Novel Tryptophan Catabolic Enzyme IDO2 Is the Preferred Biochemical Target of the Antitumor Indoleamine 2,3-Dioxygenase Inhibitory Compound d-1-Methyl-Tryptophan

Tryptophan catabolism by indoleamine 2,3-dioxygenase (IDO) mediates a protolerogenic mechanism that suppresses T cells, providing balance or feedback control in immune reactions (1, 2). This role for IDO was first established with the demonstration that the specific bioactive IDO inhibitor 1-methyl-tryptophan (1MT; 3) can trigger T cell–mediated rejection of allogeneic mouse concepti (4, 5). More recently, IDO has become recognized as a central mediator of immune tolerance in many settings. In cancer, IDO expression in tumor cells and antigen-presenting cells present in tumor-draining lymph nodes mediates an important mechanism of immune escape (6). IDO inhibitors trigger antitumor immunity (7, 8) and act synergistically with conventional or experimental chemotherapies (9, 10). Based on preclinical efficacy studies, the d stereoisomer of 1MT has emerged as a clinical lead inhibitor that is entering human trials. D-1MT has superior antitumor activity relative to the l stereoisomer in most preclinical models, and IDO is genetically required for the activity of D-1MT (11). However, at the level of biochemical specificity, the distinction between the two isomers is complicated, with the d isomer exhibiting little biochemical activity as an IDO inhibitor relative to l isomer (11). In dendritic cells, both isomers block tryptophan catabolism comparably but the d isomer is again relatively more active biologically (11). Two possible resolutions to this disparity in results are that D-1MT targets either an undefined cellular isoform of IDO, for example, an alternate spliced or modified isoform, or a different target. Here, we corroborate the latter possibility with the discovery of a novel IDO-related enzyme that is a preferential target for biochemical inhibition by D-1MT.

All materials and methods are included as online Supplementary Material.

IDO2 is a novel tryptophan-catabolizing enzyme that is preferentially inhibited by D-1MT. We discovered IDO2 by Basic Local Alignment Search Tool searches of the human genome using IDO sequences as probes, identifying a new gene on chromosome 8p12 just downstream of the IDO gene INDO. At the time of discovery, genome annotation in this region referred to an anonymous gene termed LOC169355 that was changed later to a misannotated partial gene termed INDOL1 (IDO like-1; Hs.122077). By trial and error, we identified exons permitting assembly of a full-length IDO-related gene termed IDO2. This nomenclature was chosen to distinguish it from INDOL1, which remains misannotated as incomplete gene in the database. By homology searching, we also identified the mouse orthologue Ido2.

Oligonucleotide primers specific to murine and human coding regions were used to amplify cDNAs by reverse transcription-PCR (RT-PCR) from total RNA isolated from various tissues (Supplementary Figs. S1 and S2). In this manner, we obtained full-length cDNAs with complete coding regions including four alternatively spliced variants of each gene. The primary human transcript is derived from 11 exons encompassing a 74 kb region of chromosome 8p12 (Fig. 1A and Supplementary Fig. S3). In three of the five splice isoforms of IDO2 mRNA we identified, introduction of an out-of-frame stop codon causes a premature truncation of IDO2 protein. Transcripts are initiated only 5 to 7 kb downstream of the INDO gene. The mouse gene seems to differ in its lack of the alternate exon 1a found in the human gene; otherwise, exon positions are conserved, indicating gene duplication during evolution of this region of the genome. Human and mouse IDO2 proteins are 420 and 405 amino acids, respectively, and are more conserved (72% identical, 84% similar) than IDO proteins (62% identical, 77% similar). Alignments between IDO and IDO2 sequence reveal highly conserved features that mediate heme and substrate binding (Fig. 1B), although the overall level of sequence conservation is not particularly high (43% identical, 63% similar for human). Significantly, residues determined by IDO mutagenesis and crystallographic analysis to be critically important for catalytic activity are highly conserved in IDO2 (Fig. 1B).

To confirm the expectation that IDO2 catabolizes tryptophan, we expressed it in a doxycycline-regulated T-REX cell system where formation of the enzymatic product N-formyl-kynurenine (Kyn) was monitored. Stable cell lines expressing V5 epitope–tagged or untagged proteins with similar levels of doxycycline-induced expression were used for analysis (Fig. 2A; Supplementary Fig. S4). As expected, both human and murine IDO2 catabolized tryptophan effectively as measured by Kyn production (Fig. 2B). Based on IDO-IDO2 similarity, we compared the ability of known IDO inhibitors to block the activity of IDO2 in T-REX cells. For reasons mentioned above, the IDO inhibitor D-1MT was of particular interest based on uncertainties about its biochemical target (9, 11). Therefore, we evaluated how IDO1 or IDO2 activity was affected by the d or l stereoisomers of 1MT, or by a third inhibitor MTH-trp (9). Consistent with previous observations (11), we found that IDO activity was modestly inhibited by L-1MT but not D-1MT. In contrast, IDO2 activity was inhibited by D-1MT but not L-1MT. This pattern of inhibition was specific to these 1MT isomers insofar as MTH-trp inhibited the activity of both enzymes (Fig. 2C). These results identify IDO2 as a relevant target for biochemical inhibition by D-1MT, may explain its well-documented antitumor effects.

