O-GlcNAcylation/Phosphorylation Cycling at Ser¹⁰ Controls Both Transcriptional Activity and Stability of Δ-Lactoferrin

Abstract

Δ-Lactoferrin (ΔLf) is a transcription factor that up-regulates DcpS, Skp1, and Bax genes, provoking cell cycle arrest and apoptosis. It is post-translationally modified either by O-GlcNAc or phosphate, but the effects of the O-GlcNAc/phosphorylation interplay on ΔLf function are not yet understood. Here, using a series of glycosylation mutants, we showed that Ser¹⁰ is O-GlcNAcylated and that this modification is associated with increased ΔLf stability, achieved by blocking ubiquitin-dependent proteolysis, demonstrating that O-GlcNAcylation protects against polyubiquitination. We highlighted the ³⁹¹KSQQSSDPDPNCVD⁴⁰⁴ sequence as a functional PEST motif responsible for ΔLf degradation and defined Lys³⁷⁹ as the main polyubiquitin acceptor site. We next investigated the control of ΔLf transcriptional activity by the O-GlcNAc/phosphorylation interplay. Reporter gene analyses using the Skp1 promoter fragment containing a ΔLf response element showed that O-GlcNAcylation at Ser¹⁰ negatively regulates ΔLf transcriptional activity, whereas phosphorylation activates it. Using a chromatin immunoprecipitation assay, we showed that O-GlcNAcylation inhibits DNA binding. Deglycosylation leads to DNA binding and transactivation of the Skp1 promoter at a basal level. Basal transactivation was markedly enhanced by 2–3-fold when phosphorylation was mimicked at Ser¹⁰ by aspartate. Moreover, using double chromatin immunoprecipitation assays, we showed that the ΔLf transcriptional complex binds to the ΔLf response element and is phosphorylated and/or ubiquitinated, suggesting that ΔLf transcriptional activity and degradation are concomitant events. Collectively, our results indicate that reciprocal occupancy of Ser¹⁰ by either O-phosphate or O-GlcNAc coordinately regulates ΔLf stability and transcriptional activity.

Introduction

O-GlcNAcylation is a ubiquitous post-translational modification consisting of a single N-acetylglucosamine moiety linked to Ser or Thr residues (1). It is a dynamic and reversible process mediated by the combined actions of O-GlcNAc transferase (OGT)² and O-GlcNAcase (OGA). Disruption of β-O-linked N-acetylglucosamine (O-GlcNAc) cycling through inhibitors or gene manipulations results in cellular defects (2, 3), and alterations of the O-GlcNAc status are associated with type-2 diabetes, neurological disorders, and cancer (4).

Because numerous proteins, such as transcription factors, signaling components, and metabolic enzymes are modified, O-GlcNAcylation is critical to normal cell homeostasis and gene regulation (5). It notably modulates gene expression, depending on the promoter and its associated transcription initiation complexes. For instance, the C-terminal domain of RNA polymerase II and a subset of general transcription factors are O-GlcNAcylated at transcription initiation (6). Gene silencing may be effected via the recruitment of OGT onto promoters by transcriptional corepressors. It then catalyzes the O-GlcNAcylation of specific transcription actors, modulating their activity. For instance, the association of OGT with the co-repressor mSin3A leads to the recruitment of histone deacetylase, thereby increasing transcriptional down-regulation (7, 8). OGA may favor gene transcription, not only by reducing the level of glycosylation but also via its intrinsic histone acetyltransferase domain (9). O-GlcNAcylation may also modulate the activity of transcription factors via the regulation of their trafficking, binding affinity either to protein partners or DNA, and/or turnover (8, 10–13).

Increasing evidence links O-GlcNAcylation to the proteasome pathway. It has been shown that O-GlcNAcylation is associated with lower proteasomal susceptibility of transcription factors, such as Sp1 (14, 15), p53 (16), and the estrogen receptor β (17). Most of these proteins have high PEST scores, and phosphorylation of their PEST (Pro-Glu-Ser-Thr) domain targets them for polyubiquitination (18) and subsequent degradation by the proteasome, whereas O-GlcNAcylation prolongs their half-lives. The proteasome is itself regulated through O-GlcNAcylation of both its regulatory and catalytic subunits (19, 20) as well as the ubiquitin (Ub)-activating enzyme E1 (21). Reduced degradation of O-GlcNAcylated proteins might also be due to their specific interaction with chaperones, such as Hsp70 family members that display lectin activity toward the O-GlcNAc motif, protecting them from proteolysis (22).

In many O-GlcNAcylated proteins, a phosphate group can alternatively occupy the same or adjacent sites (16, 17, 23, 24). This O-GlcNAc/P interplay, which leads to a rapid response mechanism and high molecular diversity and fine tunes protein interactions and functions, may also target Δ-lactoferrin (ΔLf) and regulate its transcriptional activity and stability. ΔLf is a transcription factor that was first discovered as a transcript, the expression of which was observed only in normal cells and tissues (25). Its absence from cancer cells (25, 26) is due to genetic and epigenetic alterations (27, 28). ΔLf messenger is therefore a healthy tissue marker, and we previously showed that its expression level is correlated with a good prognosis in human breast cancer, high concentrations being associated with longer overall survival (26). ΔLf is transcribed from the alternative promoter P2 in the lactoferrin (Lf) gene (29), and the use of the first available AUG codon in frame produces an alternative N terminus. Thus, compared with Lf, its secretory counterpart, ΔLf is a 73-kDa cytoplasmic protein able to enter the nucleus (30). Potential DNA-binding domains have been suggested for Lf, implicating the strong concentration of positive charges at the C-terminal end of the first helix, which is truncated in ΔLf, and at the interlobe region (31, 32).

ΔLf expression provokes anti-proliferative effects and induces cell cycle arrest in S phase (33). It is a transcription factor interacting via a ΔLf response element (ΔLfRE) found in the Skp1 and DcpS promoters (30, 34). ΔLf is also at the crossroads between cell survival and cell death because we recently linked ΔLf overexpression to up-regulation of the Bax promoter and apoptosis (35). Because ΔLf has several crucial target genes and may act as a tumor suppressor, modifications in its activity or concentration may have marked effects on cell survival, and its transcriptional activity should be strongly controlled. Results of screening ΔLf for O-GlcNAcylation and phosphorylation sites showed that the protein potentially undergoes both post-translational modifications. Four putative O-GlcNAc/phosphorylation sites were found at Ser¹⁰, Ser²²⁷, Ser⁴⁷², and Thr⁵⁵⁹, and their mutation led to a constitutively active ΔLf^M4 mutant (34). Here, we map the major O-GlcNAc/P site to Ser¹⁰, the PEST sequence (amino acids 391–404), and the main poly-Ub site to Lys³⁷⁹. We also report that O-GlcNAcylation at Ser¹⁰ down-regulates ΔLf transcriptional activity and up-regulates its stability by abrogating Ub-mediated proteolysis, whereas phosphorylation activates both transcription and degradation.

