T Cells Specific for an Unconventional Natural Antigen Fail to Recognize Leukemic Cells

The repertoire of HLA-presented peptides is dependent on antigen-containing protein expression and turnover. Therefore, translational events including premature termination of translation and protein misfolding are opportunities for regulation of antigen presentation. In tumor cells, protein production is often elevated, which is likely to result in increased premature termination and enhanced misfolding. The rapidly degraded subset of aberrant proteins is called defective ribosomal products (DRiP). The various classes of aberrant proteins are a potential source of (neo-)antigenic peptides and may provide specific targets for T-cell–based immunotherapy (1–4). This hypothesis is based on findings with viral and model antigens (5). In humans, natural T cells can recognize an antigen derived from an aberrant protein on insulin-producing β cells. These T cells may contribute to the induction of autoimmune diabetes (6). Similar responses may exist or be induced against natural antigens processed from aberrant proteins in tumor cells.

Donor lymphocyte infusion (DLI) after allogeneic stem cell transplantation (alloSCT) is an example of effective immunotherapy for the treatment of hematologic malignancies. The graft-versus-tumor effect of this therapy is mediated by allo-reactive donor T cells recognizing HLA-presented peptides, so-called minor histocompatibility antigens (minor H Ags), on the tumor cells. The immunogenicity of minor H Ags is a result of natural genetic variation between donor and recipient (7). Although donor T cells can mount potent responses against patient cells that carry minor H Ags, the transplanted cells (donor-derived and thus lacking minor H Ags) are spared (8, 9). Donor T cells may also attack healthy patient tissues causing graft-versus-host disease. Consequently, many studies have focused on the identification of new minor H Ags to understand how tissue-specific antigen presentation is controlled (10–12).

We now report the identification of a minor H Ag that is selectively encoded by an alternatively spliced transcript of the TTK gene. This transcript contains a premature termination codon leading to production of a truncated protein. The isolated T cells specific for this antigen fail to recognize the majority of leukemic cells expressing the alternative TTK transcript. The discovery of this antigen shows that leukemic cells, contrary to what is thought from model studies, are not necessarily immunogenic targets for T cells recognizing (neo-)antigens from aberrant proteins. Our data suggest that concern is warranted for immunotherapies targeting such antigens.

The CD8⁺ minor H Ag-specific cytotoxic T-cell (CTL) clones 10-4 and 10-5, designated as clone type H4, and here identified as recognizing LB-TTK-1D, were previously generated (13). T cells were isolated from a patient with myelodysplastic syndrome (patient 5852) who converted to 100% donor chimerism after treatment with alloSCT and DLI. T-cell clones were generated by single-cell sorting of HLA-DR⁺ CD8⁺ T cells by flow cytometry. T cells were cultured in IMDM supplemented with 5% FBS, 5% human serum, IL2 (100 IU/mL), and restimulated every 14 days with irradiated allogeneic peripheral blood mononuclear cells (PBMC) and phytohemagglutinin (0.8 μg/mL, Murex Biotec Limited). PBMCs and bone marrow mononuclear cells (BMMC) were obtained from patients and healthy individuals after approval by the Leiden UMC Institutional Review Board and informed consent according to the Declaration of Helsinki. Patients were HLA-typed using serologic analyses (n = 6) or DNA-sequencing (n = 9) by our HLA-lab that supports the allogeneic transplantation center. Mononuclear cells were isolated by Ficoll–Isopaque separation and cryopreserved. Phoenix Amphotropic Cells (Φnx-A; provided by Dr. J. Neefjes, The Netherlands Cancer Institute, Amsterdam, the Netherlands, 2013), HAP1 cells (provided by Dr. T. Brummelkamp, The Netherlands Cancer Institute, Amsterdam, the Netherlands, 2011), 2A14 melanoma cells (provided by Dr. S.H. van der Burg, LUMC, Leiden, the Netherlands, 2016), MZ1257 renal cell carcinoma cells (provided by Dr. B. Seliger, Johannes Gutenberg University, Mainz, Germany), fibroblasts (FB), keratinocytes (KC), EBV-transformed B cells (EBV-LCLs), CD40L-cultured B cells, PHA T blasts, and allo-HLA-A*02:01 reactive T cells (clone HSS12; ref. 14) were cultured as described previously (15–18). FB, KC, EBV-LCLs, and primary B and T cells were isolated and generated in our labs between 2005 and 2015. Cell lines were Mycoplasma negative as determined in bimonthly tests. Cell lines were not authenticated in the past year.

