Sialic Acid Capping of CD8β Core 1-O-Glycans Controls Thymocyte-Major Histocompatibility Complex Class I Interaction*

Bidentate interaction of a T-cell receptor and CD8αβ heterodimer with a peptide-MHCI complex is required for the generation of cytotoxic T-lymphocytes. During thymic development, the modification of CD8β glycans influences major histocompatibility complex class I binding to T-cell precursors called thymocytes. ES mass spectrometry (MS) and tandem MS/MS analysis were used to identify the changes occurring in the CD8β-glycopeptides during T-cell development. Several threonine residues proximal to the CD8β Ig headpiece are glycosylated with core-type 1O-glycans. Non-sialylated glycoforms are present in immature thymocytes but are virtually absent in mature thymocytes. These results suggest how sialylation in a discrete segment of the CD8β stalk by ST3Gal-1 sialyltransferase creates a molecular developmental switch that affects ligand binding.

Within complex biological systems, glycans serve as key structural and functional elements (1). Cell surface glycans are altered during cell differentiation and activation in conjunction with changing glycoprotein expression pattern (2). The precise chemistry of glycan modification requires that the vertebrate genome encodes a variety of enzymes that modify various classes of glycans. For example, more than a dozen sialyltransferases with unique substrate specificities and expression patterns operate at the level of the Golgi apparatus (3). Glycosylation has particular relevance in the immune system where cell surface proteins and lipids involved in immune recognition and regulation are typically modified by various glycan structures during cell differentiation and activation (4,5). One particular example is the T-cell surface glycoprotein, CD8.
The CD8 cell surface molecule is critical for the development and activation of T-cells whose T-cell receptors (TCR) 1 recognize peptides bound to major histocompatibility complex class I (MHCI) molecules (6,7). This co-receptor is encoded by two distinct genes, ␣ and ␤, whose polypeptide products are expressed in one of two forms, CD8␣␣ homodimers or CD8␣␤ heterodimers (8,9). Most T-cells mature within the thymus and express cell surface CD8␣␤ receptors. Previously, we and others (10,24) have shown that immature thymocytes bind peptide-MHCI (pMHCI) tetramers more avidly than mature thymocytes. The binding difference is the result of a developmentally regulated glycosylation modification involving sialic acid residues. Evidence of this is the increased CD8␣␤-MHCI avidity of mature thymocytes following treatment with neuraminidase, an enzyme that removes sialic acid residues from cell surface molecules. Moreover, the sialyltransferase ST3Gal-1, which specifically sialylates core 1-O-glycans, is involved in controlling the differential binding as evidenced by decreased CD8␣␤-MHCI avidity after induction of ST3Gal-1 (10). Given that CD8␤ glycans change during thymic development (10,11), we examined the physical nature of CD8␤ O-glycosylation. Through the application of recent advances in mass spectrometry (12), we have been able to identify a developmental change in CD8␤ stalk glycosylation, which functions as a molecular switch to critically affect ligand binding.

EXPERIMENTAL PROCEDURES
CD8␤ Sample Preparation-Unfractionated thymocytes from C57BL/6 mice (ϳ2-4 ϫ 10 9 /experiment) were lysed in buffer containing 1% Triton X-100. Resulting lysates were processed as described previously (10) using anti-CD8␤ mAb YTS 156.77 for immunoprecipitation. After separation on two-dimensional non-reducing/reducing SDS-PAGE, the proteins were stained with Gel-Code Blue reagent and the double positive (DP) and single positive (SP) CD8␤ bands were excised. The gel slices were digested with either trypsin or N-glycanase (PNGase) followed by trypsin using conditions described previously (13).
Mass Spectrometry Analysis-The tryptic peptide extracts were cleaned by reverse-phase (C18) trapping and eluted by 1:1 MeOH:HOH with 1% acetic acid or acetonitrile:0.01% trifluoroacetic acid or acetonitrile:0.01% formic acid in various experiments. The eluted volume (2-3 l) was loaded in a nanospray tip and analyzed by ES mass spectrometry (MS) and tandem MS/MS analysis on either ABI QStar Pulsar or Micromass Q-TOF quadrupole time-of-flight mass spectrometers. In some experiments, the same procedure was applied to gel pieces with the addition of PNGase prior to tryptic digestion.
