A paradigm shift of PD-L1 immunotherapy based on a tracking-to-triggering immunoediting effect of 2[ 18 F]FDG

Efforts have been devoted to select eligible candidates for PD-1/PD-L1 immune checkpoint blocker (ICB) immunotherapy based on radiopharmaceuticals. Here, we have observed a tracking-to-triggering (referred to as “T2T”) immunoediting effect of the employed radionuclides. In particular, we found the usefulness of 2-[ 18 F]FDG in cancer ICB therapy. Given that the PD-L1 expression is upregulated after the administration of various radiotracers, the predictive result from PET/SPECT imaging should be treated with caution. Viewed positively, radiotracers are potential immunomodulators to create an immune-favorable microenvironment for tumor immunotherapy. Improving αPD-L1 mAb utilization and signicant tumor growth delay are observed when the personalized therapeutic alliance of radiotracer stimulation and ICB are employed. This new paradigm has the potential to expand the traditional tumor theranostic model and implement precision cancer immunotherapy. results demonstrated that prevention recurrence by T2T-based immunotherapy credited to the activation immunological memory were confocal Each was xed with 100 μL of 4% paraformaldehyde for 10 After that, the cells were washed three times with PBS and incubated with 10% goat serum for 30 min to reduce nonspecic binding. Cells were stained with the rst antibody PD-L1 (primary antibody, Abcam, USA) overnight, rinsed three times with PBS. After cells were stained for 1 h with secondary antibody Alexa Fluor® Plus 488-conjugated IgG and washed three times with PBS, cell nuclei were stained blue with DAPI (Invitrogen Molecular Probes, USA). For histological analysis, tissue specimens were xed with 10% buffered formalin, dehydrated in ethanol, embedded with paran and stained with H&E. Immunohistochemistry on frozen or paran-embedded mouse tissues was performed using antibodies directed against PD-L1, CD4, IFN-γ, CD8. For paran-embedded samples (PD-L1, CD4, IFN-γ, CD8), samples were dewaxed in ethanol, followed by antigen retrieval with 0.01-M sodium citrate with 0.05% Tween. Immunouorescence staining on frozen mouse tissues was performed using antibodies against PD-L1, ki67, caspase3 and DAPI. Immunouorescence images were acquired using the Zeiss LSM880 confocal microscope with ZEN 2010 Histological and Immunohistochemistry images were acquired using Leica DM4 B upright digital research microscopes with Leica Application Suite (LAS All the images were quantitatively analyzed with ImageJ and data were shown by a with


Introduction
The great success of immunotherapy has initiated a new phase in cancer treatment, and an increased understanding of mechanisms leads to the development of various immune checkpoint blockers (ICBs).
ICB antibodies targeting the immune checkpoint programmed death receptor 1 (PD-1) expressed on Tcells and its immune regulatory ligand PD-L1 have revolutionized the oncology 1,2 . Unfortunately, immunotherapy with PD-1/PD-L1 ICBs is less pronounced in high percentage of malignancies. There is a broad consensus that targeting PD-L1 is not equally successful in PD-1/PD-L1 ICB immunotherapy and should have tumor PD-L1 expression as a prerequisite. Thereby, tumor PD-L1 expression is deemed to be a predictive biomarker for pinpointing potential candidates who might bene t most from PD-1/PD-L1 immunotherapy 3,4 .
Early studies used ex vivo immunohistochemical (IHC) staining to evaluate PD-1/PD-L1 status, which was affected by different antibodies, varying protocols, scoring systems and positive/negative staining thresholds [5][6][7]  In this study, we tested 2-[ 18 F]FDG, as an early immunomodulator, to remodel the PD-L1 expression. We demonstrate for the rst time to our knowledge, across multiple tumor cells and mouse models including patient-derived xenografts (PDXs) derived from NSCLC, that 2-[ 18 F]FDG-based PET imaging may result in deceitful and temporary upregulation of PD-L1 expression. This means the role of 2-[ 18 F]FDG for tumor immune microenvironment (TIME) is much more like an immunomodulator than an indicator. Although controversial, with growing evidences 17,18 that upregulation of PD-L1 expression is bene cial to PD-1/PD-L1 immune modulation therapy, we would like to see if imaging radiotracer is a reliable coagent for building an immune-favorable microenvironment for enhancing the e cacy of anti-PD-L1.

