The Krüppel-like factor 2 transcription factor gene is recurrently mutated in splenic marginal zone lymphoma.

Splenic marginal zone lymphoma (SMZL) is an indolent B-cell tumor involving the spleen, and is characterized by recurrent deletion of chromosome 7q and biased usage of the immunoglobulin heavy variable (IGHV) allele 1-2*04.1 Genomic studies have partially unraveled the typical SMZL-coding genome, which is characterized by lesions affecting genes involved in the physiological homeostasis of marginal zone (MZ) B cells, including mutations of NOTCH2.2, 3, 4, 5 However, the full spectrum of lesions that contribute to the malignant transformation of SMZL remains unknown. In order to identify novel genetic lesions recurrently associated with SMZL, we extended mutation analysis to genes that, though known to be well-established regulators of MZ B-cell differentiation, had not emerged as significantly mutated from genomic studies of SMZL (Supplementary Tables S1 and S2). This analysis disclosed recurrent mutations of the Krüppel-like factor 2 (KLF2) zinc-finger gene, a transcription factor important for the homeostasis and differentiation of peripheral B-cell subsets,6, 7 in 20% (19/96) of SMZL (Figure 1a). To investigate the full complement of KLF2 mutations in the spectrum of lymphoid neoplasia, we extended the analysis to 547 mature B- and 113 T-cell tumors. In addition to SMZL, KLF2 mutations recurred also in 16% (4/24) of hairy cell leukemia, 11% (17/154) of diffuse large B-cell lymphoma (DLBCL), 9% (5/56) of nodal MZ lymphoma and 8% (5/61) of extranodal MZ lymphoma, whereas they were rare or absent (frequency=0–4%) in the remaining entities (Figure 1a). KLF2 mutations in lymphoid tumors. (a) Prevalence of KLF2 mutations in lymphoid tumors. HCL, hairy cell leukemia; NMZL, nodal marginal zone lymphoma; EMZL, extranodal marginal zone lymphoma; FL, follicular lymphoma; MCL; mantle cell lymphoma; CLL, chronic lymphocytic leukemia; BL, Burkitt lymphoma; MM, multiple myeloma; vHCL, variant hairy cell leukemia; MZL-like MBL, marginal zone-like monoclonal B-cell lymphocytosis; PTCL, peripheral T-cell lymphoma; T-ALL, T-cell acute lymphoblastic leukemia. (b) Prevalence of the various types of non-silent KLF2 mutations identified in lymphoid tumors. (c) Schematic representation of the human KLF2 protein, with its key functional domains and NLS (one within the ZnF domains and the other in a cluster of basic amino acids immediately before the first ZnF). Symbols depict distinct types of KLF2 mutations. Non-silent mutations of confirmed somatic origin are color coded in red. Non-silent mutations of unknown somatic origin are color coded in black. (d) Distribution of the variant allele frequency of KLF2 mutations (blue circles). Full size image Overall, 80 KLF2 mutations were identified (Supplementary Table S3). Of these, 26% (n=21) were truncating mutations (frameshift indels=12; nonsense substitutions=5; splice-site variants=4) (Figure 1b). By complementary DNA analysis, splice-site mutations caused aberrant transcripts that retained intronic sequences and lost their coding potential. On the basis of their distribution along KLF2, truncating mutations were predicted to either disrupt the entire protein or to remove the zinc-finger (ZnF) domains of KLF2, including the putative nuclear localization signal (NLS) sequences (Figure 1c).8, 9, 10 Missense mutations (n=55) and in frame indels (n=4) accounted for 74% of the variants (Figure 1b). According to PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2/), the majority of KLF2 missense substitutions were predicted to be deleterious (Supplementary Table S3). A large fraction of the missense mutations (n=27) clustered in the region encoding the first two ZnFs of KLF2, suggesting that they may affect its nuclear localization, as well as the transcriptional function of the protein (Figure 1c).8, 9, 10, 11 Notably, amino-acid changes recurrently affected evolutionarily conserved codons of the first ZnF, including recurrent substitutions at positions 288 (n=9) and 291 (n=5), which are involved in DNA recognition by KLF2 (Supplementary Figure S1).11 The 5′ basic NLS of KLF2, which maps before the ZnF, was also recurrently targeted by missense substitutions (n=7) (Figure 1c).8, 9, 10, 11 The remaining missense mutations distributed in the N-terminal activation domain (n=7, including three somatic mutations targeting a hotspot at codon 37) and in the inhibitory domain (n=14) (Figure 1c).9, 10 Targeted next generation sequencing analysis revealed that KLF2 mutations (assessable=42) were clonally represented in the tumor (median variant allele frequency=35%) (Figure 1d). Whenever possible, complementary DNA sequencing confirmed that KLF2 mutations (n=21) were always expressed at the transcript level (Supplementary Figure S2). The somatic origin of mutations was confirmed in all tested cases by analysis of paired normal DNA, which was available for most of the variants (n=52, including 36 missense