Neoadjuvant Radiation Therapy and Surgery Improves Metastasis-Free Survival over Surgery Alone in a Primary Mouse Model of Soft Tissue Sarcoma

This article is featured in Highlights of This Issue, p. 1

Soft tissue sarcomas (STS) represent a diverse group of mesenchymal malignancies with over 70 different subtypes that account for approximately 1% of cancer in the United States (1). In patients with large and high-grade sarcomas, radiotherapy plays a critical role in improving local control (2, 3), but approximately 50% of patients develop lung metastases and succumb to the disease within 18 months (4). Therefore, alternative therapeutic approaches are urgently needed to reduce STS metastases and thereby improve overall survival. Immune checkpoint inhibitors (ICI), such as antibodies against cytotoxic T-lymphocyte antigen-4 (CTLA-4) and programmed cell death protein-1 (PD-1), reinvigorated the field of immuno-oncology and showed promising results in a subset of melanoma and non–small cell lung carcinoma patients (5–7). More than 3,000 clinical trials are currently evaluating the effect of T-cell–targeted immunomodulators as first or second-line treatment for 50 cancer types (8). The recently concluded SARC028 trial (NCT02301039) showed activity of the anti–PD-1 (α-PD-1) antibody pembrolizumab in a subset of patients with metastatic undifferentiated pleomorphic sarcoma (UPS) or dedifferentiated/pleomorphic liposarcoma (LPS; ref. 9). In addition, studies have suggested that radiotherapy can augment antitumor response by inducing the production of immunostimulatory cytokines and chemokines, increasing T-cell priming through the release of damage-associated molecular patterns, upregulating major histocompatibility complex-I, and exposing novel tumor antigens (10). Furthermore, preclinical and clinical studies suggest that radiotherapy has the potential to act synergistically with immune checkpoint blockade to enhance antitumor immune response outside of the radiation field, that is, an abscopal effect (11–13). To determine whether the addition of α-PD-1 therapy to radiotherapy and surgery improves disease-free survival for high-risk STS, we are currently enrolling subjects in phase II clinical trial SU2C-SARC032 (NCT03092323), in which patients with stage III UPS or LPS are randomized to receive neoadjuvant radiotherapy and surgical resection, with or without neoadjuvant and adjuvant pembrolizumab (Supplementary Fig. S1).

SU2C-SARC032 presented the opportunity to implement a co-clinical trial that aims to: (i) construct robust preclinical imaging protocols with MRI and micro-CT; and (ii) employ these imaging modalities to assess the efficacy of immunotherapy combined with preoperative radiotherapy to reduce local recurrence and metastatic disease in a carcinogen-induced and genetically engineered autochthonous mouse model of STS that resembles human UPS (14, 15). MRI is employed to evaluate the primary tumor site. The preclinical MRI imaging protocol and performance for the optimized extremity MRI in mice with sarcomas have been reported previously (16, 17). Serial CT imaging of the lungs of mice (16, 17) mimics a critical metric of therapeutic response in patients on SU2C-SARC032 because the lungs are the most common site of metastasis for patients with UPS or LPS. Here, we focus on the implementation and results of the preclinical arm of the co-clinical trial, which demonstrates that the combination of α-PD-1 treatment with radiotherapy and surgery improves disease-free survival compared with surgery alone by improving local control. Furthermore, neoadjuvant radiotherapy improves local recurrence-free and metastasis-free survival compared with surgery with or without α-PD-1 treatment.

All animal procedures were approved by Duke University Institutional Care and Use Committee (IACUC) and adhere to the NIH Guide for the Care and Use of Laboratory Animals. Rag2^tm1.1Flv (Rag2^−/−) and 129/SvJae wild-type mice were obtained from The Jackson Laboratory and bred at Duke University (Durham, NC). Trp53^fl/fl (18) mice were obtained from Anton Berns at the Netherlands Cancer Institute and were bred at Duke University. All animals were kept in individually ventilated cages in a room with a 12-hour light:dark cycle and constant temperature (72 ± 2ºF). All studies were performed using age-matched male and female littermate controls.

