Sumoylation Inhibits the Growth Suppressive Properties of Ikaros

The Ikaros transcription factor is a tumor suppressor that is also important for lymphocyte development. How post-translational modifications influence Ikaros function remains partially understood. We show that Ikaros undergoes sumoylation in developing T cells that correspond to mono-, bi- or poly-sumoylation by SUMO1 and/or SUMO2/3 on three lysine residues (K58, K240 and K425). Sumoylation occurs in the nucleus and requires DNA binding by Ikaros. Sumoylated Ikaros is less effective than unsumoylated forms at inhibiting the expansion of murine leukemic cells, and Ikaros sumoylation is abundant in human B-cell acute lymphoblastic leukemic cells, but not in healthy peripheral blood leukocytes. Our results suggest that sumoylation may be important in modulating the tumor suppressor function of Ikaros.


Introduction
Sumoylation is a post-translational modification that involves the conjugation of small ubiquitin-like modifiers (SUMO1-3 in mammals) to target proteins. SUMO proteins function by modulating the activity and processes of target proteins, such as nuclear localization, transcriptional regulation and protein stability [1][2][3]. Indeed, sumoylation has been shown to modulate the function of transcription factors [4][5][6]. SUMO targets usually contain a consensus sumoylation CKxE/D motif, where C is a hydrophobic amino acid [7].
The Ikaros zinc finger transcription factor is important for multiple aspects of hematopoiesis. Ikaros has been shown to act both as a transcriptional repressor and activator, by interacting with chromatin remodeling complexes like NuRD, PRC2 or SWI/SNF [8][9][10]. However, it remains largely unknown why Ikaros activates some genes and represses others. A potential mechanism may involve post-translational modifications. Indeed, phosphorylation has been shown to be important for Ikaros function in several systems [11][12][13][14]. Ikaros has also been reported to be sumoylated, and sumoylation has been proposed to prevent Ikaros from functioning as a repressor by preventing its association with transcriptional co-repressors [15].
Here we investigated the nature and function of Ikaros protein modifications in lymphocytes. We show that sumoylation is a major post-translational modification and identify three essential lysines in this process. Nuclear localization and DNA binding are required for sumoylation, and sumoylation reduces the ability of Ikaros to inhibit cell proliferation. Finally, we show that human leukemic cells exhibit high levels of sumoylated Ikaros.

Microarray analysis
Total RNA was extracted with the RNeasy Micro kit (Qiagen), and 150 ng was used for transcriptome analysis on GeneChip 1 Mouse Gene 1.0 ST arrays (Affymetrix) using standard procedures. Data were normalized with the Robust Multiarray Average algorithm. Probe sets that did not correspond to an identified gene were not included for analysis.

Antibodies
Mouse monoclonal anti-estrogen receptor (ERa-F3) and rabbit polyclonal anti-Ikaros (Cterminal) antibodies were generated by the antibody facility of the IGBMC. Rabbit polyclonal anti-Ikaros (N-terminal) antibody (ab26083) was purchased from Abcam. Polyclonal rabbit anti-SUMO-1 and monoclonal rabbit anti-SUMO2/3 (18H8) antibodies were from Cell Signaling. Goat anti-mouse and goat anti-rabbit antibodies coupled to horseradish peroxidase were from Life Technologies.

Western blotting
Protein extracts were mixed with 5x sample buffer to a final volume of 50 μl and loaded on SDS polyacrylamide gels (6, 8 or 10%) or NuPAGE Novex 3-8% Tris-Acetate gels (Invitrogen). Proteins were transferred to PVDF membranes (Immobilon P IPVH00010, Millipore), saturated with 5% non-fat milk for 1 hour at room temperature (RT) and probed overnight at 4°C with the indicated antibodies. Membranes were then washed three times (TBS; 0.1% Tween) and incubated with the appropriate secondary antibody in 5% non-fat milk for 1 hour at RT. After three washes, proteins were revealed using Western ECL Substrate (from Biorad or Pierce). To detect sumoylated proteins on membranes previously probed with anti-Ikaros or anti-ER antibodies, membranes were stripped in 62.5mM Tris pH 6.8, 100mM β-Mercaptoethanol, 2% SDS for 30 min at 50°C with agitation, washed three times, re-saturated and reprobed with anti-SUMO antibodies.
Immunoprecipitation 500 μl of total protein extract prepared from 50x10 6 cells were pre-cleared with 20 μl of 50% slurry protein A or protein G sepharose beads (Sigma) for 30 min at 4°C with rotation. Beads were pelleted by centrifugation for 1 min at 13 000 rpm (4°C) and extracts were transferred into tubes containing 100 μl of 50% slurry protein A or protein G sepharose beads and incubated overnight at 4°C with 5 μg of the indicated antibody. Beads were then spun (3 min, 2000 rpm, 4°C) and washed 5 x with RIPA lysis buffer (300 mM NaCl) without inhibitors. Beads were then boiled in 40 μl of 3x Laemmli sample buffer and loaded on the gel.

