GCNT1-Mediated O-Glycosylation of the Sialomucin CD43 Is a Sensitive Indicator of Notch Signaling in Activated T Cells

This article is featured in In This Issue, p.1419

Notch signaling is an evolutionarily conserved juxtracrine signaling pathway. Interaction of Notch receptors (Notch1–4) with agonistic ligands of the Delta-like or Jagged families leads to proteolytic release of the Notch intracellular domain, followed by its translocation to the nucleus where it mediates the transcriptional activation of Notch target genes. T cell development requires Delta-like 4 (Dll4)/Notch1 interactions in the thymus. In addition, Notch has emerged as a critical regulator of mature CD4⁺ and CD8⁺ T cell activation, differentiation, and effector function in the periphery (1). T cells incapable of receiving Notch signals showed reduced acute protective functions in mouse models of intracellular bacterial (2), viral (3), fungal (4), and parasitic infection (5) as well as in selected tumor models (6, 7). Notch also drives pathogenic T cell functions in mouse models of T cell–mediated acute graft-versus-host disease (GVHD) (8–11), chronic GVHD (12), organ rejection (13–15), and multiple sclerosis (16, 17). Despite the multiple immunological effects of Notch signaling in T cells, it has remained impossible to prospectively identify individual cells experiencing active Notch signaling, a major limitation to studying the detailed impact and regulation of the Notch pathway in vivo.

In mouse models of allogeneic bone marrow transplantation (allo-BMT) after myeloablative conditioning, donor T cells received critical Delta-like 1 (Dll1)– and Dll4-mediated Notch signals within the first 48 h after transplant. The cellular niche providing Dll1/4 ligands was restricted to nonhematopoietic Ccl19-Cre⁺ fibroblastic stromal cells rather than hematopoietic cells in secondary lymphoid organs (10). Additionally, both Ccl19-Cre⁺ fibroblastic stromal cells (18) and CD11c⁺ classical dendritic cells (19) have been shown to provide Dll4 signals to drive T follicular helper cell differentiation. These findings highlight fine spatial and temporal regulation of Notch signaling induction to mature T cells in these contexts. However, outside of these observations, little is known about when and where T cells receive Notch signals and about the essential cellular sources of Notch ligands in a broader range of immune responses. Identifying a common specific and sensitive surface indicator of Notch signaling across different activated T cell subsets and immunologic contexts would aid in answering these important questions.

CD43 (sialophorin, leukosialin) is a transmembrane sialomucin with an extensively O-glycosylated extracellular domain that is highly expressed on T cells (20). Dynamic regulation of glycosyltransferases during T cell development and activation results in the expression of two CD43 glycoforms (21). Mature naive T cells express a 115-kDa glycoform characterized by core-1 O-glycans capped by sialic acid. During T cell activation, upregulated expression of Gcnt1 and its protein product, the core-2 GlcNAc transferase-1, occurs. In concert with reciprocal downregulation of St3gal1, which encodes the sialyltransferase that caps core-1 glycans, these changes generate an activation-associated 130-kDa glycoform of CD43 characterized by core-2 O-glycans (22). The switch from core-1 to core-2 O-glycosylation of surface glycoproteins, including CD43 and P-selectin glycoprotein ligand-1 (PSGL-1), increases the affinity of immune cells for P- and E-selectins, thus allowing initial rolling of activated (23) and memory T cell subsets (24, 25) on inflamed endothelium. Glycoform-specific Abs to CD43 enable the detection of unique glycosylated forms of CD43, correlating with the overall O-glycosylation status of surface proteins. The mAb S11 recognizes CD43 regardless of glycosylation (20), whereas 1B11 specifically recognizes the core-2 activation-associated O-glycoform (22). Others have used core-2 O-glycoform CD43 reactivity to identify recently activated effector, as opposed to memory CD8⁺ T cells in response to lymphocytic choriomeningitis virus (LCMV) infection (26). Furthermore, alloreactive T cells upregulate the core-2 O-glycoform of CD43 but not other glycoforms of CD43 during GVHD (27, 28). Regulation of Gcnt1 expression and core-2 O-glycosylation, in turn, is context dependent. Central memory CD8⁺ T cells upregulate Gcnt1 expression and subsequent CD43 core-2 O-glycoform reactivity in a TCR-independent but IL-15–dependent manner in vivo (24). In vitro studies have shown that naive CD4⁺ and CD8⁺ T cells require TCR signals in cooperation with diverse cytokines, such as IL-2 and IL-12, to drive Gcnt1 expression (29–31). However, IL-2 and IL-12 were dispensable in vivo for activated CD8⁺ T cells (23). Thus, our understanding of the essential inputs that regulate Gcnt1 expression and CD43 glycosylation remains limited.

Recently, we reported a transcriptional program activated by Notch signaling in alloantigen-specific CD4⁺ T cells in a mouse model of allo-BMT and GVHD (32). Notch signaling was required for inflammatory cytokine production and lethal T cell alloreactivity (8, 10, 33). Interestingly, induction of Gcnt1 mRNA depended on both alloantigenic stimulation and Notch signals. These findings suggested that core-2 O-glycosylation of CD43 and other surface proteins may serve as a new surface marker to indicate the receipt of functional Notch signals. To establish whether core-2 O-glycosylation of CD43 depended on Notch signals, we used flow cytometric analysis with glycoform-specific Abs to CD43. Blockade of the Dll1/4 Notch ligands decreased the core-2 O-glycosylation of CD43 in a dose-dependent manner, whereas overall CD43 expression remained stable or even increased slightly. Regulation was cell intrinsic, depended on canonical Notch signaling, and was observed across all recently activated T cell subsets. However, Gcnt1-deficient T cells with intact Notch still drove lethal GVHD, indicating that Notch signaling instructs a wider pathogenic program than core-2 O-glycosylation. Instead, core-2 O-glycosylation serves as a sensitive and simple flow cytometric indicator of Notch signaling in recently Ag-activated T cells. Core-2 O-glycosylation status successfully predicted whether T cells had received critical Notch signals in multiple models of allotransplantation, immunization, bacterial infection, and viral infection. Together, these data show that Notch signals are critical to drive Gcnt1-dependent core-2 O-glycosylation in recently activated T cells, although Gcnt1-dependent core-2 O-glycosylation is not required for GVHD lethality. We propose that the core-2 O-glycoform of CD43 can be used as a simple flow cytometric readout to map how Notch signaling regulates T cell activation, differentiation, and effector function, opening new avenues for a deeper understanding of Notch signaling’s contribution to immune responses.