Primary p53/MCA sarcomas were induced in mice with different genetic backgrounds to implement the preclinical arm of the co-clinical trial and to test hypotheses related to the control of metastases in the p53/MCA model. Primary p53/MCA sarcomas were induced in Rag2^+/−, Rag2^−/−, and 129/SvJae wild-type mice between the age of 6 and 10 weeks by intramuscular injection of adenovirus expressing Cas9 endonuclease and sgRNA targeting Trp53 (Adeno-sgp53-Cas9; Viraquest; ref. 19) followed by intramuscular injection of 3-methylcholenthrene (MCA) into the gastrocnemius muscle as described previously (14). Rag2^−/− and Rag2^+/− mice were used to evaluate the role of lymphocytes in p53/MCA-induced sarcoma metastasis. Rag2^−/− mice lack mature B and T cells, while immunocompetent Rag2^+/− mice served as littermate controls for the experiment. After sarcomas reached 75 to 150 mm³, 129/SvJae mice were randomized into four treatment groups: isotype control antibody (ISO), anti–PD-1 antibody (α-PD-1), ISO + 20 Gy (ISO + RT), or α-PD-1 + 20 Gy (α-PD-1 + RT; Fig. 1A). α-PD-1 antibody (Bio X Cell, Cat # BE0146) and isotype control antibody (Bio X Cell, Cat # BE0086) were administered by intraperitoneal injection of 200 μL per dose at 1 mg/mL diluted in InVivoPure dilution buffer (Bio X Cell) on day 0 (preamputation) and day 7 (postamputation). Sarcomas were irradiated with 20 Gy (single fraction) on day 0 with parallel-opposed anterior and posterior fields using the Precision Xrad 225Cx small animal image-guided irradiator with mice anesthetized with continuous 2% isoflurane at a rate of 2 L oxygen/minute (20). The irradiation field was centered on the tumor via fluoroscopy with 40 kilovolt peak (kVp), 2.5 mA X-rays using a 0.3-mm copper filter, followed by irradiation with an average dose rate of 300 cGy/minute prescribed to midplane at treatment isocenter with 225 kVp, 13 mA X-rays using a 0.3 mm copper filter and a collimator with a 40 × 40 mm² radiation field. Amputation of tumor-bearing hind limb was performed on day 3 as described previously (21–23). Mice were followed up to 7 months with micro-CT imaging performed at 3 months and 6 to 7 months postamputation. For mice that developed local recurrence or otherwise met a humane endpoint before 6 to 7 months postamputation, micro-CT was performed before euthanasia.

Primary p53/MCA sarcomas were induced in Trp53^fl/fl mice using intramuscular injection of adenovirus expressing Cre recombinase (Adeno-Cre; University of Iowa Viral Vector Core) followed by injection of MCA (24). Trp53^fl/fl mice were randomized into two groups (ISO + RT or α-PD-1 + RT) when sarcomas reached 75 to 125 mm³. α-PD-1 antibody (Bio X Cell, BE0146) or isotype control antibody (Bio X Cell, BE0086) treatment was administered intraperitoneally as above on day 0 (preamputation), day 7 (day of amputation), and day 14 (postamputation). Sarcomas were irradiated with 20 Gy (single fraction) on day 0 without image guidance. Mice anesthetized with continuous 2% isoflurane at a rate of 2 L oxygen/minute were positioned under lead shielding that exposed only the tumor-bearing hind limb. Irradiation was performed 50 cm from the radiation source with a dose rate of 204 cGy/minute with 320 kVp X-rays, using 12.5mA and a filter consisting of 2.5 mm Al and 0.1 mm Cu (X-RAD 320 Biological Irradiator, Precision X-ray). Micro-CT scans were performed every three weeks postamputation until the end of the study.

All micro-CT imaging was performed using a micro-CT system developed in-house (25). Animals were scanned while breathing freely under anesthesia using 2% to 3% isoflurane delivered by nose-cone. A pneumatic pillow was positioned on the animal's thorax and connected to a pressure transducer to monitor respiratory motion and inform gating. Body temperature was maintained with heat lamps and a feedback controller. The X-ray source parameters were 80 kVp and 40 mA with 10 ms exposures. A total of 360 views were acquired over a full 360° rotation. The acquisition time was 3 to 5 minutes per set depending on respiratory rate and time needed for rotation to the next angle. We used prospective gating to synchronize our acquisition with respiration (16). With prospective gating, the X-ray exposure is triggered once each respiratory cycle in the desired respiratory phase, for example, peak inspiration. The reconstruction resulted in a 63 μm isotropic voxel size using the Feldkamp algorithm (26) followed by bilateral filtration to reduce noise. The micro-CT images were converted to Hounsfield units (HU) and saved in DICOM format. The radiation dose for each micro-CT scan was approximately 0.017 Gy per mouse, which is approximately 294 to 411 times less than the LD50/30 lethal dose (5–7 Gy) in mice (27). We have previously demonstrated that our micro-CT with prospective respiratory gating facilitates sensitive detection of lung nodules down to 1.0 mm³ in size (16).

At the end of the study, mice were sacrificed using an IACUC-approved CO₂ euthanasia technique. Gross necropsy was performed and any abnormal organ/mass, including local tumor recurrence at the amputation site, was collected and fixed in 10% neutral buffered formalin (Ricca, Cat # R319000–20F) overnight. Mouse lungs were inflated with 10% neutral buffered formalin before collection. Tissues were embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H&E) for histologic evaluation. H&E slides were blinded for the treatment group before histologic analysis by a veterinary pathologist (J. Everitt).

