Biosynthesis of O-N-acetylgalactosamine glycans in the human cell nucleus

Biological functions of nuclear proteins are regulated by post-translational modifications (PTMs) that modulate gene expression and cellular physiology. However, the role of O-linked glycosylation (O-GalNAc) as a PTM of nuclear proteins in the human cell has not been previously reported. Here, we examined in detail the initiation of O-GalNAc glycan biosynthesis, representing a novel PTM of nuclear proteins in the nucleus of human cells, with an emphasis on HeLa cells. Using soluble nuclear fractions from purified nuclei, enzymatic assays, fluorescence microscopy, affinity chromatography, MS, and FRET analyses, we identified all factors required for biosynthesis of O-GalNAc glycans in nuclei: the donor substrate (UDP-GalNAc), nuclear polypeptide GalNAc -transferase activity, and a GalNAc transferase (polypeptide GalNAc-T3). Moreover, we identified O-GalNAc glycosylated proteins in the nucleus and present solid evidence for O-GalNAc glycan synthesis in this organelle. The demonstration of O-GalNAc glycosylation of nuclear proteins in mammalian cells reported here has important implications for cell and chemical biology.

The nucleus is one of the most important structures of eukaryotic cells. This complex organelle stores the chromosomes and also regulates their duplication, segregation, repair, and expression through a series of specific processes. The cell's biological information is saved and transferred within the nucleus by three types of biopolymer molecules: DNA, RNA, and proteins (1). Proteins play crucial roles in nuclear scaffolding, DNA assembly, replication, transcription, and transport of molecules. The biological activity of proteins is directly modulated by their conformation, and changes in protein conformation are controlled mainly by post-translational modifications (PTMs). 3 The common PTMs of nuclear proteins are acetyla-tion, sumoylation, phosphorylation, long-chain fatty acid conjugation, and glycosylation. A given PTM adds a tag to proteins that can be recognized by specific molecules (e.g. bromodomain for acetyl residue; lectin for glycan) as a trigger of biological effect (2,3). Thus, biological functions of nuclear proteins can be regulated via PTMs. PTMs of nuclear proteins play a central role in epigenetic physiology, i.e. modulation by environmental factors of cellular phenotype other than by the effects of genetic encoding of information.
Glycosylation is the most common PTM of proteins; Ͼ50% of cellular proteins are potential targets of glycosylation. Protein O-GlcNAc glycosylation (biosynthesis of O-GlcNAc glycans) occurs in the nucleus, cytoplasm, and mitochondria (4). O-GlcNAc is added by multiple alternatively spliced isoforms of the enzyme GlcNAc transferase (OGT), which have different intracellular localizations (5). There are three major OGT isoforms: nucleocytoplasmic OGT, short OGT, and mitochondrial OGT. Nucleocytoplasmic OGT and short OGT are localized in the nucleus and cytoplasm, whereas mitochondrial OGT is present in the mitochondrial matrix. O-GlcNAcase, a glycosidase that removes O-GlcNAc from proteins, functions in a cycling fashion with OGT. OGT acts as an epigenetic "writer," whereas O-GlcNAcase acts as an "eraser" of this PTM by modulating biological activity of relevant nuclear proteins such as histones, RNA polymerase II, and transcription factors. The O-GlcNAc PTM thus plays a crucial role in nuclear homeostasis (6). The hexosamine biosynthesis pathway generates UDP-GlcNAc and UDP-GalNAc from glucose (Glc), acetyl-CoA, ATP, uridine, and amino acids (7). The UDP-hexosamine level thus depends on Glc concentration as well as salvage pathways of GalNAc and GlcNAc. O-GlcNAc glycan biosynthesis is affected by metabolic diseases (notably diabetes) in which Glc concentration is altered (8). O-GlcNAc glycosylation of nuclear proteins has a major effect on altered transcription in diabetes (9).
O-GalNAc glycans are the second most common glycan structures on secreted proteins, after N-glycans. Their biosynthesis is initiated by action of a multigene family of enzyme polypeptide-N-acetylgalactosaminyltransferases (ppGalNAc-Ts) promoting covalent linkage of GalNAc from UDP-GalNAc donor to Ser/Thr of acceptor, yielding GalNAc␣1-O-Ser/Thr (10). Twenty members of the ppGalNAc-T family have been found in humans. The initial step of O-GalNAc glycosylation is critical in defining the amino acid position of PTM on the protein and in facilitating subsequent monosaccharide incorporation. The second monosaccharide linked to GalNAc␣1-O-Ser/Thr may be galactose (Gal) or N-acetylglucosamine (GlcNAc) yielding core 1 (Gal␤3GalNAc␣1-O-Ser/Thr) or core 3 (GlcNAc␤3GalNAc␣1-O-Ser/Thr) glycan, respectively. The core structures are extended by action of specific glycosyltransferases to generate complex O-GalNAc glycans, which are synthesized mainly in Golgi (11). Truncated O-GalNAc glycans are commonly synthesized by epithelial tumor cells with overexpression of GalNAc␣1-O-Ser/Thr (Tn antigen) or Gal␤3GalNAc␣1-O-Ser/Thr (T antigen) residues, possibly by deletion or loss of the glycosyltransferases elongating Tn or T antigens (12,13), and play crucial roles in cell adhesion during the process of metastasis (14).
We examined O-GalNAc glycan biosynthesis in the nucleus of human cells as a PTM of nuclear proteins that may play an important role in regulating their functions. Our focus was the identification of all factors in the cell nucleus necessary for initiation of O-GalNAc glycan biosynthesis: the sugar donor substrate, the nuclear polypeptide GalNAc-transferase activity, an enzyme (polypeptide GalNAc-T3), and O-GalNAc-glycosylated proteins (the products of glycan biosynthesis).

