Transcription factor NFIC undergoes N-glycosylation during early mammary gland involution.

Expression of a 74-kDa nuclear factor I (NFI) protein is triggered in early involution in the mouse mammary gland, and its expression correlates with enhanced occupation of a twin (NFI) binding element in the clusterin promoter, a gene whose transcription is induced at this time (Furlong, E. E., Keon, N. K., Thornton, F. D., Rein, T., and Martin, F. (1996) J. Biol. Chem. 271, 29688-29697). We now identify this 74-kDa NFI as an NFIC isoform based on its interaction in Western analysis with two NFIC-specific antibodies. A transition from the expression of a 49-kDa NFIC in lactation to the expression of the 74-kDa NFIC in early involution is demonstrated. We show that the 74-kDa NFIC binds specifically to concanavalin A (ConA) and that this binding can be reversed by the specific ConA ligands, methyl alpha-D-mannopyranoside and methyl alpha-D-glucopyranoside. In addition, its apparent molecular size was reduced to approximately 63 kDa by treatment with the peptide N-glycosidase. The 49-kDa lactation-associated NFIC did not bind ConA nor was it affected by peptide N-glycosidase. Tunicamycin, a specific inhibitor of N-glycosylation, blocked formation of the 74-kDa NFI in involuting mouse mammary gland in vivo when delivered from implanted Elvax depot pellets. Finally, the production of the ConA binding activity could be reiterated in "mammospheres" formed from primary mouse mammary epithelial cells associated with a laminin-rich extracellular matrix. Synthesis of the 74-kDa NFIC was also inhibited in this setting by tunicamycin. Thus, involution triggers the production of an NFIC isoform that is post-translationally modified by N-glycosylation. We further show, by using quantitative competitive reverse transcriptase-PCR, that there is increased expression of the major mouse mammary NFIC mRNA transcript, mNFIC2, in early involution, suggesting that the involution-associated change in NFIC expression also has a transcriptional contribution.

Transcription factor, nuclear factor I (NFI), 1 is a family of heterogeneous transcription factors encoded by four independent genes, NFIA, NFIB, NFIC (also called CCAAT-binding transcription factor (CTF)), and NFIX (1)(2)(3)(4)(5)(6). NFI is required for the initiation of adenovirus DNA replication (7) and acts as a transcriptional regulator targeting a wide range of promoters (reviewed in Ref. 8). NFI binds as a dimer to a specific dyad symmetric binding sequence on duplex DNA (consensus: 5Ј-TTGGC/ A(N 5 )GCCAA-3Ј) (7, 9 -18). NFI proteins share a highly conserved DNA binding and dimerization domain (18 -20). However, considerable variation occurs within the amino acid sequence of the C-terminal domain of these proteins due to the existence of multiple alternately spliced transcripts from each of the four genes (3,(21)(22)(23). In total some 19 isoforms of NFI gene products have now been identified for NFIA (2,5,24), NFIB (1,5,25), NFIC/CTF (3-5, 21, 26, 27), and NFIX (2,6,23,29). Many of these transcripts have orthologous forms in multiple species. Thus, the alternative splicing in the 3Ј regions of the NFI transcripts is phylogenetically conserved suggesting conserved biological functions for each isoform (21). Complexity is further generated by alternative translational start codons generating different N termini (24) and alternative polyadenylation sites generating different 3Ј-untranslated regions (25). Alternative processing of NFI transcripts thus results in a myriad of possible NFI proteins, each of which may influence the regulation of gene expression, depending on whether it possesses transcriptional activating or repressing properties. Post-translational modification further expands the possible molecular and functional range for the NFI family; for instance, specific instances of modification by phosphorylation (29) and modification by O-glycosylation have been reported (30).
NFI proteins have been shown to contribute to different stages of the cycle of mouse mammary gland development. Lactose production, an exclusive activity of the lactating mammary gland, requires ␤1,4-galactosyltransferase. The NFIbinding site in the promoter of the ␤1,4-galactosyltransferase gene (31) may be important for its expression in the lactating gland. NFI has also been implicated in lactation-associated gene expression of milk protein genes. The whey acidic protein (WAP) gene promoter contains binding sites for NFI (32), and point mutations in the NFI sites dramatically decreased WAP gene expression in the transgenic mammary gland (33). In addition, different forms of NFI, present in nuclear extracts from the lactating mammary gland, were demonstrated to bind to at least five sites in the sheep ␤-lactoglobulin gene promoter (34). Laminin-rich extracellular matrix-triggered induction of NFI activity was shown in primary cultures of mammary epithelial cells that required this matrix interaction both for cell survival and terminal differentiation (35).
At weaning the lactating mammary gland undergoes a reductive remodeling termed involution. This is accompanied by up-regulated expression of genes such as clusterin, TGF-␤, c-fos, junD, and IGFBP-5 (36 -39). Earlier studies from our laboratory (40) reported the characterization of involution-enhanced occupation of a twin NFI-binding element in the clusterin promoter. Associated studies led to the detection of a 74-kDa NFI protein whose expression is triggered in early involution in the mouse mammary gland (40). Thus, a specific NFI activity may be required in early involution, and it may differ from NFI activities that subserve transcription during lactation. We now report that this involution-associated NFI isoform is an NFIC/CTF and most likely arises from an involution-triggered post-translational modification that generates an N-glycosylated form of this transcription factor. As far as we are aware, this is the first N-glycosylated transcription factor that has been detected.