Luciferase reporter assay
The luciferase reporter assay was performed as described previously [20].

Ikaros proteins are modified in normal and leukemic T cells
To study Ikaros modifications in developing T cells, we analyzed the relative size of these proteins in total extracts of CD4 -CD8 -CD25 + CD44 -(DN3), CD4 -CD8 -CD25 -CD44 -(DN4) and CD4 + CD8 + (DP) thymocytes by Western blot (Fig 1a). Besides the Ikaros1 (Ik1) and Ikaros2 (Ik2) isoforms, three proteins of higher molecular (MW) weight (I-III) were detected. These latter bands were detected in all three thymocyte subsets but were most abundant in DN4 cells, suggesting that Ikaros is post-translationally modified in primary T cells.
To better understand the nature of these modifications, we studied the ILC87-Ik1-ER cell line, derived from an Ikaros null T cell tumor, and engineered to overexpress the full-length Ik1 isoform fused to the ligand-binding domain (LBD) of the estrogen receptor (Ik1-ER) and green fluorescent protein (GFP) [10]. Upon treatment with 4-hydroxytamoxifen (4OHT), an ER antagonist, for 24h, GFP + cells exhibited enhanced nuclear translocation of Ik1-ER (Fig  1b), and little evidence of differentiation in terms of CD4 and CD8 expression (S1 Fig) [10]. As Ik1 expression also resulted in cell cycle exit [20,21], thus recapitulating the described antiproliferative effects associated with Ikaros, we performed biochemical analyses using this system. Interestingly, while extracts from 4OHT-treated ILC87-Ik1-ER cells showed the presence high molecular weight Ikaros proteins, those from vehicle-treated cells did not (Fig 1b), suggesting that Ikaros modifications occur in the nucleus.
Modified Ikaros proteins were labile, and required the presence of enzymatic inhibitors to increase their abundance in the extract preparations. We tested different inhibitors, and found that N-ethylmaleimide (NEM) and iodoacetic acid (IAA), which irreversibly modify thiol groups and inhibit the action of ubiquitin, Nedd8 and SUMO-specific isopeptidases [22][23][24], efficiently prevented the loss of the higher weight Ikaros proteins during protein preparation, compared with a single inhibitor or no treatment (Fig 1c). In contrast, DTT, which inhibits NEM and IAA activity, diminished the quantity of the modified proteins. Thus, protein extracts prepared in the presence of NEM and IAA, and in the absence of DTT, were studied in the experiments described below.

Ikaros sumoylation requires DNA binding
Because Ik1 was previously reported to be sumoylated when overexpressed with SUMO1 in 293T cells [15], we looked for sumoylated Ikaros proteins in ILC87-Ik1-ER cells. Nuclear extracts from 4OHT-treated cells were immunoprecipitated (IP) with an anti-ER antibody, and then analyzed by Western blot with anti-Ikaros, anti-SUMO1 or anti-SUMO2,3 antibodies (Fig 1d). Both anti-SUMO antibodies recognized proteins of similar high MW size as the anti-Ikaros antibody, suggesting that Ikaros is sumoylated by SUMO1 and SUMO2,3 in leukemic T cells.
To determine why nuclear localization is required for Ikaros sumoylation, we next investigated the influence of DNA binding and dimerization, by constructing Ik1-ER mutants lacking the N-terminal domain (aa 1-114), the DNA binding domain (DBD) (aa 119-223) [25], a large region between the DBD and the dimerization domain (aa 258-404), or the dimerization domain (aa 457-508) (Fig 2a). Although the ability of Ikaros to localize to the nucleus was unaffected by these mutations (S2 Fig), we found that loss of the DBD or the dimerization domain prevented the appearance of modified Ikaros (Fig 2b;