Results
Tumor PD-L1 expression is upregulated after radionuclide or radiotracer stimulation in vitro First, we validated the radionuclide-induced PD-L1 upregulation in multiple tumor cell lines. For quick reference, Fig. 1A lists the radionuclides used in this article. 18 F, 99m Tc, 177 Lu, 64 Cu and 131 I were compared on multiple tumor cell lines (melanoma, breast and colorectal cancer cells) in the immuno uorescence assay (Fig. 1B), which revealed that different radionuclides upregulated PD-L1 expression to different degrees. This stimulation was also embodied prominently through the ow cytometric analysis. As shown in Fig. S1, the proportions of PD-L1-positive cells in the CT26, MC38, 4T1 and B16F10 tumor cells were signi cantly increased after co-incubation with radionuclides.
The expression of PD-L1 was elevated to a greater extent by a higher dose of radiotracer, which was further con rmed by ow cytometric analysis in Fig. 1D, clearly indicating that PD-L1 was upregulated in a dose-dependent manner. The expression levels of PD-L1 mRNA and protein in MC38 and CT26 cell lines after stimulation with 2-[ 18 F]FDG were further evaluated by Western blot (WB) (Fig. 1E). As expected, PD-L1 expression was signi cantly increased in response to radionuclides.
Differentially expressed genes (DEGs) and potential mechanisms of radionuclide-induced PD-L1 upregulation Transcriptomic analysis and WB study were performed to explore the potential mechanism of PD-L1 upregulation stimulated by radionuclides. From the volcano plot (Fig. 1F), there were a total of 2002 DEGs which had changed in 2-[ 18 F]FDG-treated MC38 cells compared to the control group, with 1223 upregulated genes and 779 downregulated genes (|log 2 (FC)| > 1.0, P-value < 0.05). For 2-[ 18 F]FDG-treated CT26 tumor cells, the changed number was 2167 (1357 upregulated genes and 810 downregulated genes).
A total of 21725 genes and 21144 genes were identi ed in 2-[ 18 F]FDG treated MC38 cells and CT26 cells, respectively. As shown in Fig. 1G, Fos, Stat3, Nfkbia, Nfkbib, Nfkbie and Cd274 (PD-L1) genes in 2-[ 18 F]FDG treated MC38 cells were signi cantly upregulated compared with the untreated cells. Note that Nfkbia, Nfkbib and Nfkbie genes belong to the NF-kappa-B (NF-κB) inhibitor family, which has been reported to upregulate PD-L1 transcription in tumor cells, such as ovarian cancer, gastric carcinoma and lung cancer [19][20][21] . As veri ed in previous studies 22,23 , the IκBα kinases (IKK) is a key regulator of the NF-κB pathway and TANK-binding kinase 1 (TBK1) is closely related to the phosphorylation of IRF3.
We also showed the similar results that the radionuclide-induced PD-L1 upregulation was positively correlated with phosphorylated NF-κB P65 (p-NF-κB P65) and phosphorylated IRF3 (p-IRF-3) in radionuclide-treated MC38 cells (Fig. 1H). Intriguingly, we found that the PD-L1 upregulation in radionuclide-treated MC38 cells could be blocked by the inhibitors of IKK or TBK1. All these data suggested that the activation and phosphorylation of NF-κB and IRF3 have contributed to promoting PD-L1 expression in radiation-induced MC38 murine colon carcinoma cells.
DEGs were mapped into the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database to further explain the individual function analysis. Herein, transcriptome analysis using the RNA-seq technology was applied to compare DEGs between 2-[ 18 F]FDG and saline-treated MC38 cells. As shown in Fig. 1I, many receptor-interaction signaling pathways and metabolic pathways were signi cantly enhanced, including the cytokine-cytokine receptor interaction, the NOD-like receptor signaling pathway, necroptosis, TNF signaling pathway, IL-17 signaling pathway, and the Jak-STAT signaling pathway.
According to the literature reports, the NF-κB signaling pathway is one of the major NOD-like receptor signaling pathways, that Nod1 and Nod2 stimulation induces NF-κB activation 24 . Besides, TNF (also known as TNF-α) could induce the expression of NF-κB target genes and trigger the activation of NF-κB signaling pathway, which indirectly upregulated PD-L1 expression 25,26 . These results suggest that multiple in ammatory signaling pathways and metabolic process participate in 2-[ 18 F]FDG induced MC38 cells.

