Non-IG Aberrations of FOXP1 in B-Cell Malignancies Lead to an Aberrant Expression of N-Truncated Isoforms of FOXP1

The transcription factor FOXP1 is implicated in the pathogenesis of B-cell lymphomas through chromosomal translocations involving either immunoglobulin heavy chain (IGH) locus or non-IG sequences. The former translocation, t(3;14)(p13;q32), results in dysregulated expression of FOXP1 juxtaposed with strong regulatory elements of IGH. Thus far, molecular consequences of rare non-IG aberrations of FOXP1 remain undetermined. Here, using molecular cytogenetics and molecular biology studies, we comprehensively analyzed four lymphoma cases with non-IG rearrangements of FOXP1 and compared these with cases harboring t(3;14)(p13;q32)/IGH-FOXP1 and FOXP1-expressing lymphomas with no apparent structural aberrations of the gene. Our study revealed that non-IG rearrangements of FOXP1 are usually acquired during clinical course of various lymphoma subtypes, including diffuse large B cell lymphoma, marginal zone lymphoma and chronic lymphocytic leukemia, and correlate with a poor prognosis. Importantly, these aberrations constantly target the coding region of FOXP1, promiscuously fusing with coding and non-coding gene sequences at various reciprocal breakpoints (2q36, 10q24 and 3q11). The non-IG rearrangements of FOXP1, however, do not generate functional chimeric genes but commonly disrupt the full-length FOXP1 transcript leading to an aberrant expression of N-truncated FOXP1 isoforms (FOXP1NT), as shown by QRT-PCR and Western blot analysis. In contrast, t(3;14)(p13;q32)/IGH-FOXP1 affects the 5′ untranslated region of FOXP1 and results in overexpress the full-length FOXP1 protein (FOXP1FL). RNA-sequencing of a few lymphoma cases expressing FOXP1NT and FOXP1FL detected neither FOXP1-related fusions nor FOXP1 mutations. Further bioinformatic analysis of RNA-sequencing data retrieved a set of genes, which may comprise direct or non-direct targets of FOXP1NT, potentially implicated in disease progression. In summary, our findings point to a dual mechanism through which FOXP1 is implicated in B-cell lymphomagenesis. We hypothesize that the primary t(3;14)(p13;q32)/IGH-FOXP1 activates expression of the FOXP1FL protein with potent oncogenic activity, whereas the secondary non-IG rearrangements of FOXP1 promote expression of the FOXP1NT proteins, likely driving progression of disease.


Introduction
The FOXP1 (Forkhead box P1) gene located at 3p13 (previously assigned to 3p14.1) codes for a transcriptional regulator belonging to the FOX transcription factor family which is implicated in a wide range of biological processes [1,2]. Multiple alternative splicing variants of FOXP1 have been annotated (www. ensemble.org). The FOXP1 protein is widely expressed in human tissues. It harbors an N-terminal poly-Gln region, C2H2-type zinc finger and leucine zipper motifs, and a C-terminal DNA binding forkhead or winged helix domain [3]. FOXP1 is an essential factor in cardiac, lung, neural, monocyte and lymphocyte development and maturation, stem cell biology and in immune responses [4][5][6][7][8][9][10][11][12][13][14]. Multiple lines of evidence indicate that FOXP1 plays an important role in tumorigenesis [15]. Initial studies of Banham et al. [16] suggested that FOXP1 acts as a tumor suppressor in epithelial malignancies recurrently characterized by D3p13p14/FOXP1 and loss or decreased expression of the FOXP1 protein. Very recent work of Krohn et al. [17] on prostate cancers supports this concept. Interestingly, subsequent studies postulated an oncogenic role of FOXP1 in lymphoma, particularly in an activated B-cell subtype of diffuse large B-cell lymphoma (ABC-DLBCL) with a poor clinical outcome [18][19][20][21][22][23], and extranodal marginal zone lymphoma (MZL), where a strong expression of FOXP1 protein correlates with a risk of a high grade transformation [24][25][26][27]. Further investigations showed that FOXP1-positive ABC-DLBCL [28], as well as follicular lymphoma [29] and primary central nervous system lymphoma [30], preferentially express shorter FOXP1 isoforms, which in non-malignant conditions, may be induced by B-cell activation [28]. It has been hypothesized that the role of FOXP1 as oncogene is due to expression of short FOXP1 isoforms, while the full length FOXP1 (FOXP1 FL ) acts as tumor suppressor [15,28].
In the reported study, we performed genetic and molecular analysis of four lymphoma cases with non-IG translocations of FOXP1 and compared these with cases harboring t(3;14)(p13;q32)/IGH-FOXP1 and DLBCLs with a strong expression of FOXP1 and with no apparent structural aberrations of the gene (FOXP1-positive DLBCL). Our study demonstrates that non-IG rearrangements of FOXP1 do not generate chimeric transcripts but activate an aberrant expression of transcriptional variants of FOXP1 and N-terminally truncated FOXP1 isoforms (FOXP1 NT ). In addition, our data suggest that non-IG translocations of FOXP1 are implicated in progression of various B-cell neoplasms, including chronic lymphocytic leukemia (CLL).

