The myxoid liposarcoma specific TLS-CHOP fusion protein localizes to nuclear structures distinct from PML nuclear bodies


  • Sofia Thelin-Järnum,

    1. The Lundberg Laboratory for Cancer Research, Department of Pathology, Göteborg University, Gothenburg, Sweden
    Search for more papers by this author
    • The first two authors contributed equally to this article.

  • Melker Göransson,

    1. The Lundberg Laboratory for Cancer Research, Department of Pathology, Göteborg University, Gothenburg, Sweden
    Search for more papers by this author
    • The first two authors contributed equally to this article.

  • Alondra Schweizer Burguete,

    1. The Lundberg Laboratory for Cancer Research, Department of Pathology, Göteborg University, Gothenburg, Sweden
    Search for more papers by this author
  • Anita Olofsson,

    1. The Lundberg Laboratory for Cancer Research, Department of Pathology, Göteborg University, Gothenburg, Sweden
    Search for more papers by this author
  • Pierre Åman

    Corresponding author
    1. The Lundberg Laboratory for Cancer Research, Department of Pathology, Göteborg University, Gothenburg, Sweden
    • Lundberg Laboratory for Cancer Research, Department of Pathology, Göteborg University, Sahlgrenska University Hospital, 41345 Gothenburg, Sweden
    Search for more papers by this author
    • Fax: +46-31-827194


CHOP in 12q13, also called GADD153 or DDIT3, encodes a transcription factor of the C/EBP type. As a result of t(12;16) translocations, CHOP is rearranged and fused to TLS in 16p11 in about 90% of myxoid liposarcomas/round cell liposarcomas (MLS/RCLS). The TLS-CHOP protein consists of the N-terminal half of TLS juxtaposed to the N-terminal of the entire CHOP. It is capable of forming dimers with the natural dimer partners of CHOP. Here we report that recombinant TLS-CHOP-green fluorescence protein localizes to nuclear structures, similar to, but distinct from, PML nuclear bodies. The TLS-CHOP-green fluorescent protein nuclear structures are resistant to high salt concentration and nuclease treatment. Transfection of TLS-CHOP to normal fibroblasts causes a rapid down regulation and relocation of PML nuclear bodies. An abnormal extra nuclear localization of PML bodies was also found in TLS-CHOP carrying cell lines established from myxoid liposarcomas. Transfection of TLS-CHOP induced a rapid disappearance of PCNA. TLS-CHOP may disturb the nuclear machinery by binding and sequestering important factors from their natural sites. © 2001 Wiley-Liss, Inc.

The C/EBP Homologous Protein “CHOP”, also named GADD 153 or DDIT3, is a stress-induced basic leucine zipper type of transcription factor that forms DNA binding dimers with C/EBPβ.1CHOP, located in 12q13, is rearranged as a result of t(12;16) translocations found in about 90% of myxoid liposarcomas/round cell liposarcomas (MLS/RCLS)2 and fused to TLS, also called FUS, in 16p11.3, 4TLS encodes an RNA binding protein with extensive similarities to EWS, which is rearranged in Ewing as well as other sarcomas.5 In about 10% of MLS/RCLS, EWS replaces TLS as a fusion partner to CHOP.6 The fusion gene TLS-CHOP encodes a protein consisting of the N-terminal half of the TLS protein linked to 26 amino acids from the normally non-transcribed CHOP exon 2 and parts of exon 3, followed by the entire CHOP protein. The TLS-CHOP protein lacks the RNA binding parts of TLS, which are replaced by the DNA binding and leucine zipper parts of CHOP. TLS-CHOP localizes to the nucleus of MLS/RCLS cells and it maintains the capacity to form dimers with other C/EBP zipper regions.7 These observations suggest that TLS-CHOP may act as an abnormal transcription factor and several studies using experimental reporter systems have shown that the N-terminal half of TLS acts as a strong transcriptional activator when fused to different transcription factors.8, 9 We have constructed TLS-CHOPGFP (green fluorescent protein), TLSGFP and CHOPGFP expression vectors in order to study the cellular localization and effects of TLS-CHOP, TLS and CHOP in live human cultured cells. The cellular distribution of the fusion protein and its normal counterparts differs dramatically. The abnormal localization may be important for the oncogenic activities of TLS-CHOP.


