Altered Nuclear Export Signal Recognition as a Driver of Oncogenesis

Proper nucleocytoplasmic partitioning of macromolecules is essential to cellular homeostasis and is regulated by the nuclear pore complex and transport proteins that mediate trafficking of molecules across the nuclear envelope. Exportin-1 (XPO1; previously known as CRM1) is the major nuclear export receptor for proteins larger than approximately 40 kDa. To perform its exporting function, XPO1 identifies signals known as nuclear export signals (NES) embedded in the amino-acid sequence of cargo proteins. Protein cargoes bind XPO1 in the nucleus cooperatively with the small GTPase Ran, and the formation of ternary XPO1–RanGTP–cargo complexes begins the export process (1–3).

Numerous cellular pathways are regulated by controlling the subcellular localization of intermediate signaling proteins. Mutations that alter this nucleocytoplasmic localization in cis have been shown to drive transformation in a number of cancers (4–8). For example, mutations in nucleophosmin 1 (NPM1) that generate an NES and result in aberrant NPM1 cytoplasmic localization are among the most common genetic alterations in acute myeloid leukemia (9, 10). Similarly, mutations generating or eliminating an NES in BRCA1 (11), VHL (12), and other recurrently mutated genes in cancer have been described. In addition, alterations in the machinery regulating nuclear export globally have been observed in cancer. Notably, increased expression of XPO1 has been described in numerous cancers (13–16), and many clinical trials have studied the efficacy of drugs inhibiting XPO1 in patients with cancer (16–21).

Despite interest in XPO1 as a therapeutic target in cancer, mechanistic evidence for cancer-specific derangements in XPO1 function has not been demonstrated. Mutations in XPO1 occur in a variety of cancers (22–25), but the functional contributions of these mutations to tumorigenesis are also unknown. Further, there are no biomarkers of response to XPO1 inhibitors in advanced-phase clinical trials. Given that XPO1 is an essential protein in eukaryotic cells, defining biomarkers of response may define patients most likely to respond to these agents and potentially limit toxicity. Here, we find that XPO1 mutations are positively selected for in a variety of cancers and sensitize cells to XPO1 inhibitors. Specific mutant residues in XPO1 directly promote malignant transformation by altering the ability of XPO1 to engage protein cargo for nuclear export. These data identify a novel model of oncogenesis: tumorigenesis driven by mutations that change the global distribution of proteins across the nuclear membrane. In addition, these data implicate mutant XPO1, as well as its consequently mistrafficked protein cargo, as targets for anticancer therapy.

Recent genomic analyses have identified recurrent heterozygous XPO1^E571 mutations in a variety of cancer types, including chronic lymphocytic leukemia (CLL; refs. 22, 23), Hodgkin lymphoma (24), and esophageal carcinoma (25), among others. To systematically quantify the frequency and significance of XPO1 mutational hotspots, we performed a large-scale analysis of whole-exome and genome sequencing data from 42,793 patients with cancer spanning 322 cancer types, using an extensively validated computational framework (26, 27). We identified highly recurrent and statistically significant hotspots in XPO1 at E571 (n = 99 patients; q value =10⁻²⁰⁴), D624 (n = 6 patients; q value = 10⁻¹¹), and R749 (n =21 patients; q value = 10⁻²²) residues (Fig. 1A), the latter two hotspots being previously unidentified. In addition, we found remarkable lineage specificity of XPO1 hotspot mutations. E571K, the most common XPO1 mutation, was enriched in B-cell lymphomas, including primary mediastinal B-cell lymphoma (PMBL; 33%), classic Hodgkin lymphoma (14%), and diffuse large B-cell lymphoma (DLBCL; 2%), and in CLL (3%; Fig. 1B and C; Supplementary Fig. S1A). In contrast, XPO1 D624 mutations were found only in CLL, whereas those at R749 were found mostly in nonlymphoid solid-tumor malignancies (Supplementary Table S1). Strikingly, XPO1 mutations were found only in the heterozygous state and never occurred as homozygous or hemizygous (i.e., LOH) mutations. We confirmed this in a separate cohort of 2,877 patients with hematologic malignancies in which we analyzed both mutations and copy-number state of XPO1. We found only a single patient who had both mutation and copy-number alteration in XPO1 (a gain of mutant XPO1). Even in this case, the wild-type (WT) XPO1 allele was retained (Supplementary Fig. S1B). These findings altogether suggest a novel change of function in XPO1 caused by mutation.

