The proper regulation of ICOS and ICOS ligand (ICOSL) has been shown to be essential for maintaining proper immune homeostasis. Loss of either protein results in defective humoral immunity, and overexpression of ICOS results in aberrant Ab production resembling lupus. How ICOSL is regulated in response to ICOS interaction is still unclear. We demonstrate that a disintegrin and metalloproteinase (ADAM)10 is the primary physiological sheddase of ICOSL in mice and humans. Using an in vivo system in which ADAM10 is deleted only on B cells, elevated levels of ICOSL were seen. This increase is also seen when ADAM10 is deleted from human B cell lines. Identification of the primary sheddase has allowed the characterization of a novel mechanism of ICOS regulation. In wild-type mice, interaction of ICOS/ICOSL results in ADAM10-induced shedding of ICOSL on B cells and moderate ICOS internalization on T cells. When this shedding is blocked, excessive ICOS internalization occurs. This results in severe defects in T follicular helper development and TH2 polarization, as seen in a house dust mite exposure model. In addition, enhanced TH1 and TH17 immune responses are seen in experimental autoimmune encephalomyelitis. Blockade of ICOSL rescues T cell ICOS surface expression and rescues, at least in part, T follicular helper numbers and the abnormal Ab production previously reported in these mice. Overall, we propose a novel regulation of the ICOS/ICOSL axis, with ADAM10 playing a direct role in regulating ICOSL, as well as indirectly regulating ICOS, thus controlling ICOS/ICOSL-dependent responses.
A disintegrin and metalloproteinases (ADAMs) are a family of zinc-dependent proteinases that can mediate intramembrane proteolysis and ectodomain shedding of membrane proteins. Of the ADAM family proteins, the proteolytic domains of ADAM10 and ADAM17 share the highest homology, often resulting in the ability to cleave overlapping substrates (1, 2). ADAM10 has been shown to act in many paracrine signaling mechanisms and is responsible for cleaving numerous substrates, including Notch receptors, delta-like 1, IL-6R, CXCL16, and CD23 (3, 4). We have shown that loss of ADAM10 on B cells (ADAM10B−/−) results in loss of the marginal zone B cell compartment, disorganized secondary lymphoid architecture, decreased Ag-specific Ab (5), and decreased airway hyperresponsiveness and eosinophilic infiltration in two models of allergic airway disease (6, 7).
ICOS on T cells and ICOS ligand (ICOSL), which is expressed on APCs, have been shown to be essential for T follicular helper (TFH) and TH2 development and activity (8–11). TFH cells are essential for productive germinal center (GC) responses, providing help to B cells undergoing class-switch recombination and somatic hypermutation, as well as being critically involved in GC B cell differentiation into memory B cells and long-lived plasma cells (8, 12). Deficiency of ICOS or ICOSL essentially abolishes T-dependent humoral immune responses (9, 11). There have been several studies illustrating the regulation of ICOS (13–15), particularly at the mRNA level, as well as the cleavage of ICOSL. In particular, ADAM17 was shown to cleave ICOSL in response to PMA and BCR cross-linking (16). However, ADAM17 was not involved in ICOS-induced shedding of ICOSL, and constitutive ICOSL levels were unchanged. This indicates that a second, unknown protease is involved in physiological B cell activation in GCs, as well as the cross-talk between ICOS and ICOSL. Given these data, understanding the regulation of these proteins is quite important.
In this article, we identify the physiologically relevant ICOSL sheddase to be ADAM10. We show that, although recombinant ADAM10 and ADAM17 can cleave recombinant ICOSL, only ADAM10B−/− mice have significantly elevated ICOSL on B cells. Loss of both proteases in B cells (ADAM10/17B−/−) marginally increases ICOSL levels over the loss of ADAM10 alone, suggesting a secondary role for ADAM17 in ICOSL regulation. In these mice, the overexpression of surface ICOSL results in the internalization and degradation of T cell ICOS in the absence of TCR stimulation. As a result, the mice lack proper TFH and TH2 effector cell populations postimmunization, explaining the defective humoral immunity previously reported in ADAM10B−/− mice (5, 6). In addition, increased ICOSL resulted in enhanced TH1 and TH17 T cell activation, as demonstrated by a model of experimental autoimmune encephalomyelitis (EAE). Overall, these studies identify the sheddase of ICOSL following ICOS interaction, as well as present a novel mechanism of ICOS regulation at the posttranslational level. We hypothesize that ligand:receptor interaction causes ICOS internalization following ICOSL shedding by ADAM10. Interfering with this normal regulation gives rise to a phenotype similar to that seen in ICOS−/− mice.
