A Single Chondroitin 6-Sulfate Oligosaccharide Unit at Ser-2730 of Human Thyroglobulin Enhances Hormone Formation and Limits Proteolytic Accessibility at the Carboxyl Terminus

We localized the site of type D (chondroitin 6-sulfate) oligosaccharide unit addition to human thyroglobulin (hTg). hTg was chromatographically separated into chondroitin 6-sulfate-containing (hTg-CS) and chondroitin 6-sulfate-devoid (hTg-CS0) molecules on the basis of their d-glucuronic acid content. In an ample number of hTg preparations, the fraction of hTg-CS in total hTg ranged from 32.0 to 71.6%. By exploiting the electrophoretic mobility shift and metachromasia conferred by chondroitin 6-sulfate upon the products of limited proteolysis of hTg, chondroitin 6-sulfate was first restricted to a carboxyl-terminal region, starting at residue 2514. A single chondroitin 6-sulfate-containing nonapeptide was isolated in pure form from the products of digestion of hTg with endoproteinase Glu-C, and its sequence was determined as LTAGXGLRE (residues 2726-2734, X being Ser2730 linked to the oligosaccharide chain). In an in vitro assay of enzymatic iodination, hTg-CS produced higher yields of 3,5,5 ′-triiodothyronine (T3) (171%) and 3,5,3′,5′-tetraiodothyronine (T4) (134%) than hTg-CS0. Unfractionated hTg behaved as hTg-CS. Thus, chondroitin 6-sulfate addition to a subset of hTg molecules enhanced the overall level of T4 and, in particular, T3 formation. Furthermore, the chondroitin 6-sulfate oligosaccharide unit of hTg-CS protected peptide bond Lys2714-Gly2715 from proteolysis, during the limited digestion of hTg-CS with trypsin. These findings provide insights into the molecular mechanism of regulation of the hormonogenic efficiency and of the T4/T3 ratio in hTg. The potential implications in the ability of hTg to function as an autoantigen and into the pathogenesis of thyroidal and extra-thyroidal manifestations of autoimmune thyroid disease are discussed.

Thyroglobulin (Tg), 2 the molecular site of thyroid hormone formation, is a large homodimeric glycoprotein with an M r of 660,000. It is also a major antigen, involved in the pathogenesis of thyroid autoimmunity. In fact, experimental autoimmune thyroiditis (EAT) can be induced in genetically susceptible mice by immunization with human thyroglobulin (hTg) in complete adjuvant, and autoantibodies to hTg are found in the blood of humans affected with autoimmune thyroid disease (AITD), even though their pathogenic significance is unclear (1). Several post-translational modifications contribute to the molecular microheterogeneity of hTg, including iodination, glycosylation, phosphorylation, and sulfation (reviewed in Ref. 2). Iodine addition and hormone formation at specific sites have the most obvious implications for thyroid function, but the effects of glycosylation on the hormone-forming efficiency at specific sites and on the antigenicity of Tg have also been documented (3,4). hTg is modified with the addition of several oligosaccharide units of different kinds, among which the N-linked type A (high-mannose) and type B (complex) units have been characterized best as to their composition and localization (5)(6)(7). Type C units are linked to serine and threonine by O-glycosidic bonds and contain D-galactosamine (5). Another O-linked type D oligosaccharide unit was described, which was composed of a repeating D-glucuronic acid-N-acetyl-D-galactosamine sulfate disaccharide attached to the polypeptide chain through a D-galactosyl-D-xylosyl-serine linkage region (8). Further study indicated that the repeating disaccharides were of the chondroitin 6-sulfate type (9,10). However, the number and localization of type D oligosaccharide units in hTg were not determined.
A number of studies have documented the influence of oligosaccharide chains in the processing and/or presentation of glycoproteic antigens by antigen-presenting cells (APCs) (11)(12)(13)(14)(15)(16), and the involvement of chondroitin 6-sulfate oligosaccharide chains in the modulation of cellular immune responses (17)(18)(19)(20). Prompted by these observations, we began to determine in detail the number and localization of chondroitin 6-sulfate oligosaccharide unit(s) in hTg as a basis for further studies on their influence on the immunopathogenicity of hTg in a murine model of EAT. We found that the addition of chondroitin 6-sulfate unit(s) is a major source of molecular microheterogeneity of hTg. In dozens of hTg preparations examined, it was regularly found in a percentage of hTg molecules varying from 32.0 to 71.6% of total. Moreover, we determined that a single chondroitin 6-sulfate unit per hTg polypeptide chain is linked to Ser 2730 . We also observed changes in the hormone-forming efficiency of hTg and in the proteolytic accessibility of the extreme carboxyl-terminal region of hTg, associated with the presence of the chondroitin 6-sulfate oligosaccharide chain. The former observation contributes to delineate a general mechanism, by which modifications in the composition and number of N-linked and O-linked oligosaccharide units determine changes in the hormone-forming efficiency of the main hormonogenic domains at both hTg extremities, whereas the latter finding supports the potential role of the chondroitin 6-sulfate oligosaccharide unit in modifying the susceptibility of hTg to processing by APCs and the repertoire of hTg epitopes thus generated.
Purification of hTg-hTg was prepared as described (21) from informed euthyroid patients hemilaryngectomized for non-thyroidal disease and patients undergoing thyroidectomy for nonfamilial, simple, or multinodular goiter. Protein concentration was assayed by measuring the optical absorbance at 280 nm, using an extinction coefficient of 10 cm Ϫ1 for a 1% solution. Iodine content was assayed as described, using L-thyroxine as the standard (22).
Ion-exchange Chromatography of hTg on HiTrap TM Q-Sepharose HP (Q-IEC)-hTg molecules containing type D (chondroitin 6-sulfate) oligosaccharide units (hTg-CS) were separated from residual hTg molecules, devoid of type D units (hTg-CS 0 ), by ion-exchange chromatography on trimethylamino-substituted Q-Sepharose (Q-IEC) using 5-ml HiTrap TM Q-Sepharose HP columns, equilibrated in 0.025 M Tris/HCl, pH 7.4 (buffer A). Up to 20 mg of hTg in buffer A, plus 0.05 M NaCl, were applied to a column. After washing with buffer A, a linear gradient from 0 to 100% of buffer B (1.2 M NaCl in buffer A) was developed in 24 min at the flow rate of 2.5 ml/min. One-ml fractions were analyzed or stored at Ϫ80°C until use.