Expression vectors

PCR fragments containing the full length coding regions of TLS-CHOP type II,10TLS or CHOP were generated with primer pairs EcoTLSF: GTTGGAATTCGTTGCTTGCTT, BamCHOPR: TATGGATCCGCTTGGTGCAGATTCA, TLS35FXhoI: TATACTCGAGCGCGGACATGGCCTCAAACG, BamTLSR: TATGGATCCATACGGCCTCTCCCTGCGATCCTG, and EcoCHOPF: TGTTGGAATTCCAGGAGAATGAAAGGAAAGTGG, CHOPBamHIL: TATGGATCCGCTTGGTGCAGATTCA, respectively, containing XhoI or EcoRI or BamHI sites in their 5′ ends. The PCR-products were digested with the selected restriction enzymes and cloned into the pEGFP-N1, pIRESGFP or pDsRed1-N1 vectors (Clontech, Palo Alto, CA) in frame with the fluorescent protein sequences. The constructs were transferred to and grown in bacteria and plasmids were purified using a kit for endotoxin free plasmid preparations (Qiagene, Chatsworth, CA). Constructs used for transfection experiments were sequenced to exclude mutated clones.

Western blot analysis

Cell extracts were prepared from cultured cells in the presence of a protease inhibitor mixture (Roche 1836153, Bromma, Sweden). The proteins were size fractionated on PAGE gels11 and electro transferred to PVDF membranes. The TLS, CHOP and GFP fusion proteins were visualized by incubation of the membranes with specific antibodies followed by an alkaline phosphatase coupled secondary antibody and development in NBT/BCIP substrate solution.

Cell culture and transfection

All cultured cells, human passage 8-16 skin fibroblasts, fibrosarcoma line HT1080, MLS cell lines 402–91, 1765–92 and 2146–942 and COS-1 cells,12 were kept frozen in liquid nitrogen or cultivated in RPMI 1640 with 8% fetal calf serum. Cultures were split 1:3 every week. The cells were harvested by trypsination, and 3 × 105 cells were seeded into each flaskette and transfected the following day using 10 μl of FuGENE™6 transfection reagent (Roche) and 3 μg of plasmid DNA per flaskette.

Immunofluorescence analysis

The cultures were washed twice with PBS and fixed in 4% paraformaldehyde in PBS (for staining of Ki67 and PML) or in methanol (for staining of PCNA). After 2 more washes in PBS, the slides were stained with antibodies specific for Ki67; M7187 1:20, PCNA; M 0879 1:20, or PML; M7207 1:25, all from Dako A/S, Denmark, with Cy3 conjugated secondary antibodies (Pharmacia Amersham, Gaithersburg, MD). Before staining of the TLS-CHOP nuclear structures, the cells were permeabilized for 5 min in ice-cold permeabilization buffer (PB) (20 mM NaH2PO4, 150 mM NaCl, 0.5% Triton X-100 (Sigma Chemical Co., St. Louis, MO) and 5 mM MgCl2), incubated 5 more min in cold PB with 500 mM NaCl and followed by a wash in PB to reduce salt strength. TLS-CHOP, TLS and CHOP proteins were detected using 2 rabbit anti sera, GADD 153 (R20) and GADD 153 (490), from Santa Cruz Biotechnology (Santa Cruz, CA) used at 1:100 in PB, and a rabbit anti serum developed against the 18 N-terminal a.a of TLS. The slides were mounted in an anti fade mount containing the DNA binding dye DAPI (4,6-diamidino-2-phenylindole, dihydrochloride). For analysis of PML nuclear body distribution, nuclear bodies were counted and their cellular localization was noted in 50 cells from each cell line.

In situ protein extraction experiments

Transfected monolayer cells in flaskettes were washed 2 times in ice-cold PBS and permeabilized by adding ice-cold PB containing 50 mM NaCl, 5 mM MgCl2, 0.5% triton X 100 (Merck, Darmstadt, Germany), 10 mM Tris pH7.2 and protease inhibitor mix (Roche 1697498). The cells were exposed to increasing salt concentrations of 150, 200, 300, 600 and 1,000 mM in PB and to 20 mM EDTA in the same buffer but without MgCl2. RNase, 10 μg/ml (ROCHE number 1 119 915) or RNase free DNase (Amersham-Pharmacia) 100 U/ml treatment was made in PB at 37°C for 10 min. DAPI staining of the nuclei was followed in microscope to monitor the DNA-digestion. The permeabilized and salt/enzyme treated cells were inspected on an inverted fluorescence microscope. For Western blot analysis, cells were seeded out and transfected in parallel flaskettes. The cells were harvested for SDS gel electrophoresis after permeabilization and salt extraction in 150 and 600 mM salt. All extraction experiments were repeated at least 2 times with identical results.


