The radioisotope ¹¹C was produced by a ¹⁴N (p, α) ¹¹C nuclear reaction using a CYPRIS HM-18 cyclotron (Sumitomo Heavy Industry, Tokyo, Japan). Introducing ¹¹C into the indole ring of 1-Methyl-l-tryptophan (l-1MTrp) produced ¹¹C-l-1MTrp as injectable solutions in 40±2 min according to our previously reported method.16 These radiotracers had a molar activity of 47–130 GBq/μmol and >98% radiochemical purity (n=80) and were provided for all experiments.

Human tumor xenograft model: to facilitate comparison of the IDO1 levels and ¹¹C-l-1MTrp uptake, two human cell lines, the NCI-H69 small cell lung cancer and MDA-MB231 breast cancer, were selected for PET/CT imaging and biodistribution studies. Tumor xenograft models were established via subcutaneous injection of NCI-H69 (2.5×10⁶) and MDA-MB231 (2.5×10⁶) cells into the left and right flanks, respectively, of immunodeficient BALB/c nude mice (Japan SLC, Shizuoka, Japan) in a total volume of 0.1 mL serum-free medium containing 50% Matrigel (BD Biosciences, San Jose, California, USA). The cells were analyzed by quantitative real-time reverse transcriptase-polymerase chain reaction (qRT-PCR) before injection to confirm their IDO1 expression patterns. When the tumors grew to a maximum tumor diameter of 9–12 mm, the mice were selected randomly for further studies.

Whole-body dynamic PET scans were performed after intravenous injection of ¹¹C-l-1MTrp (16.2–19.5 MBq/0.1 mL, corresponding to 0.73–1.45 nmol of l-1MTrp) using a small-animal Siemens Inveon PET scanner (Siemens, Knoxville, Tennessee, USA) to acquire 159 transaxial slices with 0.796 mm (center-to-center) spacing, a 10 cm transaxial field of view (FOV), and a 12.7 cm axial FOV. Emission scans of the tumor models were acquired in three-dimensional list mode with an energy window of 350–650 keV under isoflurane anesthesia from 0 to 75 min after the injection.

After ¹¹C-l-1MTrp (7.1–7.5 MBq/0.1 mL, corresponding to 0.44–0.88 nmol of l-1MTrp) intravenous injection, human xenograft tumor-bearing mice were sacrificed by cervical dislocation at 5, 15, 30, 60, and 90 min, while combinatorial immunotherapy-treated syngeneic mice were sacrificed by cervical dislocation at 60 min. Tumors and major tissues (blood, tumor, MLN, heart, lung, thymus, liver, pancreas, spleen, kidney, adrenal, intestine, muscle, testis, epididymis, and brain) were promptly excised, harvested, and weighed. Radioactivity was counted using a γ-counter, and results are expressed as %ID/g tissue. All radioactivity measurements were corrected for decay.

l-1MTrp is a well-known IDO1 inhibitor showing specific binding to the catalytic pocket of IDO1, and has been indicated to inhibit tumor growth in diverse tumor types in vivo (figure 1A).20 Radiotracer ¹¹C-l-1MTrp was synthesized with satisfactory radio-characteristics (figure 1B and C).16 To study the potential of ¹¹C-l-1MTrp to specifically visualize IDO1 in human tumors, we performed ¹¹C-l-1MTrp PET imaging in mice inoculated with human tumors with different IDO1 expression levels. Immunohistological staining for IDO1 in tumor sections indicated that the NCI-H69 small cell lung cancer and MDA-MB231 breast cancer markedly differed in their IDO1 expression levels, with higher IDO1 expression observed in the NCI-H69 than in the MDA-MB231 (figure 1D). This trend was also observed at the mRNA level by qRT-PCR (figure 1E).

PET/CT imaging showed strong uptake of ¹¹C-l-1MTrp in the NCI-H69 xenografts but negligible tracer uptake in the MDA-MB231 xenografts (figure 1F–I), in good accordance with their IDO1 expression patterns. We also quantified the radioactivity levels in the tumor site based on the dynamic PET and corresponding CT images at 0–75 min after injection. The tracer uptake in NCI-H69 showed as 5.36%±0.51 % ID/g weight (the percentage value is quantified for t=75 min after injection, similarly hereinafter) when the radioactivity in tumor is steady (figure 1G), significantly higher than that in MDA-MB231 (1.51%±0.23% ID/g weight) (figure 1H).

