Targeting CD47 in Anaplastic Thyroid Carcinoma Enhances Tumor Phagocytosis by Macrophages and Is a Promising Therapeutic Strategy

Formalin-fixed, paraffin-embedded (FFPE) tissues from 19 patients (14 females; M_age ± SEM = 72.5 ± 2 years) with a diagnosis of ATC between 2005 and 2018 were identified in the archives of the Institute of Pathology, University of Bern (Bern, Switzerland). Cases were reviewed in detail by three board-certified surgical pathologists with a special interest in thyroid pathology (A.P., A.M.S., and M.S.D.), and were reclassified/restaged according to the 2017 World Health Organization Classification of Tumors of Endocrine Organs and the 2017 American Joint Committee on Cancer/Union for International Cancer Control tumor-node-metastasis guidelines (30–32). Patient characteristics are detailed in Table 1 and Supplementary Table S1. The study was approved by the local Ethics Committee of the Canton of Bern (KEK 200/14, KEK 2018-01502). A fresh sample for fluorescence-activated cell sorting (FACS) analysis was obtained after surgery and diagnosis of ATC in frozen sections, with written informed consent from the patient.

All sections were cut to 2 μm thickness. Hematoxylin and eosin–stained sections were obtained from each FFPE block. Immunohistochemistry (IHC) staining of full slides from FFPE blocks was performed on a Leica BOND RX automated immunostainer using Bond primary antibody diluent and Bond Polymer Refine DAB detection kit according to the manufacturer's instructions (Leica Biosystems). Details on antibodies, clones, manufacturers, and staining conditions for IHC are listed in Supplementary Table S2. Analysis and interpretation of the staining results were performed by two board-certified surgical pathologists (C.M.S and M.S.D.) and one pathologist in training (S.F.) in accordance with the “REporting recommendations for tumor MARKer prognostic studies” guidelines (33). Tumor cells were morphologically identified by cell size, shape, and nuclear configuration. CD47 staining in tumor cells was classified microscopically as 0 (absence of any membranous or cytoplasmic staining), 1+ (weak or incomplete membranous and/or cytoplasmic staining), 2+ (complete membranous staining of intermediate intensity), and 3+ (complete membranous staining of strong intensity). The calreticulin staining pattern was mostly granular and cytoplasmic and was classified microscopically as 0–3+. For CD68, CD163, PD-1, and PD-L1 staining, the positive cell frequencies were estimated by microscopy and were quantified by QuPath analysis, as described below. The concordance of microscopical estimation and QuPath quantification was in the range of ±10% for all cases, except for PD-1 and PD-L1 staining in 7 and 10 cases, respectively, which could not be evaluated adequately by automated QuPath analysis due to the predominantly weak membranous staining pattern. Therefore, for PD-1 and PD-L1 staining, only the values from microscopical estimation were used. All results are detailed in Supplementary Table S1.

Slides were scanned using an Aperio Scanscope CS digital slide scanner (Leica Biosystems) and analyzed using QuPath software v0.1.2. (34). For each sample, a selected and defined tumor area (at least 1 mm²) was analyzed. For detection of macrophages (CD68, CD163), T cells (CD3, CD4, CD8), granulocytes (CD15), NK cells (CD56), plasmacytoid dendritic cells (CD123), vasculature (CD31), as well as PD-1⁺ and PD-L1⁺ cells, the QuPath positive cell detection algorithm was used with the following setup parameters: detection image, hematoxylin OD for CD68, CD163, PD-1, and PD-L1; optical density sum for CD3, CD4, CD8, CD15, CD56, CD123, and Ki-67; requested pixel size, 0.5 μm; nucleus parameters—background radius 8 μm, median filter radius 0 μm, sigma 2.0 μm, minimum area 10 μm², and maximum area 400 μm²; intensity parameters—threshold 0.02, maximum background intensity 2.0, split by shape yes, exclude DAB (membrane staining) no; cell parameters—cell expansion 3 μm include cell nucleus yes; general parameters—smooth boundaries yes, make measurements yes; and intensity threshold parameters—score compartment: cell, DAB OD mean, threshold 1 + 0.2, single threshold yes. For Ki-67 staining, the scoring compartment in intensity threshold parameters was switched to nucleus: DAB OD mean. For samples showing a stronger background staining (especially CD163 IHC), setup intensity parameters were modified as follows: intensity threshold parameters—score compartment: cell, DAB OD mean, threshold 1 + 0.2, threshold 2 + 0.4, threshold 3 + 0.5, single threshold no. The quality of segmentation and positive and negative cell detection was visually analyzed and confirmed for each case.