Rag2^+/− and Rag2^−/− mice were euthanized when they developed sarcomas measuring 75 to 150 mm³. Whole tumors were harvested and mechanically homogenized. Genomic DNA was extracted from 25 mg of the homogenized tumor using the DNeasy Blood and Tissue Kit (Qiagen, Cat # 69504). The Trp53 barcode region was amplified via PCR with high-fidelity Taq polymerase (Thermo Fisher Scientific, Cat # 12337–016) and primers (ATGCTAAGCAAGTGTTGGGC, GCTGGGGAAGAAACAGGCTA). The PCR product was submitted to Genewiz for Amplicon EZ NGS sequencing with variant detection and enumeration of the barcode region. To avoid counting PCR error-related reads as Trp53 barcodes, unique sequences were defined as Trp53 barcodes if they contributed greater than 1% of the total reads. In all samples, wild-type Trp53 was identified due to the presence of stroma within the tumor; this was removed from the barcode enumeration but used to calculate total reads for each sample.

LSL-Kras^G12D/+; Trp53^fl/fl (KP) (24) and Trp53^fl/fl (14) mice on a 129/SvJae background were used to generate primary sarcomas from which KP and p53/MCA cell lines were made. Sarcomas were induced by intramuscular injection of Adeno-Cre in LSL-Kras^G12D/+; Trp53^fl/fl mice (KP model) or injection of Adeno-Cre and MCA in Trp53^fl/fl mice as above (p53/MCA model). Tumor tissue was excised and digested using a tumor dissociation kit according to the manufacturer's protocol (Miltenyi Biotec, Cat # 130–096–730). Red blood cells were lysed using ACK lysis buffer (Lonza, Cat # BP10–548E), washed with PBS, and filtered using a 70-μm filter to get a single-cell suspension. Subsequently, cells were plated in complete DMEM media containing 4.5 mg/dL glucose, 110 mg/L pyruvate, 1% antibiotic/antimycotic, 10% fetal bovine serum, and 4 mmol/L L-glutamine. Cells were passaged 5 times to deplete all stromal and immune cells before they were used in subsequent experiments. Cell lines were frequently tested for Mycoplasma using MycoAlert Mycoplasma Detection Kit (Lonza, Cat # LT07–318) as per the company's protocol. However, mouse primary sarcoma cell lines were not authenticated using mouse cell-specific short tandem repeat (STR) profiling. For each experiment, cell lines were passaged 3 to 4 times before using them for the experiment and never used beyond passage 10 for any experiments. To generate transplant sarcomas, we injected 100,000 p53/MCA (402230 or 403865, generated from two independent primary tumors) or KP (403460 or 403492, generated from two separate tumors) mouse primary sarcoma cells resuspended in 50 μL of a 1:1 mixture of DMEM (complete media) and Matrigel (Corning, Cat # 354262) into the gastrocnemius muscle. We measured tumors three times a week with a caliper and sacrificed mice when tumors reached 1,500 mm³.

To study the metastatic potential of sarcoma cell lines, we used primary mouse sarcomas cell lines generated in-house, as mentioned previously, and passaged 3 to 4 times before using them for the lung metastasis experiment. Briefly, we injected 10,000 p53/MCA (402230 or 403865) and KP (403460 or 403492) mouse primary sarcoma cells in 100 μL of PBS through an intravenous tail vein in 6 to 8 syngeneic 129/SvJae mice. Animals were sacrificed at 21 days after tail vein injection using CO₂ asphyxiation. Lungs were inflated with 10% neutral buffered formalin, and visible nodules were counted followed by overnight fixation. Lungs were embedded in paraffin, sectioned, and stained with H&E for histological verification of sarcoma metastases by an investigator blinded to tumor model/cell line.

A total of 24 sarcoma samples (two control liver samples, six amputated sarcomas with matched locally recurrent sarcomas from the ISO antibody treatment group, and five amputated sarcomas with matched locally recurrent sarcomas from α-PD-1 antibody + RT treatment group) were used for the whole-exome sequencing (WES). DNA was extracted using a Quick-DNA Miniprep Plus Kit (Zymo Research, cat # D4068) and sent to Novogene Co., Ltd for WES and data analysis. Initial sample purity, integrity, and quantity were measured by Agilent 5400, and samples with OD260/280 ≥ 1.8 were selected for the next step. Next, genomic DNA was randomly fragmented into small pieces (180–280 bp). DNA fragments were end-repaired, A-tailed, and ligated using Illumina adapters followed by DNA fragment PCR amplification, size selection, and purification. Finally, the prepped libraries were hybridized with biotin-labeled probes and the targeted exons were captured by streptavidin magnetic beads. After quality check, enriched captured libraries were pooled and sequenced on the Illumina platform using the PE150 strategy (6 Gb raw data per sample). Raw data quality control, sequence alignment, variant detection, and somatic mutation detection were performed by the bioinformatics group at Novogene Co., Ltd., while statistical analysis was performed by QuantBio LLC.

