Nucleophosmin (NPM1) is a highly conserved nucleo-cytoplasmic shuttling protein that shows a restricted nucleolar localization. Mutations of NPM1 gene leading to aberrant cytoplasmic dislocation of nucleophosmin (NPMc+) occurs in about one third of acute myeloid leukaemia (AML) patients that exhibit distinctive biological and clinical features. We discuss the latest advances in the molecular basis of nucleophosmin traffic under physiological conditions, describe the molecular abnormalities underlying altered transport of nucleophosmin in NPM1-mutated AML and present evidences supporting the view that cytoplasmic nucleophosmin is a critical event for leukaemogenesis. We then outline how a highly specific immunohistochemical assay can be exploited to diagnose NPM1-mutated AML and myeloid sarcoma in paraffin-embedded samples by looking at aberrant nucleophosmin accumulation in cytoplasm of leukaemic cells. This procedure is also suitable for detection of haemopoietic multilineage involvement in bone marrow trephines. Moreover, use of immunohistochemistry as surrogate for molecular analysis can serve as first-line screening in AML and should facilitate implementation of the 2008 World Health Organization classification of myeloid neoplasms that now incorporates AML with mutated NPM1 (synonym: NPMc+ AML) as a new provisional entity. Finally, we discuss the future therapeutic perspectives aimed at reversing the altered nucleophosmin transport in AML with mutated NPM1.
Nucleophosmin is a nuclear-cytoplasmic shuttling protein
Nucleophosmin (NPM1), also named B23 or Numatrin, is an ubiquitously expressed protein belonging to the nucleoplasmin family of nuclear chaperones.1 It is encoded by the NPM1 gene that, in humans, maps to chromosome 5q35.2 NPM1 is essential for embryonic development3 and is frequently translocated or mutated in haematological malignancies.4
Three nucleophosmin isoforms are generated through alternative splicing. NPM1 (or B23.1), the dominant isoform (accession n. NM_002520),5 is a 294-amino acid phosphoprotein of about 37 kDa that is expressed in all tissues. NPM1.2 (or B23.2) uses an alternate 3′-terminal exon compared with variant 1, resulting in a 259-amino acid protein (accession n. NM_001037738) with a distinct C-terminus; it accounts for minimal nucleophosmin content in tissues. NPM1 (B23.1) and NPM1.2 (B23.2) have different subcellular distribution patterns:5 NPM1 protein is localized only in the nucleolus6, 7 and NPM1.2 mainly in the nucleoplasm.8 A third variant (accession number NM_199185) lacks an alternate in-frame exon compared with variant 1, resulting in a shorter protein whose functions and expression pattern are unknown.
Experiments on NPM1/B23 migration in interspecies (chick-mouse) heterokaryons and studies on nuclear accumulation of cytoplasmically microinjected anti-nucleophosmin antibodies provided conclusive evidence that NPM1 (B23.1), despite its nucleolar localization,6, 7 shuttles constantly back and forth between nucleus and cytoplasm.9 As NPM1 gene mutations in acute myeloid leukaemia (AML) always result in traffic alterations of the most common isoform 1, we will refer to it constantly as NPM1 in this review.
Functional domains regulating cellular traffic of wild-type NPM1
Nuclear-cytoplasmic shuttling of wild-type NPM1 (NPM1wt)9, 10 is critical for most of its functions, including regulation of ribosome biogenesis and control of centrosome duplication.11 NPM1wt shuttles across cytoplasm and nucleoplasm as well as between nucleoplasm and nucleolus. Distinct NPM1wt functional domains regulate this complex shuttling activity.
NPM1wt transport between cytoplasmic and nuclear compartments
This process requires energy and closely depends on two NPM1wt functional domains (Figure 1). A bipartite nuclear localization signal (NLS),12 which is located between the two acidic regions in the nucleophosmin mid-region, drives NPM1wt from cytoplasm to nucleoplasm. Moreover, NPM1wt contains two leucine-rich nuclear export signal (NES) motifs13, 14 with the generally accepted loose consensus L-x(2,3)-(LIVFM)-x(2,3)-L-x-(LI) (that is, a core of closely spaced leucines or other large hydrophobic amino acids). One is found within residues 94–102 with the sequence I-xx-P-xx-L-x-L13 and the other at the N-terminus within amino acids 42 to 49 (sequence L-x-L-xx-V-x-L), with leucines 42 and 44 being critical nuclear export residues.14 These NES motifs mediate NPM1wt nuclear export by interacting with the evolutionary conserved export receptor Crm116 (also called exportin 1) (Figure 1). Fluorescence recovery after photobleaching (FRAP) showed Crm1 travels unimpeded within the nuclear compartment and interacts with NES-containing proteins both in nucleoplasm and nucleolus.17
NPM1wt traffic between nucleoplasm and nucleolus
Intranuclear trafficking of NPM1wt (and other nuclear proteins) is mostly driven by diffusion, a rapid, non-directional process that usually occurs through energy-independent mechanisms.18 FRAP analysis revealed that, in a continuous exchange with nucleoplasm,18, 19 nucleophosmin (like other typical nucleolar proteins) rapidly associates with, and dissociates from the nucleolar component, which is in keeping with lack of nucleolus membrane. Indeed, the structure of the nucleolus is thought to be determined by protein on-rate versus off-rate ratio.18 According to this model, nucleophosmin contributes to build-up the nucleolar compartment in which it is one of the most abundant components20 among the approximately 700 proteins identified by proteomics.21 In fact, flux of nucleophosmin at the nucleolus–nucleoplasm interface is definitively unbalanced towards the nucleolus. This is, in turn, the consequence of the very efficient nucleolar-binding domain (NoLS) that NPM1wt contains at its C-terminus.
