Cmas ^fl/fl mice (15) were crossed with lck^cre/+ mice (14) to generate T cell-specific Cmas KO mice. Complement-deficient C3 ^−/− mice (16) were crossed with Cmas ^fl/fl mice × lck^cre/+ mice.

CD45.1 congenic mice (17) were kindly provided by Thomas Winkler, University of Erlangen, Erlangen, Germany. All mouse lines were maintained on a C57BL/6 background. Experiments were performed in accordance with the German Animal Welfare Act and after approval by the Animal Welfare Committee.

Cell preparation for flow cytometry single-cell suspensions of bone marrow, thymus, spleen, and lymph nodes was prepared in phosphate-buffered saline (PBS) (Life Technologies, Carlsbad, CA, USA). Erythrocyte depletion was conducted using an ACK lysis buffer.

Cells were stained for 25 min at 4°C with antibodies (Abs), diluted in PBS containing 1% bovine serum albumin (BSA) (Roth, Karlsruhe, Germany), 2 mM ethylenediaminetetraacetic acid (EDTA) (Roth), and 0.01% sodium azide (Sigma-Aldrich, Darmstadt, Germany). The following Abs were used: anti-CD4 (clone: GK1.5; BD Pharmingen, San Diego, CA, USA), anti-CD8 (clone: 53–6.7; BioLegend, San Diego, CA, USA), anti-TCRβ (clone: H57–597; BioLegend), anti-TCRγδ (clone: eBioGL3; eBioscience, San Diego, CA, USA), anti-CD44 (clone: IM7; eBioscience), anti-CD62L (clone: MEL-14; eBioscience), anti-CD45.1 (clone: A20; BD Biosciences, San Jose, CA, USA), anti-CD45.2 (clone: 104; BioLegend), anti-B220 (clone RA3–6B2; BioLegend), anti-IgM (clone II/41; eBioscience), anti-IgD (clone 11–26c; BioLegend), anti-CD25 (clone PC61; BioLegend), and anti-CD3e (clone145–2C11; eBioscience). Fc-block (2.4G2; hybridoma generated in-house) was added to prevent unspecific binding via the Fc-part of the Abs to the cell’s Fc receptors. To stain dead cells, fixable viability dye eFL506 (eBioscience) was used. Biotinylated Abs were detected by an additional staining of the cells with streptavidin fluorochrome conjugations. Cells were washed and diluted in PBS containing 0.1% BSA, 2 mM EDTA, and 0.01% sodium azide for flow cytometry via CytoFLEX S (Beckman Coulter, Brea, CA, USA).

For intracellular activated caspase 3 staining, cells were incubated for 10 min at room temperature to allow spontaneous apoptosis. Subsequently, extracellular staining was performed as described above. Cells were fixed and permeabilized with a Cytofix/Cytoperm kit (BD Biosciences) according to the manual, and intracellular staining was conducted using the Ab anti-activated caspase 3 (clone C92–605; BD Biosciences). For the identification of regulatory T cells, manual and reagents of the anti-Mouse FoxP3 Staining Set (eBioscience) were used. After an extracellular staining as described above, thymus and splenic cells were fixed and permeabilized according to protocol. The cells were then intracellularly blocked and stained with Fc-block and anti-Foxp3 (clone: FJK-16s; eBioscience). After a final wash step, the cells were resuspended in 1× Permeabilization Buffer and analyzed via CytoFLEX S (Beckman Coulter).

For glycosylation analysis, cells were first extracellularly stained as described above. Then, the cells were fixed with 2% paraformaldehyde and stained with biotinylated lectins from VectorLabs [Newark, CA, USA; Sambucus nigra agglutinin (SNA), Maackia amurensis agglutinin (MAA II), Erythrina cristagalli agglutinin (ECA), and peanut agglutinin (PNA)] for 20 min at 4°C. Subsequently, cells were washed and incubated with streptavidin conjugates for 20 min at 4°C.

Thymus and splenic T cells and non-T cells were sorted via AriaII (BD Biosciences). The sorted cells were then lysed with ddH₂O 0.3 µg/µL Proteinase K overnight at 56°C. After the inactivation of the enzyme by incubation at 94°C for 10 min, a PCR for the deleted CMAS exon 4 allele was performed. The following primers were used: BWF59 fwd (5′-AGCGCCTGTGTACCCCTCTTA-3′), BWB58 rev (5′-GCGAGCAGCAAGTGAGCA-3′), and AMB40 rev (5′-TCAAGTTCAGAGGCTCAGTCACTTCACG-3′).

The amplified DNA fragments were loaded onto a 2% agarose gel.

