Loss of Effector Function of Human Cytolytic T Lymphocytes Is Accompanied by Major Alterations in N- and O-Glycosylation*

Background: Most human tumors escape the immune system because of the immunosuppressive tumor microenvironment. Results: N-Acetyllactosamine units, the ligands for galectin-3, and branching on N-linked glycans are increased in the recently activated cytolytic T lymphocytes. Conclusion: The increased galectin-3-lattices decrease the motility of key surface glycoproteins. Significance: Loss of effector functions on cytolytic T lymphocytes may be linked to reduced motility of surface glycoproteins. Most human tumors are not eliminated by the immune system, and therapeutic vaccination shows poor results, a fact that can be explained at least partially by an immunosuppressive tumor microenvironment that is abundant in galectin-3. On cytolytic T lymphocyte (CTL) clones, maintained in culture by regular stimulation, recently activated CTLs present low effector functions. However, these functions are restored after a short treatment with LacNAc. The latter, which is in agreement with the glycoprotein-galectin lattice concept involving reduced motility, poses the question why galectin-3 ligands improve effector functions. We employed ultrasensitive MALDI-TOF-MS on resting and recently activated CTL clones combined with various glycosidase digestions and GC-MS linkage analyses. Our results showed that compared with the resting CTLs, the N-glycans of the recently activated CTLs consisted of (i) larger LacNAc oligomers of which a significant portion was longer than four-units and (ii) more multi-antennary structures. Interestingly, our results showed that the poly-LacNAc appeared to be equally distributed on all available N-glycan branches and not selectively enriched on a specific branch. The above structural alterations in the recently activated CTLs are expected to increase the galectin-3-LacNAc lattices and multivalent interactions and, therefore, reduce the motility of surface glycoproteins, such as the T-cell receptor. These findings suggest that the loss of effector functions on CTLs may be linked to reduced motility of surface glycoproteins. In addition, our results showed that recently activated CTLs had a reduced abundance of NeuAcα2,6-linked N-glycans and an increased abundance of disialylated core 1 and monosialylated core 2 O-glycan structures.

Human tumors are usually not spontaneously eliminated by the immune system, and therapeutic vaccination of cancer patients with defined antigens is followed by tumor regressions only in a small minority of patients (1,2). The poor vaccination effectiveness could be explained by an immunosuppressive tumor microenvironment. Because T cells that do infiltrate tumor metastases have an impaired ability to lyse target cells or to secrete cytokines when they are stimulated ex vivo, i.e. tested on untreated and uncultured samples (3)(4)(5)(6)(7), many researchers are trying to decipher the underlying immunosuppressive mechanisms.
We suggested that extracellular galectin-3, abundant in solid tumor and carcinoma ascites, impairs human tumor-infiltrating T lymphocyte (TIL) 2 function (8). The motility of surface molecules, such as the T-cell receptor (TCR), which are necessary for T cell activation and function, can be reduced because they are trapped in a lattice of glycoproteins clustered by extracellular galectin-3 (9). Several of our results support this hypothesis. First, human tumor-infiltrating CD8 ϩ TILs from diverse histological origins, in contrast with CD8 ϩ blood cells, have an impaired ability to secrete IFN-␥ upon ex vivo stimulation (6). We documented by FRET the physical dissociation of TCR and CD8 on these TILs. Second, TILs harbor surface galectin-3, which can be detached by treatment of CD8 ϩ TIL with an anti-galectin-3 antibody. A 2-h treatment of TIL with this antibody followed by overnight stimulation with beads coated with anti-CD3 antibodies, which mimic an antigenic stimulation, boosted their secretion of IFN-␥ (6). Third, two galectin ligands, N-acetyllactosamine (Gal␤1,4GlcNAc; LacNAc) and GCS-100, a modified citrus pectin, detached galectin-3 from TIL, promoted co-localization of TCR and CD8 co-receptor, and increased their cytotoxicity and their ability to secrete IFN-␥ upon stimulation (6). It has to be noted that both galectin ligands are also able to detach galectin-1 from cells.
We also analyzed the effect of galectin ligand LacNAc on cytolytic T lymphocyte (CTL) clones maintained in culture by regular antigenic stimulation, i.e. every 2 weeks. Compared with resting CTL collected 14 days after the last stimulation, recently activated CTL that were collected 4 days after stimulation secreted low levels of IFN-␥ upon a further antigenic stimulation (8). We documented by FRET and by confocal microscopy the physical dissociation of TCR and CD8 at the surface of recently activated lymphocytes, whereas TCR and CD8 were co-localized on most of the resting cells. In agreement with the hypothesis of a glycoprotein-galectin lattice on recently activated CTL, a short treatment of these CTL with LacNAc increased the proximity between TCR with the CD8 and restored their ability to secrete high levels of IFN-␥.