UDP-GalNAc in cell nucleus
To examine nuclear localization of several molecules, we purified HeLa cell nuclei as described previously (15). Nuclei were separated from the cytoplasm of whole cells using Nonidet P-40 detergent in appropriate buffer and by centrifugation. Purified nuclei were obtained by several washes with buffer without detergent (Fig. 1), placed in hypertonic buffer, and sonicated for disruption of nuclear membrane, and the soluble nuclear fraction (nucleoplasm) was obtained by centrifugation. Quality of purified nuclei was evaluated by confocal microscopy and Western blotting (WB) with molecular markers of cellular organelles: Golgin97 (Golgi), calreticulin (ER), tubulin (cytoplasm), and histone H3 (nucleus). Each of these methods revealed high levels of nuclear markers (PI and histone) in purified nuclei or nucleoplasm but minimal levels of cytoplasm, Golgi, and ER markers. This finding indicated that the method used for purification of nuclei was appropriate.
Biosynthesis of nucleotide sugars such as UDP-GalNAc occurs in the cytoplasm (16). UDP-GalNAc is the sugar donor substrate required for ppGalNAc-T reaction. To examine the presence of this sugar donor in the nucleus, we developed an enzymatic assay for measurement of UDP-GalNAc substrate, in which the standard curve showed an appropriate linear concentration range (Fig. S1A). High specificity of this enzymatic method for UDP-GalNAc was demonstrated by comparison with UDP-GlcNAc, which was found not to participate in the reaction. UDP-GalNAc level was next measured in various sub-cellular fractions of HeLa cells (Table 1). UDP-GalNAc concentration was very different in the nucleoplasm (0.330 M) than in the last nuclear wash (0.015 M), confirming the presence of this sugar nucleotide in the nucleus.
To examine transport of UDP-GalNAc through the nuclear membrane, we added UDP-GalNAc to purified nuclei and measured the amount that passed inside. A substantial concentration (0.510 M) of UDP-GalNAc was found in the nucleoplasm, reflecting the ability of this sugar nucleotide to enter the nucleus (Table 1).

Nucleus synthesizes O-GalNAc glycans
(HMC) using naked MUC1 and MUC2 as acceptor substrates and excessive UDP-GalNAc as donor substrate. The enzymatic product, ␣GalNAc residues, was evaluated using the VVL probe, and ppGalNAc-T activity was determined by extrapolation from the standard curve of purified MUC1␣GalNAc (Fig.  S1C). ppGalNAc-T activity was detected in the cytoplasm and nucleoplasm of analyzed cells (Table 2), whereas catalytic activity was not observed in the last nuclear wash fraction. Important ppGalNAc-T activity was detected in all analyzed nucleoplasms, reflecting the enzyme's capacity for O-GalNAc glycosylation of naked mucins.
ppGalNAc-T activity was next evaluated in the intact nuclei of HeLa cells. Purified nuclei were incubated with added UDP-GalNAc for 1 h at 37°C, and the yielded glycans were detected by WB and confocal microscopy using labeled lectins. Increased numbers of terminal ␣-linked GalNAc residues in multiple nuclear proteins were demonstrated by WB with HPA ( Fig. 2A). Terminal ␣-linkage of GalNAc to proteins was demonstrated using ␣-N-acetylgalactosaminidase (␣GalNAc glycosidase). Purified nuclei previously incubated with UDP-GalNAc (OG nuclei) were washed, sonicated for nuclear membrane disruption, and incubated without or with ␣GalNAc glycosidase. The number of GalNAc residues in glycosylated nuclear proteins was greatly reduced in the presence of ␣GalNAc glycosidase, indicating that terminal GalNAc was ␣-anomeric. The findings that intact nuclei display GalNAc-T activity to glycosylate nuclear proteins and that this process is reversed by ␣GalNAc glycosidase indicate the capacity of nuclei for ␣GalNAc glycoprotein biosynthesis.
A comparison of the effects of UDP-GalNAc and UDP-GlcNAc preincubation on glycosylation capacity of intact nuclei is shown in Fig. 2B. Labeled VVL recognized an increased level of terminal O-GalNAc glycans when purified nuclei were added with UDP-GalNAc, whereas UDP-GlcNAc addition was not recognized by VVL. When nuclei were added with UDP-Gal-NAc, WB analysis with detection by WGA did not recognize an increase in O-glycosylation level. These findings indicate a clear distinction between initiation of O-GalNAc versus O-GlcNAc glycosylation of proteins in purified nuclei.
We examined O-GalNAc glycosylation inside the nucleus by confocal microscopy of purified intact nuclei of HeLa cells.
Constitutive O-GalNAc residues in these nuclei are shown using VVL (Fig. 3A), as well as HPA and anti-Tn antibody ( Fig.  S2) (top). The interaction of VVL with constitutive O-GalNAc residues of intact nuclei was inhibited in the presence of GalNAc (Fig. S3). Incubation of nuclei with added UDP-Gal-NAc enhanced the product of O-GalNAc glycosylation ( Fig. 3A and Fig. S2, bottom). Z-stacks of purified nuclei showed that in    Finally, we studied ppGalNAc-T nuclear activity in intact CHO ldlD cells. This cell line lacks a functional UDP-Gal-4epimerase and therefore relies on GalNAc salvaged from the medium for the synthesis of UDP-GalNAc. CHO ldlD cells were grown in media supplemented without (ϪGalNAc) or with (ϩGalNAc) GalNAc, and the yielded glycans were detected by confocal microscopy using labeled lectins (Fig. 4). Incubation of cells with GalNAc in growing media enhanced the product of O-GalNAc glycosylation (Fig. 4A), resulting in a 5.8-fold increase in nuclear signal in ϩGalNAc CHO ldlD cells (Fig. 4B). The subcellular distribution of terminal GalNAc in the cells reveals that a high portion of the signal coincides with the nuclear marker DAPI (Fig. 4C), indicating nuclear ppGal-NAc-T activity in intact cells.