C57BL/6 × BALB/c F1 (CBF1, H-2^b/d, Thy1.2), C57BL/6 (B6, H-2^b/b, Thy1.2), and C57BL/6-Thy1.1/2 (B6, H-2^b/b, Thy1.1/2) mice were bred at the University of Michigan and the University of Pennsylvania. BALB/cJ mice were purchased from The Jackson Laboratory. Mx1-Cre, Ccl19-Cre (34), CD4-Cre, Dll1^f/f, Dll4^f/f, Notch1^f/f, Notch2^f/f, and ROSA26^dnMAMLf alleles (abbreviated dnMAML) were described previously (8, 10). 4C Rag1^−/−TCR transgenic mice on the B6 background reactive to I-A^d were previously described (35). Gcnt1-deficient mice were previously described (36). OT-I TCR transgenic mice recognize the OVA peptide SIINFEKL and were maintained on the B6 background. E8I-Cre⁺Notch1^f/f Notch2^f/f lack Notch1/2 genes in CD8⁺ T cells, as described (2). For transplantation experiments, cohoused littermate controls were used for donors and recipients at the University of Michigan or the University of Pennsylvania, per protocols approved by the University of Pennsylvania’s Office of Regulatory Affairs and the University of Michigan’s Committee on Use and Care of Animals. For Listeria monocytogenes infection, experimental mice were housed at the Maisonneuve-Rosemont Hospital Research Center Facility and treated in accordance with the Canadian Council on Animal Care guidelines.

For transplantation, CBF1 recipients were irradiated using a [¹³⁷Cs] source with titrated doses: high-intensity myeloablative conditioning (11 Gy, split into two doses separated by 3 h), reduced-intensity nonmyeloablative conditioning (3 Gy), or no conditioning. Recipients were 8–16 wk old. Both female and male mice were used as recipients and equally distributed among experimental groups. T cell–depleted bone marrow (TCD BM) prepared with anti-Thy1.2 Abs and complement (Cedar Lane Laboratories) was injected i.v. with or without splenocytes and/or purified T cells, as described (9). Cell numbers infused for each experiment are noted in the text and figure legends. For experiments with 4C CD4⁺ T cells, T cells were purified with negative magnetic selection for CD4⁺ T cells (Stemcell Technologies) and labeled with eFluor 450 cell proliferation dye. For polyclonal experiments, wild-type, Gcnt1-deficient, or dnMAML T cells were purified with negative magnetic selection for T cells (Stemcell Technologies) and labeled with eFluor 450 cell proliferation dye (Invitrogen). For long-term experiments, mice received noted doses of splenocytes and/or lymph node cells with TCD BM. Clinical GVHD scores and weight changes were monitored at least once weekly, as described (37). In selected experiments, recipient mice received humanized IgG1 mAbs specific for Dll1 or Dll4 (5 mg/kg i.p.) (9). A human IgG1 Ab to HSV gD glycoprotein was used as an isotype control. Each Ab batch was tested for specificity and efficacy by assessing the loss of Dll4-dependent T cell progenitors (38) or Dll1-dependent marginal zone B cells in vivo (39).

The generation of bone marrow chimeras was previously described (10). Eight- to twelve-week-old CBF1 Mx1-Cre⁺ Dll1^f/f Dll4^f/f or Cre-negative littermate control donor mice received five i.p. injections of 50 μg poly(I:C) (Amersham Biosciences) every other day. Two weeks after the last dosage, bone marrow was harvested and transplanted into lethally irradiated (11 Gy) 8–12-wk-old CBF1 Ccl19-Cre⁺ Dll1^f/f Dll4^f/f or Cre-negative littermate controls. Mice were rested for 6 wk to generate four-way chimeras lacking Dll1/4 genetically in the hematopoietic compartment, Ccl19-Cre⁺ fibroblastic compartment, both, or neither. Mice were then subjected to an infusion of 60 × 10⁶ lymph node cells from MHC-mismatched C57BL/6 mice.

Bone marrow–derived dendritic cells (BMDCs) were generated with GM-CSF and IL-4 as previously described (2), matured with LPS (100 ng/ml; Sigma) plus the TLR7/8 agonist R848 (100 ng/ml; Sigma), and loaded with OVA_257–264 peptide (SIINFEKL, 2 μg/ml; Sigma). A total of 5 × 10⁵ dendritic cells (DCs) were i.v. injected. For experiments with OT-I cells, 5 × 10⁴ purified (Stemcell Technologies) CD45.1⁺ OT-I CD8⁺ T cells were adoptively transferred on the day before DC immunization. To assay anti-infectious responses, L. monocytogenes expressing OVA (Lm-OVA) was grown, as described (2). A sublethal dose of 2 × 10³ CFUs was i.v. injected. The endogenous CD8⁺ T cell response was monitored with K^b-OVA staining. To monitor the response to chronic viral infection, mice were infected with LCMV clone 13 (2 × 10⁶ PFU i.v.), and virus-specific CD8⁺ T cells were assayed at day 8 and day 30 postinfection with K^b-gp33 tetramer staining.

