OGT (O-GlcNAc Transferase) Selectively Modifies Multiple Residues Unique to Lamin A

The LMNA gene encodes lamins A and C with key roles in nuclear structure, signaling, gene regulation, and genome integrity. Mutations in LMNA cause over 12 diseases (‘laminopathies’). Lamins A and C are identical for their first 566 residues. However, they form separate filaments in vivo, with apparently distinct roles. We report that lamin A is β-O-linked N-acetylglucosamine-(O-GlcNAc)-modified in human hepatoma (Huh7) cells and in mouse liver. In vitro assays with purified O-GlcNAc transferase (OGT) enzyme showed robust O-GlcNAcylation of recombinant mature lamin A tails (residues 385–646), with no detectable modification of lamin B1, lamin C, or ‘progerin’ (Δ50) tails. Using mass spectrometry, we identified 11 O-GlcNAc sites in a ‘sweet spot’ unique to lamin A, with up to seven sugars per peptide. Most sites were unpredicted by current algorithms. Double-mutant (S612A/T643A) lamin A tails were still robustly O-GlcNAc-modified at seven sites. By contrast, O-GlcNAcylation was undetectable on tails bearing deletion Δ50, which causes Hutchinson–Gilford progeria syndrome, and greatly reduced by deletion Δ35. We conclude that residues deleted in progeria are required for substrate recognition and/or modification by OGT in vitro. Interestingly, deletion Δ35, which does not remove the majority of identified O-GlcNAc sites, does remove potential OGT-association motifs (lamin A residues 622–625 and 639–645) homologous to that in mouse Tet1. These biochemical results are significant because they identify a novel molecular pathway that may profoundly influence lamin A function. The hypothesis that lamin A is selectively regulated by OGT warrants future testing in vivo, along with two predictions: genetic variants may contribute to disease by perturbing OGT-dependent regulation, and nutrient or other stresses might cause OGT to misregulate wildtype lamin A.


Mice
Male mice (mixed genetic background based on C57BL/6J and 129/Sv strains, backcrossed with C57BL/6J for at least six generations) were housed in a temperature-controlled (22 • C) facility with a strict 12 h light/dark cycle. Mice had free access to food and water at all times. STZ (S0130; Sigma-Aldrich) was prepared in a sodium citrate buffer (50 mmol/L, pH 4.5) immediately before injections. STZ-treated mice were treated with two intraperitoneal injections of STZ (100 mg/kg) with a 1-day interval. Seven days after the first STZ injection, mice were included in a fasting-re-feeding experiment. Mice were either fasted for 24 h, or fasted for 24 h and re-fed for 12 h on a diet containing 64% carbohydrates, 31.5% protein, 4.5% fat, and no cholesterol (SDS RM no.1 maintenance, Special Diets Services, Witham, UK). The mice were euthanized by cervical dislocation at 8:00 a.m. and liver tissues were snap-frozen in liquid nitrogen and stored at −80 • C until further analysis. All use of animals was approved and registered by the Norwegian Animal Research authority.
Data from MS/MS analyses were searched against human lamin A using the Open Mass Spectrometry Search Algorithm (OMSSA; [35]). OMSSA search tolerances were ±0.01 Da and ±0.35 Da for precursor and product ion masses, respectively. For ETD spectra, search parameters were set to exclude reduced charge species from peak lists prior to searching. Database searches were completed using either AspN or no enzyme specifications and allowing up to three missed cleavages. Specified variable modifications: carbamidomethylation of Cys, oxidation of Met, and O-GlcNAcylation of Ser and Thr. While OMSSA searches were used as a guide, all spectra and O-GlcNAc site modifications were validated by manual interpretation of the raw data.

Native Lamin A is O-GlcNAcylated in Human Hepatoma Cells and Mouse Liver
To assess potential O-GlcNAcylation of A-type lamins, we cultured human hepatoma (Huh7) cells for 24 h in medium containing physiological (5 mM) or high (25 mM) glucose. A-type lamins were immunoprecipitated from sonicated Huh7 nuclear lysates using monoclonal antibody 5G4, which recognizes the Ig-fold domain (residues 466-484; Supplemental Figure 1A) in both lamin A and lamin C. Immunoprecipitates (50% per lane) were resolved by SDS-PAGE with input lysate controls (8%), and immunoblotted ( Figure 1A). One set of samples was probed using antibody CTD 110.6 , which recognizes the O-GlcNAc modification, and then stripped and re-probed for lamins A/C using antibody 5G4 ( Figure 1A). The~73 kD lamin A band was consistently O-GlcNAc-positive in both concentrations of glucose (asterisk; Figure 1A; n = 5), demonstrating O-GlcNAc modification of lamin A in cultured human hepatoma cells. However, the level of lamin A O-GlcNAcylation in 5-vs. 25-mM glucose, and the extent to which other O-GlcNAcylated proteins co-immunoprecipitated with lamin A in 25 mM glucose, varied between experiments and was not studied further. O-GlcNAc signals were not detected on lamin C (~60 kD; Figure 1A). These results demonstrated O-GlcNAcylation of endogenous lamin A in proliferating hepatoma cells, consistent with previous results from mitotic HeLa cells [19].