Nuclear localization of ppGalNAc-T3
GalNAc-Ts are localized mainly in the Golgi; however, a few ppGalNAc-T isoforms have been reported in other locations, e.g. ER (17). We examined subcellular localization of isoforms ppGalNAc-T2 (T2) and ppGalNAc-T3 (T3), with emphasis on T3. Fluorescence microscopy assays with anti-human T2 and T3 antibodies was performed to study subcellular localization of these isoforms in HeLa, MCF-7, T47D, SK-N-AS, HEK-293, MRC-5, Vero, and HMC cells (Fig. S4). T2 showed a characteristic Golgi staining pattern in these cells (Fig. S4A). T3 showed a nuclear staining pattern (overlap with nuclear marker DAPI) in HeLa, MCF7, SK-N-AS, HEK-293, MRC-5, Vero, and HMC, but not T47D. Distribution of T3 coincided mainly with the nuclei, whereas T2 showed a Golgi distribution. Evidence for the specificity of anti-T3 antibody was provided by WB and competitive immunofluorescence assay using purified soluble T3. Recombinant human soluble T3 was expressed in insect cells and purified by affinity chromatography. We demonstrated the purity of T3 by SDS-PAGE with Coomassie Brilliant Blue (CBB) staining, which revealed a protein of appropriate molecular mass (72 kDa) (Fig. S4B). Anti-T3 antibody, on WB, recognized a protein with corresponding molecular weight. This antibody also recognized a unique protein with molecular weight corresponding to that of T3 in HeLa cell homogenate. In competitive immunofluorescence assay, recognition of T3 by the antibody in HeLa cells was inhibited by the presence of purified soluble T3 (Fig. S4C). The presence of soluble T3 strongly reduced both the dot-shaped nuclear pattern and the Golgi staining pattern of T3, indicating the ability of antibody to specifically recognize endogenous T3 in HeLa cells.
T3 nuclear distribution was analyzed by confocal microscopy in purified nuclei, compared with whole HeLa cells (Fig. 5). Delimitation of the nucleus was detected by staining of nucleic acids with PI, and T3 was detected with anti-T3 antibody. Purification of nuclei removed T3 localized in Golgi but conserved nuclear T3, as observed in many purified nuclei (Fig. 5A). Colocalization of T3 with PI was observed in zoomed regions of whole-cell nuclei and purified nuclei. Orthogonal views (x, y and x, z views) of nuclear Z-stacks reveal co-localization of T3 with PI through the Z planes (Fig. 5B), confirming localization of T3 inside the nucleus of HeLa cells.

Identification of O-GalNAc-glycosylated nuclear proteins
Proteins with O-GalNAc glycans were identified using a combination of affinity chromatography of immobilized VVL to select O-GalNAc glycoproteins and MS for protein identifi-

Nucleus synthesizes O-GalNAc glycans
cation. Nucleoplasm from purified nuclei (constitutive O-GalNAc glycosylated nuclear proteins) and nucleoplasm from OG nuclei of HeLa cells were analyzed. Each nucleoplasm was divided into two equal parts: one part was loaded onto streptavidin-agarose column (ϪVVL column) to identify unspecific adsorption, and the other part was loaded onto VVL-biotin/streptavidin-agarose column (ϩVVL column) to retain O-GalNAc-glycosylated proteins. Differences of retained proteins between the ϩVVL and ϪVVL column were evaluated by SDS-PAGE with colloidal CBB staining (Fig. S5). Proteins from gel bands were digested, and peptides were identified by MS. The criterion for the presence of O-GalNAc glycans on proteins was the identification of Ն2 high-quality peptides per protein, and peptide spectrum matches (PSM) ratio Ն3 (see "Experimental procedures"). Identified O-GalNAcglycosylated proteins in order of decreasing PSM ratio, their subcellular localization, and main protein function according to UniProt (http://www.uniprot.org), 4 are listed in Table 3. Two proteins were identified as endogenously O-GalNAcglycosylated in the nucleoplasm: 40S ribosomal protein S6 (RPS6) and S3 (RPS3) ( Table 3A). They are both localized in nuclei, according to UniProt. Twenty five glycoproteins were detected in OG nucleoplasm, indicating the predominance of nuclear localization in this group (Table 3B). PSM ratios of RPS6 and RPS3 were 4 and 3, respectively; in contrast, some proteins in the OG nucleoplasm group had much higher PSM ratios (30 in the case of prelamin-A/C). These findings suggest a direct correlation between O-GalNAc over-glycosylation, major glycoprotein retention by ϩVVL column, and increased PSM ratio of proteins in the nucleoplasm. Incubation of purified intact nuclei with UDP-GalNAc led to detection of OG proteins localized predominantly in the nucleus, indicating that O-GalNAc glycosylation occurs inside the nucleus. The identified nuclear proteins with the highest O-GalNAc glycosylation levels were prelamin-A/C (PSM ratio ϭ 30) and lamin B1 (LMNB1) (PSM ratio ϭ 22). We therefore examined nuclear O-GalNAc glycosylation of lamins in more detail.