Materials and Methods
Mice were maintained at the Virginia Commonwealth University Animal Facility in accordance with guidelines by the National Institutes of Health and American Association for the Accreditation of Laboratory Animals Care. C57BL/6 ADAM10B−/− mice were generated as previously described (3). In short, loxP sites were inserted to flank exon 9 of Adam10, and these mice were crossed to Cd19-Cre mice (006785; The Jackson Laboratory). Adam17 floxed mice were purchased from The Jackson Laboratory (009597) and crossed to Cd19-Cre mice (006785). ADAM17B−/− mice were purchased from The Jackson Laboratory. ADAM10/17B−/− mice were generated by crossing ADAM10B−/− and ADAM17B−/− mice. Cre-recombinase− littermates were used as wild-type (WT) controls.
RNA isolation and quantitative PCR
For RNA isolation from cells, cell suspensions were centrifuged at 500 × g for 5 min, washed once with PBS, and centrifuged again. TRIzol (Invitrogen) was used for RNA extraction, per the manufacturer’s instructions. For RNA isolation from tissues, tissues were flash frozen in liquid nitrogen. TRIzol was added to frozen tissue and homogenized. Subsequent RNA isolation was performed according to the manufacturer’s instructions. RNA was quantified using an ND-100 NanoDrop spectrophotometer. One microgram of total RNA was reverse transcribed using SuperScript IV (Thermo Fisher) with oligo(dT20). Primers used in quantitative PCR (qPCR) analysis are as follows: Hprt_forward 5′-CAGGGATTTGAATCACGTTTGTG-3′, Hprt_reverse 5′-TTGCAGATTCAACTTGCGCT-3′, Icosl_forward 5′-CAGCGGCATTCGTTTCCTTC-3′, Icosl_reverse 5′-GTCAGGCGTGGTCTGTAAGT-3′, Icos_forward 5′-TCTAGACTTGCAGGTGTGACC-3′, and Icos_reverse 5′-CAGGGGAACTAGTCCATGCG-3′. In short, qPCR was conducted using a QuantStudio 3 Real-Time PCR System with 45 cycles using PowerUp SYBR Green Master Mix (both from Applied Biosystems). Primers were tested for specificity using melt curve analysis.
Flow cytometry and ImageStream analysis
Single-cell suspensions were created and passed through a 40-μm strainer. For flow cytometry of spleens, RBCs were lysed using ACK lysis buffer (Quality Biological). Two micrograms of anti-mouse CD16/32 (clone 93; BioLegend) was used to block Fc receptors. The following fluorophore or biotinylated Abs were used at the manufacturers’ recommended quantity: PE anti–IL-4 (clone 11B11; BioLegend), PE/Cy7 anti–IL-13 (clone eBio13A; eBioscience), allophycocyanin and PE/Cy7 anti-CD45R/B220 (clone RA3-6B2; BioLegend), allophycocyanin and FITC anti-CD4 (clone GK1.5; BioLegend), PE anti-ICOS (clone C398.4A; BioLegend), PE anti-ICOSL (clone HK5.3; BioLegend), Biotin anti-CXCR5 (clone L138D7; BioLegend), Streptavidin PE/Cy7 (BioLegend), BV421 anti-PD1 (clone 29F.1A12; BioLegend), PE anti-human ICOSL (clone 2D3; BioLegend), PE anti-CD40 (HM40-3; BioLegend), allophycocyanin anti-OX40L (clone HM40-3; BioLegend), PE/Cy7 anti-CD86 (clone GL-1; BioLegend), PE/Cy5 anti-CD80 (clone 16-10A1; BioLegend), PE/Cy7 anti-CD8a (clone 53-6.7; BioLegend), PE anti-CD3e (clone 145-2C11; BioLegend), FITC anti-CD44 (clone IM7; BioLegend), allophycocyanin anti-CD45.1 (clone A20; BioLegend), Alexa Fluor 647 anti-GL7 (clone GL7; BioLegend), PE anti-CD95 (clone 15A7; eBioscience), Alexa Fluor 647 anti–IFN-γ (clone XMG1.2; BioLegend), and BV650 anti–IL 17A (clone TC11-18H10; BD). Flow cytometry data were analyzed using FSC Express 5. For intracellular cytokine staining from lung and draining lymph nodes (dLNs), cells were stimulated using an anti-CD3–coated plate (1 μg/ml) for 4 h in the presence of monensin. Cells were fixed and permeabilized and stained for the indicated targets. Data were acquired using an LSR Fortessa for flow cytometry and Amnis ImageStream. Internalization index was calculated using IDEAS 6.0 software.
B cells isolated from spleens of the indicated strains were stimulated with the following: 25 ng/ml PMA, 2 μM ionomycin, 50 μg/ml α-hemolysin (HLA), 2.5 μg/ml plate-bound ICOS or vehicle controls for each. Cells were stimulated for 1 h at 37°C in complete RPMI. Following stimulation, cells were washed two times with cold PBS. Cells were stained for surface ICOSL for 30 min on ice following Fc receptor blocking. Cells were washed twice with staining buffer, fixed using 2% paraformaldehyde, and analyzed by flow cytometry. Percentage of ICOSL shed was calculated as [(unstimulated ICOSL mean fluorescence intensity [MFI] − stimulated ICOSL MFI)/(unstimulated ICOSL MFI)].