EXPERIMENTAL PROCEDURES
Preparation of Tissue Extracts-Nuclear extracts were prepared from mammary gland tissue from lactating, post-weaning (involuting), or "resuckled" CD-1 mice, as described previously (40). For the resuckling experiments, pups from another lactating mother were removed and placed with mice whose mammary glands had been allowed to involute for 40 h, for the desired period, between 3 and 36 h.
Elvax pellets containing 10.0 g of tunicamycin mixed with 0.5 mg of bovine serum albumin were prepared, and one or two pellets were implanted into the fourth inguinal mammary gland of anesthesized lactating mice from which the pups had just been removed to initiate involution, all exactly as described previously (40). After 24 h, mammary tissue was harvested from proximal to and distal to the site of pellet insertion and from the contralateral gland, and nuclear extracts were prepared.
Cell Culture, Transient Transfection, and Whole Cell Extract Preparation-Jeg3 cells, a human choriocarcinoma cell line (41)(42)(43)(44), were obtained from the American Type Culture Collection (ATCC HTB-36) and were cultured in Eagle's minimum essential medium containing Earle's salts, supplemented with 1.5 g/liter sodium bicarbonate, 0.1 mM non-essential amino acids, 1.0 mM sodium pyruvate, 2 mM L-glutamine, 100 units/ml penicillin, 0.1 mg/ml streptomycin, and 10% fetal bovine serum at 37°C with 5% CO 2 in a humidified atmosphere. The culture medium was changed every 2nd day. The cells were transiently transfected with the hemagglutinin (HA)-tagged NFI expression vectors (pCHmNFI-A, -B, -C, and -X) (45), a generous gift from Dr. R. Gronostajski (Lerner Research Institute, Cleveland, OH), using LipofectAMI-NE TM Reagent (Invitrogen) according to the manufacturer's instructions. Following transfection (48 h) the cells were harvested by removing the culture medium and washing with ice-cold phosphatebuffered saline and were then scraped into 1 ml of phosphate-buffered saline and briefly centrifuged to pellet the cells for whole cell extract preparation.
Mammary epithelial cells for primary culture were harvested from mid-pregnant mice, and mammosphere cultures were prepared exactly as described by Blatchford et al. (46). After 5 days culturing was continued for a further 2 days in the presence or absence of prolactin (1 g/ml). Tunicamycin (10 g/ml in medium) was added on day 7 of culture for the last 6 h. Cells were harvested by removing the culture medium from the mammospheres and digesting the extracellular matrix using 1 ml of Dispase (BD PharMingen) per 35-mm plate. This was incubated at 37°C for 1 h, and the reaction was then stopped by the addition of 10 mM EDTA, pH 8.0. The mammosphere suspension was briefly centrifuged to pellet the mammospheres, and they were washed once with phosphate-buffered saline. The mammospheres were finally pelleted by gentle centrifugation. Whole cell extracts were prepared from Jeg3 cells and primary mammary epithelial mammospheres by resuspending pelleted cells in 50 -100 l of extraction buffer (10 mM HEPES, pH 7.9, 1 mM EDTA, 0.5 mM dithiothreitol, 400 mM KCl, 1 mM Na 3 VO 4 , 1 mM ␤-glycerophosphate, 5 mM benzamidine, 10% glycerol, and 1 g/ml each of aprotinin, leupeptin, and pepstatin A). The samples were allowed to settle on ice for 15 min before being disrupted by trituration. The homogenate was then incubated on ice for 30 min before centrifugation at 14,000 rpm at 4°C for 20 min. The supernatant constituted the whole cell extract.
Solid Phase Lectin Affinity Chromatography-Lectins used in this study were immobilized on 4% cross-linked agarose (Sigma); 1.0 mg of each agarose-bound lectin was washed twice in 500 l of its binding buffer as follows: for concanavalin A lectin (ConA), 20 mM Tris-HCl, 0.5 M NaCl, 1 mM MnCl 2 , 1 mM CaCl 2 , pH 7.4; for wheat germ agglutinin (WGA) lectin, 20 mM Tris-HCl, 0.5 M NaCl, pH 7.4. Each agarose-bound lectin was then resuspended in another 500 l of its binding buffer and incubated, rotating to ensure homogeneity of the suspension, with 150 -450 g of protein extract at 4°C overnight. To inhibit competitively the interaction of the lectin with the glycosylated proteins, specific inhibitory sugars were incubated with the lectin and the protein sample as follows: for ConA, 250 mM methyl ␣-D-mannopyranoside ϩ methyl ␣-D-glucopyranoside; for WGA, 500 mM GlcNAc. The complexes were then washed three times by centrifugation at 14,000 ϫ g for 5 min followed by gentle resuspension in 1 ml of the appropriate binding buffer. The final pellet was resuspended in concentrated electrophoresis buffer (2% SDS, 10% glycerol, 10 mM Tris-Cl, pH 6.8, 2% ␤-mercaptoethanol, 0.02% bromphenol blue) for boiling and SDS-PAGE analysis. To release the 74-kDa NFIC from the ConA-agarose, the washed ConAagarose pellet was resuspended in the ConA binding buffer with 500 mM methyl ␣-D-mannopyranoside and rotated for 1-2 h at 4°C. The lectinagarose complexes were then pelleted by centrifugation at 14,000 rpm for 5 min, and the supernatant was used in SDS-PAGE and Western analysis. To release the NFIC activity for EMSA analysis, the washed agarose pellet was resuspended in 1 ϫ "bandshift" binding buffer (10 mM MgCl 2 , 1 mM EDTA, pH 8.0, 0.75 mM dithiothreitol, 7.5% glycerol, 0.05% Nonidet P-40 (40) containing poly(dI-dC) (200 ng/l) and 500 mM methyl ␣-D-mannopyranoside.