Radiotracers cause PD-L1 upregulation in tumor tissue
With the help of the IHC technique, PD-L1 expression in tumor tissue was compared between different groups. We rst investigated the biodistribution pro le of 2-[ 18 F]FDG and 131 I-αPD-L1 in CT26 and MC38 tumor types at different time points using a small-animal PET scanner. The signals in the tumor sites were well delineated from that of other tissues, indicating the high a nity of the radiotracers to tumor lesion ( Fig. 2A). Strikingly, as shown in Fig. 2B,C, PD-L1 levels in the tumor region with radiotracers uptake were more strongly positive compared to control samples.
IFN-γ and active CD4 + /CD8 + T cells in TIME play important roles in mediating antitumor immunity. For this reason, we used 2-[ 18 F]FDG as a stimulus to observe the responses of TIME and determine the optimized immunotherapy time window for the administration of αPD-L1 mAb. From the IHC in Fig. 2D,E, we could see that 2-[ 18 F]FDG gradually upregulated the expression of IFN-γ, and enhanced the in ltration of CD4 + and CD8 + T cells over time. Prior studies have demonstrated that the expression of PD-L1 has been an inclusion criterion for selecting patients of non-small cell lung cancer (NSCLC) for anti-PD-L1 treatment 27,28 . To date, PD-1/PD-L1 ICBs have shown promise in advanced NSCLC without driver oncogene mutations, but wider use is restricted to the low objective response rate 29,30 . In this study, we established NSCLC-PDX models to clarify this T2T immunoediting effect further. After 2-[ 18 F]FDG PET imaging (Fig. 3A,B), the PDXs were divided into high and low 2-[ 18 F]FDG uptake (denoted as PDX H-FDG and PDX L-FDG ) groups to evaluate the in vivo biological behavior of radiotracers and predict the PD-L1 response of NSCLC to the radionuclide. Flow cytometry analysis revealed that the PD-L1 positive population in PDX L-FDG (24.9 ± 3.5%) was lower than that in PDX H-FDG (76.9 ± 4.9%) at 4 h p.i. (Fig. 3C,D).
Representative histograms of the PD-L1 expression after radionuclides stimulation were shown in Fig. 4B.
T2T effect sensitizes the TIME to immunotherapy and enhances the immunological memory To further explore the T2T potential of radiotracers for enhancing immunotherapy, we subsequently investigated the effect of αPD-L1 mAb on MC38 tumor growth delay in cooperation with 2-[ 18 F]FDG. As shown in Fig. 5A and Fig. S2A, tumor models were treated with either 2-[ 18 F]FDG, αPD-L1 mAb, or their combination in speci c treatment sequences. In the combination groups, αPD-L1 mAb was tail vein injected into the tumor-bearing mice at different intervals (simultaneous injection, 4-h and 24-h; hereinafter referred as @ 4 h and @ 24 h) after administration of the radiotracer. Fig. 5B,C and Fig. S2B illustrated the tumor volumes, time-dependent tumor growth curves, weight changes and survival curves for each group. In the control groups of αPD-L1 mAb and saline alone, the tumor sizes developed uncontrollably. Also, single-administration of 2-[ 18 F]FDG did not signi cantly alter MC38 tumor growth. We then compared the therapeutic effect of αPD-L1 mAb which was administered simultaneously, 4 h or 24 h post radiotracer injection. Notably, the 4-h interval turned out to be the most optimal treatment sequence, and administration of 37 MBq 2-[ 18 F]FDG + 400 mg αPD-L1 mAb @ 4 h resulted in the maximum therapeutic e cacy (5/8 of the tumor mice were completely cured), clearly indicating that the T2T antitumor immunotherapy was regulated in a dose and time-dependent manner. As shown in Fig. 5D, we performed 2-[ 18 F]FDG-PET imaging on day 0 and day 90 to provide visualization for evaluating therapeutic effect in the best-performing group. Moreover, the body weights of mice were almost identical for all groups during the therapy period ( Fig. 5C) and no obvious side effects were observed in the fully recovered mice (Fig. S3), indicating that the T2T antitumor therapeutic strategy was well tolerated.