The recurrence, more often than expected in the absence of selection, of mutations at specific amino-acid residues in XPO1 suggests that XPO1 hotspot mutations functionally promote B-cell malignancies and other cancers. To evaluate the functional effects of XPO1 mutations, we used genome editing to generate isogenic B-cell leukemia cells (NALM-6 cells; ref. 28) with physiologic expression of the XPO1^E571K mutation from the endogenous XPO1 locus (Supplementary Fig. S1C). There was an equal ratio of XPO1^E571K mutant allele to XPO1^WT allele at the level of genomic DNA and mRNA (50% variant allele frequency). Expression of the heterozygous XPO1^E571K/WT mutation in NALM-6 cells resulted in increased proliferation in vitro in both noncompetitive and competitive growth assays relative to XPO1^WT counterparts (Supplementary Fig. S1D–S1F). Enhanced growth of XPO1^E571K/WT mutant cells was also seen in vivo through xenotransplantation of equal numbers of luciferase-labeled XPO1^E571K/WT mutant or XPO1^WT/WT cells (Fig. 1D and E; Supplementary Fig. S1G). The effects of XPO1 mutation were distinct from XPO1 genetic depletion, which completely suppressed cell growth (Supplementary Fig. S1D), a finding consistent with the known essentiality of XPO1 for cell survival (29, 30).

Given the high conservation of XPO1 across mammals (Supplementary Fig. S2A), we next generated a conditional knock-in mouse model to express the Xpo1^E571K mutation from the endogenous Xpo1 locus to determine its in vivo effects (Supplementary Fig. S2B and S2C). Heterozygous expression of Xpo1^E571K in the B-cell compartment using CD19-Cre Xpo1^E571K/WT mice resulted in selective expression of the Xpo1^E571K mutation in B cells at approximately 50% allelic ratio and had no effect on XPO1 protein expression in B cells relative to Cre-negative Xpo1^E571K/WT controls (Supplementary Fig. S2D–S2F).

CD19-Cre Xpo1^E571K/WT mice were born at expected Mendelian ratios (Supplementary Fig. S3A) but were smaller in body weight relative to littermate controls (Supplementary Fig. S3B). Despite the smaller body size, CD19-Cre Xpo1^E571K/WT mice had significantly larger spleens at 12 weeks of age (Fig. 1F). Colony-forming assays in IL7 using bone marrow (BM) from mice with tamoxifen-inducible pan-hematopoietic Xpo1^E571K/WT expression (CAG-CreERT Xpo1^E571K/WT mice) revealed increased numbers of B-cell colonies with enhanced clonogenicity upon replating (Fig. 1G), a finding restricted to the B-cell compartment (Supplementary Fig. S3C) and lost in the absence of IL7 supplementation (Supplementary Fig. S3D). CD19-Cre Xpo1^E571K/WT mice died more rapidly than littermate controls and had increased B-cell proliferation, B-cell lymphocytosis, anemia, and thrombocytopenia (Fig. 2A; Supplementary Fig. S3E–S3H), and had increased numbers of mature (CD19⁺ IgM⁺ IgD⁺) B cells in the BM and transitional T1 and T2 B cells (CD19⁺ CD21^lo IgM^hi CD93⁺ CD23⁻ and CD19⁺ CD21^lo IgM^hi CD93⁺ CD23⁺, respectively) in the spleen as well as splenomegaly (Fig. 2B–D). CD19-Cre Xpo1^E571K/WT mice developed greater splenomegaly (Fig. 2E) as a result of increased generation of germinal centers and germinal center (B220⁺ FAS⁺ GL7⁺) B cells upon alloimmunization with sheep red blood cells (SRBC), an effect that was evident 7 days after a single SRBC injection (Fig. 2F; Supplementary Fig. S3I and S3J).

A portion of CD19-Cre Xpo1^E571K/WT mice developed a lethal B-cell disease infiltrating lymph nodes, spleen, lungs, and BM, with coexpression of CD5 and CD23, characteristic features of human CLL, a cancer type frequently carrying XPO1^E571 mutations (Fig. 2G; Supplementary Fig. S4A–S4E). Next-generation sequencing of VDJ sequences from splenic mononuclear cells of these mice at the time of death identified productive clonal sequences occupying >98% of the total reads, indicating the presence of a highly clonal B-cell malignancy (Supplementary Fig. S4F).

The above data reveal that expression of cancer-associated XPO1 mutations can initiate lymphomagenesis when the correct microenvironmental stimulus is present (alloantigen from SRBC immunization). However, more aggressive lymphomas featuring XPO1^E571K mutations may require additional cooperating mutations to fully manifest the disease phenotype. To evaluate this hypothesis, we crossed Xpo1^E571K/WT mice to mice with overexpression of either c-MYC (using the Eμ-Myc mouse model; ref. 31) or BCL2 (using Vav-BCL2 transgenic mice; ref. 32) to model the aggressive lymphomas marked by overexpression of either of these oncogenic proteins in addition to XPO1 mutations.