Read pairs were excluded when one or both of the reads in the pair contained adapter contamination, >10% of the bases were ambiguous (N), or >50% of the bases had Phred score <5. BWA-MEM (v0.7.17; ref. 28) was run with default parameters to align cleaned reads to the GRCm38 (mm10) genome (29). After sorting with samtools (v1.8; ref. 30), PCR duplicates were marked with Picard tools MarkDuplicates (v2.18.9, http://sourceforge.net/projects/picard/). Somatic mutations (SNPs and indels) in tumor samples were called independently with each of the two normal liver samples. muTect (v1.1.4; ref. 31) was used to call somatic single-nucleotide variants (SNV) from bam files utilizing tumor and normal samples. Somatic indels were identified with Strelka (v2.9.4; ref. 32). SNVs and indels were annotated with ANNOVAR (v2015Mar22; ref. 33) and with dbSNP v138 mm10 SNPs. SNPs were included in the analysis if they passed default filtering thresholds. Somatic copy number variants (CNV) were detected utilizing CNVkit (v 0.9.6; ref. 34) with the batch command after constructing a copy number reference utilizing both normal samples. Gene level log₂ copy-number ratios (log₂ CNR) were calculated using the genemetrics command without thresholding or filtering the data. Log₂ CNR normalized to copy-number reference was converted to absolute calls by exponentiating and assuming a normal diploid reference.

For each sample, variants were called independently with each of the separate liver samples. SNPs were retained for analysis if they met the following criteria: (i) identical SNP was identified when run with each of the normal samples; (ii) had an allele frequency of >10%; and (iii) was not annotated as intergenic, intronic, silent, found in RNA, or unknown; and (iv) had a minimum of four reads for the alternate allele. For SNP and CNV analysis, paired t tests were utilized to determine differences in the number of SNPs, copy number gains, and copy number losses between amputation (Amp) and local recurrence (LR) samples. For SNP overlap analysis, mutations were included if they were observed in at least three samples per tumor status/treatment group combination [e.g., Amp (ISO) vs. Amp (LR)]. For SNP pathway analysis, a hypergeometric test was utilized to test for overlap in shared mutations within each tumor status/treatment group in combination with MSigDB Hallmark pathways translated to mouse genes obtained from MSSigDB (35). All statistical analyses were conducted in the R statistical programming environment (v4.1.2). Amp_2221 was excluded from all WES analyses given its low-quality reads and outlier status on principal component analysis. In addition, LR_2221 was excluded from analyses where paired tests were conducted.

All amputated mice were followed for 6 to 7 months for signs of lung metastasis. Metastasis-free survival was defined as the time from amputation to euthanasia/death with histologic confirmation of distant metastases. Local recurrence-free survival was defined as the time from amputation to euthanasia/death with histologic confirmation of local recurrence at the amputation site. Disease-free survival was defined as the time from amputation to euthanasia/death with histologic confirmation of local recurrence and/or distant metastases. Survival curves were estimated after surgery using the Kaplan–Meier method, and pairwise significance was determined by log-rank test with the control arm depending on the experiment: (i) ISO group compared with α-PD-1, ISO + RT, or α-PD-1 + RT for 129/SvJae mice with p53/MCA tumors; (ii) Rag2^+/− mice with p53/MCA tumors compared with immunodeficient Rag2^−/− mice; (iii) ISO + RT group compared with α-PD-1 + RT for Trp53^fl/fl mice with p53/MCA tumors. We performed Student t test to compare time to tumor initiation and the number of tumor-initiating clones between Rag2^+/− and Rag2^−/− mice. We also performed Student t test to compare the number of visible lung nodules found between KP and p53/MCA cell lines. A two-sided P < 0.05 was considered to be statistically significant. All statistical analyses were conducted in SAS 9.4 software.

WES dataset is available for download from a publicly accessible database, Sequence Read Archive (SRA), using BioProject ID # PRJNA873947. Any other data will be available from the corresponding authors upon request.

To mimic the phase II clinical trial SU2C-SARC032 (Supplementary Fig. S1), we generated autochthonous sarcomas with high tumor mutational burden (14, 15) in immunocompetent 129/SvJae wild-type mice using CRISPR/Cas9 technology (Fig. 1A). Micro-CT imaging was performed to detect lung metastasis using our previously published protocol (16), which provides reliable sensitivity to detect lesions ≥1.0 mm³ with respiratory gating. A representative micro-CT image of STS lung metastasis is shown in Fig. 1B and was verified by lung histology (Fig. 1C). In addition to lung metastases, we also detected other lung abnormalities such as inflammation (Supplementary Fig. S2A and S2B). Interestingly, a recent study showed that pneumonitis developed in roughly one-sixth of non–small cell lung cancer patients who received α-PD-1 treatment (36). However, we found that the frequency of lung inflammation was not associated with the treatment group (ISO vs. α-PD-1) in this autochthonous STS mouse model (Supplementary Fig. S2B). Micro-CT images also identified lung infiltrates, nodules, or suspicious opacities that corresponded with abnormalities other than sarcoma metastases, such as lymphoma, pulmonary adenoma, or hemorrhage (Supplementary Fig. S2C).