To verify the immunological memory of T2T-based immune checkpoint therapy, the effector memory T (T EM ) cells (CD8 + CD44 + CD62L − and CD4 + CD44 + CD62L − ) in the spleen were detected and analyzed ( As shown in Fig. S5, compared to the saline group and αPD-L1 group, the enhanced change of tumor PD-L1 level following 2-[ 18 F]FDG alone or combined immunotherapy strongly predicted response to radionuclide stimulus. Then this indicator showed a tendency to decrease during the later period. Contrary to the trend of PD-L1, we observed increased level of IFN-γ for radiotracer-induced immunotherapy (Fig. S6A,B). Further results indicated that both the CD4 + Th1 (IFN-γ + CD4 + T cells) and CD8 + cytotoxic T lymphocytes (IFN-γ + CD8 + CTLs) in TIME were enhanced from day 1 to day 7 in the group of 2-[ 18 F]FDG + αPD-L1 mAb @ 4 h, whereas the levels of these indicators were unaltered in the saline group. Meanwhile, tumor samples were harvested for detecting proliferation and apoptosis by immuno uorescence staining of Ki67 and Caspase3. As depicted in Fig. S6C, the dynamic change of PD-L1 expression from day 1 to day 7 in the combination therapy groups were further validated. As expected, at the corresponding time points, the Ki67 indexes were signi cantly higher in saline groups. While the positive rate of Caspase3 expression in the combination group was signi cantly higher than that in saline groups.
Similar to the IHC and immuno uorescence results, tumor PD-L1 expression measured by ow cytometry showed a decreasing trend during the period of day 1 to day 7 in groups containing radiotracer (Fig.  6C,D). ELISA assays were performed to measure the levels of immunostimulatory cytokines in the serum of mice. The combination of 2-[ 18 F]FDG and αPD-L1 mAb increased the production of IFN-γ, TNF-α and IL-6, and maintained for a long period in blood, which might also explain for the unexpected synergistic anticancer e cacy (described in Fig. 6E). Additionally, the ow cytometric results in Fig. S7A,B showed that intratumoral CD4 + Th1 and CD8 + CTLs become exhausted on day 3 and day 7 in the 2-[ 18 F]FDG group. However, the addition of αPD-L1 mAb @ 4 h signi cantly increased numbers of CD4 + Th1 and CD8 + CTLs compared to the other groups. However, as one type of CD4 + T cells, the immunosuppressive CD4 + FOXP3 + regulatory T cell (Treg) in tumors showed a decrease in 2-[ 18 F]FDG + αPD-L1 mAb @ 4 h group. Speci cally, further comparative analysis showed a signi cant increase of CD4 + Th1/Treg and CD8 + CTLs/Treg ratios in the combination group (Fig. 6F). In the 2-[ 18 F]FDG group, we observed slight increase in CD4 + Th1/Treg and CD8 + CTLs/Treg ratios on day 1. Over time, these ratios seem to be on a downward trend.
Other alterations of immune cells are also notable. M2-like macrophages, M1-like macrophages, myeloidderived suppressor cells (MDSCs) and dendritic cells (DCs) were detected through ow cytometry ( Fig. 6G and 7A. The most signi cant differences between the saline and 2-[ 18 F]FDG + αPD-L1 @ 4 h groups were found in the antigen processing and presentation, phagosome, cell adhesion molecules, the NOD-like receptor signaling pathway, cytokine-cytokine receptor interaction and Th17 cell differentiation. Moreover, we went a step further to con rm that the external radiotracers would affect the expression of PD-1, another key component of immune checkpoint blockade (Fig. 7B). Together, these pro les further con rmed that the radionuclide-induced PD-1/PD-L1 immunotherapy could in ame the TIME and activate the immune system.
Coincidentally, consistent with the aforementioned heat map of DEGs (Fig. 1G), subsequent therapeutic trials con rmed that MC38 tumor was more susceptible to 2-  (Fig. S9D-F). Therapeutically, the activation of CD4 + and CD8 + T cells in the tumor and increased CD4 + Th1/Treg and CD8 + CTLs/Treg levels highlighted the potential of the effective coordination to enhance antitumor immunity (Fig. S9G). Previous studies suggested that the less-immunogenic and microsatellite-instable CT26 model did not respond to irradiation with increased PD-L1 expression 31,32 . To some extent, this study describes a new method to overcome this setback via radionuclide-induced immunotherapy.