First, we crossed Eμ-Myc mice (which express human c-MYC transgenically from the immunoglobulin locus) to CAG-CreERT Xpo1^E571K/WT mice, allowing induction of mutant XPO1 only after exposure to tamoxifen (Fig. 3A; Supplementary Fig. S5A). Mice transplanted with Eμ-Myc–driven lymphomas conditionally expressing Xpo1^E571K/WT in hematopoietic cells following engraftment (CAG-CreERT Xpo1^E571K/WT Eμ-Myc mice exposed to tamoxifen) succumbed to lethal disease more rapidly than controls (CAG-CreERT Xpo1^E571K/WT Eμ-Myc mice exposed to vehicle; Fig. 3B and C). Expression of Xpo1^E571K/WTfurther increased B-cell lymphocytosis driven by Eμ-Myc throughout the blood, spleen, and BM, resulting in more rapidly lethal spleen and BM infiltration (Fig. 3D–F). Here again, sequencing of VDJ sequences from splenic mononuclear cells confirmed that death was due to infiltration of B cells which were >90% clonal (Supplementary Fig. S5B). Notably, a subset of the B cells gained aberrant CD5 expression (Supplementary Fig. S5C), and Eμ-Myc CAG-CreERT Xpo1^E571K/WT mutant cells were larger as measured by flow cytometry as well as cytologic analysis with Wright–Giemsa stain compared with controls (Supplementary Fig. S5D and S5E). Whole transcriptomic analysis by mRNA sequencing (RNA-seq) of tumors from CAG-CreERT Xpo1^E571K/WT Eμ-Myc and CAG-CreERT Eμ-Myc controls revealed that Xpo1^E571K expression in the context of Eμ-Myc tumors increased expression of established MYC targets as well as gene expression programs related to cell cycle, nuclear export, and NFκB activation (Fig. 3G).

Next, we crossed Vav-BCL2 transgenic mice (which overexpress human BCL2 throughout the hematopoietic system) to CD19-Cre Xpo1^E571K/WT mice (Fig. 4A). CD19-Cre Xpo1^E571K/WTVav-BCL2 mice had increased peripheral blood leukocytosis, splenomegaly, and a greater absolute number of B cells in the blood, BM, and spleen compared with CD19-Cre Vav-BCL2 controls (Fig. 4B; Supplementary Fig. S6A and S6B). In vivo bromodeoxyuridine (BrdU) uptake assays also confirmed that mice with combined Xpo1 mutation and BCL2 overexpression had greater splenomegaly and higher numbers and proliferation of BM and splenic B cells relative to controls (Fig. 4C and D; Supplementary Fig. S6C). In vitro proliferation and self-renewal (measured by colony-forming assays in IL7) was also greater in CD19-Cre Xpo1^E571K/WT Vav-BCL2 than in CD19-Cre Xpo1^WT/WTVav-BCL2 mice (Fig. 4E). Overall, CD19-Cre Xpo1^E571K/WT Vav-BCL2 mice had increased B-cell invasion in the spleen, BM, and lungs (Fig. 4F) and succumbed to lethal B-cell malignancy sooner than controls (Supplementary Fig. S6D). Gene set enrichment analysis of genes differentially expressed between CD19-Cre Xpo1^E571K/WT Vav-BCL2 and CD19-Cre Xpo1^WT/WT Vav-BCL2 mice was significant for cytokine receptor signaling via NFκB, a pattern that was seen in the CD19-Cre Xpo1^E571K/WT versus WT mice in the absence of BCL2 (Fig. 4G).

To evaluate how mutations in XPO1 might drive lymphoid malignancy development, we next sought to identify patterns of mutational co-occurrence or exclusivity in patients bearing XPO1^E571K mutations. In an additional cohort of 2,877 patients with hematologic malignancies who were uniformly sequenced by the MSK HemePACT assay (33), we identified 29 XPO1-mutant patients from patients with CLL (n = 288), Hodgkin lymphoma (n = 16), PMBL (n = 9), B-cell acute lymphoblastic leukemia (n = 142), or DLBCL (n = 284). Although we identified numerous genetic alterations coexisting with XPO1 mutations in each of these histologic subtypes of B-cell malignancies (Supplementary Fig. S7A–S7D), there were no statistically different mutational co-occurrences between XPO1-mutant versus WT cases in each of these disease categories. We also similarly sequenced a cohort of splenocytes at time of death from 3 CD19-Cre Xpo1^E571K/WT mice and 3 CD19-Cre Xpo1^WT/WT mice of similar age (using a sequencing panel targeting the entire coding region of the mouse orthologs of 578 genes recurrently mutated in solid tumors, lymphomas, and leukemias; Supplementary Table S2). This revealed several somatic mutations in 1 of the 3 CD19-Cre Xpo1^E571K/WTmice including compound p53 mutations (Trp53^Y120S, Trp53^M157V), an Msh2 splice-site mutation (Msh2 c.793-8A>T), and a missense mutation in Tcf7l2 (Tcf7l2^S512L), a transcription factor through which β-catenin functions.

Given the known role of XPO1 in nuclear export, we next took an unbiased proteomics approach to detect pathways and individual XPO1 cargoes affected by the E571K mutation. We quantified the partitioning of proteins between the cytoplasm and nucleus of human B-cell malignant cells (NALM-6 cells) with or without knock-in of XPO1^E571K/WT. We performed stable isotope labeling with amino acids in cell culture (SILAC)–based mass spectrometry on fractionated nuclear and cytoplasmic lysates from XPO1^E571K/WT and XPO1^WT/WT NALM-6 cells (Fig. 5A). Several proteins were identified as significantly differentially exported in XPO1^E571K/WTversus XPO1^WT/WT cells [fold change ratios >|2| and −log₁₀ (P value) ≥ 1.30; Fig. 5B; Supplementary Tables S3–S5], including numerous members of the K63-ubiquitination, TLR4, and NFκB pathways, among others (Fig. 5C).

