IDO2 Modulates T Cell–Dependent Autoimmune Responses through a B Cell–Intrinsic Mechanism

This article is featured in In This Issue, p.4421

Autoimmune diseases such as rheumatoid arthritis and lupus that are generally poorly managed clinically pose a growing challenge in developed countries. At present, there is little understanding of the pathogenic etiology of autoimmune disease or the modifier pathways that may affect the course and kinetics of its clinical development or severity. At present, major efforts focus on whole-genome genetic and epigenetic screens to elucidate etiologic drivers, but there has been less attention on novel principles of immunomodulation that may function as disease modifiers. Such efforts may be useful in illuminating questions about individual variations in the kinetics and severity of disease development, as well as offering new therapeutic directions to attenuate disease.

The enzymes IDO1 and IDO2 catabolize tryptophan (Trp) and various Trp-related compounds that modify inflammatory state and immune tolerance. These two enzymes resulted from an ancient gene duplication of an ancestral IDO with relatively low Trp catalytic activity (1). The immunoregulatory properties of IDO were first revealed in pharmacological studies of an IDO pathway inhibitor, which suggested a critical role in maintaining maternal–fetal tolerance through a T cell–dependent mechanism (2). Subsequently, numerous pharmacological and genetic studies linked the IDO pathway to immune escape in cancer (3–5) and as a contributor to autoimmunity (6–8). IDO1, the better characterized of the two enzymes, modulates the immune system primarily through alterations in regulatory T cell populations, an effect likely mediated via a population of IDO1-expressing dendritic cells (DCs) (9). In addition, a role for IDO1 in B cells in regulating T-independent responses has recently been reported (10). Mechanistically, IDO1 signals through the GCN2 and mammalian target of rapamycin (mTOR)-mediated stress response pathways in response to Trp depletion (11–13). IDO2, a low-efficiency Trp-catabolizing enzyme, was only recently directly connected to immunomodulation (14–16), and less is known about the cellular and molecular mechanisms through which it influences immunity, though it is clear that IDO2 does not simply serve a redundant function to IDO1 (15). IDO2 expression is more restricted than IDO1, with high expression levels limited to liver, kidney, and cerebral cortex (17). IDO2 is also expressed in APCs, particularly DCs (16), as well as macrophages and B cells (15). Notably, the relative contributions of IDO1 and IDO2 to various immunological phenomena are somewhat convoluted given that many published studies inhibit IDO through the use of the small-molecule inhibitor 1-methyltryptophan (1MT), which influences both IDO1 and IDO2 (5). In some reports, blocking IDO with 1MT was observed to exacerbate autoimmune disease (6, 18, 19), whereas in other reports, it was found to alleviate disease (8, 20). Although the basis for these conflicting observations is unclear, they highlight the importance of genetic knockouts rather than nonspecific small-molecule inhibitors in isolating the inflammatory roles played by the IDO enzymes in different disease settings.

Recently, we created an IDO2-deficient (knockout [ko]) mouse (15) to isolate the immunologic contributions of the two IDO enzymes. Using these mice, we have defined a critical role for IDO2 distinct from IDO1 in mediating inflammation in murine models of contact hypersensitivity (CHS) and autoimmune arthritis (14, 15). Despite the clear role of IDO2 in modulating autoimmune and inflammatory responses, little is known about the mechanism by which it acts. Initial studies using the KRN model of arthritis demonstrated a reduction in autoreactive T and B cell responses, resulting in attenuated joint inflammation in IDO2-deficient mice (14). Although there was a pronounced defect in Th cells, reciprocal adoptive transfer experiments demonstrated that the effect of IDO2 was extrinsic to T cells. In this study, we define the cellular mechanism through which IDO2 mediates inflammatory autoimmunity, demonstrating that IDO2 acts directly in B cells to drive inflammation in models of arthritis and CHS.

KRN TCR transgenic (tg) (21) and IDO2 ko (15) mice on a C57BL/6 background have been described. C57BL/6 IDO2 wild-type (wt) and ko mice lacking the TCR α-chain (TCR ko) and carrying a single copy of the MHC class II I-A^g7 allele (TCR ko B6.g7/b and TCR ko IDO2 ko B6.g7/b) were generated as recipient mice for adoptive transfer of KRN T cells. Mice must carry both an I-A^g7 and an I-A^b allele to prevent rejection of the injected KRN T, which are I-A^b. T cell donor mice were IDO2 ko KRN TCR tg (IDO2 ko KRN B6) carrying two copies of the I-A^b allele. B cell donor mice for addback experiments were wt or IDO2 ko I-A^g7/b, wt or IDO2 ko I-A^b/b, glucose-6-phosphate isomerase (GPI)–specific B cell tg mk147 I-A^g7/b, or hen egg lysozyme (HEL)–specific B cell tg MD4 I-A^g7/b C57BL/6. Additionally, C57BL/6 Rag ko mice were crossed with NOD Rag ko mice to generate B6 × NOD Rag ko I-A^g7/b recipients for adoptive transfers. For CHS experiments, IDO2 wt and ko BALB/c mice were used as both B cell donor and recipient mice. All mice were bred and housed under specific pathogen-free conditions in the animal facility at the Lankenau Institute for Medical Research. Studies were performed in accordance with National Institutes of Health and Association for Assessment and Accreditation of Laboratory Animal Care guidelines with approval from the Lankenau Institute for Medical Research Institutional Animal Care and Use Committee.