The human ATC cell lines 8505C (35), 8305C (35), HTH-104 (36), BHT-101 (37), CAL-62 (38), C643 (39), and SW-1736 (40) were obtained from Prof. Dr. Martin Walter (Department of Nuclear Medicine, Inselspital, Bern University Hospital, Bern, Switzerland), and have been described before. The human ATC cell line OCUT-2 (41) was obtained from R.-P.C. and has been described before.

NOD.Cg-Prkdc^scid Il2rg^tm1WjI/SzJ (NSG) mice (42) were purchased from Charles River Laboratories. Braf^CA/+; Pik3ca^Lat/+; ThyroglobulinCre^ERT2 (Thyro-DT) mice from a mixed FVB/C57BL6/F129 background have been previously described (43). Mice were housed under specific pathogen-free conditions in isolated ventilated cages on a 12-hour/12-hour cycle of light and dark, fed ad libitum, and regularly monitored for pathogens. All mouse experiments were licensed by the Canton of Bern and were performed in compliance with Swiss Federal legislation.

The in vitro phagocytosis assay was performed as described before (23). Buffy coats and human serum were obtained from the Swiss Blood Bank (Interregionale Blutspende SRK, Bern, Switzerland) under the signed consent of the donors and in agreement with local legislation. Peripheral blood mononuclear cells (PBMCs) were enriched from buffy coats by density centrifugation using Lympho Spin Medium (pluriSelect). Monocytes were isolated from PBMCs using the EasySep Human CD14 Positive Selection Kit II (Stemcell Technologies) according to the manufacturer's instructions. Monocytes (4–5 × 10⁶ per well) were differentiated into macrophages for seven days at 37°C, 5% CO₂, on six-well tissue culture plates in Iscove's Modified Dulbecco's Medium supplemented with 10% human serum, 1% L-glutamine, and 1% penicillin/streptomycin. 8505C cells were harvested using non-enzymatic cell dissociation buffer (Sigma–Aldrich), washed in phosphate-buffered saline (PBS) three times, and then labeled with 5(6)-carboxyfluorescein diacetate N-succinimidyl ester (CFSE) in PBS at a final concentration of 20 μM. Macrophages were starved in serum-free medium for two hours, and 1 × 10⁶ CFSE-labeled 8505C cells per well were added. Cells were co-cultured for two hours in the presence of 10 μg/mL mouse anti-human CD47 (clone B6H12.2; Thermo Fisher Scientific) or isotype control (mouse IgG1; Thermo Fisher Scientific). Then, wells were imaged at 20 × magnification on a digital inverted fluorescence microscope (EVOS™ FL imaging system; Thermo Fisher Scientific). Phase contrast and green fluorescent protein (GFP) overlay images were exported, and the phagocytosis index and CFSE⁺ cells per macrophage were manually determined by two pathologists in training (S.F. and F.B.) in an independent fashion for each experiment. After microscopic imaging, cells were harvested using trypsin/EDTA solution (Sigma–Aldrich), stained for anti-human CD45 and CD14, and analyzed by FACS.

293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM)/F12 supplemented with 10% tetracycline-free fetal calf serum (FCS), penicillin (100 IU/mL), and streptomycin (100 μg/mL), and grown to 80% confluency in T150 flasks. The pLVX-EF1α-TS-EGFP-IRES-Puro vector was cloned by introducing a TS-EGFP DNA string (GeneArt^® Elements™; Thermo Fisher Scientific) into the EcoRI and BamHI sites of the pLVX-EF1α-IRES-Puro vector (Takara Bio, Inc./Clontech). The construct was verified by sequencing. For transient transfection of 293T cells, 20 μg pLX-EF1α-TS-EGFP-IRES-Puro and 60 μL Lenti-X packaging mix (GE Healthcare) were mixed with 240 μL Dogtor (OZ Biosciences) in a total volume of 500 μL Opti-MEM (Thermo Fisher Scientific) followed by 20 minutes of complex formation at room temperature and incubation with the cells. Twenty-four hours post transfection, the medium was exchanged. Lentiviral supernatants were harvested at 48, 72, and 96 hours post transfection and filtered through a 0.45 μm PES filter, followed by sixfold concentration with Lenti-X Concentrator (Takara Bio, Inc./Clontech) according to the manufacturer's instructions. 8505C cells cultured in T25 flasks in RPMI-1640 supplemented with 10% tetracycline-free FCS, penicillin/streptomycin, and 1% L-glutamine were transduced three times with 1 mL viral supernatant at a 1:5 dilution. Forty-eight hours post transduction, cells were selected with puromycin at a final concentration of 3 μg/mL for nine days. Finally, GFP^hi-expressing cells were purified by FACS sorting.