Phenotypically, XPO1^E571K/WT cells demonstrated higher baseline and postactivation NFκB and NFAT transcriptional responses to stimulation with TNFα or PMA/ionomycin, respectively, than XPO1^WT/WT cells (Fig. 5D and E). Differential gene-expression analysis of mRNA sequencing of pro-B (CD19⁺ CD43⁺ CD24⁺ BP-1⁻), pre-B (CD19⁺ CD43⁻ IgM⁻ IgD⁻), immature (CD19⁺ CD43⁻ IgM⁺ IgD⁻), and mature B cells (CD19⁺ CD43⁻ IgM⁺ IgD⁺) from CD19-Cre Xpo1^E571K/WTand littermate CD19-Cre Xpo1^WT/WT control mice identified prominent upregulation of pathways associated with cytokine signaling and immune responses in Xpo1^E571K/WT mutant mature B cells relative to controls (Fig. 5F; Supplementary Fig. S8A and S8B). Consistent with this finding, serum from 12-week-old CD19-Cre Xpo1^WT/WT and CD19-Cre Xpo1^E571K/WT mice contained increased IgM autoantibodies to histone H1, a finding associated with dysregulated immune signaling (Supplementary Fig. S8C). Moreover, we identified a similar transcriptional signature of increased cytokine immune response and inflammation in patients with human CLL with XPO1 mutations compared with patients with CLL without XPO1 mutations (Fig. 5G; Supplementary Table S6). Differential export of NFκB and NFAT signaling proteins was confirmed by Western blot, which revealed enhanced nuclear phosphorylated p65 and NFAT2 in XPO1 mutants relative to WT (Fig. 5H and I). Other proteins whose subcellular localization was altered by XPO1^E571K/WT in our SILAC experiment were confirmed by Western blot, including TRAF2 (increased in nuclear export in the setting of mutant XPO1^E571K) and p120 catenin (which was decreased in nuclear export in the setting of mutant XPO1^E571K; Supplementary Fig. S8D). Finally, the increased NFκB and NFAT signaling and differential export of their member proteins by XPO1^E571K/WT were validated in a CLL cell line (MEC1) engineered to express WT or E571K-mutant XPO1 as well as a series of mature B-cell lymphoma cell lines naturally carrying endogenous XPO1^E571K mutations (Supplementary Fig. S8D–S8H).

XPO1 binds to its cargo by recognizing 8- to 15-residue amino-acid motifs known as leucine-rich NESs (1–3, 34), which show a characteristic spacing of hydrophobic residues separated by other amino acids (described as Φ¹-X_(2,3) − Φ²-X_(2,3) − Φ³-X-Φ⁴, where “Φ” represents leucine, isoleucine, valine, phenylalanine, or methionine). Atomic-level characterization of the XPO1/NES interaction has revealed that XPO1 binds NES of cargo proteins in a hydrophobic cleft on the outer convex surface of XPO1 (35–38). We therefore modeled the position of XPO1 hotspot mutations on the three-dimensional structure of XPO1 to understand the mechanism by which these mutations alter nuclear export of the proteins we identified. Interestingly, the two hotspot mutations in XPO1 arising in CLL and B-cell lymphomas (D624 and E571), although separated in two-dimensional amino-acid sequence, were both adjacent in the cognate folded protein structure of the NES binding cleft of XPO1 (Fig. 6A). Further structural modeling predicted that the change of a negatively charged glutamic acid at XPO1^E571 to a positively charged lysine would promote XPO1 interaction with protein cargo bearing negatively charged amino-acid residues C-terminal to the NES (Fig. 6B). Simultaneously, XPO1^E571K mutation would be predicted to repel interactions with proteins bearing positively charged amino-acid residues C-terminal to the NES.

As described above, analysis of our SILAC-based mass spectrometry data identified a number of NES-bearing proteins as depleted in the nucleus and/or enriched in the cytoplasm of XPO1^E571K/WT mutant versus XPO1^WT/WT cells beyond NFκB and NFAT proteins. Analysis of the amino-acid sequence of these proteins with an NES prediction tool (39) identified several putative NESs in these proteins in addition to the known NESs. In proteins showing enhanced export by mutant XPO1, the proportions of predicted NESs with a negative net charge in their C-terminal end were more frequent (Fig. 6C and D). Conversely, proteins showing reduced export by mutant XPO1 more frequently had NESs with a net positive charge in their C-terminal end.