OP9 cell lines overexpressing Dll1 and Dll4 have been described before (40, 41). The night before the experiment, 1 × 10⁴ OP9 cells were seeded onto 96-well flat-bottom plates in 20% FCS plus αMEM. 4C wild-type CD4⁺ T cells and 4C dnMAML CD4⁺ T cells were purified, labeled with eFluor 450 proliferation dye, and plated in a ratio of 2:2:1 with BMDCs from polyinosinic-polycytidylic acid [poly(I:C)]–induced Mx1-Cre⁺ Dll1^f/f Dll4^f/f CBF1 mice in complete RPMI 1640 medium (10% FCS) plus 2-ME. Blocking Abs (anti-Dll1, anti-Dll4, both, or isotype control) were provided at a concentration of 25 μg/ml. Cells were harvested at day 4 postplating.

Single-cell suspensions of cells from spleen, mesenteric lymph nodes, or pooled peripheral lymph nodes (cervical, brachial, axial, inguinal) were prepared by physical disruption through 70-μm cell strainers. Single-cell suspensions were prepared from liver (12), intestinal lamina propria, and intestinal epithelium (8) as previously described.

The following Abs were from BioLegend: anti-CD4 (clone GK1.5), CD8α (clone 53-6.7), CD44 (clone IM7), CD25 (clone PC61), CD45.1 (clone A20), CD45.2 (clone 104), Thy1.1 (clone OX-7), Thy1.2 (clone 30-H12), H-2K^b (clone AF6-88.5), H-2K^d (clone SF1-1.1), IFN-γ (clone XMG1.2), IL-17A (clone TC11-18H10.1), pan-CD43 (clone S11), core-2 O-glycosylated CD43 (clone 1B11), KLRG1 (clone 2F1), and CD127/IL7Rα (clone A7R34). Anti-Foxp3 (clone FJK-16s) was from eBioscience. Anti-Vβ13 (clone MR12-3) was from BD Biosciences. SIINFEKL peptide loading on H-2K^b MHC and conjugation to PE was done as previously described (42). Nonviable cells were excluded from analysis with Zombie Aqua Fixable Viability Dye (BioLegend) or DAPI (Sigma-Aldrich). For intranuclear staining, Foxp3 Transcription Factor Staining Buffer (eBioscience) was used per manufacturer’s protocol. Assessment of intracellular T cell cytokine production after short ex vivo stimulus was previously described (10). Flow cytometric analysis was performed using a four-laser Fortessa (Becton Dickinson). Sorting of T cells was performed using a four-laser FACSAria II/III (Becton Dickinson). Alloreactive 4C T cells were defined as Thy1.1⁺ CD4⁺. Naive OT-I CD8⁺ T cells (CD8⁺CD44^lo) and day 3 OT-I effector CD8⁺ T cells (CD8⁺CD45.2⁺CD44^hi) were sorted on a BD FACSAria II.

Donor 4C Thy1.1⁺ CD4⁺ T cells were FACS purified directly into TRIzol Reagent (Invitrogen). Total RNA was extracted with phenol/chloroform and purified with RNeasy Micro Kit (QIAGEN). cDNA was generated with SuperScript II (Invitrogen) using oligo(dT) primers and subjected to quantitative PCR with TaqMan reagents (Applied Biosystems). Gene expression analysis was performed using the following primers: Gcnt1 (Mm02010556_s1), St3gal1 (Mm00501483_m1), and Hprt (Mm03024075_m1). Relative gene expression was determined using the ΔΔ cycle threshold method, with normalization to Hprt.

Sample size for in vivo mouse experiments was determined empirically based on prior experience of known effect sizes and variation and used to calculate power with “pwr” statistical package in R. All statistical tests were performed using Prism software (GraphPad Prism version 8). Unless otherwise noted, data with more than one group were analyzed for differences with one-way ANOVA or two-way ANOVA depending on the number of factors. If a factor significantly explained the variation of the data, multiple comparisons between groups were made with Tukey test, assuming α = 0.05, and adjusted p values were reported. The multiple comparisons performed are represented in the figures by lines between groups. Data with only two groups were compared with an unpaired, two-tailed Student t test. If variance between the two groups differed significantly by an F-test, Welch correction for t tests was applied. Survival curves were compared using a log-rank (Mantel–Cox) test. Unless otherwise noted, graphs were generated in GraphPad Prism and presented as mean ± one SD. Unless otherwise noted, adjusted p values for comparisons were reported as *p < 0.05, **p < 0.01, and ***p < 0.001.

Hallmarks and transcriptional consequences of Notch signaling have been investigated systematically in developing T cells and in T cell leukemia (43, 44) but only recently in mature T cells in the context of allo-BMT (32). To identify potential functional targets of Notch signals in Ag-activated T cells that can be used as indicators of Notch signaling, we built on our discovery that Notch signaling plays a critical role to drive T cell pathogenicity and GVHD after allo-BMT (8–11, 32, 33). To capture the effects of Notch signaling specifically in alloantigen-reactive T cells and not in bystander T cells, we studied the transcriptome of 4C CD4⁺ TCR transgenic T cells transplanted in syngeneic or allogeneic mouse recipients, with or without blocking Abs to Dlll/4 Notch ligands. In this model, 4C CD4⁺ T cells are activated by the host MHC class II I-A^d alloantigen, whereas intact Dll1/4-mediated signaling is critical to drive lethal GVHD (32). Transcriptomic profiling showed that the expression of canonical Notch target genes, such as Hes1, Dtx1, and Il2ra, was downregulated with pharmacologic Delta-like ligand blockade (32) (Fig. 1A). Interestingly, upregulation of Gcnt1 transcripts induced by allogeneic T cell activation was completely abrogated by Delta-like ligand blockade (32) (Fig. 1A). This was confirmed by quantitative PCR validation in a separate experiment (Fig. 1B), indicating that increased Gcnt1 mRNA expression required both alloantigenic stimulation and Notch signals. Conversely, Delta-like ligand blockade induced a trend for increased expression of St3gal1, which encodes the sialyltransferase that adds a terminal sialic acid cap on core-1 O-glycans, making them inaccessible for modification by core-2 glycosyltransferases (Supplemental Fig. 1A, 1B). Other core-2 GlcNAc transferases were not highly expressed in 4C alloreactive T cells (Supplemental Fig. 1C).

