B Cell ADAM10 Controls Murine Lupus Progression through Regulation of the ICOS:ICOS Ligand Axis

The role of the ICOS of T cells (ICOS) has been well characterized and has been shown to be essential for follicular helper T cell (T_FH), T effector cell (T_eff), and Th2 functions (1–4). Loss of ICOS has been shown to result in blunted thymus-dependent (T-dependent) Ab responses as well as absent germinal center (GC) formation. The only known ligand for ICOS (ICOSL) is expressed on APCs, such as dendritic cells, B cells, and other hematopoietic and nonhematopoietic cells (5). ICOSL-deficient mice have a similar phenotype to ICOS-deficient mice, suggesting the importance of their cognate interaction (6).

Studies regarding the protein and transcriptional regulation of ICOS and ICOSL in autoimmune disease have been limited. In steady-state mice, Icos transcriptional levels have been shown to rapidly increase upon TCR stimulation (1, 4). Additionally, several studies have shown the importance of ICOS posttranscriptional regulation through Roquin-1 and mir146a (7–9). A loss-of-function mutation in Roquin-1 (sanroque mice) or loss of mir146a in mice results in elevated ICOS levels, ultimately resulting in exaggerated GC responses and Ab production (8, 9). Additionally, sanroque mice develop an autoimmune phenotype resembling lupus with autoantibody production (7, 8). These studies clearly show the importance of proper ICOS regulation in maintaining homeostasis between productive GC responses and autoimmunity.

Disease progression in lupus-prone B6.MRL-Fas^lpr/J (Fas^lpr) mice and other lupus-prone models is also associated with altered levels of ICOS and ICOSL (10–13). Several groups have shown that ICOS is required for class-switched autoantibody production in MRL.Fas^lpr and Sle1 mice (10, 13). The origin of the T cells responsible for this B cell help and autoantibody production are thought to be extrafollicular in nature but resemble T_FH in gene expression and cytokine production, mainly IL-21 and CD40L (14–18). Several studies have contradictory results regarding the role of ICOS in T_eff function in the Fas^lpr model. More recently it has been suggested that ICOSL on CD11c⁺ cells promotes T cell survival and effector function in the kidneys (12). However, these mice were not protected from the development of autoimmune Abs, whereas B cell conditional ICOSL knockout mice developed reduced autoantibody, suggesting differential roles for B cell and dendritic cell ICOSL (12) and further suggesting a multifaceted role for ICOSL.

We have recently shown that mice that conditionally lack a disintegrin and metalloproteinase 10 (ADAM10) on B cells (A10B) have elevated ICOSL on this cell due to the inability to shed ICOSL from the cell surface (19). These mice having decreased GC responses and Ab production. The mechanism for this defect in humoral immunity was the finding that the increase in ICOSL on the B cell surface led to a posttranslational downregulation of surface ICOS levels on T cells. This regulation was examined in naive, 4-hydroxy-3-nitrophenylacetyl hapten conjugated to keyhole limpet hemocyanin (NP₃₁-KLH)–immunized, experimental autoimmune encephalomyelitis, and house dust mite–challenged mice and was able to effectively downregulate ICOS levels to block T_FH responses and affinity-matured Ab production (19). These studies suggested that in addition to proper translational regulation of ICOS and ICOSL, proper posttranslational regulation of these proteins is just as important for regulating humoral immunity.

In this study, first we show that B cell ADAM10 is necessary for the enhanced ICOS and T_FH expression that is associated with the B6^{mir146a−/−} mice (9) and that loss of B cell ADAM10 ablated the increased T_FH accumulation seen in these mice. Additionally, we show that loss of B cell ADAM10 in the lupus-prone Fas^lpr mouse model results in decreased T_FH accumulation and more importantly a decrease in anti-dsDNA Abs. Our results indicate that B cell ADAM10 represents a novel mechanism of ICOSL and ICOS regulation in the context of humoral autoimmunity in models where enhanced immune responses are seen, and this novel mechanism even extends to the lupus model (one of the most severe of autoimmune diseases).