Patients
Lymphoma cases with and without FOXP1 rearrangements were selected from the database of the Center for Human Genetics and Department of Pathology, KU Leuven, Leuven, Belgium. One case with t(non-IG/FOXP1) was provided by Dr. E. Haralambieva (Institute of Pathology, University of Würzburg, Würzburg, Germany). Morphology, immunophenotype and clinical records of the studied cases were reviewed. DLBCL subtyping followed immunohistochemical (IHC) algorithm of Hans et al. [36]. The study was approved by the institutional review board ''Commissie Medische Ethiek'' of the University Hospital. For this retrospective study the ''Commissie Medische Ethiek'' waived the need for written informed consent from the participants.

Cytogenetic and Fluorescence in situ Hybridization Analysis
G-banding chromosomal analysis and fluorescence in situ hybridization (FISH) followed routine methods. Probes applied for FISH are listed in Table S1. Non-commercial probes were labeled with SpectrumOrange-and SpectrumGreen-d-UTP (Abbott Molecular, Ottigne, Belgium) using random priming. FISH images were acquired with a fluorescence microscope equipped with an Axiophot 2 camera (Carl Zeiss Microscopy, Jena, Germany) and a MetaSystems ISIS imaging system (MetaSystems, Altlussheim, Germany). One to ten abnormal metaphases and/or 200 interphase cells were evaluated in each experiment.

Immunohistochemistry
For routine IHC ready-to-use antisera against CD20, CD10, BCL6 and MUM1 from DAKO (Glostrup, Denmark) were used and stainings were performed using the automated DAKO stainer Anti-FOXP1 antibody (ab16645) was purchased from Abcam (Cambridge, UK) and used at dilution 1:200. We additionally used anti-FOXP1 serum from Roche Diagnostics (SP133, Roche Diagnostics, Vilvoorde, Belgium) and staining followed the manufacturer's recommendations. IHC results were visualized using the OptiView DAB IHC Detection Kit (Ventana, Oro Valley, Tucson, Arizona). Image acquisition was done through a Leica microscope at 2006 and 1006 magnification. Images were assembled using Adobe Photoshop CS5.

RNA Extraction and cDNA Preparation
TRIzol LS Reagent (Invitrogen, Carlsbad, CA, USA) was used for a total RNA extraction. cDNA was synthesized from 1 mg of total RNA using reverse transcriptase Superscript II (Invitrogen, Carlsbad, CA, USA) and random primer (Invitrogen, Carlsbad, CA, USA).

Quantitative RT-PCR
QRT-PCR was performed using LC480 SYBR Green I Master and the LightCycler 480 Real-Time PCR System. Data were analyzed with the LC480 software (Roche Diagnostics, Indianapolis, IN, USA). Primers representative for exons 3-18 (Table S2) were designed using Light Cycler Probe Design Software 2.0. C T values were averaged for triplicate reactions and used to calculate DC T values for each sample. HPRT1 was used as a reference gene for normalization. RPMI-8402, a T-cell leukemia cell line (www. dsmz.de), was used to normalize the relative expression of FOXP1 between the samples.