As others showed that Gcnt1 expression is critical for core 2 O-glycosylation of CD43 (22), we hypothesized that Notch signals would regulate cell surface expression of the core-2 glycoform of CD43 in recently activated T cells. We used a pan-CD43 (S11) and a core 2 O-glycoform–specific (1B11) anti-CD43 Ab (Fig. 1C) to evaluate if Notch signaling regulates core 2 O-glycosylation in 4C CD4⁺ T cells after transplantation into lethally irradiated (11 Gy) CBF1 recipients (H-2^b/d). By flow cytometric analysis, the proportion of 4C CD4⁺ T cells reactive with 1B11 decreased upon Dll1/4 inhibition in a dose-dependent manner (Fig. 1D, 1E). However, mean S11 reactivity, which reflects total CD43 levels, was not negatively affected by Notch blockade, instead increasing slightly with the highest dose of Dll1/4 inhibitors (Fig. 1F). This correlated with increased mRNA levels of Spn, the gene encoding CD43, with Notch blockade (Supplemental Fig. 1D). To correct for any overall changes in CD43 expression, the ratio of 1B11 to S11 reactivity was calculated in individual cells. The 1B11/S11 reactivity ratio was dependent on Delta-like Notch signaling (Fig. 1F, 1G). To test if this regulation depended on the transcriptional functions of Notch signaling, we transplanted CBF1 recipients with wild-type 4C cells or 4C cells expressing the pan-Notch inhibitor dnMAML, which blocks canonical RBPJk/MAML-dependent signaling downstream of all Notch ligands and receptors (45). Importantly 4C and 4C–dnMAML cells showed similar levels of alloantigen-specific activation as indicated by similar CD44 levels (Supplemental Fig. 1E, 1F). However, 4C–dnMAML cells had decreased 1B11 reactivity (Fig. 1H) and 1B11/S11 ratios (Fig. 1I) down to levels barely higher than isotype control staining levels. In contrast, cell surface CD25 in 4C–dnMAML cells, although lower than in wild-type 4C cells, remained above isotype control staining levels (Fig. 1J, 1K). This is consistent with our transcriptional profiling showing that Notch blockade decreased but did not eliminate all alloreactivity-induced Il2ra expression (Fig. 1A). To test which Delta-like ligand could drive 1B11 reactivity, we used an in vitro system in which Notch ligands were provided by OP9 stromal cell lines expressing either Dll1 (OP9-DL1) or Dll4 (OP9-DL4). Cocultured wild-type 4C and 4C–dnMAML T cells were plated at a 1:1 ratio on this stroma. Alloantigenic stimulus to 4C T cells was provided by BMDCs from CBF1 Mx-Cre⁺ Dll1^f/fDll4^f/f lacking Delta-like Notch ligands. When plated on either OP9-DL1 or OP9-DL4 stromal lines, wild-type 4C cells showed increased 1B11/S11 ratios, which was blocked by ligand-specific Abs (Fig. 1L). Thus, in concert with alloantigenic stimulus, either Dll1 or Dll4 could drive core-2 O-glycosylation of CD43 in vitro. Altogether, measuring 1B11 or the 1B11/S11 reactivity ratio quantified Notch signaling intensity in T cells through a simple flow cytometry–based assay of cell surface glycoproteins.

Next, we evaluated the abundance of CD43 core-2 O-glycoforms in multiple polyclonal donor-derived T cell subsets after allo-BMT with high-intensity conditioning (11 Gy). We transplanted 1 × 10⁶ TCD BM plus B6 (H2^b/b, Thy1.1/2) wild-type or B6–dnMAML T cells labeled with the eFluor 450 cell proliferation dye into MHC-mismatched CBF1 (H2^b/d, Thy1.2) hosts (Fig. 2A). 1B11 reactivity in proliferated eFluor 450 dilute CD4⁺ conventional T cells (Tconv) and regulatory T cells (Tregs) was significantly reduced among dnMAML T cell subsets from day 4 until day 14 posttransplant. 1B11 reactivity was also significantly reduced in dnMAML CD8⁺ T cells but to a lesser extent than in CD4⁺ T cells. Maximum differences occurred on day 4, suggesting higher dependence on Notch signaling at early time points for core-2 glycosylation (Fig. 2A, 2B). Notch-dependent regulation of core-2 glycosylation was seen in both secondary lymphoid organs (Fig. 2C) where Notch signals are first received but also in GVHD target tissues (Fig. 2C). These signals were imprinted early during T cell activation as a single dose of Dll1/4 blocking Abs at the time of transplant was sufficient to prevent upregulation of 1B11 reactivity but not when delayed to day 2 (Supplemental Fig. 2A–C). To establish if the regulation of 1B11 reactivity by Notch was cell intrinsic, we cotransplanted wild-type and Notch1/2-deficient T cells or wild-type and dnMAML T cells into MHC-mismatched CBF1 recipients. Both wild-type and Notch-deficient T cells proliferated similarly posttransplant (Supplemental Fig. 2D, 2E), as shown previously (8, 9). However, compared with cotransplanted wild-type T cells, both Notch1/2-deficient and dnMAML T cells had significantly diminished 1B11 reactivity (Fig. 2D, 2E). Furthermore, 1B11 reactivity predicted the ability of proliferated wild-type CD4⁺ T cells to produce IL-17 and IFN-γ (Fig. 2F) and the ability of CD8⁺ T cells to produce IFN-γ (Fig. 2G). Thus, across allogeneic polyclonal T cell subsets, core-2 O-glycosylation of CD43 requires cell-intrinsic canonical Notch signals and identifies alloreactive T cells with high cytokine-producing potential.

