Thymic CD4 T cell selection requires attenuation of March8-mediated MHCII turnover in cortical epithelial cells through CD83

CD83, an evolutionary conserved immunoglobulin superfamily member, was discovered as an activation marker on DCs. Besides, CD83 is also induced in activated T and B cells and is constitutively expressed by thymic epithelial cells (TECs). The physiological significance of CD83 expression and its mode of action in these diverse cellular contexts are only beginning to emerge (Prechtel and Steinkasserer, 2007; Breloer and Fleischer, 2008; Prazma and Tedder, 2008).

Remarkably, CD83 harbors the capacity to exert distinct cell-intrinsic and -extrinsic functions through discrete protein domains. In DCs, CD83’s transmembrane (TM) domain stabilizes surface MHCII by opposing the association of MHCII with the ubiquitin ligase March1, thereby interfering with MHCII ubiquitination and internalization (Tze et al., 2011). An analogous mechanism explains how CD83 enhances CD86 expression on mature DCs (Baravalle et al., 2011). Of note, the full spectrum of targets that are controlled by CD83 in this manner remains to be determined and may vary between cell types. Besides this cell-autonomous function, CD83 may transmit immune regulatory signals in trans during intercellular interactions or, when present in soluble form, even systemically. Loss- or gain-of-function approaches suggested a costimulatory function of CD83 (Kruse et al., 2000; Prechtel et al., 2007). Experiments using various species of recombinant soluble CD83 (sCD83) indicated that CD83’s extracellular (EC) domain can modulate several biological processes. sCD83 interfered with DC maturation in vitro and inhibited DC-dependent allogeneic and peptide-specific T cell proliferation (Lechmann et al., 2001). When administered in vivo, the EC domain of CD83 prevented the induction of experimental autoimmune encephalomyelitis and other autoimmune diseases (Zinser et al., 2004; Starke et al., 2013; Eckhardt et al., 2014). Importantly, sCD83 of natural derivation is found in DC and B cell supernatants and in the serum, whereby its origin, for instance via generation of alternative transcripts or shedding from the membrane, remains to be clarified (Hock et al., 2001; Dudziak et al., 2005). The nature of the presumed CD83 receptor remains enigmatic.

The most striking phenotype of CD83-deficient mice is a dramatic defect in thymic CD4 T cell selection (Fujimoto et al., 2002; Kuwano et al., 2007). An identical phenotype was reported for defective CD83 alleles generated through N-ethyl-N-nitrosourea mutagenesis (García-Martínez et al., 2004; Tze et al., 2011). Impaired CD4 T cell selection in Cd83^−/− mice reflects a requirement for CD83 in TECs (Fujimoto et al., 2002). Importantly, control of MHCII was considered an unlikely explanation for CD83’s role in CD4 T cell selection because MHCII hemizygous mice, despite showing a similar reduction in MHCII on TECs as Cd83^−/− mice, generate normal CD4 T cell numbers (Kuwano et al., 2007). Hence, it was suggested that CD83, conceivably through interacting with a CD83 receptor on thymocytes or via an as yet undefined cell-intrinsic function in TECs, provides a thymocyte differentiation signal that is mechanistically separated from MHCII interactions (Lüthje et al., 2006; Kuwano et al., 2007; Breloer and Fleischer, 2008). In the present study, we have addressed whether CD83 might directly transmit essential signals from TECs to developing thymocytes via its EC domain or whether CD83’s essential role in the thymus reflects a TEC-intrinsic effect through the control of MHCII or possibly also other modulators of CD4 T cell selection.

Because we used a novel Cd83^−/− strain (Krzyzak et al., 2016), we first verified the reduction of CD4 single-positive (SP) thymocytes (Fig. 1 A). The thymic phenotype of Cd83^−/− mice was recapitulated in WT→Cd83^−/− BM chimeras, confirming the crucial requirement for CD83 in TECs (Fig. 1 B). We assessed two alternative explanations for the diminution in CD4 SP cell numbers: (1) CD4 T cells selected in the absence of CD83 may be subject to excessive negative selection, or (2) the paucity of CD4 SP cells may reflect a genuine defect in positive selection. In MHCII^−/− (H2-ab1^−/−)→WT BM chimeras, lack of negative selection by hematopoietic APCs results in an enlarged CD4 SP compartment (Fig. 1 C). MHCII^−/−→Cd83^−/− chimeras displayed a slight increase in the CD4 compartment as compared with MHCII^+/+→Cd83^−/− chimeras. However, their CD4 compartment was not restored to the values observed in WT recipients (Fig. 1 C). Selection of CD4 T cells expressing two anti-foreign TCR transgenes specific for human C-reactive protein or OVA was dramatically reduced in Cd83^−/− BM recipients (Fig. 1, D and E). Together, these findings indicated that CD83 deficiency did not result in excessive negative selection but affected the size of the CD4 SP compartment through a defect in positive selection.