To determine whether lamin A was O-GlcNAcylated in a non-proliferating tissue, we analyzed liver lysates from wildtype mice that were either fasted for 24 h, or fasted for 24 h and then re-fed for 12 h on a regular chow diet to boost overall protein O-GlcNAcylation (see Methods). O-GlcNAc-modified proteins from each nuclear lysate (two livers per condition) were selectively affinity-purified using succinylated Wheat Germ Agglutinin (sWGA; does not bind sialylated proteins). The resulting O-GlcNAc-modified proteins were then resolved by SDS-PAGE and western blotted for O-GlcNAc (antibody CTD 110.6 ) and lamin A (antibody L1293; Figure 1B). Control input lysates were analyzed in parallel. Lysates contained similar levels of input mature lamin A (~73 kD; Figure 1B, Input). As expected, re-feeding caused an overall increase in O-GlcNAcylation of nuclear proteins (CTD 110.6 Ab; Figure 1B). Probing for specific proteins in the affinity-purified samples revealed enrichment for two lamin A-specific bands: the expected~73 kD mature lamin A band and a slower-migrating (~74 kD) band (sWGA; Figure 1B). Pull-downs from mouse liver nuclear lysates, using either WGA (QF & LMGW, unpublished observations) or sWGA ( Figure 1B), all revealed two lamin A bands. The single lamin A band detected in inputs ( Figure 1B) corresponded to the faster-migrating band in sWGA ( Figure 1B). We speculate that the second band represents differentially-modified forms of lamin A that were enriched by sWGA pull-down. Supporting this possibility, two lamin A bands were also detected in mouse liver after enrichment via immunoprecipitation with lamin A/C antibodies (TD & KLW, unpublished observations). , and then stripped and re-probed with antibody 5G4, specific for lamins A and C (Lamin A/C). Control input lysates, and proteins immunoprecipitated using lamin A/C antibody 5G4 or nonspecific IgG as the negative control, were resolved in parallel gels. Asterisk indicates the O-GlcNAc signal corresponding to immunoprecipitated lamin A; (B) Western blots of mouse liver nuclear proteins, before (Input) or after affinity purification on succinylated Wheat Germ Agglutinin (sWGA), which binds O-GlcNAc-modified proteins. We analyzed two livers for each condition. Male mice were either fasted (Fasted), or fasted and then refed for 12 h ('Refed'), and were either pretreated with streptozotocin to induce hyperglycemia ('STZ'), or not ('Vehicle'). Input controls were probed with lamin A-specific antibody L1293. Affinity-purified samples were probed first for the O-GlcNAc modification (antibody CTD110.6), and then stripped and re-probed for lamin A (antibody L1293). Control input lysates, and proteins immunoprecipitated using lamin A/C antibody 5G4 or nonspecific IgG as the negative control, were resolved in parallel gels. STZ-streptozocin.
We found similar sWGA-purification results in liver lysates from mice treated with streptozocin (STZ), which kills pancreatic β-cells and models diabetes [36]: two lamin A bands, one or both of which were presumably O-GlcNAc-modified (STZ; Figure 1B). The relative abundance of the two lamin A bands could not be interpreted, because the core epitope recognized by antibody L1293 (residues 604-611; Supplemental Figure 1) adjoins a potential O-GlcNAc site (S612; [19]) and potential phosphorylation sites [37,38], and may be differentially accessible on native lamin A. Collectively, these results from hepatoma cells and mouse livers support the hypothesis that lamin A is O-GlcNAcmodified, and hence O-GlcNAc-regulated, in the liver. To explore whether OGT targets lamin A , and then stripped and re-probed with antibody 5G4, specific for lamins A and C (Lamin A/C). Control input lysates, and proteins immunoprecipitated using lamin A/C antibody 5G4 or nonspecific IgG as the negative control, were resolved in parallel gels. Asterisk indicates the O-GlcNAc signal corresponding to immunoprecipitated lamin A; (B) Western blots of mouse liver nuclear proteins, before (Input) or after affinity purification on succinylated Wheat Germ Agglutinin (sWGA), which binds O-GlcNAc-modified proteins. We analyzed two livers for each condition. Male mice were either fasted (Fasted), or fasted and then re-fed for 12 h ('Refed'), and were either pretreated with streptozotocin to induce hyperglycemia ('STZ'), or not ('Vehicle'). Input controls were probed with lamin A-specific antibody L1293. Affinity-purified samples were probed first for the O-GlcNAc modification (antibody CTD 110.6 ), and then stripped and re-probed for lamin A (antibody L1293). Control input lysates, and proteins immunoprecipitated using lamin A/C antibody 5G4 or nonspecific IgG as the negative control, were resolved in parallel gels. STZ-streptozocin.
We found similar sWGA-purification results in liver lysates from mice treated with streptozocin (STZ), which kills pancreatic β-cells and models diabetes [36]: two lamin A bands, one or both of which were presumably O-GlcNAc-modified (STZ; Figure 1B). The relative abundance of the two lamin A bands could not be interpreted, because the core epitope recognized by antibody L1293 (residues 604-611; Supplemental Figure 1) adjoins a potential O-GlcNAc site (S612; [19]) and potential phosphorylation sites [37,38], and may be differentially accessible on native lamin A. Collectively, these results from hepatoma cells and mouse livers support the hypothesis that lamin A is O-GlcNAc-modified, and hence O-GlcNAc-regulated, in the liver. To explore whether OGT targets lamin A selectively, we compared the tail domains of A-and B-type lamins as biochemical substrates for OGT in vitro.