O-GalNAc glycosylation of LMNB1
Software programs are available with accepted algorithms that allow theoretical predictions regarding O-GalNAc glycosylation sites on proteins. Prediction of O-GalNAc glycosylation on LMNB1 was made using the NetOGlyc 4.0 Server (www.cbs.dtu.dk/services/NetOGlyc/) 4 (18). LMNB1 had a large number of sites with a high likelihood of O-GalNAc glycosylation (Table S1), consistent with our identification of

Nucleus synthesizes O-GalNAc glycans
LMNB1 as a highly O-GalNAc-glycosylated protein following incubation of purified nuclei with UDP-GalNAc.
O-GalNAc glycosylation of LMNB1 was evaluated by co-localization and FRET assays using fluorescence confocal microscopy. Labeled Cherry-LMNB1 was overexpressed in HeLa cells. Purified nuclei from these cells were incubated without or with UDP-GalNAc, and the resulting O-GalNAc glycosylation was studied using biotin-VVL and Alexa 488 -streptavidin.
Correlation analysis between VVL (channel 1) and LMNB1 (channel 2) was performed for control nuclei and OG nuclei (Fig. 6A). Correlation of signals was stronger in the OG nuclei, as a consequence of nuclear GalNAc-T activity. Fluorescence profiles of the two channels were analyzed and showed greater accompaniment of fluorescent intensities in the OG nuclei. Comparison of fluorograms showed enhanced correlation between LMNB1 and O-GalNAc residues following incubation with UDP-GalNAc. The correlation was quantified by Pearson's correlation coefficient for nuclei without or with UDP-GalNAc preincubation and glycosylation detected with biotin-VVL (␣GalNAc) or biotin-WGA (␤GlcNAc) (Fig. 6, B and C). The correlation coefficient between LMNB1 and O-GalNAc residues showed a 4.1-fold (0.630/0.151) increase after UDP-GalNAc incubation and biotin-VVL detection, but no significant change after biotin-WGA detection. These findings suggest that LMNB1 is O-GalNAc-glycosylated in the nucleus.
O-GalNAc glycosylation of LMNB1 was further evaluated by acceptor photobleaching/FRET. Again, purified nuclei from Cherry-LMNB1-overexpressing cells were incubated without or with UDP-GalNAc to study LMNB1 O-GalNAc glycosylation, and biotin-VVL or biotin-WGA with Alexa 488streptavidin were used for O-glycan detection. FRET index for each experimental condition was calculated as described under "Experimental procedures." A map of FRET index was prepared, corresponding to representative images of nuclei preincubated without or with UDP-GalNAc and detected with biotin-VVL (Fig. 7A) or biotin-WGA (Fig. 7B). FRET index was significantly higher for OG nuclei with O-GalNAc termini of LMNB1 detected with biotin-VVL (Fig. 7C). Detection with biotin-WGA gave no notable difference for control nuclei versus OG nuclei (Fig. 7, B and D). The increase in FRET index with biotin-VVL detection confirms that O-GalNAc residues are added to LMNB1 in the nucleus, showing the ability of nuclei to synthesize O-GalNAc glycans.

Discussion
PTMs are essential modulators of cell homeostasis, play key roles in protein function and localization, and regulate cell

Nucleus synthesizes O-GalNAc glycans
interactions in a variety of biological processes. Many nuclear proteinsundergoPTMs(notablyacetylation,methylation,phosphorylation, ubiquitination, and glycosylation in mammalian cells) that affect gene expression and nuclear physiology (7,19,20). We studied biosynthesis of O-GalNAc glycans in the nucleus of human cells and documented the presence of all factors necessary for initiation of O-GalNAc protein glycosylation: the sugar donor substrate, catalytic ppGalNAc-T activity, an enzyme (polypeptide GalNAc-T3), and products of the enzymatic reaction. O-GalNAc glycosylation of proteins is initiated by ppGalNAc-Ts catalyzing covalent linkage of GalNAc to the hydroxyl residue of Ser or Thr. This enzymatic reaction requires two substrates: a sugar donor (UDP-GalNAc) and a protein acceptor substrate. Mucin-type domains are more frequently O-GalNAc-glycosylated in amino acid sequences having predominant presence of Ser and Thr, surrounding Pro. Proteins with mucin-type domains are widely distributed in cells, including the nucleus. We developed a highly specific and sensitive quantification method to detect UDP-GalNAc substrate in nuclei, and we used it to measure a 0.330 M concentration of UDP-GalNAc in HeLa cell nucleoplasm. Hart and Akimoto (21) reported a similar concentration (0.50 M) of UDP-GlcNAc in the nucleoplasm, whereas Bond and Hanover (22) described UDP-GlcNAc concentrations ranging from 2 to 30 M in the cytoplasm, nuclei, and mitochondria. As in glycosylation reactions, two substrates are required, and the relevance of this UDP-GalNAc concentration in the nucleus is conditioned by enzyme's K m value of donor substrate, which is depending on the nuclear amino acid sequence of the acceptor substrate. When we incubated purified intact nuclei with added UDP-GalNAc, we observed a substantial concentration (0.510 M) of UDP-GalNAc in the nucleoplasm. These findings reflect the ability of UDP-GalNAc to reach the nucleoplasm by crossing the nuclear membrane. UDP-GalNAc synthesized in cytoplasm is thus able to cross the nuclear membrane and be available as a sugar donor substrate for nuclear ppGalNAc-Ts. Bond and Hanover (22) described similar nuclear membrane permeability and the ability of UDP-GlcNAc to enter the nucleus. Concentrations of UDP-GalNAc and its epimer UDP-GlcNAc are directly affected by nutrient availability. Similarly to previous findings for UDP-GlcNAc concentration, metabolic alterations that affect the UDP-GalNAc concentration may alter nuclear ppGalNAc-T activity and thereby regulate levels of O-GalNAc glycosylation on nuclear proteins.
We demonstrated significant ppGalNAc-T activity in all nucleoplasms of several cells, reflected by the catalytic capacity to link GalNAc to naked MUC1 and MUC2 (Table 2). HMC nucleoplasm showed ppGalNAc-T activity with MUC2 accep-