Spleen and lymph node tissue from C57BL/6 mice was harvested and passed through a 70-μm nylon strainer to generate a single-cell suspension. CD4⁺ T cells from KRN TCR tg (KRN B6) or IDO2 ko KRN TCR tg (IDO2 ko KRN B6) mice were purified by positive selection with anti-CD4 mouse MACS microbeads (Miltenyi Biotec). For T cell purification, elutant was purified over a second column to achieve higher purity (∼90%). Following purification, 3.5 × 10⁵ CD4⁺ T cells were adoptively transferred i.v. into TCR ko B6.g7 or TCR ko IDO2 ko B6.g7 hosts. For B cell addback experiments, B cells from spleens of IDO2 wt or ko I-A^g7/b were purified by anti-CD43 negative selection with MACS beads (Miltenyi Biotec). B cell purity was routinely ∼97%. A total of 0.5–10 × 10⁶ B cells was transferred with T cells. Arthritis was measured as described below. Mice were sacrificed after 2 wk.

To purify subpopulations of B cells, spleens were harvested from IDO2 wt or ko I-A^g7/b donor mice and B cells isolated by negative selection with anti-CD43 microbeads (Miltenyi Biotec). Following purification, B cells were stained with appropriate surface markers, either B220 and CD11c for isolation of the plasmacytoid DC–like population or B220, CD21, and CD23 for isolation of marginal zone (MZ) and follicular (FO) B cells. Stained populations were sorted with a BD FACSAria III flow sorter (BD Biosciences). Following isolation, 5 × 10⁵ of the appropriate population of B cells was coinjected with 3.5 × 10⁵ KRN T cells into TCR ko IDO2 ko B6.g7 hosts. Arthritis was measured as described below. Mice were sacrificed after 2 wk.

The two rear ankles of experimental mice were measured starting at the day of adoptive transfer. Measurement of ankle thickness was made above the footpad axially across the ankle joint using a Fowler Metric Pocket Thickness Gauge (Fowler; A&M Industrial Supply, Rahway, NJ). Ankle thickness was rounded off to the nearest 0.05 mm. Change in ankle thickness was defined as (measured ankle thickness) − (thickness prior to adoptive transfer of B and T cells).

IDO2 wt or ko BALB/c recipient mice were sensitized with 3% oxazolone (Sigma-Aldrich) in 100% ethanol on their shaved abdomen (100 μl) and hind footpads (10 μl each) 5 d prior to the start of the experiment. Three days later, IDO2 wt or ko B cell donor mice were sensitized using the same protocol. After 24 h, spleen and inguinal lymph nodes were harvested from donor mice, B cells isolated with the Pan-B cell bead kit (Miltenyi Biotec), and 5 × 10⁵ B cells transferred into each recipient mouse. At 24 h following B cell transfer (5 d after initial sensitization), recipient mice were elicited with 20 μl 1% oxazolone in 100% ethanol or 100% ethanol alone as a control, 10 μl on each side of the ears. After 24 h, ear thickness was measured using a dial gauge (Fowler; A&M Industrial Supply), and harvested ears were fixed in 10% buffered formalin, embedded in paraffin, sectioned, and stained with H&E. Histology sections were imaged using a Zeiss Axioplan microscope with a Zeiss Plan-Apochromat ×10/0.32 objective and Zeiss AxioCam HRC camera using AxioVision 4.1 software (Carl Zeiss). The images were then processed using Adobe Photoshop CS2 software (Adobe Systems). Trials performed in multiple mice were replicated at least three times.

For in vitro experiments, B cells were harvested from spleens of C57BL/6 mice and purified by negative selection using anti-CD43 microbeads (Miltenyi Biotec). Purified B cells were cultured at 2 × 10⁶ cells/ml in IMDM plus 10% FCS plus 5 μM 2-ME, 2 mM glutamax (Life Technologies), and 50 μg/ml gentamicin. Stimuli and cytokines were added to a final concentration of: LPS (25 μg/ml; Sigma-Aldrich), anti-CD40 (2 μg/ml; eBioscience), anti-IgM F(ab′)₂ (20 μg/ml; Jackson ImmunoResearch Laboratories), CpG (ODN 1826; 1 μM; Invivogen), IL-4 (50 ng/ml; eBioscience), IL-6 (10 ng/ml; eBioscience), IL-21 (100 ng/ml; eBioscience), and IFN-γ (500 U/ml; PeproTech).