In Vitro O-GlcNAcylation of Recombinant Lamin Tails
To identify O-GlcNAc sites, we incubated recombinant lamin tail domains with purified OGT in vitro. We tested nine different tail constructs: lamin B1, lamin C, mature lamin A residues 385-646 (wildtype, S612A, T643A, or S612A/T643A), and prelamin A residues 385-664 bearing the progeria-associated ∆35, ∆50, or ∆90 deletions (Figure 2A). Each purified recombinant T7-tagged lamin tail polypeptide (1 µg protein) was incubated with UDP-GlcNAc and calf intestinal phosphatase (CIP), plus or minus purified recombinant His-tagged OGT enzyme, for 2 h at 22-24 • C, and then overnight at 4 • C. The reactions were quenched with SDS sample buffer, resolved by SDS-PAGE and immunoblotted first with antibody CTD 110.6 , specific for the O-GlcNAc moiety, then stripped and re-probed with antibodies against the T7-tag on lamins. These in vitro O-GlcNAcylation assays revealed robust and specific modification of the recombinant mature lamin A tail domain ( Figure 2B). O-GlcNAcylation of mature lamin A tails mutated at one or both previously reported O-GlcNAc sites (S612A, T643A, S612A/T643A) was similarly robust ( Figure 2B), suggesting OGT targeted alternative O-GlcNAc sites. The O-GlcNAc signals were specific: potential antibody cross-recognition of recombinant proteins was ruled out by probing duplicate membranes with the O-GlcNAc antibody plus 100 mM free sugar (+GlcNAc; Figure 2C). In vitro O-GlcNAcylation was greatly reduced by deletion ∆35 (loss of residues 622-658) [16]; trace signals were detected only in long exposures ( Figure 2B; 'long exp'). There was no detectable O-GlcNAcylation of the HGPS-associated deletion ∆50 (∆608-658; [15]) or deletion ∆90 (Restrictive Dermopathy-associated loss of residues 567-658; [39]) in vitro ( Figure 2B,C; n = 3). There were also no specific O-GlcNAc signals on lamin C or lamin B1 tails ( Figure 2B), or on full-length C. elegans lamin ('ce-lamin'; Figure 2D), which has a short 'B-like' tail. Potential O-GlcNAc modification of the head or rod domain of human lamins was not tested. These results did not rule out potential O-GlcNAc modifications of lamin C or B-type tails since negative results can be artifacts of recombinant proteins. However, selective O-GlcNAcylation of lamin A tails in vitro was consistent with our results in hepatoma cells ( Figure 1A). These results suggested OGT targets the tail domain of lamin A selectively in vitro and is sensitive to progeria-associated deletions.
Cells 2018, 7, x 7 of 21 selectively, we compared the tail domains of A-and B-type lamins as biochemical substrates for OGT in vitro.

In Vitro O-GlcNAcylation of Recombinant Lamin Tails
To identify O-GlcNAc sites, we incubated recombinant lamin tail domains with purified OGT in vitro. We tested nine different tail constructs: lamin B1, lamin C, mature lamin A residues 385-646 (wildtype, S612A, T643A, or S612A/T643A), and prelamin A residues 385-664 bearing the progeriaassociated Δ35, Δ50, or Δ90 deletions ( Figure 2A). Each purified recombinant T7-tagged lamin tail polypeptide (1 μg protein) was incubated with UDP-GlcNAc and calf intestinal phosphatase (CIP), plus or minus purified recombinant His-tagged OGT enzyme, for 2 h at 22-24 °C, and then overnight at 4 °C. The reactions were quenched with SDS sample buffer, resolved by SDS-PAGE and immunoblotted first with antibody CTD110.6, specific for the O-GlcNAc moiety, then stripped and reprobed with antibodies against the T7-tag on lamins. These in vitro O-GlcNAcylation assays revealed robust and specific modification of the recombinant mature lamin A tail domain ( Figure 2B). O-GlcNAcylation of mature lamin A tails mutated at one or both previously reported O-GlcNAc sites (S612A, T643A, S612A/T643A) was similarly robust ( Figure 2B), suggesting OGT targeted alternative O-GlcNAc sites. The O-GlcNAc signals were specific: potential antibody cross-recognition of recombinant proteins was ruled out by probing duplicate membranes with the O-GlcNAc antibody plus 100 mM free sugar (+GlcNAc; Figure 2C). In vitro O-GlcNAcylation was greatly reduced by deletion Δ35 (loss of residues 622-658) [16]; trace signals were detected only in long exposures ( Figure  2B; 'long exp'). There was no detectable O-GlcNAcylation of the HGPS-associated deletion Δ50 (Δ608-658; [15]) or deletion Δ90 (Restrictive Dermopathy-associated loss of residues 567-658; [39]) in vitro ( Figure 2B,C; n = 3). There were also no specific O-GlcNAc signals on lamin C or lamin B1 tails ( Figure  2B), or on full-length C. elegans lamin ('ce-lamin'; Figure 2D), which has a short 'B-like' tail. Potential O-GlcNAc modification of the head or rod domain of human lamins was not tested. These results did not rule out potential O-GlcNAc modifications of lamin C or B-type tails since negative results can be artifacts of recombinant proteins. However, selective O-GlcNAcylation of lamin A tails in vitro was consistent with our results in hepatoma cells ( Figure 1A). These results suggested OGT targets the tail domain of lamin A selectively in vitro and is sensitive to progeria-associated deletions.