Clone 10-5 was tested for recognition of 90 single-nucleotide polymorphism (SNP)-genotyped EBV-LCLs by IFNγ ELISA. OD₄₅₀ > 0.15 was considered to identify cells expressing minor H Ags. T-cell recognition was investigated for association with 1.1 × 10⁶ SNPs by whole-genome association scanning (WGAs) using Fisher exact test as described previously (13).

Because haplotype blocks may differ per ethnic population (19) and because the original WGAs was based on minor H Ag phenotypes of Dutch individuals, we selected 1000 Genomes Project (1000GP) individuals of European descent for additional correlation calculations. We retrieved the genotypes for the highest associating SNP as identified by WGAs (rs608962) of 50 Caucasian (CEU) individuals from the 1000GP (20). The likely minor H Ag phenotype for each of those individuals was inferred from these genotype data and used to statistically determine which other SNPs on chromosome 6 (on which rs608962 is located) were in linkage disequilibrium with rs608962 (and thus also with the minor H Ag) using Fisher exact tests performed with our previously described and freely available algorithm in the software R (21). The top correlating SNPs were further analyzed for translational consequences using Ensembl GRCh38.p12 (https://www.ensembl.org). Peptide HLA–binding predictions were made using the NetMHC 3.2 server (22).

Cryopreserved PBMCs from patient 5852 obtained after alloSCT 6 weeks before (148 days after alloSCT) and 6 and 7 weeks after DLI (231 and 237 days after alloSCT) were thawed and stained with PE-conjugated UV-exchange tetramers containing LB-TTK-1D and APC-conjugated tetramers containing LB-ERAP1-1R. Tetramers were constructed as described previously (23, 24). Acquisition was performed on a FACS Calibur Analyzer (BD Biosciences) using CellQuest software and analyzed using FlowJo (Tree Star).

Stimulator cells (5–15 × 10³ cells/well) and T cells (2 × 10³ cells/well) were coincubated overnight in 384-well plates. IFNγ release was measured by ELISA (Sanquin). To identify the antigenic epitope, peptides were synthesized, dissolved in DMSO, and pulsed at titrated concentrations in IMDM on EBV-LCLs from the stem cell donor, or on primary leukemic cells obtained from patients. After 2 hours of pulsing at 37°C, T cells were added, and IFNγ production in the supernatant was measured after overnight coincubation by IFNγ ELISA. T-cell recognition of healthy and malignant cells, which included PHA-blasts, CD40L-cultured B cells, dendritic cells, FB, KC, and leukemic cells of different origins (ALL, AML, CLL, and CML), as well as of donor EBV-LCLs either transduced with TTK-207 construct or a mock vector, was tested in a similar manner.

To determine the genetic positivity for the LB-TTK-1D antigen, we performed genotyping of SNP rs608962 in intron 20 of TTK, which correlates with the minor H Ag encoding SNP rs240226 in European individuals of the 1000GP, using the KASPar SNP Genotyping System (KBioscience). The specific forward primers that were used for allele 1 and 2 were, respectively, 5′-GAAGGTGACCAAGTTCATGCTGTGTATTTACAATGGACTTATATTTTTCTTAT-3′, 5′-GAAGGTCGGAGTCAACGGATTGTGTATTTACAATGGACTTATATTTTTCTTAC-3′ and the shared reverse primer was 5′-TCTTCAGTGGTTCCCTTGGCCATTT-3′. Genomic DNA (10 ng) was amplified according to the manufacturer's instructions and fluorescent signals were quantified on a 7900HT device running with SDS Software (Applied Biosystems).