Discussion
ICB antibodies targeting the PD-1/PD-L1 pathway have revolutionized the oncology. Because patients with positive PD-L1 expression generally have better objective response rates (ORRs), practitioners have very high expectation for selecting eligible candidates for ICB therapy based on SPECT/PET imaging.
However, to best realize such intention, it is important, as a rst step, to nd out whether the PD-L1 expression would be in uenced by a T2T effect of the employed radiotracers (not just 2-[ 18 F]FDG, also other radiotracers in the immunoimaging toolbox). Based on our observations in this study, the dose-and time-dependent T2T effect directly in uences the accuracy and rationality of radioimmunoimagingguided PD-L1 identi cation and therefore should not be overlooked. Fig. 6A summarizes the relationship between the radiotracer and multifaceted immunologic responses (including dynamic changes of CD4 + Th1 and CD8 + CTLs, etc.), which can help us reconsider cautiously the role of radiotracers including 2-[ 18 F]FDG in tumor imaging and immunotherapy. In brief, the dynamic interferences and misleading recommendations from radiotracers need to be rede ned. Unfortunately, to date, no data are available from clinical or preclinical trials to discuss this intractable question. In general, from our point of view, a multidimensional assessment based on biopsy specimens before radiotracer injection (initial TIME status) and radioimmunoimaging feedback (T2T-based immune response) should be emphatically considered prior to initiation of immunotherapy to guide clinicians in adjusting regimens of ICB therapy as appropriate.
Therapeutically, T2T responses in multiple tumor types observed in our experiments have spurred us to investigate the radionuclide-induced immunotherapeutical effects and made us fully aware that immune cells need certain amount of time to be activated and PD-L1 expression needs to be stimulated in tumor cells rstly 31 . Subsequent studies provide evidence on the necessity and timeliness of the participation of αPD-L1 mAb. Taking MC38 tumor model receiving 2-[ 18 F]FDG plus αPD-L1 for example, of all the groups, the overall therapeutic outcomes of the 6-h interval were most impressive. That is to say, an immunefavorable TME is activated by targeted radionuclide and therapeutic bene t is enhanced by αPD-L1 mAb subsequently. Meanwhile, it has been shown that upregulation of PD-L1 in the tumor can result in an increased demand for antibody, which seems to be a reasonable approach.
The radiotherapeutic isotopes ( 177 Lu, 90 Y, 111 In, 131 I, etc.), which generally have blast of radiation, are actively pursued in targeted radionuclide therapy to destroy the tumor cells. Whilst the therapeutic potential of positron-emitting agents is still underexplored. Ideally, based on the consideration of radionuclide characteristics, 18 F-based nuclear therapy will lead to a reduction in cost and radiation.