The structure of nucleophosmin C-terminal domain has recently been revealed by nuclear magnetic resonance spectroscopy.22 It consists of a well-defined 3-helix right-handed bundle22 that appears similar to the archeal ribosomal protein S17e23 (Figure 1). Maintenance of C-terminus tertiary conformation closely depends on tryptophans 288 and 290 that form part of the folded domain's hydrophobic core (Figure 1), although phenylalanines 268 and 273 may also play a role.22 Stabilization of the C-terminus 3-helix structure by the two tryptophans is, in turn, critical for NPM1wt binding to the nucleolus.22 Accordingly, earlier findings that artificially induced mutations of tryptophans 288 and 290 in NPM1wt prevented its binding to the nucleolus24, 25 are now understood to be the result of loss of C-terminal domain folding.22 C-terminus tryptophan mutations that consistently occur in natural NPM1 mutants in AML patients prevent or reduce mutant binding to the nucleolus in the same manner,25, 26 thus contributing to their aberrant accumulation in leukaemic cell cytoplasm (see below).
The ability of NPM1wt protein to form homodimers also helps drive it to the nucleolus. Indeed, artificial NPM1 mutants that contain the nucleolar-binding domain but are unable to form oligomers cannot enter the nucleolus.27 Thus, it is not surprising that NPM1wt is isolated from cells mainly as oligomers,28 which are usually hexamers.29 The NPM1wt oligomer-forming property resides mainly in its hydrophobic N-terminal core structure (Figure 1), which it shares with other nucleoplasmin family members.1 Core amino acids in its two physiological N-terminal NES motifs (residues 44 and 47; residues 100 and 102) appear to be involved in NPM1 homo- and hetero-dimerization.27 It is still unclear how wild-type nucleophosmin oligomerization favours nucleolar targeting. The simplest explanation is that a NPM1 oligomer might result in a ‘high NoLS load’ that reaches the nucleolar targeting threshold. Alternatively, oligomerization could confer structural properties on the entire molecular complex that facilitate its targeting to the nucleolus.
It is still unknown whether binding of nucleophosmin to the nucleolus occurs through interaction with a protein or a nucleic acid, although very recent evidence suggests that other nucleolar proteins (that is, rpL23) are required.30 Avrainvillamide, a naturally occurring alkaloid with anti-proliferative activity, binds to the NPM1wt C-terminal domain through Cysteine-275.31 Interestingly, Cysteine-275 is located on the same face as Lysine residues 263 and 267.22 Orientated freely on the C-terminus domain surface, these Lysines appear to play an important functional (rather than structural) role in NPM1wt nucleolar localization,22 suggesting this protein domain area is involved in ligand binding.
Native NPM1 is a nucleolar protein with ubiquitous tissue expression
At immunocytochemistry, NPM1wt localizes selectively in the nucleolus,6, 7 the most prominent nuclear substructure.20 When observed at electron microscopy, the nucleolus is morphologically divided into three main components that reflect the multi-step ribosome biogenesis process: the fibrillar centres (containing DNA sequences encoding ribosomal RNA);38 the dense fibrillar component (in which nascent pre-rRNA transcripts enter to undergo a series of early processing steps); and the granular component, radiating out from the dense fibrillar component, in which intermediate or late stages of RNA maturation occur.39 Notably, nucleophosmin expression is restricted to the granular component,20 reflecting its involvement in later processing and assembly steps and in ribosomal subunit export.40 At first glance, the restricted NPM1wt nucleolar localization appears surprising in the light of its continual shuttling between nucleus and cytoplasm.9, 10 However, analysis of the dynamics determined by the functional domains regulating NPM1wt nuclear-cytoplasmic and intranuclear traffic (see above), explains this staining pattern.
The following scenario can be envisaged. Because of its NLS, NPM1wt migrates from cytoplasm to nucleoplasm and from there it can move to the nucleolus because of its strong nucleolar-binding domain. Interaction of Crm1 (exportin 1) with the two NPM1wt N-terminus NES motifs13, 14 (which presumably occurs both in nucleoplasm and nucleolus) ensures NPM1wt nuclear export. Using a Rev(1.4)-based shuttling assay, we found that both NPM1wt NES motifs have low export efficiency41 and, therefore, little native protein is exported from nucleus to cytoplasm. As NLS-mediated nuclear import of NPM1wt greatly predominates over export,41 most NPM1wt resides within the nucleolus (Figure 2). In contrast, the rare NPM1.2 (or B23.2) truncated isoform localizes to nucleoplasm,8 as it lacks the C-terminus nucleolar-binding domain and cannot, therefore, bind efficiently to the nucleolus.24 Knowledge of how the nucleophosmin functional domains regulate its steady state cellular traffic helped elucidate the molecular mechanisms underlying aberrant cytoplasmic expression of nucleophosmin in AML with mutated NPM1 (see below).