Upregulated Gcnt1 expression is critical for core-2 O-glycosylation of mucin-like glycoproteins including CD43 as well as other selectin ligands such as PSGL-1. In fact, core-2 modifications are required for PSGL-1 to bind P-selectins (46) and perhaps E-selectins (30, 47). Interestingly, P-selectin–deficient mice were protected from GVHD, but T cells from PSGL-1–deficient mice still caused GVHD (48). This suggests that core-2 O-glycosylation of other selectin ligands, such as CD43, may mediate trafficking of alloreactive T cells in GVHD target organs. Loss of Notch signaling also blunted trafficking of alloreactive T cells into target GVHD tissues, including the intestinal lamina propria in a model of acute GVHD (8) and skin in a model of chronic GVHD (12). Because Notch signaling regulates Gcnt1 expression and core-2 glycosylation of CD43 and likely other selectin ligands, we evaluated if Gcnt1 expression mediated trafficking to target tissues or other pathogenic effects of Notch signaling in T cells during GVHD. To compare the impact of Notch inhibition and Gcnt1 loss in T cells, we transplanted wild-type, dnMAML, or Gcnt1-deficient B6 T cells into CBF1 recipients. dnMAML CD4⁺ Tconv phenocopied Gcnt1^−/− T cells, whereas dnMAML Tregs and CD8⁺ T cells nearly fully phenocopied Gcnt1^−/− T cells in terms of 1B11 reactivity and 1B11/S11 ratio after allo-BMT (Fig. 3A, 3B). These effects were seen both in lymphoid organs and in target organs including liver, gastrointestinal (GI) epithelium, and lamina propria (Supplemental Fig. 3A, 3B). Thus, Notch signals were the dominant input driving Gcnt1 expression and core-2 O-glycosylation in this context. Consistent with previously published results, dnMAML CD4⁺ (Fig. 3C) and CD8⁺ (Fig. 3D) T cells had impaired accumulation in target tissues but not lymphoid tissues. However, Gcnt1^–/- T cells accumulated in similar numbers as wild-type T cells in both lymphoid organs and target tissues, except for the small intestine epithelium. Thus, although trafficking and/or accumulation of alloreactive T cells in liver or GI target tissues relies on Notch signals, these effects of Notch signaling are not driven by Gcnt1 expression and core-2 O-glycosylation. Furthermore, the production of inflammatory cytokines was blunted in dnMAML but not Gcnt1^−/−CD4⁺ (Fig. 3E, Supplemental Fig. 3C) and dnMAML but not Gcnt1^−/− CD8⁺ T cells (Fig. 3F, Supplemental Fig. 3D). Foxp3⁺ Tregs were also significantly expanded among dnMAML but not Gcnt1^−/− CD4⁺ T cells (Supplemental Fig. 3E). Consistent with these results, Gcnt1^−/− T cells transplanted into wild-type recipients mediated GVHD lethality and morbidity with similar if not increased kinetics as compared with wild-type T cells (Fig. 3G, solid lines). Notch appeared to drive GVHD pathogenicity independently of core-2 O-glycosylation as mice receiving Gcnt1^−/− T cells could be protected from GVHD by inactivating the Dll1/4 ligand genes in Ccl19-Cre⁺ host fibroblastic stromal cells (Fig. 3G, dotted lines), as shown previously with wild-type T cells (10). Together, these data show that Notch signals are crucial for Gcnt1-mediated core-2 O-glycosylation of CD43 in alloreactive T cell subsets. However, Notch can drive GVHD pathogenesis through mechanisms independent of Gcnt1.

Because Notch signals were necessary for cell-intrinsic upregulation of 1B11 reactivity, we hypothesized that 1B11 reactivity could be used as a surrogate marker of Notch signaling in activated T cells. As a corollary, 1B11 reactivity should predict which cellular subset is the critical source of Notch ligands that drive lethal T cell alloreactivity across multiple allo-BMT models. We previously reported that in fully MHC-mismatched allogeneic BMT with prior myeloablative conditioning, nonhematopoietic lymphoid tissue fibroblastic stromal cells lineage traced by the Ccl19-Cre transgene provided the critical Dll1/4 ligands to drive GVHD (10). However, nothing is known about the source of Notch ligands that drive lethal alloimmunity after transplantation with reduced-intensity conditioning (49), a clinically relevant scenario. Indeed, radiation-sensitive hematopoietic cells, eliminated in myeloablative models, may provide a critical source of Notch ligands to drive lethal T cell alloreactivity. To test this question, we used a “parent into F1” model of allo-BMT in which expression of parental alloantigens by the recipient causes tolerance to donor cells, allowing transplantation with titratable conditioning intensity (37). In a model with high-intensity radiation-based conditioning (11 Gy), mice that received dnMAML T cells were protected from GVHD, similarly to recipient mice lacking Dll1/4 in Ccl19-Cre⁺ fibroblastic stromal cells (Fig. 4A), in concordance with our previous results in a fully MHC-mismatched model (10). To test if 1B11 reactivity could predict this outcome, we cotransplanted wild-type and dnMAML T cells into Ccl19-Cre⁺Dll1^f/fDll4^f/f or littermate control mice. Wild-type recipients showed striking differences between cotransplanted dnMAML and wild-type CD4⁺ Tconv in 1B11 expression and 1B11/S11 ratios (Fig. 4B). In contrast, these differences disappeared in Ccl19-Cre⁺Dll1^f/fDll4^f/f recipients, indicating that no additional source of Notch ligands was available to CD4⁺ Tconv beyond the critical source in fibroblastic stromal cells (Fig. 4B, 4C). Alloreactive CD8⁺ T cells showed similar effects, with the vast majority of Notch inputs driving 1B11 reactivity from Ccl19-Cre⁺ cells (Fig. 4B, 4C). We next studied GVHD in a parent into F1 model with nonmyeloablative radiation-based conditioning (3 Gy). The differences in 1B11 reactivity and the 1B11/S11 ratio of cotransplanted wild-type and dnMAML T cells indicated that Notch signals were received by wild-type T cells in wild-type recipients. However, in Ccl19-Cre⁺Dll1^f/fDll4^f/f recipients, these differences in 1B11 reactivity disappeared, except for CD8⁺ T cells in skin-draining peripheral lymph nodes in which minor differences in 1B11 expression remained (Fig. 4D, 4E), suggesting another source of Notch ligands available to CD8⁺ T cells in these lymphoid organs. Consistent with these flow cytometric results, Ccl19-Cre⁺ Dll1^f/f Dll4^f/f recipient mice were protected from lethal GVHD (Fig. 4F). Thus, even after allotransplantation with reduced-intensity conditioning, nonhematopoietic fibroblastic stromal cells were the critical source of Notch ligands that drive GVHD.

