Immuno-PET Imaging of Siglec-15 Using the Zirconium-89-Labeled Therapeutic Antibody, NC318

Objective . Sialic acid-binding immunoglobulin-like lectin 15 (Siglec-15) is overexpressed in various cancers which has led to the development of therapeutic anti-Siglec-15 monoconal antibodies (mAbs). In these preclinical studies


Introduction
Sialic acid-binding immunoglobulin-like lectin 15 (Siglec-15) is a cell surface type-1 transmembrane protein that is overexpressed in a variety of tumor types as well as in selected immune cells such as tumor-associated macrophages and dendritic cells. Siglec-15 expression in the tumor microenvironment (TME) is thought to be associated with permissive host immunity conducive to disease progression [1][2][3][4]. In addition to modulating tumor immune responses, Siglec-15 is expressed on osteoclasts and plays a role in osteoclast regulation and bone remodeling [5]. These diverse modulatory functions are activated upon Siglec-15 binding to sialoglycan structures such as Neu5Acα2-6GalNAcα (sialyl Tn) which is overexpressed in many cancer types including gastric, breast, lung, and ovarian [6]. CD44, a glycoprotein overexpressed on human hepatoma cells, has modified sialoglycans that can also serve as ligands for Siglec-15 thereby promoting tumor cell migration [7,8]. Hence, Siglec-15 has been identified as an important immunomodulator and potential target for the treatment of osteoporosis and cancer offering certain advantages over other immunotherapeutics. Although Siglec-15 shares structural homology with programmed cell death ligand-1 (PD-L1), the immune regulatory mechanism is distinct and may offer an alternative therapeutic option in cancer patients unresponsive to PD-L1 immunotherapies [9].
In preclinical osteoporosis and tumor mouse models, antibody blockade of Siglec-15 resulted in increasing bone density and tumor regression (by reversing the immunosuppression in the TME), respectively [9][10][11][12]. These findings have prompted further development of anti-Siglec-15 monoclonal antibodies (mAbs) for evaluation as immune checkpoint inhibitors in clinical trials which include Next-Cure's NC318 for solid tumors, Medimmune's mAb for acute myeloid leukemia, and Daichi Sankyo's DS-1501 for osteoporosis [9]. NC318 in a phase I clinical trial which included patients with non-small-cell lung carcinoma (NSCLC), melanoma, ovarian, colorectal, breast, and other types of cancer demonstrated efficacy primarily in patients with NSCLC (20%). This phase I trial was conducted without any biomarker assessment, but the NSCLC patient response rate of 20% was found to correlate with the Siglec-15 positivity of 25.7% (immunohistochemistry, IHC) observed in tumor biopsy stained sections of NSCLC patients (241) from a subsequent study [4,9]. These results would indicate that tumor Siglec-15 expression may influence the therapeutic response; however, the molecular interactions remain to be elucidated. Hence, a biomarker has yet to be identified that would be predictive of therapeutic responses in patients and aid in patient selection for anti-Siglec-15 mAb therapies.
Radiolabeling of these therapeutic anti-Siglec-15 mAbs for immuno-positron emission tomography (immuno-PET) imaging could prove useful in quantitating in vivo expression levels of Siglec-15 in tumors in real time potentially aiding in patient selection, monitoring changes in Siglec-15 over a treatment time course, and determining the relationship of Siglec-15 expression levels to patient therapeutic responses. Although these therapeutic mAbs possess the high affinity and specificity required for a successful PET imaging agent, long lived PET radionuclides such as zirconium-89 (t 1/2 = 78:4 h) are required to match the long biological half-life of mAbs. In this report, a therapeutic humanized IgG 1 mAb which recognizes both human and murine Siglec-15, NC318, was labeled with zirconium-89 for preclinical evaluation of Siglec-15 targeting and potential to serve as a biomarker for patient selection and therapeutic responses. These preclinical studies included both in vitro binding assays with human melanoma and NSCLC cancer cells with varying Siglec-15 expression levels and in vivo biodistribution studies in tumor-bearing mouse xenograft models using the same cell lines to determine the clinical potential of [ 89 Zr]Zr-DFO-NC318.