EXPERIMENTAL PROCEDURES

Cell Culture, Reagents, and Transfection

HEK 293 cells (ATC CRL-1573) were grown in monolayers and transfected as described (2 μg of DNA for 2 × 10⁶ cells) (30) using DreamFect^TM (OZ Biosciences). The amounts of ΔLf expression vectors were adjusted to maintain ΔLf amounts similar to those found in normal NBEC cells (26). Transfections were done in triplicate (n ≥ 4). Cell viability was assessed by counting using trypan blue 0.4% (v/v). Cell culture reagents were from Dutscher, Cambrex Corp., and Invitrogen. Other reagents were from Sigma. Antibodies against the 3xFLAG epitope (mouse monoclonal anti-FLAG M2 antibody, Sigma), HA epitope (goat HA tag polyclonal antibodies, BD Biosciences; mouse monoclonal HA.11 antibody 16B12, Covance Research Products), O-GlcNAc motif (mouse monoclonal RL2 antibody, Affinity Bioreagents; mouse monoclonal CTD110.6, Covance Research Products), Ser(P) motif (rabbit polyclonal antibodies, Millipore), and actin (goat polyclonal antibodies I-19, Santa Cruz Biotechnology, Inc. (Santa Cruz, CA)) were used for immunofluorescence, immunoprecipitation, and/or immunoblotting.

Immunofluorescence and Microscopy

HEK 293 cells were transfected by ΔLf C-terminal fused GFP expression vector 24 h prior the 4′,6-diamidino-2-phenylindole (Sigma) staining. The pΔLf-N-EGFP vector was kindly provided by Dr. C. Teng (National Institutes of Health, Research Triangle Park, NC). Immunofluorescence and microscopy were performed as described (30). Fluorescent microscopy images were obtained at room temperature with a Zeiss Axioplan 2 imaging system (Carl-Zeiss S.A.S., Le Pecq, France) equipped with appropriate filter cubes using a ×40 objective lens.

Purification of DNA, RNA, and Poly(A)⁺ RNA

Genomic DNA was purified using the QIAprep Spin Miniprep Kit (Qiagen), total RNA using the RNeasy minikit (Qiagen), and poly(A)⁺ RNA using the polyATrack® mRNA isolation system (Promega). The purity and integrity of each extract were checked using the nanodrop ND-1000 spectrophotometer (Labtech International) and the Bioanalyzer 2100 (Agilent Technologies).

qPCR Conditions

qPCR analyses were performed as described (30). The primer pairs used for the detection of ΔLf (forward, 5′-AAGCCAGTGGACAAGTTCA-3′; reverse, 5′-GCTTTGTTGGGATTTGTAGTT-3′; annealing temperature, 55 °C), ribosomal protein, large, P0 (34), and hypoxanthine-guanine phosphoribosyltransferase (forward, 5′-GACCAGTCAACAGGGGACAT-3′; reverse, 5′-AACACTTCGTGGGGTCCTTTTC-3′; annealing temperature, 55 °C) were purchased from Eurogentec.

Plasmid Construction and Site-directed Mutagenesis

pGL3-S1^Skp1-Luc, pcDNA-ΔLf, and p3xFLAG-CMV10-ΔLf were constructed as described (30). p3xFLAG-CMV10 (Sigma) and pcDNA3.1 (Invitrogen) were used as null vectors. The Ub-HA expression vector was a gift from Dr. C. Couturier (IBL, Lille, France). The pcDNA-OGT expression vector was constructed using OGT cDNA isolated from the pShuttle-OGT vector (36) (a kind gift of Dr. J. Hart, The John Hopkins University School of Medicine (Baltimore, MD)) and further cloned into the pcDNA3.1 vector. Mutants were generated using the QuikChange® site-directed mutagenesis kit (Stratagene) with pcDNA-ΔLf as template and primer pairs listed in Table 1. The constructs in which several sites were mutated were done sequentially. Following sequencing, the HindIII-NotI digests were cloned either into pcDNA3.1 for reporter gene assays or into p3xFLAG-CMV10 for protein experiments.

TABLE 1.

Names of mutants, amino acid modification, location, and oligonucleotides used for mutagenesis

^a Single-letter amino acid codes are used.

Reporter Gene Assay

Reporter gene assays were performed using the pGL3-S1^Skp1-Luc reporter vector and the different pcDNA-ΔLf mutant constructs or a null vector as described (34). Cell lysates were assayed using a luciferase assay kit (Promega) in a Tristar multimode microplate reader LB 941 (Berthold Technologies). Relative luciferase activities were normalized to basal luciferase expression and protein content as described (30) and expressed as a percentage; 100% corresponds to the relative luciferase activity of ΔLf^WT. Basal luciferase expression was assayed using a null vector and was determined for each condition (OGT, okadaic acid (OA), and glucosamine (GlcNH₂)) at each concentration. Each experiment represents at least three sets of independent triplicates.

In Vivo DNA Binding Assays

Chromatin immunoprecipitation (ChIP) and double ChIP (re-ChIP) assays were performed as described (34, 37) with some modifications introduced for re-ChIP. Briefly, ChIP complexes (8 × 10⁶ cells) were immunoprecipitated with M2, RL2, HA tag, or anti-Ser(P) antibodies all used at 1:250 and twice eluted with 100 μl of 1 mm dithiothreitol for 30 min at 37 °C. After centrifugation, pooled eluted fractions were diluted 40 times to reduce the dithiothreitol concentration to 25 μm with ChIP dilution buffer and then immunoprecipitated with either M2 or rabbit anti-IgG (GE Healthcare) or without antibodies. Amplification conditions of Skp1 and albumin promoters were as described (34). ChIP or re-ChIP results presented in Fig. 5 correspond to one representative experiment among three. qPCR was performed only for the ChIP assay. Amplification was carried out in triplicate (n = 3) in the presence of 2 μl of purified DNA, primer pairs used to amplify the Skp1 promoter fragment (34), and Brilliant SYBER Green QPCR Master Mix (Stratagene) according to the manufacturer's instructions. Samples were then submitted to 40 cycles of amplification (denaturation, 30 s at 90 °C; hybridation, 30 s at 55 °C; elongation, 30 s at 72 °C) in a thermocycler Mx4000 (Stratagene). Data presented in Fig. 5D are expressed as a percentage of input.

FIGURE 5.

O-GlcNAcylation at Ser¹⁰ negatively controls DNA binding and ΔLf transcriptional activity. A, ΔLf is a phosphorylated protein. HEK 293 cells were transfected by different ΔLf constructs in the presence or not of GlcNH₂ for 24 h prior to lysis. M2 immunoprecipitates were immunoblotted with anti-Ser(P), and as a loading control, input was immunoblotted with M2 (left). Phosphatase treatment (1 unit/IP) confirms anti-phosphate antibody specificity (AP, right). B, relative luciferase activity of ΔLf and its O-GlcNAc mutants. HEK 293 cells were co-transfected with pGL3-S1^Skp1-Luc vector and pcDNA3.1-ΔLf^WT or Ser¹⁰ mutant constructs. Relative luciferase activities are expressed as described under “Experimental Procedures” (n ≥ 9; **, p < 0.01). C and D, O-GlcNAcylation inhibits DNA binding. The in vivo binding of ΔLf and its Ser¹⁰ mutants to the Skp1 promoter fragment was examined in HEK 293 cells treated or not with GlcNH₂ (n = 3). Cross-linked DNA-ΔLf complexes were immunoprecipitated, and precipitated DNA fragments were PCR-amplified (C) or real time PCR-amplified (D) with specific primers covering the ΔLfRE present in the Skp1 promoter. The PCR-amplified DNA purified from the sonicated chromatin was used as input and loading control. ChIP assays were performed using M2, anti-rabbit IgG antibodies as a nonspecific control (irrelevant; IR) and without antibody (NIP). Amplification of the albumin promoter region was used as a negative control. E, ΔLf transactivation complex is not O-GlcNAcylated. Re-ChIP was performed as above for the ChIP assay with some modifications. The first immunoprecipitation was performed using M2, RL2, anti-Ser(P) or anti-HA antibodies. Then, prior to reversal of protein-DNA cross-linking, the chromatin fragments were subjected to reprecipitation using M2, irrelevant antibody, or no antibodies (n = 3). Error bars, S.D. IB, immunoblot; IP, immunoprecipitation.