NPM1wt expression in normal and neoplastic human tissues was investigated by immunohistochemistry with specific monoclonal antibodies.7, 26, 42, 43 The protein is found ubiquitously,7, 44 and it is usually expressed at higher amounts in the nuclei of proliferating, than resting or differentiated cells. For example, we have found that in human bone marrow, early progenitors, such as proerythroblasts or promyelocytes, express much more nuclear nucleophosmin than terminally differentiated cells, such as normoblasts or neutrophils. The usual positivity of NPM1wt in paraffin-embedded samples is ‘nucleus-restricted, diffuse’ rather than nucleolar,26, 45 reflecting the limits of immunohistochemical staining of paraffin sections in detecting subnuclear structures, such as nucleoli or nuclear bodies.46 In normal human tissues, immunohistochemistry also identifies a small percentage of cells that express nucleophosmin in cytoplasm in the absence of NPM1 mutations. These represent cells in mitosis as well as post-mitotic and apoptotic elements in which NPM1wt localizes in the cytoplasm because of loss of nuclear membrane (mitosis) or diffusion of the protein from nucleus to cytoplasm (apoptosis) (Falini B, unpublished observation).
Nucleophosmin traffic is perturbed in lymphomas and leukaemias carrying NPM1 gene alterations
The NPM1 gene is translocated or mutated in several haematological malignancies,4 and these genetic alterations usually perturb the cellular traffic of nucleophosmin. In this review, we will mainly focus on alterations in nucleophosmin traffic in AML with mutated NPM1.
Human haematological malignancies carrying translocations involving the NPM1 gene
Anaplastic large cell lymphoma (ALCL) expressing the anaplastic lymphoma kinase (ALK) protein is now included as a distinct entity in the 2008 World Health Organization (WHO) classification of lymphoid and haemopoietic tissues.47 ALCL cells harbouring the t(2;5) translocation (about 85% of cases) express both NPM1–ALK fusion protein48 and NPM1wt (encoded by the residual NPM1 allele). Because of its ALK moiety, at least a proportion of the NPM1–ALK protein anchors to ALCL cell cytoplasm.46 This accounts for the aberrant cytoplasmic expression of nucleophosmin that is revealed with antibodies against the NPM1 N-terminus (which is retained in the fusion protein).46, 49 In contrast, NPM1wt that can be detected with an antibody against NPM1 C-terminus (that is not retained in NPM1–ALK), maintains its expected nucleolar expression in ALCL with t(2;5),49 suggesting that localization is not perturbed by the NPM1–ALK fusion product.
In the rare AML/myelodysplasia carrying the t(3;5)(q25;q35) translocation, leukaemic cells express both NPM1–MLF1 (myelodysplasia/myeloid leukaemia factor 1) fusion protein50 and NPM1wt (encoded by the normal NPM1 allele). Like ALCL with t(2;5), AML with t(3;5) shows aberrant nucleophosmin expression in cytoplasm as, due to its MLF1 moiety, the NPM1–MLF1 fusion protein localizes in leukaemic cell cytoplasm.51 However, unlike ALCL with t(2;5), NPM1wt is also dislocated into cytoplasm of AML cells carrying the t(3;5)51 through a mechanism that still needs to be deciphered.
The NPM1–RARα fusion protein was detected in extremely rare cases of acute promyelocytic leukaemia carrying the t(5;17) translocation.52 In two patients investigated by immunocytochemistry,53, 54 nucleophosmin showed diffuse nuclear positivity that clearly differed from the expected nucleolar expression of NPM1wt.
The transforming role of ALK, MLF1 and RAR-α fusion protein moieties in the above malignancies is well established.46 Nevertheless, the view that the NPM1 moiety of the chimeric protein only provides a dimerization substrate for the C-terminal onco-protein has recently been challenged by in vivo evidence that NPM1 is an haploinsufficient tumour suppressor gene.55 Therefore, its heterozygous loss could also contribute to the pathogenesis of the disease.11, 55
AML with mutated NPM1 (NPMc+ AML)
Having observed that in ALCL with t(2;5) the presence of the NPM1–ALK fusion protein was associated with ectopic expression of nucleophosmin in tumour cell cytoplasm,49 we decided to adopt immunohistochemical detection of cytoplasmic nucleophosmin, as a simple, rapid screening test for putative NPM1 gene alterations in a wide range of human malignancies. This approach led, in 2005, to the discovery that about 35% of adult AML had aberrant nucleophosmin expression in leukaemic cell cytoplasm.26
We initially named such AML cases NPMc+ (cytoplasmic positive) to distinguish them from AML with nucleus-restricted NPM expression26 (named NPMc−). As NPMc+ AML was closely associated with normal karyotype and other distinctive features,26 an unidentified genetic lesion seemed a strong possibility as underlying cause of nucleophosmin dislocation into cytoplasm. As known NPM1 fusion gene/transcripts were not found in NPMc+ AML leukaemic cells,26 we sequenced the entire coding sequence of NPM1 and found heterozygous insertions at exon-12 of the gene causing a reading frame-shift that resulted in a slightly longer protein with a different C-terminal amino-acidic sequence.26
Whole genomic sequencing is emerging as a powerful tool for solving complex problems, such as identification of underlying genetic lesions in AML with normal karyotype.56 On the other hand, the story of how NPM1 mutations were discovered in AML teaches that sometimes solutions to complex problems may derive from a simple observation at the microscope using inexpensive techniques, like immunohistochemistry. The crucial step was taken when immunoscreening for nucleophosmin expression was extended to tumours other than ALCL with t(2;5). However, had we not, several years previously, chosen to generate monoclonal antibodies directed against fixative-resistant epitopes of the NPM1 molecule,7 aberrant nucleophosmin dislocation in AML cell cytoplasm might have been difficult to recognize, as this staining pattern is optimally detected only in fixed, paraffin-embedded sections.26, 45 Another factor also played a role in the discovery. Even though NPM1 is a nuclear-cytoplasmic shuttling protein,9 the very small fraction that is physiologically present in cytoplasm is fortunately below the immunohistochemistry detection threshold. Therefore, immunohistochemical detection of cytoplasmic nucleophosmin could be exploited for specific recognition of an NPM1 mutation in AML cells.