Recombinant cleavage assays
Recombinant His-tagged mouse ICOS-L (100 ng, 8127-B7) and recombinant mouse (rm)ADAM10 (1 μg, 946-AD) or ADAM17 (1 μg, 2978-AD; all from R&D Systems) were incubated together for 24 h at 37°C. To remove N-linked glycosylation, cleaved samples were treated with PNGase F under denaturing conditions, according to the manufacturer’s instructions (New England Biolabs). SDS samples were reduced and run on 12% Bis-Tris gels before transfer onto nitrocellulose blots. Anti-ICOSL (50190-RP01; Sino Biological) and anti-His Tag (clone J099B12; BioLegend) were diluted 1:1000 and incubated for 1 h at room temperature. Secondary Abs (goat anti-mouse IgG1, Goat anti-Rabbit IgG; both from SouthernBiotech) were conjugated to HRP and diluted 1:10,000 in 5% milk block and incubated for 1 h at room temperature. Blots were developed using ECL Western blotting Substrate (Pierce) and read using a Bio-Rad Gel Doc system.
NP31–keyhole limpet hemocyanin immunizations
For footpad immunizations, 10 μg of NP31 conjugated to keyhole limpet hemocyanin (NP31-KLH; Biosearch Technologies) was dissolved in PBS and 4 mg alum (Imject; Sigma) in a volume of 25 μl and injected into each hind footpad. For i.p. immunizations, 10 μg of NP31-KLH was dissolved in 4 mg alum. NP4-BSA and NP25-BSA (15 μg/ml; Biosearch Technologies) were used to coat ELISA plates in PBS for high- and low-affinity NP-specific IgG1. For blocking Ab experiments, mice were administered the indicated amounts of a monoclonal ICOSL blocking Ab (MIL-5733) (17) i.p.
House dust mite model
House dust mite (HDM; Dermatophagoides pteronyssinus) extracts were obtained from Greer Laboratories. Mice were administered 25 μg of HDM extract dissolved in 25 μl of PBS intranasally (i.n.) daily for 4 d, with a 10-d break, followed by an additional four i.n. immunizations. For restimulation of dLNs with HDM extract, mediastinal lymph node (medLN) cells were isolated at day 20 and restimulated with the indicated HDM concentrations at 2 × 106 cells per milliliter. Tag-it Violet (BioLegend) was used to measure proliferation by flow cytometry. For restimulation of T cells from lung tissue, lungs were digested using collagenase I, DNase, and dispase I (Sigma) for 1 h at 37°C. Digested tissue was passed through a 40-μm mesh and washed twice with PBS. T cells were restimulated with plate-bound anti-CD3 (2 μg/ml) for 4 h in the presence of monensin. For T cell–proliferation studies, T cells (1 × 106 per milliliter) were stimulated with 1 μg/ml plate-bound anti-CD3 for 72 h.
Isolated mouse and human B cells were cultured in complete RPMI 1640 containing 10% FBS, 2 mM l-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mM HEPES, 1 mM sodium pyruvate, 2 mM 2-ME, and 1 μg/ml anti-CD40 (5). M12 and 8866 cell lines were cultured in complete RPMI 1640 containing 10% FBS, 2 mM l-glutamine, 100 U/ml penicillin, 100 μg/ml streptomycin, 1 mM HEPES, and 1 mM sodium pyruvate. GI254023X (Sigma), used at 1 μM, is considered specific for ADAM10 and is above the published IC50 of 5 nM (18, 19).
Spleens were homogenized and passed through a 40-μm mesh. Erythrocytes were lysed using ammonium-chloride-potassium lysing buffer (Quality Biological). B cells were isolated using B220-positive selection (Miltenyi Biotec, Auburn, CA), according to the manufacturer’s protocol. CD4+ T cell isolation was conducted using negative selection (BioLegend), according to the manufacturer’s protocol. Human tonsillar B cells were isolated by creating a single-cell suspension, followed by IgD-positive selection.
Ten-micron frozen sections were cut from the excised mouse lymph nodes, fixed in absolute acetone, air-dried, and blocked with serum-free protein block (X0909; Dako), as described. Sections were labeled with PE-conjugated rat anti-mouse ICOSL (clone HK5.3, 107405; BioLegend) and unconjugated rabbit anti-mouse ICOS (ab138354; Abcam), followed by PE-conjugated F(ab)′2 donkey anti-rabbit IgG, Alexa Fluor 647–conjugated rat anti-mouse CD45R/B220 (clone RA3-6B2, 103226; BioLegend), Alexa Fluor 488–conjugated rat anti-mouse CD4 (clone GK1.5, 100423; BioLegend), anti-mouse Lyve-1 eFluor 660 (clone ALY7, 50-0443-82; eBioscience), and FITC-conjugated GL7 (GL7, 562080; BD Pharmingen). The Ab concentrations ranged between 5 and 10 μg/ml. Sections were mounted with VECTASHIELD Antifade Mounting Medium (Vector Laboratories), coverslipped, and examined with a Leica TCS SP2 AOBS confocal laser-scanning microscope. Three lasers (488, 543, and 633 nm) were used; far red emission is shown as pseudoblue. Parameters were adjusted to scan at 1024 × 1024-pixel density and 8-bit pixel depth. Emissions were recorded in three separate channels, and digital images were captured and processed with Leica Confocal Software LCS Lite.
