Lunatic, Manic, and Radical Fringe Each Promote T and B Cell Development Free

Lunatic, Manic, and Radical Fringe are glycosyltransferases that transfer N-acetylglucosamine to O-linked fucose (O-fucose) present at a particular consensus site of epidermal growth factor-like repeats (1, 2). Mammalian Fringe genes Lfng, Mfng, and Rfng were identified based on their sequence homology to Drosophila Fringe (3, 4), originally identified as a gene that modifies Notch signaling (5). Subsequently, mice lacking Lfng were shown to have severe skeletal defects and disrupted Notch signaling during somitogenesis (6, 7). The finding that Fringe modification of Notch receptors alters their binding of, and response to, Notch ligands (8–10), identified a mechanistic basis for the regulatory effects of Fringe glycosyltransferases on Notch signaling.

The first indication that Fringe could affect the regulation of T cell development was obtained when Lfng was misexpressed in thymus under the control of the lck-proximal promoter (11). Large numbers of B cells are generated in the thymus of lck-Lfng transgenic mice. Lfng is normally expressed in CD4⁻CD8⁻ double-negative (DN) T cell progenitors, expressed poorly in CD4⁺CD8⁺ double-positive (DP) T cell precursors, and expressed at high levels in CD4⁺ and CD8⁺ single-positive (SP) T cells (12, 13). Misexpression of Lfng in lck-Lfng DP T cell precursors leads to their increased binding to Notch ligands on stromal cells, which blocks the access of DN T cell progenitors to thymic stroma, thereby allowing the differentiation of early T cell progenitors to B cells (14). Consistent with this, inactivation of Lfng causes reduced competitiveness in mixed repopulation experiments, and reduced T cell development from fetal liver cells (12) or from thymocytes expressing shRNA-targeted Lfng (13). NOTCH1 was implicated directly as a substrate of LFNG by showing that T cell development in thymus from Notch1^(12f/12f):lck-Lfng mice, in which NOTCH1 lacks the O-fucose site in the Notch ligand binding domain, is less affected by lck-Lfng (15). Roles for Mfng and Rfng in T cell development have not been reported, nor have roles for Rfng during B cell development. However, both Lfng and Mfng are important for optimal marginal zone (MZ) B cell development in spleen (16). All three Fringe genes are expressed in DN T cell progenitors and mature T and B cells of the mouse (17–19).

In this study, we investigate T and B cell development in mutant mice with inactivated Fng genes (20), including mice lacking a single Fng gene, all three Fng genes, or expressing only a single Fng (i.e., lacking two of the three Fng genes). Although the loss of Lfng can cause perinatal lethality, Lfng null homozygotes in a FVB/C57BL/6 mixed genetic background live for several months, although they are small, lack a tail, and are infertile (20–22). Deletion of Mfng or Rfng separately or together has no obvious effects on development or fertility (20, 23, 24). In this study, we show that DN T cell progenitors lacking expression of all three Fng genes (triple knockout [Fng tKO]) had reduced binding of Notch ligand DLL4 and reduced expression of the Notch targets Deltex1 and CD25. Fng tKO cells had altered frequencies of several T and B cell subsets in thymus and spleen, and this phenotype was transferable by bone marrow transplantation. Mice expressing only a single allele of Lfng, Mfng, or Rfng were rescued in the major T and B cell subset frequencies. Finally, splenic T and B cell responses to various stimulants were reduced in Fng tKO mice.

Mice null for Mfng and Rfng and heterozygous for Lfng on a mixed C57BL/6/FVB background were a gift from Susan Cole (University of Ohio) and are described in Moran et al. (20). The mice were intercrossed to obtain Fng tKO mice, in which all three Fng genes were inactivated. They were also crossed with FVB mice to generate mice expressing all three Fng genes (Fng LMR). The latter expressed one allele of each Fng gene or were Lfng^+/+Mfng^+/−Rfng^+/−. Fng LMR and Fng tKO mice were also generated by crossing Lfng^+/−Mfng^−/−Rfng^−/− mice to Fng LMR mice. Rfng^+/− mice (23) on a mixed background were obtained from the Jackson Laboratory (Bar Harbor, ME) and backcrossed for five to six generations to C57BL/6 mice before intercrossing. CD45.2⁺ C57BL/6 mice were also obtained from the Jackson Laboratory. Genotyping was performed by PCR of genomic DNA using primers that distinguish wild type and mutant alleles as described previously (20). Mice were housed in a barrier facility, allowed to eat and drink ad libitum, and used in experiments at 6–8 wk of age. All experiments were performed with permission from the Albert Einstein Institutional Use and Animal Care Committee. Euthanized mice were weighed, and isolated thymus and spleen were also weighed before making single-cell suspensions.