Experiments in mice models have revealed the influence of N-glycans on T cell activation. First, T cells from mice lacking N-acetylglucosaminyltransferase V (Mgat5), an enzyme essential for the generation of complex tetra-antennary N-glycans, have a lower activation threshold than T cells from wild-type mice. The tetra-antennary N-glycans can carry poly-LacNAc chains, the natural ligands of galectin-3. Mgat5-deficient mice showed evidence of immune dysfunction by increased signs of proliferative glomerulonephritis and were more susceptible to develop autoimmune encephalomyelitis (9). Second, T cells from mice lacking ␤1,3-N-acetylglucosaminyltransferase 2 have reduced poly-LacNAc on their N-glycans and were much more sensitive to activation upon stimulation with anti-CD3 coated beads (10). Finally, Kuball et al. (11) introduced mutations in a TCR coding sequence that destroy N-glycosylation sites. These modified sequences were grafted in polyclonal T cells, which acquire a better avidity for the antigen recognized by the grafted TCR than T cells carrying the wild-type TCR. The authors proposed that glycosylation affects the flexibility, movement, and interactions of surface molecules, but it cannot be excluded that these mutations also affect the number of galectin ligands and the possibility to form TCR-galectin lattices.
Why do galectin ligands improve effector functions of human TIL and CTL clones? Our working hypothesis is that, as a consequence of antigen-stimulated activation of T cells, the expression of enzymes of the N-glycosylation pathway is modified, and this changes the structures of N-glycans exposed at the cell surface. In agreement with that hypothesis, when resting CTL were stimulated in the presence of swainsonine, an inhibitor of ␣-mannosidase II, which is involved in the N-glycosylation pathway, recently activated CTL collected at day 4 after stimulation compared with untreated cells showed a higher TCR-CD8 FRET efficiency and a higher ability to release IFN-␥ upon antigenic stimulation (8). We surmise that the recently activated TILs compared with resting T cells harbor a set of glycans that are either more numerous or better ligands for galectin-3 and possibly for other galectins, such as galectin-1. In vivo, tumor cells, macrophages, and possibly activated T cells can secrete galectin-3. In in vitro cultures of CTL clones, galectin-3 can be secreted by the tumor cells used as stimulators, the immortalized B cells used as feeder cells, and possibly by the activated T cells themselves. Indeed, expression of galectin-3 is increased after murine T cell activation by the mitogen concanavalin A. Also, galectin-3 secretion is enhanced in response to calcium ionophore when T cells are previously activated (12).
N-Glycans harbored by murine splenocyte T cells have been previously compared with those found at the surface of splenocyte T cells, which were activated by anti-CD3 antibody and collected 3 days later. Activation resulted in a decreased expression of the sialyltransferase ST6Gal1 and increased expression of the galactosyltranferase ␣1,3GalT (13). Accordingly, activation resulted in reduction of sialylated biantennary N-glycans carrying the terminal NeuGc␣2,6Gal sequence and a corresponding increase in glycans carrying the Gal␣1,3Gal sequence. Murine T cells, however, are not good models for potential galectin-mediated processes in the human immune system because they are poor in tri-and tetra-antennary poly-LacNAc glycans (14).
Here, we characterized global changes in N-and O-glycans of two human CTL clones that were analyzed either in a resting state or 4 days after antigenic stimulation. Applying ultra-sensitive matrix-assisted laser desorption-ionization time-of-flight mass spectrometry (MALDI-TOF-MS) combined with various glycosidase digestions and gas chromatography (GC)-MS linkage analyses, we present direct evidence that galectin-3 ligands in recently activated human CTLs, compared with the resting state CTLs, are increased. N-Glycans of recently activated CTLs exhibited poly-LacNAc chains of which a portion was found longer than four LacNAc units. Moreover, recently activated CTL clones present more branched (multi-antennary) N-glycans. Differences in the non-reducing N-glycan terminal epitopes and O-glycan structures are also discussed.
Cell Pellets-Cells were collected and washed. Dead cells were removed by a depletion selection strategy using the Dead Cell Removal Kit (AutoMACS system, Miltenyi Biotec). Cells were then washed, and the remaining Epstein-Barr virus immortalized B cells were removed by a positive selection strategy using magnetic beads coated with anti-CD19 (AutoMACS system). Cells were finally washed in PBS, dry-pelleted, and stored at Ϫ80°C.
Mass Spectrometry-MS data were acquired using either a Voyager-DE STR MALDI-TOF or a 4800 MALDI-TOF/TOF (Applied Biosystems, Darmstadt, Germany) mass spectrometer. MS/MS data were acquired using a 4800 MALDI-TOF/ TOF mass spectrometer. Permethylated samples were dissolved in 10 l of methanol, and 1 l of dissolved sample was premixed with 1 l of matrix (20 mg/ml 2,5-dihydroxybenzoic acid in 70% (v/v) aqueous methanol), spotted onto a target plate, and dried under vacuum. For the MS/MS studies the collision energy was set to 1 kV, and argon was used as collision gas. The 4700 Calibration standard kit, calmix (Applied Biosystems), was used as the external calibrant for the MS mode, and [Glu1] fibrinopeptide B human (Sigma) was used as an external calibrant for the MS/MS mode.