Sumoylation occurs on lysines 58, 240 and 425
Our bioinformatic analyses predicted that the Ikaros protein sequence contains three CKxE/D consensus sumoylation sites at lysines 58, 240 and 425. K58 and K240 were previously found to be targets of sumoylation [15]. We therefore mutated all 3 lysines to arginines (K!R), individually or in combination, and expressed the resulting proteins fused to the ER LBD in ILC87 cells. These cell lines were treated with 4OHT, and the nuclear extracts were immunoprecipitated with the anti-ER antibody and analyzed for Ikaros, SUMO1 and SUMO2,3 (Fig 2c and S3  Fig). These experiments showed that the anti-Ikaros, anti-SUMO1 and anti-SUMO2,3 antibodies recognized similar bands in the different extracts, as expected (S3 Fig). Of the seven detectable bands (labeled a-g) in the Ik1-ER extracts, mutation of K58 led to the loss of bands a, d and e, mutation of K240 led to the loss of bands b, d and f, and mutation of K425 led to the loss of bands c, e and f (Fig 2c). This suggested that bands a, b and c corresponded to single sumoylation events at K58, K240 and K425, respectively, while bands d, e and f were the result of bi-sumoylation. These results were confirmed with extracts from cells expressing double mutant Ik1 proteins where only one of the lysines was left untouched. The triple mutation (TM) of K58, K240 and K425 led to the loss of all higher MW Ikaros proteins, including band g which was detected in cells expressing single K mutants, but not in those expressing the double or triple mutants (see overexposed blot in Fig 2c). This band may correspond to a mixture of Ikaros proteins with multiple SUMO moities on 2 or 3 of the lysine residues. Importantly, the absence of modified proteins in cells expressing the TM-ER indicated that the ER end of the fusion protein is not a significant target of sumoylation. Thus, K58, K240 and K425 all contribute to Ikaros sumoylation.

Sumoylation of Ikaros has little effect on gene expression
To begin to understand the functional consequences of sumoylation on Ikaros, we first investigated its effects on the ability of Ikaros to regulate gene expression. ILC87-Ik1-ER or ILC87-Ik-TM-ER cells were purified to express similar levels of GFP and Ikaros (Fig 3a), then treated or not with 4OHT, and their transcriptomes were analyzed in two independent experiments (Fig  3b). In ILC87-Ik1-ER cells, 89 genes were upregulated and 237 were downregulated after 4OHT treatment (fold-change >2 in each experiment). Unexpectedly, similar results were obtained with 4OHT-treated ILC87-Ik-TM-ER cells. A few genes appeared to be changed in the ILC87-Ik-TM-ER cells, but these differences were not reproduced in a third independent  transcriptome experiment (not shown). Our results thus suggested that sumoylation does not significantly affect the way Ikaros regulates gene expression at the genome-wide level in ILC87 cells.
At the single gene level, we analyzed the mRNA levels of Mpzl2 and Scn4b, two genes shown to be repressed by Ikaros [10], in ILC87-Ik1-ER and ILC87-Ik-TM-ER cells. No detectable effect of sumoylation was observed by RT-qPCR, as the expression of these genes was also similarly downregulated in both cell lines after 4OHT treatment (Fig 3c). Finally, we evaluated the ability of the TM protein to repress the activity of the Hes1 promoter following Notch pathway activation in transient luciferase reporter transfection experiments, as had been shown for Ik1 [20]. HeLa cells were transfected with a Hes1-luciferase reporter plamid, and expression vectors encoding the intracellular domain of Notch1 (ICN1), and Ik1 or TM (S4 Fig). These results indicated that loss of sumoylation had little detectable impact on the ability of Ikaros to repress the transcription of the Hes1-luciferase reporter.
Altogether, our results indicate that sumoylation does not play a major role in the ability of Ikaros to modulate gene expression in this system.

Sumoylated Ikaros is less effective at inhibiting cell growth
We then evaluated the consequences of Ikaros sumoylation on leukemic cell expansion. As Ik1 re-expression in Ikaros null leukemic cells inhibits their expansion [20,21] (S5 Fig), we compared the capacity of TM-ER and Ik1-ER to inhibit leukemic cell expansion in competitive coculture assays. ILC87-Ik1-ER or ILC87-Ik-TM-ER cells with similar GFP levels were co-cultured with Ikaros null ILC87 cells (GFP -) at a 1:1 starting ratio over a 6 day period, and the proportions of GFP + cells were analyzed in the presence or absence of 4OHT (Fig 3d and 3e). In vehicle-treated cultures, the ratio of GFPand GFP + cells remained steady over time (Fig 3e). However, in 4OHT-treated cultures, the proportion of GFP + cells decreased more in ILC87-Ik-TM-ER than ILC87-Ik1-ER co-cultures. This difference was confirmed when we calculated the growth inhibition (GI) index between the percentage of GFP + cells in the 4OHT-treated and vehicle-treated cultures (Fig 3f, S1 Table), which showed that the GI index of Ik-TM-expressing cells was higher than that of Ik1-expressing cells. These results indicated that unsumoylated Ikaros inhibits cell growth better than sumoylated Ikaros.