Compared with mice treated with surgery and ISO, mice treated with surgery and α-PD-1 + RT showed significantly improved local recurrence-free survival (P = 0.035), but not metastasis-free survival (Fig. 1D and E). The improved local recurrence-free survival for mice treated with α-PD-1 + RT and surgery translated into significantly better disease-free survival compared with the surgery and ISO control group (P = 0.005; Fig. 1F). However, we noted that α-PD-1 + RT and surgery was not significantly better in improving local recurrence-free, metastasis-free, and disease-free survival rate than ISO + RT and surgery, which represents the clinical standard of care for patients with large and/or high-grade sarcomas (Supplementary table S1). The rate of metastases in the ISO control group was unexpectedly low (12%, Fig. 1E), as we had previously observed a rate of lung metastasis in the LSL-Kras^G12D/+; Trp53^fl/fl (KP) model (24) of approximately 40% (21–23). The unexpectedly low rate of metastases in the p53/MCA model limited the power to detect significant differences in metastases across the four treatment groups. To evaluate the effect of α-PD-1 and RT separately on outcomes after surgery, we combined the treatment groups as follows: ISO groups (ISO and ISO + RT), α-PD-1 groups (α-PD-1 and α-PD-1 + RT), - RT groups (ISO and α-PD-1), and + RT groups (ISO + RT and α-PD-1 + RT). Mice that received radiation treatment prior to surgery (ISO + RT and α-PD-1 + RT) showed significantly higher local recurrence-free survival (P = 0.043) and metastasis-free survival (P = 0.007) compared with unirradiated treatment groups (ISO and α-PD-1; Fig. 1G and H). Moreover, mice that received radiotherapy prior to surgery (ISO + RT and α-PD-1 + RT) showed significantly higher disease-free survival (P = 0.006) compared with mice in unirradiated treatment groups (ISO and α-PD-1; Fig. 1I). In contrast, the addition of α-PD-1 (α-PD-1 and α-PD-1 + RT) did not improve local recurrence-free survival, metastasis-free survival, or disease-free survival over isotype control antibody (ISO and ISO + RT; Supplementary Fig. S3A–S3C). These results indicate that in the p53/MCA model of STS, radiotherapy to the tumor-bearing hind limb may contribute to reducing metastases by decreasing local recurrence. Given the low rate of lung metastases in the p53/MCA autochthonous mouse model, we lack sufficient power to determine whether neoadjuvant and adjuvant α-PD-1 therapy with or without radiation therapy impacts the rate of metastases.

To identify possible genomic differences between primary and locally recurrent sarcomas, we performed WES of amputated (Amp) and matched locally recurrent (LR) tumors from mice treated with ISO antibody or α-PD-1 + RT. We analyzed WES of matched amputated and locally recurrent tumors to identify somatic mutations and copy number alterations within each tumor compared with two control liver samples. Interestingly, we found that LR tumors from ISO and α-PD-1 + RT treatment groups had significantly higher single-nucleotide polymorphisms (SNP) and copy number alterations (CNA) compared with matched Amp tumors from ISO and α-PD-1 + RT treatment groups (Fig. 2A and B). Because Amp and LR tumors from both treatment groups were sequenced at similar depth, it is possible that lower clonality in LR tumors was responsible for the detection of more mutational events compared with Amp tumors. Next, we performed a hypergeometric test to identify significant overlaps between genes mutated within treatment groups and evaluated frequently mutated hallmark pathways. We found that myogenesis genes were predominantly mutated in LR tumors (regardless of treatment), while Kras pathway genes were predominantly mutated in Amp and LR tumors from the ISO treatment group (Fig. 2C). Interestingly, we noted that epithelial–mesenchymal transition (EMT) genes were frequently mutated in all tumors (Fig. 2C), possibly contributing to the low rate of metastasis in mice with high mutational burden p53/MCA-induced sarcomas.

One difference between the KP sarcoma model with rates of lung metastasis of approximately 40% (21–23) and the p53/MCA sarcoma model with a much lower rate of metastasis (12%, Fig. 1C) is that the p53/MCA sarcomas described above were generated with CRISPR/Cas9 technology, which could potentially activate the host immune system (37). To investigate whether the low rate of lung metastasis was a consequence of initiating sarcomas with Cas9, we utilized Cre-LoxP technology to generate a high-mutational burden p53/MCA sarcomas in Trp53^fl/fl mice. We induced sarcomas by injecting the gastrocnemius muscle with an adenovirus expressing Cre recombinase and then with MCA. We randomized tumor-bearing mice into two treatment groups prior to surgery: isotype control antibody + 20 Gy (ISO + RT) or anti–PD-1 antibody + 20 Gy (α-PD-1 + RT; Supplementary Fig. S4A and S4B). As in the CRISPR/Cas9–induced p53/MCA model (Fig. 1C), we observed a low rate of metastases in both treatment groups, suggesting that initiating tumors with CRISPR/Cas9 did not contribute to the low rate of lung metastasis (Supplementary Fig. S4C). Given the low rate of metastases in the Cre/loxP–induced p53/MCA model, we were underpowered to determine if metastasis-free survival was different between the α-PD-1 + RT and ISO + RT groups.