Molecular mechanism of cytoplasmic accumulation of nucleophosmin in AML with mutated NPM1
Aberrant cytoplasmic expression of nucleophosmin is the immunohistochemical hallmark45 of AML with mutated NPM1. All NPM1 mutations, despite molecular heterogeneity (about 50 mutation variants identified so far), result in a shift of the reading frame leading to common changes at the very end of the NPM1 protein C-terminus25 that are responsible for cytoplasmic mislocalization of the NPM1 leukaemic mutants. Mutation-related changes causing this effect, shown in Figure 1, are (i) generation of a new NES motif,57 which reinforces the Crm1-dependent nuclear export of NPM1 protein and (ii) loss of tryptophan residues 288 and 290 (or residue 290 alone),58 which cause the 3-helix structure of NPM1 C-terminal domain to unfold,22 thus preventing or decreasing NPM1 binding to the nucleolus.
Both alterations are critical for perturbing traffic of NPM1 mutants in AML cells25 and an hypothetical scheme of how this occurs is shown in Figure 2. As NPM1 leukaemic mutants retain their NLS, they move efficiently from cytoplasm to nucleoplasm. In the nucleus, however, the capability of NPM1 mutants to bind nucleolus is hindered either completely (when both tryptophans 288 and 290 are mutated) or partially (when only tryptophan 290 is altered).41 Consequently, a significant fraction of NPM1 mutant protein roams in nucleoplasm in which it is more prone to bind to Crm1 through the additional C-terminus NES motif introduced by the mutation41 (Figure 1). Under these circumstances, the mutation-driven NES-Crm1-mediated nuclear export is more efficient than import into the nucleolus and, therefore, the NPM1 mutants accumulate in cytoplasm (Figure 2).
NPM dislocation into cytoplasm: a critical event for leukaemogenesis?
Several observations clearly point to nucleophosmin dislocation into cytoplasm as the most distinguishing functional consequence of NPM1 mutations in AML. They include the following.
Nucleophosmin dislocation into cytoplasm is specific for AML
Aberrant nucleophosmin expression in cytoplasm was detected by immunohistochemistry in about 35% of adult AML.26 In contrast, NPM1 is not mutated nor aberrantly expressed in cytoplasm in any other human haemopoietic or extra-haemopoietic neoplasm.26, 44, 59 Moreover, ectopic nucleophosmin expression is usually observed in the entire leukaemic population. One exception is NPM1-mutated AML with M5b morphology, in which cytoplasmic NPM1 is seen mainly in monoblasts but not in the more mature monocytic forms.26 This expression pattern is in keeping with experimental in vitro observations of NPM1 downregulation during maturation.60 Finally, in AML, aberrant NPM1 expression in cytoplasm was associated with a distinct gene expression profile61 and microRNA signature.62
Generation of a NES motif through duplication appears unique to the NPM1 mutation
About 80% of AML with mutated NPM1 bear the so-called mutation A, that is, a duplication of a TCTG tetranucleotide at positions 956–959 of the NPM1 gene.58 A genome-wide computational analysis designed to evaluate whether duplication of 3–6 nucleotide combinations could lead to generation of a known NES motif in any human gene showed that duplication-linked generation of a NES motif was unique to the NPM1 mutation,44 an AML-specific genetic event.
Nucleophosmin cytoplasmic expression is a stable event in AML with mutated NPM1
Many observations indicate that NPM1 mutations are stable.63, 64, 65 Accordingly, in patients with NPM1-mutated AML who were investigated by immunohistochemistry during the course of the disease, cytoplasmic nucleophosmin was consistently detected at relapse not only in the bone marrow but also in extramedullary sites. Persistence of NPM1 expression in leukaemic cell cytoplasm over a period of many years was demonstrated in patients at late relapse66, 67 and in a xenotransplant model of AML with mutated NPM1 in immunodeficient mice.65 Moreover, the OCI-AML3 cell line, first established about 20 years ago, stably harbours NPM1 mutation A and displays aberrant nucleophosmin expression in cytoplasm.68 It should be noted that loss of NPM1 mutation at relapse has been rarely observed in NPM1-mutated AML.69, 70 Discrepancies with other reports may be due to technical reasons or one might speculate that these cases represent treatment-related or even second de-novo AMLs rather than relapses from previous leukaemia.71 Extensive investigation by FISH and other molecular techniques are warranted to clarify the issue.
NPM1 leukaemic mutants are born to be exported
Two major observations suggest NPM1 mutations act to maximize mutant export into cytoplasm underscoring this event as critical for leukaemogenesis.