Nucleus synthesizes O-GalNAc glycans
tor, but it was not observed on the MUC1 acceptor. It could be explained because this enzyme activity is conditioned by the ppGalNAc-T isoforms present in the nucleoplasm of each cellular type. The fact that ppGalNAc-T activity is present in nontumoral (HEK-293, Vero, and HMC) and tumoral (HeLa, T47D, and MCF-7) nucleoplasms, as well as in nuclei of CHO ldlD cells suggests that the presence of O-GalNAc glycosylation machinery in the nucleus would be a normal physiological process. ppGalNAc-T activity was also observed in HeLa purified nuclei, and incubation of purified nuclei with added donor substrate (UDP-GalNAc) enhanced the O-GalNAc glycosylation of several proteins. Studies of ␣-anomeric GalNAc linkage to nuclear proteins, and linkages of various controls (including UDP-GlcNAc and WGA), showed that the N-acetylhexosamine linkage in this case is not ␤GlcNAc, as described previously for other nuclear proteins. Following preincubation of purified nuclei with UDP-GalNAc, confocal microscopy assays revealed O-GalNAc glycosylation inside the nucleus. Another key point in this study was the demonstration of nuclear ppGal-NAc-T activity in intact cells. CHO ldlD cells are a robust model for studying the initiation of O-GalNAc glycosylation given that they are deficient in UDP-Gal and UDP-GalNAc 4-epimerase and are therefore unable to synthesize UDP-Gal or UDP-GalNAc. The addition of GalNAc to CHO ldlD cell culture medium allowed us to demonstrate that the incorporation of O-GalNAc terminals is also occurring in the nuclei of intact cells.
Identification of O-GalNAc-glycosylated nuclear proteins indicated the occurrence of a PTM not previously described in the nuclei and suggested that O-GalNAc glycosylation may modulate functions of certain nuclear proteins. Accordingly, we studied constitutive and over-glycosylated O-GalNAc nuclear proteins. O-GalNAc-glycosylated proteins from HeLa nucleoplasm were purified on an affinity chromatography column using immobilized VVL, and retained glycoproteins were identified by MS. Two proteins (RPS6 and RPS3), both known

Nucleus synthesizes O-GalNAc glycans
to be localized in nucleus and cytoplasm, were identified as constitutively O-GalNAc-glycosylated. It is likely that numerous other nuclear proteins are constitutively O-GalNAcglycosylated but were not retained on the VVL column because this lectin recognizes glycoproteins with terminal GalNAc residues. Constitutive O-GalNAc glycosylation could thus continue the glycan biosynthetic pathway with capping of the initial ␣GalNAc residue and evade VVL recognition.
Incubation of purified nuclei with UDP-GalNAc allowed identification of numerous O-GalNAc-glycosylated proteins. Of 25 such proteins (Table 3B), 22 are localized in the nucleus. This finding confirms that biosynthesis of O-GalNAc glycans occurs in the nucleus, because the approach involves incubation of purified nuclei with UDP-GalNAc. Among the 22 proteins as above, PSM ratios were highest for pre-LMNA/C and LMNB1. These two proteins, and LMNB2, are essential components of the nuclear lamina, a filamentous layer located between the inner nuclear membrane and peripheral heterochromatin that plays crucial roles in nuclear organization and interactions with the genome, including promoter regions that modulate gene expression (27,28). Other nuclear proteins identified as O-GalNAc-glycosylated (Table 3B) include the following: (i) splicing factor, proline-and glutamine-rich (SFPQ) and non-POU domain-containing octamer-binding protein (NONO), characteristic components of nuclear body paraspeckles, which are involved in control of gene expression during cellular processes such as differentiation, viral infection, and stress responses (29); (ii) ribosomal proteins that are involved in translation, DNA repair (RPS3), and mRNA catabolism (RPS18 and RPS3A) (30 -32); (iii) RNA-binding motif protein, X chromosome, and ATP-dependent RNA helicase A (DHX9) involved in gene expression, transcriptional activation, and RNA processing (33,34). In view of the varied functions of identified proteins as above, nuclear O-GalNAc glycosylation may play essential roles in nuclear homeostasis.
Additional methods were applied for the study of LMNB1, which had a high PSM ratio, and a prediction of highly probable O-GalNAc glycosylation. Co-localization analysis of O-GalNAc residues and LMNB1 revealed an increase of Pearson's correlation coefficient as a consequence of O-GalNAc over-glycosylation. Similarly, FRET index between O-GalNAc glycans and LMNB1 was increased by O-GalNAc over-glycosylation. Co-localization and FRET studies thus confirmed the nuclear O-Gal-NAc glycosylation of LMNB1, in agreement with findings from affinity chromatography and MS. Nuclear lamins play key roles in nuclear organization, nuclear physiology, and gene expression (35,36). They undergo extensive PTMs (e.g. phosphorylation and sumoylation) that help determine their localization and dynamics (37,38). Lamins can also be ADP-ribosylated or N-glycosylated (7,39). Wang et al. (40)  LMNB1 plays crucial roles in high-order chromatin organization, DNA replication, and transcriptional activity (42,43).
In conclusion, this study provides solid evidence of O-GalNAc glycan biosynthesis machinery in the nucleus of human cells. All factors necessary for initiation of such biosynthesis are present in the nucleus: the required donor substrate (UDP-GalNAc); nuclear ppGalNAc-T activity; an enzyme (T3); and the identified O-GalNAc-glycosylated proteins in the nucleus. This is the first demonstration of O-GalNAc glycosylation of nuclear human proteins. Because UDP-GalNAc biosynthesis is sensitive to cellular Glc concentration, biosynthesis of O-GalNAc glycans in the nucleus may be significantly altered in situations involving changes in sugar metabolism, e.g. diabetes. Increased levels of O-GalNAc residues (e.g. Tn antigen) are often observed in epithelial cancer cells and likely may reflect changes in nuclear O-GalNAc glycan abundance. Diabetes and epithelial cancer are both associated with widespread changes in gene expression, which likely involve O-GalNAc glycans. PTMs of nuclear proteins play key roles in many nuclear physiological processes. It is therefore important to further elucidate the functions of O-GalNAc glycans in nuclear proteins. This is the focus of our ongoing studies.