Proteasomal Degradation Assay

Proteasomal activity assay was performed according to the assay instructions (Chemicon International) on HEK 293 cell lysates. Lactacystin was used as a 20 S proteasome inhibitor. Fluorescence data were collected using a Tristar multimode microplate reader LB 941 (Berthold Technologies) using 380-nm excitation and 460-nm emission filters.

Immunoblotting and Immunoprecipitation

Proteins were extracted from frozen cell pellets in radioimmune precipitation buffer as described (30). For direct immunoblotting, samples mixed with 4× Laemmli buffer were boiled for 5 min. 20 μg of protein from each sample were submitted to 7.5% SDS-PAGE and immunoblotted. For immunoprecipitation experiments, 1 mg of total protein was preabsorbed with protein G-Sepharose 4 Fast Flow (GE Healthcare). RL2 (1:250), M2 (1:500), or anti-HA polyclonal (1:100) antibodies were mixed with Protein G-Sepharose beads for 1 h prior to an overnight incubation with the preabsorbed lysate supernatant at 4 °C. The beads were then washed five times with lysis buffer. Proteins bound to the beads were eluted with 4× Laemmli buffer and analyzed by immunoblotting as above. Blots were probed at room temperature with primary antibodies (M2, 1:2000; CTD110.6, 1:3000; HA.11, 1:4000; anti-Ser(P), 1:500; RL2, 1:2000; and anti-actin, 1:10,000) for 2 h and with either secondary anti-IgG antibodies conjugated to horseradish peroxidase (GE Healthcare) or secondary anti-IgM antibodies conjugated to horseradish peroxidase (Chemicon International) for 1 h before detection by chemiluminescence (ECL Advance, GE Healthcare). Each result in which immunoblots are presented corresponds to one representative experiment among at least three.

Densitometric and Statistical Analyses

The densitometric analyses were performed using Quantity One version 4.1 software (Bio-Rad), and acquisition was carried out with a GS710-calibrated densitometer (Bio-Rad). M2 densitometric values were normalized to actin and expressed as R = D^M2/Dact. The -fold stability (X) is expressed as this ratio related to the wild type ratio and to the t₀ value as follows for ΔLf^PEST. X = ^PESTR_t/^PESTR_t0/^WTR_t/^WTR_t0. All of the statistical analyses were done using Origin® 7 software (OriginLab Corp.). Means were statistically analyzed using the t test or analysis of variance, and differences were assessed at p < 0.05 (*) or p < 0.01 (**).

RESULTS

Impact of the O-GlcNAc/P Interplay on ΔLf Transcriptional Activity and Stability

Investigation of the O-GlcNAc function has mainly relied on the manipulation of the hexosamine biosynthesis pathway via an increased production of UDP- GlcNAc, the substrate for OGT (38). Thus, cells exposed to increased concentrations of GlcNH₂ or overexpressing OGT exhibit enhanced levels of protein O-GlcNAcylation (39). On the other hand, the use of OA, an inhibitor of PP2A and PP1 phosphatases, is a valuable tool for inducing protein hyperphosphorylation (40, 41).

Prior to investigating whether ΔLf transcriptional activity is regulated via O-GlcNAc/P interplay, we first established that HEK 293 cells possess rapid, inducible O-GlcNAc/P mechanisms at the OGT, GlcNH₂ and OA concentrations employed (42). First of all, we verified that cell viability was not perturbed (Fig. 1A). At the concentrations usually used in the literature, such as 40 mm GlcNH₂ and 50 nm OA, cell viability was markedly decreased in HEK 293 cells. For this reason, we used lower concentrations, such as 10 mm GlcNH₂ and 10 nm OA, that did not affect cell viability but at which modulation of the O-GlcNAc/P status was visible (Fig. 1B). Co-transfection of ΔLf (1 μg DNA/10⁶ cells) and OGT (2.5 μg of DNA/10⁶ cells) expression vectors did not significantly affect cell viability (Fig. 1A). Fig. 1B shows that ΔLf is indeed sensitive to OA and GlcNH₂ or OGT but with opposite effects. Treatment with OA led to decreased ΔLf glycosylation, whereas treatment with GlcNH₂ or co-transfection with OGT increased it. The same GlcNAcylation pattern was observed using either the RL2 or the CDT110.6 antibody. This result demonstrates clearly that ΔLf possesses O-GlcNAc site(s). OA treatment, which favors phosphorylation, decreases the ΔLf glycosylation level, suggesting that glycosylation site(s) may exist in balance with phosphorylation site(s). Because RNA polymerase II activity is also controlled by this interplay, we next verified that transcription of ΔLf; ribosomal protein, large, P0; or hypoxanthine-guanine phosphoribosyltransferase was indeed not altered under OGT, GlcNH₂, or OA treatment (Fig. 1C). Our control experiments showed that modulation of the O-GlcNAc content does not impair cell functions at the concentration of OA, GlcNH₂, or OGT we used.

FIGURE 1.

O-GlcNAc/P interplay regulates ΔLf transcriptional activity. HEK 293 cells were incubated with GlcNH₂ or OA, or transfected with an OGT construct (OGT) to assess the impact of the O-GlcNAc/P interplay on ΔLf. A, cell viability. Cell viability of 10⁴ HEK 293 was assayed 24 h after GlcNH₂ or OA treatment or after transfection with pcDNA-OGT at 2.5 or 5 μg of DNA/10⁶ cells (n = 9). B, ΔLf O-GlcNAcylation status. Treated and untreated 3xFLAG-ΔLf-expressing HEK 293 cell extracts were M2-immunoprecipitated prior to SDS-PAGE and CTD110.6 or RL2 immunodetection. Input was used as the loading control (n = 3). C, gene expression is not altered under GlcNH₂ and OA treatment or OGT overexpression. Poly(A)⁺ RNA was purified from total RNA of ΔLf-expressing cells treated with OGT, GlcNH₂, or OA 24 h after transfection and assayed using real-time PCR. RPLPO and hypoxanthine-guanine phosphoribosyltransferase are internal controls (n = 3). D–F, ΔLf transcriptional activity is modulated by OGT, GlcNH₂, or OA treatment. Cells were co-transfected with pcDNA-ΔLf and pGL3-S1^Skp1-Luc and incubated with GlcNH₂ or OA or co-transfected with pcDNA-ΔLf, pcDNA-OGT, and pGL3-S1^Skp1-Luc for 24 h prior to lysis. Relative luciferase activities are expressed as described under “Experimental Procedures” (n ≥ 9). G, the endoproteolytic activity of the proteasome is not altered under OGT, GlcNH₂, and OA treatment. The histograms represent the proteasomal activity assayed by following the fluorescence emitted during the degradation of a synthetic fluorescent peptide (69). Lactacystin is a proteasome inhibitor (n = 3). H, Ub-dependent degradation of ΔLf is GlcNH₂-sensitive. Cells were co-transfected with or without 3xFLAG-tagged ΔLf (ΔLf-3xFLAG) and Ub-HA vectors, treated or not with GlcNH₂, or transfected or not with pcDNA-OGT. Cells were incubated 2 h with 10 μm MG132 before lysis in order to inhibit proteasomal degradation. Total cell extracts were immunoprecipitated with anti-HA polyclonal antibodies or used as input. Samples were immunoblotted with M2 (top and bottom) or HA.11 (middle) antibodies. I, ΔLf traffic is not affected by GlcNH₂ or OA treatment. Cells were transfected with pEGFP empty or pEGFP-ΔLf vector and incubated with GlcNH₂ or OA. Fluorescent microscopy was performed after 4′,6-diamidino-2-phenylindole (DAPI) staining. Error bars, S.D. IB, immunoblot; IP, immunoprecipitation.