O-GlcNAc Site Identification by Mass Spectrometry
O-GlcNAc sites were identified by mass spectrometry analysis of three in vitro-modified samples: wildtype mature lamin A tail, S612A/T643A-mutated mature lamin A tail, and pre-lamin A bearing the Δ35 HGPS-associated deletion. Samples were digested with either AspN alone or AspN followed by chymotrypsin. The generated peptides were analyzed by on-line HPLC-MS/MS on a high-resolution mass spectrometer (see Methods). Peptides were fragmented by both collisionally activated dissociation (CAD) and electron transfer dissociation (ETD) in a data-dependent mode. We achieved nearly 90% peptide coverage of both the wildtype and S612A/T643A-mutated lamin A tails. Coverage of Δ35 was lower (71%), but included 48 of 57 total Ser/Thr residues in the Δ35 tail ( Figure  3A). The O-GlcNAcylated peptides were identified from CAD spectra by either the O-GlcNAc oxonium signature peak at m/z 203 amu, or they exhibited the charge reduced product ion with the loss of 203 amu, as shown in Figure 4A. ETD spectra were used to identify sites modified by O-GlcNAc. We detected a total of 11 unique O-GlcNAc sites in the wildtype lamin A tail, four of which were more abundant, based qualitatively on the relative abundance of the ions corresponding to the O-GlcNAc sites: S612, S613, S618, and T643 ('more abundant'; Figure 3B). The seven less abundant sites were S603, S619, T623, and four others (S615, S616, T621, and S628), and were detected only on heavily O-GlcNAc-modified peptides ('less abundant'; Figure 3B). These results validated previously reported sites S612 and T643 [19] and identified nine more O-GlcNAc sites in the mature lamin A tail.
Lamin A tail peptides were frequently modified at multiple sites (up to seven; Figure 3B,C). Mass spectrometry analysis of the AspN peptide containing residues 596-638 from wildtype lamin A revealed that 37.3% were unmodified, 9% had one O-GlcNAc, 35.8% were di-GlcNAcylated, 14% were tri-GlcNAcylated, and the remaining 3.9% had four to seven O-GlcNAc modifications ( Figure  3B,C). Percentages are based on the chromatographic peak area of each species compared to the chromatographic peak area of all species combined. Only two di-GlcNAcylated pairs were detected: S612 + S613 and S612 + S603. The ETD spectrum of the S612 + S613 di-GlcNAcylated pair, with near complete sequence coverage, is shown in Figure 4B. The most frequent triple combinations were S612

O-GlcNAc Site Identification by Mass Spectrometry
O-GlcNAc sites were identified by mass spectrometry analysis of three in vitro-modified samples: wildtype mature lamin A tail, S612A/T643A-mutated mature lamin A tail, and pre-lamin A bearing the ∆35 HGPS-associated deletion. Samples were digested with either AspN alone or AspN followed by chymotrypsin. The generated peptides were analyzed by on-line HPLC-MS/MS on a high-resolution mass spectrometer (see Methods). Peptides were fragmented by both collisionally activated dissociation (CAD) and electron transfer dissociation (ETD) in a data-dependent mode. We achieved nearly 90% peptide coverage of both the wildtype and S612A/T643A-mutated lamin A tails. Coverage of ∆35 was lower (71%), but included 48 of 57 total Ser/Thr residues in the ∆35 tail ( Figure 3A). The O-GlcNAcylated peptides were identified from CAD spectra by either the O-GlcNAc oxonium signature peak at m/z 203 amu, or they exhibited the charge reduced product ion with the loss of 203 amu, as shown in Figure 4A. ETD spectra were used to identify sites modified by O-GlcNAc. We detected a total of 11 unique O-GlcNAc sites in the wildtype lamin A tail, four of which were more abundant, based qualitatively on the relative abundance of the ions corresponding to the O-GlcNAc sites: S612, S613, S618, and T643 ('more abundant'; Figure 3B). The seven less abundant sites were S603, S619, T623, and four others (S615, S616, T621, and S628), and were detected only on heavily O-GlcNAc-modified peptides ('less abundant'; Figure 3B). These results validated previously reported sites S612 and T643 [19] and identified nine more O-GlcNAc sites in the mature lamin A tail.
Lamin A tail peptides were frequently modified at multiple sites (up to seven; Figure 3B,C). Mass spectrometry analysis of the AspN peptide containing residues 596-638 from wildtype lamin A revealed that 37.3% were unmodified, 9% had one O-GlcNAc, 35.8% were di-GlcNAcylated, 14% were tri-GlcNAcylated, and the remaining 3.9% had four to seven O-GlcNAc modifications ( Figure 3B,C). Percentages are based on the chromatographic peak area of each species compared to the chromatographic peak area of all species combined. Only two di-GlcNAcylated pairs were detected: S612 + S613 and S612 + S603. The ETD spectrum of the S612 + S613 di-GlcNAcylated pair, with near complete sequence coverage, is shown in Figure 4B. The most frequent triple combinations were S612 + S613 plus one of S616, T623, or S603; together these five sites accounted for all triples in the wildtype tail.
We also identified sites in the corresponding AspN peptide from the double-mutant (S612A/T643A; 'AA') tail to determine how OGT responded to substrates lacking both previously identified O-GlcNAc sites [19]. The mono-GlcNAc species (invariably at S613) predominated (50.1%) over the unmodified species (24.5%; Figure 3B). The di-and tri-GlcNAc species were also relatively abundant (18.7% and 5%), and the tetra-and penta-GlcNAc species were present at <2% each ('AA'; Figure 3B). Collectively, >25% of double-mutant tails had two or more modifications ( Figure 3B). Mass spectrometry analysis of the double mutant revealed O-GlcNAc modifications at three predominant sites (S613, S615, and S616) and four less abundant sites (S618, S619, T621, and T623; Figure 3B). We were unable to site-map the mono-O-GlcNAcylated peptide corresponding to residues 639-646 due to poor sequence coverage; however, given the T643A mutation, this peptide had only one potential site, S645, suggesting O-GlcNAc-modification of residue S645 in the double mutant. Modification at S645 was not detected in wildtype or ∆35 tails. We concluded that the S612A/T643A double-mutant tail remained attractive as a substrate for OGT.
Cells 2018, 7, x 9 of 21 + S613 plus one of S616, T623, or S603; together these five sites accounted for all triples in the wildtype tail.