59-Rapid Amplification of cDNA Ends PCR (59RACE-PCR)
First strand cDNA was synthesized from 3 mg of total RNA as previously described but using the oligonucleotide FOXP1/ex8R-Race1 designed on exon 8 of FOXP1. The first strand cDNA was tailed with deoxyadenosine triphosphate (dATP). Second strand was synthesized using Klenow DNA polymerase (Promega, Fitchburg, WI, USA) and the primers mix TV8. Anchored PCR was performed for 35 cycles with primers FOXP1/ex7R/race-1 and 467, and for nested PCR we used the primers FOXP1/ex7R/ race-2 and 468. PCR products were cloned to pGEM-T Easy (pGEM-T-easy Vector system I, Promega, Fitchburg, WI, USA) and sequenced.

RT-PCR
First strand cDNA was synthesized as described above. PCR was performed on the cDNA with the primers FOXP1/ex7R/ race-2 and PLEKHG1/forward primers to confirm the FOXP1/ PLEKHG1 fusion in case 7 (Table S2).

Western Blotting
Sections from frozen lymph node samples were lysed and processed for Western blotting according to standard procedure using the following antibodies: anti-FOXP1 (1:100 dilution; JC12/ ab32010; Abcam, Cambridge, UK) and anti-beta-actine (1:300 dilution; AC-15; Sigma Aldrich, St. Louis, MO, USA). Protein detection was performed with Image Quant Las4000. The AIDA software (Advanced Image Data Analyzer, version 4.15.025, Raytest GmbH, Straubenhardt, Germany) was used for a densitometric analysis of Western blots.

Library Preparation for Paired-end RNA-sequencing
The Illumina standard kit (IlluminaH TruSeq TM RNA Sample Preparation Kit, San Diego, CA, USA) was used for the mRNAsequencing sample preparation according to the manufacturer's protocol. The quality of the libraries was checked by Agilent Technologies 2100 Bioanalyzer, using the Chip (Agilent DNA 1000 Kit).

Analysis of RNA-sequencing Data
The fastq files were mapped to the reference human genome (Human.B37.3) using the Ensembl gene model (Homo_sapiens.GRCh37.67). The mapping and downstream analysis were performed with the software ArrayStudio, version 6.0 [37]. The mapped reads were used to calculate read counts per gene as well as fragments per kilobase of exon per million fragments mapped (FPKM). The read counts were used as input for the application DEseq [38], specially designed to find differentially expressed genes in RNA-sequencing data and for comparisons of single cases. DEseq returns a fold change in expression of every gene as well as the associated p-value and false discovery rate (FDR). We performed a pair-wise comparison between every sample expressing FOXP1 NT and the sample expressing FOXP1 FL . Additionally, we compared all possible combinations of two samples expressing FOXP1 NT versus the sample expressing FOXP1 FL . The three samples expressing FOXP1 NT were grouped and also compared against the sample expressing FOXP1 FL . In each comparison, we selected the genes with FDR lower than 0.2 and uploaded them to the Ingenuity Pathways Analysis software (IPA, Ingenuity Systems Inc., Redwood City, CA). In IPA, we retrieved genes that were common in the individual DEseq tests. The gene sets were used for an IPA core analysis aimed at identification of affected networks using only the two highest confident levels (experimentally observed data or highly predicted interaction). The RNAsequencing dataset was also explored to detect somatic mutations. Only mutations with a frequency of at least 15% were retrieved. Reads from PCR duplicates and sigletons were not considered in the frequency calculations. SNP variants from DBSBP137 were removed and the following categories of mutations were reviewed: Del 39UTR, Del Non-synonymous, Del-splicing, Non-synonymous, Splicing, StopGain, StopLoss. Mutations were annotated using Polyphen and SIFT prediction algorithms. RNA-sequencing data are available at GEO (Accession number: GSE50514).