Materials and Methods
Humanized IgG 1 monoclonal Ab, anti-Siglec-15 mAb (NC318), was kindly provided by Dr. Ido Weiss (NextCure, Beltsville, MD, USA). The p-isothiocyanatobenzyldesferrioxamine (DFO-Bz-NCS) was purchased from Macrocyclics, Inc. (Plano, TX, USA). Sodium acetate and Tris-HCl were purchased from Thermo Fisher Scientific (Waltham, MA, USA). The lyophilized whole human serum was obtained from MP Biomedicals, LLC (Solon, OH, USA) and dissolved in 2 mL saline. This serum solution was directly used without inactivation for the stability study of [ 89 Zr]Zr-DFO-NC318. All other chemicals and reagents were purchased from Sigma-Aldrich (St. Louis, MO, USA) and used without further purification. PD-10 desalting columns were obtained from GE Healthcare Biosciences (Pittsburgh, PA, USA). Zirconium-89 oxalate was obtained from 3D Imaging (Little Rock, Arkansas, USA). Analytical highperformance liquid chromatography (HPLC) analyses were performed on an Agilent 1200 Series instrument equipped with a multiwavelength UV detector connected in series with a Bioscan flow count radiodetector. HPLC was performed using a size exclusion column (SE, 4:6 : mm ID × 30 cm, 4 μm), TSKgel SuperSW3000, obtained from Tosoh Bioscience LLC (King of Prussia, PA, USA) and an eluent system comprised of 0.1 M sodium phosphate (pH 6.8), 0.1 M sodium sulfate, 0.05% sodium azide, and 10% isopropyl alcohol at a flow rate of 0.3 mL/min. The gel filtration standard (cat #151-1901) for calibration of the size exclusion column was obtained from Bio-Rad (Hercules, California, USA  (20 μL,~5 mg/mL in water, pH adjusted to 7 with 2 M Na 2 CO 3 solution) was added followed by a solution of DFO-NC318 (0.4 mg, 5.7 mg/mL, 70 μL). The reaction mixture was incubated for 1 h at room temperature and challenged with DTPA (5 μL, 0.1 M, pH 7) for an additional 10 min. The radiolabeled conjugate was purified by PD-10 column using 0.9% NaCl (pH 7). The molar activity and the purity of the radiolabeled conjugate were determined by SE-HPLC (t R = 7:4 min). The identity of the [ 89 Zr]Zr-DFO-NC318 was confirmed by comparing the retention time (based on UV 280) with DFO-NC318 (t R = 7:4 min) and the gel filtration standard.  Figure S1 and Table S1). To determine serum stability, whole human serum (500 μL) was added to a solution of [ 89 Zr]Zr-DFO-NC318 (~500 μCi in 500 μL of saline, pH 7.0) and kept at 37°C for up to 7 d. The radiochemical stability was determined every 24 h by directly injecting an aliquot of the solution into the HPLC and by iTLC (Supplementary Information Figure S2, Figure S3, and Table S2).

In Vitro Studies.
Saturation binding studies were performed to determine the K d and B max with 624-MEL WT/ +S15 and LOX-IMVI cells in plates (2-10 × 10 5 cells/well) or tubes (2-10 × 10 5 cells/tube) to which increasing concentrations of [ 89 Zr]Zr-DFO-NC318 (0.25-25 nM) were added to duplicate wells or tubes; nonspecific binding was determined by adding nonradioactive NC318 mAb (10 -6 M) to another set of duplicates. For competition studies, increasing concentrations (0-1000 nM) of nonradioactive NC318 were added to a constant concentration of [ 89 Zr]Zr-DFO-NC318 (0.75 to 2.0 nM) and 624-MEL+S15 cells. After incubation (2 h, 4°C), the cell bound [ 89 Zr]Zr-DFO-NC318 was separated from the free radiolabeled antibody either: (1) plated cells were washed with phosphate-buffered saline (PBS), treated with trypsin, and collected in vials; or (2) cells in tubes were pelleted by centrifugation, washed twice (PBS), and supernatants removed. The cell bound radioactivity for these samples was determined by gamma counting (Perkin Elmer 2480 Wizard3, Shelton, CT). From the saturation studies, the K d and B max were determined from 6 to 8 concentrations of [ 89 Zr]Zr-DFO-NC318 and analyzed using nonlinear regression curve fitting (one-site specific binding); from the competition studies, K i 's were determined from 8 to 10 competitor concentrations (Prism (version 5.04 Windows), GraphPad software, San Diego, CA).