Nano-LCMS and MS/MS experiments on SP and DP in-gel digest extracts were carried out on a Q-TOF instrument using a 75-m C18 reverse-phase column eluted with a gradient of acetonitrile in 0.01% formic acid at a flow rate of 200 nl/min. Data-dependent acquisition of MS/MS spectra was controlled by setting threshold ionization values for doubly, triply, and quadruply charged ions, and collision energies were programmed in relation to the values seen to produce good fragmentation in the earlier nanospray experiments.

RESULTS
The Peanut Agglutinin (PNA) Lectin Detects Glycosylation Differences on Developing Thymocytes and Binds CD8␤-Within the thymus, T-cell precursors move through a series of developmental stages distinguishable by different CD4 and CD8 cell surface expression patterns (14). CD4 Ϫ CD8 Ϫ double negative (DN) cells progress to the CD4 ϩ CD8 ϩ DP stage in the thymic cortex, and upon successful selection, mature into either CD4 ϩ or CD8 ϩ SP T-cells in the thymic medulla (Fig. 1a). The DP to SP transition is dependent on TCR ligation by pMHC molecules containing self-peptides (15). Varied T-cell surface glycosylation patterns detected by the plant lectin, PNA, also mark thymic developmental progression. PNA binds to core 1-O-glycans bearing terminal galactose residues (Gal␤1-3GalNAc␣Ser-Thr), staining immature cortical thymocytes strongly (PNA high ) and mature medullary thymocytes weakly (PNA low ) (Fig. 1b) (16,17). The change in PNA reactivity is attributable to the induction of the ST3Gal-1 sialyltransferase within the hematopoietic compartment that catalyzes the addition of sialic acid (Sia) residues in a ␣2-3 linkage to terminal galactose (Sia␣2-3Gal␤1-3GalNAc␣Ser-Thr), capping the PNA binding site in medullary thymocytes (Fig. 1, b and c) (18). Genetic disruption of ST3Gal-1 causes PNA high reactivity to persist into the medullary thymocyte compartment ( Fig. 1b) (2). Five likely possibilities for O-glycan structures that could bind PNA in the cortex are presented in Fig.  1c. Among the small group of thymocyte cell surface molecules identified as being PNA-reactive are CD45, CD43, and CD8 (19). In particular, the CD8␤ chain is a major component of the differential PNA binding observed on immature thymocytes (Fig. 1d).
Mass Spectrometry Analysis of CD8␤ Glycans-To identify N-and O-linked glycan sites on CD8␤ and define glycosylation changes associated with the DP to CD8 SP thymocyte transition, mass spectrometry was used to analyze tryptic peptides of CD8␤ prepared from immunoprecipitates. CD8␣␤ proteins were immunoprecipitated from lysates of cell surface-labeled DP and CD8 SP thymocytes sorted by MoFlo, using Sepharose- coupled anti-CD8␤ mAb and separated on two-dimensional non-reducing/reducing SDS-PAGE gels as described previously (10). Whereas three distinct pairs of CD8␣␤ heterodimers (␣38Kd ␤30Kd, ␣38Kd ␤29Kd, and ␣Ј33Kd ␤29Kd) are evident on DP thymocytes, CD8␤ heterogeneity is reduced upon DP to SP maturation (Fig. 2a). By the CD8 SP stage, a single 38-kDa CD8␣ subunit is paired with a major 30-kDa CD8␤ glycoform. Note that aside from the 33-kDa CD8␣Ј cytoplasmic RNA splice variant found in DP thymocytes, CD8␣ is not detectably altered during thymic maturation as assessed by the two-dimensional gel analysis. A composite pattern is obtained from silver staining of proteins immunoprecipitated from unfractionated thymocytes and run in the two-dimensional gel system (see "Experimental Procedures"). Since DP thymocytes comprise 80% of thymocytes while the CD8 SP fraction accounts for merely 3-5%, the total thymocyte CD8␣␤ immunoprecipitation pattern is most similar to that of the isolated DP thymocytes. The two-dimensional gel pattern of CD8␤ proteins precipitated from sorted DP and CD8 SP thymocytes provided a ready basis to obtain "DP" and "CD8 SP" thymocyte-derived gel slices as indicated in the Fig. 2b (inset). The excised gel slices were then digested with either trypsin or PNGase followed by trypsin using the conditions described previously (13). Following the in-gel digestion and extraction, the purified peptide mixture was analyzed using ES on Q-TOF geometry tandem hybrid instrumentation (20) in both MS and MS/MS modes (see "Experimental Procedures").
