In Vivo Imaging With Two-Photon Microscope For Assessing Tumor Selective Binding Of Anti-CD137 Switch Antibody

STA551, a novel anti-CD137 switch antibody, binds to CD137 in an extracellular ATP (exATP) concentration dependent manner. Although STA551 was assumed to show higher target binding in tumor than normal tissues, quantitative detection of the target binding of switch antibody in vivo is technically challenging. In this study, we investigated the target binding of STA551 in vivo using intravital imaging with two-photon microscopy. Tumor-bearing human CD137 knock-in mice were intravenously administered 1 mg/kg of uorescent-labeled antibodies at day 0 and 3. Flow cytometry analysis of antibody-binding cells and intravital imaging using two-photon microscopy was conducted at day4. Higher CD137 expression in tumor than spleen was detected by ow cytometry analysis, and T cells and NK cells were major CD137 expressing cells. In the intravital imaging experiment, conventional and switch anti-CD137 antibody showed binding in tumor. However, in spleen, the uorescence of switch antibody was much weaker than conventional anti-CD137 antibody and comparable with isotype control. In conclusion, we could assess switch antibody biodistribution in vivo through intravital imaging with two-photon microscopy. These results suggested that the tumor selective binding of STA551 leads to a wide therapeutic window and potent antitumor ecacy without systemic immune activation.


Introduction
CD137 is co-stimulatory receptor, and the stimulation of CD137 promotes T cell survival, proliferation and effecter function 1,2 . Several anti-CD137 agonist antibodies are being developed for the treatment of cancer 3 . In clinical stages, two monoclonal antibodies, Urelumab (BMS-663513) and Utomilumab (PF-05082566), have been administered to patients with tumors 3 . However, Urelumab caused serious hepatotoxicity in the phase I and II study 4 . Utomilumab showed lower toxicity but had less anti-tumor e cacy 5 . To overcome the issues facing conventional anti-CD137 antibodies, we generated an anti-CD137 agonist switch antibody, STA551 6 . STA551 only binds to CD137 in the presence of ATP, but the binding is not detectable in the absence of ATP. A previous study reported that anti-CD137 switch antibody showed anti-tumor e cacy in various tumor models without systemic immune activation.
Conventional anti-CD137 antibody induced splenomegaly, lymphadenopathy, and activation of T cells in normal tissues. On the other hand, anti-CD137 switch antibody did not show these responses in normal tissues. The data suggested that STA551 is a novel antibody which shows CD137 agonistic activity in tumor selectively.
Antitumor e cacy data in various mouse models and ex vivo analysis data suggested the selective binding of STA551 to CD137 in tumor, whereas quantitatively detecting conventional and switch antibody biodistribution in tumor and normal tissues under physiological condition has been a challenge 6 . ATP exists inside and outside the cell, and intracellular ATP concentration is much higher than extracellular ATP (exATP) 7 . Also, exATP levels are different between tissues. ExATP in tumor interstitial uid is reported to be approximately 100 µM, whereas plasma and normal tissues contain low exATP levels (10-100 nM) [7][8][9] . When animals are autopsied and tissues are sampled in ex vivo experiments, ATP concentration changes based on the physiological condition. ExATP and/or intracellular ATP might be degraded spontaneously or enzymatically, or intracellular ATP might be released from the cells during the process of ex vivo analysis. Thus, it is di cult to detect STA551 binding to the target quantitatively under physiological conditions using ex vivo analysis. Clarifying the minimal binding of STA551 in normal tissues in vivo would provide clear evidence of the less systematic immune activation of STA551. Therefore, we aimed to nd a working method for elucidating STA551 biodistribution in vivo.
Confocal uorescence microscopy has been widely used for observing cell physiology 10 . This tool allows us to observe high-resolution uorescent images of cells and tissues. Since confocal microscopy uses single photon absorption processes to produce images, it can only visualize tissues at depths of up to 100 µm 11 . In the last two decades, two-photon microscopy has been developed, which is an advanced form of microscopy which uses two-photon absorption processes to visualize images 12,13 . Since twophoton microscopy uses long wavelength lasers which has low energy, it causes low phototoxicity and allows long term imaging in a living animal. In addition, long wavelength light can penetrate deeper into tissues, allowing us to observe not only the surface but also deep inside tissues. Another advantage of this imaging technology is high resolution. Compared to PET and/or SPECT technology, two-photon imaging technology gives high-resolution images. High-resolution images allow us to observe detailed cell physiology, such cell morphology, movement, and cell-cell interaction 14,15 . Since two-photon microscopy is much better at intravital imaging, this technology is currently utilized to detect cell physiology in several organs including bone, brain and tumor [16][17][18][19] .
Intravital imaging using two-photon microscopes may be most appropriate way to detect switch antibody binding to the target cell for three reasons. First, two-photon microscopy can visualize deep inside various tissues, including tumor, spleen and lymph node. The difference in distribution between tumor and normal tissues should be detected. Second, we can obtain images in a living animal with physiological ATP concentrations in each tissue. Compared to ex vivo analysis, intravital imaging allows us to evaluate the antibody binding without injuring the cells or signi cantly interfering with ATP concentration. It was expected that the binding of switch antibody under physiological condition could be detected by intravital imaging. Third, high-resolution images allow us to observe antibody binding to the target cell.
In this study, we aimed to clarify the target binding of STA551 in tumor and normal tissues in vivo. First, we investigated CD137 expression in tumor-bearing human CD137 knock-in mice (hCD137 KI mice) and identi ed cell populations which express CD137. Second, we administered conventional CD137 agonistic antibody to tumor-bearing hCD137 KI mice and detected antibody-binding cells by owcytometry. Finally, we administered switch or conventional CD137 agonistic antibody to mice and con rmed the different distributions of these antibodies to tumor and spleen using two-photon microscopy.