Human
Siglec-15 Dosimetry Estimation. Human dosimetry estimates extrapolated from the mouse biodistribution studies were calculated using OLINDA V1.1 (Vanderbilt University, TN) with the mouse to human fractional organ extrapolation of the mean residence times of the ligand measured by the biodistribution described above. The %ID/g values (determined from the biodistribution studies described above) for a set of organs determined over a 7 d time course were used to extrapolate human dosimetry in the same organs. The whole organ was dissected from the carcass and counted to measure the organ's radioactive content. For the bone, skin, muscles, and blood samples, a sample was dissected, weighed, and counted in the gamma counter. Because of the relatively small uptake in the skin, muscle, and blood, these tissues were not included in the kinetic input form of the OLINDA dosimetry estimation software.
Biodistribution data showed radioconjugate uptake in the mouse skeleton above the background. To account for this, a special case was made for bone dosimetry estimation. Instead of including the bone activity in the body remainder volume, the whole bone activity was estimated and entered into the trabecula bone input field in OLINDA. The bone tissue %ID/g and the %ID/organ were estimated using a murine bone fraction model of 53.3 (g/kg) [16]. This calculation gives an estimated bone mass of 1.07 g for a 20 g mouse. The %ID/g at each time point for each mouse was multiplied by the bone mass fractional estimate for each mouse.
From the %ID/organ, time activity curves (TAC) were generated from PET images and residence times were calculated in units of hours. Between the time points of the TAC, a trapezoidal model was used to estimate the area under the curve. For the last time point, an exponential decay curve with the half-life of zirconium-89 was used to extrapolate the tail of the TAC. Since the %ID/organ of the whole intestine was measured (including the contents), the absorbed activity between the large and small intestines was estimated by the MOBY fractional mass model for a 25 g mouse [17]. The result was that 75% of the activity was assigned to the small intestine and 25% to the large intestine. The large intestine was further separated using the ICRP 80 standard in which 57% of the radioactivity was assigned to the upper and the remaining 43% assigned to the lower large intestine.
Whole slide imaging (WSI) was performed with an Aperio ScanScope XT (Leica) at 200x in a single z-plane. Digital pathology for biomarker quantification was performed following WSI with thresholds for positivity determined using known positive controls. Tumor necrosis was estimated using random forest machine learning algorithms on H&E images. Microvessel density was estimated using CD31 stained tissue sections with an object detection algorithm. CD45 and Iba1 positive cells are reported as number of positive cells per mm 2 . Cell detection algorithms were run to quantify positive cells, which are expressed as the number of positive cells per mm 2 of tissue and the percent of CD45 and Iba1 positive cells. Siglec-15 expression is reported as an H-score in which the proportion of all cells (tumor, spleen, or lymph node) found to express Siglec-15 was determined and then multiplied by the staining intensity score to obtain a final semiquantitative H-score (maximum value of 300 corresponding to 100% of cells positive for Siglec-15 with an overall staining intensity score of 3). Dual immunofluorescence for NUMA1 (human marker) and Siglec-15 was performed as Siglec-15 is reactive with both mouse and human. Immunofluorescence staining of human tumor cells was differentiated from mouse cells based on a positive NUMA1 nuclear signal. 4 Molecular Imaging