An early comparative study of CD8␤ SP and DP preparations run by nanospray ES-MS from a formic acid/acetonitrile solution showed a clear quadruply charged signal at m/z 823.88 (corresponding to M ϭ 3,291.49) in the SP sample, which was virtually absent in the corresponding DP analysis. Fig. 3 shows the MS/MS spectrum of this ion at moderate collision energies of 30 -50 eV. The spectrum is the sum of data obtained at 30, 40, and 50 eV. The spectrum shows definitive evidence of glycosylation via major signals at m/z 204 (HexNAc), 366 (Hex-HexNAc), 290 (NeuGc minus H 2 O), 308 (NeuGc), and 673 (NeuGcHexHexNAc). Hex denotes any six-carbon neutral sugar, including glucose, galactose, and mannose, whereas HexNAc is a six-carbon sugar with an N-acetylated amino group at position 2. NeuGc is formed by an enzyme that catalyzes the hydroxylation of the N-acetyl group attached to C 5 of the nine-carbon sialic acid backbone. The mouse CD8␤ sialic acids identified were of the glycolyl variety as expected rather than the N-acetyl (NeuAc)-type that predominates in mammalian brain tissue (1). The collision energies were chosen to provide both carbohydrate and peptide backbone fragmentation (21) in an effort to identify the CD8 peptide sequence carrying the glycosylation. Signals observed at m/z 215, 314, 413, 528, 627, and 740 were interpreted as N-terminal peptide fragments, b ions (20), assignable to a sequence . . . VVDV(L/I) . . . , which is present in CD8␤ in tryptic peptide-(112-125), LTVVDVLPTTAPTKK (Figs. 3 and 4). The threonines at positions 113, 120, 121, and 124 represented possible sugar attachment sites via O-glycosylation. A free (non-glycosylated) peptide of this sequence would be expected to show intense C-terminal ammonium ion (yЉ) fragmentation (20,21) corresponding to fragmentation at the labile L-P and A-P bonds, giving calculated nominal masses of 715 and 345, respectively. Neither of these signals is present in Fig. 3. However, with increasing collision energy, which causes preferential cleavage of sugar residues, prominent signals at m/z 843 and 473 begin to appear, which are 128 Da higher in mass. These signals were assigned to C-terminal proline cleavage fragments PTTAPTKK and PTKK, respectively, thus proving that the peptide backbone for the 823.88 4ϩ -peptide is in fact the CD8␤-(112-126) sequence LTVVDVLPTTAPTKK. The fact that the 473 ion was only created at higher collision energies showed that at least threonine 124 is glycosylated.
Subtracting this peptide mass (1581.93 Da) from the experimentally determined mass of the glycopeptide (3291.49 Da) then allowed the total carbohydrate mass to be calculated as 1709.56 corresponding to NeuGc 2 Hex 3 HexNAc 3, in agreement with the sugar fragment ions described earlier. The b ion peptide fragment series (m/z 215, 314, 413, 528, 627, and 740) is visible at relatively low collision energies, which in our experience would not cause total carbohydrate elimination from the fragments. This strongly suggests that threonine 113 is not glycosylated (Fig. 4). In additional variable collision energy experiments, the proline yЉ ion fragments at m/z 843 and 473 were seen to carry glycosyl substituents via signals at m/z 1046 (843 ϩ HexNAc), 1208 (843 ϩ HexHexNAc), 1249 (843 ϩ Hex-NAc 2 ), 676 (473 ϩ HexNAc), and 838 (473 ϩ HexHexNAc). A consideration of these data together with the sugar fragment ions observed (both at low mass and as neutral losses from the quasimolecular ion) suggested that threonines 120, 121, and 124 are each O-linked to core 1 (HexHexNAc) structures, two of which are capped with N-glycolylneuraminic acid. The virtual absence of the m/z 823.88 signal from the corresponding DP preparation provided the first molecular evidence of differential sialylation of the CD8␤ SP/DP glycoproteins.