Analyses of MALDI Data-The MS and MS/MS data were processed using Data Explorer 4.9 Software (Applied Biosystems). The processed spectra were subjected to manual assignment and annotation with the aid of a glycobioinformatics tool, GlycoWorkBench (18). The proposed assignments for the selected peaks were based on 12 C isotopic composition together with knowledge of the biosynthetic pathways. The proposed structures were then confirmed by data obtained from MS/MS and linkage analysis experiments.
GC-MS Linkage Analysis-Partially methylated alditol acetates were prepared as previously described (15). Linkage analysis of partially methylated alditol acetates was performed on a PerkinElmer Life Sciences Clarus 500 instrument fitted with a RTX-5 fused silica capillary column (30 m ϫ 0.32 mm inner diameter; Restek Corp.). The sample was dissolved in ϳ20 -50 l of hexanes and injected manually (2-3 l). Injector temperature was set at 250°C. Partially methylated alditol acetates were eluted with the following linear gradient oven. Initially the oven temperature was set at 65°C for 1 min, heated to 290°C at a rate of 8°C per min, held at 290°C for 5 min, and finally heated to 300°C at a rate of 10°C per min.

RESULTS
CTL clones A10 and F3.2 were both directed against MAGE-3 peptide EVDPIGHLY presented by HLA-A1. They were expanded in vitro by stimulation every 14 days using peptide-pulsed HLA-A1 cells. To assess the effector capability of the CTL at various points of the stimulation cycle, cells were collected at either 4 days (activated) or 14 days (resting) after antigenic stimulation and co-culturing with peptide-pulsed HLA-A1 cells. Amounts of IFN-␥ secreted by the CTL were estimated by ELISA in the supernatant of overnight co-cultures. Compared with CTL collected on day 0, CTL collected on day 4 secreted lower levels of IFN-␥ upon re-stimulation ( Fig. 1 and supplemental Fig. S1). But when recently activated CTLs, collected on day 4, were treated for 2 h with galectin ligand LacNAc, their ability to secrete high levels of IFN-␥ upon restimulation was restored. On the contrary, LacNAc treatment did not affect IFN-␥ secretion on resting CTL clones (supplemental Fig. S2). Extracellular galectin-1 and galectin-3 were both detected at the surface of recently activated CTL. The latter CTL contains more galectin-1 than the resting CTL. No detectable galectin-3 was found on the resting CTL. A treatment of 2 h with LacNAc on the recently activated CTL partially detached both galectins (supplemental Fig. S3). To analyze the N-and O-glycans at the cell surface of resting and activated CTL, cells were collected at the indicated time, dead cells and stimulator cells were removed using a depletion strategy, and CTL were subsequently pelleted and frozen at Ϫ80°C.
Glycomic Analysis of CTLs Revealed Differences on N-and O-Glycans-The N-glycomic profile of resting CTL clone A10 consisted of high mannose and complex structures ( Fig. 2A). Compositional and MALDI-TOF-TOF MS/MS analyses suggested that the latter were mainly core fucosylated and nonbisected, and corresponded to bi-, tri-, and tetra-antennary N-glycan structures occasionally extended with LacNAc repeats. The non-reducing ends of these glycans either remained unmodified (m/z 2244, 2489, 2938, 4490) or were further modified, mainly with terminal NeuAc residues (m/z 2605, 3416, 4402, 5300). The sLe x terminal epitope was also detected in bi-antennary (m/z 3140, 3314) and tri-antennary (m/z 3951 and 4125) N-glycan structures. However, it was of low abundance compared with other terminal structures.
The N-glycomic profile of recently activated CTL clone A10 exhibited similar N-glycans to the resting state CTLs but with the following differences being observed (Fig. 2B). (i) N-Glycans with extended poly-LacNAc structures were of higher abundance compared with the resting state, resulting in higher molecular weight N-glycan structures reaching up to 7000 mass units; (ii) sLe x -terminated glycans were of higher abundance. For example, comparing the ratios of m/z 3140 with m/z 2966 on both the resting and recently activated states (Fig. 2, A and  B), it was concluded that the sLe x structures in the recently activated state were of higher abundance. In addition, the sLe x containing N-glycans were overlapped with other most abundant structures. For example, the tri-antennary sLe x structure detected at m/z 3951 (Fig. 2B) was observed as a minor peak in the m/z 3953-ion cluster that corresponded to a monosialylated core fucosylated N-glycan.
The O-glycomic profile of the resting CTL clone A10 mainly exhibited the monosialylated core 1 sequence (m/z 895) together with a tiny amount of a non-sialylated core 2 (m/z 983) structure (Fig. 3A). Recently activated A10 CTLs (Fig. 3B) con-tained, in addition to the core 1 O-glycan at m/z 895, disialylated core 1 O-glycans (m/z 1256) and a higher abundance of core 2 O-glycans, either unmodified (m/z 983) or terminated with one NeuAc residue (m/z 1344) or with a Le x terminal epitope (m/z 1157).