Theoretically, 2-[ 18 F]FDG emits positrons and should kill cancer cells in the same manner as electrons.
In previous studies, high-dose 2-[ 18 F]FDG (111-222 MBq per mouse) was used for the radiomolecular therapy of cancer in mouse models and effective therapeutic response (incomplete cure) was observed [33][34][35] . Meanwhile, radiotoxicity of 2-[ 18 F]FDG (222 MBq per mouse) was not found 33 . It is well known that radiation can enhance antitumor immunity by damaging DNA and inducing apoptosis/necrosis of the tumor cells. On this basis, with the synergy of αPD-L1 mAb and properly scheduled time window, it is possible to further drive 18 F dosage lower, which could be favorable for clinical application. However, to date, the potential of immune-mediated response enhanced by 18 F remains largely unexplored. Fortunately, on basis of our ndings, we reason that the smart T2T immunoediting effect induced by 2-[ 18 F]FDG has the potential to boost immune activation in a manner that is synergistic with PD-1/PD-L1 ICB immunotherapy. In MC38 tumor model, the optimal dose of 2-[ 18 F]FDG is 37 MBq per mouse, which is far below the physiological human maximum tolerated dose.
More interestingly, the relationship between radionuclides, PD-L1 upregulation and timeline could be drawn directly into a tumor-nuclide-time code (similar to Fig. 1C) and used to differentiate tumor types that are sensitive to T2T immunotherapy, which aligns well with precision cancer immunotherapy. Besides the radionuclide, dose, timing and sequencing setting discussed above, the target site of the probe is an important consideration for the design of the therapy project. Of note is that targeting PD-L1 on tumor cells is not the only choice to pave the way for anticancer-T2T-based ICB immunotherapy. The other ICB targets, such as CTLA-4, STING, VISTA etc., may also be used in the radiotracers combined immunotherapy schema. Various radiolabeled speci c targeting agents (FDG, RGD, broblast-activating protein (FAP), prostate-speci c membrane antigen (PSMA), octreotate (TATE), folic acid (FA), bombesin (BBN), etc.), nanoprobes, albumin binders and biomacromolecule can be used to enrich the T2T toolbox. And the application scenarios could even be extended to allow the multi-targets or multi-isotopes strategy if necessary. Compared with radiotherapy nuclides, such as 177 Lu and 131 I, diagnostic radiotracers as immunomodulators have shorter half-life, quicker in vivo clearance rate and higher biological safety (without βradiation), which could be widely used in the patients. All these smart T2T strategies will unquestionably contribute to the vision of precision cancer immunotherapy. Compared with the traditional local radiotherapy only for the irradiated area (within the irradiated eld), the addition of appropriate radiotracer and inhibitor seems to be a more attractive option with "cross-re" effect, which can augment the depth and duration of responses, boost the systemic antitumor immune response, especially for multiple lesions, distant metastases and tiny foci. Such a breakthrough will clinically put the nuclear medical diagnosis and ICBs together in a series, and thus truly build an image-guided platform for tumor theranostics.
Studies have shown that one of the main challenges for ICB immunotherapy lies in "cold" tumor with limited T-cell responses 36,37 . Fortunately, as summarized in Fig. 6F, we found that radionuclide-based ICB immunotherapy synergistically enhanced antitumor immunity by promoting critical parameters of CD4 + Th1/Treg and CD8 + CTLs/Treg ratios in TIME. As we know, CD4 and CD8-positive T-cells are important hallmarks of transition from "cold" TIME to "hot", which is critical to realize ideal ICB outcomes. And Tregs are de ned as immunosuppressors for maintaining immunological tolerance, which is associated with a poor prognosis. Optimistically, this smart T2T effect seems to be an attractive option to eliminate treatment resistance of patients with low or no PD-L1 expression in pretreatment tumors and improve clinical outcomes.
In summary, we have demonstrated that imaging radionuclides induce signi cant PD-L1 upregulation in tumor cells. From a predictive perspective, it is necessary to develop reliable imaging strategies for precisely monitoring PD-L1 expression. Besides, further studies are also underway to take corrective actions for improving the accuracy of immuno-PET/SPECT in guiding PD-L1 expression in patients. MC38 tumor tissue samples of different groups were collected for high-throughput sequencing. Reference genome and gene model annotation les were downloaded from the genome website directly. Paired-end clean reads were aligned to the reference genome using HISAT2 v2.1.0 (hierarchical indexing for spliced alignment of transcripts), which is a highly e cient system for aligning reads from RNA sequencing experiments.
Immuno uorescent, Histology, Immunohistochemistry and Microscopy For immuno uorescent analysis of singe-cell, CT26, MC38, 4T1 and B16F10 tumor cells were seeded in confocal dishes. Each sample was xed with 100 μL of 4% paraformaldehyde for 10 minutes. After that, the cells were washed three times with PBS and incubated with 10% goat serum for 30 min to reduce nonspeci c binding. Cells were stained with the rst antibody PD-L1 (primary antibody, Abcam, USA) overnight, rinsed three times with PBS. After cells were stained for 1 h with secondary antibody Alexa Fluor® Plus 488-conjugated IgG and washed three times with PBS, cell nuclei were stained blue with DAPI (Invitrogen Molecular Probes, USA). For histological analysis, tissue specimens were xed with 10% buffered formalin, dehydrated in ethanol, embedded with para n and stained with H&E.
Immunohistochemistry on frozen or para n-embedded mouse tissues was performed using antibodies directed against PD-L1, CD4, IFN-γ, CD8. For para n-embedded samples (PD-L1, CD4, IFN-γ, CD8), samples were dewaxed in ethanol, followed by antigen retrieval with 0.01-M sodium citrate with 0.05% Tween. Immuno uorescence staining on frozen mouse tissues was performed using antibodies against PD-L1, ki67, caspase3 and DAPI. Immuno uorescence images were acquired using the Zeiss LSM880 confocal microscope with ZEN 2010 software. Histological and Immunohistochemistry images were acquired using the Leica DM4 B upright digital research microscopes (Leica, Germany) with Leica Application Suite X (LAS X). All the images were quantitatively analyzed with ImageJ 7.0 software and data were shown by a histogram with Graphpad prism 7.0 software.