Exoglycosidase Treatment of Protein Samples with PNGase F (New England Biolabs Catalog Number P0704)-Protein samples were denatured in 1ϫ denaturing buffer (0.5% SDS, 1% ␤-mercaptoethanol) at 100°C for 10 min. The denatured samples were then incubated in 1ϫ G7 buffer (50 mM sodium phosphate, pH 7.5), 1% Nonidet P-40, and 1,000 -10,000 units of PNGase F in a final volume of 30 l overnight at 37°C. Other glycosidases were used as follows: for neuraminidase (New England Biolabs catalog number P0720), the protein samples were incubated in 1ϫ G1 buffer (50 mM sodium citrate, pH 6.0) and 50 -250 units of neuraminidase in a final volume of 30 l overnight at 37°C; for ␤ϪN-acetylhexosaminidase (New England Biolabs catalog number P0721), protein samples were incubated in 1ϫ G2 buffer (50 mM sodium citrate, pH 4.5) and 5-25 units of ␤-N-acetylhexosaminidase in a final volume of 30 l overnight at 37°C.
Potential N-glycosylation sites within the mNFI-C2 amino acid sequence were sought using the PIR pattern search program (49).
RNA Extraction, RT-PCR, and Quantitative Competitive PCR-Total RNA was isolated from tissues using the procedure of Chomczynski and Sacchi (50). Total RNA was isolated from cultured cells using TRIZOL Reagent (Invitrogen) according to the manufacturer's instructions. Reverse transcription and limiting cycle PCR were carried out essentially as described previously (47). Depending on the abundance of the transcript being detected, 15-40 PCR cycles were carried out. In some cases it was necessary to carry out a nested PCR to increase the specificity of the reaction. This entails diluting the initial PCR product 1/1000 or greater and re-amplifying using a second set of primers, one or both of which lies internal to the initial set of primers used. 1-5 l of the dilution was used in a PCR similar to the primary PCR.
Competitive PCR was carried out to analyze quantitatively the levels of mNFIC2 expression. RNA competicons (competitors) were constructed for this study using an RT-PCR Competitor Construction Kit

Differential Gene Expression and Transcription
Factor Activity in Mammary Gland-Differential gene expression in the mammary gland associated with separable functional stages (e.g. lactation and involution) has been well characterized (36,37) as has the differential presence or activation pattern of a range of specific transcription factors (53). We could therefore show (Fig. 1A) by using limited cycle RT-PCR that the expression of the milk protein genes WAP and ␤-casein is predominant in lactation, less so after 2 and 3 days of involution, but is again clearly induced on resuckling of 40-h involuting glands for 36 h, showing re-establishment of the "differentiated"/lactational phenotype (54). In contrast, the involution-associated expressed genes TGF-␤1 (36), clusterin (36,37,55), and IG-FBP-5 (38) are minimally expressed in lactation and are expressed by day 2 of involution, and their expression is repressed by resuckling. Equal expression of GAPDH at all time points reflected equal input amounts of RNA in the RT reaction. Similarly, by using EMSA analysis (Fig. 1B)  , third panel), was apparent in nuclear extracts from lactating and resuckled glands but not in the involutional gland extracts (Fig. 1B, lanes 2 and 3), further verifying our model in which we have examined transcription factor NFIC expression (see below). Thus, by EMSA ( Fig. 1B) we show that the transition from lactation to involution is associated in a change in the dominant species of NFI DNA-binding complex from a relatively smaller (Fig. 1B, lane 5) to a relatively larger species (Fig. 1B, lanes 6 and 7) and that this transition is reversed in the re-differentiated resuckled gland (Fig. 1B, lane 8). We have reported previously that this involution-associated complex (Fig. 1B, lanes 5 and 6) is supershifted by an anti-NFI-specific antibody and that it contains, predominantly, a 74-kDa NFI isoform (40).