The glycomic profiles of resting and recently activated CTL clone F3.2 were similar to the CTL clone A10 structures with complex N-glycans extended with LacNAc repeats and elaborated with the same terminal epitopes, whereas O-glycans consisted of core 1 and core 2 structures terminated with NeuAc, Le xv and minor abundances of sLe x terminal epitopes (supplemental Figs. S4 and S5).
Endo-␤-galactosidase Digestion Revealed Differences in Relative Abundance of Various Terminal Epitopes and Polylactosamine-type N-Glycans-The N-glycome of CTL clones A10 and F3.2 suggested differences in the relative abundance of the terminal epitopes and LacNAc repeats. Therefore, we performed an endo-␤-galactosidase digestion on the N-glycans of both CTLs (supplemental Fig. S6) with a 2-fold aim: (i) to verify whether any of the N-glycans of the aforementioned CTLs contained linear polylactosamine chains and (ii) to characterize and reveal any potential differences in the relative abundances on the non-reducing terminal epitopes between resting and recently activated CTLs. The existence of the above should be revealed by the new molecular ions in the low mass region corresponding to antennae fragments released by the enzyme. The low mass permethylated products of the CTL clone A10 in both states (resting and recently activated) are depicted in Fig. 4. Disaccharides released from within the polylactosamine chains gave the signal at m/z 518 corresponding to the GlcNAc-Gal sequence (Fig. 4, A and B). The molecular ions at m/z 722, 896, 1083, and 1257 were consistent with the terminal sequences Gal-GlcNAc-Gal, Gal-(Fuc)GlcNAc-Gal, NeuAc-Gal-GlcNAc-Gal, and NeuAc-Gal-(Fuc)GlcNAc-Gal, respectively, which were predicted digestion products of uncapped, fucosylated, sialylated, and sLe x polylactosamine chains, respectively. Evidence for the m/z 896 and 1257 components being terminated by Le x and sLe x epitopes was provided by MALDI-TOF-TOF MS/MS that yielded major fragments at m/z 690 and 1051, respectively (data not shown). Molecular ions at m/z 967, 1141, 1171. and 1345 were assigned to pauci-mannose N-glycan structures Man 2 GlcNAc 2 , Man 2 GlcNAc-(Fuc)GlcNAc, To assess effector functions, 10,000 CTL were collected either at the resting state (Ϯ2 weeks after stimulation) or in an activated state (4 days after stimulation). CTL were incubated (or not) for 2 h with LacNAc (2 mM) and subsequently stimulated with peptide-pulsed HLA-A1 cells. IFN-␥ was measured by ELISA in the supernatant of overnight cocultures.
Analysis of the relative abundances of the above terminal epitopes revealed that compared with the resting state, the recently activated CTL clone A10 had higher levels of unmodified, Le x , and sLe x terminal epitopes and lower levels of sialylated terminal epitopes. Summing up the relative abundances of all the terminal epitopes at m/z 722, 896, 1083, and 1257, we found that compared with the resting state, recently activated CTL clone A10 was characterized by a decrease in the abundance of sialylated LacNAc repeats at m/z 1083 (ϳ24%) and an increase in the abundance of unmodified LacNAcs at m/z 722 (ϳ10%). Endo-␤-galactosidase digestion on CTL clone F3.2 revealed similar abundances to the CTL clone A10 fragments in both states (supplemental Fig. S8). Uncapped, Le x , sialylated, and sLe x polylactosamine chains characterized the resting and recently activated CTL clone F3.2. Again, compared with the resting state, recently activated CTL clone F3.2 was characterized by a decrease in the abundance of sialylated LacNAc repeats (ϳ36%) and an increase in the abundance of unmodified LacNAcs (ϳ196%). The reduction in glycan complexity on both CTL clones A10 and F3.2 afforded by the endo-␤-galacto-sidase digestion revealed additional signals that corresponded to N-glycans terminated with multiples of sLe x epitopes (supplemental Fig. S6). These signals existed in the original glycomic profiles but were in minor abundance, and the isotopic cluster of more abundant ions partly masked them. For example, the ions at m/z 4761 and 4935 corresponded to tetra-antennary N-glycans with a single and a double sLe x epitope (supplemental Fig. S6, A and B).
Taken together, the endo-␤-galactosidase digestion data on N-glycans of both CTL clones A10 and F3.2 confirmed the presence of polylactosamine chains that remained unmodified or were terminated with Le x , sialyl, and sLe x epitopes. The data also suggested that, compared with the resting state, in the recently activated state there was an increase in the relative abundance of poly-LacNAc and a decrease in the relative abundance of sialylated terminal epitopes.