Nuclei and nucleoplasm purification
Nuclei and nucleoplasm were purified as described by Shechter et al. (15) with some modifications. In brief, cultured cells were trypsinized, harvested, and washed with PBS by centrifugation at 300 ϫ g for 10 min at 4°C. 1 ϫ 10 7 cells were incubated with 1 ml of extraction buffer (10 mM HEPES, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 0.34 M sucrose, 10% glycerol) containing 0.2% Nonidet P-40 and protease inhibitors for 10 min on ice, with occasional rotation for cell lysis. The sample was centrifuged at 6,500 ϫ g for 5 min at 4°; the supernatant (cytoplasm) was recovered, and the pellet (intact nuclei) was washed three times with 1 ml of extraction buffer (without Nonidet P-40). Purity of nuclei was evaluated by immunofluorescence confocal Nucleus synthesizes O-GalNAc glycans microscopy using mouse anti-Golgin 97 and mouse anti-calreticulin antibody as membrane organelle markers and Alexa 488-labeled rabbit anti-mouse IgG antibody as secondary reagent. Nuclei were stained with PI.
To obtain soluble the nuclear fraction (nucleoplasm), purified nuclei as above were lysed by incubation with 400 l of high-salt solubilization buffer (50 mM Tris-HCl, pH 8.0, 2.5 M NaCl, 0.05% Nonidet P-40) for 20 min at 4°C and then sonicated. The sample was centrifuged at 16,000 ϫ g for 10 min at 4°C, and supernatant (nucleoplasm) was recovered. A portion of each fraction equivalent to 1.5 ϫ 10 5 cells was subjected to 4 -20% SDS-PAGE. Purity of nuclei and nucleoplasm was evaluated by Western blotting with organelle molecular markers. Samples were also analyzed by CBB staining as loading control.
For glycan detection, fixed, permeabilized, and blocked cells or nuclei were incubated with biotinylated lectins: H. pomatia agglutinin (biotin-HPA; 1:1,000; L6512; Sigma); V. villosa lectin (biotin-VVL; 1:2,500; B-1235; Vector Laboratories); or wheat germ agglutinin (biotin-WGA; 1:500; B-1025, Vector Laboratories) for 2 h at room temperature. Monoclonal IgM anti-Tn antibody (5F4) was incubated overnight at 4°C, washed, and then incubated with biotinylated goat anti-mouse IgM antibody (1:1,000, BA-2020, Vector Laboratories) in PBS, 1% BSA for 2 h at room temperature. Samples were washed and stained for 1 h at room temperature with Alexa 488 -streptavidin (1:2,000; 532354; Life Technologies, Inc.). Controls for immunostaining specificity were included with Alexa-conjugated anti-rabbit or -mouse IgG antibody or Alexa-streptavidin, but without primary antibody or biotinylated lectin. Nuclei were stained with DAPI and PI. Samples were mounted onto glass slides using Flu-orSave (Calbiochem). Images of ppGalNAc-T2 (T2) and ppGal-NAc-T3 (T3) in the four human cell lines were obtained with a fluorescence microscope (Carl Zeiss, Axioplan) using PlanApoN 60 ϫ 1.42 NA oil immersion objective. Images showing subcellular distribution of T3 and O-GalNAc glycosylation were obtained with a confocal microscope (model FV-1000, Olympus) using Pla-nApoN objective. Confocal images were acquired in sequential mode to avoid bleed-through between channels. Images were obtained in x, y and x, y, z scan modes. For zoomed images, region mode clip was performed, and stacks of equidistant (0.10 or 0.17 m) planes were acquired. Zoom was adjusted to achieve 0.045 m per pixel. Images were processed using the Fiji software program (45).

Expression and purification of recombinant human ppGalNAc-Ts
Human soluble ppGalNAc-T2 (T2) and ppGalNAc-T3 (T3) cDNAs were cloned into baculovirus expression vector pAcGP67 as described previously (47). Secreted, soluble proteins were purified from supernatant of Sf9 cell culture medium following centrifugation at 2,000 ϫ g for 30 min at 4°C. The supernatant was dialyzed (membrane MWCO Ͻ10 kDa) against PBS and centrifuged at 2,500 ϫ g for 30 min at 4°C. Proteins were purified using HisPur TM cobalt resin (Thermo-Fisher Scientific), eluted with 150 mM imidazole, dialyzed three times against PBS, and concentrated by a centrifuge filter device (MWCO Ͻ10 kDa; Millipore). Total proteins were measured by bicinchoninic acid assay with BSA as standard (Pierce; ThermoFisher Scientific). Purity of recombinant human ppGalNAc-Ts was assessed by 10% SDS-PAGE and CBB staining.