Results
To clarify the target binding of STA551 in tumor and normal tissues in vivo, we took three steps (Fig. 1). STA551 binds to CD137 of human and cynomolgus monkey but does not bind to murine CD137 6 . Therefore, we used hCD137 KI mice 6 to investigate STA551 binding to the target.
To investigate CD137 expression in hCD137 KI mice, we created a LLC1/OVA/hGPC3 model and detected CD137 expression by analyzing owcytometry (Fig. 2). We rst analyzed CD137 expression on CD45 + cells in tumor and spleen. CD45 + cells in tumor showed higher CD137 expression than in spleen ( Fig. 2A).
CD137 expression was observed in 2-8% of CD45 + cells in tumor and in less than 2% of CD45 + cells in spleen. Among CD45 + cells, CD8 + T cells and NK cells had high CD137 expression (Fig. 2B). CD137 expression on CD4 + T cells and CD11b + cells were also detected (Fig. 2B). These data suggested that anti-CD137 antibody would distribute more to tumor than to spleen and binds to CD45 + cells, especially CD8 + cells and NK cells.
In previous research, to evaluate in vivo antitumor e cacy of STA551, tumor-bearing mice were treated with Sta-MB and Ure-MB. Sta-MB has the same variable region as STA551 and has MB as the constant region, which is an engineered constant region of mouse IgG1 to increase binding activity to mouse Fc gamma receptor II (FcγRII). Ure-MB has urelumab-like Fab and MB as the constant region. Ure-MB was used as a conventional CD137 agonist antibody in contrast to Sta-MB, an anti-CD137 switch antibody 6 .
To investigate anti-CD137 antibody binding in tissues, we administered Alexa Fluor 488-labeled Ure-MB and isotype control antibody (anti-KLH-MB) to tumor-bearing hCD137 KI mice. These antibodies were administered at the doses of 1 mg/kg twice. Sta-MB was not used because it might dissociate from the target during the process of ex vivo analysis. Ure-MB was detected not only in tumor but also in spleen ( Fig. 3A-B). Ure-MB bound to 5-30% of CD45 + cells in tumor and 10-25% of CD45 + cells in spleen. In addition, Ure-MB bound to CD4 + T cells, CD8 + T cells and CD11b + cells (Fig. 3D). NK cells in tumor were not detected in this experiment. In tumor, isotype control antibody bound to CD45+ cells, especially CD11b+ cells ( Fig. 3B-C). CD137 expression was not evaluated because Ure-MB binds to the same site of CD137 as the detection antibody. Given that the Ure-MB binding population was consistent with the CD137 expressing population shown in Fig. 2B, Ure-MB would bind to CD137-expressing immune cells. Taken together, these results indicate that Ure-MB binds to targets in both tumor and normal tissues.
To verify the target binding of STA551 in vivo, we investigated the distribution of Sta-MB in tumor and spleen. Alexa Fluor 488-labeled Ure-MB, Sta-MB, and isotype control antibody were administered to tumorbearing hCD137 KI mice and detected by two-photon microscopy. In tumor, uorescence from all antibodies were detected (Fig. 4A, C, Suppl Fig. 2). However, Sta-MB showed weaker uorescence in spleen than Ure-MB (Fig. 4B, C), and the Sta-MB uorescence was comparable to isotype control antibody. These data suggested that Sta-MB distributes and binds differently in tumor and spleen in vivo and shows little binding to CD137 in spleen.