Cytogenetic and FISH Analysis
We collected four B-cell lymphoma cases with the FISH proven 3p13/FOXP1 chromosomal aberrations not involving IG loci (further referred to as index cases) (Table 1, Figure 1). Two of them, case 1 with t(2;3)(q36.1;p13) and case 4 with unknown t(non-IG/FOXP1) detected in cases of MZL and non GCB-DLBCL, respectively, were previously reported [33,35]. Two novel aberrations, t(3;10)(p13;q24) and inv(3)(p13q11), were found respectively in a case of CLL at time of Richter transformation (case 2) and in a case of progressive MZL (case 3). The 3p13 rearrangements in index cases were initially characterized by FISH using a set of Bacterial Artificial Chromosome (BAC) clones spanning FOXP1 ( Figure 2, Table S1). In all cases the breakpoints were mapped in the terminal (coding) region of FOXP1 flanked by RP11-135M05 and -79P21 spanning exon 6 and exon 8 onwards, respectively. In contrast, two cases with t(3;14)(p13;q32)/IGH-FOXP1 (cases 5 and 6) included in this study, revealed the 3p13 breakpoints within the 59 untranslated region of FOXP1 flanked by RP11-713J07 and -905F6. Of note, two out of 16 cases with t(FOXP1) studied by Goatly et al. also showed the 3p13 breakpoints within a coding region of FOXP1 [32].
The reciprocal partner breakpoints of cases 1, 2 and 3 were investigated by BAC-walking FISH using sets of probes for 2q36, 10q24 and 3p11-3q13, respectively. Detailed results of FISH analysis are shown in Table S1. Briefly, the 2q36.1 breakpoint of t(2;3)(q36;p13) (case 1) was mapped in the region spanned by RP11-183N07 ( Figure 1b). Of note, this clone harbors the entire AP1S3 gene located in an opposite transcriptional orientation to FOXP1. The 10q24 breakpoint of t(3;10)(p13;q24) (case 2) was narrowed down to the approximately 40 kb area bordered by RP11-346A7 and RP11-2F13 ( Figure 1d). This small region harbors three genes: TMEM180, located in the same transcriptional orientation as FOXP1, and C10orf95 and ACTRIA, both oriented opposite to FOXP1. Notably, the t(3;10)(p13;q24) was found in one of two cytogenetically related subclones which were identified at time of a high grade transformation of CLL (Table 1, 2b), but were not detected in a diagnostic sample characterized by D13q14 (RBI1/D13S319) ( Table 1, 2a). Cytogenetic analysis of case 3 was performed on two diagnostic samples, bone marrow (BM) ( Table 1, 3a) and lymph node (LN) ( Table 1, 3b). The former sample revealed a relatively simple karyotype with t(3;3)(p21;q26) and del(6)(q23). The karyotype of LN was related but more complicated. Among others, it displayed secondary rearrangements of one der(3) (later referred to as inv(3)(p13q11) involving FOXP1 (Figure 1e). The reciprocal breakpoint of inv(3) was investigated by FISH and eventually mapped in the nearcentromeric region at 3q11 flanked by two consecutive probes: CTD-2234G15, which stayed at 3p11 and RP11-778P17, the first available 3q11 probe, which moved to 3p13 ( Figure 1f). Notably, no genes have been mapped to this region of approximately 3.2 Mb annotated as a golden path gap (http://www.ensembl. org). Given that the 3p13/FOXP1 aberration in case 4 was not  identified by cytogenetics, further FISH studies of this case were limited to application of probes flanking the 2q36, 10q24 and 3q11 breakpoints in cases 1, 2, 3, respectively. All these probes, however, showed a normal hybridization pattern in this case. Relevant clinical features of the index cases are presented in Table S3. There were two female and two male patients in age ranging from 50 to 73 years (on average 63). The patients developed MZL, CLL and gastric non-GCB-DLBCL. Both MZLs progressed during follow-up and CLL underwent high grade transformation (Richter syndrome). Three documented patients (cases 1-3) died due to lymphoma-related disease after 2.5-26 months (on average 12) after genetic detection of a 3p13/FOXP1 aberration. The appearance of FOXP1 rearrangements during follow-up (cases 2 and 3) or in a subclone at time of diagnosis (case 4) indicates that non-IG rearrangements of FOXP1 were secondary hits acquired during the clinical course of disease.