All NPM1 mutations result in dislocation of the encoded proteins into cytoplasm
Although NPM1 mutations in AML usually occur at exon-12 of the gene,26 they may very occasionally involve exon-972 or exon-11.73, 74 Yet, these very rare NPM1 mutants localize aberrantly in leukaemic cell cytoplasm73 through the same mechanism described in AML carrying typical exon-12 NPM1 mutations, that is, through the concerted action of tryptophan(s) changes and a new NES motif at the C-terminus.25 In particular, mutations involving exon-11 generate a truncated NPM1 protein (274 instead of 294 amino acids) with consequent loss of tryptophans at positions 288 and 290,73, 75 which are critical for nucleolar binding.24 Deletion of the two C-terminus tryptophans has the same detrimental effect on the nucleolus-binding capability of the truncated protein as their replacement in AML carrying typical exon-12 NPM1 mutations.26 Even though the reduced nucleolus-binding capability increases mutant affinity for Crm1/Exportin,25 it is not sufficient to cause aberrant mutant accumulation in cytoplasm. As with exon-12 NPM1 mutations, the additional force contributing to nuclear export and ectopic cytoplasmic accumulation of the leukaemic mutant is a new NES motif at the same C-terminus end of NPM1.73, 75 Interestingly, the only NPM1 mutation that was found to involve the splicing donor site of NPM1 exon 972 also caused deletion of C-terminus tryptophans and creation of a new NES motif. These findings in AML also explain why B23.2, the physiologically truncated NPM1 isoform, localizes in nucleoplasm.8 In fact, even though B23.2 lacks tryptophans 288 and 290, there is no new NES motif at its C-terminus to drive accumulation of the protein in cytoplasm. These observations clearly show that all NPM1 mutations inevitably cause export of the leukaemic mutants into cytoplasm.
In rare NPM1 leukaemic mutants with residual nucleolar targeting ability, stronger NES are selected
In NPM1-mutated AML, >95% mutants lack both tryptophans 288 and 290, whereas rare mutants retain tryptophan 288 and show residual nucleolar targeting ability.58 Despite this difference, all mutants aberrantly accumulate in leukaemic cell cytoplasm, implying that their nuclear export occurs even when tryptophan 288 is in place. As NES sequences may differ in export efficiency of their cargo proteins,76, 77 we hypothesized that these rare NPM1 mutants could have stronger NES motifs to ensure their efficient cytoplasmic accumulation even in the presence of a residual force driving the proteins to the nucleolus. Indeed, when both tryptophans were mutated, we found NPM1 mutants always bore the most common NES motif (that is, L-xxx-V-xx-V-x-L).41, 58 In contrast, when tryptophan 288 was retained, NPM1 mutants consistently carried variant NES motifs (that is, Valine at the second position was replaced with Leucine, Phenyl-alanine, Cysteine or Methionine).41, 58 Interestingly, we demonstrated that these variant NES motifs were endowed with stronger export efficiency than L-xxx-V-xx-V-x-L motif,41 which counteracted the force of tryptophan 288 driving mutants to the nucleolus.24
The above findings strongly suggest that NPM1 mutants are ‘born to be exported’,41 and that aberrant nucleophosmin accumulation in AML cell cytoplasm may play, through a yet unknown mechanism, a critical role in leukaemogenesis. As very low levels of NPM1wt are transiently present in cytoplasm under physiological conditions9 (due to shuttling activity), the amount of NPM1 mutant that accumulates in leukaemic cell cytoplasm and/or the highly increased shuttling rate of the mutated protein may be crucial for transforming activity.
NPM1 leukaemic mutants dislocate diverse protein partners into cytoplasm
How mutated NPM1 contributes to AML development is still unclear. One interesting hypothesis is that the NPM1 mutants exert their oncogenic action by binding and dislocating into leukaemic cell cytoplasm NPM1wt (encoded by the normal allele)25, 58 and other protein partners, including tumour suppressor p19Arf,78, 79 thereby interfering with their functions. Proteins whose subcellular distribution and function are altered by NPM1 mutants are discussed below.
Delocalization of NPM1wt
In vitro transfection studies25, 78 and immunohistochemical observations in samples from AML patients25 clearly demonstrated that NPM1 mutants recruit NPM1wt from nucleoli and delocalize it into nucleoplasm and cytoplasm. In the same way as NPM1wt with NPM1 fusion proteins,4, 46, 80 NPM1 mutants form heterodimers with NPM1wt25, 58 through the N-terminal dimerization domain, which is conserved in all NPM1 mutants.4, 58 Core amino acids in the two physiological amino-terminal NES motifs seem to play a major role in mediating NPM1 homo- and hetero-dimerization, by dictating in a ‘dose-dependent tug of war’ fashion the subcellular distribution of wild-type and mutated NPM1 proteins.27 In fact, transfected cells showed that, when NPM1wt and mutant proteins were expressed at equimolar doses, a fraction of NPM1wt was consistently retained in the nucleoli (as is most often observed in AML samples). On the other hand, in the presence of excess of mutant NPM1, all NPM1wt protein was pulled out of the nucleoli into nucleoplasm and cytoplasm,27 as seen in some patients with NPM1-mutated AML. On the contrary, excess doses of NPM1wt relocated NPM1 leukaemic mutant from cytoplasm to the nucleoli of transfected cells,27 a pattern never observed in NPM1-mutated AML patients in whom the mutant was always cytoplasmic restricted.81 This evidence points at a dose-dependent relationship between wild-type and mutated NPM1 proteins, and supports the view that NPM1 leukaemic mutants are ‘born to be exported’.41