TLS-CHOPGFP protein localizes to distinct nuclear structures

Live CHOPGFP transfected fibroblasts showed a smooth or fine grainy nuclear fluorescence (Fig. 1a). TLSGFP was found both in the nuclei and cytoplasm of transfected fibroblasts, with a weaker staining of the cytoplasm. In some cells, brighter regions on the nucleo-cytoplasmic borders were seen together with a granular staining in the cytoplasm (Fig. 1b).

Figure 1.

(a) CHOPGFP in the nuclei of live human fibroblasts. (b) TLSGFP in live human fibroblasts. (c) TLS-CHOPGFP in nuclei of live human fibroblasts. (d) TLS-CHOPGFP (green) and anti PML Cy3 staining (red) in fixed human fibroblasts. (e) Anti PML Cy3 staining in fixed myxoid liposarcoma cells 402–91. DAPI DNA staining in blue. (f) TLS-CHOPGFP and anti CHOP immunostaining in nuclei of TLS-CHOPGFP transfected HT1080 fibrosarcoma cells. Top: TLS-CHOPGFP (green) Bottom: The same 2 nuclei with Anti-CHOP Cy3 (red). (g–i) Anti-CHOP Cy3 staining in nuclei of fixed normal fibroblasts (control, g), myxoid liposarcoma cell lines 402–91 (h) and 1765–92 (i). DAPI DNA staining in blue.

Live TLS-CHOPGFP transfected fibroblasts showed a strong nuclear expression of the GFP fusion protein. In most cells the protein localized to between 5 and 15 small distinct brightly fluorescent structures. A proportion of the transfected cells exhibited, in addition to the structures, a weaker, smooth nuclear staining (Fig. 1c). Staining of TLS-CHOPGFP transfected HT1080 cells with 2 different CHOP antibodies showed a similar pattern (Fig. 1f). A similar staining pattern but with only 1–6 smaller structures per cell was seen in myxoid liposarcoma 402–92 and 1765–92 stained with antibodies against CHOP (Fig. 1gi). Several fixation protocols had to be tested to find conditions for antibody detection of the TLS-CHOPGFP nuclear structures. The anti-TLS serum failed to stain the TLS-CHOP nuclear structures in any cell type.

Western blot analysis of transfected COS cells showed that the cells contained proteins of the expected sizes for GFP-fusions (GFP and linker adds approximately 30 kDa to the natural protein weights) with no evidence for accumulation of GFP-containing degradation products (Fig. 2).

Figure 2.

Western blot analysis of COS cells transfected with CHOPGFP, TLSGFP and TLSCHOPGFP and developed with anti CHOP + anti TLS (a) and with anti GFP (b,c). Positions of sizemarkers are given in kDa. The natural TLS protein band is indicated by an asterix. GFP and linkers add approximately 30 kDa to the weights of the natural proteins Lanes for (a) and (b) are 1, Control cells; 2, TLS-CHOPGFP; 3, CHOPGFP and 4, TLSGFP tranfected cells. (c) Salt extraction of recombinant proteins. Protein preparations from TLSGFP, CHOPGFP and TLS-CHOPGFP transfected cells, extracted with 150 or 600 mM salt. C, Controls are from transfected untreated cells.

TLS-CHOP structures are resistant to salt extractions and nucleases

As judged by fluorescence microscopy, CHOPGFP disappeared from the nuclei as soon as they were permeabilized in 150 mM salt. However, Western blot analysis showed that small amounts of CHOPGFP remained after 150 mM salt extraction (Fig. 2c). TLSGFP was also washed out of the cells at low salt concentration as observed by both fluorescence microscopy and Western blot analysis. In contrast, as shown by fluorescence microscopy and Western blot analysis, the TLS-CHOPGFP nuclear structures resisted increasing salt washes up to 600 mM NaCl (Fig. 2c). Furthermore, fluorescence microscopy studies showed that RNase, DNase or EDTA treatment had no effect on the TLS-CHOPGFP structures. Some DNase-treated cells exhibited remaining DAPI stained heterochromatin but no DNA remained within the TLS-CHOPGFP structures.