Radiochemistry.
Zirconium-89-labeled DFO-NC318 conjugate was prepared following the literature method with minor modifications [13]. The isolated radiochemical yields were in the range of 85-95% (n = 20) with radiochemical purity > 95% ( Figure 1). The molar activities of the radioimmunoconjugates were 22,200-70,300 MBq/μmol (n = 15). To determine the storage stability, a saline solution of [ 89 Zr]Zr-DFO-NC318 was kept at 4°C and monitored by SE-HPLC every 24 h. A slow decomposition was observed (91% intact after 48 h; Supporting information Figure S1 and Table S1). Whole human serum stability determined by SE-HPLC at 37°C indicated 48% decomposition in 4 days (Supporting information Figure S2 and Table S2) and remained unchanged from 4 d to 7 d. As free zirconium-89 is often trapped on a SE-column to a significant degree, iTLC analysis was performed on days 3-7 to assess the presence of free zirconium-89. The iTLC results indicated the formation of~15% free zirconium-89 in 4 d which remained relatively constant until 7 d. The amount of intact [ 89 Zr]Zr-DFO-NC318 based on iTLC after 7 d was 60%, similar to the SE-HPLC results (Supporting information Figure S3). Nonspecific binding (B ns ) comprised the greater part of the binding with the 624-MEL WT and HCC-827 cells ranging from 57% to 96%. The moderate and low Siglec-15 expression levels of LOX IMIV and HCC-827, respectively, found in these binding studies were consistent with published results using flow cytometry [4,19]. These in vitro results indicate that [ 89 Zr]Zr-DFO-NC318 distinguishes between minimal to moderate Siglec-15 expression levels and would be appropriate for in vivo imaging of tumors with high to moderate Siglec-15 expression levels as was observed with the 624-MEL+S15 cells and potentially LOX IMIV. Conversely, tumors with lower Siglec-15 expression levels (<27,000 sites per cell) as in the case of 624-MEL WT and HCC-827 cells may not be clearly discernable from the background.
In addition to the presence of murine Siglec-15+ positive cells in the TME, murine Siglec-15+ positive cells were generally observed in the spleen and lymph nodes in IHC sections from 624-MEL+S15 xenografts further confirming the specific uptake of [ 89 Zr]Zr-DFO-NC318 in these tissues (Figure 9).