Linkage Analysis Revealed Differences in Polylactosaminetype Abundances-The N-glycans of both the resting and recently activated CTLs contain linear LacNAc units whose abundances appeared to be modified in the resting-activation CTLs cycle. Therefore, to express the percent difference of the polylactosamine-type chains between the resting and recently activated CTLs, we performed GC-MS linkage analysis. Comparing the relative abundances of the 3-linked galactose and 4-linked GlcNAc residues (that together make-up the LacNAc unit in poly-LacNAc) between resting and recently activated CTLs, we are able to relatively quantify the different abundance of the LacNAc units between the aforementioned CTLs. Linkage analysis was performed on desialylated N-glycans after digestion with broad specificity sialidase-A. This is because ␣2,3-linked NeuAc residues potentially could be linked to galactose residues and supplementing the 3-linked galactose from the polylactosamine-type chains. Results from linkage analysis from both CTLs clones are depicted in Table 1.
Recently activated CTL clone A10 presented a 121 and 344% increase of the 3-linked galactose and 4-linked GlcNAc abundance, respectively, compared with resting CTLs, whereas for the CTL clone F3.2 the increase was 48 and 69%, respectively. These data show that in both CTLs, the N-glycans in the recently activated state contained more LacNAc chains than in the resting state. Additional linkage data confirmed structural differences between the resting and recently activated CTLs. For example, the increase of 3,4-GlcNAc residues on the recently activated CTLs was in accordance with the increase of sLe x epitopes. This increase was 363 and 29% that for the CTL clones A10 and F3.2, respectively (Table 1).

Endo-␤-galactosidase Digestion Revealed Differences in
Multi-branched N-Glycans-We next investigated whether there was a difference in the abundance of complex bi-, tri-, and tetra-antennary N-glycans between the resting and recently activated CTLs that carry LacNAc repeats. Endo-␤-galactosidase digestion on resting CTL clone A10 showed that the most abundant N-glycans carrying LacNAc extensions corresponded to bi-antennary (m/z 1836) and to a less extent to tri-(m/z 2081 and 2285) and tetra-antennary (m/z 2326) N-glycan structures (Fig. 5A). On the contrary, recently activated CTL clone A10 showed a decrease in the relative abundance of biantennary structures and an increase in the abundance of triand tetra-antennary N-glycans (Fig. 5B). Summing-up the relative abundances of the ions at m/z 1836, 2081, 2285, 2326, and 2530, we found that compared with the resting state, the triand tetra-antennary N-glycans (m/z 2081, 2285, 2326, and 2530) on the recently activated clone A10 were increased by ϳ89%. The same decrease of the bi-antennary and increase of tri-and tetra-antennary N-glycans was observed when resting and recently activated CTLs clone F3.2 were compared  (supplemental Fig. S9). By similar calculations we found that the tri-and tetra-antennary N-glycans on the recently activated clone F3.2 were increased by ϳ90%. Taken together, the data suggest that recently activated CTLs contain more branched N-glycans that carry LacNAc chains than the resting CTLs.
More Polylactosamine Chains Are Found on Recently Activated CTLs-All previous experiments showed that the N-glycans exist as multibranched structures extended with LacNAc repeats and terminated with various epitopes and that all these are modified in the resting-activation CTLs cycle. However, none of above experiments gave a cumulative indication on the length of these LacNAcs repeats. This was an important factor to consider because the binding affinity of galectin-3 is expected to be affected by the length of these LacNAc repeats as well as by the relative abundance of these terminal epitopes. Therefore, we determined the minimum length of the LacNAc chains on the resting and recently activated CTL clone F3.2 and additionally where these poly-LacNAcs were mainly formed, i.e. on bi-, tri-, or tetra-antennary N-glycans (Fig. 6).
We removed all fucose and sialic acid residues substituting the antennae by hydrofluoric acid treatment and by a nonspecific sialidase enzymatic digestion, respectively, of the N-glycans of the CTL clone F3.2 leaving only the core-fucose and the LacNAc extensions (Fig. 6A, upper structure). Then, performing two cycles of enzymatic digestions with subsequent incuba-tions of ␤-galactosidase and ␤-N-acetylhexosaminidase, we reduced the N-glycan LacNAc chains of the CTL clone F3.2 by 2 LacNAc repeats. Finally, we incubated the resultant N-glycans with endo-␤-galactosidase (Fig. 6A, upper structure). The new signals appearing in the spectra or the ones that were greatly enhanced corresponded to N-glycans deriving from the above process, and their structures depended on the number of LacNAc repeats originally present in each branch (Fig. 6A,  lower structure).
