Membrane-Anchored and Tumor-Targeted IL12 (attIL12)-PBMC Therapy for Osteosarcoma

Osteosarcoma is a rare malignant tumor that arises primarily in children and adolescents (1). Even with the combination of surgery and adjuvant chemotherapy, the overall 5-year survival rate is about 20% for patients with advanced osteosarcoma, a rate that has remained virtually unchanged over the past 30 years (2). Therefore, novel therapeutic strategies are urgently needed to treat advanced osteosarcoma.

In light of the emerging in-depth understanding of the molecular mechanisms of osteosarcoma, numerous molecularly targeted inhibitors and chimeric antigen receptor (CAR) T-cell therapies have been investigated in clinical trials (3, 4). However, the outcomes of antigen-targeted therapy for osteosarcoma have not been impressive. For example, no objective response was observed in a phase II clinical trial of cixutumumab, an inhibitor of insulin-like growth factor (3), in osteosarcoma patients. Similarly, ERBB2 (HER2) inhibitor treatment did not lead to significantly different outcomes in patients with ERBB2-positive and ERBB2-negtive osteosarcoma (3). In fact, ERBB2-targeted CART-cell treatment induced objective responses in only 3 of 16 patients with ERBB2-positive osteosarcoma (4). The important reason for these unsatisfactory responses to targeted treatments is the heterogeneity of osteosarcoma tumors, which consist of multiple cell lineages and therefore cannot be cured by single-target-based approaches. To address the challenges posed by tumor heterogeneity, CART-cell approaches targeting multiple antigens have been investigated (5–7), but these approaches were frequently associated with toxic effects (8–11). Finally, the long duration of T-cell expansion in vitro not only is expensive but also may miss the optimal time window for intervention. Thus, alternative approaches are urgently needed to make progress in treating osteosarcoma.

In this study, we determined the extent to which a cell-membrane-anchored tumor-targeted IL12 (attIL12) could arm unexpanded peripheral blood mononuclear cells (PBMC) to treat mice bearing patient-derived xenograft (PDX) tumors or metastatic osteosarcoma tumors in the lungs. This approach does not require the costly and time-consuming steps of T-cell isolation and expansion (12, 13), unlike a recently described attIL12 T-cell therapy (14). Moreover, the attIL12-modified PBMC treatment yielded no observable toxic side-effect. Notably, we discovered that attIL12-modified PBMCs impeded osteosarcoma tumor growth by triggering terminal differentiation of cancer into bone-like cells instead of by direct killing of tumor cells. Single-cell RNA sequencing analysis supported this finding of osteosarcoma cell differentiation into a bone-like phenotype at the molecular level and uncovered a bone-cell gene signature.

Six- to 8-week-old NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice of both sexes were purchased from The Jackson Laboratory. The mouse care and handling procedures were approved by the Institutional Animal Care and Use Committee of The University of Texas MD Anderson Cancer Center.

Patient-derived OS1, OS2, and OS17 osteosarcoma tumor cells (generously provided by Dr. Richard Gorlick, the Pediatric Preclinical Testing Consortium, The University of Texas MD Anderson Cancer Center, Houston, TX) were implanted subcutaneously into C.B-Igh-1^b/IcrTac-Prkdc scid mice (Taconic). Tumors were measured with calipers twice weekly after implantation. Tumor volume was calculated by the formula: V = (π/8) × (a × b²), where V, tumor volume in cubic centimeters; a, maximum tumor diameter, and b, diameter at 90° to a. Survival curves were generated by the Kaplan–Meier method. attIL12 PBMCs (2.5 × 10⁶ cells/mouse) or control PBMCs (2.5 × 10⁶ cells/mouse) were intravenously injected into the mice 3 times at intervals of 2 weeks.

To establish the osteosarcoma lung metastasis model, human OSD cells (5 × 10⁵ cells/10 μL/mouse) were injected into the proximal tibia of 6-week-old NSG mice. Tumor growth was monitored by measuring tumor volume as described above. Lung metastases were allowed to develop for 4 weeks after OSD cell injection. The leg into which OSD cells had been injected was amputated 4 weeks after injection. attIL12 PBMCs (2.5 × 10⁶ cells/mouse) or control PBMCs (2.5 × 10⁶cells/mouse) were injected intravenously into the mice 3 times at intervals of 2 weeks.

OSD, OSO, SAOS2, and U2OS human osteosarcoma cells were kindly provided by Dr. Dennis PM Hughes (The University of Texas MD Anderson Cancer Center). OSD, OSO, SAOS2, and U2OS cells were cultured in high-glucose DMEM containing 10% FBS and supplemented with antibiotics, non-essential amino acid solution, and minimum essential medium vitamin mixture and cultured in an incubator maintained at 5% CO₂ and 37°C. Patient-derived OS1, OS2, and OS17 osteosarcoma cells were dissociated from osteosarcoma xenograft tumors and cultured in high-glucose DMEM supplemented with 15% FBS and 1% penicillin/streptomycin. Detailed information about OS1, OS2, and OS17 osteosarcoma cell lines has been previously published (15, 16). All current available data on these models can be found at PedcBioPortal: (https://pedcbioportal.kidsfirstdrc.org/study/summary?id=pptc). All the tumor cell lines were characterized by DNA fingerprinting within 6 months of initiating the experiments at MD Anderson Cancer Center's Characterized Cell Line Core Facility and were treated with a Mycoplasma removal agent from Bio-Rad to ensure that cells were free of Mycoplasma before inoculation.