Thymus cells of 1 × 10⁷ were resuspended in 700 µL of Roswell Park Memorial Institute (RPMI) media (Life Technologies) containing 5% fetal calf serum (FCS) (PAN-Biotech, Aidenbach, Germany) and loaded with 1.5 mM Indo-1 AM plus 15% pluronic acid F-127 (Molecular Probes, Eugene, OR, USA) by shaking for 25 min at 30°C. Afterward, 700 µL of RPMI media (Life Technologies) containing 10% FCS (PAN-Biotech) was added, and cells were incubated by shaking for another 10 min at 37°C. Cells were washed twice and extracellularly stained with anti-CD4, anti-CD8, and an unconjugated hamster IgG anti-TCRβ antibody as described above. Cells were washed, resuspended in Krebs-Ringer solution (DeltaSelect, Munich, Germany), and incubated at 37°C for 3 min, and calcium mobilization was measured via LSRII (Becton Dickinson, Franklin Lakes, NJ, USA). The baseline Ca²⁺ level was determined for 50 s. Stimulation of the TCR was conducted by crosslinking TCRs with 3.75 or 7.5 µg/mL anti-hamster IgG (Jackson ImmunoResearch, West Grove, PA, USA), and Ca²⁺ mobilization was measured for another 2 min. Loading efficiency was determined by stimulation with ionomycin (Sigma-Aldrich).

The TCRβ chain repertoire of thymus and splenic T cells was performed by extracellularly staining as described above. The used antibodies were anti-CD3, anti-CD4, anti-CD8, and antibodies from the Mouse Vβ TCR Screening Panel (Vβ 2, 3, 4, 5.1 and 5.2, 6, 7, 8.1 and 8.2, 8.3, 9, 10b, 11, 12, 13, 14, and 17a; BD Pharmingen).

Recipient CD45.1 wt mice were sublethally irradiated with 9 Gy. After 24 h, they were intravenously reconstituted with a bone marrow mixture from CD45.1 wt donor mice (10%) and CD45.2 lck^cre/+xCMAS^fl/fl donor mice (90%). Seven weeks later, the mice were sacrificed and extracellularly stained for B-cell markers (B220 and IgM) and T-cell markers (CD3, CD4, CD8, and TCRγδ) CD45.1 and CD45.2 as described above.

Statistical analyses were performed using GraphPad Prism software. Unpaired Mann–Whitney U-test was used to evaluate significance, which was plotted only for the relevant data. Statistical data are presented as mean ± SD.

We generated T cell-specific Cmas KO mice by crossing the Cmas^fl/fl mice (15) with lck-cre transgenic mice. The lck-cre line is a transgenic mouse line in which the cre is driven by the proximal lck promoter that is turned on in the DN stage of T-cell development in the thymus. It was reported that the cre expression in this mouse line affects T-cell development on its own (14). We therefore used for most of our experiments two control mice: Cmas ^fl/fl mice and lck-cre mice (with a wild-type Cmas gene). We indeed detected a mild reduction of peripheral αβ T-cell numbers in mice where the lck-cre allele was present on its own ( Figure 1B ). In the thymus of T cell-specific Cmas KO mice, we did not observe any changes in DN, DP, or SP T-cell populations, when compared to the two controls ( Figure 1A ). In the spleen, however, we found a reduction of CD4⁺ and CD8⁺ T cells when compared to both controls ( Figure 1B ). A similar, even stronger reduction of CD4⁺ and CD8⁺ T cells was observed in the lymph nodes of T cell-specific Cmas KO mice ( Supplementary Figure 1 ). The T-cell deficiency specifically affected the classical αβ T cells, but not the γδ T cells, which showed rather higher numbers in the spleen ( Figure 1B ). This phenotype can be attributed to the presence of the lck-cre allele, as it was described before that in this transgenic line, γδ T cells are elevated in relative proportion, when compared to αβ T cells (18).

We performed lectin staining by flow cytometry to determine the degree of loss of Sia on different T-cell populations of T cell-specific Cmas KO mice. For this analysis, T-cell subpopulations of Cmas ^fl/fl, Cmas ^wt × lck-cre, and Cmas ^fl/fl × lck-cre mice were stained either with SNA, specific for α2,6-linked Sia or with MAA II binding α2,3-linked Sia. We observed downregulation of both α2,6-linked Sia and α2,3-linked Sia during T-cell differentiation in the thymus of T cell-specific Cmas KO mice ( Figures 2A, C ). While DN cells still expressed near normal levels of sialoglycans, cell surface Sia was increasingly lost in the DP stage. SP CD4⁺ or CD8⁺ thymic T cells of T cell-specific Cmas KO mice had lost almost all α2,6-linked Sia, and thus resembled a control mouse with an ST6Gal-I deficiency, the sialyltransferase responsible for this linkage (12). Cmas KO SP T cells additionally lost α2,3-linked Sia ( Figures 2A, C ). The loss of sialoglycans in the thymus affected only SP αβ T cells, but not γδ T cells ( Figure 2A ). It is known that the proximal lck promoter is not effective in γδ T cells, therefore these cells are spared from cre deletion of target genes (18).