We utilized a cellular reporter containing one of the NESs to evaluate the sequence-specific effects of XPO1 hotspot mutations on nuclear export, which were suggested by the structural and proteomic analyses above. We used a modified version of a previously described nuclear-export reporter (40) where the first 100 amino acids of the XPO1 substrate survivin (bearing its NES) are fused to a 3× FLAG epitope and two nuclear localization signals (Supplementary Fig. S9A). We generated three versions of this reporter: (1) the native NES, (2) a mutant NES where amino-acid residues C-terminal to the last hydrophobic residue were enriched in positive charges, and (3) a mutant version where amino-acid residues C-terminal to the last hydrophobic residue were enriched in negative charges. Cotransfection of each of these 3 NES reporters with either WT or E571K-mutant XPO1 revealed that the XPO1^E571K mutant significantly enhanced export of the negatively charged NES construct, while reducing export of the positively charged NES construct (Fig. 6E–G; Supplementary Fig. S9B and S9C). We validated these findings with a second reporter based on the HIV Rev protein NES [the Rev(1.4)-GFP reporter (41) with insertion of the IκBα NES] with an additional step of adding cycloheximide to inhibit new protein translation in the cytoplasm (Supplementary Fig. S9D). This experiment once again revealed that the XPO1^E571K mutant greatly enhanced nuclear export of negatively charged NES. Overall, these experiments confirmed that XPO1^E571K alters nuclear export on the basis of the charge properties of amino-acid residues surrounding the NES of XPO1 cargo proteins (Supplementary Fig. S9E and S9F). Interestingly, the XPO1^D624G hotspot mutation was predicted to disrupt hydrogen bonding between XPO1 and cargo protein at a site farther removed from the NES (Supplementary Fig. S10A). Consistent with this prediction, the XPO1^D624G mutant decreased export compared with XPO1^WT, a finding which was not altered by changing the C-terminal sequence of the NES (Supplementary Fig. S10B and S10C).

Inhibitors of nuclear export that bind to the C528 of XPO1 and occupy the NES-binding pocket of XPO1, thereby blocking XPO1 function, have recently been tested in phase II and III clinical trials. The XPO1 inhibitor selinexor recently received Orphan Drug and Fast Track FDA designation for relapsed multiple myeloma. However, biomarkers of response to XPO1 inhibition are not known, and it is not known if cancer-associated mutations in XPO1 modify response to XPO1 inhibitors. Interestingly, leukemia or lymphoma cells with naturally occurring XPO1^E571K mutations were among the most sensitive to the XPO1 inhibitor KPT-185 across 75 cell lines in publicly available cell line dependency data (depmap.org; Supplementary Fig. S11A). This finding was also seen in XPO1^E571K mutant B-cell lymphoma cell lines exposed to selinexor compared with those WT for XPO1 (Supplementary Fig. S11B), where XPO1^E571K mutant cell lines were all more sensitive to selinexor compared with WT cells. The enhanced sensitivity of XPO1^E571K mutant cells to selinexor was further seen in isogenic human B cells (NALM-6 cells), B cells from CD19-Cre Xpo1^E571K/WT knock-in mice, and mice with combined Xpo1^E571K mutation and BCL2 overexpression relative to XPO1^WT counterparts in vitro (Fig. 7A–C). Similar enhanced sensitivity of XPO1 mutant cells to selinexor was seen in vivo: Selinexor extended the survival of mice bearing Eμ-Myc Xpo1^E571K mutant lymphomas compared with vehicle to a far greater extent than mice bearing Eμ-Myc Xpo1^WT lymphomas (Fig. 7D; Supplementary Fig. S11C). Selinexor also decreased the spleen size and percentage of CD5⁺ B cells in the spleens of secondary recipients engrafted with CLL cells from CD19-Cre Xpo1^E571K/WTmice (Supplementary Fig. S11D–S11F).

To evaluate the basis for preferential effects of XPO1 inhibition on cells bearing cancer-associated XPO1 hotspot mutations, we next assessed XPO1 protein abundance and nuclear/cytoplasmic protein compartmentalization by XPO1 mutant genotype in the context of XPO1 inhibition. Interestingly, selinexor treatment resulted in enhanced dose-dependent degradation of XPO1 in XPO1 mutants relative to WT cells (Fig. 7E). The close proximity of the XPO1^E571 residue to the known binding residue of selinexor on XPO1 at C528 (Supplementary Fig. 11G) suggested that the XPO1^E571K mutation might influence the binding affinity of XPO1 inhibitors to the mutant XPO1 protein. Consistent with this prediction, microscale thermophoresis assays with the XPO1 inhibitor KPT-185 identified that the XPO1^E571K mutant protein bound KPT-185 with greater affinity than XPO1^WT protein (Fig. 7F; Supplementary Fig. S11H). Within cells, XPO1 inhibition resulted in preferential degradation of the XPO1^E571K protein as identified in MEC1 cells expressing XPO1^WT or XPO1^E571K FLAG-tagged protein treated with selinexor (Supplementary Fig. S12A). Finally, proteomic analysis by tandem mass tag mass spectrometry of cells treated with XPO1 inhibitor compared with vehicle also revealed mutant-selective changes in the localization of many proteins after XPO1 inhibition (Supplementary Fig. S12B and S12C). Certain proteins, including p53, although retained in the nucleus upon XPO1 inhibition in both WT and mutant cells, were retained at higher levels in XPO1^E571K mutant cells (Supplementary Fig. S12D). However, other proteins had increased nuclear retention upon XPO1 inhibition only in XPO1^E571K cells, suggesting potential synthetic lethal interactions. These proteins included members of critical cell pathways such as cell cycle, translation, mRNA splicing, and innate immune signaling (Fig. 7G).