We also identified sites in the corresponding AspN peptide from the double-mutant (S612A/T643A; 'AA') tail to determine how OGT responded to substrates lacking both previously identified O-GlcNAc sites [19]. The mono-GlcNAc species (invariably at S613) predominated (50.1%) over the unmodified species (24.5%; Figure 3B). The di-and tri-GlcNAc species were also relatively abundant (18.7% and 5%), and the tetra-and penta-GlcNAc species were present at <2% each ('AA'; Figure 3B). Collectively, >25% of double-mutant tails had two or more modifications ( Figure 3B). Mass spectrometry analysis of the double mutant revealed O-GlcNAc modifications at three predominant sites (S613, S615, and S616) and four less abundant sites (S618, S619, T621, and T623; Figure 3B). We were unable to site-map the mono-O-GlcNAcylated peptide corresponding to residues 639-646 due to poor sequence coverage; however, given the T643A mutation, this peptide had only one potential site, S645, suggesting O-GlcNAc-modification of residue S645 in the double mutant. Modification at S645 was not detected in wildtype or Δ35 tails. We concluded that the S612A/T643A double-mutant tail remained attractive as a substrate for OGT.   With the S612A/T643A-mutated substrate, OGT apparently compensated by targeting three predominant sites adjoining residue 612 (S613, S615, and S616) plus nearby residues S618, S619, T621, and T623. However, there were fewer 'multiples': the peptide comprising residues 596-638 was more likely to have one modification (50.1% versus 9% of wildtype), and less likely to have two (18.7% vs. 35.8% of wildtype), three (5% vs. 14% of wildtype), or four or more modifications (1.5% vs. 3.9% of wildtype; Figure 3B). In the peptide comprising residues 596-626, we detected di-GlcNAcylation at residues S613 + S616 and S613 + S615, as well as tri-GlcNAcylation at S613 + S616 + S618 and S613 + S615 + T621. In a longer version of this peptide (residues 596-638), we detected di-GlcNAcylation mainly on S613 + S615 and less often on S613 + (S616, S618, or S619). Figure 5A shows the CAD spectrum of the di-GlcNAcylated species, with the most abundant peaks being the charge-reduced product ion with the loss of 203 and 406, one and two O-GlcNAc moieties, respectively. Figure 5B shows the ETD spectrum of the S613 and S618 di-GlcNacylated peptide with almost full sequence coverage. Tri-GlcNAcylation occurred on S613 plus different combinations of S615, S616, T621, With the S612A/T643A-mutated substrate, OGT apparently compensated by targeting three predominant sites adjoining residue 612 (S613, S615, and S616) plus nearby residues S618, S619, T621, and T623. However, there were fewer 'multiples': the peptide comprising residues 596-638 was more likely to have one modification (50.1% versus 9% of wildtype), and less likely to have two (18.7% vs. 35.8% of wildtype), three (5% vs. 14% of wildtype), or four or more modifications (1.5% vs. 3.9% of wildtype; Figure 3B). In the peptide comprising residues 596-626, we detected di-GlcNAcylation at residues S613 + S616 and S613 + S615, as well as tri-GlcNAcylation at S613 + S616 + S618 and S613 + S615 + T621. In a longer version of this peptide (residues 596-638), we detected di-GlcNAcylation mainly on S613 + S615 and less often on S613 + (S616, S618, or S619). Figure 5A shows the CAD spectrum of the di-GlcNAcylated species, with the most abundant peaks being the charge-reduced product ion with the loss of 203 and 406, one and two O-GlcNAc moieties, respectively. Figure 5B shows the ETD spectrum of the S613 and S618 di-GlcNacylated peptide with almost full sequence coverage. Tri-GlcNAcylation occurred on S613 plus different combinations of S615, S616, T621, and/or T623 as the second and third sites. We speculate that the first O-GlcNAc modification (e.g., S612 in wildtype; S613 in the AA mutant) strongly favors further modification. and/or T623 as the second and third sites. We speculate that the first O-GlcNAc modification (e.g., S612 in wildtype; S613 in the AA mutant) strongly favors further modification. The Δ35 tail, with nine hypothetical sites (see Figure 6A), was modified surprisingly inefficiently: 97.7% of peptides were unmodified, with only single modifications detected at either S612 (<3% of peptide 596-664) or less frequently at S601 or S618 ( Figure 3A,B). Peptides that included residue S613 were recovered, but unmodified, supporting the hypothesis that Δ35 greatly impairs substrate recognition or modification by OGT in vitro.
These results identify residues 601-645 as a broadly defined O-GlcNAc 'sweet spot' (Figure 6A), since OGT targeted this region robustly even when two abundant sites (S612 and T643) were missing ('AA' mutant; Figures 2B and 3B). This sweet spot overlaps the region required for ZMPSTE24dependent cleavage of prelamin A, raising the possibility that OGT might influence prelamin A maturation. However, the 'sweet spot' is also a permanent feature of mature lamin A filaments (Figure 7), suggesting OGT regulates roles unique to lamin A. The ∆35 tail, with nine hypothetical sites (see Figure 6A), was modified surprisingly inefficiently: 97.7% of peptides were unmodified, with only single modifications detected at either S612 (<3% of peptide 596-664) or less frequently at S601 or S618 ( Figure 3A,B). Peptides that included residue S613 were recovered, but unmodified, supporting the hypothesis that ∆35 greatly impairs substrate recognition or modification by OGT in vitro.