Dosimetry Estimation for [ 89 Zr
]Zr-DFO-NC318. The extrapolation of radioconjugate residence times in humans was determined from the radioactivity content of the organs    Siglec-15 has emerged as a novel immune inhibitor which utilizes a pathway that is distinct from the PD-1/ PD-L1 immune checkpoint pathway and therefore may represent the next generation of immunotherapeutics. Preliminary results in mouse tumor models have shown tumor regression in both PD-1 sensitive and insensitive tumors suggesting that Siglec-15-targeted immunotherapeutics may offer an alternative to patients resistant to PD-1/PD-L1 therapies [22,23]. In a NC318 clinical trial with NSCLC patients who were refractory for anti-PD-1 mAb therapy, 20% had complete or partial responses and 30% had stable disease [9]. Although the results from this clinical trial are encouraging, the patients were not selected based on their PD-1/PD-L1 or Siglec-15 expression levels, and therefore, more definitive results may be possible with selected populations. Development of diagnostic agents and a predictive biomarker are needed to select patients for Siglec-15targeted therapeutics and to gain a better understanding of the relationship of the interactions between Siglec-15 positive tumor cells and immune cells in the tumor TME to therapeutic responses. An IHC assay has been developed for patient selection for Siglec-15-targeted therapies which detects Siglec-15+ tumor and immune cells in tumor biopsies [24]. These IHC results reflect the Siglec-15 positivity of the tumor at the time and location of the biopsy and may not reflect changes that have occurred in the TME from the time of the biopsy to the start of therapy or during the therapeutic time course. Since preclinical studies would suggest that the overexpression of Siglec-15 in the TME plays a role in the immunosuppression of the tumor and a therapeutic such as NC318 would act to reverse this immunosuppression in the TME, assessing the dynamic changes in the TME of Siglec-15 expression levels in response to therapy is needed [9]. Immuno-PET imaging with [ 89 Zr]Zr-DFO-NC318 could provide a real-time readout of Siglec-15 expression levels in all lesions of the patient and surrounding tissues as well as monitor changes in expression levels with treatment.
Generally, the uptake of [ 89 Zr]Zr-DFO-NC318 was higher in the lymphoid tissues of the athymic tumorbearing mice compared to the immunocompetent Balb/c 13 Molecular Imaging mice indicating that the athymic mice had increased Siglec-15+ expressing immune cells. Athymic mice are known to have impaired T-cell function with fewer circulating leukocytes and depletion in "thymus dependent" areas of the spleen, lymph nodes, and bone marrow compared to normal mice but have normal cytotoxic responses to T-independent antigens [25,26]. In contrast, natural killer (NK) cell and macrophage cytotoxic activities have been reported to be enhanced in athymic mice compared to immunocompetent mice suggesting that these immune cell types may compensate for the T-cell deficiency [27]. In particular, athymic mice transplanted with human tumor cells have been found to have allograft cytotoxic responses mediated by macrophages and NK cells [27,28]. As the tumor progresses, a state of chronic inflammation ensues resulting in the accumulation of myeloidderived suppressor cells (MDSCs) which is a heterogeneous immature myeloid cell population with immunosuppressive functions that operate by a wide variety of mechanisms [29]. The abnormal production of growth factors and cytokines by the tumor cells and stroma includ-ing resident macrophages causes expansion of MDSCs by inhibiting normal myeloid differentiation [30]. The expansion of MDSCs has been found to extend beyond the tumor to the spleen of tumor-bearing mice as well as bone marrow cells cultured with tumor cells [31]. Within the cell groups comprising immature MDSCs, a small group phenotypically similar to monocytes (M-MDSCs) can differentiate into macrophages, dendritic cells, and osteoclasts that primarily have immunosuppressive functions both at the tumor site and periphery [31,32]. Similarly, macrophages induced with growth factors and other innate inflammatory mediators as well as tumor-associated macrophages (TAM) have been found to overexpress Siglec-15 suggesting that a subset of the M-MDSC cells are Siglec-15+ [9]. Taken together the increased cytotoxic activity of macrophages and then the allograft immune response resulting in the expansion of Siglec-15+, M-MDSCs in the spleen, lymph node, and bone marrow of athymic tumor-bearing mice may be expected to account for the increased [ 89 Zr]Zr-DFO-NC318 uptakes observed in these lymphoid tissues. 14 Molecular Imaging The in vivo [ 89 Zr]Zr-DFO-NC318 uptakes of the 624-MEL+S15, LOX-IMIV, HCC-827, and 624-MEL WT tumors exhibited similar rank order compared to the in vitro (B max ) and IHC results; however, the magnitude of the differences in Siglec-15 expression levels between the tumor types in vivo was less than would have been predicted from the in vitro results (B max ). Most likely the [ 89 Zr]Zr-DFO-NC318 increased tumor uptakes in LOX IMIV, HCC-827, and 624-MEL WT cell lines can be attributed to the contribution of Siglec-15+ murine immune cells in the TME which was confirmed with IHC. Further LOX IMIV, HCC-827, and 624-MEL WT tumor sections exhibited high levels of CD45+ and Iba1 + immune murine cells indicating the presence of infiltrating immune cells in the TME which in tumorbearing athymic mice would be expected to include dendritic cells, NK cells, and macrophages. Further, these IHC results indicate that macrophages comprise a majority of the immune cell type in the TME since Iba1 is a marker specific for macrophages whereas the CD45+ marker cells include not only macrophages but T-cells, B-cells, dendritic cells, NK cells, stem cells, and granulocytes [33]. Both macrophages and dendritic cells are known to express Siglec-15; therefore, in the TME, a subset of these immune cell types would be capable of Siglec-15 expression. These preclinical studies would suggest that although tumors have low expression levels of Siglec-15, the presence of Siglec-15+ immune cells in the TME may make possible detection with PET imaging in a clinical setting. However, these preclinical studies are not predictive of the immune cell types nor the Siglec-15+ subset that would comprise the TME in human patients and would require [ 89 Zr]Zr-DFO-NC318 imaging studies in human subjects.
These promising preclinical results would suggest that the [ 89 Zr]Zr-DFO-NC318 would be appropriate for Siglec-15 immuno-PET imaging in human subjects to establish the predictive value of Siglec-15 expression as a reliable biomarker for patient selection in clinical trials, monitoring therapeutic responses and evaluating the efficacy of this new class of immunotherapeutics. Siglec-15 expression in humans is generally absent from normal tissue and confined to myeloid cells and osteoclasts in lymphoid tissues [9,34]. In normal human spleen and lymph nodes, Siglec-15 expression was found on a small number of dendritic cells and macrophages suggesting that Siglec-15 expression occurs on a subset of dendritic cells or macrophages [35]. In contrast, Siglec-15 expression levels are upregulated in tumors and tumor infiltrating macrophages which may also extend to the spleen, lymph nodes, and other lymphoid tissues. Therefore, in PET images, lesions of cancer patients in which Siglec-15 has been upregulated should be discernible from normal tissue. In addition with whole body PET imaging, changes in Siglec-15 expression levels in the spleen, lymph nodes, and other lymphoid tissues can be assessed to determine the role Siglec-15 plays in cancer progression and M-MDSC biology [31]. Additionally, since NC318 is currently in clinical trials, [ 89 Zr]Zr-DFO-NC318 could be used as a companion diagnostic imaging agent as well as assist in establishing dosing and tracking in vivo the tissue distribution of NC318 to gain a better understanding of off-target side effects. Since NC318 has been found to be safe in on-going clinical trials, clinical translation of [ 89 Zr]Zr-DFO-NC318 can be completed relatively more easily for "proof of concept" PET imaging studies. Providing these studies corroborate the value of Siglec-15 as a predictive biomarker further