Our data present a novel model of tumorigenesis: change-of-function mutations in nucleocytoplasmic trafficking machinery driving transformation. Systematic quantification of XPO1 mutations across large-scale sequencing data from numerous cancer types enabled the identification of specific hotspot mutations in XPO1 as among the most common mutations in patients with CLL, Hodgkin lymphoma, or PMBL, showing lineage specificity and occurring as heterozygous point mutations. Prior work suggests that XPO1 mutations in CLL are clonal at diagnosis and likely to be present at relapse as well, suggesting an early founder mutation (23). Consistent with this notion, XPO1^E571 mutations have been tracked in plasma cell-free DNA in patients with Hodgkin lymphoma (24), where their presence correlates with tumor regression or progression, and the frequency of XPO1 mutations in CLL does not appear to be significantly different between the treatment-naïve or relapsed/refractory state (42).

Although the discovery of hotspot mutations in XPO1 suggests an oncogenic driver function, XPO1 is expressed in every eukaryotic cell and is essential for cell survival. The essential housekeeping function of XPO1 has presented a barrier to manipulating endogenous XPO1 in vivo. Consequently, there has never been an animal model of any alteration in XPO1 previously. Here, we overcame this technical challenge and identified that a single amino acid–residue substitution in XPO1 in vivo increased B-cell proliferation and resulted in the development of clonal B-cell malignancies resembling human CLL. These models thereby provide one of the few genetically accurate murine models of CLL and establish the causality of XPO1 mutations in cancer development. Moreover, expression of the XPO1^E571K mutant in vivo also promoted development of other types of B-cell malignancies in concert with oncogenes such as MYC and BCL2.

The biochemical, structural, proteomic, and molecular studies here indicate that XPO1 mutations alter NES recognition in a sequence-specific manner, favoring the export of cargoes with negatively charged C-terminal NES sequences as a result of changes in charge near the NES binding cleft of XPO1. The generation of genetic models of mutant XPO1 enabled us to characterize the malignant mutant XPO1 exportome. Hundreds of proteins in various pathways were affected by altered NES recognition by mutant XPO1, including changes in the nucleocytoplasmic distribution of proteins involved in inflammatory signaling, DNA repair, RNA export, and chromatin remodeling pathways. Although it might be expected that the abundance of individual proteins detected in nuclear and cytoplasmic fractions would yield symmetrical results, with proteins enriched in nuclear analysis showing up as depleted in the cytoplasmic fraction and vice versa, this was not seen here or in prior analyses (43–45). In contrast, the vast majority of shuttling proteins assayed by mass spectrometric analysis of nuclear and cytoplasmic fractions show changes in only one cellular compartment without the expected reciprocal change in the other compartment. Potential explanations for this lack of overlap include the fact that many proteins are exported from the nucleus as part of their regulation and are degraded after nuclear export. In addition, the balance between nuclear export and import is important, and changing nuclear localization could result in changes in protein stability.

Drugs that inhibit XPO1 binding to cargo proteins have been demonstrated to have therapeutic efficacy for a number of cancer types (18–21, 46). As a result, altered function of XPO1 has been suggested to be common in cancer, but there has been no demonstration of XPO1 driving tumor development previously. The genomic data from patients as well as these functional data show that XPO1 mutations are selected for in cancer cells and promote malignant transformation by altering activation of pathways such as NFκB and NFAT signaling. This discovery that mutant XPO1 can functionally drive tumorigenesis may be particularly useful for the development of targeted therapies for malignancies bearing XPO1 hotspot mutations; indeed, XPO1^E571K was associated with increased sensitivity to XPO1 inhibitors currently in clinical trials. Thus, these data establish oncogenic mutations in nuclear export machinery and identify these mutations as novel drivers of tumorigenesis and potential targets of anticancer therapy.

See also Supplementary Materials for additional details.

All animals were housed at Memorial Sloan Kettering Cancer Center (MSKCC). All animal procedures were completed in accordance with the Guidelines for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committees at MSKCC. Generation and genotyping of the Xpo1^E571K/WT are described below.

Xpo1^E571K/WT mice on a pure C57BL/6 background were crossed to CD19-Cre mice (47), tamoxifen-inducible CAG-CreERT mice (48), Eμ-Myc transgenic mice (31), and/or Vav-BCL2 transgenic mice (32), each of which was also maintained on a pure C57BL/6 background.

Eight-week-old female CD45.1 C57BL/6J mice (The Jackson Laboratory) were used as recipients for BM transplantation assays, and 8-week-old female NSG mice (The Jackson Laboratory) were used for cell-line xenografting. Blood was collected by submandibular bleeding using heparinized microhematocrit capillary tubes (Thermo Fisher Scientific). For assessment of germinal center formation, age- and sex-matched C57BL/6J mice were immunized intraperitoneally at 8 to 12 weeks old with 0.5 mL of a 2% SRBC suspension in PBS (Cocalico Biologicals) or PBS as control. Automated peripheral blood counts were obtained using a ProCyte Dx Hematology Analyzer (IDEXX).