To verify the specific binding of ¹¹C-l-1MTrp to IDO1, we then treated the same mouse model using two well-known IDO1 inhibitors, l-1MTrp (50 mg/kg) and INCB024360 (10 mg/kg),8 and acquired PET images after blockade by coinjection with the radiotracer. As expected, the uptake of ¹¹C-l-1MTrp was significantly decreased in the NCI-H69 xenografts (3.17%±0.54% ID/g weight) of INCB024360-pretreated mice (figure 1Fiii,G). Surprisingly, l-1MTrp pretreatment had little effect on the uptake of the tracer in NCI-H69 xenografts (4.77%±0.09% ID/g weight, figure 1Fii,G). As l-1MTrp is an analog of the essential amino acid Trp,3 this suggests that l-1MTrp administration did not interfere with ¹¹C-l-1MTrp PET while visualizing IDO1 in vivo.

An ex vivo biodistribution study in this xenograft model validated the PET results (figure 1I and online supplemental table 1). Peak uptake in NCI-H69 tumors was achieved at 60 min, and the tumor-to-blood ratio was estimated to be 2.5–3 (online supplemental figure 1A,B). These results demonstrated that ¹¹C-l-1MTrp could specifically bind to human tumor-derived IDO1 and that its accumulation mimicked the trends in IDO1 expression observed in vivo.

An IDO1-specific radiotracer that noninvasively and quantitatively tracks dynamic IDO1 expression during IDO1-blockade therapy would be expected to identify any divergence in individual responses to IDO1-targeted treatments. To test this hypothesis, we initially used ¹¹C-l-1MTrp to evaluate therapeutic efficacy in B16F10 tumor-bearing immunocompetent mice, a commonly used preclinical tumor model with successful immunotherapies, receiving one of two clinically relevant IDO1 blockade-containing combinatorial therapy regimens, l-1MTrp (5 g/kg) plus cyclophosphamide (CPA, 150 mg/kg, an immunomodulatory chemical toxin21 or l-1MTrp plus paclitaxel (PTX, 13.3 mg/kg) (figure 2A). The results showed that at the end of therapy, the combination of CPA (intravenously, 150 mg/kg) with l-1MTrp (orally, 5 mg/kg feed) achieved >90% tumor growth inhibition in all mice. In contrast, PTX (intravenously, 13.3 mg/kg) plus l-1MTrp (orally, 5 mg/kg feed) showed only moderate antitumor efficacy (~50%) compared with vehicle or their respective monotherapies (figure 2B–E and online supplemental figure 2A–D), indicating the l-1MTrp+CPA combinatorial strategy induced a more effective antitumor response than that of l-1MTrp+PTX in the difficult-to-treat melanoma. Hematoxylin and eosin stain (H&E) staining of tumor sections revealed a large zone of cell necrosis in tissue from the l-1MTrp+CPA-treated mice, but this phenomenon was rarely observed in tumor samples from the l-1MTrp+PTX and vehicle-treated mice (figure 2F). We then stained tumor sections for the IDO1 protein and found that the highest IDO1 protein expression was detected in the vehicle group, while the lowest IDO1 protein level was in the l-1MTrp+CPA group (figure 2G and online supplemental figure 3A), indicating an obvious decrease in tumor IDO1 expression after effective IDO1 blockade-containing combinatorial therapy.