The 74-kDa Involution-associated NFI Is a CTF/NFIC Isoform-mNFI-C2 has a DNA binding domain (amino acids 1-221), almost entirely encoded by a single exon, exon 2 (4,8), that is strongly conserved in all NFI family members and a unique C-terminal transactivating/repressing domain (amino acids 222-439) that diverges considerably from NFIA, NFIB, and NFIX family members (5,18,21). Fig. 2A shows the epitopes against which the two anti-human nuclear factor IC C-terminal directed antibodies, ␣NFI 8199 and ␣NFI 2902, used in this study are directed. 3 The specificity of these antibodies was tested using whole cell extracts of Jeg3 cells (which are relatively NFI-deficient (45)) transfected with NFI expression vectors (pCHmNFIA, pCHmNFIB, pCHmNFIC and pCHmNFIX), that encode each of the four mNFI gene-specific proteins downstream of an N-terminal hemagglutinin (HA) tag. Analysis of extracts from mock-transfected cells served as a negative control (Fig. 2B, lane 1). Western blot analysis using an anti-HA antibody demonstrated that each of the transfected Jeg3 cell lines expressed an NFI protein of the predicted size (Fig. 2B, left-hand panel) (45). The same cell extracts blotted with the C-terminally directed antibodies (␣NFI 8199 and ␣NFI 2902) demonstrate that both of these C-terminal-specific antibodies detect NFIC only and not NFIA, NFIB, or NFIX (Fig. 2B, middle and right-hand panel). We note that a weak band in the NFIX lane was detected with both these antibodies, but this was not at the expected size for NFIX (see Fig. 2B, left-hand panel, lane 5). Western analysis of nuclear extracts from lactating and 2-day involuting mammary gland with the ␣NFI 8199 and ␣NFI 2902 antibodies was then carried out. This showed the involution-associated 74-kDa NFI (40) to be a CTF/NFIC (Fig. 2C, lanes 2 and 4), and the antibody of higher apparent titer, ␣NFI 8199, detects a prominent 49-kDa NFIC in the lactating extracts (Fig. 2C, lane 1).
A time course Western analysis using the ␣NFI 8199 antibody (Fig. 2D, top panel) showed that once involution is initiated, there is a gradual loss in expression of the 49-kDa NFIC isoform complemented by the appearance (to a significant level by 18 h) of the larger 74-kDa isoform. This transition in NFIC isoform expression from the 49-to the 74-kDa species in early involution is reversed by resuckling of the 40-h involuting gland (Fig. 2D, 2nd panel) with reappearance of expression of the lactation-associated 49-kDa protein occurring quite rapidly (ϳ3 h) (Fig. 2D, lane S 3 ) and the 74-kDa protein being cleared from the gland in 18 h (Fig. 2D, lane S 18 ). Finally, in Fig. 2D (third panel) we demonstrate by Western "shift" analysis that "activated" and phosphorylated forms of the prolactin-modulating transcription factor, Stat 5, are detectable in the lactating mammary gland (Fig. 2D, third panel, lane L) but are lost from the mammary nuclear extracts in early involution (Fig. 2D, third panel, lanes 48 and 96) but regenerated on gland resuckling (Fig. 2D, third panel, lane 56) as would be predicted (54); this latter analysis was carried out to control and confirm the potential reversibility of the changes triggered by early involution in this model system.
Thus, a switch in expression of NFIC isoforms occurs in the transition from lactation to early involution. However, the longest mouse NFIC open reading frame reported to date encodes 439 amino acids (45) and has an apparent molecular mass, as judged by SDS-gel electrophoresis, of ϳ49 kDa. Thus, the involution-associated isoform (apparent molecular mass, 74 kDa) is too big to be a direct translation product of any NFI gene described to date and may therefore be generated by an involution-associated post-translational modification event(s).
The Involution-associated 74-kDa NFIC Protein Is Post- translationally Modified by N-Glycosylation-Transcription factors including NFIC have been reported previously to be modified by a unique type of O-glycosylation (a single N-acetylglucosamine residue bound by O-linkage (O-GlcNAc) to serine or threonine (30)). Interestingly, the PIR pattern search program (49) predicts the mNFIC2 sequence to have two potential N-glycosylation sites, NFSL and NWTE (Fig. 3A). Both predicted sites conform to the permissive consensus sequence, Asn-X-Ser/Thr (X indicates any amino acid except proline) (58). It is noteworthy that the first of these two sites is conserved in three of the four NFI amino acid sequences A, X, and C).
Lectins have been used to isolate and characterize transcription factors based on their glycosylation profile. For instance, because transcription factor Sp1 is modified by O-GlcNAc, it has been isolated and detected by its association with wheat germ agglutinin (WGA) (30). The lectin, concanavalin A (ConA), recognizes ␣-linked mannose, which occurs in the core structure of N-linked glycans (59). By using Western analysis, with the anti-NFIC-specific ␣NFI 8199 antibody as end point, we could show (Fig. 3B) that under non-denaturing conditions the 74-kDa NFIC protein in 2-day involuting mammary gland nuclear extracts (2Di) binds to ConA (Fig. 3B, 1st panel, lane 2) but not to WGA (Fig. 3B, 1st panel, lane 4). It did not bind to ConA in the presence of an excess of the specific competing sugars, methyl ␣-mannopyranoside and methyl ␣-glucopyranoside (Fig. 3B, 1st panel, lane 3). No such association of the 49-kDa lactation-associated NFIC with lectin was detectable (Fig. 3B, 2nd panel). The presence of the potential N-glycosylation target sites in the mNFIC primary sequence and the selective binding to ConA suggest that the 74-kDa NFIC has been modified by N-glycosylation. We repeated the analysis under denaturing conditions (in 1% ␤-mercaptoethanol and 0.5% SDS; Fig. 3B, 3rd panel) and could again show the association of the 74-kDa NFIC with ConA, but we now saw some binding to WGA. This suggests that the 74-kDa NFIC also contains some of the smaller O-GlcNAc modifications whose binding may be masked in the non-denatured protein, or denaturation has un-masked interacting sugars on the N-linked carbohydrate. Again, these interactions were blocked by an excess of specific competing sugars, and no such interactions  5 and 6). The arrowhead indicates the position of the involution-associated complex. P marks the position of the free probe.