Compositional Analysis of Q-IEC Fractions of hTg-D-Glucuronic acid was assayed in Q-IEC fractions by the meta-hydroxybiphenyl method (23), using the same compound (0 -5 g) as the standard. Duplicate samples of 30 -500 g of hTg were dialyzed against 0.01 M NH 4 HCO 3 , dried in centrifugal evaporator, and redissolved in 0.2 ml of double-distilled H 2 O in borosilicate Pyrex tubes. Samples were processed by the addition of 1.2 ml of 0.0125 M Na 2 B 4 O 7 ⅐10H 2 O in H 2 SO 4 and, after heating in boiling water bath for 5 min and cooling on ice, 0.02 ml of 0.15% meta-hydroxybiphenyl in 0.5% NaOH, after which the optical absorbance at 520 nm was read. Correction for aspecific color development was provided by a replicate set of samples, which were treated identically, with the exclusion of meta-hydroxybiphenyl in the 0.5% NaOH reagent. Total neutral hexoses were assayed in duplicate samples of 20 -300 g of hTg by the anthrone method (24), using D-galactose (0 -20 g) as the standard. Sialic acid was assayed in duplicate samples of 40 -500 g of hTg by the thiobarbituric acid method (25), using N-acetylneuraminic acid (0 -15 g) as the standard. Iodine was assayed in duplicate samples of 30 -200 g of hTg, using the method cited (22).
Enzymatic in Vitro Iodination of hTg and Analysis of the Iodine, T 3 , and T 4 Content-hTg from goiters, with an iodine content not exceeding 0.09% on a weight basis, was iodinated enzymatically in vitro. Bulk iodination of 20 mg of unfractionated hTg, at the concentration of 0.45 g/liter, was performed in 0.02 M imidazole/HCl, pH 7.0, using 2 g/ml of LPO from bovine milk, 4 ϫ 10 Ϫ5 M potassium iodide, 1 ϫ 10 Ϫ3 M D-glucose, and 0.19 g/ml glucose oxidase from A. niger. Iodination was stopped with 0.05 M 2-mercapto-1-methylimidazole (MMI). Comparative iodination of unfractionated hTg and of hTg-CS 0 and hTg-CS subfractions of the same hTg preparation was performed with 0.65 mg of each protein, under identical conditions, except that 7.5 ϫ 10 Ϫ5 M potassium iodide, and 0.21 g/ml glucose oxidase were used. Aliquots of 0.1 mg of hTg were withdrawn at 5,15,30,50,70, and 90 min, supplemented with MMI, dialyzed against 0.01 M NH 4 HCO 3 , 5 ϫ 10 Ϫ3 M NaCl, and assayed in duplicate for protein content, using the BCA protein assay reagent and bovine serum albumin as the standard, and for iodine content, as already described. For the assay of T 3 and T 4 , 10 -15 g of hTg were hydrolyzed at 37°C with Pronase, at the enzyme/substrate weight ratio of 1:1, in 0.2 ml of 0.1 M Tris/ HCl, 0.05 M MMI, pH 8.0, to which 15 l of toluene were added. After 24 h, aminopeptidase M was added, at the enzyme/substrate weight ratio of 1:10, and digestion was prolonged for 24 h at 37°C, after which T 3 and T 4 were measured by solid-phase radioimmunoassays in antibody-coated tubes (26). In both assays, 125 I-labeled T 3 or T 4 competed with the respective hormone in the test samples for antibody sites. After the tubes were decanted and radioactivity was measured, the T 3 or T 4 concentration was obtained by interpolation from a calibration curve.
Enzymatic Digestion of Chondroitin 6-Sulfate Oligosaccharide Units of hTg, hTg Subfractions, and hTg Proteolytic Fragments with Chondroitinase ABC-hTg, or its proteolytic fragments in 0.1 M NaCl, 0.05 M Tris/HCl, pH 7.4, were supplemented with equal volumes of 0.1 M sodium acetate, 0.1 M Tris/HCl, pH 8.0. Pooled fractions from the Q2 peak of the Q-IEC of hTg were dialyzed against the same buffer. All samples were supplemented with 200 milliunits/ml of chondroitinase ABC from P. vulgaris, which degrades chondroitin 4-sulfate and 6-sulfate oligosaccharide chains into sulfated disaccharides by hydrolyzing the ␤-(134)-glycosidic bonds between N-acetyl-D-galactosamine 4-and 6-sulfate and D-glucuronic acid in the repeating disaccharide units. The protease inhibitors already indicated with regard to limited proteolysis were added, when not already present, and the samples were incubated at 37°C for 4 h. Samples to be analyzed by SDS-PAGE were immediately precipitated in methanol/chloroform/water (27), although those to be used for enzymatic iodination in vitro or limited tryptic digestion, following chondroitinase ABC digestion, were subjected to Q rechromatography, as already described, concentrated in Centriprep 30 concentrators, and dialyzed against the appropriate buffer.
Two replicas of each gel were prepared. One was stained with 0.1% Coomassie Brilliant Blue R-250 in 25% (v/v) 2-propanol, 10% (v/v) acetic acid and destained in 25% (v/v) methanol, 10% acetic acid. The other one was fixed in 25% 2-propanol, 10% acetic acid, thoroughly rinsed in double-distilled H 2 O, stained with 0.005% DMTCC in 50% formamide for 48 h, and finally destained in 5% (v/v) glycerol in tap water (28). Bands were identified on the basis of their mobilities, according to the detailed characterization of their NH 2 -terminal peptide sequences provided in a previous study (21). Proteolytic fragments to be identified ex novo were separated by reducing or nonreducing SDS-PAGE in 4 -17% acrylamide gradient gels and transferred to polyvinylidene difluoride (PVDF) membranes (Immobilon P, Millipore) by semi-dry blotting in 0.025 M Tris, 0.01 M glycine, at the constant current of 0.8 mA/cm 2 for 1 h. Membranes were rinsed in double-distilled H 2 O, stained with 0.1% Coomassie Brilliant Blue R-250 in 50% methanol, and destained in 50% methanol, 10% acetic acid. Bands were subjected to NH 2 -terminal peptide sequencing at the Molecular Structure Facility, University of California, Davis. Peptide sequences were aligned with the cDNA-derived sequence of hTg (29).