To elucidate whether the antitumor effect is caused by the re-establishment of anticancer immunity, we then analyzed the mRNA expression levels of several immune-related cytokines and proteins in tumors. IFN-γ expression was detected to be approximately fourfold higher in l-1MTrp+CPA-treated mice than in vehicle-treated and l-1MTrp+PTX-treated mice (figure 2H), indicating the activation of cytotoxic T cell-mediated and natural killer cell-mediated antitumor immunity.22 The mRNA expression of the immune mediator stat1 (figure 2I) and effector molecule granzyme B (figure 2J), both of which are known to be responsive to IFN-γ treatment,23 24 was increased in the l-1MTrp+CPA-treated mice, leading to B16F10 tumor cell death. For mice treated with l-1MTrp+PTX, the mRNA expression of these genes was comparable to that in vehicle-treated mice. Collectively, our results are consistent with the idea that the l-1MTrp+CPA regimen is a robust therapeutic strategy for cancer immunotherapy, in which l-1MTrp blocks the tumor-derived IDO1 protein. CPA can augment the immune response by regulating cytotoxic T cells and natural killer cells,20 generating a synergistic effect between CPA and l-1MTrp, which contributes to reversing the immunosuppressive cancer-immune state and produces superior antitumor outcomes (online supplemental figure 3B).25

To examine whether ¹¹C-l-1MTrp PET imaging can reflect the different treatment outcomes among the three treatment strategies (vehicle, l-1MTrp+PTX, and l-1MTrp+CPA), we performed whole-body dynamic PET imaging in B16F10-bearing mice treated with the three strategies. Figure 3A shows representative coregistered ¹¹C-l-1MTrp PET/CT images, and very low background radioactivity was found in the brain, heart, lungs, liver, spleen, and intestine (online supplemental figure 4A,B); the radioactivity was rapidly eliminated via the bladder (online supplemental figure 4C), indicating a favorable pharmacokinetic profile that would be practical for daily practice. We detected no significant difference in ¹¹C-l-1MTrp uptake by tumors among the three groups (figure 3Afigure 3Aand online supplemental figure 4D), although the l-1MTrp+CPA regimen exhibited a better antitumor efficacy than the other regimens, suggesting that direct imaging of IDO1 in tumors via ¹¹C-l-1MTrp is not a suitable approach for predicting therapeutic outcomes. Surprisingly, in the mice treated with l-1MTrp+CPA, which showed the best antitumor efficacy (figure 2D), we observed extremely high tracer uptake in the MLNs (figure 3A,B and online supplemental video 1), an organ that has never been evaluated by PET imaging in cancer immunotherapy (figure 3C). In stark contrast, in vehicle-treated, single agent-treated and l-1MTrp+PTX-treated mice, only low background signals were found in the MLNs (figure 3A,B and online supplemental figure 2E,F). We conducted further ex vivo biodistribution assays to anatomically confirm the whole-body PET results. The radioactivity in the main organs was measured at 60 min postinjection of ¹¹C-l-1MTrp and was consistent with the PET images (online supplemental table 2); the biodistribution results unambiguously demonstrated the highest uptake of ¹¹C-l-1MTrp in the MLNs of the mice treated with l-1MTrp+CPA (23.93%±4.03% ID/g tissue), followed by the MLNs in the vehicle-treated mice (7.52%±3.86% ID/g tissue), and the lowest tracer uptake in the l-1MTrp+PTX-treated mice (4.28%±0.42% ID/g tissue) (figure 3D, online supplemental figure 2G and online supplemental table 2).

IDO1 plays crucial roles in balancing intestinal immune tolerance and gut microbiota maintenance, and IDO1 expression in the MLNs is inducible and dynamically changed to cope with challenges with foreign antigens derived from gut nutrients or the microbiota.26 In this context, we envisioned that connections exist between the accumulation of ¹¹C-l-1MTrp in the MLNs and active responses to immunotherapies (online supplemental figure 4E), as a counter-regulatory mechanism to maintain the necessary immune privilege in the intestines that evolved for the perpetuation of the species in the ‘flaming’ body, and speculated that IDO1 expression in the MLNs could be an unprecedented off-tumor biomarker candidate for treatment outcomes in cancer immunotherapy. To test this assumption, we first performed competitive ¹¹C-l-1MTrp PET imaging of B16F10-bearing mice treated with the IDO1 inhibitor INCB024360 as a blocking agent to demonstrate that ¹¹C-l-1MTrp specifically binds to IDO1 in the MLNs. A significant decline in ¹¹C-l-1MTrp uptake in the MLNs (5.12%±0.62% ID/g tissue) was observed in the mice treated with the inhibitor (figure 3E,F, online supplemental figure 4F, online supplemental table 2 and online supplemental video 2), supporting the conclusion that IDO1 is an unequivocal target of ¹¹C-l-1MTrp in the MLNs.