IDO2 expression is more restricted than IDO but includes dendritic cells. By RT-PCR analysis, we found IDO2 is expressed in a subset of tissues expressing IDO. Primers spanning the complete human coding region detected full-length mRNAs only in placenta and brain, whereas primers specific to exon 10 (found to be common to all human IDO2 cDNAs) detected IDO2 mRNAs in human liver, small intestine, spleen, placenta, thymus, lung, brain, kidney, and colon (Fig. 3A and B). Although RT-PCR reactions spanning exons 1 to 8 might not have been sensitive enough to detect low-level transcripts, exon 1a–specific primers gave similar results (data not shown), implying that other transcription start sites may exist in human IDO2. Northern analysis of mouse tissue RNAs confirmed a more narrow range of expression, revealing detectable IDO2 transcripts only in liver and kidney (Fig. 3C). In a query of the NCBI SAGEmap database with a sequence tag to IDO2, the top 4 hits in terms of tag count prevalence were all identified as bone marrow–derived dendritic cell libraries. Because D-1MT inhibits kynurenine production in dendritic cells and block their ability to activate T cells (11), we examined IDO2 expression in an established predendritic mouse cell line (JAWII) that matures to dendritic cells after treatment with IFN-γ. IDO2 mRNA was expressed in unstimulated JAWII cells, and IFN-γ treatment and, to a lesser extent, IL-10 or lipopolysaccharide treatment increased levels modestly (Fig. 3D). Using an IDO2-specific monoclonal antibody, we confirmed expression of IDO2 protein in JAWII cells by Western blotting and indirect immunofluorescence microscopy, the latter of which revealed a generally cytoplasmic pattern of expression like IDO (data not shown). Although we could not detect Kyn production in JAWII cells, we confirmed that full-length cDNAs cloned from these cells encoded a fully active enzyme in T-REX cells (data not shown). Together, these observations defined a pattern of expression for IDO2 that includes dendritic cells.

Common genetic polymorphisms in human IDO2 compromise or abolish enzymatic activity. During characterization of IDO2 cDNAs, we identified two single nucleotide polymorphisms (SNP) that abolished enzymatic activity. One C-T SNP affecting R248 in human IDO2 was structurally analogous to R231 in human IDO, which makes a critical contact with the indole ring of tryptophan (12). The nonsynonymous substitution (R248W) reduced catalytic activity ∼90% in T-REX cells (Supplementary Fig. S5). A second T-A SNP affecting Y359 generated a premature stop codon (Y359X), which completely abolished activity (Supplementary Fig. S5). Strikingly, both SNPs were commonly found in human genomic DNAs in public databases, with the C-T SNP being highly represented in individuals of European descent, the T-A SNP being highly represented individuals of Asian descent, and neither SNP being as prevalent in individuals of African descent (Supplementary Fig. S6). Thus, as many as 50% of individuals of European or Asian descent and 25% of individuals of African descent may lack functional IDO2 alleles. This analysis implicates these SNPs as having a broad effect on IDO2 activity in human populations, which may have a significant bearing on the interpretation of clinical responses to drug-like inhibitors of IDO2 like D-1MT.

IDO2 and IDO each activate LIP, an inhibitory isoform of immune regulatory transcription factor NF-IL6, but IDO2 produces a tryptophan-independent signal. Tryptophan catabolism by IDO triggers GCN2-dependent phosphorylation of the translation initiation factor eIF-2α (13). Activation of this pathway inhibits translation of most messages with the exception of certain messages essential for stress-related functions. Additional contributions of IDO to tolerogenesis are imparted by Kyn and other downstream catabolites (14–16). We evaluated the ability of IDO2 to activate this pathway in T-REX cells. In IDO-expressing cells, Kyn production was constant for 4 days postinduction after which cell growth rate slowed appreciably. This effect related to tryptophan depletion rather than Kyn elevation, because supplementing the culture medium with tryptophan rescued the effect (Supplementary Fig. S7). In IDO2-expressing cells, tryptophan consumption was slower such that cell growth was not affected (Supplementary Fig. S7). Nevertheless, induction of IDO2 caused GCN2-dependent phosphorylation of eIF-2α like IDO (data not shown). To compare downstream effects, we examined how IDO or IDO2 activation affected translation of LIP, an inhibitory isoform of the transcription factor NF-IL6/CEPBβ that is up-regulated by amino acid deprivation by a switch to an alternate translational start site (17). Both enzymes up-regulated LIP strongly, however, restoring tryptophan to culture medium reversed LIP induction only when stimulated by IDO (Fig. 4). Thus, IDO2 produced a distinct signal for LIP activation that was independent of tryptophan availability. This signal required catalytic activity because it was inhibited by D-1MT (Fig. 4). These findings implied that IDO2 has a distinct signaling role in cells compared with IDO.