Clinical studies have shown that high tumor mutational load is associated with a better response to immune checkpoint blockade therapies and consequently improved overall survival (38, 39). Previously, we have shown that the p53/MCA murine STS model exhibits a high mutational burden, leading to the expression of neoantigens (14, 15). Novel neoantigen expression could make p53/MCA sarcoma cells highly immunogenic and thus susceptible to immune-mediated elimination, reducing the development of lung metastases. Therefore, we hypothesized that immunosurveillance by lymphocytes in immunocompetent 129/SvJae wild-type mice contributed to a low rate of lung metastases (12%) through the elimination of immunogenic clones with the potential to form metastases. To test this hypothesis, we used CRISPR/Cas9 technology (19) to induce p53/MCA sarcomas (15) in lymphocyte-deficient Rag2^−/− mice and littermate control immunocompetent Rag2^+/− mice (Fig. 3A). We have previously shown that primary p53/MCA sarcomas initiated in Rag2^−/− mice have approximately a 2-fold higher number of nonsynonymous mutations and a higher level of neoantigen expression compared to p53/MCA sarcomas initiated in Rag2^+/− mice (15). Time to tumor initiation did not differ between Rag2^+/− versus Rag2^−/− mice (Fig. 3B). When sarcomas reached 75 to 125 mm³, we amputated the tumor-bearing hind limb and observed mice for up to 7 months for signs of lung metastases. CRISPR-barcoding was performed to determine the number of tumor-initiating clones from the resected tumor. The number of tumor-initiating clones did not differ between Rag2^+/− versus Rag2^−/− mice, suggesting that mature B and T cells were unable to eliminate tumor-initiating clones from high-mutational burden p53/MCA sarcomas (Fig. 3C). Next, we compared local recurrence-free and metastasis-free survival after surgery alone in Rag2^+/− and Rag2^−/− mice. We found that Rag2^−/− mice had similar local recurrence-free survival (Fig. 3D) and metastasis-free survival (Fig. 3E) compared with Rag2^+/− mice. These results indicate that the presence of mature lymphocytes does not affect local recurrence-free and metastasis-free survival of mice bearing high-mutational load p53/MCA sarcomas. Thus, the low rate of metastasis in the p53/MCA-induced STS mouse model is not secondary to lymphocyte-mediated elimination of tumor cells with metastatic potential.

Others have similarly observed the low metastatic potential for MCA-induced sarcomas in mice (40). To determine whether the low metastatic potential of p53/MCA sarcomas relative to KP sarcomas is due to tumor cell-intrinsic factors, we generated STS cell lines from p53/MCA (Trp53^fl/fl + MCA) and KP (LSL-Kras^G12D/+; Trp53^fl/fl) mouse models using adenovirus expressing Cre recombinase. We transplanted MCA-induced (p53/MCA: 402230 and 403865) and non–MCA-induced (KP: 403460 and 403492) STS cell lines orthotopically into the hind limb of syngeneic mice. While transplantation of both MCA-induced and non–MCA-induced STS cell lines led to tumor formation, the p53/MCA cell lines had a slightly lower penetrance and tumor growth rate (Fig. 4A and B) compared with KP cell lines (Fig. 4C and D). Next, we performed intravenous tail vein injection of KP or p53/MCA STS cells and sacrificed mice 21 days later to determine the metastatic potential of these cell lines. KP cell lines had a significantly higher potential to metastasize to the lung compared with p53/MCA cell lines, as evidenced by the quantification of visible lung nodules at the time of euthanasia (Fig. 4E, P < 0.0001). Histologic analysis of lungs from mice with tail vein injection of p53/MCA cell lines showed no metastases, whereas sarcoma metastases were confirmed in the lungs of all mice with tail vein injection of KP cell lines (Fig. 4F). These results suggest that the low rate of metastases in the p53/MCA model is due to tumor cell-intrinsic processes, potentially related to mutations induced by MCA.