Next, we sought to determine the spatial localization of ¹¹C-l-1MTrp and IDO1 in the MLNs. We performed ex vivo autoradiography (ARG) and histopathology with MLN tissues harvested from treated mice described above. ARG of MLN sections indicated strong binding of ¹¹C-l-1MTrp in the medulla of the MLNs of l-1MTrp+CPA-treated mice (figure 4A,E), which was consistent with the expression pattern of the IDO1 protein (figure 4B,F). In contrast, the IDO1 protein expression in the same zones of the MLNs was lower in samples from vehicle-treated and l-1MTrp+PTX-treated mice (figure 4B), resulting in poor tracer binding in those tissue samples (figure 4A,E). H&E staining of MLN sections did not reveal any noticeable tissue morphology variation (figure 4C), demonstrating that the treatment regimens had good biosafety. Although CD11b⁺ myeloid cells have been reported to inducibly express IDO1 in the tumor microenvironment,27 interestingly, double staining for IDO1 and CD11b revealed that CD11b⁺ cells in the MLNs of l-1MTrp+CPA-treated mice generally did not express IDO1 (figure 4D), indicating that IDO1 induction primarily occurred in CD11b-negative cells. Conversely, IDO1 and CD11b showed perfect colocalization in MLN sections from vehicle-treated and l-1MTrp+PTX-treated mice (figure 4D). The CD11b expression level exhibited undistinguishable differences in the MLNs among the three cohorts (figure 4G), further demonstrating a CD11b⁺ positive cell independent upregulation of IDO1 in MLNs. Overall, IDO1 expression in the MLNs is inducible in mice with IDO1 blockade-containing combinatorial immunotherapy and can be visualized by ¹¹C-l-1MTrp PET.

Next, to unveil the inter-relationship between the dynamic IDO1 expression in the MLNs and the cancer-immune set point in individuals and to establish a PET imaging-guided precision medicine paradigm for cancer immunotherapy, we performed a longitudinal ¹¹C-l-1MTrp PET imaging study throughout the entire treatment process in mice treated with l-1MTrp+CPA (figure 5A). Mice were subcutaneously injected with B16F10 tumor cells on day 0 and treated with l-1MTrp+CPA on days 7, 10, 13, and 16. Tumor growth was successfully inhibited until day 30 postinoculation but then increased from day 30 to day 40 (figure 5B). We employed the specific growth rate (SGR, the percentage change in tumor size per day, %/d) to define the real-time antitumor efficacy of the combinatorial immunotherapy. The SGR curves of the tumors revealed a three-phase process: phase I showed a gradual decline in the SGR that was attributed to the initiation of therapy and lasted until the end of the therapy. Although the therapeutic intervention was terminated on day 16, the therapeutic effect was sustained throughout phase II, which was characterized by inhibited tumor growth with a constant small SGR; in phase III, the therapeutic effect was gradually exhausted, resulting in tumor relapse, as demonstrated by the continuously increasing SGR (figure 5C).

¹¹C-l-1MTrp PET imaging was carried out at multiple timepoints to closely monitor the therapeutic immune response (figure 5D). Before treatment, we determined the baseline uptake of ¹¹C-l-1MTrp in the MLNs on days 0 (without tumor) and 7 (with a subcutaneous B16F10 tumor) and found that tracer uptake in the MLNs did not change depending on whether a tumor was engrafted. PET imaging showed that the tracer uptake in the MLNs on day 10 was similar to that on day 7, whereas the uptake increased drastically on day 13 and peaked on approximately day 16 (figure 5D,E); the percentage uptake on day 16 was double that at baseline (from ⁓5% to ⁓10% ID/g weight, figure 5E). The plateau in tracer uptake was maintained for approximately 1 week until day 23, even though the last drug injection was performed on day 16. Afterward, the tracer uptake in the MLNs showed a steep decline from day 23 to day 30 and finally demonstrated a relatively consistent decline to the baseline level from day 30 to day 40 (figure 5D,E).