were detectable with the 49-kDa lactation-associated NFIC (Fig. 3B, 4th panel). Fig. 3B (5th panel) shows that the ConAbound 74-kDa NFIC (lane 3) can be eluted by incubation with an excess of specific competing sugar (lane 4); lane 5 shows the residual protein still associated with the ConA. Finally (Fig.  3C), we could demonstrate that the 74-kDa NFIC is reduced in apparent molecular size (to 63 kDa) after incubation with a specific N-glycosidase, PNGase F (which cleaves between the innermost GlcNAc and asparagine residues of N-linked glycoproteins), but not by ␤-N-acetylhexosaminidase f (which catalyzes the hydrolysis of terminal GalNAc and GlcNAc residues from oligosaccharides, typically O-linked) or by neuraminidase (which catalyzes the hydrolysis of terminal ␣2-3 and ␣2-6 and ␣2-8 sialyl linkages on glycoproteins).
We have demonstrated previously (40) that the 74-kDa NFIC is the NFI species in the complex generated by 2-day involuting mammary gland nuclear extracts with the NFI-specific DNAbinding element (as detected by EMSA). We now show that treatment of 2-day involuting nuclear extracts with ConA results in the elimination of the 2-day involution-associated shifted complex (Fig. 3D, lanes 1 and 2). We also demonstrate that treatment of the "ConA-74-kDa NFIC" complex (prepared from 2-day involuting nuclear extracts) with an excess of competing sugar (methyl ␣-D-mannopyranoside) releases an activity that is capable of generating a complex identical to the 2-day involuting NFI EMSA-shifted complex (compare lanes 4 and 6). Thus, ConA harvests from the 2-day involuting nuclear extract the complete biochemical activity we have associated previously with the 74-kDa NFIC protein, that is an ability to generate a complex of a specific size with the double-stranded DNA NFI-binding element (as judged by EMSA). The activity associated with the lactation-associated NFI DNA-binding complex could not be harvested from nuclear extracts by ConA (Fig. 3D, compare lanes 3 and 5).
Tunicamycin Inhibits Formation of the 74-kDa NFIC in Vivo-Tunicamycin is a potent inhibitor of N-glycosylation. It blocks the initial step of formation of the precursor oligosaccharide in the dolichol pathway, in the endoplasmic reticulum. It is a sugar-nucleotide analogue and competitive inhibitor of UDP-GlcNAc-transferase. This enzyme catalyzes the first step in dolichol pyrophosphate oligosaccharide synthesis (59). One (containing 1.0 g) or two Elvax depot pellets (containing 0.5 g of tunicamycin each) were placed in the right-hand 4th inguinal mammary gland immediately after initiating involution with a view to inhibiting N-glycosylation of the involutionassociated NFIC. The mice carried these tunicamycin Elvax depots for 24 h, but the exposure did have systemic effects as reflected by a very significant increase in fluid intake over this 24-h period (data not shown). Western analysis of nuclear extracts from the mice carrying both a single (Fig. 4A, lanes  4 -6) and two pellets (Fig. 4A, lanes 7-9) show a loss of the 74-kDa NFIC from both the glands carrying the pellet(s) (Fig.  4A, lanes 4 and 5 and 7 and 8) and from the contralateral glands (Fig. 4A, lanes 6 and 9). In order to confirm the specificity of the loss of the 74-kDa NFI by tunicamycin treatment, we measured the expression of two other transcription factors whose levels in the nucleus are induced in early involution, Stat 3 (60, 61) and AP-1 (37). We could show, by Western analysis, that nuclear Stat 3 levels were similar to that seen in a 1-day involuting control animal (appropriate time point/lane for comparison, Fig. 4B, lane 1) in the mice carrying both one and two pellets (Fig. 4B, lanes 3-5 and 6 -8). The AP-1 levels, as measured by DNA binding activity (using EMSA (37)) were also as seen in a 1-day involuting control animal (Fig. 4C). These results show that the synthesis of the involution-associated 74-kDa NFIC in the early involuting mammary gland can be inhibited, in vivo, by the inhibition of N-glycosylation with tunicamycin and that this effect of tunicamycin is specific, as the nuclear levels of two other transcription factors whose activity is associated with early involution, Stat 3 and AP-1, remained unchanged.