The current ongoing phase II clinical trial SU2C-SARC032 (NCT03092323) provided a unique opportunity to perform a co-clinical trial using a genetically engineered mouse model of STS. The goal of the co-clinical trial was (i) to optimize and standardize preclinical imaging techniques, such as micro-MRI and micro-CT, and (ii) to test the efficacy of neoadjuvant and adjuvant α-PD-1 treatment in combination with neoadjuvant radiotherapy and surgical resection. With prospective respiratory gating, metastatic lung nodules as small as 1.0 mm³ were detectable by micro-CT, but histologic verification was required as imaging could not differentiate a lung nodule representing sarcoma metastasis from lung inflammation or lung adenoma. This is a clear limitation of anatomical CT imaging due to the low contrast capabilities of this X-ray–based modality. Nevertheless, in vivo micro-CT is still a useful screening method to survey for the development of lung metastases using low-dose scans because the same mice can be imaged at multiple time points rather than increasing the number of mice to perform histology at multiple time points. To address the challenge of detecting lung nodules in a large number of scans from mice, we have also recently implemented a deep learning method (41).

In this preclinical study, we found that treatment with neoadjuvant radiotherapy with or without α-PD-1 therapy significantly improved local recurrence-free survival and metastasis-free survival compared with surgery alone. Because the rate of metastasis was unexpectedly low, we were unable to determine the impact of adding neoadjuvant and adjuvant α-PD-1 treatment to neoadjuvant radiotherapy and surgery on metastasis-free survival. In the clinical management of STS, radiotherapy plays a crucial role to improve local control for patients with high-risk STS, and it also plays an increasing role in nonresectable and oligometastatic disease (42, 43). Similarly, our preclinical study in mice revealed that radiotherapy decreased the rate of local recurrence (Fig. 1E) and remarkably also decreased the rate of metastases (Fig. 1F). Although two randomized clinical trials (44, 45) have not shown a statistically significant improvement in overall survival with the addition of radiotherapy to surgery for STS, they were relatively small studies that were not powered for an overall survival endpoint. Larger database studies (46, 47) have suggested that neoadjuvant or adjuvant radiotherapy is associated with significantly improved overall survival. Although these studies are limited by their retrospective nature and lack of data on cause-specific or metastasis-free survival, our preclinical results showing improved local control and reduced metastases in mice receiving radiotherapy support these clinical data, suggesting a survival benefit with radiotherapy for patients with STS. Although limited by the low number of metastatic events (particularly in mice receiving radiotherapy), the available data from this study suggest that α-PD-1 treatment lacks an abscopal effect in this primary p53/MCA model of STS.

Our previously published data revealed that a single fraction of 20 Gy radiotherapy dramatically reshaped the distribution of myeloid cells compared with unirradiated p53/MCA–induced sarcomas. Radiotherapy significantly increases macrophage infiltration in the TME, and these infiltrating macrophages express high levels of Ccl2 and Ly6C (consistent with active phagocytosis), Fcgr2B, Fcgr1, and Fcgr3 (consistent with antibody-dependent cell-mediated cytotoxicity), and B2m, Ctsl, Ctsd, Ctsb, and Ctss (consistent with antigen/protein processing; ref. 15). In addition, radiotherapy affects the expression of transcription factors responsible for the increased expression of interferon response genes, such as Mx1, Ifit1, Ifit2, Ifit3, and Ifit3b. Thus, radiotherapy treatment repolarizes myeloid cells via activation of type I and II interferon response pathways. Studies have shown that highly inflamed TME caused by radiotherapy correlates with a better response to immune checkpoint blockade therapy (48). Furthermore, WES data from the p53/MCA model revealed frequent mutations in genes involved in the EMT pathway, regardless of treatment status. Previous studies have shown that mutations in EMT genes reduce circulating cancer cells’ ability to extravasate from blood vessels to colonize distant sites, thus reducing their metastasizing potential (49). In future studies, we will utilize single-cell clones generated from this model system to work towards identifying specific genes regulating sarcoma metastasis.

We recently reported an in-depth characterization of the p53/MCA model (15), which included WES, bulk RNA sequencing (RNA-Seq), and single-cell analysis of the CD45⁺ hematopoietic cells using single-cell RNA-Seq and mass cytometry. CIBERSORTx analysis of the RNA-Seq data demonstrated that the immune infiltrates in the autochthonous p53/MCA sarcomas resemble most human STS. However, a subset of human sarcomas has an immune-enriched microenvironment (termed sarcoma immune class E), which may be a biomarker for response to α-PD-1 immunotherapy (50). We previously showed that injection of p53/MCA cell lines into the gastrocnemius muscle of syngeneic mice creates an inflamed tumor immune microenvironment that resembles human STS with sarcoma immune class E and these transplanted mouse p53/MCA sarcomas show the synergy between radiotherapy and α-PD-1 therapy (15). Therefore, it is conceivable that in the subset of human STS with an immune-enriched microenvironment, the combination of radiotherapy and α-PD-1 may lead to a systemic immune response that decreases the rate of lung metastases. We will investigate this possibility in planned correlative studies from SU2C-SARC032.