Active EAE was induced, as previously described (20). Briefly, 200 μg of myelin oligodendrocyte glycoprotein (MOG)35–55 was dissolved in PBS and emulsified in CFA with 4 mg/ml heat-inactivated Mycobacterium tuberculosis extract. Mice were immunized at four locations (50 μl each) on the hind flanks. Pertussis toxin was administered i.p. on days 0 and 2. Mice were observed daily and scored as follows: 0, no clinical signs; 1, paralyzed tail; 2, hind limb paresis/loss of coordinated movement; 3, both hind limbs paralyzed; 4, forelimbs paralyzed; and 5, moribund. For restimulation experiments, cells were isolated from the dLNs and cultured with increasing concentrations of MOG35–55. Cells were incubated at 37°C for 90 h. Tag-it Violet (BioLegend) was used to measure proliferation by flow cytometry. For T cell intracellular cytokine staining following MOG restimulation, monensin was added during the last 6 h of culture.
RPMI 8866 cells were transfected with a modified pX330-U6-Chimeric_BB-CBh-hSpCas9 (Feng Zhang [Addgene plasmid #42230]) plasmid (21) containing GFP. Cells were transfected using Lipofectamine and single cell sorted 3 d later. Colonies were grown to confluency and screened for ADAM10 by flow cytometry. The short guide RNA sequences are as follows: Adam10_sg1, 5′-AATTCTGCTCCTCTCCTGGG-3′ and Adam10_sg2, 5′-TTTCAACCTACGAATGAAGA-3′.
ADAM10 is the constitutive regulator of ICOSL levels
Chaimowitz et al. (5) have previously shown that ADAM10B−/− mice have disrupted GC morphology and decreased TFH development following immune activation. Given the importance of costimulatory signals in T cell activation, abnormal expression of CD80, CD86, CD40, ICOSL, and OX40L was investigated using our previously described ADAM10B−/− mice. Only ICOSL was found to be dysregulated. Expression was ∼15-fold higher on ADAM10−/− B cells compared with WT (Fig. 1A, Supplemental Fig. 1A). Interestingly, although we found an increase in ICOSL surface expression, mRNA levels for Icosl dropped significantly (Fig. 1B). Given that ADAM17 has been reported to be involved in ICOSL shedding (16), we also examined B cells that lacked ADAM17 (ADAM17B−/−), as well as B cells from ADAM10/17B−/− mice. Consistent with previous results, there was no difference in ICOSL levels on ADAM17−/− B cells compared with WT (Supplemental Fig. 1B); however, a modest, but significant, additional increase in ICOSL was observed in ADAM10/17−/− B cells (Fig. 1A).
We next wanted to investigate ICOSL regulation in human B cells. The human lymphoblastoid cell line 8866 (22) was found to have high surface expression of ADAM10 but little ICOSL expression (Fig. 1C); thus, this cell line was a good candidate to test the effect of ADAM10 deletion. Using CRISPR-Cas9 8866 cells to delete ADAM10 (Supplemental Fig. 1C), ICOSL levels were enhanced above control 8866 cells when using two guide RNAs to limit the chance of off-target effects (Fig. 1C). Next, human tonsillar B cells or WT mouse B cells were treated with an ADAM10-specific inhibitor (GI254023X) (18), and a time-dependent increase in surface ICOSL (Fig. 1D, F) was seen. CD40 stimulation is known to induce an increase in ICOSL expression (23) and, thus, the data are normalized to CD40 stimulation alone. This may explain why the increase in ICOSL with inhibitor treatment was only 1.5-fold greater in ADAM10B−/− mice compared with 15-fold greater in WT mice (Fig. 1A, 1C versus Fig. 1E, 1F).