Except where noted, Abs and Ab conjugates to FITC, PE, or allophycocyanin were obtained from eBioscience (San Diego, CA) as follows: CD4-FITC rat IgG_2a clone RM4-5; CD8a-allophycocyanin rat IgG_2a clone 53-6.7; CD25-FITC rat IgM κ clone 7D4 (BD Biosciences, San Jose, CA); CD44-Alexa Fluor⁷⁰⁰ rat IgG_2b κ clone 1M7; IgM-allophycocyanin rat IgG_2a clone II/41; B220-Alexa Fluor⁷⁰⁰ rat IgG_2a, clone RA3-6B2; CD21-FITC rat IgG_2b clone 7G6 (BD Biosciences); CD23-PE rat IgG_2a clone B3B4; allophycocyanin rat IgG_2a isotype; Alexa Fluor⁷⁰⁰ rat IgG_2a isotype; Alexa Fluor⁷⁰⁰ rat IgG_2b isotype; FITC rat IgG_2b isotype; PE rat IgG_2a isotype; R-PE AffiniPure F(ab′)₂ Frag goat-anti-human IgG, Fcγ Frag Spec (Jackson ImmunoResearch, West Grove, PA); rat-anti-mouse CD16/CD32 clone 2.4G2 (mouse Fc block; BD Biosciences); CD45.2-FITC mouse IgG_2a clone 104; CD45.1-PE-Cyanine 7 mouse clone A20; CD4-allophycocyanin mouse clone Gk1.5; anti-CD8a-PE mouse clone 53-6.7; CD25-PerCPCy5.5 rat IgG₁κ clone PC61.5; CD44-PE clone IM7. CD117-allophycocyanin rat IgG_2b κ clone 2B8; CD11b-FITC rat IgG_2b clone M1/70 (BD Biosciences); CD11c-PE hamster IgG₁ clone HL3 (BD Biosciences); Gr1-allophycocyanin rat IgG_2b clone RB6-8C5 (BD Biosciences); Ly6G-PerCPCy5.5 rat IgG_2a κ clone 1A8; CD62L-PerCPCy5.5 rat IgG_2a κ clone MEL-14; CD122-PECy7 rat IgG_2b κ clone TM-b1 (BD Biosciences); CD69-PE Armenian Hamster IgG clone H1.2F3; Foxp3-allophycocyanin rat clone FJK-16s; Ag-purified polyclonal sheep anti-NOTCH1 (aa 19-526) AF5267 (R&D Systems, Minneapolis, MN); rabbit polyclonal IgG anti-NOTCH2 (aa 25-255) sc-5545 (Santa Cruz, Dallas, TX); Rhodamine Red-X-conjugated donkey anti-sheep IgG (Jackson ImmunoResearch); and anti-rabbit IgG-PE (Jackson ImmunoResearch).

Single-cell suspensions from thymus or spleen were prepared using homogenization by inserting a 3-ml syringe and passage through a 70-μm strainer. Thymocytes were washed in cold FACS binding buffer (FBB; HBSS, 2% BSA, 0.05% sodium azide, pH 7.2–7.4), resuspended, and counted in a Coulter counter. Splenocytes were incubated in 3 ml RBC lysis buffer (0.15 M NH₄Cl, 10 mM KHCO₃, 0.1 mM EDTA, pH 7.2–7.4) for 1.5 min before adding 20 ml FBB. After centrifugation and resuspension in FBB, splenocytes were counted in a Coulter counter. For Ab binding to unfixed cells, cells were incubated with fluorochrome-conjugated Abs according to standard protocols. Briefly, 5–10 × 10⁵ cells were washed with 1 ml FBB, resuspended in 90 μl FBB containing 1 μl Fc block (rat-anti-mouse CD16/CD32), and incubated for 15 min on ice. Ab diluted in FBB (10 μl) was added and the tube was incubated for 30 min at 4°C. Cells were washed twice in 1 ml FBB and transferred to a 5-ml Falcon tube in 300–500 μl FBB, to which was added 5 μl 7-actinomycin D (7-AAD, BD Biosciences) in 100 μl FBB. After 10 min on ice, cells were subjected to flow cytometry. Damaged cells that were 7-AAD⁺ were excluded by gating. For cells that had been fixed in 4% PBS-buffered paraformaldehyde at room temperature for 15 min and stored at 4°C, no 7-AAD was added. For fixation followed by permeabilization, the Fixation/Permeabilization solution from eBioscience was used according to the manufacturer’s instructions. For all samples, immunofluorescence was analyzed using FACSCalibur or FACScan flow cytometers (BD Biosciences), and data files were analyzed using FlowJo software (Tree Star, Ashland, OR).

