abnormalities). All AML cases with other abnormalities or a normal karyotype formed an intermediate risk category.⁴⁸

Detection of FLT3 and NPM1 Mutations

Genomic DNA was extracted from BM aspirate specimens and FLT3 was assessed for internal tandem duplication (ITD) in the activation loop of the tyrosine kinase domain using polymerase chain reaction (PCR) methods followed by capillary electrophoresis.⁴⁶ To detect NPM1 mutation, exon 12 was amplified by PCR using the following primers: 5’-GATGTTGAACTATGCAAAGAGACA-3’ (forward) and 5’-AACCAAGCAAAGGGTGGAGTT-3’ (reverse). The PCR products were purified using a MinElute PCR purification Kit (Qiagen, Valencia, CA) and directly sequenced using a 5’-GGCATTTTGGACAACACA-3’ (reverse) primer, fluorescence dye chain-terminator chemistry (Sanger sequencing), and an ABI Prism 3100 or 3130 genetic analyzer (Applied Biosystems, Foster City, CA). The ABI GeneMapper software program (Applied Biosystems) was used to analyze the raw data.

Immunohistochemical Analysis

Expression of CXCR4 and pCXCR4 by blasts was analyzed using immunohistochemical methods as described previously.⁴⁹ Briefly, we used formalin-fixed, formic acid-decalcified, paraffin-embedded BM biopsy specimens. Total CXCR4 expression was assessed using a polyclonal rabbit antibody specific for CXCR4 (clone 2074; Abcam). A formalin-fixed, paraffin-embedded cell block of the Jurkat cell line, known to be positive for CXCR4,⁵⁰ was used as a positive control. Phosphorylated (p) CXCR4 was assessed using a rabbit polyclonal pCXCR4-specific antibody (kindly provided by Dr. Joshua Rubin, Washington University, St. Louis, MO).⁵¹ A formalin-fixed, paraffin-embedded glioblastoma multiforme tissue array (kindly provided by Dr. Greg Fuller, The University of Texas MD Anderson Cancer Center, Houston, TX) was used as a positive control for pCXCR4 expression. CXCR4 and pCXCR4 expression by blasts was assessed by counting 5 random high-power microscopic fields (x400) and calculating the mean percentage of positive blasts (positive blasts/all blasts). A BM specimen was considered positive if at least 10% of blasts exhibited a cytoplasmic pattyern of expression. Immunohistochemistry slides were reviewed independently by two authors (SK and PL) and discrepant interpretations were reviewed together using a multiheaded microscope to reach a consensus.

Statistical Analysis

The Pearson chi-square test (or Fisher exact two-tailed test) was used to assess associations among categorical variables. Kaplan-Meier methods were used to estimate overall survival (OS) and relapse-free survival (RFS), the latter computed starting from the date of diagnosis. Relapse or death was censored at the date of last follow-up if relapse or death was not observed. Univariate and multivariate Cox proportional hazards models were used to determined the association between tumor characteristics and survival (OS and PFS) according to age, sex, race, presence of antecedent hematological disorder,⁵² achievement of complete remission (CR), and laboratory data at the time of diagnosis. Specific laboratory data assessed included leukocyte and platelet counts; hemoglobin level; percentage of blasts in peripheral blood and BM; serum bilirubin, creatinine, and albumin levels; cytogenetic results; presence or absence of FLT3 and NPM1 mutations; and presence or absence of expression of CXCR4 and pCXCR4. In the multivariate Cox proportional hazards model, all potential covariates were included in the full multivariate Cox proportional hazards model. A backward elimination method was used to determine the final model. Statistical analysis was performed using the STATA/SE software program (version 11.2; StataCorp LP).

RESULTS

The study group included 117 AML patients, 62 (53%) men and 55 (47%) women, with median age of 61 years (range, 18-88). Twenty-six (22%) patients had an antecedent hematological disorder. All patients received intensive chemotherapy according to institutional protocols. Fifty-nine (50%) patients achieved CR, of whom 18 patients relapsed; 58 (50%) patients failed to achieve CR. Thirty-five (30%) patients were alive and 82 (70%) patients were dead at last follow-up. The median follow-up was 8 months ranging from less than a month for patients with early death up to 117 months. The 3-year OS rate was 32% (95% CI: 24-41%) and the 5-year OS rate was 27% (95% CI: 19-36%).