ADAM10 and ADAM17 directly cleave ICOSL but differentially regulate it at the cell surface
Given the above results and a previous report (16) that ADAM17 may be involved in the regulation of ICOSL, we sought to conclusively determine whether ADAM10 and ADAM17 can directly cleave ICOSL. To test this, rmICOSL was incubated with rmADAM10 or rmADAM17. After cleavage, N-linked glycosylation was removed with PNGase, and blotting was performed with anti-ICOSL (Fig. 2B) or anti-His Tag (present on the C terminus, Fig. 2A). Both blots indicated the release of a small 3-kDa fragment from the N terminus (fragment 1, Fig. 2A). Additionally, there were several sequential cleavage events, because there are multiple band shifts present, with the N-terminal fragments estimated at 3.3 kDa, 6 kDa (fragment 2, Fig. 2A), and 10 kDa (fragment 3, Fig. 2B). rADAM17 was also His tagged and explains the multiple bands present in lane 3 (Fig. 2A). These small fragments were not recognized by either Ab; they are either further degraded or do not have epitopes for the anti-ICOSL Ab. Additionally, an ADAM10 inhibitor (GI254023X) blocked the cleavage of ICOSL by ADAM10 (data not shown). Thus, ADAM10 and ADAM17 can cleave ICOSL when purified components are used. Blotting with anti-ICOSL (Fig. 2B) resulted in an additional 16-kDa fragment being visible (Fig. 2B, fragment 3).
Although ADAM10 and ADAM17 were able to cleave recombinant ICOSL, we used various stimuli to tease apart the roles of each protease in the regulation of cell-bound ICOSL. Consistent with a previous report (16), treatment of purified WT B cells with PMA resulted in reduced ICOSL shedding in ADAM17−/− and ADAM10/17−/− B cells compared with WT (Fig. 2C), confirming a role for ADAM17 in the PMA-stimulated ICOSL shedding. The increase in shedding for ADAM10B−/− versus WT was expected because ADAM17 levels are increased on these cells (24). Ionomycin is known to activate ADAM10, whereas it only slightly activates ADAM17 (1). Correspondingly, shedding of ICOSL on ionomycin-treated ADAM10−/− B cells was significantly less than on WT B cells. Shedding on ADAM10/17−/− B cells was reduced even further (Fig. 2D). Next, we stimulated purified B cells with HLA, which selectively binds to and activates ADAM10 (25, 26). Treatment with HLA triggered ∼50% shedding of ICOSL in WT B cells, whereas shedding was essentially absent in ADAM10−/− B cells (Fig. 2E).
Of the stimulations known to cause ICOSL ectodomain shedding, engagement of ICOSL with ICOS is the most physiologically relevant, and, indeed, ICOS−/− mice have elevated ICOSL levels (27, 28). This interaction is known to cause rapid shedding of ICOSL from the cell surface (29). WT and ADAM17−/− B cells incubated for 1 h with plate-bound ICOS caused ∼27% of ICOSL to be shed from the cell surface (Fig. 2F). In contrast, ADAM10−/− B cells shed ∼5% of ICOSL (Fig. 2F). Interestingly, ADAM10/17−/− B cells were completely protected from ICOSL shedding in response to ICOS engagement (Fig. 2F). Taken together, these results suggest that ADAM10 is the primary sheddase in response to ICOS engagement. However, ADAM17 appears to have a more limited role in the regulation of ICOSL in response to ICOS engagement in the absence of ADAM10.
ICOSL shedding is required for proper T cell ICOS regulation
Given the importance of the ICOS/ICOSL interaction for humoral immunity, it was counterintuitive that elevated ICOSL on ADAM10−/− B cells would result in the reduced TFH differentiation and Ab response seen in these mice (5). Thus, we considered that the inability to shed ICOSL may be altering ICOS expression. Indeed, splenic T cells from ADAM10B−/− mice have undetectable surface ICOS expression (Fig. 3A). Within the different T cell subsets, ICOS is most highly expressed on TFH cells. When we examined the few TFH cells present in dLNs from ADAM10B−/− or ADAM10/17B−/− mice (Supplemental Fig. 2A, 2B), in contrast to total T cell analysis (Fig. 3A), ICOS surface levels were detectable, but dramatically reduced, compared with WT, whereas loss of ADAM17 alone had no effect on TFH levels. Regulatory TFH cells have been shown to alter TFH responses (30), possibly explaining the decreased TFH cells seen in ADAM10/17B−/− mice. However, we saw no differences in regulatory TFH cell numbers following immunization with NP31-KLH in alum (Supplemental Fig. 2C). When WT CD45.1+ T cells were adoptively transferred into CD45.2+ ADAM10B−/− recipient mice and ICOS surface expression levels were monitored, donor CD45.1+ T cells lost ICOS surface expression within 24 h, and this loss of surface expression was maintained for up to 7 d (Fig. 3B, 3C). Because ICOS expression on developing T cells in the thymus was not altered (Supplemental Fig. 2D, 2E), we concluded that changes in expression occur in the periphery and require continual interaction with elevated B cell ICOSL. Further evidence for regulation by ligand:receptor interaction is seen when T cell ICOS from ADAM10B−/− mice is restored to WT levels after ex vivo stimulation with anti-CD3 for 3 d (Fig. 3D). Next, we wanted to examine whether elevated ICOSL on B cells from ADAM10B−/− mice would directly downregulate WT T cell ICOS levels in vitro. Coculturing WT T cells with ADAM10−/− B cells led to a significant decrease in T cell ICOS levels within 24 h (Fig. 3E). This decrease was blocked by addition of MIL-5733 (17, 31), a nondepleting ICOSL-blocking Ab (Fig. 3E).