Fresh thymocytes were resuspended in isolation buffer (PBS without cations, pH 7.2-7.4, containing 0.1% BSA and 2 mM EDTA) on ice. For CD4⁺ CD8⁺ T cell depletion, 5 × 10⁷ thymocytes were incubated with 20 μg anti-CD4 (rat IgG_2b clone GK1.5; BioXCell, West Lebanon, NH) and 37.5 μg anti-CD8a (rat IgG_2a clone 53-6.72; BioXCell) in 5 ml isolation buffer for 20 min at 4°C with tilted rotation, centrifuged, resuspended in 5 ml isolation buffer, and incubated twice with 250 μl sheep anti-rat IgG Dynabeads (Thermo Fisher Scientific, Waltham, MA) for 30 min at 4°C with tilted rotation. After each incubation, the tube was placed in a magnet for 2 min, unbound DN T cell progenitors were combined in a new tube and centrifuged, and RNA was extracted from the cell pellet with 1 ml TRIZOL (Ambion, Carlsbad, CA) as described below.

DN T cell progenitors from 5 × 10⁷ thymocytes were pipetted vigorously in 1 ml TRIZOL and incubated for 5 min at room temperature before adding 0.2 ml chloroform. Tubes were vortexed for 15 s, incubated at room temperature for 2–3 min, and centrifuged at 12,000 × g for 15 min at 4°C. The aqueous phase was transferred to a fresh tube, and 0.5 ml isopropanol was added. Samples were incubated at room temperature for 10 min. Following centrifugation at 12,000 × g for 10 min at 4°C, the RNA pellet was washed once with 1 ml 75% ethanol. Samples were vortexed and centrifuged at 7500 × g for 5 min at 4°C. The RNA pellet was air-dried for 5–10 min and dissolved in 25–30 μl RNase-free water. RNA concentration was determined by Nanodrop and cDNA was prepared from 500 ng RNA using the Verso cDNA synthesis kit (Thermo Fisher Scientific) following the manufacturer’s protocol. The product was diluted in RNase-free water to 7.5 ng/μl, and 2 μl was used for quantitative RT-PCR using the following primers:

• Lfng-F: 5′-CTGCACCATTGGCTACATTG; Lfng-R: 5′-ATGGGTCAGCTTCCACAGAG

• Mfng-F: 5′-ATGCACTGCCGACTTTTTCG; Mfng-R: 5′-CCTGGGTTCCGTTGGTTCAG

• Rfng-F: 5′-TGCTGCTGCGTACCTGGATCTC; Rfng-R:5 ACAGCAGAGCAATTGGTGTTGA

• Hes1-F: 5′-AAGGCAGACATTCTGGAAAT; Hes1-R: 5′- GTCACCTCGTTCATGCACTC

• Dtx1-F: 5′-CATCAGTTCCGGCAAGAC; Dtx1-R: 5′ATGGTGATGCAGATGTCC

• cMyc-F: 5′-AGTGCTGCATGAGGAGACAC; cMyc-R: 5′ GGTTTGCCTCTTCTCCACAG

• CD25-F:5′-GGAATTGGTCTATATGCGTTGCTTA;

• CD25-R:5′-CATGTCTGTTGTGGTTTGTTGCTCT

• Actb-F: 5′-TTCTACAATGAGCTGCGTGTG; Actb-R: 5′-GGGGTGTTGAAGGTCTCAAA

• Gapdh-F: 5′-AAGGTCATCCCAGAGCTGAA; Gapdh-R: 5′-CTGCTTCACCACCTTCTTGA

• Hprt-F: 5′-GGACCTCTCGAAGTGTTGGATAC;