These results identify residues 601-645 as a broadly defined O-GlcNAc 'sweet spot' (Figure 6A), since OGT targeted this region robustly even when two abundant sites (S612 and T643) were missing ('AA' mutant; Figures 2B and 3B). This sweet spot overlaps the region required for ZMPSTE24-dependent cleavage of prelamin A, raising the possibility that OGT might influence prelamin A maturation. However, the 'sweet spot' is also a permanent feature of mature lamin A filaments (Figure 7), suggesting OGT regulates roles unique to lamin A. Cells 2018, 7, x 12 of 21 The starting points of fragments that were cleaved by ZMPSTE24 with 100%, ~50%, or 0% efficiency in cells [41] are indicated below the sequence. Mutation G608S is sufficient to block cleavage in a different assay [42]. The region targeted by OGT in both wildtype and mutated lamin A tails (residues 601-645) is broadly designated the 'sweet spot', and includes 'supersweet' residues 612-623 (O-GlcNAc detected at every Ser and Thr); (B) amino acid sequence of the OGT-binding region in mouse Tet1 compared to homologous regions of conspecific Tet2 and Tet3 [43], two regions in human lamin A, and human HCF-1 and Sin3A. Bold residues are important for mTet1 binding to OGT [43]. Box indicates best conservation between mTet proteins and human lamin A residues 639-645. Human protein sequences were aligned against the consensus sequence found in mouse Tet1 using EMBOSS water pairwise alignment tool v.6.6.0 under default conditions for short sequence alignments (gap penalty: 10.0, extend penalty: 0.5). The upstream VTXS motif in lamin A (residues 609-635) was aligned manually against HCF-1 and Sin3A.

Discussion
Lamin glycosylation was first suggested nearly 30 years ago [44]. The nature of this modification was unknown until O-GlcNAc sites were identified in mitotic HeLa cells (S612 and T643; [19]) and mouse brains (sites S611 and S613, homologous to human S612; [45]). Our evidence that endogenous lamin A is O-GlcNAcylated in human hepatoma cells and mouse liver demonstrates biological OGT targeting of lamin A in liver. However, further studies are needed to identify O-GlcNAc sites in liver O-GlcNAc sites are located in the C-terminal region unique to prelamin A, expanded below (dotted lines) to show the amino acid sequence. Abundant or less-abundant O-GlcNAc sites are symbolized by large ('major') or small ('minor') black rods, respectively. The ZMPSTE24-dependent proteolytic cleavage site is indicated. The starting points of fragments that were cleaved by ZMPSTE24 with 100%, 50%, or 0% efficiency in cells [41] are indicated below the sequence. Mutation G608S is sufficient to block cleavage in a different assay [42]. The region targeted by OGT in both wildtype and mutated lamin A tails (residues 601-645) is broadly designated the 'sweet spot', and includes 'supersweet' residues 612-623 (O-GlcNAc detected at every Ser and Thr); (B) amino acid sequence of the OGT-binding region in mouse Tet1 compared to homologous regions of conspecific Tet2 and Tet3 [43], two regions in human lamin A, and human HCF-1 and Sin3A. Bold residues are important for mTet1 binding to OGT [43]. Box indicates best conservation between mTet proteins and human lamin A residues 639-645. Human protein sequences were aligned against the consensus sequence found in mouse Tet1 using EMBOSS water pairwise alignment tool v.6.6.0 under default conditions for short sequence alignments (gap penalty: 10.0, extend penalty: 0.5). The upstream VTXS motif in lamin A (residues 609-635) was aligned manually against HCF-1 and Sin3A.

Discussion
Lamin glycosylation was first suggested nearly 30 years ago [44]. The nature of this modification was unknown until O-GlcNAc sites were identified in mitotic HeLa cells (S612 and T643; [19]) and mouse brains (sites S611 and S613, homologous to human S612; [45]). Our evidence that endogenous lamin A is O-GlcNAcylated in human hepatoma cells and mouse liver demonstrates biological OGT targeting of lamin A in liver. However, further studies are needed to identify O-GlcNAc sites in liver and explore their biological impact. Our biochemical analysis shows that OGT modifies lamin A tails selectively and robustly in vitro. Interestingly, all 13 identified primary and 'compensatory' O-GlcNAc sites are located within a 'sweet spot' (residues 601-645) unique to mature lamin A. This region includes 'super-sweet' residues 612-623, where O-GlcNAc was detected on every available Ser and Thr ( Figure 6A). Two of our most abundant in vitro O-GlcNAc sites, S612 and T643, were previously reported in mitotic HeLa cells [19] and one was also seen in mouse brain [45], strongly supporting their biological relevance. However, their functional significance in mitosis or the brain remain completely unstudied. Our results also independently confirm two previously reported in vitro sites: S623 and S628 [46]. We further identified seven novel in vitro O-GlcNAc sites in mature wildtype human lamin A (S603, S613, S615, S616, S618, S619, and T621) and two novel compensatory sites: S601 (identified in ∆35) and S645 (identified in the S612A/T643A mutant). Our results do not rule out possible O-GlcNAc sites in the neck or Ig-fold domains, because we recovered the majority of-but not all-Ser/Thr-containing peptides from these regions. However, the lack of O-GlcNAcylation of either the ∆50, ∆90, or lamin C tails strongly implies that OGT selectively targets residues 601-645 in the lamin A tail. One limitation of this study is that we did not examine the head or rod domains, or precursor residues 647-664. These regions warrant further study, because residue T91, in the coil-1B domain shared by all A-type lamins [47], is O-GlcNAc-modified in human embryonic stem cells [48]. and explore their biological impact. Our biochemical analysis shows that OGT modifies lamin A tails selectively and robustly in vitro. Interestingly, all 13 identified primary and 'compensatory' O-GlcNAc sites are located within a 'sweet spot' (residues 601-645) unique to mature lamin A. This region includes 'super-sweet' residues 612-623, where O-GlcNAc was detected on every available Ser and Thr ( Figure 6A). Two of our most abundant in vitro O-GlcNAc sites, S612 and T643, were previously reported in mitotic HeLa cells [19] and one was also seen in mouse brain [45], strongly supporting their biological relevance. However, their functional significance in mitosis or the brain remain completely unstudied. Our results also independently confirm two previously reported in vitro sites: S623 and S628 [46]. We further identified seven novel in vitro O-GlcNAc sites in mature wildtype human lamin A (S603, S613, S615, S616, S618, S619, and T621) and two novel compensatory sites: S601 (identified in Δ35) and S645 (identified in the S612A/T643A mutant). Our results do not rule out possible O-GlcNAc sites in the neck or Ig-fold domains, because we recovered the majority of-but not all-Ser/Thr-containing peptides from these regions. However, the lack of O-GlcNAcylation of either the Δ50, Δ90, or lamin C tails strongly implies that OGT selectively targets residues 601-645 in the lamin A tail. One limitation of this study is that we did not examine the head or rod domains, or precursor residues 647-664. These regions warrant further study, because residue T91, in the coil-1B domain shared by all A-type lamins [47], is O-GlcNAc-modified in human embryonic stem cells [48].