ICOS is internalized in response to ICOSL engagement
To determine the mechanism of this ICOS downregulation, Icos transcription levels were examined; no difference was seen between WT and ADAM10B−/− T cells (Fig. 4A), suggesting that regulation occurs at the protein level. When surface and internal ICOS levels were examined by flow cytometry, T cells from ADAM10B−/− mice exhibited an elevated internal ICOS ratio relative to WT T cells (Fig. 4B, 4C), with essentially all of the ICOS internalized. Using Amnis ImageStream analysis, we confirmed internalization of ICOS (Fig. 4D) by WT and ADAM10B−/− T cells. ICOS staining for WT T cells indicates some punctate internal spots, but most ICOS is surface expressed, whereas ADAM10B−/− T cells have bright internal spots for ICOS staining. Examination of the internalization index with ImageStream software confirms an increased internalization in ADAM10B−/− T cells compared with WT T cells (Fig. 4E). Together, these results indicate a novel regulation of ICOS through internalization following ligand interaction. This interaction occurs as the cells circulate through the secondary lymphoid system. In WT T cells, this interaction causes ICOSL cleavage with some ICOS internalization, whereas in ADAM10B−/− mice, ICOS internalization is enhanced in T cells as a result of the inability of ADAM10 to cleave ICOSL.
Increased ICOSL on ADAM10−/− B cells is responsible for the attenuated Ab production
We wanted to directly demonstrate that altered ICOS/ICOSL interactions in ADAM10B−/− mice were responsible for the decreased humoral response by performing rescue studies. MIL-5733 is a blocking anti-ICOSL Ab (32) that can completely block ICOSL in vivo when administered at a dose of 150 μg every other day (17). We administered a suboptimal dose (10 μg per mouse) of MIL-5733 every other day, which we hypothesized would reduce the level of exposure of ICOSL to T cells in ADAM10B−/− mice (Fig. 5A). Within 48 h of this dosing regimen, unblocked ICOSL (Fig. 5B) and ICOS (Fig. 5C) were returned to levels seen in WT mice. To determine whether the humoral response in ADAM10B−/− mice was restored, MIL-5733 was administered 2 d prior to footpad and i.p. injections of NP31-KLH and continued every other day. At 14 d postimmunization, GC B cells were significantly higher in ADAM10B−/− mice receiving the blocking Ab compared with isotype control (Fig. 5D, 5E). TFH numbers were also significantly increased in ADAM10B−/− mice receiving the blocking Ab compared with isotype control (Fig. 5F, 5G), and within the recovered TFH population, ICOS levels were similar to those seen in WT mice (Fig. 5H, 5I). Immunofluorescent microscopy analysis of MIL-5733–treated ADAM10B−/− dLNs indicates at least partial recovery of GCs by number and size (Supplemental Fig. 3). Furthermore, Ab production was rescued, because total (NP25) and high-affinity (NP4) IgG1 levels were not significantly different between MIL-5733–treated ADAM10B−/− and WT mice, whereas ADAM10B−/− mice receiving the isotype had significantly less total and high-affinity IgG1 (Fig. 5J).
Shedding of ICOSL is necessary for proper T cell responses
We have previously reported decreased TH2 responses in ADAM10B−/− mice (6). With the above results, we hypothesized that this defect in TH2 responses may be a result of elevated ICOSL in these mice. To investigate this, we decided to use ADAM10/17B−/− mice because they lack both physiological sheddases and, therefore, have the most elevated ICOSL levels. Using the HDM model of Ballesteros-Tato et al. (33) (Supplemental Fig. 4A), we demonstrated reduced TFH cells in medLNs on day 6 (Supplemental Fig. 4B), as well as reduced TH2 responses in lung T cells on day 20 (Fig. 6A–D). Thus, both TFH and TH2 are reduced in ADAM10/17B−/− mice, as was seen in ADAM10B−/− mice. Furthermore, when day-20 medLN cells were cultured with increasing doses of HDM Ag, T cells from ADAM10/17B−/− mice had decreased proliferation (Fig. 6E), indicating less T cell activation. Together, these results suggest that the inability to shed ICOSL is the primary mechanism behind the dysfunctional TH2 responses in an HDM model.