Two different modes of action were conceivable of how CD83 may support CD4 T cell selection: (1) in trans, through binding and signaling of its EC domain to an as yet unknown receptor on thymocytes or (2) in cis, through protecting substrates of ubiquitin ligases in TECs (Tze et al., 2011). To address this issue, we established an approach to express truncated or chimeric CD83 variants in Cd83^−/− TECs in vivo (Travers et al., 2001; Aichinger et al., 2012). Reaggregation thymus organ cultures (RTOCs) from fetal TECs that had been transduced with bicistronic lentiviral vectors encoding CD83 variants and GFP were transplanted under the kidney capsule of recipient mice (Fig. 2 A). Such organoids are seeded by host-derived progenitors so that a steady-state flux through T cell differentiation was established within ∼4 wk (Anderson and Jenkinson, 2007). We achieved transduction rates of 20−60%, with equal efficiency in cortical TECs (cTECs) and medullary TECs (mTECs; Fig. 2 B) and stable expression of virus-encoded GFP for at least 5 wk (Fig. 2 C).

The coexistence of transduced and nontransduced TECs in these RTOCs represented a caveat for the interpretation of gain-of-function experiments with CD83 or variants thereof. To ask whether CD83 expression by only a fraction of TECs was sufficient for normal CD4 T cell selection, we generated mixed RTOCs with titrated ratios of WT and Cd83^−/− TECs to emulate different transduction efficiencies (Fig. 2 D). Even at the lowest ratio of WT TECs tested (1:9), a marked increase in CD4 SP cells occurred.

The CD4 SP compartment in transplanted RTOCs transduced with full-length CD83 (construct #2) was indistinguishable in size from the CD4 SP population in WT RTOCs (Fig. 2, E and F). The respective roles of the EC, TM, and cytoplasmic domains of CD83 were then dissected through introduction of truncated or chimeric forms of CD83 (Fig. 2 F). Expression of a chimeric CD83 molecule harboring the TM and cytoplasmic domain, yet bearing the EC portion of human CD4 (hCD4; construct #3), rescued CD4 T cell selection. Likewise, a truncated CD83 molecule (construct #6) lacking the cytoplasmic domain restored the CD4 SP compartment to normal size. In contrast, a construct (#4) containing the EC domain of CD83 yet bearing a TM domain derived from hCD4, despite a higher transduction efficiency and very similar expression level, failed to restore CD4 T cell selection (Fig. 2, F and G). The crucial requirement for the TM domain was confirmed with a second construct (#5). These observations identified the TM domain as the minimal functional unit through which CD83 supports CD4 T cell selection.

We next addressed the expression pattern of CD83 in cTECs and mTECs. Surface CD83 was substantially more abundant on cTECs compared with mTECs (Fig. 3 A). This differential expression was confirmed at the level of RNA expression (Fig. 3 B). Consistent with a cTEC-specific function of CD83, surface MHCII was diminished on cTECs but not mTECs from Cd83^−/− mice (Fig. 3 C). Surface MHCI was not affected by lack of CD83 (Fig. 3 C).

DCs and B cells from mice lacking CD83 or carrying a mutation in CD83’s TM domain display reduced surface MHCII owing to accelerated MHCII turnover (Kuwano et al., 2007; Tze et al., 2011). We hypothesized that the reduced MHCII on CD83-deficient cTECs may likewise stem from increased MHCII internalization. To test this idea, purified cTECs from WT or Cd83^−/− mice were cultured in vitro with or without brefeldin A (BFA; Fig. 3 D). BFA is an ER to Golgi transport inhibitor that prevents the delivery of newly synthesized proteins to the cell surface, thus abrogating de novo MHCII routing to the surface. With WT cTECs, BFA did not have any discernable effect on the surface MHCII, consistent with the previously reported remarkably long t_1/2 of MHCII complexes on TECs (Müller et al., 1993). In contrast, surface MHCII on Cd83^−/− cTECs was significantly reduced after culture in the presence of BFA (Fig. 3 D). BFA treatment resulted in a similar degree of MHCII reduction on WT and Cd83^−/− mTECs (Fig. 3 E). Thus, CD83 controls MHCII surface levels on cTECs but not mTECs through opposing MHCII turnover. However, the mechanistic basis of CD83’s influence on MHCII remained open, and most importantly, it was unclear whether impaired CD4 T cell selection in the absence of CD83 was causally (and solely) related to its effect on MHCII.