p19Arf protein delocalization
The tumour suppressor p14ARF, a crucial positive regulator of the p53 response to oncogenic stimuli, closely interacts with NPM1wt.82 Most studies used murine cell lines to study nucleophosmin interaction with ARF, thereby focusing on its murine ortholog p19Arf. P19Arf co-localizes with NPM1wt protein in the nucleolus in which it forms high molecular complexes.83 As p19Arf has a stable structure only when bound to NPM1wt,84 the NPM1wt–p19Arf interaction appears not only to protect p19Arf from the rapid proteasome-mediated degradation85, 86 (which many misfolded proteins are subjected to87), but also to dictate its nucleolar location.84 As p19Arf stability and subcellular distribution are essential for maintenance of its basal levels and functional activity,88 cytoplasmic NPM1 mutant may contribute to AML development by inactivating p19Arf. In fact, in NIH-3T3 fibroblasts that were engineered to produce a zinc-inducible p19Arf protein, NPM1 mutant A delocalized NPM1wt and p19Arf from nucleoli to cytoplasm,78 which reduced p53-dependent (Mdm2 and p21cip1 induction) and p53-independent (sumoylation of NPM) p19Arf activities.78 When complexed with the NPM1 mutant, p19Arf stability was greatly compromised and the p53-dependent cell-cycle arrest at the G1/S boundary was weaker.79
HEXIM1 protein delocalization
Hexamethylene bis-acetamide inducible protein 1 (HEXIM1) inhibits the positive transcription elongation factor b (P-TEFb), which is a key RNA polymerase II (Pol II) transcriptional regulator. Gurumurthy et al89 reported that NPM1wt binds to HEXIM1 in vitro and in vivo, and functions as a negative HEXIM1 regulator. In transfected cells, mutated NPM1 associated with, and sequestered, HEXIM1 in cytoplasm, resulting in higher transcription of RNA pol II target genes, among which were some positive regulators of cell-cycle progression such as cyclin D1 and anti-apoptotic proteins such as Mcl-1.89
Activated nuclear factor kappaB may confer resistance to chemotherapy in AML patients.90 Cilloni et al91 showed that mutated NPM1 bound to NF-kappaB and sequestered it in cytoplasm, leading to its inactivation. The authors hypothesized that this mechanism underlies the high sensitivity of AML with mutated NPM1 to chemotherapy.91
Delocalization of proteins involved in control of c-Myc and p15 and p21
In MEF transfected cells, Bonetti et al92 found that the NPM1 mutant stabilized cMyc by dislocating to cytoplasm and degrading the F-box protein Fbw7γ, a component of the E3 ligase complex, which is involved in c-Myc ubiquitination and proteasome degradation.93 Thus, NPM1 mutant-mediated enhanced proliferative activity (by increasing c-Myc protein levels) may be a potential player in leukaemogenesis.94
Wanzel et al30 found that the NPM1 leukaemic mutant interacted with the Myc-associated zinc-finger protein Miz1, dislocating it into cytoplasm. When complexed with the NPM1 mutant in cytoplasm, Miz1 no longer exerted its physiological action of negatively regulating cell proliferation by inducing expression of cell-cycle inhibitors p15 and p21.
The role of NPM1 mutant-induced protein mislocalizations in leukaemogenesis, however, still remains unclear. In fact, information on the altered subcellular distribution of NPM1 partners (ARF, HEXIM1, Fbw7γ and Miz1) was derived only from the study of transfected non-haemopoietic cells, and no immunohistochemical evidence was provided that these proteins were really expressed aberrantly in the cytoplasm of NPMc+ AML cells from patients. Strong in vitro and in vivo evidence of a transforming role of mutated NPM1 in haematopoietic cells is indeed still lacking. The only positive cell transformation assay was achieved in mouse embryonic fibroblasts by combining NPM1 mutant with the adenoviral protein E1A, as expression of mutated NPM1 alone only conferred a senescent phenotype in this cell model.95
Diagnostic and clinical implications of aberrant cytoplasmic expression of nucleophosmin in AML with mutated NPM1
Ectopic expression of nucleophosmin in leukaemic cell cytoplasm has not only a biological impact but also diagnostic and clinical implications.
Cytoplasmic expression of nucleophosmin as immunohistochemical marker for AML
Many haematological centres identify AML patients with mutated NPM1 by directly searching for NPM1 mutations by means of molecular techniques.96 Immunohistochemistry is another valuable diagnostic approach, as detection of aberrant nucleophosmin expression in leukaemic cell cytoplasm is fully predictive of NPM1 mutations.45
Optimal detection of nucleophosmin using anti-NPM antibodies is easily achieved in paraffin-embedded samples, while fresh material (smears or cytospins) is not suitable for this purpose.26 For routine immunohistochemical diagnosis of NPMc+ AML, the most reliable reagents are monoclonal antibodies directed against fixative-resistant epitopes that are shared by wild-type and mutated NPM1 proteins7, 26, 45 (Figure 3). When stained with these reagents, AML cells harbouring an NPM1 mutation appear positive in the nucleus (which contains a fraction of NPM1wt) and in cytoplasm that contains mutated NPM1 and NPM1wt (recruited by the mutant through heterodimer formation)45 (Figures 3 and 4). Although polyclonal antibodies recognizing only mutant NPM1 protein45, 68 are very useful for investigating lineage involvement81 and for diagnosing AML with mutated NPM1 by western blotting,97 they are of limited value in immunohistochemical detection of cytoplasmic nucleophosmin on paraffin sections81 because antigenic epitopes are frequently denaturated by fixation/decalcification procedures.