The laboratory data are summarized in Table 1. We detected CXCR4 expression in 75 (64%) patients and pCXCR4 expression in 31 (26%) patients, hence CXCR4 was found to be phosphorylated in 41% of cases. All p-CXCR4 positive were also positive for CXCR4, and all CXCR4-negative patients were also negative for pCXCR4. We detected NPM1 mutations in 63 of 116 (54%) patients assessed. We did not identify NPM1 mutations in patients with AML associated with t(8;21) or inv(16). In patients with NPM1 mutations, 40 (63%) expressed CXCR4 and 15 (24%) expressed pCXCR4. In the patients with wild type NPM1, 25 (47%) and 15 (28%) expressed CXCR4 and pCXCR4, respectively. NPM1 mutations were not associated with CXCR4 (p = 0.212) or pCXCR4 (p = 0.355) expression. FLT3-ITD mutations were detected in 22 of 114 (19%) patients tested. In 21 patients with FLT3-ITD mutations assessed, 13 (62%) had CXCR4 expression and 7 (33%) had pCXCR4 expression. In a group of 77 patients with wild type FLT3, 50 (65%) had CXCR4 expression and 22 (29%) had pCXCR4 expression. NPM1 mutations were significantly associated with FLT3/ITD mutations (p = 0.040). FLT3-ITD did not correlate with CXCR4 (p = 0.797) or pCXCR4 (p = 0.672) expression. There was no statistically significant correlation between therapeutic modalities and CXCR4 expression, pCXCR4 expression, NMP1 mutations, or FLT-3 ITD.

Table 1.

Summary of Laboratory Data

Variable	Median (range)
White blood cell count, × 109/L	23.8 (0.3-200.5)
Platelet count, × 109/L	51 (5-306)
Hemoglobin level, g/dL	9.2 (4.1-14.8)
Peripheral blood blasts count, %	44 (0-99)
Bone marrow blasts count, %	64 (20-96)
Albumin level, g/dL	3.6 (1.4-4.9)
Bilirubin level, mg/dL	0.5 (0.1-2.7)
Creatinine level, mg/dL	1.0 (0.5-4.4)
Cytogenetics	No. of Patients (%)
Favorable	19 (16)
Intermediate	70 (60)
Unfavorable	28 (24)
No analyzable metaphases	3 (3)
NPM1
Mutated	63 (54)
Wild-type	53 (45)
Not done	1 (1)
FLT3
ITD	22 (19)
Wild-type	92 (79)
Not done	3 (3)
CXCR4 expression
Positive	75 (64)
Negative	40 (34)
Not done	2 (2)
pCXCR4 expression
Positive	31 (26)
Negative	70 (60)
Not done	16 (14)

In the univariate Cox proportional hazards model, shorter OS was significantly associated with older age (hazard ratio [HR]: 1.07; p < 0.001), history of antecedent hematologic disorder (HR: 2.21; p = 0.006), thrombocytopenia (HR: 1.01; p = 0.002), increased serum creatinine level (HR: 2.14; p < 0.001), decreased serum albumin level (HR:1.38; p = 0.031), failure to achieve CR (HR: 5.26; p < 0.001), CXCR4 expression (HR: 1.75; p = 0.042), wild type NPM1 (HR: 1.74; p = 0.023), and unfavorable cytogenetics (HR: 7.16; p < 0.001). We did not find a statistically significant association between OS and sex, leukocyte or platelet count, hemoglobin level, serum bilirubin level, FLT3/ITD, or pCXCR4 expression. The Kaplan-Meier curves for CXCR4 expression and NPM1 status are presented in figures 1A and 1B, respectively. Of interest, when we analyzed the impact of CXCR4 expression separately in patients with wild type NPM1 or mutated NPM1, CXCR4 expression had impuct on OS only in patients with wild type NPM1 (p = 0.028, figure 1C). CXCR4 expression in patients with NPM1 mutations appeared to have worse OS but the difference did not reach the statistical significance (p = 0.139, figure 1D).

Figure 1.

A) Adverse effect of CXCR4 expression on overall survival in the entire study group; B) Adverse effect of wild type NMP1 on overall survival in the entire study group; C) Adverse effect of CXCR4 expression on overall survival in patients with wild type NMP1; D) Adverse effect of CXCR4 expression on overall survival in patients with mutated NMP1; all curves are drawn using Kaplan-Meier method.