Isolation and Sequencing of Glycopeptide hTg-CSgp-Forty seven mg of hTg-CS were denatured and reduced in 15 ml of 0.3 M Tris/HCl, pH 8.0, 6.0 M guanidine/HCl, 1 ϫ 10 Ϫ3 M EDTA, 0.01 M dithiothreitol at 37°C for 2 h. The reduced protein was carboxymethylated with a 5-fold molar excess of iodoacetamide, with respect to total -SH groups, at room temperature for 30 min in the dark. Alkylation was stopped with excess dithiothreitol. The sample was dialyzed against 0.05 M sodium phosphate, pH 7.8, and digested with endoproteinase Glu-C (protease V8) from S. aureus at the enzyme/substrate weight ratio of 1:100 at 37°C for 18 h. The sample was adjusted with concentrated solutions to 0.025 M Tris/HCl, 0.1 M NaCl, 2.0 M urea, pH 7.4 (buffer C), and loaded onto a 5-ml HiTrap TM Q-Sepharose HP column equilibrated in the same buffer. After washing with buffer C, a gradient was started from 0 to 100% of buffer D (1.2 M NaCl in buffer C) in 55 min, followed by 100% buffer D for 10 min at the flow rate of 1 ml/min. One-ml fractions were monitored for the optical absorbance at 280 nm and D-glucuronic acid content. A single D-glucuronic acid-containing peak was subjected to size-exclusion chromatography on a 1.5 ϫ 100-cm column of Bio-Gel P-2 in 0.01 M NH 4 HCO 3 . A D-glucuronic acid-containing peak, eluted in the void volume, was lyophilized and further purified by gel chromatography on a 0.5 ϫ 40-cm column of Sephadex G-50 fine in 0.01 M NH 4 HCO 3 . One-ml fractions were monitored for peptide content, by measuring the optical absorbance at 220 nm, and D-glucuronic acid content, and a single peptide-and D-glucuronic acid-containing peak was lyophilized. A solubilized aliquot was subjected to NH 2 -terminal peptide microsequencing. Purity was checked by PAGE in Tris/ Tricine/SDS in a 16.5% total acrylamide, 6% N,NЈ-methylenebisacrylamide gel, containing 6.0 M urea (30). Molecular mass standards (marker kit MW-SDS-17S, Sigma) were as follows: myoglobin fragments 1-153 (16,950 Da), 1-131 (14,440 Da), 56 -153 (10,600 Da), 56 -131 (8,160 Da), 1-55 (6,210 Da), glucagon (3,480 Da), and myoglobin fragment 132-153 (2,510 Da). SDS-PAGE was followed by semi-dry transfer to Immobilon P, as described above, for 25 min. The membrane was stained with DMTCC in 50% formamide and destained in tap water (28).
Purification of Carboxyl-terminal Tryptic Fragments of hTg by HPLC-The fragments obtained from the limited digestion of 20 mg of goiter hTg (0.03% of iodine on a weight basis) with trypsin for 20 min, under the conditions described above, were freed from trypsin immediately after digestion by filtration through a 10-ml column of DEAE-Sepharose Fast Flow in 0.025 M Tris/HCl, pH 7.2. hTg fragments were eluted with 0.15 M NaCl in 0.025 M Tris/HCl, pH 7.2, and desalted by gel filtration on a 1.5 ϫ 100-cm column of Bio-Gel P-2, equilibrated with 0.01 M NH 4 HCO 3 . One-mg aliquots of the fragments were fractionated by reverse-phase HPLC with a Vydac C-4 column (250 ϫ 10 mm, 5 m) equilibrated in 0.1% (v/v) trifluoroacetic acid in H 2 O (solvent A) and containing 2% of 0.07% trifluoroacetic acid in acetonitrile (solvent B). After 2 min at 35% of solvent B, the fragments were eluted with a two-step linear gradient from 35 to 46% of solvent B over 35 min and from 46 to 55.5% over the following 35 min. The flow rate was 1 ml/min. Fractions were analyzed by SDS-PAGE in 4 -17% total acrylamide gradient gels under reducing and nonreducing conditions. Corresponding peaks from repeated runs were pooled for further mass spectrometric analysis.
Characterization of Carboxyl-terminal Tryptic Fragments of hTg by Electrospray Mass Spectrometry-ES-MS of peptide h4bis TR were recorded on a Q-TOF Ultima hybrid mass spectrometer (Waters), equipped with an electrospray ion source and operating in positive ion mode. A nanoflow high pressure pump system model CapLC (Waters) was used to deliver a 1-l/min flow rate of 10% acetonitrile to the mass spectrometer. The spectra were scanned at the speed of 10 s/scan. The source conditions were the following: capillary voltage, 3000 V; cone voltage, 100 V; extractor, 0 V; RF lens, 60. Raw data were processed using the MassLynx 3.5 software. Mass calibration was carried out using the multiple charged ions from a separate introduction of horse heart myoglobin (average molecular mass of 16,950.5 Da). The quantitative analysis was performed by integration of the multiple charged ions of the single species. Molecular masses are reported as average values. The mass signals recorded were associated with the corresponding peptides, on the basis of the expected molecular masses, using the Biolynx software (Waters).

A Relevant Fraction of hTg Is Regularly Composed of hTg Molecules Provided with Type D (Chondroitin 6-Sulfate)
Oligosaccharide Units-Type D (chondroitin 6-sulfate) oligosaccharide unit-containing hTg molecules were separated from residual hTg molecules by using ion-exchange chromatography on trimethylamino-substituted HiTrap TM Q-Sepharose HP (GE Healthcare). The Q-IEC of 40 hTg preparations, mostly from nonfamilial, simple, and multinodular goiters but also from normal thyroids, using a NaCl gradient from 0 to 1.2 mol/liter in 0.05 M Tris/HCl, pH 7.4, regularly resulted in the elution of two peaks (henceforth referred to as peaks Q1 and Q2, in elution order) at NaCl concentrations of 0.45 Ϯ 0.01 and 0.84 Ϯ 0.03 M, respectively ( Fig. 1). In 18 chromatograms, the area under the Q2 peak varied from 32 to 71.6% of total (mean, 52.8%; median, 51.7%), with no apparent relation with the iodine content ( Table 1). The hTg subpopulations thus separated were stable, and separation was satisfactory for practical purposes, as shown by pooling and rechromatography of the two peaks (Fig. 2, panel A). We traced chondroitin 6-sulfate oligosaccharide units in hTg fractions by assaying their D-glucuronic acid content, using a meta-hydroxybiphenyl-based method (23). D-Glucuronic acid was almost exclusively restricted in the Q2 peak, which coincided with a fraction of hTg, marked by the presence of D-glucuronic acid-containing type D (chondroitin 6-sulfate) oligosaccharide units (henceforth referred to as hTg-CS) (Fig. 1). Negligible amounts of D-glucuronic acid in the Q1 peak possibly reflected the presence of hTg molecules with very short, nascent type D oligosaccharide chains. The neutral hexose content remained constant or increased slightly along the elution profile of the chromatograms, while the sialic acid content showed no change or only moderate changes in either sense ( Fig. 1 and Table 1). Instead, the number of moles of D-glucuronic acid per mol of hTg regularly