Human B-ALL cells show high levels of sumoylated Ikaros
The above experiments suggested that sumoylation reduces the ability of Ikaros to inhibit cell proliferation. We therefore evaluated the sumoylation state of Ikaros in human B-ALL cells, where loss of Ikaros function is a major event [26]. Total cell extracts from three B-ALL cell lines and one primary B-ALL sample were analyzed with an anti-Ikaros antibody by Western blot, and compared with peripheral blood mononuclear cells from 2 healthy donors (Fig 4a  and not shown). The B-ALL samples used possessed either wild-type IKZF1 alleles (patient #4524, ACC42) or monoallelic IKZF1 deletions that resulted in haploinsufficiency (Tom-1, RS4;11), as determined by multiplex ligation-dependent probe amplification (not shown). Interestingly, Ikaros was abundantly modified in B-ALL cells, but not in PBMCs. These modified Ikaros proteins were sumoylated because they were also detected with antibodies to SUMO1 and SUMO2,3, in extracts immunoprecipitated first with an anti-Ikaros antibody (Fig  4b). The complex pattern of sumoylated bands may reflect sumoylation of both the Ik1 and Ik2 isoforms (S6 Fig). These results showed that B-ALL cells contain high levels of sumoylated Ikaros, regardless of whether their IKZF1 alleles are mutated, suggesting an important role for Ikaros sumoylation in leukemic cell development.

Discussion
Sumoylation is a major post-translational modification that influences protein activity, and we provide evidence that sumoylation may negatively affect the tumor suppressor function of Ikaros through, in part, a decrease in cell growth inhibition. Ikaros is markedly sumoylated in human B-ALL cells, and sumoylation occurs at three conserved lysine residues that act as SUMO1 and/or SUMO2,3 acceptor sites. Further, sumoylation requires nuclear localization and most likely DNA binding.
Our results confirm and extend previous findings in embryonic kidney and osteosarcoma cell lines that identified K58 and K240 as SUMO acceptor sites on the Ikaros protein [15]. Interestingly, K425 was not sumoylated in these cells, perhaps because the GFP-SUMO1 protein transfected into these cells preferentially sumoylates K58 and K240 due to its larger size (38 vs. 11kD for endogenous SUMO1). Alternatively, K425 sumoylation may require cell typespecific co-factors, as has been shown for the glucocorticoid receptor [27].
Sumoylation of Ikaros requires the DNA binding and dimerization domains. Since the three sumoylated lysines lie outside these domains, our results suggest that Ikaros sumoylation may be linked to DNA binding. The requirement for DNA binding in sumoylation has been observed for the yeast GCN4 protein, where sumoylation occurs after DNA binding to promote GCN4 clearance from promoter regions [28]. The concept that sumoylation may act to reduce transcription at specific gene subsets was also recently observed in human fibroblasts where sumoylation was inhibited [29]. Whether sumoylation affects Ikaros similarly remains to be investigated.
Ikaros sumoylation was previously associated with a loss of repressive properties, as Ikaros proteins conjugated with GFP-SUMO1 interacted poorly with co-repressors [15]. We were therefore surprised to find that cells expressing sumoylated Ikaros and those with unsumoylated Ikaros exhibit similar gene expression profiles by different criteria (transcriptome, RT-qPCR, luciferase assays), suggesting that sumoylation has minimal impact on gene activation or repression by Ikaros, and may be more modulatory than absolute in its influence. Alternatively, the transcriptional effects of sumoylation might be masked be the heterogeneity of the cell population that we studied, in particular with respect to clonal diversity and cell cycle status. It is for instance possible that sumoylation may play a role only at a certain stage of the cell cycle. This may be especially relevant as other studies have also revealed an inhibitory role of sumoylation in the transcriptional regulation of cell proliferation [29,30]. Future studies will therefore be required to better understand the molecular importance of sumoylation on Ikaros in the regulation of genes associated with cell proliferation.
Nonetheless, our results indicate that sumoylation limits certain physiological functions associated with Ikaros. As Ikaros is abundantly sumoylated in B-ALL cells, it is thus tempting to speculate that leukemic cells might hijack this mechanism to reduce Ikaros activity during cancer development.  Table. Growth inhibition by Ik1-ER and TM-ER. The Table provides the percentage of GFP-positive and negative cells at day 6 in 4 competition experiments between ILC87-Ik1-ER or ILC87-TM-Ik1-ER cells and empty ILC87 cells or mock-transduced ILC87-NGFR cells (see Fig 3c for experimental setup). Values in the "growth inhibition" columns correspond to the ratio of the percentages of GFP + cells in 4OHT-over those in EtOH-treated samples. (DOCX)