In most NPMc+ AML samples, immunostaining results are striking and easy to interpret. Problems may occasionally be encountered in the following circumstances: (i) weak nucleophosmin positivity in leukaemic cell cytoplasm; (ii) cytoplasmic expression of nucleophosmin in only a percentage of leukaemic cells, as usually occurs in AML cases with M5b morphology26 and (iii) determining nucleophosmin positivity in small-sized blasts with limited cytoplasmic rim. Corollary evidence is obtained by staining parallel paraffin sections with a monoclonal antibody against a different shuttling nucleolar protein such as C23/nucleolin.9 Nucleus-restricted localization of this protein in the above situations confirms the diagnosis of NPMc+ AML26 (Figure 3). Finally, non-specific nucleophosmin expression in cytoplasm (in the absence of NPM1 mutation) may occur in AML cells in association with apoptosis and extensive bone marrow necrosis. These cases are easily recognized because immunostainings show that C23/nucleolin is also aberrantly dislocated into cytoplasm in a non-specific manner.
Immunohistochemical detection of cytoplasmic nucleophosmin in AML has several diagnostic applications. It is a first-choice assay for recognizing NPM1-mutated AML in cases of dry-tap or in extramedullary sites,66, 98 and for detecting haemopoietic lineage involvement81 (Figure 4). As cytoplasmic nucleophosmin is fully predictive of NPM1 mutations,45 immunohistochemistry could be used as a first-line screening to rationalize cytogenetic and molecular studies in AML. A potential algorythm is shown in Figure 5. As immunohistochemistry is a cheap, easy, and readily available technique, applying it to detect cytoplasmic nucleophosmin should facilitate implementation of the 2008 WHO classification of myeloid neoplasms, which now incorporates AML with mutated NPM1 (synonym: NPMc+ AML) as a new provisional entity.99
Trephine biopsies are not, however, performed in all countries and cases of acute leukaemia. Under these circumstances, PCR-based screening for NPM1 mutations in peripheral or bone marrow blast cells from AML patients is the first-choice approach (Figure 5). A method for rapid detection of cytoplasmic nucleophosmin using intracellular flow cytometry has recently been proposed100 and may emerge in the future as a valuable alternative procedure for the initial screening of AML samples. Finally, for well-equipped laboratories, both molecular and immunohistochemical studies can be applied in parallel, to complement each other (Figure 5).
It should be emphasized that immunohistochemistry is not suitable for monitoring minimal residual disease in NPM1-mutated AML because in normal bone marrow a few cells (mitoses, post-mitotic and apoptotic cells) may show cytoplasmic positivity for NPM1wt (in the absence of NPM1 mutation). Thus, quantitative PCR analysis is the first-choice assay for detecting minimal amounts of NPM1 mutants.101, 102 This procedure is ideally carried out on bone marrow aspirates.
In conclusion, availability of several techniques for detecting NPM1 mutations offers a great advantage as it may facilitate application of the WHO classification. Investigators can choose an appropriate technique (or a combination of techniques) on the basis of their experience, equipment, preferred diagnostic procedure (for example, bone marrow biopsy and/or aspirate) and type of analysis (initial mutational screening, monitoring of minimal residual disease, etc.).
Impact of mutated cytoplasmic nucleophosmin on clinical management of AML
The flexible diagnostic approach depicted in Figure 5 has important prognostic and clinical implications. AML with mutated NPM1 usually carries normal karyotype and, according to several studies, is associated with good outcome when FLT3-ITD is absent.103, 104, 105, 106, 107, 108 About 15% of NPM1-mutated AML harbour chromosomal aberrations (other than the recurrent ones), which are likely to represent secondary events. In a recent study designed to establish the prognostic value of these additional genetic abnormalities, we found that prognosis in AML with mutated NPM1 is not influenced by karyotype.109
Given its favourable prognosis, it is recommended that NPM1-mutated/FLT3-ITD negative AML be treated by conventional chemotherapy regimens with or without autologous stem cell transplantation. Schlenk et al108 recently reported that NPM1-mutated AML patients without FLT3-ITD did not appear to benefit from allogeneic stem cell transplantation in first complete remission.
The role of allogeneic transplantation in AML patients with mutated NPM1 and a concomitant FLT3-ITD remains controversial.107 As the mutant NPM1 might be recognized by immune cells,110 it would be interesting to explore whether it elicits a specific GvL effect. It should be added that detection of minimal residual disease may emerge in the future as an important assay for assessing quality of therapy response and for guiding therapy.102, 111
In patients >60 years old, the favourable impact of NPM1 mutations may be less important.112 A recent report113 suggested that elderly patients may benefit from the addition of all-trans retinoic acid (ATRA) (see below).