exhibited a linear increase across the Q2 peak, ranging from a minimum of about 3 to a maximum of 13.2 mol in the hTg preparation with the lowest fraction of hTg-CS, and of 24.0 mol in some hTg-CS-rich goiter hTg preparations (Table 1). In particular, preparations Mi. and Mi. N. had elution profiles (preparations are named using last initial(s) of patients), in which a shoulder in the trailing edge of the Q2 peak indicated the presence of a secondary component, eluting at 0.96 M NaCl (Fig. 2, panel B). We hypothesized that the complexity and extended span of D-glucuronic acid content of these Q2 peaks might reflect the coexistence of heterodimeric hTg-CS, with a single chondroitin 6-sulfate-bearing subunit, and homodimeric hTg-CS. To prove this, pooled fractions of the Q2 peak from the Q-IEC of hTg Mi. N. were dissociated in 0.04 M Tris/HCl, 2.5 M urea, pH 9.0, at 20°C overnight, as described previously (31), and subjected again to Q-IEC, using a NaCl gradient from 0 to 1.2 mol/liter in 2.5 M urea, 0.055 M Tris/HCl, pH 7.4. As shown in Fig. 2, panel B, the hTg monomers produced by the dissociation of hTg-CS in urea were separated again into peaks Q1 and Q2, eluting at NaCl concentrations of 0.44 and 0.86 mol/liter, respectively. Once again, D-glucuronic acid was restricted in the Q2 peak, and its concentration showed a linear increase from 3.2 to 14.8 mol/hTg mol across the peak span, as opposed to the increase from 3.6 to 24.0 mol/mol of undissociated hTg-CS (Table 1). The Q2 peak of dissociated hTg-CS was superimposed onto the leading component of the Q2 peak of native hTg-CS, with no secondary shoulder at the trailing edge. The Q2/Q1 ratio in hTg-CS dissociated in urea was 72.7 versus 27.3%. These data indicate that the hTg-CS fraction of hTg Mi. N. included both an earlier eluting subfraction of heterodimers, composed of nonchondroitinated (27.3%) and chondroitinated (27.3%) monomers in the 1:1 ratio, and a later eluting homodimeric subfraction, composed only of chondroitinated subunits, and representing the remaining 45.4% of hTg-CS. Thus, hTg-CS was microheterogeneous, both because of the variable number of repeating disaccharide units per chondroitin 6-sulfate chain, which is typical of glycosaminoglycans (GAGs) (32), and because of the coexistence of heterodimeric and homodimeric hTg-CS molecules in hTg-CSrich goiter hTg preparations. When the pooled fractions of the Q2 peak from the Q-IEC of hTg preparation Ca. were digested with chondroitinase ABC and subjected to Q rechromatography, they eluted in a peak superimposed onto the Q1 peak of the initial Q-IEC of native hTg (Fig. 2, panel C). Fig. 1, panels A and B, also shows the native PAGE of the Q-IEC fractions of hTg preparations O. and Ma., evidencing the anodal shift, going from the Q1 to the Q2 peak, because of the added negative charge of D-glucuronic acid residues and sulfate groups of the chondroitin 6-sulfate unit(s).
hTg-CS Has a Higher Efficiency of T 3 and T 4 Formation than hTg-CS 0 and Enhances the Overall Hormonogenic Efficiency of hTg-We investigated the possible influence of the chondroitin 6-sulfate oligosaccharide unit(s) on the hormonogenic function of hTg-CS, in comparison with chondroitin 6-sulfate-devoid hTg (henceforth referred to as hTg-CS 0 ), and unfractionated hTg. Iodination of Tg at the apical pole of thyroid follicular cells occurs by the action of thyroid peroxidase (TPO) (reviewed in Ref. 2). Lactoperoxidase (LPO) is a common alternative for TPO in reconstituted systems of iodination and coupling in vitro. TPO and LPO are equally effective in the selective iodination of a subset of tyrosyl residues, which are able to undergo coupling to form T 3 and T 4 . TPO and LPO also catalyze the coupling of iodotyrosines with similar efficiencies at pH 7.4, even though LPO is significantly less efficient at lower pH (33). Iodine-poor unfractionated hTg, hTg-CS 0 , and hTg-CS from nonfamilial goiters (with no more than 0.09% of iodine, on a weight basis) were iodinated enzymatically in vitro by using 2 g/ml bovine LPO, 7.5 ϫ 10 Ϫ5 M potassium iodide, 0.21 g/ml glucose oxidase from A. niger, and 1 ϫ 10 Ϫ3 M D-glucose over 90 min at 25°C (34) . Fig. 3 shows the means Ϯ S.E. of the amounts of iodine incorporated into hTg, expressed as percent of hTg on a weight basis, and of the levels of T 3 and T 4 formed as a func- a hTg preparations were named after the patients (using the last initial(s)). Preparation O. was from a euthyroid individual, hemilaryngectomized for a non-thyroidal disease; the other preparations were from nonfamilial, simple, or multinodular goiters. b Values were measured in the fractions with the maximal protein concentration in the Q2 peak. c Data indicate the D-glucuronic acid concentration range across the span of the Q2 peak, starting from the valley between the Q1 and the Q2 peaks. tion of iodine bound during four experiments of time course of iodination in vitro. Each experiment was performed using unfractionated hTg, together with hTg-CS 0 and hTg-CS subfractions deriving from the same iodine-poor hTg preparation. The average fraction of hTg-CS in the hTg preparations used was 58.1%. Similar amounts of iodine were incorporated at plateau in hTg and hTg-CS, and slightly higher amounts in hTg-CS 0 (Fig. 3, panel A). However, the number of T 3 millimoles (panel B) and T 4 moles (panel C) synthesized per mol of hTg-CS were in the ratios of 1.70 and 1.34, respectively, with those synthesized per mol of hTg-CS 0 . Instead, hTg and hTg-CS did not differ, as for the efficiency of formation of T 3 and T 4 . These data indicated that hTg-CS had a higher efficiency of T 4 and, especially, T 3 formation than hTg-CS 0 and that the entire population of unfractionated hTg molecules benefited from this property of hTg-CS. A rationale for this may be that chondroitin 6-sulfate chains exerted an influence not only on the hTg-CS molecules they were linked to but also on the hTg-CS 0 molecules they interacted with or on the peroxidase. If so, one would not expect to see differences of hormone yields between hTg-CS and hTg-CS 0 when these were iodinated altogether in the pool of unfractionated hTg. Such a prediction was verified by the lack of significant variations in the amounts of T 3 and T 4 formed, between the Q-IEC fractions of hTg preparation O., physiologically iodinated in vivo (Fig. 4). Previous digestion of hTg-CS with 200 milliunits/ml of chondroitinase ABC from P. vulgaris, at pH 8.0 and 37°C for 4 h, was not associated with reductions of the T 3 and T 4 yields, with respect to native hTg-CS, and of the differences of hormoneforming efficiency