Most O-GlcNAc Sites in Lamin A Are Not Predicted by Current Algorithms
Substrate recognition by OGT is complex, involving both its TPR (protein-protein interaction) and catalytic domains, and can be influenced by post-translational modifications and partners [23,49]. There is no strict consensus site for O-GlcNAc modification [49]. For example, a study of pre-selected 13-mer peptides showed that O-GlcNAc sites are influenced by size and conformational restrictions in the −3 to +2 positions [46,50], and reported O-GlcNAc modifications at three sites (T623, S625, and S628) in lamin A peptide 618 SSVTVTRSYRSVG 630 and at one site (T399) in lamin B1 peptide 389 KLSPSPSSRVTVS 401 [46]. Even though our lamin B1 tail polypeptide included T399, O-GlcNAcylation was not detected ( Figure 2B), possibly because this fragment began at residue 394 and was likely insufficient for OGT recognition. For lamin A, we identified T623 and S628 (but not S625) as low-abundance sites, in the context of our much longer (>250-residue) tail polypeptides. Overall,

Most O-GlcNAc Sites in Lamin A Are Not Predicted by Current Algorithms
Substrate recognition by OGT is complex, involving both its TPR (protein-protein interaction) and catalytic domains, and can be influenced by post-translational modifications and partners [23,49]. There is no strict consensus site for O-GlcNAc modification [49]. For example, a study of pre-selected 13-mer peptides showed that O-GlcNAc sites are influenced by size and conformational restrictions in the −3 to +2 positions [46,50], and reported O-GlcNAc modifications at three sites (T623, S625, and S628) in lamin A peptide 618 SSVTVTRSYRSVG 630 and at one site (T399) in lamin B1 peptide 389 KLSPSPSSRVTVS 401 [46]. Even though our lamin B1 tail polypeptide included T399, O-GlcNAcylation was not detected ( Figure 2B), possibly because this fragment began at residue 394 and was likely insufficient for OGT recognition. For lamin A, we identified T623 and S628 (but not S625) as low-abundance sites, in the context of our much longer (>250-residue) tail polypeptides. Overall, very few identified O-GlcNAc sites in lamin A were predicted by the Pathak et al. (2015) [46] consensus. A shorter consensus, proposed by Liu and colleagues [51], fared slightly better: most 'sweet spot' sites match at positions −1 and +2, but only three of our 13 sites match this consensus at position −2.

Second Role for Residues Near the ZMPSTE24-Dependent Cleavage Site
We were puzzled by progeria-causing deletion ∆35, which retained nine potential O-GlcNAc sites (including seven of eight 'supersweet' sites), but remained 97.7% unmodified in vitro ( Figure 3B). Wondering if OGT might associate with this region, either during substrate recognition or independently as a hypothetical partner, we performed pairwise protein sequence alignments of lamin A to five proteins that bind OGT directly as partners ( Figure 6B): Tet1 [43], Tet2 [57], Tet3 [58], HCF-1 [59], and transcription regulator Sin3A, which recruits OGT to promoters [23,60]. Triple Ala-substitution mutagenesis of mouse Tet1, in particular, identified a motif in which four residues (DNVVTVS) were critical for binding to OGT [43]. All four critical residues are conserved in conspecific Tet2 and Tet3 [43], and in human lamin A residues 639-645 (DNLVTRS; boxed in Figure 6B), which adjoin the ZPMSTE24-dependent cleavage site ( Figure 6A). Whether these four residues are also important for cleavage is unknown. We found that a similar motif with three of four critical residues conserved (VTRS) is present in human HCF-1, Sin3A, and lamin A residues 622-625, which adjoin the 'supersweet' region ( Figure 6B). Both motifs are removed by ∆35, which deletes residues 622-656 ( Figure 6A). Whether lamin A associates functionally with OGT as a partner, in addition to being modified by OGT, is an open and challenging question for future work (Levine and Walker, 2016). Collectively, our findings provide novel insight into progeria mechanisms by predicting a second molecular consequence of the ∆50 deletion: loss of OGT-dependent regulation.