Spleen tissue from C57BL/6 mice was harvested and passed through a 70-μm nylon strainer to generate a single-cell suspension. RNA was extracted with TRIzol (Invitrogen) and first-strand cDNA synthesized using oligo-dT primer (Promega GoScript). IDO2 expression was measured by real-time PCR using TaqMan for detection (ABsolute qPCR ROX Mix; ThermoFisher) or SYBR Green (Sigma-Aldrich). Expression of target gene IDO2 was determined relative to β₂-microglobulin (β2M) and calculated as 2^⁻(Ct_{Target gene} − Ct_b2M) as primers had similar efficiencies. Primers for TaqMan mouse IDO2 and β2M were from Life Technologies. SYBR Green primers for IDO2 were 5′-GCCCAGAGCTCCGTGCTTCAT-3′ and 5′-TGGGAAGGCGGCATGTAGTCC-3′ and for β2M, 5′-CTCGGTGACCCTGGTCTTTC-3′ and 5′-TTGAGGGGTTTTCTGGATAGCA-3′.

IDO2 wt and ko C57BL/6 spleen cells were harvested and stimulated for 72 h with anti-IgM F(ab′)₂ (20 μg/ml; Jackson ImmunoResearch Laboratories) or anti-CD40 (2 μg/ml; eBioscience) plus IL-21 (100 ng/ml; eBioscience) in IMDM plus 10% FCS plus 5 μmol 2-ME, 2 mmol Glutamax (Life Technologies), and 50 μg/ml gentamicin. Levels of the following costimulatory markers were directly measured by flow cytometry on a BD FACSCanto (BD Biosciences) with subsequent analysis using FlowJo Software (Tree Star): CD40 (eBioscience), CD80 (eBioscience), CD86 (eBioscience), ICOS ligand (ICOSL; BioLegend), IL-21R (BD Biosciences), OX40 ligand (OX40L; BioLegend), programmed cell death ligand-1 (PD-L1; BioLegend), and MHC class II (eBioscience). A minimum of three replicate mice were analyzed per condition.

Statistical significance was determined using one-way ANOVAs followed by comparison of means with Tukey post hoc multiple comparison correction or Kruskal–Wallis nonparametric ANOVA with Dunn multiple comparison correction as appropriate using Prism6 (GraphPad Software).

The KRN T cell tg mouse model of autoimmune arthritis (21) is a tractable model system that allows us to distinguish the specific contribution of IDO2 to the autoimmune response. In this model, the KRN T cell recognizes the ubiquitous autoantigen GPI, a glycolytic enzyme, in the context of the MHC class II molecule I-A^g7 (22). Unlike other models of arthritis, no adjuvants are required for disease induction. Although it is a T cell tg model, arthritis is induced by the production of pathogenic autoantibodies resulting from the activation of endogenous B cells. KRN mice can be used to model arthritis in multiple ways. First, arthritis can occur spontaneously, by breeding the KRN T cell tg to a mouse carrying I-A^g7 (Fig. 1A) (21). The resulting KRN.g7 mice develop robust arthritis in the presence of wt IDO2, but a greatly attenuated joint inflammatory response in the absence of IDO2 (IDO2 ko KRN.g7) (14, 23). Arthritis can also be induced by adoptively transferring KRN T cells into a T cell–deficient mouse expressing I-A^g7 (Fig. 1B) (24). Transfer of KRN T cells results in a robust arthritic response in recipient mice expressing wt IDO2. Recapitulating what is seen in the spontaneous model, arthritis is greatly reduced in IDO2 ko hosts (14). To define the cell-specific role of IDO2, in this paper, we expand the T cell adoptive transfer model by performing addback of different B cell types to IDO2 ko T cell–deficient hosts (Fig. 1C). Because the host has endogenous IDO2 ko B cells, these are termed addback experiments to clarify that additional B cells are being added to an existing milieu.

Previously, we have shown that IDO2 is crucial for the B and T cell–dependent initiation phase responsible for production of autoantibodies, whereas IDO2 is dispensable for the downstream T and B cell–independent effector phase of disease (14, 25). Reduced differentiated Th subsets, coupled with a decrease in serum autoantibody titers and autoantibody-secreting cells in IDO2 ko arthritic mice, suggested that IDO2 may function at the interface between a T cell and an APC (14). Reciprocal adoptive transfer experiments demonstrated conclusively that this defect is not intrinsic to the T cell, thereby implicating a role for IDO2 in APCs (14).