When we performed similar lectin staining on CD4⁺ and CD8⁺ T cells in the spleen of T cell-specific Cmas KO mice, we surprisingly found only a very minor reduction of both α2,6-linked Sia and α2,3-linked Sia. The mean reduction of Sia on splenic T cells was between 0 and 50% ( Figure 2B ). The primary lectin staining showed sometimes two populations, one with wild-type levels of Sia, and another one with downregulation Sia ( Figure 2D ). This indicated heterogeneous populations. We also used the lectins ECA and PNA recognizing terminal β1,4-linked galactose or β1,3-linked galactose, respectively. In wild-type T cells, most galactose residues are masked by a terminal sialic acid and therefore cannot be bound by these lectins. Accordingly, we observed increased ECA and PNA binding on SP T cells in the thymus of T cell-specific Cmas KO mice, but not on control T cells ( Supplementary Figure 2A ). In contrast, ECA and PNA binding was only weakly increased on splenic T cells of T cell-specific Cmas KO mice ( Supplementary Figure 2B ).

In order to find out whether the genetic deletion of Cmas by the lck-cre is complete, DN, DP, and SP CD4⁺ or CD8⁺ T cells from the thymus or CD4⁺ or CD8⁺ T cells and CD3⁻ non-T cells were sorted and tested by a genomic PCR that detects the Cre-mediated deletion of exon 4 of the Cmas gene. As expected, the deletion was detected at approximately 50% in the DN stage of T cell-specific Cmas KO mice and at approximately 100% in the DP and SP T-cell stages of the thymus. However, only a minor Cmas exon 4 deletion was detected in CD4⁺ and CD8⁺ T cells of the spleen ( Figure 2E ). In the spleen of T cell-specific Cmas KO mice, most T cells carried a Cmas wild-type allele. This indicated a leakiness of the lck-cre-mediated Cmas deletion in the thymus and a strong selection of WT T cells with normal sialoglycan levels into the spleen.

From the lectin staining and genomic PCR analysis, we concluded that the peripheral T-cell repertoire of T cell-specific Cmas KO mice is a mixture of Sia-deficient and wild-type T cells that are preferentially selected. We determined the composition of naïve versus effector and memory T cells with CD62L/CD44 staining. We observed a strong reduction of naive T cells but normal numbers of central memory and effector memory-like T cells in T cell-specific Cmas KO mice ( Figure 3 ). The higher proportion of memory and effector memory T cells in these mice may indicate a selection of certain T-cell subpopulations into the periphery, which could potentially affect the TCR repertoire of the remaining T cells in the periphery. In order to exclude this, the TCR repertoire of thymic and splenic T cells was determined by staining with a panel of Vβ-specific antibodies. We did not detect any differences in the TCR Vβ repertoire in the CD4⁺ or CD8⁺ T-cell population in the thymus or spleen of T cell-specific Cmas KO mice, compared to the two control mouse lines ( Figure 4 ). We furthermore observed that in both the thymus and spleen, the total number of regulatory T cells was reduced in T cell-specific Cmas KO mice ( Supplementary Figure 3 ).

In order to examine whether the TCR signaling of Cmas KO T cells is changed, SP T cells of the thymus were stimulated with anti-TCR-β, followed by an anti-IgG crosslinking, and the Ca²⁺ mobilization was measured. In this experiment, we used SP cells of the thymus because the depletion of the Cmas allele is almost complete in this population ( Figure 2E ). The Ca²⁺ response in both CD4⁺ and CD8⁺ T cells was impaired in T cell-specific Cmas KO mice ( Supplementary Figure 4 ). We concluded that the loss of sialoglycans on T cells affected the TCR signaling capacity.

In order to determine the mechanism responsible for the reduced T-cell numbers in peripheral organs of T cell-specific Cmas KO mice, we first examined the role of complement in this pathway. α2,3-Linked Sia can bind to complement factor H to inhibit complement activation (4), and this inhibition could be missing in T cell-specific Cmas KO mice, explaining a T-cell loss by complement attack. In order to determine this, we crossed Cmas ^fl/fl × lck-cre mice with complement C3-deficient mice (C3^−/−). When we compared the splenic CD4⁺ and CD8⁺ T-cell numbers of Cmas ^fl/fl × lck-cre × C3^−/− mice to the T-cell numbers of Cmas ^fl/fl × lck-cre mice, we did not find a rescue of the T-cell defect ( Figure 5 ). We therefore conclude that complement attack is not the cause of the reduced T-cell numbers in the spleen of T cell-specific Cmas KO mice.