FOXP1 Expression by IHC
All four index cases were subjected to IHC with two monoclonal anti-FOXP1 antibodies, ab 16645 from Abcam and SP133 from Roche Diagnostics, both recognizing epitopes located in the Cterminus of FOXP1 protein. We also analyzed selected positive controls (cases 5-14), negative controls (cases 15-16) and three non-malignant (reactive) LN (NL1-3). Examples of IHC are shown in Figure S1. Neoplastic cells of index cases and positive controls revealed a nuclear expression of FOXP1 (Table 1). Of note, gain of FOXP1 tumoral protein expression in case 3 correlated with the acquisition of FOXP1 rearrangement (sample b). Neoplastic cells of FOXP1-negative DLBCL did not express FOXP1, but a proportion of non-malignant cells displayed positive staining. NL1-3 showed FOXP1 positivity in both T-and B-cells in paracortex and lymphocytic corona, and in a small proportion of the germinal center B-cells, as previously described [5,6].

59RACE-PCR
To identify putative partner genes of FOXP1 in the index cases, two available cases (cases 3 and 4) were subjected to 59RACE-PCR. Following QRT-PCR data, we applied primers for exon 7, which was the first commonly upregulated coding exon of FOXP1 in these lymphomas (Figure 3), and expected to identify flanking upstream sequences. The analysis was also performed on eight cases of FOXP1-positive DLBCL/MZL (cases 7-14). After 59RACE-PCR and cloning, up to 44 colonies per case were randomly sequenced. This analysis did not detect any FOXP1 fusion in the index cases, but unexpectedly, identified PLEKHG1 (6q25.1) as a fusion partner of FOXP1 in case 7 ( Figure 4). The fusion, however, occurred out of the reading frame of FOXP1 indicating that the PLEKHG1-FOXP1 rearrangement did not result in a functional chimeric product. As rearrangements of both genes were not demonstrated by FISH (Table S1), the fusion was likely generated by either e.g., a cryptic insertion, or was present in a minor clone. Of note, all cases analyzed by 59RACE-PCR revealed expression of different isoforms of FOXP1 containing exon 7 and various upstream exons. Particularly frequently expressed was transcript or transcripts expressing an alternative exon 6b, which according to Ensembl (Homo_sapiens.GRCh37.67) is shared by FOXP1-009 and -011. This observation was further confirmed by QRT-PCR, which detected a common expression of exon 6bpositive transcripts in the cases analyzed (data not shown).

Western Blot Analysis
Eight cases with available material (cases 3, 5-7, 12, 13, 15 and 16) and three non-malignant LNs were subjected to Western blot analysis with the monoclonal JC12 antibody. As shown in Figure 5a, at least four bands with a differential expression were detected in the samples analyzed: the 75 kDa band corresponding to the FOXP1 FL protein and bands of 64, 60 and approximately 45 kDa. The 75 kDa band was predominantly expressed in both cases with t(3;14)(p13;q32) (cases 5 and 6) and in three nonmalignant LNs. The index case and three FOXP1-positive DLBCLs (cases 7, 12 and 13) showed a marked expression of 64 and 60 kDa isoforms, and a various expression of band of 75 kDa. Case 7 revealed additional bands of 45 kDa and 70-73 kDa, not To evaluate a relative abundance of the full length protein and shorter isoforms in the cases analyzed, we performed at first a densitometric analysis of expressed isoforms (Table S4)