IDO2 expression has been reported in several types of APCs, including DCs, macrophages, and B cells (15–17), and we hypothesized that IDO2 exerted its critical function in one or more of these cell types to influence the T cell response. To separate the action of IDO2 in innate versus adaptive immune cells, we performed a series of adoptive transfers to induce autoimmune arthritis in Rag-deficient hosts. In this study, IDO2 wt or ko KRN T cells and IDO2 wt or ko B cells were transferred into Rag ko C57BL/6 × NOD recipient mice (Fig. 2A). Recipient mice carried one copy of the MHC class II allele I-A^g7 from the NOD parent. Components of the innate immune system were derived from the recipient mouse and were thus wt for the IDO2 allele. We found that arthritis developed only in mice that received wt B cells, regardless of whether wt or IDO2 ko KRN T cells were transferred (Fig. 2B). Mice that received IDO2 ko B cells did not develop arthritis, not even the reduced level of joint inflammation previously described in the spontaneous IDO2 ko arthritis model (14). It is unclear why arthritis was absent; however, inflammation levels in general were lower in the Rag ko recipient mice than in spontaneous arthritis, suggesting some IDO2-independent contribution of endogenous B cells to the sustained autoimmune response present in the spontaneous model. In the Rag ko recipient mice, all components of the immune system, including DCs and macrophages from the host and transferred T cells, were wt with the exception of transferred IDO2 ko B cells, demonstrating that wt IDO2 in B cells is necessary for arthritis development.

Although the experiments in the Rag ko hosts established that IDO2 expression in B cells is necessary for arthritis, they did not address whether IDO2 expression in B cells is sufficient for the complete response or if IDO2 expression in additional cell types is needed to induce robust arthritis. To test this, we performed a series of B cell addback adoptive transfers into TCR ko IDO2 ko hosts (Fig. 2C), examining whether addition of IDO2 wt B cells is sufficient to restore the arthritic response to levels seen in wt hosts. Note that in contrast to the Rag ko experiments (Fig. 2A, 2B), in this study, the host mouse and all of its immune components are IDO2 deficient. Recipient mice lack T cells but harbor an intact IDO2 ko innate immune system as well as endogenous IDO2 ko B cells. This allowed us to separate IDO2’s role specifically in B cells from that in other cell types, as well as from other critical non-IDO2 functions of B cells (e.g., establishment of proper lymphoid architecture) (26). For this series of experiments, purified IDO2 ko KRN T cells were adoptively transferred into TCR ko IDO2 ko mice. Concurrent with the T cell transfer, either wt or IDO2 ko purified B cells were transferred to the recipients so that the only source of IDO2 would be from the transferred IDO2 wt B cells—all other immune cells, including the endogenous population of B cells, being IDO2 ko. Addition of wt B cells was sufficient to restore the wt arthritic response (Fig. 2D). This outcome was not simply the result of the presence of additional B cells, as addition of IDO2 ko B cells did not restore arthritis above the attenuated level observed without the transfer of additional B cells (Fig. 2D), recapitulating the reduced joint inflammatory response previously described in IDO2 ko KRN.g7 mice (14). Together, these data demonstrate IDO2 expression in B cells is both necessary and sufficient in the KRN model to support robust arthritis development.

B cells have been shown to play both Ag-specific and nonspecific (bystander) roles in the activation and recruitment of Th cells (27, 28). To determine which of these roles requires a contribution from IDO2, we first examined whether a cognate interaction between wt B cells and other cell types is required for robust arthritis. As described, in the model of arthritis used in this study, KRN TCR tg T cells bind the autoantigen GPI in the context of the MHC class II molecule I-A^g7. To test whether the B cells in our addback experiments were directly involved in this TCR/MHC interaction, we transferred B cells purified from C57BL/6 mice that either could (I-A^g7/b, cognate) or could not (I-A^b/b, noncognate) present Ag to KRN T cells. Only wt B cells expressing I-A^g7 were able to restore arthritis to wt severity, suggesting that IDO2 is required for efficient Ag presentation (Fig. 3A).

To determine if IDO2-expressing B cells need to be Ag specific, we used BCR tg models with defined Ag specificities as a source of IDO2-expressing B cells in the addback strategy described above. Purified B cells from anti-GPI BCR tg mice [mk147 I-A^g7/b (29)] were used as a source of Ag-specific B cells and anti-HEL BCR tg mice [MD4 I-A^g7/b (30)] as a source of non-Ag–specific B cells. GPI Ag specificity in the IDO2 wt B cells was required for the complete arthritic response as only the mk147 B cells were able to restore arthritis in an IDO2 ko host to wt levels (Fig. 3B). To rule out the restored arthritis response being due to differential IDO2 expression in the transferred B cells, we confirmed that the level of IDO2 in the transferred cells did not differ between the different donor B cell types prior to adoptive transfer (Fig. 3C). Together, these data suggest that IDO2 is acting in a pathway that directly affects the Ag-presentation function of B cells to an interacting T cell.