In the multivariate Cox proportional hazards model (Table 3), shorter OS was associated with antecedent hematological disorder (HR:2.63; p = 0.005), failure to achieve CR (HR:5.88; p <0.001), thrombocytopenia (HR:1.02; p < 0.001), unfavorable cytogenetics (HR:7.16; p < 0.001), CXCR4 expression (HR:2.31; p = 0.008), and wild type NPM1 (HR:1.82; p = 0.030). In the univariate Cox proportional hazards model, shorter RFS was associated with failure to achieve CR (HR:4.55; p <0.001), older age (HR:1.03; p < 0.001), antecedent hematological disorder (HR:1.78; p = 0.035), increased serum creatinine level (HR:1.97; p < 0.001), decreased serum albumin level (HR:1.33; p = 0.030), CXCR4 expression (HR:1.59; p = 0.060), pCXCR4 expression (HR:1.02; p = 0.0934), wild type NPM1 (HR:0.62 p = 0.028), and unfavorable cytogenetics (HR:9.76; p < 0.001). We did not observe a statistically significant association between RFS and sex, leukocyte or platelet count, hemoglobin level, serum bilirubin level, or FLT3 mutation. In the multivariate Cox proportional hazards model (Table 4), shorter RFS was associated with antecedent hematological disorder (HR:2.18; p = 0.043), failure to achieve CR (HR:2.94; p <.001), thrombocytopenia (HR:1.01; p = 0.009), unfavorable cytogenetics (HR:5.94; p = 0.002), CXCR4 expression (HR:3.59; p = 0.001), pCXCR4 expression (HR:2.88; p = 0.002), and wild type NPM1 (HR:3.33; p <0.001).

Table 3.

Multivariate Cox Proportional Hazards Model for Relapse-Free Survival

Variable	P	HR	95% CI
Antecedent hematological disorder	0.043	2.18	1.03-4.64
Thrombocytopenia	0.009	1.01	1.00-1.01
Unfavorable cytogenetics	0.002	5.94	1.90-18.56
CXCR4 expression	0.001	3.59	1.73-7.46
pCXCR4 expression	0.002	2.88	1.49-5.56
Wild type NPM1	<0.001	3.33	1.81-6.25

DISCUSSION

Based on data published by Zhang and colleagues ⁴¹ which suggested that the CXCR4-CXCL12 axis and NPM pathways are linked, we hypothesized that there would be a correlation or possibly synergistic or antagonistic effects between CXCR4 and NPM in AML The data we present, however, show no correlation between NPM1 mutational status and frequency of CXCR4 expression. These data suggests that NPM and CXCR4 predict prognosis independently; a finding indirectly supported by the fact that the prognostic impact of NPM1 mutations and CXCR4 expression was evident in multivariate analysis. Therefore, the NPM and CXCR4 pathways likely affect leukemic cell survival via separate mechanisms. Even if an interaction between NPM1 and CXCR4 does occur in AML, it seems unlikely to have an important effect on patient prognosis. This being stated, it is worth noting that the number of patients in this study is relatively small, 117 patients, and it is possible that larger studies might detect overlap between NPM1 and CXCR4 effects on prognosis. The potential group of interest could be a group of young patients with normal karyotype, who are known to have high frequency of NPM1 mutation ^{37, 38}. Alternatively, more precise and more quantitative detection of CXCR4 and pCXCR4 expression not achievable by immunochemistry, might uncover a relationship between the NPM and CXCR4 pathways. For example, flow cytometric assays provide quantitative evaluation of CXCR4 receptor density expression per cell including AML progenitor cells which might convey important functional consequences in CXCR4-expressing patients. In fact, recent studies by Chou et al. utilizing both murine knock-in model of AML and human AML specimens demonstrated remarkable suppression of CXCR4/CXCL12-related gene signatures, attesting to the role of bone marrow niche defect in NPM1-mutant leukemias. ⁴² Interestingly, CXCR4 is not listed as one of the differentially expressed genes in their study, suggesting that differences in CXCL12 ligand expression could possibly account for the observed functional differences. Importantly, two recent papers in Nature confirmed the critical importance of CXCL12 produced by specialized stromal and endothelial cells within the bone marrow microenvironment for the support of normal HSC.^{53, 54} Further studies including quantitative assessment of CXCF4 and CXCL12 levels, bone marrow architecture and respective contributions of endothelial vs osteoblastic niches ^{55, 56} are clearly warranted to delineate the role of perturbed microenvironment in NPM1-mutant AML including improved outcomes with cytarabine-containing regimens.