between hTg-CS and hTg-CS 0 . Chondroitin 6-Sulfate Oligosaccharide Unit(s) of hTg-CS Are Restricted in the Carboxyl-terminal Region, Downstream Thr 2514 -We were able to differentiate hTg-CS from hTg-CS 0 in SDS-polyacrylamide gels by staining them with DMTCC (Stains All TM , ICN). This dye was reported to stain GAGs metachromatically and, in particular, chondroitin sulfate purple, while staining proteins red (35). In fact, fractions in the Q1 (hTg-CS 0 ) peak were stained pinkish red with DMTCC, and those in the Q2 (hTg-CS) peak were stained purple. Side-byside comparison of the two peak fractions also revealed a subtle cathodal shift in hTg-CS in the presence of SDS. Both metachromasia and the mobility shift were abolished by digestion with 200 milliunits/ml of chondroitinase ABC, at 37°C for 4 h, prior to SDS-PAGE (Fig. 5, panel A). To identify the chondroitin 6-sulfate-containing region(s) of hTg, the Q-IEC fractions of hTg were subjected to limited digestion with thermolysin and trypsin, using DMTCC as a probe. The proteolytic fragments, separated by reducing SDS-PAGE, corresponded exactly to those characterized previously by NH 2 -terminal peptide sequencing (21) and were identified on the basis of their mobilities. Fig. 5, panel B, Fig. 1, the fractions of the Q1 and Q2 peaks obtained therefrom were pooled separately, with the exclusion of three fractions between peaks. Each pool was dialyzed against buffer A and subjected to Q rechromatography, under identical conditions as the initial Q-IEC. The optical absorbance at 280 nm of the fractions of the initial Q-IEC (continuous line) and of the Q rechromatography of the Q1 (dotted line) and Q2 peak (dashed line) was monitored. Panel B, following the Q-IEC of 20 mg of hTg preparation Mi. N., the fractions of the Q2 peak were pooled, dialyzed against buffer A, dissociated in 0.04 M Tris/HCl, 2.5 M urea, pH 9.0, at 20°C overnight, as reported (31), and subjected to Q rechromatography by using a NaCl gradient from 0 to 1.2 mol/liter in 2.5 M urea, 0.055 M Tris/HCl, pH 7.4. The symbols used are as follows: optical absorbance at 280 nm of the fractions of the initial Q-IEC (continuous line) and of the Q rechromatography of the Q2 peak dissociated in urea (dashed line); D-glucuronic acid (GlcUA) content, in mol/hTg mol, of the fractions of the initial Q-IEC (F) and of the Q rechromatography of the Q2 peak dissociated in urea (E). Panel C, following the Q-IEC of 20 mg of hTg preparation Ca., the fractions of the Q2 peak were pooled. An aliquot of 6 mg of this material was taken to the volume of 1.5 ml with Centriprep 30 concentrators, dialyzed against 0.1 M Tris, 0.1 M sodium acetate, pH 8.0, digested with O. with thermolysin. Bands h2 TL , h5 TL , and h7 TL exhibited metachromasia and a cathodal shift, whose extent was inversely related with the fragment apparent mass, going from fraction 29 in the Q1 peak to fractions 48 and 56 in the Q2 peak. This is in keeping with the notion that peptides with high negative net charges bind lower than average amounts of SDS, and they exhibit lower than average mobilities in SDS-PAGE (36). A mixed pattern was apparent in fraction 40, in the valley between the Q1 and Q2 peaks. Both changes were reverted by digestion of the proteolytic fragments of fraction 48 with chondroitinase ABC, prior to SDS-PAGE, although no changes were brought about in fraction 29 (Fig. 5, panel C). Inspection of the flow diagram of the limited proteolysis of hTg with thermolysin revealed that fragments h2 TL , h5 TL , and h7 TL were all located at the carboxyl-terminal side of hTg and shared the region downstream from Leu 1832 (Fig. 6, panel A).
On the other hand, a diffuse band (h8-CS TR ), with an average apparent mass of 41,000 Da, appeared among the tryptic fragments of hTg, going from fraction 29 to fraction 40. It was stained purple with DMTCC, but was not apparent in the gel stained with Coomassie Brilliant Blue R-250. Its intensity of staining and apparent relative mass increased going from fraction 40 to 56 (Fig. 5, panel D), in keeping with the increase of D-glucuronic acid content revealed by the analysis of the Q-IEC fractions across the Q2 peak ( Fig. 1 and Table 1). Upon digestion of the tryptic fragments of fraction 48 with chondroitinase ABC, band h8-CS TR was replaced by a well focused band, corresponding to dechondroitinated h8-CS TR (h8-deCS TR ), with an apparent mass of 36,000 Da, which was stained blue with Coomassie Brilliant Blue R-250 and red with DMTCC. No changes were caused by chondroitinase ABC in fraction 29 (Fig. 5, panel E). The formation of band h8-deCS TR was not prevented when the limited digestion of hTg-CS with trypsin was preceded, rather than followed, by digestion with chondroitinase ABC under identical conditions. Band h8-deCS TR was prepared by limited tryptic digestion of 0.5 mg of hTg-CS, followed by chondroitinase ABC digestion, reducing SDS-PAGE and transfer to PVDF. NH 2 -terminal peptide microsequencing of this   (w/w). The hTg preparation was subjected to Q-IEC, as described in the legend to Fig. 1, and the fractions obtained were analyzed for their iodine (OE), T 3 (F), and T 4 (E) content, as described under "Experimental Procedures." The continuous line indicates the optical absorbance at 280 nm, for which measuring units are not indicated as they were the same as in Fig. 1. band revealed a single sequence (TSSKTA), corresponding to hTg residues 2514 -2519 (29) (see supplemental Table 1). Naturally, this was also the sequence of peptide h8-CS TR . The flow diagram of limited proteolysis of hTg with trypsin shows the carboxyl-terminal location of peptide h8-CS TR , in keeping with the results obtained with thermolysin (Fig. 6, panel B). Because the NH 2 -terminal sequence of peptide h8-deCS TR was identical with one of two sequences found previously in band h4 TR (21), we also determined the sequence of band h4 TR prepared from hTg-CS. As expected, two sequences were found, one starting at residue 1 of hTg (h4 TR ) and the other one at residue 2514 (h4bis TR ) (supplemental Table 1). Thus, of two peptides, both starting at residue 2514, one (h8-CS TR ) contained chondroitin 6-sulfate oligosaccharide units and the other (h4bis TR ) did not. Inspection of Fig. 5, panel D, reveals that the increase of staining intensity of band h8-CS TR , going from fraction 40 to 56, was paralleled by a decrease of intensity of band h4 TR , in keeping with the finding that heterodimeric hTg-CS, yielding peptides h8-CS TR and h4bis TR , and homodimeric hTg-CS, yielding only peptide h8-CS TR , coexisted in the late portion of the Q2 peak (Fig. 2, panel B).