Mature B cells can be divided into subsets, each with distinct phenotypic and functional characteristics. To further examine the role of B cells in arthritis, B cell subsets were compared for their ability to restore wt arthritis in the addback adoptive transfer model described above. There has been extensive interest in the population of B220⁺CD11c⁺ cells, which have characteristics of both B cells and DCs and have been linked to immune regulation. This population, often considered a subset of plasmacytoid DCs, though its exact lineage remains obscured, contains a high proportion of IDO1⁺ cells that have been reported to contribute to immune escape in cancer through the activation of regulatory T cells (31–35). Age-associated B cells are also B220⁺CD11c⁺ and have been shown to be important mediators of spontaneous autoimmune responses in lupus-prone mice and human patients with rheumatoid arthritis and systemic lupus erythematosus (36–38). To examine whether this subset contributes to IDO2-mediated arthritis development, B220⁺CD11c⁺ cells were sorted and compared with B220⁺CD11c⁻ cells in a B cell addback adoptive transfer approach. B220⁺CD11c⁻ cells restored wt arthritis in IDO2 ko recipients, whereas the mice that received the B220⁺CD11c⁺ population exhibited an attenuated arthritic response that was not significantly different from the control IDO2 ko recipients that did not receive B cells (Fig. 4A). Thus, although the B220⁺CD11c⁺ population has a significant role in the interplay between IDO1 function and T cell regulation, it does not appear to be important for IDO2-mediated B cell function in autoimmunity.

Marginal zone (MZ) B cells also contribute directly to autoimmunity, a phenomenon particularly well documented in the KRN model. MZ B cells (B220⁺CD21^highCD23^low), a rare, noncirculating B cell population confined to the splenic MZ and separated from the splenic follicle by the marginal sinus, are important for the anti-GPI autoantibody response. These cells are spontaneously activated and contain a high proportion of anti-GPI B cells (29). In contrast, FO B cells (B220⁺CD21^intCD23^int) are, in general, antigenically ignorant to GPI. To assess the relevance of IDO2 expression in these two distinct B cell subtypes, sort-purified populations of IDO2 wt MZ and FO B cells were coinjected with IDO2 ko KRN T cells into TCR ko IDO2 ko recipient mice. Transfer of MZ B cells alone generated a stronger arthritic response than the no–B cell control. In contrast, arthritis in mice with FO B cells added back was reduced compared with the MZ B cell addbacks and not significantly different from the control (Fig. 4B). This may be because GPI-specific MZ B cells are effectively preactivated, with higher basal levels of activation markers including CD69, CD80, and CD86, leading to a rapid anti-GPI response (29). Alternatively, due to the requirement for IDO2 in Ag-specific B cells, the observed differential in arthritic responses could be due to a higher proportion of GPI-reactive cells in the MZ compartment. To control for GPI specificity, MZ and FO B cells were sorted from mk147 anti-GPI BCR tg mice (29). This allowed us to compare the effect of MZ versus FO B cells that all recognize the GPI autoantigen. Both MZ and FO B cells from mk147 tg mice were able to restore wt arthritis in IDO2 ko recipients (Fig. 4C), indicating that both populations can contribute to the IDO2-mediated proinflammatory effect. However, similar to non-tg B cells, the mk147 tg MZ B cells elicited a stronger arthritic response compared with the mk147 tg FO B cells, demonstrating that even when controlling for autoantigen specificity by using tg B cells, IDO2 is particularly important in the MZ B cell population.

To determine if IDO2 expression in B cells is required to drive inflammation outside the context of the KRN model, we used a second model of inflammation shown to be IDO2 dependent, the CHS model. Previous studies have shown that IDO2 ko mice exhibit significantly reduced ear swelling and associated levels of inflammatory (IL-6, TNF-α, and IFN-γ), chemotactic (MCP1/CCL2), and hematopoietic (GM-CSF and G-CSF) cytokines in response to hapten challenge (15). Although considered a classic model of T cell immunity, the CHS response has been shown to also require the involvement of anti-hapten B cells and Abs (39, 40). This allowed us to test whether IDO2 mediates inflammation in CHS responses through the same B cell–specific mechanism we described for arthritis.

The role of IDO2 was examined using a B cell addback experiment (Fig. 5A). Recipient and donor IDO2 wt and ko mice were first sensitized with oxazolone. Total B cells were isolated from donors and subsequently adoptively transferred to recipient mice. Mice were challenged with oxazolone on the ears and swelling measured after 24 h. Control transfers showed a robust CHS response when IDO2 wt B cells were transferred into wt recipients and a significantly reduced response when IDO2 ko B cells were transferred to ko recipients. As in the arthritis model, adoptive transfer of IDO2 wt but not ko B cells restored a robust CHS response in IDO2 ko recipients (Fig. 5B). The increased swelling in the hosts with transferred wt B cells was confirmed by histology (Fig. 5C). The role of IDO2 in B cells is thus not limited to autoimmunity or to the KRN model, but is a general phenomenon contributing to inflammatory immune responses in two distinct model systems.