We then investigated ΔLf transcriptional activity using reporter gene assays and a Skp1 promoter fragment containing the ΔLfRE known to be highly transactivated by ΔLf (30). ΔLf transcriptional activity increased in line with OA concentration (Fig. 1D), whereas it decreased in a dose-dependent manner in the presence of GlcNH₂ (Fig. 1E). Thus, when phosphorylation was augmented, transactivation was increased 6–7-fold compared with controls, whereas when O-GlcNAcylation was increased in ΔLf-expressing cells, transactivation of the Skp1 promoter was strongly reduced. ΔLf transcriptional activity also decreased when cells overexpressed OGT but at a lower level (Fig. 1F).

Because, as for many transcription factors, ΔLf is rapidly degraded, we next investigated whether its turnover is dependent on both the Ub-proteasome pathway and O-GlcNAc/P interplay. We first verified that treatment with OA, GlcNH₂, or OGT did not disturb the proteasome pathway. As shown in Fig. 1G, these treatments did not alter or exacerbate proteasomal degradation compared with the untreated and lactacystin-treated conditions. Fig. 1H shows that a ladder of polyubiquitinated ΔLf forms is visible (top, lanes 3 and 6). We next evaluated whether O-GlcNAcylation regulates ΔLf degradation, and the intensity of polyubiquitination was indeed decreased in a dose-dependent manner after GlcNH₂ treatment (Fig. 1H, top, lanes 4 and 5) and after OGT overexpression (Fig. 1H, top, lane 7). Equivalent loadings of Ub-HA protein (Fig. 1H, middle) and ΔLf (Fig. 1H, bottom) were confirmed by immunoblotting. These data demonstrated that ΔLf is more stable in an environment favoring O-GlcNAcylation.

We further investigated whether ΔLf traffic might be altered, leading to an exclusive nuclear targeting of the phosphoform. As previously described (29, 30), a ΔLf-GFP fused protein localizes predominantly to the cytoplasm but also to the nucleus (Fig. 1I, panel 2). Here, we showed that the subcellular localization of ΔLf-GFP was not modified with either GlcNH₂ or OA (Fig. 1I, panels 3 and 4, respectively), suggesting that ΔLf traffic is not regulated by the O-GlcNAc/P interplay.

Mapping the Key O-GlcNAc Site to Ser¹⁰

The low abundance of ΔLf, the necessity for producing a 3xFLAG-tagged protein in order to detect it, and the inherent limitation of the sensitivity of tritium labeling render the detection of carbohydrate moieties on ΔLf and the subsequent mapping of its glycosylated sites extremely difficult. Therefore, in order to confirm the presence of the O-GlcNAc sites and characterize their roles, we made a series of glycosylation mutants in which only one O-GlcNAc site is preserved, named ΔLf^S10+, ΔLf^S227+, ΔLf^S472+, and ΔLf^T559+, respectively (Fig. 2A). ΔLf^M4 (34) and ΔLf^WT were used as controls. ΔLf and its glycosylation mutants were then expressed in HEK 293 cells, and their levels of expression were compared. Fig. 2B shows that the ΔLf^S10+ mutant was expressed at the same level as ΔLf^WT (short exposure time) in contrast to the other mutants (long exposure time). ΔLf^S227+ and ΔLf^S472+ were slightly more expressed than were ΔLf^M4 and ΔLf^T559+, which were both feebly expressed. These data suggest that the post-translational modifications present on Ser¹⁰ may participate in ΔLf stability and that its absence from the other mutants leads to their rapid turnover.

FIGURE 2.

Post-translational modifications of Ser¹⁰ modulate ΔLf transcriptional activity and stability. A, schematic representation of ΔLf glycosylated mutant constructs. B, expression of 3xFLAG-ΔLf^WT and its O-GlcNAcylation mutants. ΔLf^WT and mutant constructs were transfected for 24 h prior to lysis. Whole cell extract was immunoblotted with either M2 or anti-actin antibodies used as loading control. Development at two exposure times is shown for M2. C and D, mapping of ΔLf O-GlcNAcylated sites. Cells were transfected by the above constructs and lysed 24 h later. C, lysates were immunoprecipitated with RL2 and immunoblotted with M2. D, lysates were immunoprecipitated with M2 and immunoblotted with CTD110.6. E, relative luciferase activity of ΔLf and its mutants. Cells were co-transfected with pGL3-S1^Skp1-Luc reporter vector and pcDNA-ΔLf (ΔLf^WT) vector or the O-GlcNAc mutant constructs. Relative luciferase activities are expressed as described under “Experimental Procedures” (n ≥ 9; **, p < 0.01). Error bars, S.D. IB, immunoblot; IP, immunoprecipitation.

O-GlcNAcylation was then investigated on the ΔLf isoforms. Because ΔLf mutants are feebly produced, we first immunoprecipitated ΔLf-expressing cell lysates with RL2 in order to accumulate enough O-GlcNAcylated material (Fig. 2C). A reverse immunoprecipitation was then performed using the M2 antibody in order to specifically immunoprecipitate ΔLf or its glycovariants (Fig. 2D). Fig. 2C shows that ΔLf was effectively glycosylated, whereas ΔLf^M4 was not, confirming that no other O-GlcNAc sites are present on the protein. ΔLf^S10+, ΔLf^S227+, and ΔLf^S472+ mutants were glycosylated, whereas ΔLf^T559+ was not (Fig. 2C). The reverse immunoprecipitation of the cell lysates with M2 antibody followed by O-GlcNAc immunodetection with the CTD 110.6 antibody (Fig. 2D) confirmed that ΔLf and its ΔLf^S10+ mutant were glycosylated, whereas ΔLf^M4 and ΔLf^T559+ were not. The O-GlcNAcylated signals corresponding to ΔLf^S227+ and ΔLf^S472+ mutants that were effectively visible when RL2 antibody was used, were poorly visible for ΔLf^S227+ and not visible for ΔLf^S472+ when the CTD 110.6 antibody was used. RL2 (43) and CTD110.6 (44) are the two most commonly used antibodies, with CTD110.6 described as being the most specific. Therefore, this divergent result may be due to the fact that both isoforms were too poorly expressed to be detected or that the Ser⁴⁷² site was not fully glycosylated. However, we cannot exclude the possibility that performing the immunoprecipitation first with RL2 antibody may favor immunoprecipitation of O-GlcNAcylated ΔLf partners. Nevertheless, both immunoprecipitations confirmed without ambiguity that ΔLf^S10+, ΔLf^S227+, and ΔLf^WT mutants were glycosylated, whereas ΔLf^T559+ and ΔLf^M4 were not.