substitutions) (Figure 1c; Supplementary Table S3). Most (72%; 41/57) lymphoma cases harbored one single monoallelic KLF2 mutation. Among the cases displaying more than one mutational event, sequencing analysis of KLF2 transcripts after subcloning demonstrated that the mutations were generally (60%) located on separate alleles. Beside KLF2, no other KLF family member homologous to KLF2, nor genes belonging to the KLF2 pathway were mutated in SMZL (Supplementary Table S1). Copy number abnormalities of the KLF2 locus were examined by a combination of single nucleotide polymorphism array (n=58 samples) and fluorescent in situ hybridization (n=112 samples) in 96 SMZL and 74 DLBCL cases. Monoallelic KLF2 deletions occurred in 11% (11/96) SMZL and in 9% (7/74) DLBCL (Supplementary Table S4). One 174-kb focal loss of the KLF2 locus identified KLF2 as the specific target of the deletion (Supplementary Figure S3). KLF2 mutations and deletions were mutually exclusive in most samples, revealing a predominant monoallelic distribution (Supplementary Figure S3). Thus, when considering both mutations and copy number aberrations in cases, in which both types of alterations could be assessed, 31% (30/96) of SMZL and 26% (19/74) of DLBCL harbored KLF2 structural alterations (that is, mutations or deletions) (Supplementary Figure S3). In SMZL, KLF2 lesions were enriched among the cases harboring 7q deletion (61%, 11/18 vs 22%, 16/72; P=0.0013), NOTCH2 mutations (55%, 10/18 vs 25%, 20/78; P=0.013) and IGHV1-2*04 allele usage (47%, 10/21 vs 27%, 18/66; P=0.082) (Supplementary Figure S4). In DLBCL, KLF2 lesions were restricted to cases showing a non-germinal center phenotype (Supplementary Figure S4). Cell lines and primary tumors carrying KLF2 heterozygous mutations retained expression of the unmutated allele at transcriptional (Supplementary Figure S2) and protein (Supplementary Figure S5) levels. Future experiments will be required to clarify whether KLF2 mutations act as a dominant negative or whether loss of one copy has functional impact. Western blot analysis, after fractionation of cytoplasmic and nuclear proteins, and confocal microscopy analysis were used to determine KLF2 distribution/localization across the subcellular compartments. In the SSK41 cell line, which harbors a wild-type gene, KLF2 was predominantly expressed in the nucleus (Figure 2a). In contrast, the protein was primarily cytoplasmic (Figure 2a) in cell lines harboring mutations that truncated or changed the amino-acid composition of the ZnFs, or targeted codons within or flanking the 5′ basic NLS of KLF2. Cytoplasmic KLF2 expression was also observed in primary tumor cells harboring a monoallelic KLF2 mutation of the ZnF (Figure 2a). The near complete cytoplasmic expression of KLF2 observed in cases with monoallelic mutations warrants investigations aimed at testing whether wild-type KLF2 is sequestered by the mutant protein. Figure 2 Figure 2 NLS mutations of KLF2 cause nuclear displacement of the KLF2 protein. (a) Cell fractionation experiments and confocal microscopy analyses performed in a panel of cell lines with known KLF2 mutation status (A and B represent the two alleles). In the wild-type SSK41 line, KLF2 is preferentially expressed in the nucleus, while the presence of mutations in the VL51, AS283A and OCI-Ly8 cell lines cause a predominant accumulation of the molecule in the cytosolic compartment. Primary cells from a patient with a KLF2 missense mutation (#5141) confirm predominant localization of the mutated protein in the cytosol. The nuclear/cytoplasmic expression ratio was calculated by dividing the mean pixel intensity of the green fluorescence in the nuclear area (defined by the 4',6-diamidino-2-phenylindole (DAPI) blue fluorescence) by the mean pixel intensity of the green fluorescence in the rest of the cell. Confocal images were acquired with a × 63 oil immersion objective. Wt, wild type; WCL, whole-cell lysate; N, nuclear lysate; C, cytoplasmic lysate. (b) Human embryonic kidney 293T cells were transiently transfected with flag-tagged KLF2-GFP constructs and grown on glass coverslips. After 24 h, cells were washed, fixed, stained with anti-flag (red) and counterstained with DAPI (blue). Cells were analyzed using a confocal microscope with a × 63 oil immersion objective. Representative images from three independent experiments. The nuclear/cytoplasmic expression ratio was calculated by dividing the mean pixel intensity of the red fluorescence in the nuclear area (defined by the DAPI blue fluorescence) by the mean pixel intensity of the red fluorescence in the rest of the cell. (c) Luciferase activity of human embryonic kidney 293T cells transfected to express a luciferase reporter driven by the CDKN1A/p21 promoter encompassing the KLF2-binding region together with the flag-tagged wild-type or mutants KLF2 constructs, and with the empty vector as control. KLF2 expression was induced by doxycycline (1 μg/ml) 24 h post transfection. Transactivation activity was detected