Parental or stably GFP^hi-expressing 8505C cells (4–5 × 10⁶) were injected subcutaneously (s.c.) into the flanks of six-week-old female NSG mice. Starting three days after tumor cell injection, mice were treated with 500 μg anti-human CD47 monoclonal antibody (mAb; clone B6H12; BioXCell) or mouse IgG1 isotype control mAb (clone MOPC-21; BioXCell) by intraperitoneal (i.p.) injection twice a week. Tumor growth was measured with a caliper twice a week, and tumor volumes were calculated by the formula V = (π × width × length × height)/6. After five to six weeks, mice were sacrificed by CO₂ inhalation followed by cervical dislocation, and tumors were excised, measured, and weighed.

Tumors were induced in Thyro-DT mice by daily i.p. injections of 1 mg tamoxifen diluted in 100 μL peanut oil (Sigma–Aldrich) on five consecutive days. To monitor tumor development, tumors were measured by ultrasound every three weeks. Mice were anesthetized using 5 μL/g of body weight of a mixture of 0.1 mg/mL medetomidine, 0.5 mg/mL midazolam, and 5 μg/mL fentanyl in 0.9% NaCl (Sigma–Aldrich) by i.p. injection. The fur around the neck was epilated with Veet^® hair removal cream. Images were acquired with an ESAOTE MyLab Five ultrasound device equipped with an 18 MHz LA455 Probe (Siemens). After imaging, anesthesia was reversed with 10 μL/g body weight of a mixture of 0.25 mg/mL atipamezole, 5 μg/mL flumazenil, and 20 μg/mL naloxone in 0.9% NaCl (Sigma–Aldrich) by s.c. injection. Ultrasound images were analyzed using ImageJ software. After three months of tumor growth, tumor-bearing mice were evenly divided into two groups according to tumor sizes. All mice weighed between 20 and 30 g at the start of the experiment. Mice were treated with either 500 μg (5 mg/mL) of anti-mouse/human/rat CD47 mAb (clone MIAP410; BioXCell; treatment group) or 100 μl InVivoPure, pH 7.0, dilution buffer (BioXCell; control group) by i.p. injection twice a week. During treatment, ultrasound measurements of tumor size were performed every two weeks. Tumor size was approximated by the biggest area in cross-section found for each tumor. Continuous measurements were normalized to the measurement at the start of treatment to create tumor growth curves for each individual mouse.

For sample preparation, Thyro-DT mice were anesthetized with 10 mg/mL ketamine and 1.6 mg/mL xylazine at a dose of 10 μL/g body weight by i.p. injection. When mice were unresponsive to mechanical stimuli, the thoracic and abdominal cavities were opened. The heart was punctured on the right side, and PBS supplemented with 137 mM NaCl, 2.7 mM KCl, 18 mM KH₂PO₄, and 100 mM Na₂HPO₄ (Sigma–Aldrich) at room temperature was injected to the left side of the heart for perfusion. Successful perfusion was confirmed by liver decoloration. Organs were excised and stored in ice-cold DMEM supplemented with 10% FCS, penicillin/streptomycin, 2 mM L-glutamine, and 1% MEM non-essential amino acids solution (Thermo Fisher Scientific).

The fresh human ATC sample, subcutaneous tumors of NSG mice, and thyroids of Thyro-DT mice were washed with PBS, cut into small pieces, and enzymatically digested on a shaker for 60 minutes at 37°C in 5 mL RPMI-1640 supplemented with 1 mL 0.25% trypsin and 1 ml (0.1 g/mL) collagenase IV (Sigma–Aldrich). After digestion, supernatants were filtered through 70 μm cell strainers and washed with RPMI-1640, and cells were disaggregated by slowly pushing cell suspensions through 20G followed by 24G needles. Then, cells were washed, filtered through 40 μm cell strainers, and re-suspended in RPMI-1640 supplemented with 10% FCS.