We next assayed the transcriptional activity of the mutants compared with wild type (WT) (Fig. 2E). The strong transcriptional activity of the glycosylation-null mutant confirmed that O-GlcNAcylation negatively regulates ΔLf activity, as previously indicated (Fig. 1, E and F). We compared the activity of mutants in which only one glycosylation site was preserved with that of ΔLf^M4 in order to evaluate the impact of adding only one regulatory site at a time. The ΔLf^S227+ and ΔLf^S472+ mutants showed transcriptional activities nearly 2-fold greater than WT and close to that of ΔLf^M4 (Fig. 2E), suggesting that the presence of either O-GlcNAc or phosphate on these two sites does not crucially regulate ΔLf transcriptional activity. These sites might be priming sites, as described for phosphorylation, necessary to target OGT or specific kinases to the other sites and may only be transiently O-GlcNAcylated. In contrast, the transcriptional activities of ΔLf^S10+ and ΔLf^T559+ were strongly inhibited compared with ΔLf^M4, suggesting that these two sites are primordial for regulation. The absence of response of the ΔLf^T559+ mutant might be due to the feeble expression or rapid degradation of this non-glycosylated and transcriptionally inactive mutant. However, this may not be the only explanation because the removal of all four glycosylation sites in ΔLf^M4 leads to a constitutively active isoform. Preliminary results on the ΔLf^T559+ mutant showed that it may compete with WT for DNA binding (data not shown), and further work will be necessary to understand the intrinsic role of Thr⁵⁵⁹. O-GlcNAc/P modifications on the Ser¹⁰ site led to a 5-fold inhibition of ΔLf transcriptional activity compared with ΔLf^M4 and 2-fold inhibition compared with ΔLf^WT (Fig. 2E). This site seems therefore to be a crucial target for the regulation of both ΔLf stability and transcriptional activity. We therefore focused our initial attention on Ser¹⁰ and first investigated the impact of O-GlcNAc/P modifications on ΔLf turnover by studying their relationship with the proteasome pathway.

ΔLf Turnover Is Driven through a PEST Motif (Amino Acids 384–404) and Lys³⁷⁹

Short intracellular half-life proteins frequently have a short hydrophilic stretch of amino acids termed a PEST motif. Phosphorylation of the Ser and/or Thr residues and ubiquitination, often of the flanking Lys residues, trigger degradation. Analysis of the ΔLf sequence did not allow identification of a PEST motif in the Ser¹⁰ environment but indicated one at the C terminus with three nearly contiguous Ser (Ser³⁹², Ser³⁹⁵, and Ser³⁹⁶) and two flanking Lys (Lys³⁷⁹ and Lys³⁹¹) residues as potential targets either for kinase/OGT or Ub ligase, respectively. Alignment of Lf sequences from other species to this PEST motif shows that the locus is conserved (Table 2).

TABLE 2.

Alignment of ΔLf PEST sequence to lactoferrin sequences from different species

^a Single-letter amino acid codes are used; the PEST sequence of ΔLf predicted using the PESTfind software (EMBnet Austria) is in boldface type; italic boldface letters indicate the Ser residues within the putative PEST sequence in Lf from different species; the Ub-targeted Lys residue is underlined in the ΔLf PEST sequence.

We evaluated the functionality of the PEST sequence using a ΔLf^PEST mutant in which the three Ser residues were replaced by Ala and showed that this mutation leads to a slight increase in ΔLf content of about 40% compared with WT (Fig. 3, A and B). To measure the ΔLf turnover rate indirectly, we performed incubations (0–150 min) with cycloheximide, a potent inhibitor of de novo protein synthesis (45, 46). The ΔLf content of HEK cells transfected with either ΔLf^WT or ΔLf^PEST constructs was analyzed following the addition of cycloheximide (Fig. 3C). Differences in the steady state levels of ΔLf were readily apparent after 30 min, which may correspond to the delay necessary for observing the first effects of cycloheximide treatment (Fig. 3C, panel 1). Mutation of the Ser residues in the PEST sequence conferred stability on ΔLf (Fig. 3C, panel 5). GlcNH₂ treatment of HEK cells transfected with either ΔLf^WT or ΔLf^PEST constructs was also performed (Fig. 3C, panels 3 and 7, respectively), and OGT was coexpressed with ΔLf^WT (Fig. 3C, panel 9). Actin, which is stable under the same experimental conditions, was used as an internal control (Fig. 3C, panels 2, 4, 6, 8, and 10). Densitometric data are expressed as -fold stability as described under “Experimental Procedures” (Fig. 3D). Invalidation of the PEST sequence led to a 5–6-fold gain in stability and confirmed that this sequence is determinant for ΔLf degradation. GlcNH₂ treatment or OGT overexpression led to a 2- or 3-fold increase in ΔLf^WT stability, respectively, compared with controls, visible at 90 min, confirming that increasing O-GlcNAcylation protects ΔLf from degradation. However, when ΔLf^PEST-expressing cells were submitted to the GlcNH₂ treatment, the stability of ΔLf^PEST was not significantly different in the presence or absence of GlcNH₂ suggesting that mutation of the PEST sequence is sufficient to confer stability on ΔLf. Moreover, these results also suggest that these two events could be linked because, if they were independent, a greater stability of the ΔLf^PEST isoform should be visible in the presence of GlcNH₂.

FIGURE 3.

Ub-dependent ΔLf degradation is mediated through a PEST sequence at the C terminus and Lys³⁷⁹ and is inhibited by O-GlcNAcylation. A and B, deletion of the PEST sequence slightly increases ΔLf stability. HEK 293 cells were transfected with either ΔLf^WT or ΔLf^PEST constructs for 24 h. Total protein extracts were immunoblotted with M2. C, modulation of ΔLf half-life by O-GlcNAcylation. Cells were transfected with either ΔLf^WT or ΔLf^PEST and GlcNH₂-treated or cotransfected with the OGT-construct or not and then incubated with fresh medium supplemented by 10 μg/ml cycloheximide for the indicated time 24 h after transfection. Total protein extracts were immunoblotted with either M2 or anti-actin antibodies. D, data are expressed as -fold stability as described under “Experimental Procedures.” *, p < 0.05. E, the three Ser residues of the PEST sequence are equivalent. Mutation of Ser residues was done on the 3xFLAG-ΔLf^M4 construct as template. Cells were transfected by the different constructs, and 24 h after transfection, total protein extracts were immunoblotted with either M2 or anti-actin antibodies. F, mutation of Lys³⁷⁹ rather than of Lys³⁹¹ inhibits degradation. 3xFLAG-ΔLf and 3xFLAG-ΔLf^M4 constructs were used as template to obtain Lys³⁷⁹ and Lys³⁹¹ mutants. Cells were transfected by the different constructs, and 24 h after transfection, total protein extracts were immunoblotted with either M2 or anti-actin antibodies. G, Lys³⁷⁹ is the main Ub-ligase target. HEK 293 cells were co-transfected with or without the 3xFLAG-ΔLf constructs and the Ub-HA-expressing vector for 24 h and then incubated with a 10 μm concentration of the proteasomal inhibitor MG132 for 2 h prior to lysis. Total cell extracts were immunoprecipitated with anti-HA polyclonal antibodies or used as input. Samples were immunoblotted with M2 (top and bottom) or with HA.11 (middle) antibodies. Error bars, S.D. IB, immunoblot.

We next studied the invalidation of the PEST sequence on the ΔLf^M4 mutant. This mutant is detected at low levels in transfected cells, indicating that it is either feebly expressed or rapidly degraded (Fig. 2B). Fig. 3E shows that this invalidation increased ΔLf^M4 stability and rendered this mutant more resistant to proteasomal proteolysis. We further investigated whether a particular Ser within the PEST motif was involved in this process using a series of single Ser mutants (Table 2). Whatever the Ser mutated, ΔLf expression was identical, suggesting that the three Ser residues were equivalent phosphorylation targets due to their proximity (Fig. 3E). Moreover, prediction results for putative phosphorylation sites using the NetPhos 2.0 server (CBS.DTU; available on the World Wide Web) also emphasized these three Ser residues as kinase targets, albeit with a higher score for Ser³⁹⁶ (Ser³⁹², 0.766; Ser³⁹⁵, 0.789; Ser³⁹⁶, 0.977).