RNA-sequencing Analysis
RNA-sequencing of six FOXP1-expressing lymphomas, which included three cases with FOXP1 rearrangements (case 3 with inv(3), case 5 with t(3;14)(p13;q32) and case 7 with PLEKHG1-FOXP1), and three cases with no apparent structural aberrations of FOXP1 (cases 8, 11 and 12) was performed. In addition, two FOXP1-negative DLBCLs (cases 15 and 16) were also included. The total number of reads produced by sequencing for each sample ranged from 87134076 to 115622676 with a median of 94596970. The percentage of uniquely mapped paired reads ranged from 81 to 90% with a median of 89%. RNA-sequencing detected FOXP1 transcripts in all cases analyzed, but two FOXP1negative lymphoma cases showed a significantly lower overall expression of FOXP1 transcripts than FOXP1-positive tumors (on average 486.3 FPKM vs 1190.9 FPKM) ( Figure S2). RNAsequencing data were initially used to verify IHC subtyping of seven DLBCL cases. We analyzed expression values of the predictor genes reported by Wright et al. [39] to classify DLBCLs into ABC/non-GCB and GCB subtypes. Principal component analysis and hierarchical clustering showed that all six non-GCB-DLBCL cases clustered together and were distinct from the GCB-DLBCL case 15 ( Figure S3). Further analysis of potential chimeric transcripts did not detect any fusion of FOXP1, including  PLEKHG1-FOXP1 previously identified in case 7 by 59RACE-PCR. This finding suggests that the fusion occurred in a minor clone. Mutation analysis identified a wide range of mutations in all analyzed lymphomas, but none targeted FOXP1. Of interest, nonsynonymous mutations of two known DLBCL-related genes, MYD88 and CARD11 [40,41], were detected in 3 cases (no. 3, 7 and 12) and one case (no. 12), respectively (data not shown). Looking for unique mutations in case 5 expressing FOXP1 FL , we found eight genes (excluding IGLV) affected by non-synonymous mutations and 32 genes with Del39UTR (Table S5) Notably, KLF5 was affected by two mutations, S236C and S237, both predicted as deleterious by Polyphen 2 algorithm. IPA analysis of the genes uniquely mutated in case 5 revealed a network of interactions (mostly direct interactions) with several cancer genes including MYC, TP53, BCL2, PI3K and NFkBIA ( Figure S4).
To get insights into transcriptional networks regulated by FOXP1 NT , we ran inference analyses comparing transcriptomes of cases 8, 11 and 12 expressing FOXP1 NT with the transcriptome of case 5 expressing FOXP1 FL , all diagnosed as non-GCB-DLBCL (see Material and Methods). Cases 3 and 7 were excluded from the inference analysis due to lack of an equivalent control sample in the former case, and complex proteomic pattern of the latter case (Figure 5a). Given a weak expression of short FOXP1 isoforms in case 16 (Figure 5c), this sample was not used as a negative control. IPA analysis of the datasets obtained identified 112 genes commonly dysregulated (downregulated) in all seven inference analysis performed (Table S6). The IPA core analysis of these genes identified three top networks comprising 36, 23 and 21 genes, respectively (Table S7). Top diseases and functions of the second most significantly dysregulated network are ''Cell Death and Survival, Cellular Movement, Cancer''. This network, including FOXP1, is shown in Figure 6. It comprises at least three key cancer genes, TP53, CDKN1A and MYC, recently shown to repress microRNA-34a which regulates expression of FOXP1 [42]. Of note, SERPINB5 (maspin) found to be significantly downregulated in cases 8, 11 and 12 versus case 5, is a candidate tumor suppressor in prostate cancer [43]. HIP1R, a postulated target of FOXP1 in ABC-DLBCL was not consistently downregulated in FOXP1 NT -positive DLBCLs. We cannot exclude, however, that other accessory proteins also contribute to FOXP1 repression of HIP1R, as previously argued [44].