For qRT-PCR and sequencing analysis, tRNA was isolated using the RNAqueous Micro-Kit and Small Scale Kit (Ambion, Life Technologies) for a maximum of 0.5 × 10⁶ and 10 × 10⁶ cells, respectively, following the manufacturer's instructions. Fractionation of nuclear and cytoplasmic RNA was performed on 10⁷ cells. Cells were lysed using a 0.05% NP-40 buffer (140 mmol/L NaCl, 1.5 mmol/L MgCl2, 10 mmol/L Tris–HCl, pH 8.0) containing 40 U/mL of RNAseOut (Invitrogen). After centrifugation, the nuclei (pellet) and cytoplasmic fraction (supernatant) were separated. RNA was then isolated using TRIzol and phenol–chloroform extraction, respectively. cDNA was synthesized from tRNA using Moloney murine leukemia virus reverse transcriptase (Invitrogen). For qPCR, TTK expression was measured on the Roche Lightcycler 480 using Fast Start TaqDNA Polymerase (both Roche) and EvaGreen (Biotium) with forward primer 5′-ACCTTACTGATGAACTAAGCTTGAA-3′ and reverse primer 5′-TCCCGAGTTATCTGTAGTATCAGC-3′ for the full-length TTK transcript and forward primer 5′-ACTTTGAATGGTGTCTGGCACA-3′ and reverse primer 5′-CTCTGGGTTGTTTGCCATCAT-3′ for the alternative TTK transcript. Expression was normalized to the PBGD (HMBS) reference gene.

For Sanger sequencing of the alternative transcript TTK-207, we performed two sequential rounds of PCR on cDNA of patient EBV-LCLs using two primer sets each containing one primer that was specific for the alternative (and not the full-length) transcript. In the first round, forward primer 1 5′-ATGGAATCCGAGGATTTAAGTG-3′ and reverse primer 1 5′-TGTGCCAGACACCATTCAAAG-3′ as well as forward primer 2 5′-ACTTTGAATGGTGTCTGGCACA-3′ and reverse primer 2 5′-GTAGTCACGTGCATCATCTG-3′ were used to produce two products, which were combined to one fragment in the second round of amplification using forward primer 1 and reverse primer 2. The alternative TTK transcript was confirmed by Sanger sequencing in both directions.

A construct for the alternative TTK transcript containing exon 1-5 was cloned into a pLZRS vector containing a truncated ΔNGFR marker gene linked by an IRES. The construct was cloned from patient 5852 cDNA (homozygous for the minor H Ag encoding SNP) using a two-step PCR with forward primer 1 5′-CGCGGATCCACCATGGAATCCGAGGATTTAAGTG-3′ and reverse primer 1 5′-TGTGCCAGACACCATTCAAAG -3′ as well as forward primer 2 5′-ACTTTGAATGGTGTCTGGCACA-3′ and reverse primer 2 5′-TATATACTCGAGGTAGTCACGTGCATCATCTG-3′ in the first round to produce two bands, which were combined to one fragment in the second round of amplification using forward primer 1 and reverse primer 2. Fragments were digested with BamHI and XhoI for cloning and the transcript construct was confirmed by Sanger Sequencing. Retroviral supernatant was obtained by transfecting wild-type Φnx-A packaging cells as described previously (25) with the exception that the Fugene HD Transfection Kit (Roche) was used. Viral supernatant was used for transduction on plates coated with recombinant human fibronectin CH 296 (Takara Shuzo).

We previously isolated several T-cell clones specific for a single minor H Ag from a patient after alloSCT and DLI (13). In total, 14 clones using two different T-cell receptor variable β-chain (TRBV) recognized a minor H Ag with a population frequency of 50% in HLA-A*02:01. Because minor H Ags are polymorphic in the population, they can be identified using WGAs (13, 26, 27). Therefore, we first phenotyped 90 different EBV-LCLs by testing their individual capacity to activate the minor H Ag-specific T-cell clone 10-5 (Fig. 1A). Because we previously genotyped the same panel of EBV-LCLs for 1.1 × 10⁶ SNPs, we next associated the minor H Ag phenotypes to these genotypes in a WGAs. This analysis identified 13 SNPs on chromosome 6 that associated with T-cell recognition (P < 3.0 × 10⁻¹⁶), of which the strongest associating SNP rs608962 had a P value of 3.6 × 10⁻²⁰ (Fig. 1B). Although located in close proximity to and within the TTK gene, these SNPs were all noncoding and were therefore not our preferred candidates to encode the minor H Ag. Furthermore, Sanger sequencing of the primary transcript of TTK did not identify differences between patient and stem cell donor.