Despite the reduced responses seen in the HDM model, dLN cellular expansion was not different from WT mice (data not shown). To explore potential differential T cell activation, we used the active experimental autoimmune encephalomyelitis (EAE) model, which is driven by TH1 and TH17 activation (20, 34). ADAM10/17B−/− mice succumbed to more severe disease and higher incidence than WT mice (Fig. 7A, 7B). Although CNS CD4+ T cell levels and dLN cell numbers were not significantly different (Supplemental Fig. 4C, 4D), TFH and GC B cell numbers were strongly reduced in ADAM10/17B−/− mice (Supplemental Fig. 4E–H), suggesting the possibility for enhanced TH1 and/or TH17 polarization of CD4+ T cells. In support of this, clear increases in the levels of IFN-γ+ CD4+ T cells were present in the CNS (Fig. 7C, 7D) and dLNs (Fig. 7E, 7F). Although we did not see a difference in the number of IL-17A+ T cells (Fig. 7G–J), restimulation of dLN cells with MOG35–55 led to enhanced IL-17A+ (Fig. 7K) and IFN-γ+ (Fig. 7L) T cell levels in ADAM10/17B−/− mice compared with WT. Additionally, restimulation of dLNs with MOG35–55 caused a higher dose-dependent increase in proliferation by ADAM10/17B−/− T cells (Fig. 7M), in contrast to HDM restimulation. Overall, the results demonstrate an affinity toward differentiation into TH1 and TH17 effector cells, whereas TFH and TH2 differentiation is attenuated. The overall effect of enhanced ICOSL expression is outlined in the model shown in Fig. 8.
Defective humoral immune responses in ADAM10B−/− mice (5, 6) and an earlier report of ADAM17 being important for ICOSL shedding (16) prompted us to examine whether ADAM10 was involved in the shedding of ICOSL. Our studies demonstrate a large increase in ICOSL surface expression on B cells from ADAM10B−/− mice (Fig. 1A). Although it had previously been reported that ADAM17 was involved in ICOSL release in response to PMA and BCR cross-linking (16), in agreement with Marczynska et al. (16), we did not see perturbation of ICOSL levels in ADAM17B−/− mice, supporting the presence of another protease. With interaction of ICOSL with ICOS most likely being the physiologic signal for cleaving ICOSL in vivo, we examined whether ADAM10 was the protease responsible for this regulation. ADAM10−/− B cells had reduced ICOSL shedding in response to ICOS ligation (Fig. 2F). Thus, in agreement with Marczynska et al. (16), we conclude that ADAM17 cleaves ICOSL post-PMA or BCR interaction, but ADAM10 is the protease responsible for ligand-induced, as well as constitutive, ICOSL shedding. This cleavage was further reduced in B cells from ADAM10/17B−/− mice, suggesting that ADAM17 may have the capacity to act as a backup sheddase of ICOSL in the absence of ADAM10. In support of this, ADAM10/17B−/− mice exhibited further increased ICOSL levels in vivo compared with ADAM10B−/− mice (Fig. 1A). In addition, ICOSL cleavage was shown using recombinant proteins by examination of the fragment remaining after cleavage. Based on band-shift analysis, recombinant protein assays suggest that the primary fragment shed from the N terminus of ICOSL was ∼3.3 kDa (Fig. 2A, 2B). Cleavage of this small fragment could result in disruption of tertiary structure, preventing further ligation with ICOS. At least by size analysis, ADAM10 and ADAM17 give similar-sized fragments, although exact cleavage site analysis will be required to confirm this. Intriguingly, the cleavage of ICOSL has similarities to the S2 Notch1/2 cleavage. The interaction between Notch and its ligand exposes the cleavage site for ADAM10. ADAM17 is reported to cleave Notch, but ADAM17 cleavage is primarily nonphysiological. With the rapid kinetics of ICOSL shedding following interaction with ICOS, it would suggest that the increase in ICOSL shedding is due to activation of ADAM10 at a posttranslational level. It will require further studies to determine whether interaction of ICOSL with ICOS causes signaling events to activate ADAM10 or whether ADAM10 is already associated near ICOSL, and the interaction of ICOS and ICOSL opens a cleavage site that is accessible to ADAM10, similar to Notch S2 cleavage (35). Perhaps a similar situation exists with ICOSL cleavage (i.e., interaction with ICOS further exposes the ICOSL cleavage site). We note though, that HLA activation of ADAM10 results in efficient ICOSL cleavage, indicating that highly activated ADAM10 does not need the ICOS interaction. We would argue that ADAM10 is responsible for the low-level constitutive ICOSL cleavage (Fig. 1E, 1F) and the ICOS-induced cleavage (Fig. 2F).