The findings of this study are significant and timely regarding how tryptophan catabolism suppresses T-cell immunity, how immune escape evolves during cancer progression, and how the d stereoisomer of the widely studied IDO inhibitor 1MT, presently entering phase I clinical trials, acts to elicit antitumor responses in animals. Given the striking therapeutic effects of D-1MT in preclinical models of cancer and other diseases (6), our findings point to IDO2 as an important therapeutic target and genetic modifier for understanding disease susceptibility. The existence of widely dispersed genetic polymorphisms in human populations that ablate catalytic activity argues that knowing the genetic status of IDO2 of individuals enrolled in D-1MT trials may be important for understanding clinical responses. Given the likelihood that IDO2 may contribute to immune tolerance, two implications are that individuals heterozygous or homozygous for catalytically inactive alleles may be (a) less susceptible to developing diseases driven by immune suppression, and (b) less susceptible to manifesting clinical responses to D-1MT or other IDO2 inhibitory compounds. Due to deficiencies in IDO2 activity, such individuals may be relatively less prone to immune escape and malignant progression of oncogenically initiated lesions, but relatively more prone to autoimmune disorders. Given differences in the antitumor responses seen in various preclinical cancer models to L-1MT versus D-1MT (11), it may also be interesting to evaluate the murine IDO2 gene for related polymorphisms.

In LIP, we have defined a novel component of the tryptophan catabolism signaling pathway triggered by IDO or IDO2, using it here to reveal a mechanistic difference in how translational control by these enzymes may modulate immune tolerance. As a downstream reporter, LIP could provide a useful biomarker for genetic and biochemical pathways activated by IDO1 or IDO2 in cells that express NF-IL6 (also known as CEBPβ). In essence, LIP is a dominant inhibitory isoform composed of only the DNA binding region of NF-IL6 (17). By interfering with target genes that control stress signaling, cell growth, and immune modulation, LIP is well positioned to mediate stable effects of IDO or IDO2 on immune tolerance generated by antigen-presenting cells or other cell types. Using LIP, we found that transient activation of IDO2 generates a stable signal that persists independently of tryptophan availability. The potential significance of this mechanism is that it could be used to propagate tolerance from a local to a peripheral immune environment, away from an initial site of tryptophan catabolism (18), for example, to support cancer metastasis. Differences in LIP response argue that the functions of IDO and IDO2 may be distinct, even if outcomes for eliciting immune tolerance are similar.

IDO2 may address key questions about how 1MT manifests its antitumor activity. Previous studies indicated that D-1MT can inhibit tryptophan catabolism in human dendritic cells and that the IDO gene is needed for antitumor activity, implicating IDO in the D-1MT mechanism at some level (11). Our findings do not rule out the possibility that 1MT may target an endogenous IDO protein differing at some level, for example, due to posttranslational modification (11); however, identifying IDO2 addresses a key gap in knowledge concerning the biochemical target of D-1MT. In most models, D-1MT displays much better antitumor activity than L-1MT prompting the choice made for clinical development. One implication is that compounds with dual specificity for IDO and IDO2 may exert more potent antitumor efficacy, and MTH-trp fulfills this expectation (9). Based on genetic knockout studies supporting a role for IDO in the response to D-1MT at some level (11), our findings strongly suggest cross-talk or cooperation between the functions of IDO and IDO2 in immune regulation. Consistent with this idea, IDO activity may be supported by other elements involved in tryptophan catabolism (16, 19). In future work, it will be important to examine IDO-IDO2 cooperation as well as how catabolites of tryptophan catabolism may figure into IDO2 action.

Recently we became aware of another group reporting the identification of this gene (20).

Grant support: NIH grants CA82222, CA100123, and CA10954 and the Lankenau Hospital Foundation (G.C. Prendergast); and grants from the Department of Defense Research Program (BC044350), the Pennsylvania Department of Health, and the Concern and Lance Armstrong Foundations (A.J. Muller).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We apologize to investigators whose work was not cited due to size restrictions for publication.