PCNA is downregulated in TLS-CHOPGFP transfected fibroblasts

Immunostaining showed no, or a very weak, PCNA expression in most TLS-CHOPGFP positive cells (Table I). Only very few TLS-CHOPGFP positive cells expressed traces of PCNA. CHOPGFP expressing cells exhibited no change in PCNA staining compared to control cells in the same culture.

Table I. PCNA Expression in TLS-CHOPGFP Positive Cells
Experiment PCNA positive/counted cellsPercent PCNA positive cells
  • 1

    Counted control cells were TLS-CHOP negative cells in TLSCHOPGPF transfected cultures.

  • 2

    TLSCHOPGFP and PCNA double positive cells.

  • 3

    CHOPGFP and PCNA double positive cells.

2Positive cells1/88212
2Positive cells65/743873

Approximately 80% of the TLS-CHOPGFP expressing cells stained for the Ki67 antigen (data not shown). The Ki67 antibody also stained 80% of the TLS-CHOPGFP negative cells. Similar results were obtained in CHOP-GFP expressing cells. We failed to find any dividing cells among the TLS-CHOP expressing transfectants using DAPI staining of fixed cells or by following the development of live TLS-CHOPGFP positive cells.

TLS-CHOPGFP nuclear structures are distinct from PML nuclear bodies

Staining of fixed TLS-CHOPGFP expressing cells with antibodies specific for the PML protein showed that PML and TLS-CHOPGFP are located in different but similar sized structures (Fig. 1d). The number of PML staining structures was lower in TLS-CHOPGFP expressing cells than in the surrounding nontransfected cells and the PML nuclear bodies were found at or close to the nuclear borders in the transfected cells. This observation prompted us to investigate the PML staining pattern in MLS cell lines. Analysis of 3 MLS cell lines showed that they contained in average 6–12 PML nuclear bodies/cell compared to 20/cell in cultured fibroblasts (Fig. 3a). Between 20 and 35% of the anti PML staining bodies localized outside the nuclei of the tumor cell lines, whereas the cultured fibroblasts carried only 2% extranuclear anti PML staining bodies (Figs. 1e and 3b).

Figure 3.

Number and cellular distribution of PML nuclear bodies in MLS-lines. (a) Average number of anti PML staining bodies/cell in the indicated MLS cell lines. Control is normal cultured fibroblasts. Standard deviation is shown as solid bars. (b) Percent extra nuclear anti PML staining bodies in the indicated MLS cell lines. Control is normal cultured fibroblasts.


The TLS-CHOP fusion gene and its product are found in myxoid/round cell sarcomas.2–4 Its causative role in the development of MLS/RCLS is supported by the recent successful establishment of a TLS-CHOP transgenic mouse model.13 The fusion protein is a nuclear protein and believed to act as an abnormal variant of the CHOP transcription factor. It forms dimers with the natural CHOP partners within the C/EBP family of transcription factors.3, 14

The methods used in the present study allowed for direct observation of CHOPGFP, TLSGFP and TLS-CHOPGFP transfected cells and the fusion proteins in live or fixed cells. The CHOPGFP fusion protein showed an even nuclear distribution with a lower degree of staining in the nucleoli. Similar observations have previously been reported using antibody staining of fixed cells,1 indicating that its localization is maintained after the addition of GFP to the C-terminal of CHOP. TLS was reported to be an RNA binding protein that shuttles between the nuclei and cytoplasm.15, 16 In accordance with this view, TLSGFP transfected cells exhibited fluorescence both in nuclei and cytoplasm.

In contrast to the smooth distribution patterns of CHOPGFP and variable cytoplasmic and nuclear pattern of TLSGFP, most of the fluorescence signal from TLS-CHOPGFP was found in distinct nuclear structures. An identical pattern was seen using a TLS-CHOPRFP (red fluorescent protein) vector (data not shown). The RFP and GFP proteins are structurally unrelated, suggesting that the GFP or RFP has no influence on the localization of the recombinant GFP or RFP fusion proteins.

Immune staining of TLS-CHOPGFP transfected cells with CHOP binding antibodies showed an identical distribution. Antibody staining of the TLS-CHOP nuclear structures was highly dependent on the fixation protocol, suggesting a low accessibility of the antigen. Our attempts to stain the TLS-CHOP structures were frustrated before we discovered that salt extractions of the cells unmasked the epitopes.