During the last decade, we have improved the resolution of our genome-wide association strategy by implementing data of the 1000GP; containing more than 40 × 10⁶ SNPs; refs. 21, 27, 28). We now used the 1000GP data for an in silico attempt to identify the encoding genetic variation of the minor H Ag as it could not be identified by the initial WGAs. We searched in the 1000GP data for coding SNPs that are in linkage disequilibrium with SNP rs608962 as identified by the WGAs to link with the minor H Ag encoding locus. The analysis yielded 93 fully correlating SNPs on chromosome 6 (Fig. 1C and D), of which only SNP rs240226 is a nonsynonymous coding variant, encoding a glutamic acid (E) to aspartic acid (D) change in a putative alternative TTK transcript.

SNP rs240226 is located in an exon, found in an alternative TTK transcript (TTK-207, Ensembl) but absent from other TTK transcripts (Fig. 2A). TTK-207 contains a premature termination codon in the last but one exon, which may render this transcript susceptible to nonsense-mediated decay (NMD), and define its translational product as a DRiP (3, 29, 30). To show expression of this alternatively spliced transcript, we performed a two-step PCR strategy to amplify this transcript (illustrated in Fig. 2A, see Materials and Methods). Sanger sequencing of the amplified cDNA from patient EBV-LCLs confirmed that it was alternative transcript TTK-207. Because NMD transcripts have decreased expression in the cytoplasm compared with the nucleus (31), we separately analyzed expression of TTK-207 and the full length TTK transcript (TTK-201) in the nucleus and cytoplasm by qPCR. The data showed that the ratio of nuclear-to-cytoplasmic TTK-207 was higher than for TTK-201 in only one of the three cell lines tested (Fig. 2B). However, in all three cell lines, nuclear and cytoplasmic expression of TTK-207 was low compared with the full-length transcript, which is a characteristic of NMD-degraded transcripts (Fig. 2B; refs. 32, 33).

To establish whether the minor H Ag is derived from TTK-207, we cloned this transcript from minor H Ag⁺ patient cDNA and retrovirally introduced this construct into donor EBV-LCLs. T-cell clone 10-5 recognized donor EBV-LCLs upon retroviral transduction with the construct (Fig. 2C), confirming that this transcript encodes the minor H Ag (LB-TTK-1D). On the basis of NetMHC HLA-peptide–binding predictions (22), we hypothesized that epitope RLHDGRVFV was the most likely candidate to be recognized by clone 10-5 with RLHEGRVFV being the allelic variant. Indeed, T-cell experiments confirmed that RLHDGRVFV was potently recognized with picomolar affinity, while the allelic E-variant was not recognized (Fig. 2D). Of note, two different T-cell receptors (TCR) for LB-TTK-1D represented by T-cell clone 10-5 (TRBV3-1) and clone 10-4 (TRBV19) showed similar affinity for the peptide (Fig. 2D).

The antigenic peptide recognized by T-cell clone 10-5 is only encoded by TTK-207 and not by other TTK transcripts and is therefore exclusively derived from the truncated TTK protein. TTK is a protein kinase that plays a role in mitosis. It has a kinase domain that stretches from exon 14 into exon 20 and an N-terminal tetratricopeptide repeat domain in exon 3, which is required for its subcellular localization to kinetochores (34, 35). Both domains are absent in the truncated protein encoded by TTK-207, suggesting that this protein lacks a physiologic function. Because TTK-207 is a naturally occurring transcript variant, the antigen LB-TTK-1D represents a tool to study the behavior of T-cell antigens processed from aberrant proteins.

To investigate whether T cells specific for LB-TTK-1D can undergo priming and expansion in vivo, we analyzed the peripheral blood of the alloSCT recipient 6 weeks before DLI and 6 and 7 weeks after DLI with tetramers. The frequencies of LB-TTK-1D–specific T cells were compared with minor H Ag LB-ERAP1-1R–specific T cells, which were previously identified as immunodominant in the same patient after DLI (13). Both antigen-specific T cells were undetectable prior to DLI but were detected in similar numbers after DLI (up to 0.25% and 0.58%, of total lymphocytes, respectively; Fig. 2E). This indicated that an oligoclonal T-cell response against LB-TTK-1D was induced in vivo that was similar to an immunodominant control T-cell response.