UDP-GalNAc measurement
UDP-GalNAc was measured by enzymatic assay. Microtiter plates were adsorbed with 2 g/ml MUC1 peptide in coating buffer overnight at 4°C, washed with PBS, and blocked with PBS with 0.1% Tween 20 for 1 h at room temperature. Catalytic reaction mixture containing 25 mM sodium cacodylate, pH 7.4, 10 mM MnCl 2 , 0.05% Tween 20, and 10 nM purified T2 in a total volume of 45 l was incubated with either 5 l of HeLa cytoplasm, last nuclear wash before obtaining nucleoplasm, or nucleoplasm for 15 min at 37°C. Additional multiwell plates adsorbed with 2 g/ml MUC1 were incubated with catalytic

Nucleus synthesizes O-GalNAc glycans
reaction mixture and various concentrations of UDP-GalNAc (Sigma) as standard, for construction of a reference curve. After catalytic reactions, plates were washed with PBS, incubated with biotin-VVL (1:1,000) in PBS with 0.05% Tween 20 for 60 min at room temperature, washed again with PBS, incubated with HRP-streptavidin (1:2,500) in PBS with 0.05% Tween 20 for 30 min at room temperature, and washed again with PBS. Peroxidase colorimetric reaction and absorbance reading were as described above.
To evaluate the capacity of UDP-GalNAc to enter the nucleus, 50 l of purified nuclei were incubated with 50 l of UDP-GalNAc (500 M) in TBS for 10 min on ice, and washed three times with 1 ml extraction of buffer (10 mM HEPES, pH 7.9, 10 mM KCl, 1.5 mM MgCl 2 , 0.34 M sucrose, 10% glycerol) by centrifugation at 6,500 ϫ g for 5 min at 4°C. Purified nuclei were sonicated, and the supernatant (nucleoplasm) was obtained by centrifugation at 16,000 ϫ g for 10 min at 4°C. UDP-GalNAc was measured in samples corresponding to the last nuclear wash and nucleoplasm.
For construction of the standard curve, microtiter plates were adsorbed with various concentrations of MUC1␣GalNAc in coating buffer overnight at 4°C, washed with PBS, and blocked with PBS with 0.1% Tween 20 for 1 h at room temperature. ␣GalNAc residues were detected using biotin-VVL as described in above. GalNAc-T activity was expressed as international units (1 unit ϭ 1 mol of ␣-linked GalNAc/min) and normalized relative to total proteins (units/mg).

O-GalNAc glycosylation in purified nuclei
Purified nuclei were divided into three equal samples and incubated with equal volumes of TBS with either 10 mM MnCl 2 (endogenous glycosylation), TBS with 10 mM MnCl 2 and 500 M UDP-GalNAc (O-GalNAc over-glycosylated nuclei; "OG nuclei"), or TBS with 10 mM MnCl 2 , and 500 M UDP-GlcNAc (control) for 1 h at 37°C. Nuclei were then washed three times with TBS, and the resulting purified nuclei were analyzed by fluorescence microscopy and WB. For fluorescence microscopy studies, treated nuclei were seeded on polylysine-coated coverslips, fixed, blocked, permeabilized, and incubated with biotinylated lectins (HPA and VVL) or monoclonal IgM anti-Tn antibody (5F4) to reveal ␣GalNAc residues. Nuclei were stained with PI and imaged by confocal microscopy (model FV-1000, Olympus) in x, y and x, y, z scan modes. Region mode clip was performed, and stacks of 20 -30 equidistant (0.17 m) planes were acquired. Zoom was adjusted to achieve 0.045 m per pixel. Images were acquired using identical settings of laser power, detector gain, and offset.
HeLa cells, cytoplasm, purified nuclei, purified nuclei preincubated with UDP-GalNAc (OG nuclei), and purified nuclei preincubated with UDP-GlcNAc (control) were analyzed by WB using biotin-VVL and biotin-WGA. A portion of each fraction equivalent to 1.5 ϫ 10 5 cells was subjected to 4 -20% SDS-PAGE and CBB-stained or electrotransferred to nitrocellulose membranes. O-GalNAc and O-GlcNAc glycoproteins were detected, respectively, with biotin-VVL and biotin-WGA. Purified OG nuclei were sonicated, incubated without or with chicken liver ␣-N-acetylgalactosaminidase (A9763; Sigma) in 100 mM sodium citrate, pH 5.5, for 2 h at 37°C, and analyzed by SDS-PAGE (12% acrylamide) and WB with biotin-HPA detection.

Quantification of nuclear O-GalNAc glycosylation
Purified nuclei and OG nuclei were processed for fluorescence microscopy, stained with Alexa 488 -streptavidin (negative control: without lectin) or with biotin-VVL followed by Alexa 488 -streptavidin, and finally stained with PI. Samples were imaged in x, y, z scan mode, and stacks of 3-4 equidistant (0.7 m) planes were obtained. Images were acquired with identical settings of laser power, detector gain, and offset. Z-stack images were processed by Fiji program. Signal in the 488 channel (corresponding to O-GalNAc glycosylation detected with VVL and nonspecific signal in negative controls) was measured. The z axis profile and selection of the nuclear plane (546 channel) (PI) were performed for each stack. A nuclear mask created by nuclear signal thresholding (546 channel) (Huang method) was used to determine mean intensity in the 488 channel for each nucleus, under each condition. Background was pre-subtracted for both channels.
CHO ldlD cells grown in coverslips were processed for fluorescence microscopy in the same way that purified nuclei, and

Nucleus synthesizes O-GalNAc glycans
the cell nucleus was stained with DAPI. Images of cells were acquired with identical settings and were processed in the same way using Fiji software. To measure the signal in the nuclear region, nuclear masks were generated by DAPI nuclear signal thresholding. Background was subtracted, and the mean fluorescence in the nuclear region corresponding to nuclear terminal O-GalNAc was measured in 58 -60 individual cells for each condition (ϩGalNAc/ϪGalNAc). The resulting fluorescence values were relativized to the average fluorescence in the ϪGalNAc condition (the condition with lower O-GalNAc glycosylation), and finally, measurements were expressed as relative nuclear mean fluorescence. All data were analyzed using the GraphPad Prism 5 software program.