Selective OGT Targeting of Lamin A as a Mechanism for Separate Functions and Differential Regulation of Lamin A versus Lamin C
The discovery that lamin A and lamin C form separate filaments is so new and unexpected [61,62] that few studies have even attempted to distinguish one from the other. We found that OGT selectively modifies the tail domain of mature lamin A, not lamin C, in vitro, suggesting the intrinsically-disordered 'tail' of mature lamin A is uniquely controlled by OGT ( Figure 7A). These results add to growing evidence for functional specialization of lamin A versus lamin C filaments ( Figure 7B). For example, lamin C specifically associates with nuclear pore complexes [62]. Specialization is also evident in the brain and in metabolism. Lamin C predominates in neurons, where mRNAs encoding prelamin A are eliminated by a specific microRNA [63]. When compared with wildtype controls, mice that express lamin C as their only A-type lamin have significantly longer lifespans, despite accumulating white fat and developing abdominal tumors that limited their lifespan; these mice also have reduced rates of metabolism and respiration, fewer mitochondria, higher fasting insulin levels, and higher blood glucose levels [64]. Mice that express mature lamin A only have nuclear shape defects in fibroblasts, but otherwise normal body weight, health, and fertility [65], implying appropriate regulation of mature lamin A. A-type lamins are linked to the mTOR pathway, which senses glucose and other nutrients to promote biosynthetic pathways, and inhibit or limit catabolic (breakdown) pathways, including autophagy [66][67][68]. Autophagy is significantly elevated in lmna-mutated cells [69,70]. Mice that express H222P-mutated lmna in the heart have elevated AKT and mTOR signaling that leads to cardiomyopathy and impaired fasting-induced autophagy [71]. These phenotypes are ameliorated by mTOR inhibitors, suggesting an A-type lamin(s) normally dampens mTOR signaling [72][73][74]. We speculate that OGT, which responds independently to nutrient stress, specifically influences lamin A under these conditions.
What happens if OGT cannot modify the lamin A tail? Two elements are lost by deletion in 'classic' (∆50) HGPS: functions specifically mediated by the lamin A tail, and OGT-dependent control of these functions. Neither element is likely to be restored by therapies that block farnesylation. Treating HGPS patients with the farnesyltransferase inhibitor Lonafarnib improved weight gain [18] and reduced prevalence of stroke [75], but did not affect other parameters, including insulin resistance (seen in 35% of patients; [18]). We found that progeria-associated deletions either abolished (∆50) or greatly reduced (∆35) lamin A modification by OGT in vitro. Interestingly, three missense variants that perturb metabolism are located in the 'sweet spot': p.G602S (insulin resistance syndrome [8]; type 2 diabetes [76]), p.G631D (metabolic syndrome; [12]), and p.R644C (multiple phenotypes; [77]). Variant p.L92F, identified in an obese patient with severe metabolic syndrome [13], adjoins the reported O-GlcNAc site at T91 [48]. However more work is needed to decipher mechanisms, since variants that affect metabolism, like those that cause other laminopathies, are located throughout the polypeptide.

Potential Impacts of O-GlcNAc on Mature Lamin A
OGT and lamin A each have global roles that include signaling, epigenetic regulation, and mitosis [3,78,79]. For example, lamins and key partners (LEM-domain proteins and BANF1) are required to dynamically rebuild nuclear structure during exit from mitosis [80][81][82][83][84]. Similarly, OGT is required for dividing cells to survive [21,23]. Lamins have many reported modifications, the best understood of which is phosphorylation. From this perspective, we note that 10 of our 12 total identified O-GlcNAc sites can be phosphorylated in a variety of cell types or conditions [37,38] (www.phosphosite.org). To our knowledge, O-GlcNAc is the only reported modification of lamin A residues T91 [48] and T621 (this work). Our findings suggest that lamin A-like cytoplasmic intermediate filaments-is specifically and dynamically regulated by crosstalk between OGT and multiple kinases [85].
O-GlcNAc and phosphate modifications can profoundly alter the conformations of intrinsically disordered polypeptides [86][87][88], with consequences for partners. Partners that bind the unstructured region of lamin A, besides ZPMSTE24 (discussed above), include SUN1 [89], Narf, RBBP4, PKCα, and actin [37]. Certain partners identified in a yeast two-hybrid screen are disrupted by disease-associated missense variants in the 'sweet spot', and might therefore be influenced by O-GlcNAc modifications. These partners include LAP1β (disrupted by lamin A variant p.G602S), ZNF3 (disrupted by p.G602S, p.G608S, p.T623S, p.R644C), MORF4L1 (disrupted by p.G602S, p.G608S, p.T623S), Cyclin G1 (disrupted by p.G602S, p.G608S, p.R644C), and CENP-P (disrupted by p.R644C) [90]. We speculate that 'sweet-spot' modifications might also influence association with AIMP3, a cytoplasmic scaffolding protein that can also enter the nucleus and act as a tumor suppressor [91,92]. AIMP3 binds directly to residues 641-647 in the 'sweet spot' of mature lamin A [91]. AIMP3 recruits Siah1 (an E3 ubiquitin ligase) and triggers selective degradation of mature lamin A to promote senescence [91,92]. In the context of lamin A filaments, modifications by OGT and kinases might dictate a spectrum of tail conformations unique to lamin A (modeled in Figure 7B), influencing various partners either directly or via access to the Ig-fold domain ( Figure 7B), as suggested for actin [31] and Tpr [93].
This study is significant because it identifies a novel molecular pathway that may profoundly influence lamin A function. Our results suggest OGT-dependent control of lamin A is perturbed by progeria-associated deletions. As a master regulator of metabolism [94,95], OGT has the potential to misregulate wildtype lamin A in response to nutrient (including glucose) stress or other stresses. The physiological significance of lamin A tail O-GlcNAcylation, whether it influences signaling, transcription, 3D genome organization [96][97][98][99], or novel roles, and proposed crosstalk with phosphorylation, all remain to be determined in future studies.