According to this procedure, most of the N-glycan structures detected on the resting and recently activated CTL clone F3.2 (supplemental Fig. S4, A and B) should be fully digested to the basic N-glycan core structure Man 3 GlcNAc(ϮFuc)GlcNAc (see Fig. 6A). Indeed, ions detected at m/z 1171 and 1345 corresponded to all complex bi-, tri-, and tetra-antennary N-glycans that were extended with up to 2 LacNAc units that were terminated with Le x , sialic acid, and sLe x epitopes (Fig. 6B). Ions at m/z 1416 and 1590 corresponded to complex structures where one branch was extended with 4 LacNAc repeats or more. Ions detected at m/z 1661 and 1836 corresponded to N-glycans that had two branches that originally contained 4 LacNAc repeats or more. The ions detected at m/z 1866, 2111, and 2285 had originally one branch extended with three LacNAc repeats and one or two branches (depending if they were bisected or not) with four LacNAc repeats or more (Fig. 6,

CTL clones
Relative abundance a % Difference (on activated) b  A and B). The ion detected at m/z 2530 corresponded to an N-glycan with two or three branches extended with 4-LacNAcunits or more (depending if it was bisected or not), although the ion at m/z 2775 corresponded to a bisected tetra-antennary N-glycan that had one branch extended with 3 LacNAc repeats whereas the rest were at least 4 LacNAc units long. Finally, the ions at m/z 1906, 2081, 2326, and 2571 corresponded to bi-, tri-, and tetra-antennary N-glycans that had all of their branches extended with 4 LacNAc repeats or more. The above data suggest that a part of the resting and recently activated CTL clone F3.2 N-glycans had all their branches extended with at least four LacNAc repeats.

3-Gal 4-GlcNAc 3,4-GlcNAc 3-Gal 4-GlcNAc 3,4-GlcNAc
To express the percent difference of the poly-LacNAc chains between the resting and the recently activated CTLs, we calculated the relative abundance of all the N-glycans with more than one terminal GlcNAc and compared it with the summed relative abundance of the ions at m/z 1141 and 1345 that corresponded to N-glycans with two LacNAc units or less. The results showed that compared with the resting, recently activated CTLs contained ϳ30% more N-glycans with one or more terminal GlcNAc N-glycans. Therefore, the data suggest that recently activated CTLs contained N-glycans with more branches of at least four LacNAc units.
We then performed GC-MS linkage analysis of the above resulting N-glycans where any possible variations in the levels of 2,4-linked mannose and 2,6-linked mannose would directly positively correlate with the abundance of tri-or/and tetra-antennary N-glycans but specifically for those that were elongated with more than 4-LacNAc units. Our data showed that 2,4-linked mannose and 2,6-linked mannose increased by ϳ80 and 60% on the recently activated clone F3.2 CTLs (supplemental Table S1), suggesting that more branches of at least 4-Lac-NAc units were found on the recently activated CTLs than the resting CTLs. In addition, combining the data from the MALDI-TOF-MS which indicated that the most abundant structure corresponded to triple-terminated GlcNAc N-glycan (m/z 2081) and the data from the GC-MS linkage analysis that the 2,4-linked mannose residue was more abundant than the 2,6-linked mannose residue, we concluded that the most abundant poly-LacNAc N-glycan corresponded to the tri-antennary N-glycan structures of GlcNAc␤1-2(GlcNAc␤1-4)Man␤1-3.
Recently Activated CTLs Have Lower Abundance of ␣2,6-Linked NeuAc-Endo-␤-galactosidase digestion suggested that, compared with the resting state, there was a decrease in the sialylation of the N-glycans in the recently activated state of both CTLs. However, the previous experiments did not reveal any information concerning the linkage of the NeuAc residues. Therefore, we applied ␣2-3-specific sialidase digestion on resting and recently activated CTL clone A10 to relatively quantify the ratio of ␣2,3-NeuAc/␣2,6-NeuAc residues between these different conditions. Fig. 7 depicts the MALDI-TOF mass spectra of permethylated N-glycans of CTL clone A10 in resting and recently activated states after sialidase-S (␣2-3 specific) digestion. Compar-  MARCH 30, 2012 • VOLUME 287 • NUMBER 14

Glycomics on Human Cytolytic T Lymphocytes
ing the ratio of sialylation to the base core fucosylated bi-antennary N-glycan (m/z 2244), we found that resting state CTL clone A10 was characterized by a high abundance of mono-sialylated (m/z 2605, 2850, 3054, and 3299) and di-sialylated (m/z 2966) N-glycans (Fig. 7A). On the contrary, the recently activated CTL clone A10 exhibited a lower abundance of mono-and di-sialylated N-glycans (Fig. 7B). These data suggest that the recently activated CTL clone A10 has N-glycans with lower abundance of ␣2,6-linked NeuAc residues than the resting state. This ratio was found to be ϳ106% lower in the recently activated CTLs.
The abundance of ␣2,6-linked NeuAc is directly correlated with the abundance of the 6-linked galactose residue. Therefore, we performed GC-MS linkage analysis on both the resting and recently activated CTL clone A10 to detect any possible difference in the abundance of 6-linked galactose residues. The results suggested that recently activated CTL clones A10 contained 56% less 6-linked galactose residue compared with their resting state (data not shown). Taken altogether, the data suggested that recently activated CTLs contained less ␣2,6-linked NeuAc residues than their resting state CTLs.