A Single Type D (Chondroitin 6-Sulfate) Oligosaccharide Unit Is Linked to Ser 2730 of hTg-CS-Forty seven mg of hTg-CS were reduced and carboxymethylated, as reported under "Experimental Procedures," and hydrolyzed with endoproteinase Glu-C from S. aureus, at the enzyme/substrate weight ratio of 1:100, in 0.05 M sodium phosphate buffer, pH 7.8, at 37°C for 18 h. Digestion products were subjected to Q-IEC on a 5-ml HiTrap TM Q-Sepharose HP column, using a gradient from 0.1 to 1.2 M NaCl in 0.025 M Tris/HCl, 2.0 M urea, pH 7.4, in 55 min, at the flow rate of 1 ml/min. Most protein was discarded in the flow-through, whereas a unique D-glucuronic acid-containing peak, having negligible absorbance at 280 nm, was eluted late in the gradient (supplemental Fig. 1, panel A). This peak was subjected to size-exclusion chromatography on Bio-Gel P-2 (size-exclusion limit of 1800 relative mass units) in 0.01 M NH 4 HCO 3 , which yielded a D-glucuronic acid-containing peak in the void volume. Its further purification by size-exclusion chromatography on Sephadex G-50 fine, in 0.01 M NH 4 HCO 3 , monitored at 220 nm, is shown in supplemental Fig. 1, panel B. Of the two peaks resolved, only one peak contained D-glucuronic acid. NH 2 -terminal peptide microsequencing of this material revealed a single, homogeneous nonapeptide with the LTAGXGLRE sequence, corresponding to residues 2726 -2734 of the cDNA-derived sequence of hTg (29), X being Ser 2730 linked with the chondroitin 6-sulfate oligosaccharide unit (supplemental Table 1). This glycopeptide will be henceforth referred to as hTg-CSgp. The comparison between the sequence around Ser 2730 and a consensus for the recognition of serine residues by UDP-D-xylose:proteoglycan core protein ␤-D-xylosyltransferase, deriving from the alignment of 51 chondroitin 6-sulfate attachment sites of 19 proteoglycan core proteins (37), revealed a 90% concordance, provided that two insertions of 3 and 2 residues were allowed (Fig. 7). The purified aliquot of hTg-CSgp contained 41.1 g of D-glucuronic acid and 25.2 g of protein.
On the basis of the M r of 903, calculated for the peptide moiety, this indicated an average content of 8.4 mol of D-glucuronic acid per mol of glycopeptide. Electrophoresis of 18 g of hTg-CSgp in a 16.5% polyacrylamide gel in Tris/Tricine/SDS, followed by transfer to a PVDF membrane and staining with DMTCC, revealed a homogeneous, meta-chromatic band with the diffuse migration, which is typical of GAGs (17,32), and an apparent molecular mass of 13,000 -20,000 Da, both indicating polydispersity (supplemental Fig. 1,  panel C). hTg-CSgp was labile upon freezing and thawing.
The Chondroitin 6-Sulfate Oligosaccharide Unit of hTg-CS Protects Peptide Bond Lys 2714 -Gly 2715 from Proteolysis-Next, we sought an explanation for the difference between the apparent relative molecular masses of peptides h8-deCS TR (36,000 Da) and h4bis TR (29,000 Da), by determining the carboxyl terminus of the latter. In fact, only a minor part of the mass difference could be accounted for by the residual oligosaccharide (GlcUA-Gal-Gal-Xyl), which was left in situ after chondroitinase ABC digestion. Even though peptide h4bis TR comigrated with h4 TR , in reducing SDS-PAGE of the products of tryptic digestion of hTg described above, previous observations indicated that the homologous peptide b11 TR of bovine Tg was free of comigrating species under nonreducing conditions, because it was the sole fragment not linked to other fragments by disulfide bonds (38). In fact, nonreducing SDS-PAGE of the tryptic fragments of hTg yielded only two bands instead of several bands seen with reduction (supplemental Fig. 2, panel B). NH 2terminal microsequencing of band h4 TR-NR , transferred onto PVDF, revealed the same sequence as peptide h4bis TR , starting at residue 2514, together with traces of a sequence starting at residue 2518 (supplemental Table 1). Thus, the products of limited tryptic digestion of goiter hTg (with an iodine content of 0.03%, on a weight basis), at the enzyme/substrate weight ratio of 1:100, at 30°C for 20 min, were subjected to reverse-phase HPLC in the absence of reduction, using a Vydac C-4 column  A) and trypsin (panel B) (21). Each bar represents a peptide, with the peptide number above the bar, the apparent relative molecular mass below at right, and the amino-terminal residue number below at left, according to Ref. 29. Fragments of hTg-CS, which exhibited cathodal shifts and metachromasia upon staining with DMTCC, with respect to their counterparts deriving from hTg-CS 0 , are shown in gray. Their apparent relative molecular masses, both before and after digestion with chondroitinase ABC (ABC), are indicated. Cross-hatching of bars h4 TL and h2 TR indicates that, under the digestion conditions used, the corresponding peptides were not detected in the gels shown in Fig. 5.  (29). Black circles with white lettering mark the sequence found for the purified glycopeptide hTg-CSgp (with the proviso that Ser 2730 appeared as a blank). A consensus sequence for the recognition of core protein serine residues by UDP-D-xylose:proteoglycan core protein ␤-D-xylosyltransferase (37) is aligned with the hTg sequence, and concordances are indicated. A diagram of the chondroitin 6-sulfate oligosaccharide unit attached to Ser 2730 is shown (8 -10). The number of repeating ␤-Dglucuronic acid-N-acetyl-D-galactosamine 6-sulfate disaccharide units per chondroitin 6-sulfate chain found in the present study is indicated. The two sites of tryptic cleavage between residues 2714 -2715 and 2745-2746 found in the present study and the preferential site of T 3 formation at Tyr 2747 (29,40,41) are also marked.
(250 ϫ 10 mm, 5 m) and a gradient of acetonitrile in trifluoroacetic acid (supplemental Fig. 2, panel A). Nonreducing SDS-PAGE of peak 2 revealed band h4 TR-NR . Upon reduction, this band exhibited the same relative apparent mass as peptide h4bis TR (supplemental Fig. 2, panel B). The mass values associated with this band were determined by ES-MS (supplemental Table 2) and defined a peptide having ragged amino-and carboxyl-terminal ends (residues 2512-2514 and 2713-2714, respectively) and containing a high-mannose oligosaccharide unit, composed of 2 N-acetyl-D-glucosamine and 8 or 9 D-mannose residues, linked to Asn 2563 , as reported previously (7,29,39). Furthermore, the ES-MS analysis of HPLC peak 1 revealed a mass value corresponding to hTg peptide 2715-2745 (supplemental Table 2). Thus, peptide h4bis TR was truncated at Lys 2714 , whereas peptide bond Lys 2714 -Gly 2715 was protected from proteolysis in peptide h8-CS TR , which extended through Ser 2730 , with its bound chondroitin 6-sulfate unit. Another tryptic site was located between Lys 2745 and Thr 2746 (Fig. 7).

