DDIT3/CHOP and the sarcoma fusion oncoprotein FUS-DDIT3/TLS-CHOP bind cyclin-dependent kinase 2

Background The DDIT3 gene encodes a transcription factor belonging to the CCAAT/enhancer binding protein (C/EBP) family. It is normally expressed at very low levels but is activated by cellular stress conditions and induces G1 arrest and, in some cell types, apoptosis. DDIT3 is found as a part of the fusion oncogene FUS-DDIT3 that is causal for the development of myxoid/round-cell liposarcomas (MLS/RCLS). Results In the present study, we searched for putative interaction partners of DDIT3 and the oncogenic FUS-DDIT3 among G1 cyclins and cyclin-dependent kinases. We found that FUS-DDIT3 and the normal DDIT3 bind CDK2. In addition, CDK2 showed an increased affinity for cytoskeletal proteins in cells expressing FUS-DDIT3 and DDIT3. Conclusions We conclude that DDIT3 binds CDK2 and that many of the observed biological effects of DDIT3 may involve interaction with CDK2.


Background
DDIT3 (GADD153, CHOP) encodes a transcription factor belonging to the CCAAT/enhancer binding protein (C/ EBP) family [1]. DDIT3 is normally transcribed at very low levels but is elevated upon DNA damage in cellular stress conditions [2][3][4][5][6]. The DDIT3 protein has a central role in endoplasmatic reticulum stress and DNA damage response by inducing cell cycle arrest and apoptosis [7,8]. DDIT3 has recently been implicated in the stress response leading to death of pancreatic insulin producing β-cells [9] and it may also be a part of cellular stress conditions causing neurodegenerative disorders [10]. In addition, DDIT3 is believed to be involved in growth cessation and terminal differentiation of lipoblasts, osteoblasts and erythrocytes [11][12][13]. DDIT3 forms heterodimers with several other C/EBP family members [14] as well as other leu-cine zipper carrying proteins [15,16] and the heterodimers are believed to act as dominant negative inhibitors of transcription [14].
DDIT3 is also a part of a fusion oncogene critical for the development of myxoid/round cell liposarcoma (MLS/ RCLS) [17,18]. The tumor cells carry the chromosomal translocation t(12;16)(q13;p11) that results in fusion of DDIT3 to FUS (also called TLS) or more rarely the t(12;22)(q13;q12) that fuses DDIT3 to EWSR1 [19][20][21]. The chimerical FUS-DDIT3 oncoprotein functions as an abnormal transcription factor [22] and localizes to distinct nuclear structures in cultured cells [23,24]. MLS/ RCLS cells exhibit abnormal expression profiles of cell cycle controlling factors and among them cyclins and cyclin-dependent kinases (CDKs) [25]. In addition, the DDIT3-binding C/EBPα has been shown to interact and inhibit the kinase activity of CDK2 and CDK4 [26]. Based on these observations, we searched for putative interaction partners of the oncogenic FUS-DDIT3 with G1 cyclins and CDKs as these might provide insight into the molecular mechanisms by which the chimerical oncoprotein induces malignancy.

CDK2 and cyclin E colocalize with FUS-DDIT3
HT1080 cells were transiently transfected with a FUS-DDIT3-GFP construct and colocalization between the ectopically expressed fusion oncoprotein and endogenous cyclin D1, cyclin E, CDK2 and CDK4 was investigated by immunofluorescence. Cyclin E and CDK2 showed prominent colocalization with the FUS-DDIT3 protein and were detected in FUS-DDIT3 containing granules in the majority of cells ( Figure 1). We found no signs of colocalization between FUS-DDIT3 and CDK4 ( Figure 1) or cyclin D1 (not shown). Cells transfected with the empty GFP vector showed no granules and a smooth nuclear distribution of CDK2 and cyclin E ( Figure 1).

FUS-DDIT3 binds CDK2 through its DDIT3 part
Cells were transfected with FUS-DDIT3-GFP and cellular proteins were immunoprecipitated with GFP antibodies and further analyzed by western blot. In pilot experiments, we used HT1080 cells stably expressing FUS-DDIT3-GFP and could weakly detect endogenous CDK2 in immunoprecipitates while cyclin E was not found (data not shown). To further confirm a possible interaction between FUS-DDIT3 and CDK2, we instead transiently transfected cells using cloned CDK2 cDNA expressed in frame with the fluorescent protein DsRed1. Cells co-transfected with CDK2-DsRed1 and either of FUS-DDIT3 or DDIT3 expressing constructs showed presence of CDK2-DsRed1 in anti-GFP immunoprecipitates from both FUS-DDIT3 and DDIT3 transfected cells (Figure 2a). In contrast, CDK2-DsRed1 was not found in immunoprecipitates of cells transfected with the N-terminal part of FUS present in the FUS-DDIT3 protein or in GFP-transfected control cells (Figure 2a). In a reverse experiment, HT1080 cells were transfected with FUS-DDIT3, FUS-DDIT3 LZ, DDIT3 and GFP constructs and cellular proteins in extracts were immunoprecipitated with anti CDK2 antibodies ( Figure 2b). We found that endogenous CDK2 coimmunoprecipitated with DDIT3, FUS-DDIT3 and FUS-DDIT3 lacking a leucine zipper domain. These results indicate that the region binding CDK2 is located N-terminally of the leucine zipper part of DDIT3. ClustalW alignment performed between DDIT3 and the related C/EBPα showed conservation of several amino acids in distinct regions shared by the two proteins but that DDIT3 lacks a region similar to the one that binds CDK2 in C/EBPα [26] ( Figure 2c). Figure 1 CDK2 and cyclin E colocalize with FUS-DDIT3 in transiently transfected HT1080 cells. The endogenous distribution of CDK2 and cyclin E seen in red is detected by antiserum specific for these proteins. Ectopically expressed GFP-tagged FUS-DDIT3 is shown in green. The DAPI dye is used to stain nuclei blue. CDK4 was not seen to accumulate in FUS-DDIT3-containing granules and no granules were formed in cells overexpressing the GFP protein only. Bars indicate 5 μm.