DISCUSSION
Defining the underlying immunosuppressive mechanisms of how tumor microenvironment impairs TIL function is of great importance. The level of galectins, abundant in solid tumors, has been correlated with the aggressiveness of the tumor, and it has been hypothesized that they might modulate tumor progression and influence the disease outcome (19,20). This suggestion appears to hold true because cells highly regulate the expression of endogenous galectins and their ligands to regulate cellular functions, suggesting that exogenous galectins or modification in the expression of galectin ligands could lead to pathology (21). The results presented here have been obtained with clones and, therefore, are expected to be more rigorous than polyclonal splenocytes that may contain different cell types. Applying ultrasensitive MALDI-TOF-MS combined with various glycosidase digestions and GC-MS linkage analyses, we make the novel observation that the N-glycome of the recently activated human CD8 ϩ CTLs, when compared with the N-glycome of the resting CTLs (i) consists of longer LacNAc chains of which a portion are longer that four units ((LacNAc) 4 , poly-LacNAcs) and (ii) contains more branched (multi-antennary) N-glycans. These observations verify our working hypothesis that activation of human CTLs and TILs results in the expression of surface glycoproteins that exhibit higher abundance of galectin-3 ligands and possibly other galectin ligands such as galectin-1. This combined with the high abundance of extracellular galectin-3 found in tumor microenvironments could potentially explain the loss of TIL functions through a mechanism of reduced motility of surface molecules.
Galectin-3 is a ␤-galactoside binding lectin that binds to internal LacNAc repeats, and it is not severely affected by the terminal epitopes at the non-reducing end of long-LacNAcextended glycans (22,23). Therefore, binding to the long poly-LacNAc chains is not affected by the variations of the abundance of sialic acids, sLe x , and Le x epitopes found here. Thus, our data suggest that on the recently activated CTLs the increased abundance of LacNAc repeats increases the galectin- 3-LacNAc interactions and blocks the mobilization of various glycoproteins. Simultaneously, galectin-3 is a lectin that possesses multivalent carbohydrate binding and cross-linking activity (24). Surprisingly, on the recently activated CTLs, these poly-LacNAc extensions were found to be more abundant on tri-antennary N-glycans with GlcNAc␤1-2(GlcNAc␤1-4)-Man␤1-3(GlcNAc␤1-2Man␤1-6)Man branches and less abundant on tetra-antennary N-glycans, whereas resting CTLs exhibited reduced abundance of these tri-and tetra-antennary branches. Most interestingly, our data suggest that these poly-LacNAc repeats appeared to be mainly equally distributed on all available N-glycan branches and not selectively enriched on a specific N-glycan branch. The above suggest that recently activated human CTLs can form more multivalent galectin-3-LacNAc lattices, a fact that combined with the high levels of galectin-3 found in tumor microenvironments could further explain the reduced motility of surface glycoproteins.
Our results are consistent with previous mouse studies highlighting the importance of N-glycosylation and specifically how LacNAc abundance and branching could block T cell activation by regulating the mobility of key surfaces molecules, such the TCR (9). However, those data suggested the selective extension of LacNAc repeats on the ␤1,6-GlcNAc branch, something that has not been verified from our data. Considering the data obtained here, it is possible that variations of LacNAc extensions are taking place on all available branches of mouse N-glycans, but only variations of the ␤1,6-branch were detected by the Phaseolus vulgaris leukoagglutinin binding lectin (9,25). The possibility that mouse TCR is specifically decorated either with tri-antennary N-glycans of type GlcNAc␤1-2-(GlcNAc␤1-6)Man␤1-6Man or tetra-antennary N-glycans cannot be excluded. In such a case, a possible down-regulation of the Mgat5 gene would provoke a reduced abundance of the LacNAc "density," which is expected to decrease the galectin-3-LacNAc lattices.
A major surface glycoprotein that could be affected from the above galectin-3-LacNAc lattices is the TCR. In the recently activated human CTLs, characterized by the loss of effector functions and the absence of co-localization of TCR and CD8, the TCR was found in close proximity to galectin-3. A treatment of these CTL with galectin ligands (LacNAc) resulted within 2 h in TCR/CD8 co-localization, detaching the galectin-3 from the TCR, and was accompanied by restored functions (8). This suggests that galectin-3, which is involved in distancing the TCR from CD8 (8), could potentially regulate CTLs functions. Several lines of evidence support this concept, meaning that loss of effector functions on human CTLs could be associated with the galectin-3-LacNAc-TCR lattices. First, treatment of human TIL with anti-galectin-3 antibody detached galectin-3 from cells and boosted their functions (6). Second, human TCR ␣-chain, in contrast to ␤-chain, has been shown to be occupied exclusively by complex N-glycans (26). Third, a previous study showed that the removal of a N-glycosylation site from the human TCR ␣-chain resulted in improved sensitivity of a T cell response (functional avidity) (11). Finally and most importantly, a key step in T cell effector function is the phosphorylation of immunoreceptor tyrosine-based activation motifs within the CD3 complex, which in turn is depen-dent on LCK activation (27). However, a recent study has shown that LCK is already activated in resting human T cells, enough to initiate TCR signaling (28). Moreover, the expression-activation of LCK did not appear to depend on TCR or co-receptor expression-contact with pMHC. Hence, the above collectively suggest that the increased galectin-3-LacNAc lattices occurring in the recently activated CTLs and accompanied by the loss of effector functions could be linked, although not exclusively, to the de-localization of the TCR from the CD8 molecule. This in turn could potentially prevent the immunoreceptor tyrosine-based activation motif phosphorylation by keeping the TCR-CD3 complex distant from LCK, which is stably associated with the CD8 co-receptor (29) and, therefore, prevent further signaling. This concept has been already been shown in mouse studies. Mgat5-deficient mice after stimulation exhibited increased association of the Zap-70 kinase with immunoreceptor tyrosine-based activation motifs, compared with WT mice, exhibiting a negative regulation of galectin-3-LacNAc lattices and immunoreceptor tyrosine-based activation motif phosphorylation (9).