All mice were maintained at the Virginia Commonwealth University Animal Facility in accordance with guidelines by the U.S. National Institutes of Health and the American Association for the Accreditation of Laboratory Animals Care. C57BL/6J ADAM10-floxed mice crossed to the CD19-cre mouse were generated previously (20). Fas^lpr mice were purchased from The Jackson Laboratory (000482). These mice were crossed to Adam10^fl/fl-floxed CD19cre^+/− mice. For lpr studies, Fas^lpr/lpr Adam10^fl/fl Cre^−/− are referred to as B6^lpr, Fas^lpr/lpr Adam10^fl/fl Cre^+/− are referred to as A10B^lpr, Fas^−/− Adam10^fl/fl Cre^−/− are referred to as B6, and Fas^−/− Adam10^fl/fl Cre^+/− are referred to as A10B. B6^{mir146a−/−} mice were purchased from The Jackson Laboratory (016239) and crossed to Adam10^fl/fl-floxed CD19cre^+/− mice. For Mir146a studies, mir146a^−/− Adam10^fl/fl Cre^−/− mice are referred to as B6^{mir146a−/−}, mir146a^−/− Adam10^fl/fl Cre^+/− mice are referred to as A10B^{mir146a−/−}, mir146a^+/+ Adam10^fl/fl Cre^−/− mice are referred to as B6, and mir146a^−/− Adam10^fl/fl Cre^+/− mice are referred to as A10B.

Kidneys and lungs were excised and trimmed followed by fixation in 10% formalin for 48 h. Organs were paraffin embedded, sectioned, deparaffinized, and rehydrated as described (21). Sections were stained with H&E and imaged using an Olympus BX41 microscope.

Serum was collected from mice by cardiac puncture and allowed to clot for at least 30 min at room temperature. Serum was then separated from clotted blood components by centrifugation and stored at −20°C until use. For total Ab assays, 384-well FLUOTRAC 600 (Greiner Bio-One) were coated with 20 μl of 5 μg/ml goat anti-mouse IgG₁ (1071-01; Southern Biotech), IgG_2a (1081-01; Southern Biotech), IgG₃ (1101-01; Southern Biotech), or IgM (1021-01; Southern Biotech) in borate-buffered saline overnight at 4°C. Plates were washed (PBS with 0.02% Tween-20) and blocked with 20 μl of 5% FBS in PBS and for 2 h at room temperature. Twenty microliters of serum dilutions and Ab standards were added to wells and incubated overnight at 4°C. Plates were washed and Abs were detected using 20 μl of biotinylated goat anti-mouse IgG (1036-08; Southern Biotech) for all IgG isotypes and 20 μl of biotinylated goat anti-mouse IgM (1021-08; Southern Biotech) for IgM assays at 2 μg/ml overnight at 4°C. Plates were washed and incubated with 20 μl of 2 μg/ml PE-streptavidin (405204; BioLegend) in block for 2 h at room temperature. Plates were then washed, and 30 μl of PBS was added to all wells and analyzed with a SpectraMax M5 with an excitation wavelength of 496 and emission wavelength of 574. For anti-dsDNA assays, the 384-well FLUOTRAC 600 plates were coated with 20 μg/ml protamine sulfate in BBS overnight at 4°C. Plates were washed and incubated with 5 μg/ml highly polymerized calf thymus DNA (Worthington) in PBS overnight at 4°C. Plates were washed and serum dilutions were added and incubated overnight at 4°C. After washing, detection was performed as stated for total Ab assays.