at 48 h and expressed as relative luciferase intensity compared with wild-type (WT) KLF2 (referred as 100%), in five independent experiments (mean: top of the bar; s.d.: whiskers; P-value by t-test). Untransfected cells (NT) or transfected with an empty vector construct (EV) were used as negative controls. Quantitative real-time PCR and western blot at 48 h, showing comparable expression of the constructs, are also shown. β-tubulin expression is included as protein loading control. Densitometry analysis of five independent western blots shows quantitation of KLF2 levels after normalization over β-tubulin. KLF2 mRNA expression was assessed in five independent experiments and calculated relative to ACTB using the ΔCT method. Full size image To provide a proof of principle of the functional significance of KLF2 mutations, we transfected human embryonic kidney 293T cells with flag-tagged wild-type KLF2 constructs and two mutants: (i) the p.P257* variant, representative of tumor cell mutations that truncate the C-terminal portion of KLF2, including its 5′ basic NLS sequence and the ZnFs; (ii) the p.H288Y variant, the most recurrent substitution affecting a hotspot codon that is physiologically involved in DNA recognition by KLF2 and in its nuclear localization.8, 9, 10 Confocal microscopy showed predominant nuclear localization of wild-type KLF2. In contrast, p.P257* and p.H288Y mutants were displaced from the nucleus (Figure 2b). We then assessed the relative transactivation activity of the mutant KLF2 proteins by measuring their ability to upregulate the expression of a luciferase reporter gene driven by the CDKN1A/p21 promoter, a known direct target of KLF2,12, 13 in human embryonic kidney 293T cells that lack endogenous KLF2 (Supplementary Figure S6). Expression of wild-type KLF2, but not mutant, strongly induced the reporter (Figure 2c), a finding consistent with the predicted loss of interaction between KLF2 mutants and DNA. KLF2 mutations represent one of the most frequent genomic abnormalities of SMZL, and unravel the disruption in human tumors of a previously unidentified molecular pathway. Given the involvement of KLF2 in the transcriptional regulation of an array of genes, it is difficult to predict which cellular targets/pathways may be critically affected by its mutations in lymphomagenesis. In normal lymphocytes, KLF2 binds the promoter and regulates the expression of genes involved in cell cycle/apoptosis (CDKN1A/p21) and cell trafficking (S1PR1, SELL/CD62L, ITGB7/β7-integrin and CXCR5).6, 7, 12, 13 A limited set of KLF2 domains is necessary to exploit its transcriptional activities, including three highly conserved ZnFs, which allow protein contact with DNA, and two potent, independent NLS that mediate the localization of KLF2 in the nucleus.8, 9, 10, 11 Consistent with the critical roles of the ZnFs and the NLS, ~60% of KLF2 mutations disrupt or are predicted to modify these structures of the KLF2 protein. Our data confirm that KLF2 mutants lacking the NLS and some of those harboring missense substitutions of the first ZnF (codon 288 substitutions) are displaced from the nucleus and transcriptionally defective. Due to the large number of somatic missense KLF2 mutations disclosed by our molecular investigations, additional analyses are needed to comprehensively characterize the effects of the full spectrum of these variants. According to their distribution and predicted functional consequences, it is conceivable that, in addition to nuclear localization, missense substitutions might also disturb other KLF2 functions, including DNA binding, transcriptional activation or protein interactions. As is the case for most cancer-associated genetic lesions, KLF2 inactivation may not be sufficient for malignant transformation. In fact, transgenic mice engineered to lack KLF2 in mature B cells display an expansion of the MZ at the expense of the follicular compartment, but do not develop lymphoma.6, 7 It is important to note that, however, lymphoma development may require longer times than those observed so far in mice,6, 7 in line with the indolent course of SMZL and the elderly age of patients affected by this lymphoma. Consistent with a multistep process of lymphomagenesis, KLF2 lesions frequently co-occur with IGHV1-2*04 usage, NOTCH2 mutations and 7q deletion in SMZL, suggesting a possible cooperation between genetic abnormalities and B-cell receptor configuration in promoting transformation. Emerging evidence points to the role of KLF transcription factors in human cancers.14 Although KLF family gene expression and function are altered in a large number of neoplasms, including B-cell tumors, recurrent structural alterations of these genes are exceedingly rare and sporadically restricted to solid cancers.14, 15 Our data provide the first evidence of the molecular deregulation of KLF family genes in hematologic malignancies, and suggest that selection of KLF2 mutations has a role in transformation common to several lymphoma subtypes.