The production of unconventional proteins including DRiPs may be increased in uncontrolled proliferating cells. Therefore, we investigated whether cancer cells are better recognized by LB-TTK-1D–specific T cells in comparison with healthy cells. Primary leukemic cells as well as healthy cells from various hematopoietic lineages isolated from patients and healthy donors who were positive for HLA-A*02 and the minor H Ag allele were tested for T-cell recognition. Only one of 11 leukemic tumors was recognized by the LB-TTK-1D–specific T cells (Fig. 3A). Of the various regular hematopoietic cell types [PHA T cells, (im)mature dendritic cells, CD40L-activated B cells, EBV-LCLs, full PBMCs, and BMMCs], immature monocyte-derived dendritic cells, PBMCs, and BMMCs lacked the capacity to activate LB-TTK-1D–specific T cells (Fig. 3A). For the immature dendritic cells, this could be explained by a decreased efficacy in general antigen processing and presentation, because these cells also induced lower activation of a control HLA-A*02:01–restricted T-cell clone recognizing an USP11-derived peptide (Fig. 3B). In contrast, antigen processing and presentation of conventional antigenic peptides in the majority of leukemic cells was intact, because they triggered the USP11-specific T-cell clone (Fig. 3B). Moreover, leukemic cells exogenously loaded with the TTK peptide stimulated T-cell clone 10-5, further substantiating that the leukemic cells are, in principle, functional presenters of antigens (Fig. 3C). Furthermore, T cells specific for LB-TTK-1D and USP11 exhibited a similar affinity toward their cognate peptide when exogenously loaded on T2 cells or EBV-LCLs or (both in the range of 0.2–2 nmol/L), indicating that the lack of recognition of LB-TTK-1D on leukemic cells was not skewed by differences in peptide-sensitivity between the two T-cell clones (Fig. 2D; Supplementary Fig. S1; ref. 14).

A cell line generated from one of the ALL samples was, similar to its primary tumor, not recognized by LB-TTK-1D–specific T cells, whereas regular antigen processing and presentation was sufficient to stimulate USP11-specific T cells (Fig. 3A and D). Retroviral introduction of the alternative TTK-207 transcript restored the capacity of this cell to stimulate LB-TTK-1D–specific T cells (Fig. 3D). These results show that the processing and presentation of LB-TTK-1D in this ALL cell line is not defective, but rather suggest a failure in leukemic cells to express the truncated TTK protein.

To determine whether the failure to express the truncated TTK protein could be explained by low expression of the alternative TTK-207 transcript, we quantified expression of TTK-207 as well as the full length TTK transcript by qPCR. In all cells tested, expression of the alternatively spliced TTK-207 transcript was lower compared with the regular TTK transcript, again indicating that TTK-207 may be rapidly degraded through the NMD transcript surveillance pathway (Fig. 4; refs. 32, 33). All tested leukemic tumors naturally expressed the TTK-207 transcript, at least as much as did EBV-LCLs, which were efficiently recognized by T-cell clone 10-5 (Fig. 3A and 4). In summary, these data suggest that leukemic cells have developed mechanisms to prevent aberrant protein expression and thereby antigen presentation from unconventional, possibly NMD-derived, transcripts.

In this article, we report the identification of a naturally occurring antigen encoded by an alternatively spliced transcript. This transcript contains a premature termination codon leading to the production of a truncated protein. The antigen is a polymorphic peptide recognized by minor H Ag reactive T cells, propagated only after HLA-matched allogeneic transplantation. We identified the HLA-A*02:01–restricted minor H Ag LB-TTK-1D by combining multiple genetic analyses. Currently, 1000GP-based analyses provide the most power to identify polymorphic antigens. Here, this power was required to identify the antigen recognized by T-cell clone 10-5, because conventional WGAs failed to determine the antigenic epitope. The additional in silico strategy that we used here is likely to be also applicable for identification of other minor H Ags not previously identified using SNP arrays (36).