Perturbed cellular traffic of nucleophosmin in AML as therapeutic target
The process of developing drugs targeting altered nucleo-cytoplasmic shuttling in cancer cells is hindered by several obstacles, which include (i) specificity (that is, targeting neoplastic but not normal cells) and (ii) technical difficulties in creating small molecules that are able to interfere with protein–protein interactions.16
One example of specificity-related problems are inhibitors of karyopherin-B family members, such as leptomycin B, a natural product that binds covalently and irreversibly inhibits Crm1, the export receptor for NES-containing proteins.114, 115 Another problem with Crm1 inhibitors is how to re-direct the NPM1 mutant straight to the nucleolus, as one or both C-terminus tryptophans are missing. Indeed, incubation of transfected or NPMc+ AML cells with leptomycin B causes relocalization of the mutant only in the nucleoplasm (Figure 2, bottom, c). A third obstacle to the use of leptomycin B in therapy is its toxicity, which is hardly surprising as Crm1 is critical for nuclear export of most proteins and RNAs in normal and neoplastic cells. Indeed, phase I clinical trials with leptomycin B were associated with several adverse side effects, particularly profound anorexia and malaise.116 Generation of less toxic Crm1 inhibitors with improved therapeutic windows could overcome these problems.117
A more rational approach involves developing small molecules to re-direct wrongly located proteins, like mutated NPM1, to the proper cellular compartment (that is, the nucleolus). Designing small molecules that specifically interfere with NPM1 mutant traffic will require better knowledge of the 3D structure of different nucleophosmin domains.118 Theoretically, in the future, libraries of small interfering molecules might be screened in a high-throughput format16 to search for compounds that, in an appropriate revealing assay, change the subcellular localization of mutated NPM1.
An alternative therapeutic intervention in NPMc+ AML is to interfere with the localization/function of the residual NPM1wt encoded by the normal allele.119 The rationale for this approach is that NPM1 mutations in AML are always heterozygous,58 suggesting that a certain amount of NPM1wt may be necessary for leukaemic cell survival. Indeed, experimental evidence demonstrated that complete deletion of the NPM1 gene in knock-out mice led to death during embryogenesis.3 Therefore, we predict that small molecules blocking residual NPM1wt activity in NPMc+ AML cells (for example, disassembling nucleolar structure and dislocating nucleophosmin in the nucleoplasm) may enhance the propensity of these cells to die or be killed by concomitant administration of chemotherapeutic drugs. As NPM1-mutated AML cells contain much less nucleolar NPM1wt than normal haemopoietic cells, tuning the dose of the small molecules designed to interfere with NPM1wt could also become a strategy to target leukaemic cells more selectively than normal cells.
Another potential strategy is to interfere with NPM1 mutant-induced changes in traffic of various proteins. As an example, NPM1 mutants dislocate HEXIM1 into cytoplasm and the consequent increase in P-TEFb-dependent transcription was advocated as a potential target for P-TEFb inhibitors, such as flavopiridol and CYC202.120 The current list of molecules that are mislocalized by the NPM1 mutants25, 30, 78, 79, 89, 91, 92 may be only the ‘tip of the iceberg’ and the use of mass spectrometry and immunohistochemistry is expected to identify other target proteins and possibly help defining the functional significance of their dislocation. Better understanding of the structural bases of NPM1 heterodimerization with other proteins could help development of molecules able to target this interface, thus preventing mutated NPM1 from delocalizing, and inactivating, potential tumour suppressors. Zhou et al121 generated a NPM-derived peptide fused to the HIV-1 TAT protein and demonstrated that it suppressed leukaemogenesis in mice by interfering with the NF-kB signalling pathways. Stapled peptides are also emerging as promising compounds. They are covalently stabilized helical peptides that circumvent the major problems afflicting most synthetic peptides, that is, loss of secondary structure, decreased stability and difficulty in penetrating intact cells, and have proven to be effective in vitro and in vivo. A BID-derived stapled peptide disrupted the BCL-2/BID interaction, triggering apoptosis and prolonging survival in leukaemic mice.122 A p53-derived stapled peptide disrupted p53-HDM2 interaction, activating a p53-dependent transcriptional response and triggering apoptosis in HDM2 overexpressing cells.123
Finally, it could be feasible to screen a library of pharmacological agents on the OCI-AML3 cell line68 or primary NPMc+ AML cells from patients, in a search for drug sensitivity that is specifically associated with the NPM1 mutation. Schlenk et al113 retrospectively analysed the prognostic impact of NPM1, CEBPA, FLT3 and MLL gene mutations on the effects of ATRA treatment in AML patients >60 years old (excluding APL) who were enrolled in the HD98B randomized trial. Notably, ATRA, given as adjunct to intensive induction therapy (idarubicin, cytarabine and etoposide), significantly improved relapse-free and overall survival only in patients bearing the NPM1 mutation without concomitant FLT3-ITD. On the other hand, the NPM1-mutated/FLT3-ITD negative genotype did not emerge as a predictive marker for response to ATRA in a large series of non-APL patients from the UK MRC AML12 trial.124 Different ATRA administration schedules and/or inclusion of etoposide in the regimen may account for discrepancies. Etoposide in combination with ATRA seems to exert a beneficial synergistic effect in NPM1-mutated/FLT3-ITD negative AML125 possibly because this leukaemia subtype frequently associates with myelomonocytic/monocytic differentiation which is believed to confer particular sensitivity to etoposide. Prospective clinical trials are ongoing to clarify the issue. The molecular mechanism through which ATRA exerts its effects in AML with mutated NPM1 remains elusive. Interestingly, ATRA induced apoptosis (rather than differentiation) of NPMc+ AML cells, and this effect was associated with specific downregulation of the NPM1 mutant.126
Conflict of interest
B Falini applied for a patent on the clinical use of NPM1 mutants.
This work was supported by grants from Associazione Italiana Ricerca Cancro (AIRC) and Fondazione Cassa di Risparmio di Perugia, project code 2007.0099.020. We thank Roberta Pacini and Manola Carini for performing the immunohistochemical studies and Dr Geraldine Boyd for helping editing the manuscript.