• Hprt-R: 5′-GCTCATCTTAGGCTTTGTATTTGGCT

Soluble Notch ligands DLL4-Fc, DLL1-Fc, JAG1-Fc, and Fc control were prepared from HEK-293T cells as described previously (25). JAG2-Fc was purchased from R&D Systems. Single-cell suspensions from thymus were washed in ligand binding buffer (LBB; HBSS pH 7.4, 1 mM CaCl₂, 1% [w/v] BSA, 0.05% NaN₃) and fixed in PBS-buffered 4% paraformaldehyde for 15 min at room temperature, washed twice with LBB and stored in LBB at 4°C. Fixed thymocytes were washed with LBB and 3 × 10⁶ cells were incubated with FcR blocking solution (rat-anti-mouse CD16/CD32) on ice for 15 min. Thereafter, the cells were incubated in 100 μl LBB containing anti-CD4-FITC (1:200), anti-CD8a-allophycocyanin (1:200) and 500–750 ng DLL1-Fc, DLL4-Fc, JAG1-Fc, JAG2-Fc, or Fc. After incubation at 4°C for 1 h, cells were washed twice with 0.5 ml LBB and incubated with anti-IgG-PE (Fc-specific) Ab (1:100) at 4°C for 30 min. The cells were then washed twice with 1 ml LBB and analyzed in a FACSCalibur flow cytometer (BD Biosciences). For detection of NOTCH1 and NOTCH2 at the cell surface, fixed thymocytes were incubated with FcR block rat-anti-mouse CD16/CD32 (1:100) followed by CD4-FITC mAb (1:200), CD8a-allophycocyanin mAb (1:200), sheep anti-mouse NOTCH1 Ab (1:50), or rabbit anti-human NOTCH2 Ab (1:100) at 4°C for 1 h, washed, and incubated with rhodamine Red-X-conjugated donkey anti-sheep IgG (1:100) or anti-rabbit IgG-PE (1:100) at 4°C for 30 min. Cells were washed twice with 1 ml LBB and analyzed using a FACSCalibur flow cytometer.

Cell suspensions were made from bone marrow flushed from the femur of 7–8 wk Fng LMR and Fng tKO mice into 5 ml cold HBSS, centrifuged, and resuspended in RBC lysis buffer. After 1.5 min on ice, 10 ml HBSS was added, and the cells were washed three times in HBSS and counted. Fng LMR, Fng tKO, and a 1:1 mix (3 × 10⁶ cells total for each sample) were injected retro-orbitally in 50 μl HBSS into CD45.2⁺ C57BL/6 lethally irradiated recipients. γ-Irradiation of 500 rad per recipient mouse was given twice, with a 16-h interval. After 6 wk, mice were euthanized, thymus and spleen were weighed, and thymocytes and splenocytes were analyzed for T and B cell subsets by flow cytometry after gating to remove damaged cells (7-AAD⁺), and subsequently on donor-derived cells that were DP for staining with Abs CD45.1-PE-Cyanine 7 and CD45.2-FITC.

Whole splenocytes (10⁷ cells/ml) were incubated in 10 μM CFSE (Molecular Probes, Eugene, OR) in PBS containing 5% heat-inactivated FBS for 10 min at 37°C. Uptake was inhibited by the addition of two volumes of ice-cold complete RPMI 1640 medium (RPMI 1640 containing 10% heat-inactivated FBS and 1% penicillin–streptomycin), and the cells were incubated on ice for 5 min. Cells were then washed twice in complete RPMI 1640 medium and resuspended to the desired concentration in complete RPMI 1640 medium. Fresh CFSE-labeled splenocytes were cultured at 10⁶ cells per well in a 24-well plate in 1 ml complete RPMI 1640 medium. T cells were stimulated by the addition of prewashed CD3/CD28 Dynabeads (Thermo Fisher Scientific) at 25 μl/ml and rIL2 (PeproTech, Rocky Hill, NJ) at 5 ng/ml. B cells were stimulated with LPS at 15 μg/ml (Escherichia coli, serotype O55:B5; Sigma-Aldrich, St. Louis, MO). Plates were incubated in a humidified atmosphere of 5% CO₂ at 37°C. After 3 d, cells were harvested, incubated with fluorochrome-conjugated anti-CD4 (1:200) and anti-CD8a (1:200) or anti-B220 (1:100) Abs, followed by the addition of 7-AAD (1:20), and analyzed by flow cytometry. FlowJo algorithms were used to quantitate the CFSE profiles of viable (7-AAD^–) T or B cell populations.

T cell activation in response to Concanavalin A (Con A; Pharmacia, Uppsala, Sweden) or Phaseolus vulgaris leukoagglutinin (L-PHA; Vector Labs, Burlingame, CA) were investigated by expression of CD69 using flow cytometry. Fresh splenocytes were washed twice in complete RPMI 1640 medium, counted, and added to 24-well culture plates at 10⁶ cells per well. Splenocytes were stimulated by the addition of Con A (5 μg/ml) or L-PHA (2 μg/ml) added to duplicate wells in a final volume of 1 ml. The concentrations of Con A and L-PHA that induced maximal cell proliferation were determined in preliminary experiments. After 20 h at 37°C in a humidified atmosphere of 5% CO2, cells were harvested, washed with FBB, and incubated with fluorochrome-conjugated anti-CD4 (1:200), anti-CD8a (1:200), and anti-CD69 (1:80) Abs, followed by the addition of 7-AAD (1:20). Analysis with flow cytometry was performed to determine CD69 expression on 7-AAD⁻ CD4⁺ and CD8⁺ T cells.