Specificity of rabbit anti-human T3 antibody
Purified human T3 (1 g) or HeLa homogenate (40 g) was loaded onto 10% SDS-PAGE, electrotransferred, and analyzed by WB using rabbit anti-human T3 antibody and IRDye 800conjugated goat anti-rabbit IgG antibody (1:20,000). Specificity of T3 (72 kDa) recognition was analyzed based on the molecular weight of recognized protein.
Specificity of the anti-human T3 antibody was evaluated by competitive immunofluorescence assay. HeLa cells were grown on coverslips and subjected to immunofluorescence microscopy. Fixed, permeabilized, and blocked cells were incubated for 2 h at room temperature with anti-human T3 antibody (1:500) either alone (control) or with purified human T3 (0.9 g/ml). Coverslips were washed and incubated with Alexa 546conjugated anti-rabbit IgG antibody (1:1,000), and nuclei were stained with DAPI. Samples were imaged as described under "Fluorescence microscopy," using identical settings of laser power, detector gain, and offset.

Identification of O-GalNAc-glycosylated proteins
O-GalNAc glycoproteins from HeLa nucleoplasm (endogenous glycosylation) and nucleoplasm from OG nuclei were purified by affinity chromatography using VVL-biotin/ streptavidin-agarose column (ϩVVL column). ϩVVL column was prepared by incubation of 100 g of biotin-VVL with 100 l of streptavidin-agarose CL-4B (85881; Sigma) for 1 h at 4°C, washed with 10 volumes TBS, and divided into two 50-l parts. To obtain nucleoplasm, 2 ϫ 10 7 HeLa-purified nuclei or OG nuclei (preincubated with UDP-GalNAc for 1 h at 37°C) were lysed in high-salt solubilization buffer as described previously (15). Each nucleoplasm was pre-adsorbed with 50 l of streptavidin-agarose column for 1 h at 4°C. Resulting nucleoplasms were divided into two equal parts, which were incubated, respectively, with 50 l of VVL-biotin/ streptavidin-agarose column (ϩVVL column) and with 50 l of streptavidin-agarose column without VVL (ϪVVL column) for 1 h at 4°C. Columns were washed with 20 volumes of TBS, resuspended in Laemmli sample buffer, and heated at 90°C for 10 min. Samples were loaded onto SDS-PAGE (12% acrylamide), run until 1-cm front separation, and gel stained with colloidal CBB G (B1131; Sigma). Gel regions corresponding to ϩVVL and ϪVVL columns were analyzed at a MS facility (Centro de Estudios Químicos y Biológicos por Espectrometría de Masa, Universidad de Buenos Aires, Argentina). Samples were digested with trypsin, and peptides were purified by nano-LC-MS/MS in a Q-Exactive mass spectrometer (ThermoFisher Scientific). A H 2 O/acetonitrile gradient at a flow rate of 33 nl/min was used with a C18 2-mm EASY-Spray Accucore (ES801; ThermoFisher Scientific) coupled to Q-Exactive-Orbitrap hybrid spectrometer (ThermoFisher Scientific). The top 12 peaks in each cycle were fragmented by the data-dependent MS2 method. Data analysis was performed using the Proteome Discoverer software program, version 1.4. Based on the results, the following criterion was established. Proteins were considered O-GalNAc-glycosylated for those hits in the ϩVVL column having Ն 2 high-quality peptides and peptide spectrum matches ratio (PSM ratio) of Ն3. PSM ratio was calculated as PSM for a hit identified in the ϩVVL column divided by PSM of the same hit in the ϪVVL column. Thus, the criterion was PSM ratio (ϩVVL column PSM/ϪVVL column PSM) of Ն3.

O-GalNAc over-glycosylation of nuclei expressing LMNB1
HeLa cells (80% confluence) were transfected for 2 h using Lipofectamine Transfection Reagent (ThermoFisher Scientific) with mCherry-LMNB1-10 plasmid (55069; Addgene). Transfected cells were incubated for 24 h at 37°C, harvested, and subjected to subcellular fractionation for nuclei purification. Purified nuclei were incubated without or with UDP-GalNAc in glycosylation buffer for 1 h at 37°C as described above, washed with PBS, seeded onto polylysine-coated coverslips, and subjected to fluorescence microscopy.

O-GalNAc/LMNB1 correlation analysis
HeLa-purified nuclei and OG nuclei expressing Cherry-LMNB1 were stained with biotin-VVL (1:1,000) or biotin-WGA (1:500) (control) and then with Alexa 488 -streptavidin (1:2,000). Samples were imaged by confocal microscopy (model FV-1200, Olympus) using PlanApoN objective. Images were acquired in constant acquisition settings and sequential mode to avoid bleed-through between channels and processed using the Fiji program. Background was subtracted; a Gaussian filter with ϭ 2 was applied, and correlation analysis of green (Alexa 488/channel 1) and red (cherry/channel 2) channels was performed. Intensity profiles of the two channels were obtained using BAR extension 1.1.6. Fluorograms and Pearson's correlation coefficients were obtained by the JaCoP extension, using Costes' automated thresholding method.

Acceptor photobleaching/FRET assay
HeLa nuclei and OG nuclei were treated as indicated above. FRET assay was performed using a confocal microscope (model FV-1200, Olympus) with PlanApoN objective. Images were acquired in constant acquisition settings and sequential mode before and after bleaching. Samples were subjected to bleaching for 2 ms in the acceptor channel (Cherry-LMNB1) receiving complete photobleaching. The area of bleaching region was Mean FRET index (n ϭ 5) in the bleaching area was quantified for each condition. Images were processed using Fiji and FV10-ASW 3.1 software programs.

Statistical analysis
Means were compared by Mann-Whitney test or unpaired t test, using GraphPad Prism 5. Standard error of the mean (S.E.) is shown as error bars in figures. Statistical significance of differences between means is indicated by * (p Ͻ 0.05); ** (p Ͻ 0.01); *** (p Ͻ 0.001), or ns (not significant).