Because of the complexity of the nanospray MS spectra of the total CD8␤ digests, a further preparation of SP and DP CD8␤ was then examined by nanoLC-MS and nanoLC -MS/MS (see "Experimental Procedures"). These data were important in allowing the unambiguous confirmation of the glycosylation state of the remainder of the CD8␤ stalk region. This was achieved by locating N-glycolylneuraminic acid-containing signals at m/z 1040 4ϩ , 1117 4ϩ , and 1194 4ϩ . These glycopeptides were found eluting at 24.8 min, 26.7/27.4 min (doublet), and 28.1 min respectively, compared with the CD8␤-(112-126) disialyl glycopeptide described earlier in the nanospray experiment, which eluted at 30.2/31.1 min as evidenced by signals at 823 4ϩ and 1098 3ϩ .

Key Differences in CD8␤ O-linked Glycopeptides Linked to T-cell Maturation-Since the comparison of the nanoLC-MS
and nanospray experiments then showed little or no suppression in the nanospray data and because the "residence time" for MS/MS analysis is much longer in nanospray leading to better quality data in complex studies such as these, subsequent comparative analyses were conducted using nanospray ESMS. Fig. 5 gives the comparative data for a CD8␤ SP/DP preparation. The data show an almost complete absence of non-sialylated HexHexNAc structures in the SP sample compared with the DP via signals at m/z 893, 963, and 1284, whereas the principal sialylation state in both SP and DP is mono-sialyl (m/z 995 and 1040) as seen in the other studies. Fig. 6 is an expansion of the m/z 893 region of Fig. 5 showing the comparative relative abundance of the key signal at m/z 893 for SP and DP and its absence in the SP sample. Note that the absence of non-sialylated signals in SP is not attributable to the low relative abundance of this sample, because the signals at m/z 995 and 1040 in the same approximate mass range of Fig. 5  key difference in the O-but not the N-linked glycopeptides. This difference is the almost complete absence of nonsialylated Hex 3 HexNAc 3 structures in CD8 SPs (Figs. 5 and 6). Conversely, double and triple Sia-capped Hex 3 HexNAc 3 CD8␤-(112-126)-peptides occur in greater abundance in SP rather than DP, although these differences are smaller by comparison. Unexpectedly, our structural studies have revealed that both SP and DP are mainly mono-sialylated in the stalk region despite the presence of five core-type 1 Oglycan substitutions (Thr-120, Thr-121, Thr-124, Thr-127, and Thr-128) (Fig. 4). Sialylation occurs principally within the 120 -124 sequence, and there appears to be little or no additional sialylation of residues 127 and 128. Since SP thymocytes are PNA low , this finding suggests that the majority of CD8␤ stalk O-glycans are inaccessible to this lectin. Thus, the key change between DP and SP is core-type 1 sialylation at a single site in the 121-124 stalk segment. This site is presumably that recognized by the peanut lectin. DISCUSSION CD8␣␤ rather than CD8␣␣ was previously shown to be the critical co-receptor on thymocytes for MHCI binding (10,22). In CD8␤ knock-out mice, for example, CD8␣␣ homodimers are expressed on the surface of thymocytes but fail to support significant MHCI binding activity as assessed by pMHCI tetramers using flow cytometry. Moreover, the varied glycosylation pattern of CD8␤ at different thymic developmental stages correlates with the noted change in the CD8␣␤ ligand binding activity whereby DP thymocytes interact with MHCI more avidly than CD8 SP thymocytes. Several hypotheses have been offered to account for this developmentally programmed alteration in CD8␣␤ co-receptor MHCI ligand-binding function (10,  [23][24][25]. First, the sialylation of the CD8␤ stalk may affect the orientation of the co-receptor globular head domains relative to the T-cell membrane and/or CD8␣␤ domain-domain association strength, modulating the ability of the distal binding surface of the CD8␣␤ Ig-like domain to clamp MHCI (10). Second, sialic acid residues might reduce the clustering of CD8␣␤ molecules on thymocyte surfaces because of repulsion of the negatively charged sugars, preventing MHCI binding as detected by pMHCI tetramers. Third, multivalent mammalian lectins such as galectins (26), known to control clustering of certain T-cell surface glycoproteins (27), might "pre-cluster" non-sialylated CD8 co-receptors in DP thymocytes but not sialylated glycanbearing counterparts on CD8 SP thymocytes. Fourth, the effects might be attributed to sialylation of molecules other than CD8␣␤ given the previous lack of direct evidence for developmentally controlled sialic acid addition to the CD8 co-receptor itself.