Comparisons are presented as mean ± SEM. Significance was determined by two-tailed unpaired, parametric, Student t test analysis (unless otherwise noted), using Prism software.

Mice lacking a single Fng activity were compared for T and B cell development. Lfng mutant mice are viable on a background including C57BL/6 and FVB/NJ (20). Lfng^−/− mice (FVB/C56BL/6) age 7–8 wk had reduced body, thymus, and spleen weights compared with Fng LMR, Mfng^−/−, or Rfng^−/− mice (Supplemental Fig. 1), and absolute numbers of thymocytes and splenocytes were equivalently reduced. However, the ratio of thymocytes and splenocytes to body or organ weight were similar in all Fng mutant mice and controls. Therefore, the frequencies of different T, B, and myeloid subsets in thymocyte or splenocyte populations were determined. T cell subsets were determined using Abs against CD4, CD8a, CD44, and CD25; B cell subsets using Abs against B220, CD21, CD23, and IgM; myeloid cells using Abs against Gr1, CD11b, and CD11c; and Tregs using Abs against CD4 and Foxp3. No significant differences were observed between controls and single Fng knockout mice in the relative proportions of DN or DP T cell precursors, or CD4⁺ or CD8⁺ SP T cells in thymus (Supplemental Fig. 1). There were only minor populations of B220⁺, Tregs or myeloid cells in thymus, and they were similar in frequency to controls (not shown). In spleen, the proportions of CD4⁺ and CD8⁺ T cells, B220⁺ cells (Supplemental Fig. 1), and myeloid cell subsets (not shown) were similar in single Fng knockout and Fng LMR mice. To determine whether differences caused by the absence of Rfng could be detected in a more homogeneous genetic background, Rfng^−/− mice on a mixed 129Sv/C57BL/6 background were backcrossed to C57BL/6 for five to six generations and wild type and heterozygous mutant mice were compared with homozygous mutant littermates. Only minor variations in the proportions of T and B cell subsets were observed when mice were compared on the basis of sex (Supplemental Fig. 2). Therefore, we focused on mice in which two Fng genes or all three Fng genes were inactivated.

Lfng is well expressed in DN T cell progenitors but poorly expressed in DP T cell precursors that compose the majority of the thymocyte population (12, 13). Therefore, Notch ligand binding was examined using DN T cell progenitors. Thymocytes from 6–8-wk-old Fng LMR and Fng tKO mice were fixed and incubated with CD4-FITC and CD8a-allophycocyanin mAbs, along with Notch ligand-Fc or Fc control. Notch ligand binding was determined on DN T cell progenitors using anti-Fc-PE. Consistent with previous reports (26), Fng LMR DN T cell progenitors bound DLL4-Fc (Fig. 1A) better than DLL1-Fc (Fig. 1B). However, Fng tKO DN T cell progenitors exhibited markedly reduced binding of DLL4-Fc (Fig. 1A). DN T cell progenitors from Fng dKO mice expressing only Mfng or only Rfng were partially rescued for DLL4-Fc binding (Table I). Cell surface expression of NOTCH1 and NOTCH2 was similar in DN T cell progenitors from Fng LMR and Fng tKO mice (Supplemental Fig. 3), indicating that reduced DLL4-Fc binding primarily reflects reduced interactions with Notch receptors lacking Fringe modification. Therefore, all three Fringe activities contributed to DLL4-Fc binding to DN T cell progenitors. The absence of Fringe might be expected to enhance the binding of Notch ligands JAG1 and JAG2 as Fringe may inhibit Jagged ligand binding (25, 27), but JAG1-Fc and JAG2-Fc bound similarly to DN T cell progenitors from controls and mice lacking all three Fringe activities (Fig. 1C, 1D).