Having established that it is both necessary and sufficient for IDO2 to be expressed in B cells in order promote arthritis and CHS, we proceeded to investigate the pathways responsible for regulating the expression of IDO2 in B cells. Because IDO2 ko mice have diminished T cell responses, we first tested whether B cells might upregulate IDO2 in response to T cell–mediated costimulatory molecules and cytokines. The interaction between B and T cells required to produce an Ab response is complex and multidirectional, requiring not only the direct interaction of a TCR with an MHC class II molecule, but also a wide variety of costimulatory molecules including ICOS/ICOSL, CD40L/CD40, and PD-L1/programmed cell death-1 (PD1), as well as cytokines such as IL-4 and IL-21. We sought to mimic this interaction through in vitro stimulation of B cells and subsequent examination of the effect of various stimuli on the expression of IDO2 mRNA. To mimic T cell help via costimulation, purified B cells were cultured with anti-CD40 (Fig. 6A). Alone, this stimulus has no effect on IDO2 message. Because cytokine signals are critical for effective T–B cell crosstalk, several cytokines were subsequently tested in concert with anti-CD40. IL-4, IL-6, and IL-21 are all downregulated in IDO2 ko arthritic mice, suggesting a connection between these signals and IDO2 (14). These cytokines, especially IL-4 and IL-21, are intricately involved in the interaction between B and T follicular helper (T_FH) cells (41). IFN-γ induces IDO1 expression (42, 43), thus we investigated whether it would also upregulate IDO2. Forty-eight–hour in vitro stimulation of purified B cells with anti-CD40 upregulated IDO2 expression only in conjunction with IL-4 or IL-21, but not IL-6 or IFN-γ (Fig. 6A).

Next, we investigated whether T-independent, polyclonal stimuli can upregulate IDO2 in B cells. LPS, a potent immune-stimulating pathogen-associated molecular pattern produced by Gram-negative bacteria, signals in B cells, and other APCs through TLR4, a surface-expressed TLR. LPS induced low levels of IDO2 message in purified B cells, but upregulated IDO2 message dramatically when IL-4 or IL-21 was added (Fig. 6A). CpG ODN is an agonist of TLR9, a TLR expressed in endosomal compartments of several immune cell types, including B cells. Like anti-CD40, CpG only upregulated IDO2 in conjunction with IL-4 or IL-21, suggesting IDO2 functions upstream or independently of IFN-γ and IL-6 production. Neither LPS nor CpG alone had a significant effect on IDO2 message, indicating that T-independent stimuli also require additional signaling in the for IDO2 upregulation. Thus, both signal 2 and an appropriate cytokine signal influence IDO2 transcription.

B cell signaling is initiated by the binding of the BCR with a cognate Ag (signal 1). This can be mimicked by anti-IgM, a stimulus that binds the BCR receptor and initiates the BCR signaling cascade. We tested whether anti-IgM alone or in concert with cytokine signals could upregulate IDO2. Surprisingly, anti-IgM did not upregulate IDO2 expression in B cells under any conditions tested (Fig. 6A). The increase in IDO2 seen in the anti-IgM plus IL-21 condition was not significantly different from the slight upregulation in IDO2 seen with IL-21 alone (Fig. 6A). Thus, IDO2 is not directly induced by stimulation of the BCR (anti-IgM, signal 1), but requires a T-independent or T-dependent stimulus (signal 2) plus a cytokine (signal 3).

Previously, we showed that the IDO pathway only needed to be inhibited during the onset of the autoimmune response to provide protection against arthritis (8), suggesting that IDO2 may be expressed transiently in B cells during their activation. To determine the kinetics of IDO2 expression, B cells were stimulated with anti-CD40 plus IL-21 or LPS plus IL-21, the two conditions that produced the most dramatic upregulation of IDO2. IDO2 induction occurs rapidly, with IDO2 message detectable within the first 5 h of culture. Expression peaked at ∼24 h in culture and was sustained over the next 2 d (Fig. 6B). No significant differences in the timing of induction were seen with LPS plus IL-21 versus anti-CD40 plus IL-21. Taken together, these data demonstrate that IDO2 expression in B cells is quickly and dramatically upregulated in response to specific costimulatory and cytokine signals and remains elevated for several days.