Discussion
In this study, we revealed that anti-CD137 switch antibody binds to target cells differently in tumor and spleen in vivo by using two-photon microscopy. We rst examined the CD137 expression levels of tumor and spleen in tumor-bearing hCD137 KI mice. Conventional anti-CD137 antibody was then administered to tumor-bearing hCD137 KI mice to determine the distribution of the antibody to the tumor and spleen.
Finally, two-photon microscopy was used to detect the distribution of anti-CD137 switch antibody to tumor and spleen in vivo.
To con rm the different binding ability of STA551 between tissues in vivo, we needed to detect the distribution of antibodies in a non-invasive way to minimize ATP-concentration changes in tissues. Therefore, we used two-photon microscopy to detect antibodies intravitally. In experiments using twophoton microscopy, it can be time-consuming to prepare animals, optimize imaging, and monitor animals. On the other hand, in ex vivo analysis using ow cytometry, several experimental conditions, such as dosing regimen, can be veri ed with better thruput. In addition, ex vivo studies can detect the expression of target molecules, the binding of antibodies, and identify the types of cells. Establishing optimal imaging conditions for intravital imaging by using ex vivo ow cytometric analysis would be useful in studying antibody distribution in vivo.
In this study, in vivo imaging revealed that Sta-MB distributed and bound to the cells in tumor but much less so in spleen. It has been suggested that the toxicity induced by anti-CD137 antibody is CD137 dependent, and that T-cell and macrophage in ltration and secreted cytokines by the cells are involved in the pathogenesis [20][21][22] . Ure-MB was highly distributed to not only tumor but also spleen, and it seemed to lead to systemic toxicity by inducing CD137 signaling to normal tissues. Our ndings suggest that the tumor-selective target-binding ability of STA551 avoids the systemic reaction caused by conventional anti-CD137 antibodies.
Ure-MB, a conventional anti-CD137 antibody, bound to cells in tumor and spleen. Ure-MB mainly bound to T cells and NK cells, consistent with cell populations in which CD137 expression was observed Fig. 2B or previously reported 23 . Thus, this suggests that Ure-MB bound to the cells in a CD137 dependent manner. There were two possible reasons that the binding of Ure-MB was greater than the expression of CD137. The rst reason is that the administration of antibody increased the expression of CD137 molecules.

CD137 agonist signals are known to activate T cells and other immune cells, leading to increased
expression of CD137 24 . Ure-MB administered to mice might bind to CD137, introduce CD137 agonist signals in tumor and spleen and induce increased expression of CD137 in tissues. The second reason is the Fc region-mediated binding of Ure-MB. Ure-MB, isotype control antibody and Sta-MB have engineered Fc regions that bind to murine FcγRs, particularly FcγRII 6 , and FcγRII is predominantly expressed in the myeloid lineage cells 25 . Isotype control antibody bound to CD45 + cells, especially CD11b + cells, suggesting that the antibody bound to the cell via Fc region. However, the binding of Ure-MB to CD11b + cells was comparable with that of isotype control antibody in either tumor or spleen, suggesting that Fabmediated binding is more prevalent. In addition, since the main Ure-MB binding cells were con rmed to be CD137 expressing cells, such as T cells and NK cells, binding of Ure-MB to cells was thought to be primarily Fab-mediated binding.
In the imaging experiment, uorescence from all antibodies were detected in tumor. Table1 and suppl Fig. 1 show the concentration of Ure-MB, Sta-MB, and isotype control antibody in plasma, spleen, and tumor. The concentration of isotype control antibody in each tissue, especially in tumor, was higher than the other two antibodies. In image analysis, the difference could not be distinguished between the antibodies which were binding to target molecule and non-binding antibodies and/or bound nonspeci cally in tissues. Isotype control antibody which was highly distributed to tumor might be detected as uorescent signal in the imaging experiment.
STA551 is designed to bind to CD137 strongly in the presence of 100 µM ATP but not in the absence of ATP 6 . Murine ATP levels have been reported to be approximately 100 µM of extracellular ATP in tumor and 10-100 nM in normal tissues 7,8,26 . However, it is di cult to measure the exact ATP concentration in physiological conditions because ATP concentration changes depending on the sampling and measurement conditions due to degradation of ATP and release of intracellular ATP. This study revealed that Sta-MB showed binding in tumor and little binding to spleen. This data suggested that, in physiological conditions, tumor had ATP levels of 100 µM or higher and normal tissue had lower ATP levels. The present imaging results may be useful for estimating ATP levels of tissues under physiological conditions. In addition, human ATP levels have been reported to be 10-100 nM in normal tissues 26 and more than 10 µM in tumor, and there is around 1,000-fold difference in ATP concentration between tumor and normal tissues, which is similar in mice 27 . Considering the similarity in ATP distributions in human and mouse tissues, it is anticipated that STA551 will also exhibit tumor-selective binding in humans. Thus, STA551 is expected to exert anti-tumor e cacy with tumor selective CD137 signals, while reducing systemic reaction, even in human patients.
In conclusion, we showed that STA551 distributes in tumor but little in spleen. Such STA551 distribution demonstrated more clearly the reason why STA551 work in tumor but not in normal tissues. Because of less distribution in normal tissues, STA551 could be a promising therapeutic antibody for patients with currently di cult-to-treat cancers.