Kidneys were frozen on dry ice in OCT compound (Tissue-Tek). Ten-micrometer sections were cut from frozen blocks using a cryostat (Frigocut 2800E; Jung), fixed in absolute ice-cold acetone for 10 min and air dried. Sections were rehydrated in PBS and blocked in 4% horse serum and 3% BSA in PBS for 60 min at room temperature. Sections were washed and labeled with 5 μg/ml FITC goat anti-mouse IgM (1021-02; Southern Biotech) or 5 μg/ml Alexa Fluor 647 goat anti-mouse IgG-Fc (1033-31; Southern Biotech) for 60 min at room temperature. Images were captured on a LSM700 Axio Imager 2.

Lymph nodes were excised and teased apart followed by mechanical digestion through a 100-μm cell strainer. Spleens were excised and mechanically forced through a 100-μm cell strainer followed by ACK lysis of RBCs. Kidneys were excised and cut into <1 mm pieces using a razor and then digested at 37°C for 45 min with 0.05 mg/ml Liberase TM (Roche) and 0.05 mg/ml DNAse1 (Worthington). Digested tissue was then mechanically passed through a 100-μm filter followed by ACK (Quality Biologic) lysis of RBCs. Single-cell suspensions of all organs were washed with PBS, and live/dead staining was done using Zombie Aqua (423102; BioLegend) according to the manufacturer’s protocol. Cells were washed with FACS buffer (5% FBS in PBS with 2 mM EDTA). Fc receptors were blocked with 5 μg 2.4g2 (22) for 10 min at 4°C. Abs were added at indicated concentrations (Table I) for 45 min at 4°C. Cells were washed two times with FACS buffer and fixed in Fixation Buffer (420801; BioLegend), or secondaries were added and incubated for 30 min at 4°C followed by fixation. For intracellular staining, following fixation, cells were permeabilized using Intracellular Stain Permeabilization Buffer (421002; BioLegend) according to the manufacturer’s protocol. Cells were stained for intracellular markers for 60 min at room temperature and washed and fixed. Flow cytometry data were collected on an LSR Fortessa II and analyzed in FCS Express 5. Gating strategies and cell surface phenotypes can be found in Supplemental Fig. 1.

Ten micrograms of NP₃₁-KLH (BioSearch Technologies) was added to PBS containing 0.25 mg alum/ml (Imject, Sigma-Aldrich) in a volume of 25 μl and injected into each hind footpad. Popliteal lymph nodes (popLN) were isolated from mice 14 d postimmunization.

All statistical analyses used GraphPad Prism 7.3. Individual figure legends indicate statistical tests used for different experiments. Overall, Bartlett test for equal variance was used, and if variances were not equal, equivalent nonparametric tests as indicated in figure legends were used.

We first sought to examine the ability of B cell ADAM10 to regulate elevated humoral immune responses seen in the B6^{mir146a−/−} mice. B6^{mir146a−/−} mice have elevated ICOS levels as well as elevated T_FH responses following immunization (9). ADAM10 B cell conditional knockout mice (A10B) (20) were crossed to B6^{mir146a−/−} mice (9) to make A10B^{mir146a−/−} mice to examine whether B cell ADAM10 could effectively regulate humoral immune responses in this model. We first examined ICOSL levels and saw that A10B^{mir146a−/−} mice had elevated levels of ICOSL on B cells similar to A10B mice. In both cases, ICOSL was significantly higher than wild-type C57BL/6J (B6) or B6^{mir146a−/−} mice (Fig. 1A). B6^{mir146a−/−} CD4⁺ T cells from the nondraining popLN repeated the previously reported increase in ICOS compared with B6 mice (9); however, loss of B cell ADAM10 resulted in decreased ICOS levels in both situations (Fig. 1B). Additionally, when adoptively transferred CD4⁺ T cells from B6^{mir146a−/−} mice into either B6 or A10B mice, we saw that ICOS levels on donor T cells from B6^{mir146a−/−} mice dropped significantly in the A10B recipients but not the B6 recipients (Fig. 1C).