In immature DCs, MHCII is ubiquitinated at lysine 225 of the I-Aβ chain, earmarking it for lysosomal degradation (Shin et al., 2006). During DC maturation, up-regulation of CD83 antagonizes this process and hence enhances MHCII surface expression (Tze et al., 2011). To test whether decreased MHCII on Cd83^−/− cTECs may similarly be caused by ubiquitination-dependent destabilization of MHCII, we used mice in which the I-Aβ chain is modified with a lysine to arginine substitution at position 225 (MHCII^K225R; McGehee et al., 2011). MHCII surface expression on MHCII^K225R cTECs was indistinguishable from that on WT cTECs (Fig. 4 A), and the size of the CD4 SP compartment in MHCII^K225R mice was very similar to WT controls (Fig. 4 B). Superimposition of CD83 deficiency with expression of ubiquitination-resistant MHCII not only restored normal MHCII expression on cTECs, but also rescued CD4 T cell development (Fig. 4, A and B). These findings supported the idea that CD83 opposes the ubiquitination-dependent down-modulation of MHCII in cTECs and established that the detrimental effect of CD83 deficiency on CD4 T cell selection is solely attributable to CD83’s effect on MHCII.

MHCII ubiquitination in DCs and other hematopoietic APCs is mediated by the E3 ligase March1 (Matsuki et al., 2007; De Gassart et al., 2008). We therefore asked whether CD83 controls MHCII in cTECs through opposing March1. However, unlike several hematopoietic APC subsets, neither cTECs nor mTECs expressed March1 mRNA above the detection limit (Fig. 5 A), and MHCII levels on March1^−/− cTECs were unaffected (Fig. 5 B). Moreover, combined deficiency in CD83 and March1 did not restore MHCII surface expression on cTECs, and the CD4 SP compartment in Cd83^−/−March1^−/− mice was similarly reduced as in Cd83^−/− mice (Fig. 5 C). Thus, the critical requirement of CD83 in cTECs for MHCII expression and CD4 T cell selection was not a reflection of March1 antagonism.

The March protein family member most closely related to March1 is March8 (Ishido et al., 2009). Forced expression of March8 in hematopoietic APCs reduces surface MHCII, and transgenic overexpression of March8 in TECs results in a defect in CD4 T cell selection reminiscent of that in CD83-deficient mice (Ohmura-Hoshino et al., 2006). However, the physiological targets of March8 remain unknown (Ishido et al., 2009). For instance, unlike March1, March8 is not dynamically regulated during DC maturation, and hence, it is not considered a critical player in MHCII regulation in hematopoietic APCs (De Gassart et al., 2008). Within different types of thymic APCs, March8 mRNA was most strongly expressed in cTECs (Fig. 6 A). Indeed, superimposing CD83 and March8 deficiency restored MHCII levels on cTECs (Fig. 6 B), and the size of the CD4 SP compartment in Cd83^−/−March8^−/− mice was indistinguishable from that in WT mice (Fig. 6 C).

In sum, we show that CD83’s crucial role for CD4 T cell selection reflects its capacity to attenuate MHCII turnover in cTECs by counteracting March8-mediated MHCII ubiquitination. MHCII hemizygous mice, despite a similar reduction of MHCII on TECs as in Cd83^−/− mice, generate normal CD4 T cell numbers. Thus, a relatively slow MHCII turnover rather than absolute MHCII surface levels as such may be crucial for efficient CD4 T cell selection. This adds to our understanding of how cTECs have adapted their cell biology for T cell selection and how their MHC ligandome is shaped (Klein et al., 2014). The kinetics of cellular interactions during positive selection are only beginning to emerge, but it is tempting to speculate that a long MHCII half-life on cTECs provides a platform for stable and efficient selection events (Müller et al., 1993). Intriguingly, the efficient selection of CD4⁺ but not CD8⁺ T cells seems to be contingent upon prolonged or even repetitive interactions with selecting ligands on cTECs (Singer et al., 2008).