In models with no prior conditioning, Notch also drives lethal alloimmunity, but the source of ligands remains unknown (8). To test the key cellular source of Notch ligands after allotransplantation with no prior conditioning, we created four-way bone marrow chimeric recipient mice with loss of Delta-like ligands in different cellular compartments. Wild-type or Ccl19-Cre⁺Dll1^f/fDll4^f/f CBF1 mice were transplanted after lethal irradiation with bone marrow from poly(I:C)-induced Mx-Cre⁺ Dll1^f/fDll4^f/f or Cre⁻ control CBF1 mice. Chimeras were allowed to reconstitute for 6 wk, generating recipients lacking Delta-like ligands in the hematopoietic compartment (Fig. 5A, recipient ii), fibroblastic stromal compartment (recipient iii), both (recipient iv), or neither (recipient i). These mice then received a parental donor lymphocyte infusion of cotransferred 5 × 10⁶ wild-type and 5 × 10⁶ dnMAML B6 T cells (H2^b/b) without prior conditioning, a model capable of triggering alloimmune-mediated hematopoietic failure in CBF1 recipients (H2^b/d) (Fig. 5A). To predict the critical cellular source(s) of Notch signaling, we examined 1B11 and S11 reactivity in these cotransferred T cells (Fig. 5B, 5C) at day 6 postinfusion. The presence of Dll1/4 in the hematopoietic compartment had no impact on CD4⁺ Tconv Notch-dependent 1B11 reactivity (Fig. 5B, 5C, recipient i versus recipient ii). However, loss of Dll1/4 ligands in the fibroblastic stromal compartment eliminated Notch-dependent 1B11 reactivity (Fig. 5B, 5C, recipient i versus recipient iii). Alloreactive CD8⁺ T cells also depended on a fibroblastic source of Dll1/4 to show increased Notch-dependent 1B11 reactivity (Fig. 5D, 5E). Thus, we hypothesized that a fibroblastic source of Delta-like Notch ligands would be critical to drive lethal alloreactivity. To test this hypothesis, we transplanted CBF1 recipients with no prior conditioning and with 5 × 10⁶ lymph node cells from wild-type or dnMAML parental B6 mice. In this model, lethal T cell alloreactivity depended on canonical Notch signaling and was rescued when the fibroblastic stromal source of Dll1/4 was eliminated (Fig. 5F). Thus, 1B11 reactivity predicted the functionally relevant source of Notch ligands, and fibroblastic stromal cells remained the essential source even in the absence of radiation conditioning.

To assess if Notch-dependent regulation of CD43 core-2 O-glycosylation applied to other types of Ag-activated T cells, we used a model of BMDC immunization pulsed with OVA peptide (DC-OVA). In this model, CD8⁺ T cell differentiation into KLRG1⁺IL7R⁻ short-lived effector cells (SLECs) depends on cell-intrinsic Notch signals (2, 3). However, the generation of KLRG1⁻IL7R⁻ early effector cells (EECs) and KLRG1⁻IL7R⁺ memory precursor effector cells (MPECs) remains intact. At the time of adoptive DC-OVA transfer, mice were treated with anti-Dll1/4 or isotype control Abs, as noted in Fig. 6A. On day 10, Ag-specific CD8⁺ T cells were assessed by K^b-OVA tetramer staining (Fig. 6A). The overall percentage (Fig. 6A) and total number (data not shown) of Ag-specific CD44⁺CD8⁺ T cells were not affected by Dll1/4 blockade. However, differentiation into SLECs was blunted by anti-Dll1/4 Abs (Fig. 6B). Consistent with the effects of Notch on core-2 O-glycosylation, 1B11 reactivity of OVA-specific CD8⁺ T cells depended on Dll1/4 Notch signals (Fig. 6C). The Notch-dependent 1B11 reactivity was not restricted to SLECs but was also seen in OVA-specific CD8⁺ T cell MPECs and EECs, suggesting that these cellular subsets have also received Notch signals (Fig. 6D). Concordant results were observed when the response of adoptively transferred OT-I cells to DC-OVA was examined (Fig. 6E–G). These data show that Notch regulates SLEC differentiation and 1B11 reactivity in polyclonal and monoclonal models of OVA responsiveness.