The expression of each Fringe gene during T cell development in C57BL/6 mice was extracted from published microarray data (Supplemental Fig. 4). The relative expression of Lfng and Mfng does not change significantly through early T cell progenitor development, is reduced in DP T cell precursors, and increases markedly in CD4⁺ and CD8⁺ SP T cells. By contrast, Rfng expression is relatively high in ETP progenitors, reduced in DN1 and DN2 T cells, and equivalently high in DN4, DP, CD4⁺, and CD8⁺ SP T cells. The expression of Fringe and Notch target genes in DN T cell progenitors from mice expressing all Fng genes (LMR), a single Fng gene (L, M, or R) or no Fng gene (tKO) was examined by quantitative RT-PCR. Fng expression correlated directly with genotype as expected (Fig. 2A–C). Importantly, there was no compensatory increase in expression when any combination of two Fng genes was deleted. Initial experiments to examine Notch target gene expression in cDNA from Fng LMR and Fng tKO mice (n = 3–6) revealed no change in relative expression compared with Actb of Hes1, Hes5, or Dtx2 transcripts, a reduction in CD25 to 48 ± 1% (n = 3 experiments; p < 0.0003) of Fng LMR, and a significant reduction in Dtx1 transcripts with two primer sets to 41 ± 0.8% (n = 6 experiments; p < 0.0001). Data from a more recent cohort of mice (Fig. 2) showed equivalent relative expression of Hes1 and cMyc and significantly reduced expression of Dtx1 in Fng tKO DN T cells, compared with the combined average expression of Actb, Hprt, and Gapdh (Fig. 2D–F). However, the expression of CD25 transcripts was not reduced in DN T cells from mice expressing a single Fringe, nor in Fng tKO DN T cells (data not shown). Therefore, we examined CD25 expression level by analyzing flow cytometry data from all cohorts. The mean fluorescence intensity (MFI) for CD25 was markedly reduced in Fng tKO compared with Fng LMR DN T cells, and was significantly reduced in mice expressing only Rfng or Mfng (Fig. 2G). Thus, the expression of both CD25 and Dtx1 Notch targets was consistently reduced in the absence of Fringe. Expression of a single Fng rescued Dtx1 transcript levels, but only a single Lfng substantially rescued CD25 expression at the cell surface.

Fng tKO mice had a small thymus compared with Fng LMR mice and mice lacking only Lfng (Mfng and Rfng heterozygous, designated MR), but not significantly different from mice lacking both Lfng and Mfng (designated R), or both Lfng and Rfng (designated M; Fig. 3A). Thymocyte numbers gave a similar result, but were more variable (Fig. 3B). Thus, Mfng or Rfng alone could not substitute for Lfng in the development of the thymus. However, the ratio of both thymus/body weight and the number of thymocytes/thymus weight were similar across all Fng mutant mice. Therefore, frequencies rather than absolute numbers of T, B, and myeloid cell subsets were compared. In Fng tKO thymus, the frequency of DP T cell precursors decreased slightly, whereas the frequencies of CD4⁺ and CD8⁺ SP T cells were increased (Fig. 3C, 3D). When DN T cell progenitor subsets were examined using Abs to cKit/CD117, CD44, and CD25, the proportions of DN1/cKit⁺ and DN2 T cell progenitors were reduced in Fng tKO thymus (Fig. 4). Interestingly, the proportion of B220⁺ B cells in the thymus was not increased in Fng tKO mice (0.65 ± 0.1%; n = 7) compared with Fng LMR mice (0.62 ± 0.19%; n = 3), as occurs when Notch signaling is blocked or Lfng is misexpressed (26). Fng LMR and Fng tKO thymocytes also contained similarly low proportions of regulatory T and myeloid cells expressing Gr1, CD11b, or CD11c (data not shown).

Mice expressing only Mfng, only Rfng, or both and Fng tKO mice had a smaller spleen by weight (Fig. 5A) and fewer splenocytes (Fig. 5B) than Fng LMR control mice did. However, the normalized number of splenocytes to spleen weight was similar in Fng mutant and Fng LMR mice. The proportion of B220⁺ B cells was unchanged (∼50% of the CD4⁻CD8⁻ splenocytes) by the absence of Fringe (Fig. 5C), but the proportions of both CD4⁺ and CD8⁺ T cells were decreased in Fng tKO spleen (Fig. 5D). Effector and memory T cell populations were also affected by the absence of Fringe. CD4⁺ and CD8⁺ splenocytes examined for CD44 and CD62L expression showed a reduction in the proportion of naive T cells (CD44^loCD62L^hi) as expected, an increase in T effector memory cells (CD44^hiCD62L^lo), and a reduction in central memory T cells (CD44^hiCD62L^hi) in Fng tKO mice (Fig. 6A, 6B). CD8⁺ splenocytes gave a similar result when CD44 and CD122 expression were compared: a reduced proportion of CD44^hiCD122^hi central memory T cells, and an increased proportion of CD44^hiCD122^lo effector memory T cells (Fig. 6C). For B220⁺ cells, the frequency of follicular B (Fo B) cells was increased (Fig. 7A, 7B), whereas the proportion of MZ B cells was decreased in Fng tKO splenocytes (Fig. 7C, 7D). MZ P cell frequency was concomitantly increased (Fig. 7C). These results are consistent with evidence that Lfng and Mfng are both required for optimal MZ B cell development (16). Finally, in all forward scatter versus side scatter (SSC) profiles of fresh or fixed splenocytes (including cells stored for several months at 4°C), we observed an increased population of granulated cells (SSC^hi) in Fng tKO spleen. Initial investigations revealed an increased frequency of Ly6G⁺ cells in Fng tKO splenocytes. Subsequently, we determined that Gr1⁺Ly6G⁺ neutrophils were significantly increased in splenocytes from Fng tKO compared with Fng LMR mice (Fig. 7E, 7F), a phenotype consistent with reduced Notch signaling in bone marrow (28).