The antigenic peptide is encoded by the alternative transcript TTK-207. As TTK-207 contains a premature termination codon more than 50 nucleotides (namely 193 nucleotides) before the last exon–exon junction, it is supposed to be subjected to the surveillance mechanism of NMD leading to rapid mRNA degradation (29, 30, 32, 33). Despite its low expression, we confirmed the presence of the alternative transcript in various cell types, and demonstrated that the transcript is sufficiently expressed for efficient T-cell induction in vivo and T-cell recognition in vitro. There has been much debate about the intracellular location and timing of NMD and how much protein is actually produced by NMD transcripts (37, 38). Evidence from in vitro data using model antigens indicates that NMD transcripts lead to newly synthesized but dysfunctional proteins that are rapidly degraded via the ubiquitin–proteasome pathway (32, 39, 40). Recognition of LB-TTK-1D on EBV-LCLs was indeed partially abolished by the proteasome inhibitor bortezomib (Supplementary Fig. S2), indicating that the truncated TTK protein is processed via the canonical route of the ubiquitin–proteasome system.

DRiPs are presumably generated in the pioneering rounds of translation, and artificial DRiPs encoded by NMD transcripts produce peptides for MHC class I presentation (32, 39, 40). However, the contribution of DRiPs relative to stable mature proteins to the total pool of antigenic peptides remains unclear (2, 3, 37, 39). Our data provide evidence that naturally occurring alternatively spliced transcripts encoding aberrant proteins may be a relevant source of (neo-)antigens.

Epitopes originating from DRiPs may serve as neoantigenic targets for antitumor immunotherapeutic approaches based on model antigen and compound inhibition studies (1, 4, 32, 41). The repertoire of antigenic peptides on tumor cells would increase upon inhibition of the rapid degradation process of NMD. This is supported by the finding that broad inhibition of NMD resulted in production of neoantigens and induction of specific immune responses, leading to tumor rejection in vivo (4). Despite the fact that the TTK antigen can induce robust allogeneic CD8⁺ T-cell expansion, it apparently was not able to induce autologous T-cell immunity in the patient before stem cell transplantation, implying that aberrant proteins such as DRiPs are not necessarily a class of immunogeneic neoantigens. Furthermore, the TTK-derived antigen is naturally expressed and presented by healthy cells showing that presentation of unconventional antigens is not restricted to tumor cells. Moreover, our data show that leukemic cells fail to present the TTK antigen to allogeneic CD8⁺ T cells even though they were genomically positive for the antigenic allele, expressed the alternative TTK transcript, and displayed proper antigen processing and presentation. These results may be explained by a discrepancy between the presence of the alternative TTK transcript and expression or turnover of its truncated protein product in different cell types (42). Leukemic cells may regulate these processes to minimize immunogeneic antigen presentation possibly through suppression of immunoribosomes (43, 44). These data indicate that antigens from aberrant proteins such as DRiPs are not necessarily good targets in cancer immunotherapy.

Our approach for characterizing this minor H Ag may aid future epitope identification studies. Our observation that this endogenous antigen, derived from an aberrant protein, is only occasionally expressed on tumor cells suggests potential risks to DRiP-directed immunotherapeutic applications.

T. Mutis reports receiving commercial research funding from Genmab, Takeda, Janssen Pharmaceuticals, Onkimmune, Aduro, and Novartis. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.J. Pont, R. Oostvogels, H.M. Lokhorst, J.H.F. Falkenburg, T. Mutis, M. Griffioen, R.M. Spaapen

Development of methodology: M.J. Pont, R. Oostvogels, J.H.F. Falkenburg, T. Mutis, M. Griffioen, R.M. Spaapen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.J. Pont, R. Oostvogels, C.A.M. van Bergen, E.D. van der Meijden, M.W. Honders, S. Bliss, M. Griffioen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.J. Pont, R. Oostvogels, C.A.M. van Bergen, E.D. van der Meijden, M.W. Honders, J.H.F. Falkenburg, M. Griffioen, R.M. Spaapen

Writing, review, and/or revision of the manuscript: M.J. Pont, R. Oostvogels, H.M. Lokhorst, J.H.F. Falkenburg, T. Mutis, M. Griffioen, R.M. Spaapen

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Oostvogels, M.W. Honders, M.L.M. Jongsma, M. Griffioen

Study supervision: M.J. Pont, H.M. Lokhorst, M. Griffioen, R.M. Spaapen

The authors would like to thank Michel Kester for the production of tetramers and Dr. Marieke von Lindern and Nurcan Yagci for expertise and technical assistance on the isolation of nuclear and cytoplasmic RNA. This work was supported by a NWO-VENI personal grant (016.131.047, to R.M. Spaapen) and by the Dutch Cancer Society (UL 2010-4748, to M. Griffioen).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.