We then studied whether ΔLf-mediated ubiquitination occurs predominantly through Lys³⁷⁹- or Lys³⁹¹-linked chains by constructing a series of mutants in which residue 379 or 391 was mutated to Ala either in ΔLf or ΔLf^M4. The K379A mutation led to a slightly increased expression level of ΔLf and completely restored stability to ΔLf^M4 compared with controls, whereas the K391A mutation had no effect on the ΔLf expression level and only slightly increased expression of ΔLf^M4 (Fig. 3F). This result confirms that the flanking Lys³⁷⁹, which is highly conserved among species (Table 2), is involved in ΔLf turnover and suggests that it is the major poly-Ub acceptor site. We next verified that ubiquitination of ΔLf is indeed Lys³⁷⁹-linked. Fig. 3G shows that polyubiquitination was strongly visible on ΔLf^WT (top, lane 3), was lower on ΔLf^PEST (lane 4) and ΔLf^K391A mutants (lane 6), was poorly visible on ΔLf^K379A (lane 5), and was not at all visible on the double mutant ΔLf^KK (lane 7) in which both Lys residues were mutated. Control levels of ubiquitination and ΔLf expression are shown in the middle and lower panels, respectively (Fig. 3G). These data confirmed that the Lys³⁷⁹ residue corresponds to the main Ub ligase target and that Lys³⁹¹ corresponds to a minor site. We next investigated which type of relationship may exist between the functional PEST sequence at the C terminus and the O-GlcNAc/P site at the N terminus.

O-GlcNAcylation of Ser¹⁰ Protects ΔLf from Polyubiquitination

To determine whether ΔLf protein stability was controlled via an O-GlcNAc/P switch on Ser¹⁰, other mutants were constructed (Fig. 4A), such as the ΔLf^S10A mutant that has only the Ser¹⁰ residue mutated and the ΔLf^S10D mutant in which an Asp residue was introduced in place of Ser in order to mimic constitutive phosphorylation as described previously (47, 48). Immunoblotting of the different mutants with M2 antibodies is presented in Fig. 4B. ΔLf^S10A and ΔLf^S10+ had expression levels similar to that of WT, whereas ΔLf^S10D had an extremely short half-life (Fig. 4C), suggesting that this mutant is an interesting tool for studying degradation of the ΔLf phosphoform. Due to the absence of Ser¹⁰, ΔLf^S10A was expected, like the ΔLf^M4, ΔLf^T559+, ΔLf^S227+, and ΔLf^S472+ mutants (Fig. 2B), to be less stable than WT. In ΔLf^S10A, only Ser¹⁰ is mutated. Therefore, the stability of ΔLf^S10A might be due to the other sites, which could be used as “protecting sites” in the absence of Ser¹⁰.

FIGURE 4.

Ser¹⁰ posttranslational modifications control ΔLf turnover; O-GlcNAc confers stability, whereas phosphorylation triggers degradation. A, schematic representation of the Ser¹⁰ mutant constructs. B, total protein extract of HEK 293 cells expressing 3xFLAG-ΔLf or its Ser¹⁰ mutant constructs was immunoblotted with M2 or anti-actin antibodies. C, O-GlcNAc posttranslational modification at Ser¹⁰ increases ΔLf stability. Transfected cells were incubated 24 h posttransfection with fresh medium supplemented with 10 μg/ml cycloheximide. ΔLf^WT and ΔLf^S10+ transfected cells were incubated with or without GlcNH₂. Cell lysates were immunoblotted with M2 or anti-actin antibodies. D, the graph data are expressed as -fold stability as described under “Experimental Procedures” (n = 5; *, p < 0.05). E, invalidation of Ser¹⁰ protects against Ub-dependent degradation. Cells were transiently co-transfected with or without the 3xFLAG-tagged ΔLf or Ser¹⁰ mutant contructs and the Ub-HA expression vector for 24 h and then incubated with 10 μm MG132 for 2 h prior to lysis. Total cell extracts were immunoprecipitated with anti-HA polyclonal antibodies or used as input. Samples were immunoblotted with M2 (top and bottom) or with HA.11 antibodies (middle). Error bars, S.D. IB, immunoblot; IP, immunoprecipitation.

The turnover of these different Ser¹⁰ mutants compared with WT and actin (internal control) is shown in Fig. 4C. Differences in the steady state levels of ΔLf and ΔLf^S10+ mutant were readily apparent around 30–60 min and strongly visible after 90 min (Fig. 4C, panels 1 and 5). Invalidation of the Ser¹⁰ site in ΔLf^S10A resulted in a markedly prolonged half-life (Fig. 4C, panel 3). Comparable results were obtained when cells expressing ΔLf^S10+ were cultured in the presence of GlcNH₂, confirming the crucial role of O-GlcNAcylation in ΔLf stability (Fig. 4C, panel 7). Interestingly, ΔLf^S10D had a faster turnover rate compared with WT (panel 9), indicating that mimicking phosphorylation at this locus triggers degradation. Fig. 4D summarizes the densitometric data of the ΔLf immunoblots expressed as -fold stability, as described under “Experimental Procedures.” ΔLf^S10+, which could be either phosphorylated or glycosylated, was slightly more stable than WT (1.5-fold), whereas the same mutant expressed in hyper-O-GlcNAcylation conditions was 4-fold more stable than WT. Interestingly, the mutation of Ser¹⁰ to Ala also led to ΔLf stability (3.5-fold compared with WT), which suggests that stability is not due to the presence of the O-GlcNAc moiety but to the absence of the phosphate group. Mimicking phosphate at Ser¹⁰ in the ΔLf^S10D mutant shortened its half-life. From these experiments, we conclude that phosphorylation at Ser¹⁰ accelerates ΔLf degradation, whereas O-GlcNAcylation at Ser¹⁰ controls its stability, confirming the existence of a strong link between the O-GlcNAc/P interplay and the Ub degradation pathway.

We next investigated whether ubiquitination of ΔLf is linked to Ser¹⁰ phosphorylation. Fig. 4E shows that polyubiquitination was marked on ΔLf^WT and ΔLf^S10+ (top, lanes 3 and 4), whereas it was reduced on ΔLf^S10A (lane 5). Control levels of ubiquitination and ΔLf expression are shown in the middle and bottom panels (Fig. 4E). Unfortunately, the high turnover rate of the ΔLf^S10D mutant or of ΔLf^WT under OA treatment precluded the observation of a polyubiquitination signal (data not shown).

In conclusion, our data showed that ΔLf turnover is driven through a PEST sequence located at the C terminus with polyubiquitination occurring mainly at Lys³⁷⁹. We also demonstrated that the degradation process is regulated via the O- GlcNAc/P interplay, which targets Ser¹⁰. As a glycoform, ΔLf is stable, whereas as a phosphoform, it is sensitive to degradation. Since proteasomal degradation is triggered by phosphorylation, we suggest that phosphorylation of Ser¹⁰ favors phosphorylation of the PEST sequence, whereas O-GlcNAcylation of Ser¹⁰ prevents it.