DISCUSSION
The present study shows that type D (chondroitin 6-sulfate) oligosaccharide units are a main source of molecular microheterogeneity of hTg, being regularly found in a significant and sometimes predominant fraction of it. We developed an IEC method, which permitted us to separate chondroitin 6-sulfatecontaining hTg molecules (hTg-CS) from the residual hTg molecules (hTg-CS 0 ). By exploiting the changes of electrophoretic mobility and staining properties conferred upon the products of limited proteolysis of hTg by chondroitin 6-sulfate units, we first restricted the chondroitin 6-sulfate-containing regions of hTg to a carboxyl-terminal peptide, starting at Thr 2514 . The subsequent purification of a homogeneous, D-glucuronic acid-containing nonapeptide (hTg-CSgp), corresponding to hTg residues 2726 -2734, permitted us to establish Ser 2730 as the sole site of chondroitin 6-sulfate addition in hTg. The surrounding sequence showed concordance with a consensus sequence for the recognition of core protein serine residues by UDP-D-xylose:proteoglycan core protein ␤-D-xylosyltransferase, with the exception of two insertions of three and two residues (Fig.  7), in keeping with the finding that recognition sequences of proteins, which are not modified quantitatively with chondroitin sulfate, match the consensus less well than those of proteins that are (37). The presence in hTg of O-linked oligosaccharide units, composed of 1 mol of D-xylose, 2 mol of D-galactose, 11 mol of a D-glucuronic acid-N-acetyl-D-galactosamine disaccharide, and up to 14 mol of sulfate per mol of serine was reported (8). Based on numbers of moles of D-glucuronic acid per mol of hTg varying from 4.8 to 8.7 in different hTgs, it was suggested that some hTg molecules might be devoid of type D units, although some might contain units of different sizes or multiple units (8). Subsequent studies showed that the repeating disaccharides were of the chondroitin 6-sulfate type (9,10). Our data show that a relevant hTg fraction regularly contained a single chondroitin 6-sulfate unit per polypeptide chain, linked to Ser 2730 and composed of a broadly varying number of repeating disaccharide units. In some goiter hTg preparations with high contents of hTg-CS, hTg-CS heterodimers, in which only one monomer contained a chondroitin 6-sulfate chain, coexisted with hTg-CS homodimers. In the monomers derived from the dissociation of hTg-CS in urea, the net number of disaccharide units per chondroitin 6-sulfate chain did not exceed 14.
We also show that hTg-CS had a higher efficiency of hormone formation than hTg-CS 0 and that the whole unfractionated hTg benefited from this property. Thus, chondroitin 6-sulfate addition represents an ergonomic mechanism, by which the post-translational modification of a fraction of molecules influences the overall function of hTg. The greater advantage in the formation of T 3 , with respect to T 4 , associated with the chondroitin 6-sulfate unit, and the proximity of the attachment site of the latter to the carboxyl terminus of hTg suggest that it may influence hormonogenesis by affecting the function of the site of preferential T 3 formation at Tyr 2747 (29,40,41). Normal ranges of serum T 4 concentration (64 -142 nmol/liter) and T 3 concentration (1.1-2.9 nmol/liter) have higher limits that exceed the lower ones only by a factor of 2.2 and 2.6, respectively (42), so that the increases in the rates of T 4 and T 3 formation in hTg-CS, in comparison with hTg-CS 0 , observed in the present study may be physiologically significant. The regulation of the T 4 /T 3 ratio in thyroid secretion is crucial for maintaining physiological concentrations of active hormone (T 3 ) in blood and is finely tuned by multiple mechanisms, under the control of TSH (2). The T 4 /T 3 ratio is regulated first at the biosynthetic level in Tg and decreases in experimental animals given TSH. In rabbit and guinea pig Tg, TSH stimulates T 3 formation at tyrosine 2747 and decreases T 4 synthesis at tyrosine 5 (40). The diminished T 4 formation at the amino terminus of Tg is mediated by the TSH-stimulated maturation of the N-linked oligosaccharide chain linked to Asn 91 , from the high-mannose to the complex type (3). The data that we present contribute to delineate a common mechanism, mediated by modifications of the composition and number of N-linked and O-linked oligosaccharide chains, which may be responsible for changes in the hormonogenic efficiency at both hTg termini, affecting the overall ratio of T 4 over T 3 formation in hTg. For hormone release to occur, Tg must be internalized into thyroid follicular cells and degraded in lysosomes. A part of Tg, internalized via the endocytic receptor megalin, bypasses lysosomes and is transcytosed across cells, to be secreted from the basolateral pole (43). Tg binding to megalin is facilitated by accessory interactions with heparan-sulfate proteoglycans on the surface of thyroid cells (44,45). Experiments are in progress to determine whether chondroitin 6-sulfate units may interfere with the binding of hTg to megalin. Conceivably, the addition of chondroitin 6-sulfate to hTg, besides improving the yield of active hormone, may also prevent the nonproductive internalization of hTg by transcytosis.
Furthermore, we identified two sites, one between Lys 2714 and Gly 2715 and the other between Lys 2745 and Thr 2746 , near the site of chondroitin 6-sulfate addition, which were susceptible to limited proteolysis with trypsin. Our data show that the former was protected from proteolysis in hTg-CS but not in hTg-CS 0 . Both the hormone-forming advantage of hTg-CS over hTg-CS 0 and the resistance of peptide bond 2714 -2715 to proteolysis persisted after digestion of hTg-CS with chondroitinase ABC. Such findings indicate that the residual oligosaccharide that was left in situ was sufficient for the observed effects. This is not surprising with respect to proteolysis, because O-linked monoand disaccharides were able to protect specific cleavage sites from cathepsin L (12). Moreover, the production of proteolytic fragments by an oxidative cleavage mechanism was repeatedly observed in the course of Tg iodination both in vitro (46,47) and in vivo (48). The protection from the occurrence of such cleavages in the carboxyl-terminal end of hTg, possibly afforded by the oligosaccharide chain at Ser 2730 , even in its residual form after digestion with chondroitinase ABC, might contribute to improve the hormone yields. Verification of this hypothesis will be the aim of future work. It is worth noticing that the product of digestion of hTg-CS with chondroitinase ABC differed from native hTg-CS 0 , because this enzyme hydrolyzes the ␤-(134)-glycosidic bonds between N-acetyl-D-galactosamine 6-sulfate and D-glucuronic acid, thus removing ␤-Gl-cUA-(133)-GalNAc-6S disaccharides, while leaving in situ a GlcUA-Gal-Gal-Xyl linkage oligosaccharide. Moreover, Hascall et al. (49) reported that the disaccharide closest to the linkage oligosaccharide in chondroitin 4-sulfate resisted digestion by chondroitinase ABC. Our data entail, even though some of the hTg molecules in the Q1 peak might contain short oligosaccharide chains of this kind, that all the hTg molecules from the Q2 peak digested with chondroitinase ABC contained residual stubs, in a heterodimeric or even in a homodimeric form.