Spleen
Lymph node Spleen Lymph node Figure 9: Representative Siglec-15 immunohistochemistry stained sections of the spleen and lymph nodes from 624-MEL+S15 xenografts. Spleen: Siglec-15+ cells were commonly observed in the splenic red pulp with a cytoplasmic to membranous staining pattern; lymph node: faint Siglec-15+ cells were commonly observed in the medulla but occasionally were present in the paracortex or cortical regions of the node. development of small molecules or Ab fragments targeting Siglec-15 labeled with shorter lived radionuclides (i.e., fluorine-18) would be justified. The faster pharmacokinetics of these imaging agents would allow for same day imaging rather than waiting several days to acquire images which is preferable with whole mAb-based imaging agents.
In conclusion, [ 89 Zr]Zr-DFO-NC318 would make available real-time PET images that represent Siglec-15 tumor expression levels not only for the primary tumor but metastatic lesions as well. In patients with metastatic disease, [ 89 Zr]Zr-DFO-NC318 imaging would make possible the measurement of all Siglec-15+ lesions thereby providing a potentially better tool for patient selection than a biopsy of the primary tumor. Such imaging could also potentially serve as a biomarker to monitor responses in patients undergoing Siglec-15-targeted therapies. Validated PET imaging agents that could identify and quantify tumor Siglec-15 expression levels would be beneficial not only for clinical diagnostic and prognostic applications but also for the drug development process as well.

Data Availability
The data used to support the findings of this study are included within the article and the supplementary information files.

Disclosure
The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services nor does mention of trade names, commercial products, or organization imply endorsement by the U.S. Government.

Conflicts of Interest
The authors declare that they have no conflicts of interest.

Supplementary Materials
The stability of [ 89 Zr]Zr-DFO-NC318 stored as a saline solution at 4°C was determined using SE-HPLC at 24 h and 48 h following radiosynthesis; a slow decomposition was observed after 48 h (91% intact; Supporting information, Fig. S1; Table S1). In vitro, the stability of [ 89 Zr]Zr-DFO-NC318 in whole human serum at 37°C was determined by SE-HPLC which indicated a 48% decomposition at 4 days (Supporting information, Fig. S2; Table S2) that remained relatively unchanged from 4 d to 7 d. (Supplementary  Materials)