An 8.4 kb genomic DNA used to construct the targeting vector was first subcloned from a positively identified B6 BAC clone (RP23: 123E19). The region was designed such that the long homology arm (LA) extends approximately 5.73 kb 5′ to the engineered LoxP site, and the short homology arm (SA) extends 2.15 kb 3′ to the insertion of the inverted mutant exon 26. The mutant exon 16 (GAA→AAA) plus the flanking genomic sequences for correct splicing (saEx26*sd) was inserted in the reverse direction in intron 17–18 and is 153 bp downstream of exon 16, which is immediately followed by a Lox66 site. A Lox71 site was placed 5′ to the inverted saEx16*sd sequence. The FRT-flanked neo cassette was inserted immediately upstream of the Lox71/Lox66–flanked inversion sequence. The targeting vector was confirmed by restriction analysis and sequencing after each modification. The boundaries of the 2 homology arms were confirmed by sequencing with P6 and T73 primers that read through both sides of the backbone vector into the genomic sequence. The FRT-neo cassette was confirmed by sequencing with LAN1 and iNeoN2 primers that read from the 5′- and 3′-ends of the neo cassette, respectively, into the inverted saEx16*sd (LAN1) and the genomic sequence (iNeoN2). saEx16*sd and its junction with the genomic sequence were sequencing confirmed with primer LAN1. LOX1 sequencing result confirmed the engineered single LoxP site upstream of exon 16.

The targeting vector was confirmed by restriction analysis and sequencing after each modification step. The boundaries of the 2 homology arms were confirmed by sequencing with P6 and T73 primers that read through both sides of the backbone vector into the genomic sequences. LAN1 and N2 primers read from the selection cassette into the inversion cassette (LAN1) and the 3′ end of the middle arm (N2). The 5′ LoxP site and the 5′ junction of genomic sequence/eGFP was confirmed by sequencing with Lox1 primer. Primer GFP3 sequencing confirmed the remaining eGFP-T2A sequence and its 3′ junction with the genomic sequence. Primers LAN1, mCherrySQ1, and IMBPSQ1 confirmed the entire inversion cassette sequence and its junctions.

Primers used for sequencing:

The targeting construct was linearized using AscI prior to electroporation into embryonic stem (ES) cells. After selection with G418 antibiotic, surviving clones were expanded for PCR analysis to identify recombinant ES clones. Secondary confirmation of positive clones identified by PCR was performed by Southern blotting analysis. DNA was digested with XbaI (external long arm) or EcoRV (internal short arm), and electrophoretically separated on a 0.8% agarose gel. After transfer to a nylon membrane, the digested DNA was hybridized with a probe targeted against the 5′ external region. DNA from a normal C57BL/6 (B6) mouse was used as WT control. Positive C57BL/6 FLP ES cells were expanded and microinjected into BALB/c blastocysts. Resulting chimeras with a high percentage black coat color were mated to C57BL/6J mice (The Jackson Laboratory) to remove the FLP transgene, and subsequently to CD19-Cre or CAG-CreERT transgenic mice. Mice were backcrossed for six generations to C57BL/6 mice, and tamoxifen was administered to 6-week-old mice at 12 μg/g intraperitoneally every other day for 3 doses (as previously described; ref. 49). The XPO1^E571K mice are available at The Jackson Laboratory as Stock No. 033957.

Tail DNA samples from positive mice were amplified by PCR using the LOX1 and SDL2 primers. This reaction amplifies a WT product 409 bp in size. The presence of a second PCR product 62 bp greater than the WT product indicates a positive LoxP PCR. After a 2-minute hot start at 94°C, the samples were run using the following conditions for 30 cycles: 94°C for 30 seconds, followed by 55°C for 30 seconds, and 72°C for 1 minute. The PCR product was run on a 2% gel with a 100 bp ladder as reference.

Primers for PCR screening:

• Forward oligo:

• LOX1: 5′-TGG CTA AAC CAG AGG AGG TAC TG-3′

• Reverse oligo:

• SDL2: 5′-ATG GAG CCT ATT GCC CAA CAC AAC-3′

NALM-6 cells engineered to express XPO1^E571K/WT from the endogenous locus were generated by Horizon Discovery using recombinant adeno-associated virus-mediated homologous recombination and cultured in RPMI medium with 10% FCS. MEC1 cells were purchased from DSMZ. SUDHL-5, SUDHL-6, SUDHL-16, L428, and SUP-HD1 cells were a generous gift from Dr. Laura Pasqualucci. All cell lines were tested for Mycoplasma with the MycoAlert Mycoplasma Detection Kit (Lonsa) at time of receipt and before preparing stock aliquots for storage in liquid nitrogen. Cells were thawed and passaged no more than 10 times for all experiments performed. Validation of the XPO1^E571K mutation was performed by next-generation targeted capture sequencing, and serial routine testing was performed by PCR and Sanger sequencing.