ATC cell lines were cultured to 70% confluence and harvested using non-enzymatic cell dissociation buffer (Sigma–Aldrich). All stainings were performed in PBS for 20–30 minutes at 4°C. Details on antibodies, clones, manufacturers, and staining conditions for FACS are listed in Supplementary Table S3. Fixable Viability Dye-eFluor506 (dilution 1:4000) was from eBioscience. For mouse thyroid tumors, stained cells were washed and fixed using the BD Cytoperm/Cytofix kit (BD Biosciences) according to the manufacturer's instructions. Samples were acquired on a BD LSR II flow cytometer (Becton Dickinson) and were analysed using FlowJo software (TreeStar). Samples from NSG mice injected with GFP^hi-expressing 8505C cells were also acquired on an ImageStreamX Mark II imaging flow cytometer and analyzed using IDEAS software (Amnis/EMD Millipore).

Statistical analysis was performed using GraphPad Prism^® v5.0 (GraphPad Software). Data are represented as the mean ± standard error of the mean.

To study the role of CD47 and TAMs in ATC, FFPE tissues were used from a retrospective cohort of 19 patients who underwent surgery at the authors' hospital from 2005 to 2018. In this cohort, the median survival was 3.5 months (range 0.9–61 months). Patients were treated by either surgery alone or a combination of surgery, radiotherapy, and/or chemotherapy, as well as palliative supportive care (Table 1, Supplementary Table S1, and Supplementary Fig. S1). IHC revealed that most ATCs expressed low to moderate levels of CD47 in a cytoplasmic and/or membranous staining pattern (Fig. 1A, C, and E). Flow cytometry (FACS) analysis of a dissociated fresh ATC indicated strong surface CD47 expression in the CD45⁻ cell fraction containing the tumor cells (Supplementary Fig. S2). The pro-phagocytic molecule calreticulin was moderately to strongly expressed in a dot-like cytoplasmic pattern in most tumors analyzed by IHC (Fig. 1A, C, and E).

Previous studies have shown that ATCs are heavily infiltrated by TAMs (44–46). IHC was therefore performed for the macrophage markers CD68 and CD163 on the different ATC cases, which showed a mean macrophage infiltration rate of 17% (CD68) and 23% (CD163), as analyzed by automated digital quantification (Fig. 1B, D, and F and Supplementary Fig. S3). Similar results were obtained by FACS analysis of the dissociated fresh ATC tumor, with 13.5% of total cells being positive for CD68 (Supplementary Fig. S2). CD68 and CD163 percentages correlated moderately with each other; strong correlation was observed between automated quantification and semi-quantitative microscopic analysis for each macrophage marker (Supplementary Fig. S4).

Furthermore, PD-1 was expressed in ATCs on tumor cells, as well as on cells showing the histomorphological features of macrophages. Moreover, in line with previous studies (47–49), many tumors expressed high levels of the PD-1 ligand 1 (PD-L1; Fig. 1B, D, and G).

To address possible histological differences in patients with longer compared to shorter survival, the cohort was split into two groups: those surviving <6 vs. >12 months (p < 0.01). Tumor cell mitoses, CD68⁺ or CD163⁺ macrophages, CD47, calreticulin, or checkpoint molecules were not significantly different in these two groups (Supplementary Fig. S5). Furthermore, in-depth IHC analyses of the three most extreme outliers were performed (shortest- vs. longest-surviving patients with survival of <2 vs. >18 months, respectively). These results showed a trend toward higher percentages of T cells, lower tumor cell mitoses, and lower infiltration of CD15⁺ granulocytes and CD163⁺ macrophages in longer-surviving patients (Supplementary Fig. S6).

In summary, these data indicate that ATCs express the dominant pro-phagocytic molecule calreticulin, as well as the “don't eat me” signal CD47 and the immune checkpoint molecules PD-1 and PD-L1, all of which are currently targeted therapeutically in other types of cancer. In addition, they confirm that ATCs are infiltrated by innate and adaptive immune cells, pointing toward an important role of the immune microenvironment in these tumors.

To investigate the effects of blocking anti-CD47 mAbs on phagocytosis of human ATC cell lines by macrophages in vitro, first the expression of CD47 and calreticulin was analyzed by FACS and IHC. Both CD47 and calreticulin were significantly expressed on the surface of all eight ATC cell lines analyzed (Fig. 2A–C). In addition to membranous staining, IHC revealed weak cytoplasmic CD47 staining in most cell lines (Fig. 2C). Next, human macrophages were co-cultured with CFSE-labeled 8505C ATC cells in the presence of anti-CD47 or isotype control mAb. Inverted fluorescence microscopy of co-cultures (Fig. 2D) revealed that blocking CD47 significantly increased the numbers of macrophages with phagocytosed ATC cells (phagocytosis index, Fig. 2E), as well as the number of phagocytosed ATC cells per macrophage (Fig. 2F). These results were further validated by FACS analysis of the co-cultures (Fig. 2G and H). Thus, blocking CD47 promotes the phagocytosis of human ATC cells by macrophages.