We next immunized these mice with a T-dependent immunization model using NP₃₁-KLH and examined GC responses. As previously reported, B6^{mir146a−/−} mice had a significant increase in T_FH compared with B6 mice (Fig. 1D–F). However, A10B^{mir146a−/−} mice had similar T_FH levels compared with A10B mice and significantly lower relative and absolute T_FH numbers than both B6 and B6^{mir146a−/−} mice (Fig. 1D–F). Supporting our previous findings (19), we saw significantly decreased absolute and relative GC B cells in A10B mice compared with B6 mice in the draining popLN (Fig. 1G–I). Additionally, we saw significantly reduced absolute and relative GC B cells numbers in A10B^{mir146a−/−} mice compared with B6^{mir146a−/−} mice (Fig. 1G–I). As also reported (9), B6^{mir146a−/−} mice had significantly higher T_FH ICOS levels than B6 mice (Fig. 1J, 1K). However, T_FH ICOS levels were significantly lower in A10B^{mir146a−/−} mice compared with both B6 and B6^{mir146a−/−} mice and were not significantly different from A10B mice (Fig. 1J, 1K). Thus, loss of mir146 does not overcome the suppression of ICOS or the decreased GC response that was seen in A10 animals (19, 23).

We next turned to a more severe model where overstimulation of the immune response is seen, the Fas^lpr model (24). By appropriate backcrossing, B6^lpr mice that lacked ADAM10 on their B cells (A10B^lpr) were generated. The B6 strain has a delayed onset compared with the MRL, with 6 mo of age being the common onset of disease (24). Thus, our initial studies were performed at this age. Splenomegaly was evident in both B6^lpr and A10B^lpr strains, and no significant difference in splenic leukocyte (Fig. 2A) or infiltrating kidney leukocytes (Fig. 2B) and kidney T cells (Fig. 2C) were seen. A10B^lpr mice had similar T cell subset distributions in the spleen (Supplemental Fig. 2A–D) and kidneys (Supplemental Fig. 2E–H) compared with B6^lpr mice. We also did not observe any pathologic differences in the lungs or kidneys between A10B^lpr and B6^lpr mice as observed by H&E staining (Supplemental Fig. 2I, 2J). For unknown reasons, CD8⁺ T cell numbers in B6 and A10B were lower than B6^lpr and A10B^lpr (Supplemental Fig. 2C). There was a significant reduction (∼10-fold) in total cervical lymph node (cervLN) CD45⁺ cell number in the A10B^lpr compared with B6^lpr mice (Fig. 2D).

When we examined cervLN CD4⁺ T cells for T cell activation status, we saw that there was a significantly reduced percentage of activated CD4⁺ T cells (CD44⁺ CD62L⁻) in A10B^lpr mice versus B6^lpr mice (Fig. 2E–G). Additionally, both B6 and A10B mice had significantly fewer relative and absolute numbers of activated CD4⁺ T cells compared with their lpr counterpart (Fig. 2E–G). To further examine T cell activation states, we used CD127 as a marker for differentiation between T effector memory cells (T_em) and T_eff (12) gated on CD4⁺ CD44⁺ CD62L⁻ T cells. We saw a significant decrease in absolute but not relative levels of CD4⁺ T_eff in the cervLN of A10B^lpr mice versus B6^lpr mice (Fig. 2E, 2H, 2I). Additionally, we saw significant decreases in both relative and absolute numbers of CD4⁺ T_em in the cervLN of A10B^lpr mice versus B6^lpr mice (Fig. 2E, 2J, 2K). We did not see a difference in relative or absolute CD4⁺ T_em or CD4⁺ T_eff between A10B and B6 mice, but we did see a difference in absolute and relative CD4⁺ T_em or CD4⁺ T_eff between B6 and B6^lpr mice as well as A10B and A10B^lpr mice (Fig. 2E, 2H–K).