Studies were approved by the Institutional Review Board of MSK and conducted in accordance to the Declaration of Helsinki protocol. We obtained written informed consent from patients before primary human deidentified CLL samples derived from whole peripheral blood or BM mononuclear cells were utilized.

All data are presented as mean ± SD, unless otherwise stated. The replicate for each experiment is stated in the figure legend or indicated in the figure. Statistical significance was determined by a two-sided Student t test, and a P value of <0.05 was considered statistically significant. For non–normally distributed data, a nonparametric test (Kruskal–Wallis) was used, followed by multiple group comparisons using FDR. For the Kaplan–Meier survival analysis, a Mantel–Cox log-ranked test was used to determine statistical significance. For offspring frequency analysis, a χ² test was performed to test the difference between observed and expected frequencies from different genotypes. No blinding or randomization was used. Unless otherwise noted, all immunoblot quantitation and immunofluorescence image quantitation were representative of at least three biological replicates from independent experiments. Data were plotted using GraphPad Prism 7 software.

A.R. Mato reports receiving commercial research grants from Celgene, Janssen, AstraZeneca, and AbbVie, and is a consultant/advisory board member for TG Therapeutics, Pharmacyclics, AstraZeneca, Loxo, Sunesis, Prime Oncolgy, Pfizer, and AbbVie. A. Dogan is a consultant/advisory board member for Roche, Novartis, Seattle Genetics, Celgene, PER, and Corvus. D. Wang is a member of the Adjunct Faculty at the Medical College of Wisconsin and Fujian Normal University. R.C. Hendrickson has ownership interest (including stock, patents, etc.) in Merck. O. Abdel-Wahab reports receiving commercial research support from H3 Biomedicine Inc., has received honoraria from the speakers bureau of Foundation Medicine Inc., and is a consultant/advisory board member for Foundation Medicine Inc., H3 Biomedicine Inc., Janssen, and Merck. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J. Taylor, D. Wang, B.S. Taylor, O. Abdel-Wahab

Development of methodology: J. Taylor, M. Sendino, C. Stewart, R.C. Hendrickson, G. Prieto, J.A. Rodriguez

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Taylor, M.T. Chang, C. Stewart, E.M. Melnik, F. Herrejon Chavez, L. Bitner, A. Yoshimi, S.C.-W. Lee, D. Inoue, B. Liu, X.J. Zhang, A.R. Mato, Y. Chen, R.K. Soni, R.C. Hendrickson, B.S. Taylor

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Taylor, A.N. Gorelick, A. Pastore, M.T. Chang, A.V. Penson, E.I. Gavrila, C. Stewart, A. Yoshimi, X.J. Zhang, A. Dogan, R.K. Soni, G. Prieto, B.S. Taylor, O. Abdel-Wahab

Writing, review, and/or revision of the manuscript: J. Taylor, E.I. Gavrila, A. Yoshimi, S.C.-W. Lee, A.R. Mato, A. Dogan, D. Wang, R.K. Soni, G. Prieto, B.S. Taylor, O. Abdel-Wahab

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X.J. Zhang, O. Abdel-Wahab

Study supervision: M.G. Kharas, O. Abdel-Wahab

J. Taylor is supported by the Conquer Cancer Foundation of the American Society of Clinical Oncology, the American Association for Cancer Research, the American Society of Hematology (ASH), the Robert Wood Johnson Foundation, and the NIH/NCI (1K08CA230319-01). C. Stewart is supported by an ASH HONORS award. A. Yoshimi is supported by grants from the Aplastic Anemia and MDS International Foundation and the Lauri Strauss Leukemia Foundation. A. Yoshimi, S.C.-W. Lee, D. Inoue, and O. Abdel-Wahab are supported by the Leukemia and Lymphoma Society. S.C.-W. Lee is supported by the NIH/NCI (K99 CA218896), the ASH Scholar Award, and the EvansMDS Young Investigator Award. B.S. Taylor is supported by grants from NIH/NCI (R01 CA207244, R01 CA204749, U54 OD020355), the American Cancer Society (RSG-15-067-01-TBG), the Anna Fuller Fund, and the Robertson Foundation. D. Wang is supported by grants from the NIH/NIAID (AI079087) and NIH/NHLBI (R01 HL130724). M. Sendino and J.A. Rodriguez are supported by grants from the Spanish Government MINECO-FEDER (SAF2014-57743-R), the Basque Country Government (IT634-13 grant and a predoctoral fellowship) and the University of the Basque Country (UFI11/20). O. Abdel-Wahab is supported by grants from NIH/NHLBI (R01 HL128239) and NIH/NCI (1 R01 CA201247-01A1), an MSKCC Steven Greenberg Lymphoma Research Award, the Geoffrey Beene Research Center of MSKCC, and the Pershing Square Sohn Cancer Research Alliance. This work was also supported by P30 CA008748. Research supported AACR-Conquer Cancer Foundation of ASCO Young Investigator Award for Translational Cancer Research. Editorial support in the preparation of this manuscript was provided by Hannah Rice, ELS.