Phosphorylation at Ser¹⁰ Controls ΔLf Transcriptional Activity

Because OA treatment increases ΔLf transcriptional activity, we next questioned whether the phosphoform might be responsible for gene transactivation. Using immunoprecipitation with the M2 antibody and probing the resulting blot with an anti-Ser(P) antibody, we studied the phosphorylation status of ΔLf (Fig. 5A, left). Phosphatase treatment markedly abrogated the phosphorylation signal, confirming antibody specificity (right). Immunoblotting (Fig. 5A) showed that ΔLf and its Ser¹⁰ mutants exist as phosphoforms. The decreased phosphorylation signals observed under GlcNH₂ treatment confirm that phosphorylation and O-GlcNAcylation may alternate on some of the sites. Therefore, the weaker phosphorylation signal observed with the hyperglycosylated ΔLf^S10+ isoform (lane 9) compared with control (lane 5) strongly suggests that the O- GlcNAc/phosphorylation interplay targets the Ser¹⁰ site. However, because ΔLf^M4 is phosphorylated, ΔLf is also phosphorylated on sites different from the O-GlcNAc/P interplay sites.

We next performed gene reporter analyses as described above and investigated whether phosphorylation at Ser¹⁰ controls ΔLf transcriptional activity. Fig. 5B shows that, compared with ΔLf^WT, ΔLf^S10+ transcriptional activity was inhibited 2-fold as in Fig. 2E, whereas the transcriptional activity of ΔLf^S10A was increased 1.5–2-fold, and that of ΔLf^S10D was increased 4.5–5-fold. The prevention of glycosylation of Ser¹⁰ favored transcription, suggesting that O-GlcNAcylation at this site inhibits ΔLf transcriptional activity. Mimicking phosphorylation at Ser¹⁰ rendered ΔLf even more active than ΔLf^M4 (Fig. 2E) and strongly suggests that the presence of a phosphate group on this site favors transactivation (Fig. 5B). This result reinforces the status of ΔLf^S10D as a constitutive phosphorylated mutant. Because Ser¹⁰ is present in a basic environment (¹MRKVRGPPVSCIKR¹⁴) within a putative truncated DNA-binding domain, we constructed a ΔLf^Δ1–14 mutant in which the first 14 amino acid residues were deleted. Surprisingly, this deletion did not affect ΔLf transcriptional activity (Fig. 5B), suggesting that the ΔLf DNA-binding domain must be located at the hinge region (31, 32). Because O-GlcNAcylation and phosphorylation might occur on neighboring sites, we screened the vicinity of Ser¹⁰ and identified Ser¹⁶ that might be used as a replacement target by kinases. We therefore constructed a ΔLf^S16D mutant in order to mimic phosphorylation at this site. Expression of this mutant led to a basal expression level of the reporter gene (Fig. 5B), showing that constitutive phosphorylation at this locus does not lead to increased transactivation as for ΔLf^S10D and does not take over when the major acceptor site is invalidated, confirming the key role of Ser¹⁰.

Because ΔLf transcriptional activity is altered by O-GlcNAcylation at Ser¹⁰ and an OGT·OGA complex has been described in the vicinity of transcription factors bound to their response elements (8, 9), we next considered whether glycosylated ΔLf binds DNA. Using a ChIP assay we investigated the binding of the different Ser¹⁰ mutants compared with WT. As shown in Fig. 5C, specific ChIP PCR products were detected for each mutant. It is interesting to note that the PCR product signals for ΔLf^WT and ΔLf^S10+ were equivalent, whereas treatment with GlcNH₂ led to a weaker signal for both, suggesting that fewer promoter sites were occupied. Because ΔLf^WT and ΔLf^S10+ were equivalently expressed (Fig. 4B) even under GlcNH₂ treatment (Fig. 1, B and C, respectively), we suggest that glycosylation inhibits binding to DNA and that among the ΔLf intracellular pool, only the Ser¹⁰ phosphoforms bind ΔLfRE. These results were confirmed by the detection of a PCR product signal comparable with that of WT for ΔLf^S10D, which was poorly expressed but extremely active (Fig. 4B), suggesting that a large proportion of ΔLf^S10D binds ΔLfRE (Fig. 5C). The detection of a weaker signal for ΔLf^S10A, which was expressed similarly to WT, shows that without phosphorylation and glycosylation at Ser¹⁰, ΔLf still binds DNA, but its capacity to occupy promoter sites is reduced. Real time PCR was next performed to quantify promoter site occupancy (Fig. 5D). The qPCR data confirmed the PCR results except that promoter site occupancy for ΔLf^S10+ and ΔLf^S10D was twice as high as that of WT. Treatment with GlcNH₂ led to a 0.5-fold promoter site occupancy compared with WT, confirming that favoring GlcNAcylation prevents DNA binding.

In addition, we performed a re-ChIP assay to investigate whether ΔLf or a ΔLf-associated transcriptional complex binds to the endogenous human Skp1 promoter in vivo as a phosphoform. Moreover, since the half-life of ΔLf is short as a phosphoform, we studied the possibility that ΔLf also exists as a ubiquitinated isoform on DNA. Using a re-ChIP assay, we showed that phosphorylated and ubiquitinated but not O-GlcNAc ΔLf complexes were specifically co-localized on the Skp1 promoter fragment (Fig. 5E). The slight amplification observed in panel 1 (NIP and IR) might be due to the fact that the two immunoprecipitations were performed with the same antibody, increasing the background level. Our results clearly demonstrate that phosphorylated and/or ubiquitinated ΔLf or ΔLf associated with phosphorylated and/or ubiquitinated proteins specifically binds the Skp1 promoter segment with close proximity in vivo, whereas glycosylated ΔLf or ΔLf associated with glycosylated proteins does not. Because ΔLf is ubiquitinated at Lys³⁷⁹ and phosphorylated at Ser¹⁰, we suggest that these two post-translational modifications might be concomitantly present on ΔLf bound to DNA and may both be determinant in its activity. Further work will have to be done to demonstrate such a partnership, and for that, specific antibodies against the phospho-Ser¹⁰ or the Ub-Lys³⁷⁹ or poly-Ub-Lys³⁷⁹ will be obtained.

DISCUSSION

O-GlcNAc/P modification of transcription factors modulates their transcriptional activity by regulating their turnover, traffic, binding to DNA, or cofactor recruitment. ΔLf is a transcription factor controlling the expression of key molecular actors and as such should be highly regulated. In this study, we demonstrated that it is alternatively O-GlcNAcylated or O-phosphorylated at Ser¹⁰ and that these two alternative modifications play distinct roles in modulating its turnover and transcriptional activity.

The concentration of transcription activators and the rate of their degradation are under the control of the proteasome, and there is direct evidence that a switch between O-GlcNAcylation and phosphorylation regulates the process. Phosphorylation drives proteins to degradation via the capping of PEST hydroxyl groups, whereas O-GlcNAcylation hinders it mainly by competing for and masking these hydroxyl groups from kinases. Numerous proteins, such as the transcription factor Sp1 (14), the estrogen receptor (49), the eukaryotic initiation factor eIF2a-p67 (50), or p53 (16), are protected from proteasomal degradation by O-GlcNAcylation. Here, we show that ΔLf has a short half-life compatible with its function and is stabilized when Ser¹⁰ is O-GlcNAcylated. Moreover, we showed that Ser¹⁰ is not present within a phosphodegron, which is a recognition signal for Ub ligases. The ΔLf degradation motif (³⁹¹KSQQSSDPDPNCVD⁴⁰⁴) is conserved in Lf from different species, and the mutation of all three Ser residues led to increased stability of the protein, clearly confirming the functionality of this motif. Mutation of each Ser separately indicated that they behave similarly, suggesting that they are equivalent targets of kinases due to their proximity, but we do not know whether they are also substituted with GlcNAc moieties. YinOYang l.2 server predictions indicated Ser²⁹² and Ser³⁹⁵ as OGT targets but with low scores. The ΔLf^M4 isoform is not glycosylated, suggesting that no further glycosylation sites are present, but we cannot exclude glycosylation of the PEST motif only when Ser¹⁰ is glycosylated or the possibility that the ΔLf^M4 isoform, which is extremely unstable, exists only as a phosphorylated PEST isoform.