The upregulation of IDO2 by anti-CD40 plus cytokines suggests that IDO2 expression is associated with T cell help and the crosstalk between B and T cells necessary for a robust B cell response. We thus sought to determine whether IDO2-deficient B cells have an intrinsic defect either in Ag presentation or costimulation affecting this interaction. Given the idea that IDO2 may be affecting T–B collaboration, we assessed the expression of B cell costimulatory markers in IDO2 ko compared with wt B cells in response to in vitro stimulation. Cells were stimulated with either anti-IgM alone or anti-CD40 plus IL-21 and levels of B cell costimulatory markers directly measured by flow cytometry (Fig. 7). When stimulated with anti-IgM alone, wt B cells upregulate costimulatory molecules needed for collaboration with T cells, including CD40, the B7 markers (CD80/86), ICOSL, IL-21R, OX40L, PD-L1, and the MHC class II molecule itself (Fig. 7A, 7B). IDO2 ko B cells also upregulate CD80, CD86, ICOSL, IL-21R, OX40L, PDL1, and MHC class II in response to anti-IgM stimulation. In contrast, CD40 levels are reduced on IDO2 ko B cells in response to anti-IgM by ∼2.5-fold at 72 h in culture (Fig. 7B). This differential effect of anti-IgM signaling on IDO2 ko B cells as compared with their wt counterparts indicates that IDO2 ko B cells may have a defective ability to elicit T cell help after initial encounter with Ag (signal 1) due to a reduction in the CD40 costimulatory molecule. To determine if IDO2 ko B cells are defective in their ability to upregulate costimulatory markers in response to other stimuli, IDO2 wt and ko splenocytes were cultured with anti-CD40 plus IL-21, culture conditions that mimic signals derived from T cell help. No differences between IDO2 ko and wt cells were seen under these conditions for any of the markers tested (Fig. 7C). This suggests that there is not a global defect in the ability of IDO2 ko B cells to upregulate costimulatory markers. These data demonstrate that IDO2 ko B cells can indeed become activated, express costimulatory molecules, and act as APCs when sufficient signals are received, as seen when anti-CD40 and IL-21 are supplied, but suggests that the lack of IDO2 compromises the activation process under other conditions, including anti-IgM stimulation, that may be relevant to the deficiencies observed in IDO2 ko B cells in vivo.

Novel immune modulatory mechanisms that modify the severity of autoimmune disorders may help illuminate their pathogenicity and promote insights into their clinical management. The increasingly apparent role of IDO2 as an immune modulator has led to important questions regarding its mechanism of action, particularly with regard to its better-studied family member IDO1. Little is known about the enzymatic, cellular, and molecular functions of IDO2, but there is evidence of differences at each level (44). IDO2 is expressed in DCs (15, 16), suggesting it acts like IDO1 to modulate the consequences of Ag presentation, but the precise consequences of IDO2 expression in this setting have yet to be determined. Our previous work has suggested a role for IDO2, but not IDO1, in mediating B and T cell responses in the KRN model of autoimmune arthritis, as evidenced by reduced autoantibody production, number of Ab-secreting cells, and differentiated Th cell populations in IDO2 ko arthritic mice (14). Initial experiments also conclusively demonstrated this effect to be T cell extrinsic, suggesting that IDO2 may be involved in Ag presentation to T cells. Our work in this study confirms and extends IDO2’s role in autoimmune arthritis by defining its critical function in B cells, one that is consistent with feedback control on T cell help needed for autoantibody generation. Notably, in vitro studies revealed a reduction in levels of the costimulatory marker CD40 on B cells. Given this result, along with the previously identified reduction of T_FH-associated cytokines IL-4 and IL-21 in IDO2 ko arthritic mice (14), we propose in this paper that IDO2 influences the effects of Ag presentation to T cells by modifying the costimulatory crosstalk between B and T cells.

Previously, our laboratory reported that IDO2 ko mice exhibit deficient autoreactive T and B cell responses mediating arthritis and that these defects are extrinsic to T cells (14). In our current study, we define the B cell as being the critical IDO2-expressing cell necessary for arthritis development. In addition to the secretion of pathogenic autoantibodies, B cells have many important functions during the onset and progression of immune responses (23, 45–48). In particular, they are highly potent APCs for specific Ags recognized and internalized through the BCR (49, 50), a role that is important in T cell activation. B cells can also act in a bystander role, a function not dependent on a cognate, Ag interaction with the TCR. Through a series of adoptive transfer experiments, we found that IDO2 must function in cognate Ag-specific B cells to mediate a robust arthritic response, suggesting a direct role for IDO2 in mediating the collaboration between T and B cells.

Mature B cells can be separated into subsets, based on their phenotypic and functional characteristics. IDO2 was found to exert particularly dramatic effects in MZ B cells. Adoptive transfer of IDO2 expressing MZ B cells into IDO2 ko recipients significantly increases arthritis development compared with transfer of conventional FO B cells. This effect also seems to be dependent on the B cell specificity, with autoantigen-specific mk147 tg MZ B cells exhibiting a particularly dramatic effect on arthritis development. Interestingly, the Abs produced by the mk147 tg B cells are IgM and generally do not contribute directly to disease pathogenesis, which requires anti-GPI IgG autoantibodies (22, 29). The fact that mk147 tg B cells are able to restore wt levels of arthritis suggests that these cells elicit T cell help, with these T cells then inducing endogenous IDO2 ko B cells to produce the pathogenic anti-GPI IgG, further supporting the hypothesis that IDO2 plays an important role in the cross-talk between autoreactive B and T cells.