The SDF-1α (CXCL12)-CXCR4 axis is known to play a key role in normal and leukemic hematopoiesis, and investigators are interested in this axis as a potential therapeutic target for AML patients. Tavor and colleagues have shown that inhibition of the SDF-1α–CXCR4 axis by AMD3100 reduced the proliferation and survival of AML cell lines and primary AML cells and that many genes involved in regulation of cell proliferation are differentially expressed in AML cells treated with AMD3100 compared with control cells.⁵⁷ Tavor and colleagues also showed that a neutralizing treatment targeting CXCR4 led to inhibition of the repopulation and survival of human AML cells after engraftment in mice.^{57, 58} Importantly, this effect was considerably more significant in the leukemic stem cell compartment than in the normal human stem cell compartment, suggesting a therapeutic potential for CXCR4 neutralization in AML.⁵⁹

Recently, Uy and colleagues reported that plerixafor, a bicyclam small molecule antagonist of CXCR4 binding to CXCL12, combined with mitoxantrone, etoposide, and cytarabine yielded a promising CR rate of 39% in patients with relapsed or refractory AML.⁶⁰ These data support the hypothesis that inhibition of the CXCR4/CXCL12 axis can disrupt the interaction of neoplastic cells with the bone marrow microenvironment increasing the sensitivity of neoplastic cells to chemotherapy.

CXCR4 activation is thought to be primarily ligand dependent, supported by the anti-tumor efficacy of the agent AMD 3100, an antagonist of CXCL12 binding.^61-66 CXCR4 is phosphorylated in response to ligand binding in a G-protein-coupled receptor kinase 2–dependent fashion.^20-22 Receptor phosphorylation stimulates the interaction of ß-arrestin with the carboxy terminus ceasing CXCR4-mediated activation of Gα_i, promoting dynamin-dependent, clathrin-mediated receptor endocytosis, and enhancing CXCR4-RAF-dependent signaling.^{20, 21, 23, 24} Regulation of phosphorylation and internalization has significant effects on CXCR4-mediated cellular responses, and for this reason some authors believe that the descriptions of total CXCR4 expression alone are inadequate. In this study, we demonstrated that CXCR4 is phosphorylated in 41% if AML patients, and that phosphorylated (activated) CXCR4 expression independently predicts poorer RFS but not for OS in adults with AML. This result is similar to observations in adult patients with acute lymphoblastic leukemia,⁴⁹ but the precise mechanism of this phenomenon and the upstream kinases involved remains to be elucidated. Notably, PIM1 kinase which is frequently activated in FLT3/ITD-mutant AML was reported to cause phosphorylation of serine 339 in the CXCR4 intracellular domain, ⁶⁷ however, we did not observe higher frequency of CXCR4 phosphorylation in NPM1-mutant AML which additionally carried FLT3-ITD mutations.

CONCLUSIONS

In summary, we could not demonstrate any correlation between NPM1 mutations and CXCR4 or phosphorylated CXCR4 expression suggesting that in adult AML patients. Nevertheless, both NPM1 mutations and CXCR4 expression predict prognosis. These data suggest that the CXCR4 and NPM pathways act independently. Our findings support the notion that targeting CXCR4 may improve outcomes in AML, and implicate routine testing CXCR4 expression as an additional prognostic marker in AML..

CLINICAL PRACTICE POINTS.

• CXCR4 expression and wild type NPM1 gene are associated with shorter overall and progression free survival in adult patients with AML

• Phosphorylated CXCR4 expression is associated with shorter progression free survival in adult patients with AML

• There is no correlation between NPM1 mutations and CXCR4 or phosphorylated CXCR4 expression in adult AML suggesting that CXCR4 and NPM pathways act independently.

• These data provide a rationale for independent targeting CXCR4 and NPM pathways in leukemic blasts to improve an outcome in AML patients.

Table 2.

Multivariate Cox Proportional Hazards Model for overall survival.

Variable	P	HR	95% CI
Antecedent hematological	0.005	2.63	1.34-5.16
Failure to achieve CR	<0.001	5.88	3.11-11.11
Thrombocytopenia	<.001	1.02	1.01-1.03
Unfavorable cytogenetics	<0.001	7.16	2.62-19.60
CXCR4 expression	0.008	2.31	1.25-4.29
Wild type NPM1	0.030	1.82	1.06-3-13