Discussion
The work presented here was inspired by our previous finding of the novel FOXP1-related t(2;3)(q36;p13) in case of MZL [35]. In contrast to t(3;14)(p13;q32)/IGH-FOXP1 affecting the 59 untranslated region of FOXP1 [31,34], the 3p13 breakpoint of t(2;3)(q36;p13) was mapped within the coding domain of the gene. Given that MALT1, another oncogene implicated in the pathogenesis of MZL, is involved in two types of translocations, t(14;18)(q32;q21)/IGH-MALT1 [45,46] and t(11;18)(q21;q21) generating the API2(BIRC3)-MALT1 chimeric gene [47,48], we initially hypothesized that t(2;3)(q36;p13) encodes a FOXP1-related fusion transcript [35]. However, the subsequent discovery of a predominant expression of shorter FOXP1 isoforms by FOXP1positive ABC-DLBCL led to the hypothesis that 3p13 translocations targeting the coding region of FOXP1, like t(2;3)(q36;p13), may activate expression of N-terminally truncated isoform(s) of FOXP1 [28,32]. To validate this interesting concept and to decipher the molecular consequences of non-IG aberrations of FOXP1 in general, we performed comprehensive genetic and molecular investigations of four lymphoma cases with non-IG rearrangements of FOXP1 and compared these with cases harboring t(3;14)(p13;q32)/IGH-FOXP1 and FOXP1-positive DLBCL with no apparent structural aberrations of the gene. Our study demonstrated that FOXP1 breakpoints in all index cases fall within the coding region of FOXP1 and unlike the t(3;14)(p13;q32) do not locate to the 59untranslated region of the gene. We also found that non-IG aberrations of FOXP1 promiscuously affect partner sequences at various chromosomal regions (2q26, 3q11, 10q24 and other) targeting either a gene-poor region or gene that is in the incorrect transcriptional orientation to be fused with FOXP1. Therefore these aberrations do not generate functional chimera products. Even the PLEKHG1-FOXP1 fusion identified in case 7 was not functional because it occurred out of the reading frame of FOXP1. To prove that the disruption of the FOXP1 coding sequences by these aberrations activates expression of short FOXP1 isoforms, we performed preliminary transcriptomic and proteosomic investigations of a few available cases. Using QRT-PCR, we showed that the index cases abundantly express sequences encoded by exon 7 onwards, but not by all coding exons of FOXP1 (exon 6 onwards). This finding is in line with their 3p13 breakpoints positioned between exon 6 and 8 of FOXP1. A similar pattern of FOXP1 expression was detected in FOXP1-positive DLBCL without rearrangements of the gene, confirming the previously published data [28]. In contrast, cases with t(3;14)/IGH-FOXP1 which affects the 59untranslated region of FOXP1 expressed all coding exons of the gene. Further evidence was provided by Western blotting which in general, detected three differently expressed FOXP1 isoforms: of 75 kDa, representing FOXP1 FL , and of 64 and 60 kDa, representing shorter FOXP1 isoforms. Notably, the 75 kDa molecule was highly and predominantly expressed in both lymphomas with t(3;14)(p13;q32) and in non-malignant LNs, while the 64 and 60 kDa isoforms were predominantly expressed in FOXP1-positive DLBCLs without FOXP1 rearrangements (Figure 5a). Case 3, the only one case with a non-IG aberration of FOXP1 analyzed, showed a predominant expression of both shorter isoforms as well as a significant expression of protein of 75 kDa. The latter band, however, may represent isoform FOXP1-011 with a molecular weight of 76 kDa, thus indistinguishable from FOXP1 FL . This assumption is supported by the QRT-PCR data showing preferential expression of exon 7 onwards and the 59RACE-PCR which detected a strong expression of transcript variant(s) with an alternative exon 6b which features FOXP1-011 (www.ensembl.org). Generally, our findings are in line with observations of Brown et al. [28], who were the first to demonstrate expression of shorter FOXP1 isoforms (60-65 kDa) in ABC-DLBCL cell lines and primary lymphomas. Identification of the abundantly expressed isoforms, however, is challenging. The 64 kDa band may represent FOXP1-009 (573 aa/64 kDa) which also contains the alternative exon 6b. Ensembl (Homo_sapiens.GRCh37.67), however, does not annotate any isoform of 60 kDa. Further investigations are warranted to clarify whether 64 and 60 kDa bands represent the same isoform which underwent posttranslational modifications, or two different isoforms of FOXP1.