Expression of the 74-kDa NFIC Can Be Reiterated in a Mammary Epithelial Cell Culture System, and Tunicamycin Inhibits Its Formation-Mammary epithelial cells cultured on a laminin-rich extracellular matrix can form multicellular structures termed mammospheres in which cells associate basolaterally with the matrix. Evidence of apoptosis can be found on days 3-5 of culture as the mammospheres lose non-matrixassociated internal cells and generate a lumen (46). In the presence of lactogenic hormones, cells within the mammospheres become polarized, form tight intercellular junctions, and secrete milk proteins vectorially into the luminal space. Expression of the involution-associated 74-kDa NFIC is induced in the mammosphere cultures (Fig. 5A). Mammospheres in culture for 1 day rapidly (and reproducibly) lost their expression of the 49-kDa NFIC (Fig. 5, lanes 1 and 2). There follows a very modest expression of the 74-kDa NFIC on days 2 and 3 of culture during lumen formation, but there is a clear expression of this protein by day 5 that is intensified by day 7. Although these mammospheres are viable at day 7 they would be beginning to show "involution-like" changes, triggered by "milk stasis," in their lumens. It is highly possible that the increased expression of the 74-kDa NFIC at this time is associated with this. Like the involution-associated 74-kDa NFIC, it could be shown that this mammosphere-associated protein is N-glycosylated, as it bound to ConA (Fig. 5B). Addition of tunicamycin to these cultures led to the inhibition of the expression of the 74-kDa NFIC, further suggesting that this protein is post-translationally modified by N-glycosylation (Fig. 5C, upper panel). However, we did not see this inhibition when the cultures were maintained in the presence of prolactin (data not shown). To ensure that the inhibition of the expression of the 74-kDa NFIC by tunicamycin did not arise from its general cellular toxicity, we measured total cellular extracellular signal-regulated kinase levels. These were unaffected by the treatment (Fig. 5C, lower panel).
NFI RNA Transcript Expression in the Mouse Mammary Gland-It was of interest also to establish if change(s) in NFI gene transcription, particularly NFIC gene transcription, accompanied the induction of expression of the involution-associated 74-kDa NFIC. By using NFIA, -B, and -X family-specific PCR primers, we first cloned and characterized all transcripts expressed in lactating and involuting mouse mammary gland. All identified transcripts (NFIA1 and -A2; NFIB1, -B2, and -B3; and NFIX1 and -X2) (Fig. 6A) were found to be expressed both during lactation and early involution. Thus, no evidence of discrete differential expression of transcripts of any of these three genes during the transition from lactation to involution was obtained. Nor did limited cycle semi-quantitative PCR reproducibly show differences in expression of NFIA (A1 and A2), NFIB (B1, B2, and B3), or NFIX (X1 and X2) transcripts in the transition from lactation to early involution (results not shown and see Fig. 7A). Regarding NFIC, seven human CTF/ NFIC transcripts, which arise from differential splicing, have been described (3,26,27), but only two mouse transcripts are reported so far (45). 4 By using exon 2 and 3Ј-UTR sequencespecific primers (exon 2 is DNA-binding domain encoding and to date has not been reported to be spliced out) and "nested PCR" (Fig. 6B), we amplified, subcloned, and sequence-characterized the predominant NFIC transcripts from multiple independent lactating, involuting, and resuckled mouse mammary gland samples. We found mNFIC2 and mNFIC5 to be the most abundant NFIC transcripts in mouse mammary gland by se-  1, 2, 3, 5, and 7); and control lactating (L) and 2-day involuting mouse mammary gland (2di). The filled arrowhead indicates the position of the 74-kDa band and the open arrowhead the 49-kDa band. Bottom panel, the same region of the filter shown in Western blot above stained with PonceauS as loading control (for mammosphere culture samples (lanes 0, 1, 2, 3, 5, and 7)). B, ␣NFI 8199 probed Western blot analysis of supernatant (S) from nuclear extracts of mammary epithelial cells cultured under conditions that generate mammospheres after treatment with ConA-agarose; and protein recovered from associated ConA-agarose pellet (P). The arrowhead indicates the position of the 74-kDa band. C, upper, ␣NFI 8199 probed Western blot analysis of nuclear extracts from the following: control lactating (L) and 2-day involuting mouse mammary gland (2di); and mammary epithelial cells cultured under conditions that generate mammospheres for 7 days, ϩ vehicle (Ϫ), and ϩ tunicamycin (10 g/ml) (for last 16 h of culture) (ϩ) (lanes 1-3, respectively). The filled arrowhead indicates the position of the 74-kDa band and the open arrowhead the 49-kDa band; and lower same as upper but Western blot probed with anti-Erk1/2 antibody. quencing more than 100 independent cloned cDNAs. mNFIC2 is the equivalent of hCTF-2; it has exon 9 spliced out and the first nucleotide of exon 10, thereby generating a frameshift in the translated protein (3). mNFIC1A, mNFIC1B, 4 and mNFIC (45) are the mouse equivalents, but they differ from each other in the small exon 1. Our strategy excluded identifying the exon 1 sequence. mNFIC5 is identical in exonic structure to hCTF-5 (27). Other novel, less abundant transcripts (accounting for 1-10% of transcripts sequenced) were designated mNFIC8, mNFIC9, mNFIC10, and mNFIC11 (Fig. 6B); they do not correspond to any previously described NFIC transcripts; their proposed "exon" content is depicted in Fig. 6B.