With our previous report showing that A10B mice have elevated ICOSL levels and decreased ICOS levels (19), we next investigated whether these altered ICOS and ICOSL levels were also present in the Fas^lpr model. B cells from A10B^lpr mice had a significant increase in ICOSL levels compared with B6^lpr mice; however, these ICOSL levels were significantly decreased when compared with A10B mice (Fig. 3A). We saw significantly lower ICOS levels on CD4⁺ T cells in A10B^lpr mice compared with B6^lpr mice (Fig. 3B). Interestingly, B6^lpr mice had significantly higher levels of ICOS on CD4⁺ T cells than age-matched B6 mice, and A10B^lpr mice had significantly higher levels of ICOS than A10B mice (Fig. 3B).

Because of the role of ICOS and ICOSL in GC reactions, we next examined GC responses in the cervLN in these mice. We did not see a difference in relative number of GL7⁺ cells in the cervLN (Fig. 3C, 3D) but did observe decreased absolute numbers of GL7⁺ B cells (Fig. 3E) in A10B^lpr mice, potentially due to decreased cervLN cellularity in A10B^lpr mice. When we examined the spleen, there was a significant decrease in percentage and absolute number of GL7⁺ GC B cells in the spleen (Supplemental Fig. 3A–C) in A10B^lpr mice.

T_FH were decreased in both percentage as well as absolute number in the cervLN of A10B^lpr mice compared with B6^lpr mice at 6 mo of age (Fig. 3F–H). T_FH were significantly decreased in absolute number but not relative number in the spleens of A10B^lpr mice compared B6^lpr mice (Supplemental Fig. 3D–F). In both the spleen and cervLN, T_FH were decreased in B6^lpr compared with B6 as well as A10B^lpr compared with A10B mice (Fig. 3F–H, Supplemental Fig. 3D–F). CervLN T_FH ICOS levels were significantly decreased in A10B^lpr mice compared with B6^lpr mice (Fig. 3I, 3J). This trend was similar in age-matched A10B and B6 controls, albeit with higher ICOS levels.

We saw no difference in relative or absolute numbers of CD4⁺ IFN-γ⁺ T cells between A10B^lpr mice B6^lpr mice in CD4⁺ T cells in both the spleen (Supplemental Fig. 3G–J) and cervLN (Supplemental Fig. 3K–M). However, the absolute number of CD4⁺ IFN-γ⁺ T cells in the cervLN trended down but was NS in the A10B^lpr mice because of the decrease in cellularity seen in these mice (Supplemental Fig. 3L).

Because of the decreased GC and T_FH numbers in the A10B^lpr mice, we examined whether there were any differences in total serum Abs as well as autoantibody production in these mice. Six-month-old A10B^lpr mice had a significant decrease in total IgG₁ and IgG_2a levels but had no significant difference in total IgM or IgG₃ (Fig. 4A). Anti-dsDNA–specific Abs were used to evaluate auto-Ab. We saw a modest decrease in anti-dsDNA IgM at 6 mo of age but a larger and significant decrease at 8 mo in A10B^lpr mice compared with B6^lpr mice (Fig. 4B). Anti-dsDNA IgG was clearly suppressed at both 6 and 8 mo of age in A10B^lpr mice compared with B6^lpr mice (Fig. 4C). Although 6-mo-old A10B^lpr mice did not have a noticeable difference in kidney IgM deposition, there was noticeable decrease in IgG Ab deposition in the kidneys of A10B^lpr mice. (Fig. 4D, 4E). Additionally, CD138⁺ cells in A10B^lpr mice were significantly decreased in both percentage and absolute number in the cervLN (Fig. 4F–H) as well as in the spleen (Fig. 4I–K).