Altogether, our findings support the hypothesis that non-IG aberrations affecting the coding region of FOXP1 activate an aberrant expression of the 59end-truncated variant transcripts of FOXP1. The molecular mechanisms underlying these lesions, however, remain largely unknown. As promoter substitution cannot be the common mechanism operating in these cases, we assume that non-IG rearrangements could either activate alternative internal promoter(s) of FOXP1 (http://www.humangenes.org) or lead to loss of negative regulatory elements, similar like activating rearrangements of LMO2 in T-ALL [49]. Interestingly, an alternative mechanism of generation of N-terminal deletion of foxp1 was reported in myeloblastosis-associated virus type-2induced chicken nephroblastoma [50]. This oncogenic isoform was induced by a retroviral integration in the second coding exon of foxp1, which corresponds to exon 7 of FOXP1, targeted by non-IG aberrations in human lymphoma. Activation of truncated oncoproteins by chromosomal translocations has been previously described in cancer, including B-ALL and prostate tumors harboring promiscuous translocations of PAX5 [51,52] and ETV1 [53], respectively. Molecular pathways affected by FOXP1 NT in lymphoma are largely unknown. Recent studies of Wong et al. [44], however, identified HIP1R as a potential target of FOXP1 in ABC-DLBCL, and ChIP-on-chip analysis of the human GC-like DLBCL cell line OCI-Ly1 reported by Sagardoy et al. [5] identified 279 FOXP1 target genes. IPA analysis of our RNA-sequencing expression data of 4 non-GCB-DLBCL cases detected a set of dysregulated genes in three cases expressing FOXP1 NT when compared with the case expressing FOXP1 FL . These genes, which were exclusively downregulated and code for cytoplasmic, membrane and extracellular molecules, were found to be directly linked to MYC and other key cancer genes. They may comprise direct or non-direct targets of FOXP1 NT, potentially implicated in disease progression. Their identification, however, requires further molecular studies performed on large series of FOXP1-positive lymphomas.
Collectively, we provide evidence that FOXP1 is the target of two molecularly distinct types of rearrangements in B-cell neoplasms: (i) t(3;14)(p13;q32)/IGH-FOXP1, which dysregulates expression of FOXP1 FL , and (ii) non-IG aberrations, which result in ectopic expression of FOXP1 NT . The former translocation is regarded as the primary genetic event, because like other wellknown IGH-mediated translocations in B-cell lymphoma, it occurs as a sole karyotypic aberration and is present in diagnostic samples. Consequently, the aberrantly expressed FOXP1 FL seems to play an oncogenic role in lymphoma. In contrast, non-IG rearrangements of FOXP1 are found as secondary genetic hits acquired during clinical course of various B-cell neoplasms (DLBCL, MZL and CLL), frequently heralding inferior outcome. Therefore, the overexpressed FOXP1 NT isoforms seem to be implicated in disease progression. Our new finding of the FOXP1 NT involvement in high grade transformation of CLL remains in line with the data recently published by Quesada et al. [54]. This group identified a novel C-terminally truncated FOXP1 protein aberrantly expressed in CLL and showed that this isoform is generated by mutated SF3B1 (splicing factor 3b, subunit 1). Mutations of SF3B1 were found in 9.7% of CLLs analyzed by whole-exome sequencing and were associated with an advanced disease at diagnosis and poor overall survival of affected individuals.
Although our study postulates the oncogenic role of FOXP1 FL in lymphomagenesis, constitutive expression of FOXP1 FL in transgenic mice seems to be insufficient to drive tumorigenes [5]. Therefore, we presume that FOXP1, like BCL2 [55,56], may require additional genetic hits to initiate lymphoma. To identify such hit(s), we searched for unique mutations in case 5 with t(3;14)/qFOXP1 FL . Among approximately 40 mutated genes, one gene, KLF5, was targeted by two potentially deleterious mutations. Thus far, the implication of KLF5 in lymphomagenesis is unknown, but this transcription factor is involved in several important biological processes including cell proliferation, transformation, hematopoietic stem cell homing and carcinogenesis [57][58][59].
Altogether, our data support the important role of FOXP1associated rearrangements in development and progression of Bcell lymphoma. Further studies, however, are needed to decipher complexity of FOXP1 and presumably opposing roles of the gene in the pathogenesis of lymphoid and epithelial tumors. Figure S1 Morphology and FOXP1 expression detected by IHC with SP133 antibody in index and control cases.  Figure S3 Performance of GC and non-GCB subgrouping of DLBCL cases subjected to RNA-seq using expression values of predictor genes described by Wright et al. [39]. (PPTX) Figure S4 Interaction network of genes exclusively mutated in t(3;14)-positive case 5 (in grey) with the well know cancer genes specified by IPA. Continuous and discontinues lines indicate direct and indirect interactions, respectively. Red arrows mark genes found to be upregulated in FOXP1 FL expressing case 5 when compared with cases expressing FOXP1 NT . (PPTX)