NFIC Transcript Expression Levels in Lactation and Involution-Semi-quantiative "limited cycle" RT-PCR was initially used to measure the expression of the dominant mNFIC2 and C5 transcripts (Fig. 7A). Two independent primer sets (Fig. 7A,  2nd and 3rd panels) were used, and the results indicate a probable higher level of expression of the predominant mN-FIC2 transcript in the 2-day involuting mammary gland than during lactation. The low level of mNFIC5 transcripts hindered accurate evaluation of expression levels. In contrast, NFIX1 and -X2 are expressed at equal levels during lactation and early involution (Fig. 7A, 1st panel). GAPDH transcript levels were equivalent in both samples (Fig. 7A, 4th panel). To obtain a more satisfactory quantitative estimation of the transcript level of NFIC2 during lactation, involution, and resuckling, competitive PCR was carried out using specific primers complementary to a 231-bp sequence encompassing parts of exons 8 -10 in mNFIC2. A competitor cDNA (competicon) (Fig. 7B,  upper panel) was generated that had a 20-bp deletion in this sequence and would therefore yield a PCR product of 211 bp (see "Experimental Procedures"). Increasing amounts of competicon was titrated into individual aliquots of sample first strand cDNA (prepared from lactating, 2-day involuting and 36-h resuckled mammary gland), and these were subjected to PCR. When equal amounts of PCR product were made from competicon and endogenous cDNA target, it was assumed that the initial copy number of the target in the sample matched the known value of the added competicon (here assigned arbitrary units). Repeated independent analyses showed higher levels of mNFIC2 cDNA in involution than in lactation, and a partial reversal toward lactation levels was observed in resuckled mammary gland (Fig. 7B, upper panels). A parallel analysis of GAPDH levels showed them to be equal in the three mammary gland samples (lower panels). Overall our analysis points to an increase in transcript levels, with the most abundant mNFIC transcript, mNFIC2, being triggered in involution in parallel with the involution-triggered generation of the post-translationally modified NFIC. DISCUSSION The transition from lactation to the early stages of involution involves a shutting down of the transcription of genes expressed in the terminally differentiated mammary epithelial cell, particularly that of the milk protein genes, and in contrast, an induction of expression of a new population of genes that accompany the apoptotic response, which are required for the survival of the residual epithelial cell population and for the epithelial cell contribution to the reductive remodeling of the gland that follows the induction of involution (36,53). The hypothesis that the new pattern of gene expression that accom-panies the activation/generation of a cohort of transcription factors that modulate this gene expression has been borne out by the necessary association of Stat 3 activation (60, 61), AP-1 generation (37), and NFB nuclear translocation (63)(64)(65) with the onset of involution. In addition, we have reported earlier an induction in the expression of a unique isoform of the transcription factor NFI family, the 74-kDa NFI, at this time. In this study the characterization of two NFIC-specific antibodies ( Fig.  2A) led to the identification of the involution-associated 74-kDa NFI as an NFIC and to the description of a transition in the production of NFIC molecules from a 49-kDa isoform in lactation to the 74-kDa isoform which begins to be produced some 12 h after involution initiates.
The apparent molecular weight of the involution-associated 74-kDa NFIC is greater than that of a peptide encoded by the predicted open reading frame of the largest member of the NFIC family. This suggested that it was post-translationally modified. Previous reports demonstrated O-glycosylation (30) and phosphorylation (18,29) of NFI peptides. However, sequence analysis showed the presence of potential N-glycosylation sites in mNFIC2, the most abundant of NFICs in mouse mammary gland, and whose transcription was selectively induced in early involution (Fig. 7). Supporting this we could demonstrate that the 74-kDa NFIC binds to ConA and that this binding is competitively reversed by specific carbohydrate ligands (Fig. 3B) and that its apparent molecular size can be significantly reduced by treatment with a specific N-glycosidase (Fig. 3C). However, we note that the 74-kDa NFIC is only reduced to an apparent molecular mass of 63-kDa by N-glycosidase treatment. This suggests that it retains residual modification by phosphorylation and/or O-glycosylation that is not present on the lactation-associated 49-kDa NFIC from which one assumes it arises. Furthermore, ConA chromatography purified from involuting mammary gland nuclear extracts the activity that generated the unique "bandshifted" NFI DNAbound complex previously shown (40) to contain the 74-kDa NFI (Fig. 3E). In addition, we could show that tunicamycin, which inhibits the initial step in N-glycosylation (the biosynthesis of the donor carbohydrate moiety in the Golgi apparatus), inhibited the formation of the 74-kDa NFI, in vivo, in the early involuting mammary gland (Fig. 4). Finally, we could reiterate the production of the 74-kDa NFIC in mammary epithelial cells cultured on a laminin-rich extracellular matrix and, under defined conditions, inhibit its generation by tunicamycin (Fig. 5). Taken together these data strongly suggest that involution triggers the production of an NFIC isoform that is, unexpectedly, post-translationally modified by N-glycosylation.