Compared with infection, DC-OVA immunization does not induce robust innate immune stimuli. Thus, it is possible that additional cytokine pathways (such as IL-12) that have been shown to drive core-2 O-glycosylation in vitro (31) may not be triggered in this model. To determine if Gcnt1 expression was also regulated by Notch during an anti-infectious response, we studied control versus Notch1/2-deficient OT-I CD8⁺ T cells in naive mice and on day 3 after a sublethal dose of Lm-OVA. Gcnt1 expression was induced after Lm-OVA infection in control but not in Notch-deficient OT-I cells (Fig. 7A). Next, we administered Lm-OVA to wild-type or E8I-Cre⁺ Notch1^f/f Notch2^f/f mice (lacking Notch1/2 in all CD8⁺ T cells). In parallel, we treated a group of wild-type mice with anti-Dll1/4 Abs. Eliminating Notch signals either through the loss of Notch1/2 receptors or through Ab-mediated Dll1/4 blockade did not impair or even slightly increased the overall expansion of Ag-specific CD8⁺ T cells after Lm-OVA infection (Fig. 7B). However, differentiation into SLECs depended on CD8⁺ T cell–intrinsic Notch signals, which was recapitulated by blocking Delta-like ligands (Fig. 7C). Consistently, 1B11 reactivity in bulk Ag-specific CD8⁺ (Fig. 7D) and fractionated SLEC, EEC, and MPEC subsets (Fig. 7E) depended on cell-intrinsic Notch signals and Delta-like Notch ligands (Fig. 7C–E). Compared with wild-type or E8I-Cre⁺ Notch1^f/f Notch2^f/f mice, CD44⁺CD4⁺ T cells from mice receiving anti-Dll1/4 treatment had decreased 1B11 reactivity, indicating that Notch also regulates core-2 O-glycosylation in activated CD4⁺ T cells after Lm-OVA infection (Fig. 7F). Furthermore, 1B11 reactivity was higher in wild-type compared with E8I-Cre⁺ Notch1^f/fNotch2^f/f mice at day 8 and day 30 after LCMV clone 13 infection (Supplemental Fig. 4). Altogether, our data identify core-2 O-glycosylation and 1B11 reactivity as conserved and sensitive readouts of Notch signaling activity in T cells in multiple types of T cell responses.

Multiple studies have shown that Notch signals regulate T cell differentiation and function during adaptive immune responses with context-specific effects. However, in most contexts, little is known about the cellular source, regulation, and timing of Notch signals received by T cells. Identification of a simple and generalizable cell surface marker of Notch signaling in activated T cell subsets will enhance our understanding of how Notch shapes T cell responses in specific immune contexts.

In this study, we report that Gcnt1-dependent core-2 O-glycosylation of CD43 behaves as a new cell surface marker to indicate whether a T cell has received Notch signals after Ag activation. This marker was quantitative as it directly correlated to dose-dependent pharmacologic blockade of Notch ligands and could be easily monitored by flow cytometry with the mAb 1B11. Upregulated core-2 O-glycosylation of CD43 was cell intrinsically controlled by Notch signaling within multiple T cell subsets and depended on Notch signals mediated via its canonical transcriptional activation complex. Importantly, the usefulness of this new surface marker was generalizable from alloimmunity to immunization, bacterial, and viral infection models. Furthermore, monitoring CD43 glycosylation status could predict the critical cellular sources of Notch ligands that drove lethal alloimmunity.

We and others previously relied on CD25 as a marker of Notch signaling as its expression is regulated by Notch in developing as well as mature T cells (50). However, 1B11 functioned as a superior marker for Notch signaling over CD25 for three reasons. First, blocking Notch signaling after allo-BMT expanded the proportion of Foxp3⁺ Tregs, which highly express CD25 irrespectively of Notch signals (8, 11). Instead, 1B11 reactivity in Tregs was regulated by Notch as it was in CD4⁺ Tconv. Second, 1B11 reactivity appears more sensitive and less transient than CD25 expression. For example, we had previously not observed decreased CD25 expression among dnMAML T cells in unirradiated allotransplantation models (8). Furthermore, only a small fraction of Ag-specific CD8⁺ T cells express CD25 at day 8 postinfection (26) or day 6 post-DC immunization (51). Instead, we have shown that 1B11 remains highly detectable on CD8⁺ T cells until day 14 postactivation in transplantation, whereas others have shown its presence until at least day 14 after acute viral infection (26). In fact, virus-specific Notch-deficient CD8⁺ T cells still had reduced 1B11 reactivity at day 30 after inoculation with LCMV clone 13, a model of chronic infection. Third, 1B11 reactivity had a stricter requirement for Notch signals than CD25 as we observed small but significant CD25 upregulation in dnMAML 4C T cells. Thus, 1B11 reactivity provides both more sensitive and more specific information than CD25 expression as a readout of Notch signaling activity.

In addition to its increased expression in GVHD (27, 28), 1B11 reactivity has previously been proposed as a marker to distinguish CD8⁺ effector T cells from memory T cells in response to viral infection (26). Similar to our studies in which high 1B11 reactivity correlated with cytokine production, the authors showed that high 1B11 reactivity correlated with cytolytic function. Because of the centrality of Notch signaling in driving CD8⁺ T cell effector differentiation and short-lived effector functions, we speculate that Notch orchestrates both Gcnt1-dependent 1B11 reactivity and effector T cell differentiation during early stages of T cell activation. Whether other signals can bypass Notch signaling to drive activation-associated Gcnt1 upregulation, core-2 O-glycosylation, and effector T cell differentiation remains to be seen. In contrast to in vitro experiments, in vivo findings showed that IL-12, IL-2, and IL-15 were dispensable for functional P-selectin ligand formation and thus likely for Gcnt1 induction and core-2 O-glycosylation in effector CD8⁺ T cells (23). In CD4⁺ T cells, Gcnt1 expression depended on TCR signaling in concert with STAT4-dependent IL-12 signaling or other diverse cytokines in vitro (29, 31). However, little was known about the regulation of Gcnt1 in CD4⁺ T cells in vivo. Given that dnMAML CD4 Tconv fully recapitulated and dnMAML CD8 T cells largely recapitulated the impact of Gcnt1 deficiency on 1B11 expression, Notch signals are critical for Gcnt1 upregulation and core-2 O-glycosylation, at least in the context of allo-BMT. We show that a short pulse of Notch signaling in cooperation with Ag activation drives 1B11 reactivity in effector T cells. Over time, 1B11 reactivity tends to decay in Ag-activated CD4⁺ and CD8⁺ T cells. In contrast, memory CD8⁺ T cells were reported to express Gcnt1 in an IL-15–dependent and Ag-independent manner (24). It would be interesting to test if Notch signals also cooperate with IL-15 in this context to regulate Gcnt1 expression.