Viral strategies are commonly used to manipulate gene expression in hematopoietic cell types. The transplantation of RTOCs generated with transduced embryonic TECs represents an analogous strategy to study gene function in thymic epithelium in vivo (Travers et al., 2001; Aichinger et al., 2012). By expressing chimeric CD83 variants in Cd83^−/− TECs, we identified its TM domain as the minimal functional unit that is sufficient for efficient CD4 T cell selection. The fact that CD83’s EC domain is dispensable indicates that in this context, CD83 operates in an exclusively cTEC-intrinsic, indirect manner by controlling MHCII. Previous data were interpreted to indicate that CD83 may not only quantitatively control the selection of CD4 T cells, but may at the same time also directly condition their functional properties (García-Martínez et al., 2004; Lüthje et al., 2006). Whereas it has been hypothesized that this may occur through interactions of CD83 with an unknown ligand on developing T cells, we deem it more likely that functional alterations in T cells of CD83-mutant mice reflect an indirect effect of aberrant MHCII expression.

The finding that MHCII is a physiological March8 substrate suggests an intriguing division of labor between March1 and March8 in the control of MHCII in hematopoietic APCs versus TECs. Although MHCII appears to be the sole relevant March8 substrate in the context of CD4 T cell selection, it is conceivable that March8 also targets other substrates in cTECs. Against this background, the possibility that CD83 may selectively interfere with the ubiquitination of MHCII but not of other March8 substrates needs to be considered. This might explain the paradox that cTECs constitutively express both a promoter and an attenuator of MHCII turnover, although shutting off both would appear a more economical way to achieve stable MHCII expression.

March1^−/− (Matsuki et al., 2007) and I-Aβ-K225R-EGFP mice have been described previously (McGehee et al., 2011). To generate the Cd83^−/− strain, mice carrying a floxed Cd83 allele (Krzyzak et al., 2016) were mated to germline-deleting E2a-Cre mice. March8^−/− mice were generated as described in Fig. S1. All mice were on a C57BL/6 background. Animal studies were approved by local authorities (Regierung von Oberbayern; Az 7–08 and 142–13).

The following antibodies were purchased from eBioscience, BioLegend, or BD: CD4 (GK1.5), CD8α (53-6.7), I-A/I-E (M5/114.15.2), H-2K^b (AF6-88.5.5.3), epithelial cell adhesion molecule (EpCAM; G8.8), CD45 (Ly-5), CD11c (N418, HL3), Sirpα/CD172a (P84), Ly51 (6C3), CD83 (Michel-19), CD80 (16-10.A1), CD19 (6D5), and CD317/PDCA-1 (927), conjugated to different fluorochromes or biotin. Dead cells were excluded by gating on DAPI-negative cells. Surface stainings were performed according to standard procedures with 1–2 × 10⁶ cells in a volume of 50 µl. FACS measurements were performed on a cell analyzer (FACS Canto II; BD) and analyzed using FlowJo software (Tree Star). Cell sorting was performed on a FACS Aria III cell sorter (BD).

Truncated or chimeric CD83 constructs have been described before and were provided by C. Goodnow (Garvan Institute of Medical Research, Darlinghurst, Australia) and K. Horikawa (John Curtin School of Medical Research, Australian National University, Acton, Australia). Construct numbers in Fig. 2 correspond to the following construct names in Tze et al. (2011): #1, hCD4; #2, CD83 WT; #3, hCD4 chimera 1; #4, CD83 chimera 1; #5, CD83 chimera 2; and #6, CD83 ΔC. Constructs were subcloned into the lentiviral vector FUGW followed by a T2A peptide and enhanced GFP (EGFP; pFUGW-T2A-GFP) provided by T. Brocker (Ludwig-Maximilians-University, Munich, Germany). For lentivirus production, confluent HEK293FT cells in a 10-cm culture dish were transiently transfected with 8 µg of the respective lentiviral vector, 6 µg PAX2 packaging plasmid, and 6 µg of VSVG envelope plasmid using a standard calcium phosphate protocol. The supernatant was collected and replaced with fresh medium at 48 and 72 h after transfection, pooled, and centrifuged at 14,000 rpm for 4 h to pellet the virus particles. The pellets were resuspended in 3–4 ml of fresh DMEM and stored at −80°C.

Single cell suspensions of E14–E16 fetal thymic lobes were prepared by collagenase/dispase I (Roche) digestion. CD45⁺ cells were depleted using magnetic-activated cell-sorting beads (Miltenyi Biotec) according to standard procedures. Aliquots of 10⁶ cells were infected with lentivirus (mean fluorescence intensity [MFI] of 1) in 1 ml DMEM (8% FCS) supplemented with 10 µg/ml polybrene. After 3 h, cells were washed three times, spun down, and resuspended in ∼1 µl. The cell slurry was deposited onto 0.45-µm nylon membranes (EMD Millipore) floating in 6-well plates containing 6 ml DMEM (8% FCS). RTOCs were incubated for 48 h before transplantation.