To validate the in vitro phagocytosis assays in vivo, xenotransplantation studies were performed by injecting 8505C cells subcutaneously into the flanks of NOD.Cg-Prkdc^scid Il2rg^tm1WjI/SzJ (NSG) mice. After three days of engraftment, mice were treated with blocking anti-human CD47 mAb or IgG isotype control twice a week. Blocking CD47 resulted in significantly reduced tumor volumes and weights after six weeks of treatment (Fig. 3A–C). Frequencies of CD45⁺ tumor-infiltrating leukocytes and CD11b⁺ F4/80⁺ double-positive (DP) macrophages were significantly higher in anti-CD47-treated mice (Fig. 3D and E). Interestingly, the expression of CD11b integrin and F4/80 was also significantly increased in TAMs from anti-CD47-treated mice compared to controls (Fig. 3F and G), suggesting that blocking CD47 activates macrophage function. Importantly, the lack of signal on CD45⁻ tumor cells upon CD47 staining using the same mAb clone as for treatment indicated that the CD47 blockade was effective in vivo (Fig. 3H).

To investigate in more detail whether anti-CD47 treatment indeed improves phagocytosis by the accumulated TAMs, stably GFP-expressing 8505C cells were xenotransplanted into NSG mice. Tumor volumes and weights (Fig. 3I and J), the frequency of DP macrophages in tumors (Fig. 3K), and the blocking of CD47 on GFP-expressing 8505C cells (Fig. 3L) were comparable to conditions using parental 8505C cells. Interestingly, a higher PD-1 expression was observed on DP macrophages from anti-CD47-treated mice (Fig. 3M). More importantly, the frequency of GFP-expressing DP macrophages was significantly higher in anti-CD47-treated mice compared to controls, as analyzed by FACS (Fig. 3N). This was confirmed by ImageStream^® analysis showing GFP⁺ cellular debris inside DP macrophages (Fig. 3O). In addition, tumor weights in both experiments inversely correlated with the frequency of DP macrophages in tumors (Fig. 3P), indicating that tumor weight and volume in xenotransplanted mice were substantially affected by the presence of infiltrating macrophages.

In summary, these results indicate that anti-CD47 treatment promotes the accumulation of activated TAMs and improves their phagocytic function in ATC tumors in vivo.

To study the effects of anti-CD47 treatment in a more clinically relevant setting, we made use of the Braf^CA/+; Pik3ca^Lat/+; ThyroglobulinCre^ERT2 (Thyro-DT) mouse model (43). Thyro-DT mice start developing fatal high-grade thyroid carcinomas between three and six months after transgene induction with tamoxifen. First, the expression of CD47 and macrophage markers was analyzed in FFPE thyroid tissues from tumor-bearing Thyro-DT mice. Neoplastic epithelial cells showed intermediate to strong expression of CD47 (Fig. 4A). Staining for CD68 and F4/80 revealed numerous TAMs in the neoplastic epithelium and tumor stroma (Fig. 4A). Three months after tamoxifen induction, tumor-bearing mice were divided into two groups with similar tumor sizes, as measured by ultrasound. Mice were treated with either anti-CD47 mAb or antibody dilution buffer, respectively, and subsequent tumor growth was repeatedly measured by ultrasound. During the 10-week treatment period, mouse tumors nearly doubled in size. Yet, no significant difference in tumor size was observed between the two groups (data not shown). However, FACS analysis of spleens revealed that anti-CD47 treatment had induced significantly increased frequencies of CD8⁺ T cells and a trend toward higher frequencies of DP macrophages. Proportions of splenic CD4⁺ T cells remained comparable between the two experimental groups (Fig. 4B–D). Interestingly, the expression of CD80 on DP macrophages in the spleen was significantly increased after CD47 blockade (Fig. 4E). In thyroids, there was a trend toward higher frequencies of total tumor-infiltrating CD45⁺ cells, and the frequency of CD11b⁺ macrophages was significantly increased (Fig. 4F and G). FACS analysis using anti-mouse immunoglobulin revealed that the therapeutic CD47 mAb had effectively bound on the surface of CD45⁻ EpCAM⁺ tumor cells (data not shown).

These data indicate that murine ATCs are infiltrated by macrophages in vivo and suggest that CD47 blockade modulates the composition of the tumor immune infiltrate as well as systemic immunity in transgenic ATC-bearing mice.