CDK2 expression, phosphorylation status and turnover is not changed by FUS-DDIT3
The CDK2 expression in HT1080 cells after 42 hours of transfection with FUS-DDIT3-GFP or GFP was investigated by western blot analysis ( Figure 3). In addition, the phosphorylation status of CDK2 was analyzed using CDK2 antibodies targeting either the two inhibitory phosphorylations at threonine 14 and tyrosine 15 or the activating phosphorylation at threonine 160 ( Figure 3). No apparent difference in CDK2 phosphorylation between cells expressing FUS-DDIT3-GFP or GFP was detected and the amount of CDK2 between the two transfected cell populations was equivalent ( Figure 3). The CDK2 protein half-life did not differ between stably transfected FUS-DDIT3 cells and HT1080 control cells upon a six hour cycloheximide chase assay (not shown).

Discussion
In the present study, we used immunofluorescence microscopy to investigate putative colocalization between FUS-DDIT3 and cyclins/cyclin-dependent kinases involved in G1 cell cycle control. CDK2 and the CDK2 binding cyclin E were found to change their localization in FUS-DDIT3 expressing cells to a pattern that was identical to the FUS-DDIT3 nuclear granules reported earlier [24]. This suggests that CDK2 and cyclin E are translocated to nuclear structures defined by the FUS-DDIT3 protein. As the Expression and phosphorylation status of CDK2 Arrows indicate the differentially immunoprecipitated vimentin bands present in lanes of DDIT3 and FUS-DDIT3 cells but absent in that of GFP cells. These bands migrate slightly slower than the thicker bands inside the same region. The thicker bands were identified as rabbit heavy chain immunoglobulin molecules and these originate from the antibodies used for immunoprecipitation of CDK2 and are present in all three lanes.
DDIT3 part is considered a DNA-binding transcription factor [27], we speculate that the FUS-DDIT3 defined structures may contain active chromatin and mRNA at different stages of processing. Thus, the accumulation of CDK2 and cyclin E to such nuclear regions may result in changed phosphorylation patterns and regulation of substrates present at these foci.
DDIT3 is able to form DNA-binding heterodimers with C/ EBPα through its leucine zipper region [14] and C/EBPα is reported to bind CDK2 and inhibit its kinase activity [26]. Therefore, it is possible that the binding between DDIT3 and CDK2 is mediated through C/EBPα or another C/EBP protein. However, we here show that a DDIT3 mutant lacking the leucine zipper domain binds CDK2, which implies a binding between DDIT3 and CDK2 that is independent of other C/EBP proteins. DDIT3 and C/EBPα contain several regions with sequence similarities but the proline-histidine rich part reported to bind CDK2 in C/EBPα [26] (Figure 2c) is not present in DDIT3. This suggests that DDIT3 binds CDK2 through a different mechanism than C/EBPα.
Analyses of GFP-immunoprecipitates failed to detect cyclin E in these samples. The absence of cyclin E in CDK2 immunoprecipitates bears resemblance to previous reports showing that C/EBPα disrupts CDK2/cyclin complexes leading to growth arrest [26]. Hence, it is possible that the DDIT3 bound to CDK2 disrupts CDK2/cyclin E complexes in similar way as C/EBPα. Consequently, since cyclin E is a major regulator of CDK2 kinase activity in G1, the DDIT3 binding may alter CDK2 activity in DDIT3 and FUS-DDIT3 expressing cells. We did however not detect a change in phosphorylation status of CDK2 in FUS-DDIT3 expressing cells compared to control cells. To further analyze the functional effects of DDIT3/FUS-DDIT3 binding to CDK2, we immunoprecipitated CDK2 in cells transiently transfected with FUS-DDIT3, DDIT3 and GFP constructs. Analysis of the precipitates revealed enhanced binding of CDK2 to components of the cytoskeleton in cells expressing FUS-DDIT3 and DDIT3. An increased affinity for cytoskeletal components and crosstalk between cell cycle proteins and cytoskeletal regulatory proteins could lead to changes in cytoskeleton structure [28].

Conclusions
In conclusion, we show that CDK2 is translocated to nuclear structures defined by the FUS-DDIT3 oncoprotein and that it binds the DDIT3 part of the chimera. Cyclin E is also recruited to FUS-DDIT3 nuclear structures but can not be found in CDK2-containing DDIT3/FUS-DDIT3 precipitates. The interaction of FUS-DDIT3 and DDIT3 with CDK2 appears to alter the binding affinity of CDK2, possibly leading to changed phosphorylation patterns and regulation of cytoskeletal or other proteins. Many of the observed biological effects of DDIT3 may involve the interaction with CDK2.

ClustalW alignment
The amino acid sequences of DDIT3 and C/EBPα were aligned using the ClustalW algorithm available at the Universal protein resource (UniProt) [32].

Authors' contributions
PÅ conceived of the study. CB, MKA and PÅ designed the experiments. CB and MKA performed research. CB, MKA and PÅ wrote the manuscript. All authors read and approved the final version of the manuscript.