Overall, the previously mentioned data indicates a strong modulation to the lpr phenotype following loss of B cell ADAM10. As shown in Fig. 3A and 3B, cells from 6-mo-old A10B^lpr mice had significantly higher ICOSL levels compared with B cells from B6^lpr mice; however, the ICOSL on B cells from A10B^lpr mice was decreased compared with age-matched A10B mice. Examination of ICOSL levels at different ages in A10B^lpr mice revealed a steady decrease in ICOSL levels with increasing age (Fig. 5A). This decrease in ICOSL levels was also seen in B6^lpr mice, albeit to a much lesser extent and at a younger age (Fig. 5A). There was a drastic drop in ICOSL levels seen between 4 and 5 mo in A10B^lpr mice, which coincided with T cell ICOS levels greatly increasing during this same time period (Fig. 5A, 5B). This is further seen when examining the correlation between decreasing ICOSL levels and increasing ICOS levels (Supplemental Fig. 4A). The increasing ICOS levels correlate with increasing T cell numbers in the cervLN (Fig. 5C). There was a striking difference in ICOS:ICOSL ratios between B6^lpr and A10B^lpr mice that continued with age but did narrow between 4 and 5 mo (Fig. 5D). Interestingly, this increase in ICOS:ICOSL ratios between 4 and 5 mo of age in A10B^lpr mice (Fig. 5D) overlaps with the age in which lymphoproliferation and increased cellularity began to occur in these mice (Fig. 5E) as well as T_FH accumulation (Fig. 5F).

We have previously shown that ADAM10 is responsible for proteolytic ICOSL regulation on B cells (19). Loss of A10B results in elevated ICOSL levels and a compensatory posttranslational decrease in T cell ICOS. This dysregulation of the ICOS:ICOSL axis resulted in almost complete loss of T_FH and affinity-matured Ab production in T-dependent immunization models (19, 23). This study built on those findings and examined the ability of ADAM10 to regulate the ICOS:ICOSL axis in two separate autoimmune models in which ICOS is important for pathogenesis.

In this study, we first assessed whether B cell ADAM10 was still able to regulate the ICOS:ICOSL axis in a model known to have elevated ICOS levels due to increased transcriptional levels. B6^{mir146a−/−} mice have been shown to have elevated ICOS levels on resting T cells as well as increased T_FH numbers and T_FH ICOS levels following immunization (9). Mir146a was shown to regulate ICOS transcriptional levels in cooperation with Argonaute2 and Roquin, and its loss resulted in elevated ICOS transcriptional levels (9, 25). The loss of A10B in B6^{mir146a−/−} mice resulted in decreased ICOS on T cells and decreased T_FH and GC responses, completely reversing the phenotype seen in B6^{mir146a−/−} mice. These results further support the importance of posttranslational regulation of ICOSL and ICOS by ADAM10-mediated catabolism of ICOSL.

A number of studies have shown an important role for both ICOS and ICOSL in murine lupus models (10–14). Several reports point to ICOS being essential for the development of autoantibody production while being dispensable for the accumulation of Th17 and Th1 effector cells (10, 11, 13). In A10B^lpr mice, we did not see a difference in Th1 cells in either the spleen of cervLN (Supplemental Fig. 3). Other groups have shown that ICOS stimulation is essential for T cell–mediated interstitial nephritis in MRL^lpr mice, which is not seen in the B6 version of this model (12, 24). However, we did see decreased IgG deposition in the glomeruli of A10B^lpr mice (Fig. 4E). Studies using conditional ICOSL knockout mice have shown that B cell ICOSL is not fully necessary for autoantibody production in MRL^lpr mice and is also dispensable for the accumulation of IL-21–producing effector helper T cells (T_efh) in the kidneys and spleens (12). In addition, loss of ICOSL on dendritic cells also did not affect autoantibody production but did cause a reduction in T_efh accumulation in the kidneys as well as decreased kidney pathology (12). Other groups have shown that ICOS^−/− mice have a loss of T_eff in the kidney when crossed to MRL^lpr mice (10, 11, 14). In our study, we did not see a difference in kidney T cell infiltration (Fig. 2C).