The reported effects on proteolysis are in support of a possible modifying influence of the chondroitin 6-sulfate unit in the processing of hTg by APCs, and in the ability of hTg to function as an autoantigen, particularly because the chondroitin 6-sulfate unit was located within an epitope-rich region, harboring several T cell-related epitopes, capable of causing EAT in genetically susceptible mice, and B cell-related epitopes, recognized by circulating autoantibodies of patients with AITD (reviewed in Ref. 1). The effects of oligosaccharide chains on the activity of proteolytic enzymes involved in antigen processing and on the affinity of epitope binding to major histocompatibility complex or T cell receptor molecules have been reviewed (11). O-Linked mono-and disaccharides in tumor-associated glycoprotein MUC1 restricted the repertoire of the epitopes produced and/or presented in a site-specific manner, either by limiting the proteolytic accessibility of the protein (12) or by preventing epitope recognition by a peptide-specific T cell hybridoma (13). N-Linked oligosaccharide chains inhibited the generation of a self-epitope from glutamate receptor subunit 3 (14) and of cytotoxic lymphocyte-specific epitopes from influenza A nucleoprotein (15). Moreover, a keratan sulfate chain masked an arthritogenic T cell epitope in the G 1 domain of aggrecan, whereas the depletion of multiple chondroitin sulfate side chains generated clusters of chondroitin sulfate stubs, which activated specific B cells to function as APCs (16). We expect that the chondroitin 6-sulfate chain of hTg may inhibit the processing of hTg peptide 2731-2744 (amino acid numbering as per Ref. 29), known for its ability to stimulate in vitro strong proliferative responses and the adoptive transfer of EAT by splenic lymphocytes of CBA mice immunized with mouse Tg (50). Other epitopes might be abrogated by the chondroitin 6-sulfate unit, should their generation require cleavage in the vicinity of Ser 2730 , within a range including the Lys 2714 -Gly 2715 peptide bond on the amino side.
In addition, significant effects of chondroitin sulfate oligosaccharide chains on cellular immune responses have been documented. A small percentage (2-5%) of invariant chain molecules, associated with class II major histocompatibility complex molecules, are modified with the addition of a single chondroitin sulfate chain at Ser 291 (Ii-CS). In this form, they remain associated with class II molecules at the surface of APCs (51,52), where they act as accessory molecules in antigen presentation, facilitating the interactions between APCs and T cells, and greatly enhancing class II-dependent allogeneic and mitogenic T cell responses (17). Such effects occur through interactions of Ii-CS with CD44 on responding T cells, as they can be inhibited both by anti-CD44 antibodies, and by a soluble form of CD44 (CD44Rg), which binds Ii-CS directly (17). Treatment of spleen cells with xyloside, which inhibits GAG addition (53), interferes with their antigen-presenting capabilities (54). Serglycins, small proteoglycans stored in secretory granules of hematopoietic cells, activate the CD3-dependent release of cytokines and proteases from CD44-positive cytotoxic lymphocytic clones (18) by interacting with CD44 through their chondroitin 4-sulfate and 6-sulfate side chains (19). Moreover, CD44 binding to aggrecan, a major proteoglycan of the cartilage matrix, through the chondroitin 4and 6-sulfate side chains of the latter, can trigger the oligomerization of CD44 molecules and the activation of intracellular signaling (20). In keeping with these suggestions, preliminary results indicate that the chondroitin 6-sulfate oligosaccharide unit significantly affects the immunopathogenic capacity of hTg in CBA/J(H-2A k ) mice by a complex mechanism. 3 Finally, the close clinical association between Graves hyperthyroidism and thyroid-associated ophthalmopathy (TAO) led to the hypothesis that the latter might be the result of an autoimmune response against orbital autoantigen(s) that are also present in the thyroid. Candidate antigens, besides the TSH receptor, include Tg (55). Some authors detected hTg in the orbital tissues of patients with TAO and hypothesized that it might be transferred there via thyroid-orbit connections evidenced by lymphography (56). hTg was found prevailingly in fibroadipose tissue (57). It was proposed that the ability of hTg to bind to GAGs, including chondroitin sulfate B and C, might mediate its localization in orbital tissues, where it might function as a target of immune responses directed against the thyroid (58). Metachromatic GAGs accumulate in TAO and thyroid-associated dermopathy (55). Edematous connective perimysial tissues of patients with TAO are composed mainly of hyaluronan and chondroitin sulfate (59). Accumulation of GAGs and adipose tissue expansion is apparent in the fatty connective tissue of the posterior orbit (60). In our opinion, the presence of an integral GAG chain, in a fraction of hTg molecules, may represent a mechanism by which autoaggressive responses against hTg may spread to connective tissue antigens with shared GAG chains, particularly in the event that the synthesis of both be quantitatively and/or qualitatively dysregulated. Proving this hypothesis will require the fine structural characterization of the chondroitin 6-sulfate chains of hTg and of the GAGs of orbital connective tissues from patients with TAO, and the demonstration of cross-reacting B and/or T cell clones.
In conclusion, a single chondroitin 6-sulfate chain linked to Ser 2730 in a relevant fraction of hTg molecules influences both the hormone-forming efficiency and the proteolytic accessibility of the carboxyl-terminal region of hTg. These effects may bear consequences on thyroid homeostasis and autoimmunity. Further work will be aiming to determine the following: 1) What are the physiological limits of hTg-CS abundance in hTg? 2) How is the synthesis of the chondroitin 6-sulfate unit of hTg regulated, particularly regarding the role of TSH? May changes of hTg-CS abundance mediate thyroid adaptation to iodine deficiency or inherited defects of thyroid hormone synthesis or secretion? 3) Do any correlations exist between thyroid function and variations in the hTg-CS/hTg-CS 0 ratio or in the chondroitin 6-sulfate chain length, as analyzed by gel electrophoresis of the oligosaccharide units released from hTg by ␤-elimination, in comparison with appropriate standards? 4) How does the chondroitin 6-sulfate unit influence the processing of hTg by APCs and cellular immune responses to hTg in vivo? It is our opinion that a systematic investigation may shed light on the pathogenesis of thyroid diseases, particularly AITD.