In the high tumor mutational burden p53/MCA primary sarcomas, mature lymphocytes did not play a role in regulating metastasis by immunosurveillance. These results are consistent with other studies in the MCA-induced sarcomas (51, 52). Our results from tail vein injection experiments support the notion that MCA-driven high tumor mutation burden STS have an impaired ability to colonize the lung, possibly due to intrinsic rather than extrinsic factors. Many preclinical studies testing the impact of immune checkpoint inhibitors (ICI) have been performed using transplant tumor models in syngeneic mice, which can create an inflamed tumor microenvironment (15, 53) and may overestimate the response rate of ICIs in the clinic. Although these transplant tumor models may mimic a subset of immune-enriched human cancers (15), a primary mouse model may better reflect the response of most human cancers to ICIs because autochthonous tumors grow under constant immune surveillance and these primary tumors also model intra- and inter-tumor heterogeneity. Thus, we selected the p53/MCA autochthonous sarcoma model to study the impact of neoadjuvant and adjuvant α-PD-1 treatment on metastasis-free survival. Unfortunately, we were underpowered for this endpoint despite a robust sample size of an average of 70 mice per treatment group because of the low rate of metastasis in the p53/MCA sarcoma model (∼12%) compared with high-risk human STS patients (∼50%; ref. 4). This low rate of metastasis underpowered the preclinical study to test whether or not adding α-PD-1 to radiotherapy and surgical resection provides systemic antitumor immune effects to improve metastasis-free survival. It is also possible that CD8⁺ T cells are not responsible for eliminating tumor clones from the circulation or lung microenvironment and thus have no impact on adding α-PD-1 to the metastasis rate. In fact, a previous study reported that natural killer T cells play a critical role in immunosurveillance of MCA-induced sarcomas through interleukin (IL)-12 or α-galactosylceramide (α-GalCer; ref. 54). Thus, it is possible that natural killer T cells play a role in improving metastasis-free survival in MCA-induced sarcomas, which warrants further study.

In summary, we have successfully implemented the preclinical arm of a co-clinical trial of combined radiotherapy and immunotherapy in sarcoma. We find that radiotherapy improves the rate of local recurrence-free survival and metastasis-free survival. However, the addition of α-PD-1 to radiotherapy was unable to improve metastasis-free survival. Our work illustrates the strengths and weaknesses of the p53/MCA sarcoma model. Autochthonous mouse models can complement preclinical transplant models to study important clinical questions in radiotherapy and immunotherapy (55). This study also demonstrates the successful incorporation of optimized small animal imaging such as micro-CT into a co-clinical trial. By establishing clinically driven preclinical imaging methods to serve in a co-clinical trial, we have created a pipeline that reduces the gap between preclinical and clinical studies of sarcoma therapy.

Y.M. Mowery reports grants from Merck and SU2C, and nonfinancial support from Bayer outside the submitted work. J.E. Himes reports grants from NIH during the conduct of the study. D.G. Kirsch reports grants from XRAD Therapeutics, Bristol-Myers Squibb, Varian Medical Systems, and Eli Lilly during the conduct of the study, personal fees from Lumicell outside the submitted work, as well as a patent for Lumicell issued, licensed, and with royalties paid and a patent for XRAD Therapeutics issued. C.T. Badea reports grants from National Cancer Institute and National Institute of Health during the conduct of the study. No disclosures were reported by the other authors.

R. Patel: Software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. Y.M. Mowery: Conceptualization, resources, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. Y. Qi: Investigation, writing–review and editing. A.M. Bassil: Investigation, writing–review and editing. M. Holbrook: Software, investigation, writing–review and editing. E.S. Xu: Investigation, writing–review and editing. C.S. Hong: Investigation, writing–review and editing. J.E. Himes: Investigation. N.T. Williams: Investigation. J. Everitt: Data curation, supervision, investigation, writing–review and editing. Y. Ma: Investigation. L. Luo: Investigation. S.R. Selitsky: Data curation, methodology, writing–review and editing. J.L. Modliszewski: Data curation, methodology, writing–review and editing. J. Gao: Formal analysis, writing–review and editing. S.-H. Jung: Formal analysis, writing–review and editing. D.G. Kirsch: Conceptualization, supervision, validation, investigation, visualization, methodology, writing–review and editing. C.T. Badea: Conceptualization, resources, data curation, software, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing.

We are grateful to all our funders who provided valuable resources to complete the study. This work was funded by U24 CA220245 (Y.M. Mowery and C.T. Badea), RF1AG070149 (C.T. Badea), ASCO Young Investigator Award (Y.M. Mowery), RSNA Research Resident/Fellow Grant (Y.M. Mowery), Sarcoma Alliance for Research through Collaboration (SARC) SPORE 5U54CA168512 (D.G. Kirsch and Y.M. Mowery), R35CA197616 (D.G. Kirsch), a grant from Emerson Collective (D.G. Kirsch), and the Duke Cancer Center Support Grant (P30CA14236).

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.