Thymuses of 3–5-wk-old animals were cut into pieces, and thymocytes were mechanically released by pipetting up and down. The supernatant containing thymocytes was discarded. The thymus fragments were digested with 0.5 U/ml Liberase Thermolysin medium (Roche) at 37°C in two consecutive rounds of 15 min. Cells were washed and resuspended in 1 ml of high-density Percoll (ρ = 1.115; GE Healthcare) and overlaid with 1 ml of low-density Percoll (ρ = 1.055) followed by a layer of 1 ml RPMI. The gradient was centrifuged at 1,350 g for 30 min at 4°C (without brake). The upper interphase containing the low-density cell fraction was harvested, washed, and stained for FACS sorting. TECs, DC subsets, and B cells were sorted according to surface expression of CD45, Ly51, EpCAM, CD80, CD19, CD11c, PDCA-1 (CD317), and Sirpα (CD172a) as follows: cTECs, CD45⁻EpCAM⁺Ly51⁺CD80⁻; mTECs, CD45⁻EpCAM⁺Ly51⁻CD80⁺; B cells, CD45⁺CD11c⁻CD19⁺; plasmacytoid DCs, CD45⁺CD11c^intCD317⁺; Sirpα⁺ classical DCs, CD45⁺CD11c^highCD172a⁺; and Sirpα⁻ classical DCs, CD45⁺CD11c^highCD172a⁻.

The low-density fraction of enzymatically prepared thymic cells (see the previous section) was plated in a 96-well plate (2 × 10⁵ cells/well in 200 µl) in DMEM (8% FCS) with or without 5 µg/ml BFA (Sigma-Aldrich). After 16-h incubation at 37°C, cells were collected and stained for FACS analyses. In parallel, some cells were kept for 16 h at 4°C as controls.

BM was depleted of T cells using biotinylated CD8α and CD4 monoclonal antibodies and streptavidin magnetic-activated cell-sorting beads (Miltenyi Biotec). Recipient mice were irradiated with 2 × 550 rad and reconstituted with 10⁷ BM cells.

RNA was isolated from thymic APCs using the Arcturus PicoPure RNA isolation kit (Applied Biosystems). Purified RNA was subjected to DNase digestion (QIAGEN) and reverse transcribed using the iScript cDNA synthesis kit (Bio-Rad Laboratories) containing both oligo deoxythymidine and random hexamer primers. Quantitative PCR reactions were performed on a real-time thermal cycler (CFX96 C1000; Bio-Rad Laboratories) using the SsoFast EvaGreen Supermix (Bio-Rad Laboratories). Fluorescence was recorded at the annealing step, and relative expression levels were calculated with the comparative cycle threshold method, using β-actin as the housekeeping gene. The primers were March1-5′, AAGAGAGCCCACTCATCACACC; March1-3′, ATCTGGAGCTTTTCCCACTTCC; March8-5′, AGTAGTCCTCCATCCACGAC; March8-3′, GATGACGAGAGCCCTCTGAT; CD83-5′, GCCTCCAGCTCCTGTTTCTA; CD83-3′, AGTGTTTTGGATCGTCAGGG; β-actin–5′, GCCTTCCTTCTTGGGTAT; and β-actin–3′, GGCATAGAGGTCTTTACGG.

Unless indicated otherwise, statistical significance was assessed using the two-tailed unpaired Student’s t test with Welch’s correction for unequal variances.

Fig. S1 provides details on the targeting of the March8 gene locus.

C. Goodnow and K. Horikawa generously provided CD83 constructs.

J. von Rohrscheidt, E. Petrozziello, and L. Klein were supported by the Deutsche Forschungsgemeinschaft (DFG; grant SFB1054, project parts A01 and Integrated Research Training Group [IRTG]). E. Petrozziello was supported by a PhD fellowship from the Boehringer Ingelheim Fonds. J. Nedjic was supported by a Marie Curie postdoctoral fellowship from the European Union. L. Krzyzak and A. Steinkasserer were supported by the DFG via the Graduiertenkolleg (GK) 1660 (project part B02). A. Steinkasserer was supported by the DFG via the Sonderforschungsbereich 1181 (project part B03).

The authors declare no competing financial interests.