In the CHS model, B cells also act at the initiation of the autoimmune response, though the relevant B cell population may differ from what is observed in the KRN model. Unlike the arthritis model, in which we demonstrate that classical B-2 B cell populations are critical for the arthritic response, it is the B-1 B cell population that has been identified as important for initiation of the CHS elicitation phase response (39). In fact, B-2 B cells, including MZ B cells, may be suppressive under some conditions (51). In CHS, B-1 B cells generate IgM within 24 h following hapten challenge (52) and mediate the CHS response through a complement-dependent mechanism allowing for T cell recruitment (53). The role of IDO2 in this process remains to be discovered, but there are several important parallels with the KRN model, particularly in the role of B cells in Ag presentation and T cell recruitment. In addition, as in the KRN model, the CHS model demonstrates a unique and separate role for IDO2 as compared with IDO1 (15).

In vitro data give us insight into both upstream regulators and downstream effectors of IDO2. We found that signal 1 by itself, as triggered by anti-IgM, is insufficient to upregulate IDO2, yet IDO2 ko B cells nevertheless responded differently to this stimulus by generating a less robust costimulatory response. A mimic of T cell help, anti-CD40, also did not upregulate IDO2 on its own, but did so in concert with the T_FH-associated cytokines IL-4 or IL-21. The T-independent stimuli LPS and CpG, which with IL-4 or IL-21 also upregulated IDO2, act through TLRs to promote B cell activation. Specifically, TLR4 (LPS) and TLR9 (CpG) engage MyD88 to initiate a downstream signaling cascade that can activate the MAPK and NF-κB pathways, leading to the production of cytokines and various effector molecules (54, 55). TLR9 has been linked with autoimmunity both in the context of autoantibody production and maintenance of tolerance (56, 57). CpG is also a potent inducer of IFN-γ and IL-6 (58). IDO2 is likely upstream of both of these cytokines, as neither IFN-γ nor IL-6 affected IDO2 expression in any context measured.

Studies with autoimmune mice highlight the differences in immune regulatory functions between IDO1 and IDO2 and the importance of targeting the two molecules individually in a therapeutic context. Existing small-molecule inhibitors, particularly 1MT, likely do not directly affect IDO2 itself, but rather influence the IDO pathway downstream of Trp catabolism (5). IDO1 cannot be understood wholly as an anti-inflammatory or immunosuppressive factor, but it clearly influences production of regulatory T cell populations and in that sense acts to promote immunological tolerance, as illustrated in models of maternal–fetal tolerance and tumoral immune evasion (5). IDO2, in contrast, appears to have a more proinflammatory role in immune modulation, as evidenced by the exacerbated arthritis and CHS responses seen in IDO2 wt versus ko mice (14, 15). B220⁺CD11c⁺ cells have been shown to be an important cell type for IDO1-mediated inhibition of T cell function (31–35). Like IDO1, IDO2 appears to affect T cell help indirectly, as shown in the current study through its expression in B cells. Importantly, IDO2 wt B220⁺CD11c⁺ cells did not restore wt arthritis when adoptively transferred into IDO2 ko hosts, further clarifying the distinctive roles of IDO1 and IDO2 in mediating immune responses. Because these two immune modulatory enzymes have nonoverlapping and, in some contexts, contrasting immunoregulatory functions, it is critical that specific inhibitors be developed to target them selectively.

In considering the medicinal import that IDO2 may offer for selective treatment of autoimmunity, we have noted previously that although IDO2 ko mice exhibit an attenuated response to an autoantigen, they respond similarly to wt mice after immunization with a model neoantigen (14). Similarly, in this study, although we see strong, IDO2-dependent differences in B cell–mediated inflammatory responses in vivo, these differences can be overcome in vitro when appropriate external stimuli and cytokines are provided, as in the case of anti-CD40 plus IL-21. This suggests that IDO2 ko B cells are not intrinsically defective but that in vivo they are not receiving appropriate or sufficient stimuli to trigger pathogenic autoantibody production. This was also mimicked in our in vitro cultures stimulated with signal 1 (anti-IgM) alone. We hypothesize that adjuvants necessary for immunizations with model Ags provide signals that overcome the loss of IDO2, similar to what is seen with our in vitro data with anti-CD40 plus IL-21–stimulated cells. This further supports the idea that IDO2 is important in the initiation phase of immune responses, as adjuvant compounds can effectively bypass some of the initial signals that are necessary for a spontaneous response.

In summary, our work demonstrates that IDO2 functions as a modifier in B cells to control pathogenic inflammation and autoimmunity. The identification of the relevant cell type for the action of IDO2 in modulating immune responses is critical for determining the molecular mechanism by which IDO2 can exert its apparent selective effects on pathogenic inflammatory phenomena. Additionally, this finding may offer potential medical import with regard to IDO2 targeting strategies for autoimmunity, akin to those similarly leveraged for IDO1 inhibitors in cancer.

J.B.D., G.C.P., and A.J.M. are inventors and shareholders and G.C.P. is a grant recipient and member of the scientific advisory board for New Link Genetics Corporation, which has licensed IDO- and IDO2-related intellectual property from the Lankenau Institute for Medical Research. The other authors have no financial conflicts of interest.