Our ADAM10 B cell conditional knockout model examines the role of ICOS and ICOSL in the Fas^lpr model in a different context than genetic knockouts of either protein. Although our previous results have indicated an important role for B cell ADAM10 in the regulation of ICOSL levels (19), aberrant regulation of ICOS and ICOSL in A10B^lpr mice was still seen. With increasing age, we observed an increase in T cell surface ICOS and a corresponding decrease in B cell surface ICOSL. In B6^lpr mice, this decrease in ICOSL and increase in ICOS occurs between 2 and 3 mo, whereas in A10B^lpr mice, it occurs between 4 and 5 mo (Fig. 5A, 5B) and can be seen more drastically when examining the ICOS:ICOSL ratio (Fig. 5D). Intriguingly, this increase in the ICOS:ICOSL ratio coincides with the first signs of lymphoproliferation in these mice (Fig. 5E). With these findings, we propose a model in which disease progression in lpr mice begins with the accumulation of T cells due to loss of FAS-mediated cell death. With this increase in T cell numbers, we see a decrease in ICOSL levels. This ultimately allows increased T cell accumulation of ICOS, which eventually leads to increased T cell activation, proliferation, and the ability to provide B cell help. This B cell help then allows for increased Ab and autoantibody production.

ICOS and ICOSL interaction have been shown to be necessary for class-switched Ab responses during T-dependent immunization models (2). Mice that lack either ICOS or ICOSL have drastically decreased class-switched Ab. In our model, A10B^lpr mice have decreased total IgG₁ and IgG_2a at 6 mo compared with B6^lpr mice. We did not see a difference in IgM in A10B^lpr mice at this timepoint, which is not surprising as IgM is not a class-switched Ab. However, we did see a difference in anti-dsDNA IgM in A10B^lpr mice compared with B6^lpr mice, suggesting that A10B^lpr mice have decreased loss of tolerance compared with B6^lpr mice. We saw a significant difference in anti-dsDNA IgG at both 6 and 8 mo of age, further supporting the role for ADAM10 in regulating the ICOS:ICOSL axis in the lpr model. We were not able to examine the mice past 8 mo of age due to increased mortality in B6^lpr mice.

We hypothesize that early T cell lymphoproliferation in B6^lpr mice results in more interaction with B cell ICOSL, resulting in ICOSL shedding and ultimately decreasing B cell ICOSL levels. These decreased ICOSL levels are then responsible for the increased T cell ICOS levels seen, as we have previously shown that interaction of ICOS with ICOSL results in decreased ICOS levels through internalization of surface ICOS (19), further supporting this is the finding that ICOSL^−/− mice have elevated T cell ICOS levels (26). We propose that T cells are able to be activated more extensively and differentiate to T_FH and T_efh once their ICOS levels increase. ICOS stimulation for these cells could likely be coming from other APCs, such as dendritic cells.

An interesting finding in this study was that that 6-mo-old A10B^lpr mice had decreased B cell ICOSL levels compared with age-matched A10B controls (Fig. 2A). There are several possibilities to explain this finding. Although A10B mice have elevated ICOSL protein levels compared with wild-type mice, the transcriptional level of Icosl is greatly reduced (19), suggesting that excessive interaction of ICOS and ICOSL signals B cells to further downregulate Icosl transcription, ultimately resulting in decreased ICOSL surface expression. It is also possible that with the rapid lymphoproliferation seen in this model, we are losing penetrance of the Cre transgene in A10B^lpr mice, which could explain why the mice begin to develop lymphoproliferation and autoantibody production at later timepoints. However, if this were the case, we would expect to see a bimodal pattern of ICOSL expression on the surface of B cells in A10B^lpr mice, but that is not the case.