The current biochemical analysis favors reorientation of the globular head domains of CD8. We find that on the noted stalk threonines, CD8␤ harbors a sialic acid linked to a core 1 disaccharide that lacks a N-acetyllactosamine, indicating an absence of both core 2 O-glycans and elongated core 1 glycans, which might alter lateral mobility of CD8␣␤ in the plasma membrane by virtue of larger hydrodynamic radii. Consistent with this view, we observed no differences in the distribution of CD8␣␤ co-receptors on the surface of DP and CD8 SP thymocytes (10). Furthermore, CD8␤ remains constitutively concentrated in cholesterol-sphingolipid-rich plasma membrane microdomains due, at least in part, to CD8␤ cytoplasmic tail palmitoylation (28).
The structure of O-glycan adducts revealed herein also limits the likelihood that galectins are operating to cross-link galactose residues on neighboring CD8␤ stalk adducts. Galectin-1 has been implicated in thymocyte apoptosis through the recognition of core 2 O-glycans on CD43 and CD45 (27,29). Galectin-3 binding to ␤1-6-branched lactosamine chains produced by the Mgat5 gene on TCR N-glycans is reported to inhibit T-cell activation, perhaps by altering TCR clustering (30). Most mammalian galectins bind preferentially to galactose on polylactosamine, although some may bind to other galactose linkages (31,32). Detailed site-specific assignment of Sia adducts on the CD8␤ stalk confirms that the addition of sialic acid per se modulates CD8␣␤ co-receptor function.
Our findings show that the genetically programmed alteration of CD8␤ glycosylation during thymocyte differentiation from immature DP to mature SP stages is restricted to the O-glycans without concurrent changes in the N-linked structures. These O-linked sites (Thr-120, Thr-121, Thr-124, Thr-127, and Thr-128) localize to a segment of the CD8 stalk immediately abutting the CD8␤ Ig-like domain. O-Linked glycans are attached to none of the 14 other serine or threonine residues in the examined tryptic fragments. Three of the five threonines (Thr-120, Thr-124, and Thr-128) are conserved in all of the CD8␤ homologues sequenced to date, residing within or adjacent to the lysine-rich segment that is unique to the CD8␤ stalk (10). In view of both the weak association between CD8␣ and CD8␤ head regions (10), an uncharacteristic feature for Ig-like domain heterodimers and the almost certain requirement for participation of CD8␣ and CD8␤ CDR-like loops in the binding to the MHCI ␣ 3 domain (by extension from crystallographic analysis of two CD8␣␣⅐pMHCI complexes) (33,34), sialylation in this specific region may impact significantly on CD8␣␤ binding to MHCI. The addition of sialic acid to the CD8␤ stalk could facilitate neutralization of positive charges on the adjacent stalk lysine residues (Lys-125, Lys-126, Lys-130, and Lys-132 to Lys-135), probably permitting the stalk to assume a retracted rather than fully extended configuration or resulting in other conformational changes. By altering CD8␣␤ domain-domain association and/or disposition of the CD8 globular headpiece relative to the cell surface, CD8␤ stalk O-glycans create a molecular switch regulating MHCI binding. That sialic acid addition to core 1 O-glycans during thymic ontogeny is a conserved feature of vertebrate development (2) with CD8␤ representing a major thymic PNA-binding protein (Fig. 1d) underscores the essential nature of this molecular switch. Additional chemical details regarding the dynamic glycobiology of CD8 will be important, not only for understanding the coreceptor function of thymocytes but that of naive, memory and effector CD8 peripheral T-cells.