To show the utility of 1B11 reactivity as a surrogate surface marker of Notch signaling, we asked whether it could be used to predict the cellular source of Notch ligands that drive lethal alloimmunity. We previously showed that nonhematopoietic fibroblastic stromal cells lineage traced by the Ccl19-Cre transgene were the source of Delta-like ligands after allo-BMT with high-intensity myeloablative conditioning (10). However, in models of allotransplantation with reduced or no prior conditioning, it was unclear if viable hematopoietic cells could substitute for fibroblastic stromal cells as a source of Notch ligands. By comparing 1B11 reactivity in cotransferred wild-type and dnMAML T cells, we predicted that in both models of allotransplantation with reduced or no prior conditioning, fibroblastic stromal cells were necessary and hematopoietic cells dispensable to drive lethal Notch-mediated T cell pathogenicity. Additionally, we used 1B11 reactivity as a surrogate marker to compare Notch signals received by CD8⁺ T cells during L. monocytogenes infection. Using 1B11, we predicted that pharmacologic blockade of Dll1/4 ligands would recapitulate the cell-intrinsic loss of Notch1/2 receptors in CD8⁺ T cells. This was confirmed when studying the differentiation of Ag-specific CD8⁺ T cells into SLECs. Together, these findings show for the first time, to our knowledge, that Delta-like Notch ligands are essential, and Jagged ligands are dispensable during Notch-dependent CD8⁺ T cell differentiation. This is consistent with recent findings that dendritic cells lacking Dll1 but not Jag1 had the decreased ability to support anti-tumor T cell response (52). Furthermore, the fact that pharmacologic blockade of Dll1/4 ligands fully recapitulated the loss of 1B11 reactivity and effector differentiation seen in Notch1/2-deficient CD8⁺ T cells suggests that there is a limited role for ligand-independent Notch signaling in vivo, in contrast to suggestions by other groups based on in vitro findings (53, 54).

What is the functional significance of Notch-dependent Gcnt1-mediated core-2 O-glycosylation in activated T cell subsets during GVHD? Gcnt1 is crucial for the creation of functional ligands for P- and E-selectin in leukocytes and their trafficking into some inflamed nonlymphoid tissues. In our model of MHC-mismatched allo-BMT, Gcnt1-deficient T cells could still traffic into liver and GI target tissues and mediate lethal GVHD in a Notch-dependent manner. Thus, Gcnt1 function is likely not central to T cell pathogenicity in this context. Interestingly, others have shown that P-selectin–deficient recipients had decreased lethality in GVHD models, although PSGL1-deficient T cells did not induce reduced mortality (48). Together with our data, these findings suggest that T cells can traffic into the GI tract and liver independently of functional P-selectin ligands. Other mechanisms such as α4β7/MAdCAM-1 interactions may also be involved in T cell trafficking into the gut during GVHD (55). P-selectin–deficient mice may be protected from GVHD as the subsequent migration of donor bone marrow–derived myeloid cells into target tissues, which is necessary to sustain GVHD (56, 57), may require P-selectin to traffic into target tissues. Additionally, because morbidity in our allo-BMT model is predominantly driven by early GI damage, we did not examine skin-homing T cells, which may have a more stringent requirement for expression of P- and E-selectin ligands. In fact, we recently showed in a model of chronic GVHD that Notch-deprived T cells had significantly impaired accumulation in the skin (12).

In humans, a core-2 O-glycan–specific mAb for CD43, 1D4, recognizes a subset of what has been described as memory CD4⁺CD45RO⁺ mature T cells (58). Another mAb, T305, specific for core-2 O-glycosylated CD43 showed increased reactivity in T cells of patients with GVHD, autoimmunity, or viral infection (59). These findings suggest that similar regulation of core-2 O-glycans may exist in recently activated human effector T cells in humans.

Much remains to be understood about Notch signaling during T cell activation. Our work defining a cell-intrinsic generalizable indicator of Notch signals in activated T cell subsets will help answer pressing questions in the field. For example, the cellular source of Notch ligands remains unknown in immunization and infection models. Also, little is known about the dynamic regulation of Notch ligand expression and activity in APCs versus fibroblastic stromal cells. Assessing surface core-2 O-glycosylation of CD43, a simple marker correlated to Notch-dependent T cell function, will help decipher the immunobiology of Notch signaling in T cell immunity.

B.R.B. receives remuneration as an advisor to Kamon Pharmaceuticals, Five Prime Therapeutics, Regeneron Pharmaceuticals, Magenta Therapeutics, and BlueRock Therapeutics as well as research support from Fate Therapeutics, RXi Pharmaceuticals, Alpine Immune Sciences, AbbVie, BlueRock Therapeutics, Leukemia & Lymphoma Society, Children’s Cancer Research Fund, and Kids First Fund. He is a cofounder of Tmunity. The other authors have no financial conflicts of interest.