Methods
Cell line LLC1/OVA/hGPC3 cells were established by transfecting human GPC3 and chicken ovalbumin (OVA) expressing plasmids into LLC1, which was purchased from ATCC 6,28 .

Animals
The animal studies were carried out in compliance with the ARRIVE guidelines hCD137 KI male mice 6 . hCD137 KI mice were generated by replacing mouse CD137 with human CD137, and did not have mouse CD137 gene and protein but human CD137 gene and protein 6 .

Antibody labeling
Alexa Fluor 488 Antibody Labeling Kit (Thermo Fisher Scienti c) was used for labeling Ure-MB, Sta-MB, and isotype control antibody (anti-KLH-MB). The binding activity to human CD137 of labeled Ure-MB and Sta-MB in the presence and absence of ATP were determined by Biacore assay.
in vivo study in hCD137 KI mice In vivo study was performed according to previously reported procedures 29 with some modi cations. About 3 weeks after LLC1/OVA/hGPC3 tumor inoculation, 1 mg/kg of Alexa Flour 488-labeled isotype control antibody or PBS were intravenously administered once to investigate CD137 expression. For antibody distribution study, 1 mg/kg of Alexa Flour 488-labeled antibodies were intravenously administered at day 0 and 3. Mice were anesthetized with iso urane at 24 hours after the administration to be observed by intravital two-photon microscopy or to have their plasma and tissues collected for single cell analysis and determining antibody concentrations in tissues.  Intravital two-photon microscopy imaging As pretreatment for intravital two-photon microscopy imaging, mice were shaved, and the hair was removed with depilatory cream to prevent it being mixed into the visual eld. The spleen and tumor were then surgically exposed to be covered with the cover glass using n-butyl cyanoacrylate glue (3M Vetbond Tissue Adhesive, 3M). Qtracker 655 Vascular Labels (Thermo Fisher Scienti c) were intravenously administered into mice just before imaging of spleen and tumor in order to visualize the blood vessels in these tissues. Inverted multiphoton microscope (A1R-MP, Nikon) equipped with multi-immersion objectives (20X, Plan Fluor, numerical aperture [NA], 0.75, Nikon) was used to observe tumor and spleen. The microscope was driven by a Chameleon Vision II Ti:Sapphire laser (Coherent) tuned to 930 nm. The uorescence was detected by an external non-descanned detector with four channels (Nikon) with three dichroic mirrors (495, 560, and 593 nm) and four band-pass lters: 492 nm for the second harmonic generation (SHG) signal, 525/50 nm for Alexa Fluor 488, 575/25 nm for tdTomato, and 629/56 nm for Qtracker 655. 4 to 11 images of around 300 µm x 300 µm with a vertical step size of about 3 µm to a depth of around 100 µm were collected of each tissue and then analyzed by NIS-Elements integrated software (Nikon) to create the maximum intensity projection (MIP) images with median lters for noise reduction. The area of antibody-binding region in tumor and spleen was detected by ImageJ software. The antibody-binding region was extracted from MIP images. Area fraction was calculated by uorescence area/total area in all images.

Statistical Analysis
Statistical analyses were performed with GraphPad Prism 7.0 (GraphPad Software). CD137 expression level and antibody distribution were compared by using Student t-test. Antibody uorescence in the two groups were compared using Tukey's multiple comparisons test. Signi cant values were marked as: * p < 0.05, ** p < 0.01, *** p < 0.001, and n.s. Figure 1 Research strategy for detection of antibody binding in tissues by two-photon microscopy. In step1, human CD137 expression was examined in tumor-bearing hCD137 KI mice. In step2, uorescent labeled anti-CD137 antibody was administered to hCD137 KI mice. Antibody binding cells in tumor and spleen were detected. Finally, antibody binding cells in tumor and spleen were detected by two-photon microscopy in step3.