We next assessed whether IDO1 expression in the MLNs is a predictor of the cancer-immune set point in the immunoediting process. As the tumor SGR is a real-time indicator of immunotherapeutic outcomes,28 we evaluated the relation between the percentage uptake of ¹¹C-l-1MTrp in the MLNs and the tumor SGR (figure 5F). From the SGR curves (figure 5C), we divided the treatment process into four stages: tumor growth (I), therapeutic effect activation (II), therapeutic effect duration (III), and therapeutic effect mitigation (IV) (figure 5F). As expected, this tumor profile perfectly coincided with the percentage uptake of ¹¹C-l-1MTrp in the MLNs: the ¹¹C-l-1MTrp uptake increased when the therapeutic response was activated but decreased when the therapeutic effect was lost (figure 5F). Correlation analysis identified a significantly inverse association between the tumor SGR and percentage uptake in the MLNs (Pearson r=−0.7375, p<0.0001) (figure 5G). The unique ¹¹C-l-1MTrp PET imaging patterns in MLNs of B16F10 mice (figure 5D) were further corroborated by ex vivo biodistribution studies (online supplemental figure 5 and online supplemental table 3), which show significantly higher uptake of ¹¹C-l-1MTrp in MLNs at day 16 (29.34%±0.80% ID/g tissue) and day 23 (23.93%±4.03% ID/g tissue) postinoculation. These results indicated that ¹¹C-l-1MTrp PET/CT could visually, quantitatively and spatiotemporally track the dynamic IDO1 expression in MLNs associated with immunotherapeutic efficacy and reveal the cancer-immune set point in individuals treated with IDO1 blockade-containing combinatorial immunotherapy.

To generalize ¹¹C-l-1MTrp PET imaging for monitoring the cancer-immune status and antitumor response in mice receiving different immunotherapies, we evaluated another cohort of immunocompetent melanoma-bearing mice that received a triple-treatment regimen of anti-programmed cell death 1 (PD-1, clone RMP1-14, 10 mg/kg) + anti-cytotoxic T-lymphocyte-associated protein 4 (CTLA4, clone 9H10, 5 mg/kg) + CPA (150 mg/kg) (figure 6A). The poor responders and good responders were designated as group a (>0.5 cm³ tumor volume; n=14) and group b (<0.5 cm³ tumor volume; n=6), respectively, with the total cohort showing an overall positive response rate of 30% (6 of 20 mice) according to the tumor size measured on day 32. Group a had a significantly larger average tumor volume than group b on day 32 (2.08±0.35 cm³ and 0.40±0.14 cm³, respectively) (figure 6B and online supplemental figure 6A). The corresponding SGRs on day 32 were estimated to be 17.54%±1.75% /d in the poor responder group (n=14) and 2.99%±1.63% /d in the good responder group (n=6) (online supplemental figure 6B). As expected, ¹¹C-l-1MTrp PET imaging on days 13 and 25 after tumor engraftment showed that group a exhibited low ¹¹C-l-1MTrp uptake in the MLNs (figure 6C,E), whereas all mice in group b exhibited significantly higher tracer uptake in the MLNs (figure 6D,E), consistent with the generation of an effective antitumor response in the group b mice. In contrast, the tumor uptake exhibited little difference between the two groups (figure 6C,D). Furthermore, we evaluated the correlation between the tumor SGR and percentage uptake in the MLNs for B16F10-bearing mice in this treatment cohort. A strong negative correlation was observed between MLN radioactivity and the tumor SGR (figure 6F, Pearson r=−0.8270, p<0.0001). Collectively, these results suggest that ¹¹C-l-1MTrp PET imaging can serve as a robust tool to gauge the cancer-immune set point and predict the antitumor immune efficacy of different types of immunotherapeutic blockade in addition to combination immunotherapy targeting multiple factors by monitoring IDO1 expression in the MLNs.