Western Blot Analysis
Western Blot was performed as described previously 38 , with minor modi cations. Brie y, cell lysates were made in ice-cold RIPA buffer containing complete protease inhibitor cocktail and phosphatase inhibitor cocktail (Sigma). Total protein was quanti ed using the BCA Assay according to the manufacturer's instructions (ThermoFisher). 10% Bis-Tris polyacrylamide gels were equiloaded with 20 μg of protein, electrophoresed at 120 V, and electro-transferred to PVDF membranes. After blocking with 5% BSA, membranes were probed with primary antibodies to PD-L1, NF-κB p65, Phospho-NF-κB p65, IRF3, Phospho-IRF-3 and β-actin. Blots were developed by ECL (Thermo Fisher Scienti c).

Flow Cytometry of Cell Lines and Organs
Tumor cells were seeded in 6-well plates overnight and treated with different radionuclides ( 18 F, 99m Tc, 177 Lu, 64 Cu, 131 I). Saline-treated cells were used as control group. After incubation with different time (0. Single-cell suspensions of mouse tumors were prepared for ow cytometry as described previously 39 . In brief, samples were harvested and cut into small fragments (1-2 mm 3 ), and placed in DMEM containing Collagenase IV (1 mg/mL; Gibco, USA), trypsin inhibitor (1 mg/mL; EMD Millipore), and DNase I (2 U/mL; Promega). The fragments were then incubated at 37 °C for 60 min with gentle shaking every 10 min.
Specimens were passed through a 70 μm mesh and centrifuged at 350 g for 5 min. Red blood cells were eliminated from the samples with a hypo-osmotic red blood cell lysis buffer (Solarbio). Each sample was xed with 100 μL of 4% paraformaldehyde for 10 minutes. After that, the cells were collected and washed three times with PBS and incubated with 10% goat serum to reduce nonspeci c binding. For detecting PD-L1 expression and T cell alteration, preprocessed cells were stained with the rst antibodies (anti-PD- Cytokine Analysis Serum samples were isolated from mice after various treatments and diluted for analysis. The proin ammatory cytokines including TNF-α, IFN-γ and IL-6 were determined by using enzyme-linked immune sorbent assay (ELISA) kits according to vendors' protocols (Dakewe biotech). The data were calculated and shown in a bar chart with Graphpad prism 7.0 software.

In Vivo Anticancer E cacy
As the tumor volume reached about 50 mm 3 , the CT26 or MC38 tumor-bearing mice were randomly divided into different groups (n = 8 per group) and treated with different schemes. An additional therapeutic course was scheduled on day 4. After initiation of radiotracer-related therapy, the feeding surroundings were shielded with lead bricks to protect them from any contact with extrinsic radiation. The tumor volume and body weight were monitored at the given time points. Mice were euthanasia if the tumor volume exceeded 1500 mm 3 . The percent survival of mice in each group was measured until all the mice had been sacri ced.

Small Animal PET Imaging
Small animal PET imaging studies were performed at the given time points under the approved guidelines. The injected doses were identical to that in therapeutic trials. During the scan procedure, anaesthesia was induced with isoflurane/air mixture to maintain spontaneous breathing of mice. Wholebody microPET imaging was carried out in tumor-bearing mice injected with 2-[ 18 F]FDG. Images were acquired directly following the acquisition of the CT. PET images were reconstructed using 2D/3D ordered-subset expectation-maximization (2D/3D OSEM) algorithm and with a Maximum a Posteriori Method (MAP). Injected dose and body weight were input before imaging to accomplish normalized and decay corrected radioactivity concentration. For quantitative comparisons, the tissue uptake was acquired by selecting the ROIs on images.

Statistical Analysis
Statistical analyses were performed using an unpaired two-tailed Student's t-test with GraphPad Prism 7.0 software (GraphPad Software Inc.). Survival curve data were analyzed with the Kaplan-Meier method followed by the log-rank test with GraphPad Prism 7.0 software. Data are presented as the mean ± standard deviation. Statistical signi cance is de ned at the *p ≤ 0.05 level.

Reporting summary
Further information on experimental design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The authors declare that the data supporting the ndings of this study are available within the article and its Supplementary Information Files or from the corresponding author on reasonable request. Representative histograms were used to present the upregulation of PD-L1 after radionuclide stimulation.

Declarations
Data are expressed as mean ± SD (n = 3). Each experimental group was compared to the control (*p ≤