Another possibility for the decreased ICOSL levels in A10B^lpr mice is that a different protease, most likely ADAM17, is able to cleave ICOSL in the context of autoimmunity. A10B mice have been shown to have elevated ADAM17 levels, which may allow for more shedding of ICOSL (27). We, and others, have shown that ADAM17 has the ability to shed ICOSL in response to several stimuli, most notably PMA stimulation (19, 28). Mice that lack both ADAM10 and ADAM17 on B cells have even higher levels of ICOSL compared with just the loss of ADAM10 (19). However, this seems to only be the case in conjunction with the loss of ADAM10, as mice deficient in ADAM17 on B cells do not have elevated ICOSL levels compared with wild-type mice (19). The ability of ADAM17 to control ICOSL in the absence of ADAM10 in the lpr model warrants further investigation, and we hypothesize that loss of ADAM17 in addition to A10B in lpr mice would further protect from autoantibody development.

Because of limitations of this study, we cannot exclude the possibility that other substrates of ADAM10 may account for some of the phenotype seen in A10B^lpr mice. ADAM10 is known to shed a wide range of substrates, including chemokine receptors, as well as participate in Notch signaling (20, 29). We have previously reported that in A10B mice, reversing the dysregulation of ICOSL and ICOS by blocking excess ICOSL interaction with ICOS was able to restore humoral responses following a T-dependent immunization (19). However, the correlation we see with decreased ICOSL and increased ICOS certainly implicates this change as being relevant to the increased autoimmmunity that is seen.

Lower ICOS levels were observed on T_FH present in A10B^lpr mice, suggesting they may have decreased ability to fully offer B cell help for Ab production, particularly when isotype switching is involved (13, 30, 31). Several reports have suggested the source of autoantibody production is extrafollicular (14, 18) in the lpr model, so we cannot conclusively determine that the decrease in T_FH is directly responsible for decreased autoantibody production, at least with respect to this occurring in GCs. However, all CD4⁺ T cells in A10B^lpr mice had decreased ICOS levels compared with B6^lpr mice. These extrafollicular helper T cells may also have decreased ability to aid in extrafollicular autoantibody production, particularly in the context of IL-21 and CD40L, which are reliant on ICOS stimulation (15, 32, 33). This is further supported with the significant decrease seen in anti-dsDNA IgM and IgG levels at 8 mo.

Overall, these results show that indirectly regulating ICOS by blocking ICOSL catabolism is a potent way to control humoral immunity, even in a severe autoimmune model. Additionally, the A10B^lpr mice offer a model to study the role of ICOSL and ICOS in the development of lymphoproliferation and autoantibody production in lpr mice, without genetic deletion of either ICOS or ICOSL. Additionally, our results further exemplify the importance of tight regulation of this system to properly function and particularly the importance of ADAM10 in this regulation. However, because this model examines the genetic loss of ADAM10 on B cells before disease onset, we cannot determine whether blocking ADAM10 following onset of disease would be beneficial to reverse disease pathologies. We would exercise caution in the use of ADAM10 inhibitors for the treatment of established autoimmunity, as we cannot preclude the possibility that inhibiting ICOSL shedding through ADAM10 inhibition may actually further exacerbate disease severity when ICOS levels are already elevated.

The results further our understanding of ICOS:ICOSL regulation in the context of autoimmunity as well as other models of increased humoral immune responses. Our results suggest that proteolytic regulation of ICOSL on B cells by ADAM10 is able to extensively impact the ICOS- and ICOSL-dependent responses. These findings support the role for ICOSL catabolism in controlling ICOS levels and T cell activation, particularly in the context of autoimmunity reliant on ICOS stimulation. Additionally, this work suggests that measuring ICOS:ICOSL ratios may be a method to monitor disease progression and severity in autoimmune models reliant on these costimulatory molecules. Overall, altering the proteolytic regulation of ICOSL by ADAM10 may be a way to modulate the immune response, particularly in the context of humoral autoimmunity.

We thank Matthew Zellner for help in mouse colony management.

The authors have no financial conflicts of interest.