We next investigated Ub targets by mutating lysine residues neighboring the PEST motif and demonstrated that ΔLf ubiquitination occurs on Lys³⁷⁹ and Lys³⁹¹ with Lys³⁷⁹ as the main target. The ΔLf^KK double mutant was devoid of Ub, confirming that only these two residues are involved. The formation of Ub ladders observed with ΔLf^K379 and ΔLf^K391 also revealed that, despite the possibility of its multimonoubiquitination, ΔLf undergoes polyubiquitination. Unexpectedly, the ΔLf^PEST mutant was still ubiquitinated, suggesting the existence of other degradation motifs. ΔLf is involved in S phase control and should be ubiquitinated via the SCF complex, but it is possible that another complex, such as anaphase promoting complex/cyclosome, might be involved. Interestingly, ΔLf possesses a ⁴⁷⁵RSNLCAL⁴⁹¹ sequence, which may behave as a potential RXXLXX(L/I/V/M) D-box motif (ELM D-box entry), the target of anaphase promoting complex/cyclosome (51). The presence of two degradation motifs suggests that ΔLf may be degraded throughout the cell cycle. Nevertheless, further work has to be done in order to prove the functionality of this D-box.

The relationship linking O-GlcNAcylation and the Ub pathway has not yet been elucidated. Although Yang et al. (16) demonstrated that O-GlcNAcylation inhibits ubiquitination of p53, a recent study by Guinez et al. (21) shows that O-GlcNAc and Ub can coexist on the same protein and suggests that the Ub/O-GlcNAc ratio may send proteins either to destruction or repair. Here, we demonstrated that enhancement of the O-GlcNAc status within the cells inhibited ΔLf ubiquitination, and the absence of Ser¹⁰ as in the ΔLf^S10A mutant was accompanied by a decrease in polyubiquitination, suggesting that this modification of ΔLf^S10+ and ΔLf^WT only occurs on the phosphoforms. Phosphorylation at Ser¹⁰, by acting through the creation of a negatively charged region and/or the triggering of transient conformational changes, may lead to phosphorylation at the PEST locus, conferring a priming site role on Ser¹⁰.

The O-GlcNAc/P interplay also modulates transcriptional activity. O-GlcNAcylation directly activates FoxO1 (52), p53 (53, 54), and NF-κB (55) and Sp1 indirectly via cofactors (14, 15), whereas it inhibits c-Myc (24) and mouse estrogen receptor β (49). In this work, we have demonstrated that GlcNAcylation inhibits ΔLf transcriptional activity, whereas phosphorylation activates it, and that Ser¹⁰ is central to this regulation. An absence of modification at Ser¹⁰ leads to gene transactivation, whereas phosphomimetism increases it, confirming the inhibitory role of glycosylation. Because the expression of ΔLf^S10+ is much greater than that of ΔLf^S10D, we suggest that ΔLf exists normally in the cell as a pool of stable but inactive glycoforms that, under appropriate stimuli, become activated by phosphorylation and sensitive to degradation. However, another explanation is that only the phosphoform is present in the nucleus. Nucleocytoplasmic traffic may be regulated via O-GlcNAcylation because the modification of the O-GlcNAc status leads to a change in the cellular distribution of Tau (56), Alpha4, and Sp1 (57) but does not influence Stat5a traffic (58). Here, we showed that ΔLf-GFP traffic was not affected by GlcNH₂ or OA treatment. But even if the nucleocytoplasmic traffic is not governed by the O-GlcNAc/P interplay, because the OGT·OGA complex and kinases are present within both compartments (59, 60), nuclear ΔLf might exist only as a phosphoform.

Phosphorylated transcription factors are usually more competent to bind DNA and activate transcription than their non-phosphorylated counterparts, but there is direct evidence for the involvement of O-GlcNAcylation. PDX-1 O-GlcNAcylation increases its ability to bind DNA (61) and enhances p53 DNA binding by hiding an inhibitory domain at the N terminus (54). O-GlcNAcylation of HIC1 does not affect its specific DNA binding (62), and whatever the modifications present on Stat5, it binds its response element similarly (58). However, O-GlcNAcylation at the C terminus of Sp1 abolishes homopolymerization and dramatically affects its function (15). In this study, we demonstrated that in vivo ΔLf binding to ΔLfRE occurred with the unmodified or the phosphomimetic Ser¹⁰ isoform but decreased when O-GlcNAcylation was increased, suggesting that the glycoform is unable to bind DNA. ΔLf^S10A bound DNA and transactivated transcription at a basal level, but given the dynamic nature of the O-GlcNAc/P interplay, it is doubtful whether an unmodified ΔLf isoform exists. Nevertheless, we can infer that transactivation by ΔLf is a two-step process, starting at a basal level and increasing with phosphorylation, as depicted in Fig. 6. Moreover, using the re-ChIP assay, we were able to show that the ΔLf transcriptional complex linked to ΔLfRE is phosphorylated. Our data demonstrate that O-GlcNAcylation at Ser¹⁰ inhibits DNA binding, whereas phosphorylation favors it and promotes transactivation.

FIGURE 6.

Illustration of the regulation of Δ Lf activity and stability by the O-GlcNAc/P interplay. ΔLf exists as an inactivated and stable Ser¹⁰-O-GlcNAcylated isoform pool in cells. Depending on cellular events, such as transcription, translation, cell cycle progression, or cell signaling, or in the case of cell stress/injury, OGA will trigger ΔLf deglycosylation, unmasking the Ser¹⁰ hydroxyl group. Deglycosylated ΔLf can bind DNA and transactivate basal transcription. Phosphorylation at Ser¹⁰ by an appropriate kinase leads to a strong amplification of the process and mediates PEST phosphorylation, poly-Ub at Lys³⁷⁹, and ΔLf degradation. P, phosphate.

The O-GlcNAc/P content fluctuates during cell cycle progression. A recent study showed that increasing O-GlcNAc levels induces a slowing down of both S and G₂/M phases, whereas a reduced O-GlcNAc level impairs the G₁/S checkpoint transition (63). Because temporal control of Ub-proteasome-mediated protein degradation is critical for normal G₁ and S phase progression, ΔLf modifications may switch between glycosylation and phosphorylation, depending on the cell cycle phase. Progression of the cell cycle requires degradation of cyclins and cyclin inhibitors. At the G₁/S check point, Skp1, one of the targets of ΔLf, is involved in the process when associated with the SCF complex (64, 65). Thus, O-GlcNAcylation of ΔLf, by down-regulating Skp1 expression, may alter SCF activity, whereas phosphorylation of ΔLf may increase it. Regulation of the transcriptional activity of ΔLf by the O-GlcNAc/P interplay may therefore modulate the Ub-proteasome-mediated degradation of cell cycle regulators. Furthermore, we demonstrated that ΔLf is itself ubiquitinated; thus, its turnover could be regulated by feedback control via overexpression of Skp1. On the other hand, ubiquitination also occurs on the ΔLf·DNA complex. Modification by Ub is not only a destruction signal but also determines membrane receptor internalization, sorting at the endosomal compartment, activation of DNA repair, or transactivation of transcription factors, such as c-Myc and SRC-3 (66–68). As an example, SRC-3 is first activated by multi-(mono)ubiquitination and then polyubiquitinated prior to degradation. Therefore, ΔLf might require concomitant preubiquitination and phosphorylation as a transcriptional activation signal before being degraded as a polyubiquitinated isoform.