To determine whether the Fng tKO phenotype is cell autonomous, bone marrow transplantation experiments were performed. Two of three experiments gave >35% reconstitution of donor bone marrow. In one experiment, donor reconstitution was ∼40% for thymus and spleen, and there were no significant differences between T and B cell subset frequencies for cells derived from Fng tKO versus Fng LMR bone marrow. However, in the experiment in which Fng tKO and Fng LMR control donor cells constituted 74–88% of host thymocytes and 64–74% of host splenocytes, a cell-autonomous effect for Fng tKO recipients was observed (Fig. 8). Thus, thymus weight and DP T cell precursor frequency were reduced in recipients of Fng tKO versus Fng LMR bone marrow, or an equal mixture of Fng LMR and Fng tKO bone marrow (Fig. 8A). In addition, frequencies of donor-derived CD4⁺ and CD8⁺ SP T cells were increased in thymus (Fig. 8A), as observed in Fng tKO thymus (Fig. 3). In spleen, reconstitution from Fng tKO bone marrow resulted in a significant decrease in the frequencies of CD4⁺ and CD8⁺ T cells (Fig. 8B), as observed in Fng tKO spleen (Fig. 5). Importantly, the proportion of each T cell subset was similar in thymus and spleen from mice that received Fng LMR cells or the 1:1 mix of Fng LMR and Fng tKO mice, indicating that the differential production of T cells from Fng tKO bone marrow was overcome by coinjection with Fng LMR bone marrow.

To identify roles for individual Fringe activities, we analyzed Fng dKO mice in which only one Fng gene was active. A single allele of either Mfng or Rfng did not rescue the reduced thymocyte and splenocyte numbers observed in Fng tKO mice (Figs. 3, 5); therefore, T and B cell subset frequencies were compared. There were no significant differences in the frequencies of DN, DP, or SP T cells in thymus of mice expressing a single Fringe versus Fng LMR control mice (Fig. 9A). Therefore, each Fringe, acting alone, could restore the reduced proportion of DP T cell precursors, and did not exhibit the increased proportions of CD4⁺ and CD8⁺ T cells observed in Fng tKO thymus (Fig. 3).

In spleen, the percentage of CD4⁺ T cells from mice expressing only Rfng or Lfng was similar to Fng LMR mice (Fig. 9B), demonstrating rescue compared with Fng tKO mice (Fig. 5D). However, mice expressing only Mfng were only partially rescued (Fig. 9B). By contrast, for the CD8⁺ T cell subset in spleen, mice expressing only Mfng or only Rfng were rescued, but mice expressing only Lfng were not fully rescued (Fig. 9B). The proportion of B220⁺ and Fo B splenocytes was not changed in mice expressing a single Fringe gene compared with Fng LMR mice (Fig. 9B). However, MZ B cell frequencies in spleen from mice expressing only one allele of Mfng or Rfng were significantly reduced, whereas mice expressing only Lfng were similar to Fng LMR mice (Fig. 9B). These results show that each Fringe enzyme was largely sufficient, in the absence of the other two Fringes, to support T cell development in thymus and spleen, although MZ B cell production was not optimal in mice expressing only Mfng or only Rfng. Rescue of the Fng tKO T cell phenotype was not expected for Rfng. To demonstrate functional effects of Rfng more clearly, T cell subsets from mice expressing one allele of Rfng were compared with Fng tKO (Fig. 10). It is apparent that Rfng was able to restore T cell subset frequencies to control levels in both thymus and spleen.

T and B cell activation were investigated by stimulation with anti-CD3/CD28 Dynabeads and IL 2, or LPS, respectively. Stimulation of splenocytes with anti-CD3/CD28 Dynabeads and IL-2 showed that proliferation was reduced in both CD4⁺ and CD8⁺ T cell subsets from Fng tKO spleen (Fig. 11A, 11B). Stimulation of splenocytes by LPS revealed that B cells lacking Fringe also proliferated less well (Fig. 11C). Stimulation by the lectins Con A or L-PHA caused expression of CD69 in most CD4⁺ and CD8⁺ Fng LMR splenic T cells. However, the number of activated T cells was reduced after Con A or L-PHA stimulation of T cells lacking Fringe, particularly in the CD4⁺ subset (Fig. 11D, 11E).