The ICOS/ICOSL interaction has been shown to be particularly important in humoral immune function. Loss of this costimulatory axis largely blocks TFH and TH2 function (36). Understanding the regulation of ICOS and ICOSL is essential for developing new approaches to modulate the Ab responses. Although a number of ICOS-regulatory mechanisms have been identified at the mRNA level (13, 15, 37), our studies demonstrate the importance of ICOSL shedding and ICOS internalization in control of this interaction. Indeed, T cell Icos transcriptional levels were not different between ADAM10B−/− and WT T cells (Fig. 3A), confirming that regulation of ICOS is at the protein level. As its name implies, ICOS is an induced costimulatory molecule, but it is still present at low levels on naive T cells (11). When noncleavable ICOSL is present, then ICOS levels become undetectable. Several studies have shown that the flip side of this also occurs (i.e., that alteration in ICOS levels can lead to dysregulation of ICOSL). In ICOS−/− mice, there is a marked increase in ICOSL levels on APCs (16, 29). In this article, we show that blocking the ICOSL catabolism gives rise to a phenotype similar to ICOS−/−. Although it may be possible that soluble ICOSL is causing the apparent downregulation of surface ICOS, we do not see a difference in thymic T cell ICOS levels, which would also be exposed to soluble ICOSL (Supplemental Fig. 2E). These results suggest that soluble ICOSL is not the mechanism behind decreased surface ICOS levels in ADAM10B−/− mice. Interaction between T cells and B cells in the secondary lymphoid tissue appears to be sufficient for this mechanism of ICOS/ICOSL regulation to occur.
One of the limiting factors of the GC response is the length of the initial ICOS/ICOSL interaction (38), which induces APCs to shed surface ICOSL (30, 38). In this article, we demonstrate that if ICOSL shedding is blocked, ICOS is internalized and degraded to a significant extent. It is important to note that ICOS internalization is also seen in WT T cells, although to a lesser extent, suggesting that this is a normal control pathway for regulating surface ICOS/ICOSL expression. Loss of the capacity for this shedding, as seen in ADAM10B−/− and ADAM10/17B−/− mice, leads to a large accumulation of ICOSL on the B cell surface (Fig. 1A). To our knowledge, ICOSL is the only B7 family costimulatory molecule that is proteolytically cleaved upon receptor interaction. Although internalization of CD28 may also play an important role in its regulation (39, 40), CD28 is also negatively regulated by CTLA4, which is able to bind the same ligands as CD28; this contrasts with ICOS, with which only ICOSL reportedly interacts. The overall effect of this enhanced ICOSL expression is outlined in the model shown in Fig. 8.
The presence of such high ICOSL, caused by the loss of its primary sheddase, resulted in the loss of TFH differentiation in several immunization models. In addition, we saw defective TH2 responses in an HDM model of allergic asthma (6) (Fig. 6). Thus, although we see the same overall reduction in TH2 as Ballesteros-Tato et al. (33), it is equally plausible that low ICOS on the T-resident memory cells would mean that TFH and T-resident memory cell models (41) could be operable. The increased TH1 and TH17 responses seen in an active EAE model (Fig. 7) were reminiscent of the findings in ICOS−/− mice (9, 11). Indeed, expanded cellularity anticipated as a result of immunization was not significantly influenced with any of the immunization or disease models. Rather, increased TH1, and probably TH17, activation now occurs. Although blocking ICOSL resulted in recovery of ICOS, TFH differentiation, and GC formation, these studies were limited in their ability to determine whether loss of GC formation was due to loss of ICOS costimulatory capacity or its ability to properly orient T cells to the GC (42).
Overall, this study confirms the importance of ICOSL shedding in ICOS/ICOSL function and expression, and it identifies ADAM10 as the most important sheddase for controlling ICOSL levels. These results demonstrate how a normal regulatory pathway becomes aberrant when one of the required catabolism signals is not present. In addition, these findings demonstrate that blocking B cell ADAM10 activity represents a novel mechanism to modulate ICOSL and ICOS expression and alter the humoral immune response.
We thank Matthew Zellner for assisting with mouse colony management. We thank Julie Farnsworth for assistance with flow cytometry analysis and ImageStream analysis. The modified PX330 plasmid containing GFP was a gift from Gordon Ginder (Massey Cancer Center, Virginia Commonwealth University) and was developed by Xiaofei Yu (Massey Cancer Center, Virginia Commonwealth University). Alex Azzo (Center for Clinical and Translational Research, Virginia Commonwealth University) aided in the development of ADAM10−/− RPMI 8226 and RPMI 8866 cell lines. We thank Jared Farrar (Center for Clinical and Translational Research, Virginia Commonwealth University) for use of equipment.
This work was supported by National Institutes of Health/National Institute of Allergy and Infectious Diseases Grant R01AI18697A1-33-38 (to D.H.C.). Flow cytometry was supported in part by Massey Cancer Center Core National Institutes of Health Grant P30 CA16059.
The online version of this article contains supplemental material.
Abbreviations used in this article:
a disintegrin and metalloproteinase
loss of ADAM10 on B cells
draining lymph node
experimental autoimmune encephalomyelitis
house dust mite
intranasal(ly): medLN, mediastinal lymph node
mean fluorescence intensity
myelin oligodendrocyte glycoprotein
NP31 conjugated to keyhole limpet hemocyanin
T follicular helper
The authors have no financial conflicts of interest.