N-Glycosylation is a common fate of secreted proteins or of the extracellular domains of integral plasma membrane proteins. These are modified as they pass through the endoplasmic reticulum and Golgi apparatus to reach their final destinations. The sorting of noncytoplasmic proteins begins at the membrane of the ER. Secreted and integral membrane proteins are transported across the ER membrane at sites called translocons (reviewed in Ref. 66). Translocons are composed of several ER membrane proteins that associate to form a pore through which peptides pass from the cytoplasm to the ER lumen for subsequent post-transcriptional modification. Nuclear targeting would require a "retrotransport" from the ER/ Golgi compartment, for instance to the cytoplasm and thence to the nucleus. The NFIs contain strong, proven nuclear translocation signals, and we have reported previously (40) that the 74-kDa NFIC clearly partitions to the nucleus (relative to cytoplasm) in 2-day involuting mouse mammary gland. It has been suggested that the same translocon pore is used to move proteins in the opposite direction, from the ER lumen to the FIG. 7. NFIC expression in the mouse mammary gland. A, shown are the ethidium bromide-stained agarose gels from RT-PCR analyses of mNFIX1, -X2, and -X3 (top), mNFIC2 and C5 (middle), and GAPDH (bottom) transcripts in quantitatively standardized total RNA samples from lactating (L), 2-day involuting (2di), and 3-day involuting (3di) mouse mammary gland. The ϪRT GAPDH control analysis is shown. The RT synthesis used random primers. NFIX transcripts were generated using primers X-Fwd 2/Rev 3. NFIC transcripts were generated with primers C-Fwd 8(exon 5)/Rev 8(3Ј-UTR) and C-Fwd 9(exon 8)/Rev 9(3Ј-UTR). Product sizes are indicated. B, shown are the ethidium bromide-stained agarose gels from competitive RT-PCR analyses of quantitatively standardized total RNA samples from lactating, 2-day involuting, 3-day involuting, and 36-h resuckled mouse mammary gland for NFIC2 and GAPDH. The point of equality of analyte and competicon product indicates relative expression level and is indicated by a star. cytoplasm in a process termed retrotranslocation (66,67). The functional stages in retrotranslocation are as follows: identification of the substrate and targeting to the translocon, opening of the translocon pore, and some mechanism for powering the transport of the protein through the pore (66). A general signal for retrotranslocation, suggested by Johnson and Haigh (67), is the prolonged exposure of a polypeptide sequence, surface, or glycosylation state that would elicit chaperone binding. Cytoplasmic proteins and ATP hydrolysis are also required (68,69). The retrotranslocated protein is probably pulled into the cytoplasm by a protein located in the cytoplasm (67). This mechanism of retrotranslocation has been described for proteins that are misfolded and targeted for the proteosome. A similar retrotransport must be proposed for the newly synthesized Nglycosylated 74-kDa NFIC. Its strong nuclear translocation signals (Fig. 3A) (62) direct the transport of this protein to the nucleus once it has been relocated to the cytoplasm. Equally intriguing is how and why this transcription factor becomes a target for such post-translational modification in early involution. We have failed to find evidence that a unique NFIC transcript is generated once involution is triggered rather we can detect increased expression of the most dominant lactationassociated transcript, mNFIC2 (Fig. 7). Thus, the most likely explanation is an involution-associated induction of a chaperone-mediated event that places the 74-kDa apo-NFIC in an environment where it can be post-translationally modified by N-glycosylation and a retrotransport of the product to the cytoplasm.
Why would N-glycosylation of this transcription factor occur? Most probably the modification induces a functional change at a recognition level and at an activity level. Our data suggest that this modification does not affect the ability of this NFIC isoform to translocate to the nucleus or to recognize or bind to NFI DNA-binding elements. The N-glycosylation target sites are in the C-terminal regions of NFIC that houses its transcriptional activator/repressor domains. It is the easiest to suggest that they will be most affected by this modification. Thus, their potential to recruit co-activators or co-repressors or other nuclear regulatory targets may be the outcome of this event. Interestingly, Kannius-Janson et al. (51) have very recently demonstrated that NFIC2 is a significant transcriptional activator in mammary epithelial cells and is specifically required for activation of a milk protein gene promoter, and in contrast, it has been shown to display weak transcriptional activity in other cellular environments.
However, because the mammary gland is so heavily geared toward secretion and the associated glycosylation, it is possible that the N-glycosylation of NFIC seen may be due to the massive restructuring of the secretory apparatus that occurs during involution. This might result in an inappropriate reorganization of the glycosylation apparatus or its substrates. Thus, a number of proteins that are not normally glycosylated, but which contain glycosylation sites, may become inappropriately glycosylated.
In summary, we have characterized the involution-associated 74-kDa transcription factor NFI as a CTF/NFIC that is posttranslationally modified by N-glycosylation. Its production in early involution is paralleled by increased expression of the predominant NFIC2 transcript. It is now of particular interest to establish the physiological role played by this protein.