In this study, we show that Mfng and Rfng are required for optimal T and B cell development. Although Lfng plays a major role in promoting general development of thymus and spleen and in T and B cell development, we found unexpectedly that when Lfng was absent, a single allele of Mfng or Rfng supported the development of normal, or nearly normal proportions, of most T and B cell subsets. When all three Fng genes were inactivated, reduced Notch signaling and Notch ligand binding were evident in thymic DN T cell progenitors. It is of interest that the large reduction in DLL4-Fc binding to DN T cells was reflected in only a modest reduction in Notch signaling, presumably because DLL4-Fc soluble ligand binds weakly to NOTCH receptors, and may be more Fringe-dependent than membrane-bound DLL4, which must signal well but clearly not optimally in the absence of Fringe. Thus, the proportions of thymus ETP and DN2 T cell progenitors and DP T cell precursors were reduced. However, the frequencies of CD4⁺ and CD8⁺ SP T cells in thymus were slightly increased in mice lacking the three Fringe activities. These effects were also observed when Fng tKO bone marrow cells were transferred to irradiated hosts, showing that the effects of Fng were cell autonomous, as observed in other contexts (26).

The increased frequency of CD4⁺ and CD8⁺ SP T cells in Fng tKO thymus is intriguing. Potential and nonexclusive explanations include: 1) SP T cells lacking all Fringe activities emigrate from the thymus at a slower rate compared with mice expressing Fng; 2) the transition from DP T cell precursors to SP T cells occurs at a slightly faster rate in Fng tKO thymus compared with control thymus, as observed previously in Notch1^12f/12f thymus in which Notch1 signaling is reduced (29); and 3) SP T cells undergo clonal deletion at a slower rate in Fng tKO thymus compared with control thymus. In Fng tKO mice and recipients of Fng tKO bone marrow, the numbers of CD4⁺ and CD8⁺ T cells in spleen were reduced, suggesting they are either preferentially retained in the thymus, lost during passage to spleen, or have reduced survival upon their arrival in spleen.

Whatever cellular mechanisms are responsible for the phenotype of Fng tKO mice, it is apparent that expression of a single Fng gene is sufficient to support the generation of normal proportions of the major T and B cell populations. However, a single allele of Rfng or Mfng is not sufficient to generate normal levels of spleen MZ B cells. Thus, it is probable that mice expressing only Rfng or Mfng would exhibit defective innate immunity, and potentially altered adaptive immunity to certain pathogens (30, 31). In fact, because we show in this study that splenic T and B cells from mice lacking all Fringe activities exhibited a reduced frequency of central memory T cells, and reduced responses to stimulation by anti-CD3/CD28 beads, LPS, Con A, and L-PHA, it can be expected that immunologic functions will be uncovered for potentially each mammalian Fringe when mutant mice are subjected to immunologic challenges. For example, the Rfng gene is upregulated, whereas Lfng and Mfng genes are downregulated in naive CD4+ T cells in lungs of asthmatic rats (32). Knockdown of Rfng or overexpression of Mfng or Lfng in CD4⁺ naive T cells from asthmatic lung reduces their production of Th2 cytokines and increases their production of Th1 cytokines. Some functional consequences of loss of Mfng have also been noted in B cell and macrophage responses in a general survey of a small cohort of mutant mice (33). In that study, B cells from Mfng^−/− mice exhibited increased proliferation following stimulation with anti-IgM, and responded with infiltration of fewer macrophages to i.p. stimulation by thioglycollate. It might also be expected that mammalian Fringe activities have roles in certain lymphomas or leukemias. Thus, Lfng loss in mouse prostate leads to prostatic intraepithelial neoplasia (34), and Lfng loss in mouse mammary gland was found to co-operate with amplification of the Met/Caveolin gene to promote basal-like breast cancer (35).

Investigations into the mechanism by which Fringe affects Notch signaling have been performed in cocultures using cells expressing Notch ligands (36, 37). Interestingly, both groups show that DLL4 and JAG2 induce Notch signaling that promotes T cell development. Both groups also found increased Notch signaling and T cell development after the introduction of Lfng into fetal liver cells or human hematopoietic stem cells, respectively. Structural and in vitro Notch ligand binding and signaling experiments (38) (and references therein), as well as the data reported in this manuscript, support the interpretation that the addition of GlcNAc by Fringe to Notch in signal-receiving T and B cells increases Notch ligand binding and thereby Notch signaling, which affects the differentiation of T and B cells in thymus and spleen and the functions of mature T and B cells in response to stimulation.

We thank Susan Cole (University of Ohio) for providing mice carrying mutant alleles of Lfng, Mfng, and Rfng; Cynthia Guidos (University of Toronto and Hospital for Sick Children) for helpful comments; and Wen Dong, Huimin Shang, and Subha Sundaram for technical assistance.

The authors have no financial conflicts of interest.