Using the same staining protocol as for transfected cells, we found that the number of TLS-CHOP structures is lower, typically 1–6 in MLS cell lines, compared to the typically 5–15 structures seen in TLS-CHOPGFP transfected cells. The structures are also smaller in MLS cell lines compared to transfected cells. This difference between TLS-CHOPGFP transfected cells and MLS cells is most probably a result of the very strong expression of TLS-CHOP in the transfected cells. We conclude that TLS-CHOP is distributed partially in nuclear bodies and partially as a more evenly staining protein in the nuclei of MLS-cells.

Our antibody against the TLS N-terminal failed to stain the nuclear structures. This could indicate that the nuclear structures contain deposits of partially degraded proteins lacking the TLS epitopes. Western blot analysis of transfected cells with GFP, CHOP and TLS antibodies showed, however, no signs of degradation products. Therefore, our results suggest that the N-terminal TLS epitopes of TLS-CHOP are hidden also after high salt wash extractions. Judging from their extremely bright fluorescence emission, the structures contain large amounts of TLS-CHOPGFP, perhaps in dense structures with poor penetration possibilities for antibodies.

The TLS-CHOP structures were unchanged after DNase or RNase treatment and after incubation in high salt concentrations and in the presence of EDTA. This suggests that DNA or RNA is not needed to keep these structures together.

Previous studies have suggested that TLS-CHOP is localized to the nucleoli in cells treated with transcriptional inhibitors.15, 16 In the present study of live cells with active transcription machinery, we found, however, that the TLS-CHOPGFP is not localized to the nucleoli but to distinct nuclear structures. The partially contradictory results may depend on the different methodology, treatment of cells and reagents used for detection of the TLSCHOP protein.

The PML protein is the N-terminal fusion partner of the PML-RARA protein product found in promyelocytic leukemia with t(15;17).17, 18 PML is localized to nuclear bodies resembling those seen in TLS-CHOPGFP transfected cells. Staining of transfected cells with PML specific antibodies showed no overlap between PML staining and TLS-CHOPGFP, indicating that PML and TLS-CHOP are localized to different structures. We also noted that the number of PML nuclear bodies per cell was lower than in control cells. Furthermore, MLS cells carried anti-PML stained bodies not only in the nuclei but also in the cytoplasm, suggesting a disturbance of the PML protein localization in TLS-CHOP carrying cells. PML is considered to act as a tumor suppressor, interacting with TP53, P105RB and proteins important for genomic stability. It is also involved in control of apoptosis and transcriptional regulation.19 The TLS-CHOP induced abnormal PML nuclear body localization may therefore be an important part of the transforming mechanism in MLS/RCLS.

The KI67 and PCNA antibodies, both of them specific for proliferation associated nuclear antigens,20, 21 showed a divergent pattern in TLS-CHOP transfected cells. While the KI67 was maintained, the PCNA antigen disappeared. Our results suggest that the disappearance of PCNA was very rapid, suggesting that a protein degrading factor was activated. The PCNA was reported to be a part of the replication complex, and the rapid down regulation of PCNA may indicate an interaction between this complex and TLS-CHOP.

Staining of PCNA in MLS cell lines showed that these proliferating cells have normal PCNA patterns (data not shown). The observed difference in PCNA expression between MLS cells and TLS-CHOPGFP transfected cells may be a result of the very strong expression of TLS-CHOP in the transfected cells.

TLS-CHOP is not compatible with growth or survival of human fibroblasts when expressed at high levels after transfection with CMV-promoter driven expression vectors. Similarly, MLS-cells that carry the TLS-CHOP dies when transfected with TLS-CHOPGFP, TLS-CHOPRFP or a pIRESTLS-CHOPGFP vector (data not shown). Transfection with the latter vector results in GFP labeled cells that express the normal sized TLS-CHOP protein. The reason for the toxicity could be that the TLS-CHOP structures may attract TLS or CHOP binding proteins from their normal localization. For example, CHOP was reported to bind and form dimers with CEBPβ1 and ATF322 and the TLS N-terminal was found to bind several steroid receptor proteins.23 In MLS/RCLS that carry the TLS-CHOP fusion gene, the amount of TLS-CHOP may be expressed at moderate levels that causes disturbances but allows the nuclear machinery to work despite the presence of the fusion protein. The small size and the low numbers of TLS-CHOP nuclear structures found in MLS/RCLS cells are in accordance with this view.