Our data showed that the changes in LacNAc abundance between the resting and recently activated states of clone F3.2 were less prominent than those of clone A10. For TILs present in a tumor microenvironment of a given concentration of galectins and activated by the same peptide-MHC complex, the ones with a smaller increase in LacNAc abundance (such as clone F3.2) would be expected to present a relatively higher motility of surface molecules such as the TCR compared with other TILs with a more pronounced increase in LacNAc abundance (such as of clone A10). This is because the reduced increase of LacNAc abundance should induce a reduced increase of galectin-3-LacNAc lattices and, therefore, should have a smaller impact on the motility of surface molecules. Hence, a reduced increase of LacNAc abundance in activated T cells would be expected to result in higher motility of surface molecules and possibly effector functions. This concept has already been proposed in a previous study, suggesting that either the peptide-MHC complexes that would be predicted to overcome the negative regulation by the galectin lattice or the reduced N-glycan branching that would reduce the interaction of TCR for galectin would lead to activation (30).
Our data showed that recently activated human CD8 ϩ CTL clones exhibited a lower abundance of terminal ␣2,6-linked NeuAc residues than resting CTLs. This fact is expected to affect galectin-1-glycoprotein binding interactions. Galectin-1 binds to terminal LacNAc units either in their unmodified state or when they are decorated with ␣2,3-linked NeuAc (31,32), but this binding is blocked by ␣2,6-linked NeuAc (23,33,34). Therefore, our data suggest that more galectin-1-ligand binding interactions are expected to be formed in recently activated CTLs than in the resting CTLs. Interactions of the TCR with galectin-1, instead of galectin-3, which may contribute to lattice formation and restrain TCR function, cannot be excluded either. Indeed, we detected extracellular galectin-1 in addition to galectin-3, and both have been detected in TILs (6). However, in a previous study the anti-galectin-3 monoclonal antibody B2C10, although being able to boost TIL function, seemed unable to detach galectin-1, suggesting that detaching galec-tin-3 from TIL was sufficient to restore function while not excluding the contribution of other galectins (6). Exactly how and if this could be related to loss of effector functions related to TCR on the recently activated human CTLs is currently unknown. Moreover, the possibility of additional molecules that contribute to the mechanism, such as the CD45, with galectin-1 and/or galectin-3, cannot be excluded either (35,36). In mouse, the decreased abundance of ␣2,6-linked sialic acids has been described in CD4 ϩ and CD8 ϩ T lymphocytes during activation, and it has been suggested that this leads to increased binding with CD22 (siglec-2) and/or galectin-1 (13).
MALDI-TOF-MS suggested that O-glycans of recently activated human CTLs exhibited increased abundance of disialylated core 1 and monosialylated core two structures than the resting CTLs. The increased abundance of core 2 O-glycans on recently activated CTLs is in accordance with previous studies showing increased abundance of core 2 O-glycans in human leukosialin after activation of human T cells (37). Exactly how and if changes in O-glycosylation could be related to loss of effector functions in the recently activated human CTLs is currently unknown. Noteworthy, when our CTLs were activated and cultured for 4 days in the presence of swainsonine, an inhibitor of mannosidase II, neither their functions nor the TCR-CD8 co-localization was lost (8). Previous studies have shown that changes in O-glycan structures could affect CD8 co-receptor signaling by either altering the conformation of its stalk region (38 -40) or by regulating its avidity (41). Further studies will show if O-glycosylation affects the loss of effector functions.
In conclusion, our results show that the N-glycome of the recently activated CTL clones after antigenic stimulation result in increased abundance/density of poly-LacNAc chains that is able to sustain galectin-3 lattices without excluding the participation of other galectins and reduce the motility of key surface molecules, such as the TCR that is accompanied by the loss of effector functions of TIL. The possibility that certain surface glycoproteins do not follow the alteration of glycosylation presented here cannot, however, be excluded either.