Next, we examined whether induction of apoptosis is responsible for the peripheral T-cell defect in T cell-specific Cmas KO mice. Spontaneous apoptosis was the main mechanism for the B-cell deficiency of B cell-specific Cmas KO mice (7). In order to analyze this, thymic T cells or splenic T cells were intracellularly stained with an antibody that recognizes activated caspase 3. Caspase 3 is an effector caspase that is activated by both intrinsic and extrinsic apoptosis pathways. In the thymus, no increased apoptosis levels were detected in DN, DP, or SP T cells of Cmas ^fl/fl × lck-cre mice, when compared to lck-cre controls ( Figure 6B ). However, in the spleen, both naïve CD4⁺ and naïve CD8⁺ T cells show a strongly increased proportion of cells with activated caspase 3. Central memory and effector memory T cells hardly showed this phenotype, but a significant increase of activated caspase 3 was also detected in CD4⁺ central memory T cells, when compared to the lck-cre controls ( Figures 6A, C ). This indicated an ongoing apoptosis in naïve peripheral T cells of T cell-specific Cmas KO mice, affecting mainly this peripheral T-cell population, which is strongly reduced in these mice. Supporting these data, naïve splenic T cells of Cmas ^fl/fl × lck-cre mice showed a higher exposition of the apoptosis marker phosphatidylserine on the outer layer of the cell membrane, indicating early apoptotic cells ( Supplementary Figure 5 ). By this analysis, also a slightly higher number of apoptotic thymocytes of Cmas ^fl/fl × lck-cre mice were detected.

Peripheral T cells lacking sialoglycans may be deleted by cells bearing the asialoglycoprotein receptor (Ashwell–Morell receptor) that is involved in deleting desialylated platelets and clearing desialylated proteins. This receptor is expressed by hepatocytes in the liver (19, 20). We therefore examined whether T cells of Cmas ^fl/fl × lck-cre mice accumulate in the liver. However, we did not find higher T-cell numbers in the liver of Cmas ^fl/fl × lck-cre mice, when compared to lck-cre controls ( Supplementary Figure 6 ).

Since we observed a strong selection of wild-type T cells without deleted Cmas alleles in peripheral organs, we designed a mixed bone marrow (BM) transfer experiment with Cmas KO and wild-type BM cells to address their relative role in T-cell maturation. Since we expected a preference for wild-type precursors over Cmas KO precursor cells, we mixed the bone marrow of wild-type mice with the allelic marker CD45.1 and that of Cmas ^fl/fl × lck-cre mice with the marker CD45.2 in a ratio of 10% (WT) to 90% (KO). The mixed BM cells were transferred into sublethally irradiated recipient CD45.1 mice and analyzed 7 weeks after reconstitution ( Figure 7A ). The B-cell population served as a control population after the transfer. We observed a similar CD45.1/CD45.2 ratio of B cells in the thymus and spleen of the recipient mice 7 weeks after cellular transfer, as given as cellular input at day 0 ( Figures 7B, C ). Since we did not detect a much higher fraction of CD45.1 B cells than in the mixed BM donor cells, we conclude that the hematopoietic cells of the CD45.1 recipients were effectively deleted by irradiation; however, a small contribution of recipient cells cannot be excluded. For the thymocyte populations, we observed a similar ratio of 12% wild-type (CD45.1) cells in the DN population, as in the input. The percentage of wild-type cells in the thymus then constantly increased in further development in the DP to 30% and in the SP population to approximately 50% ( Figures 7B, C ). A strong competitive advantage of wild-type versus Cmas KO T cells was found in the spleen with a ratio of over 97% wild-type cells to less than 3% Cmas KO T cells. γδ T cells did not show this bias in cell numbers ( Figures 7B, C ). The remaining CD4⁺ T cells derived from the CD45.2 allele (Cmas KO) had an almost 100% expression level of sialic acid, as indicated by SNA staining, while the CD8⁺ T cells showed an expression level of 30% when compared to wild-type T cells ( Figure 7D ). This showed again the leakiness of the Cmas deletion in this mouse. The bone marrow chimeras demonstrated that in such a competitive situation, the Cmas KO αβ T cells are strongly out-competed by wild-type T cells. Thus, Cmas KO T cells are not selected or cannot survive in peripheral lymphatic organs.