Buffy coats from de-identified healthy blood donors were purchased from the Gulf Coast Regional Blood Center, and their acquisition was approved by the MD Anderson Institutional Review Board. PBMCs were isolated from buffy coat samples via centrifugation over Ficoll-Paque. PBMCs were cultured in medium containing 10% FBS, supplemented with recombinant human IL2 (rhIL-2; 50 U/mL), rhIL-7 (10 ng/mL), and rhIL-15 (5 ng/mL).

As previously described (14), high-titer replication-defective lentiviral vectors were produced and concentrated by the MD Anderson Functional Genomics Core Facility. Briefly, HEK293T human embryonic kidney cells were transfected with pVSV-G (a VSV glycoprotein expression plasmid), pCMV-Gag/Pol/Rev, and a transfer plasmid by using Lipofectamine 2000 (Thermo Fisher Scientific). The viral supernatant was harvested at 48 hours after transfection. Viral particles were concentrated using Lenti-X Concentrator (Takara Bio, Inc.).

As previously described (14), human PBMCs were stimulated by CD3/CD28 Dynabeads (Thermo Fisher Scientific) according to the manufacturer's instructions. On day 2, PBMCs were plated in non-tissue culture–coated 24-well plates. Lentiviral supernatant (Vector Builder) was first spun at 1,500 × g for 2 hours on retronectin (Takara)-coated non-tissue culture–treated plates. PBMCs were then plated and centrifuged at 1,000 × g for 20 minutes and incubated at 37°C. After 24 hours, the medium was changed to 45% RPMI-1640 and 45% Click's medium containing 10% FBS, supplemented with rhIL-2 (50 U/mL), rhIL-7 (10 ng/mL), and rhIL-15 (5 ng/mL).

As previously described (14), frozen tumor sections and organs were sequentially fixed with cold acetone, acetone plus chloroform (1:1), and acetone. Paraffin-embedded sections were deparaffinized and heated in antigen retrieval buffer. Tissue sections were blocked with 3% H₂O₂ in distilled water for 20 minutes and then in blocking buffer (5% normal horse serum and 1% normal goat serum in PBS). Slides were incubated with primary antibodies overnight at 4°C and secondary antibodies for 1 hour at room temperature. For IHC staining, the secondary antibody was biotin conjugated, the sections were treated with ABC reagent (Vector Labs), and the nuclei were counterstained with hematoxylin (Sigma-Aldrich). Tumor sections were mounted with Cytoseal mounting medium (Life Technologies). Quantifications of IHC images were assessed by examining 3 randomly selected low-power fields per slide. For immunofluorescence staining, tumor sections were mounted in an antifade fluorescence mounting medium with 4–6-diamidino-2-phenylindole. Slides were visualized under a Nikon Eclipse Ti fluorescence microscope.

Tumor lysates were collected from treated and control tumor-bearing mice. Tumor lysates were collected from tumors that were lysed with radioimmunoprecipitation assay buffer as described previously (17). Serum was collected from the supernatant of centrifuged mouse whole blood. Culture medium was collected at 1 mL medium/106 cells. The levels of IL12, TNFα, and IFNγ were measured by using ELISA Ready-SET-Go! kits (eBioscience).

As previously described (14), cells were sequentially incubated with primary and secondary antibodies for 30 minutes each at 4°C. Stained cells were analyzed using an Attune acoustic focusing cytometer (Applied Biosystems) or a BD LSR-Fortessa cell analyzer (BD Biosciences). Flow cytometry data were analyzed using the FlowJo software program (FlowJo, LLC).

As previously described (14), tumors were minced into 2-mm fragments, placed in 5 mL of dissociation buffer (RPMI-1640 medium with 100 U/mL collagenase type IV and 100 U/mL DNase I), and incubated at 37°C while shaking at 120 rpm for 30 minutes to 1 hour. The released cells were filtered with 70-μmol/L strainers and centrifuged at 600 × g for 5 minutes, followed by red blood cell lysis. Cells were then resuspended in fluorescence-activated cell sorting solution containing 2% FBS. Single-tumor-cell suspensions were obtained after CD45 depletion using an EasySep Human CD45 Depletion Kit (Stem Cell Technologies).

Tumor cells (1 × 10⁵) were cocultured with transduced PBMCs in 24-well plates at tumor-to-PBMC ratios of 1:5 for 48 hours. PBMCs and tumor cells were identified by staining with anti-human CD45 and GFP, respectively.

As previously described (14), a PDMS microfluidics slide chip (Abnova) was coated with 1 mg/mL streptavidin (Agilent). The chip was then washed and coated with 0.4 mg/mL biotin-vimentin (MD Anderson) for 1 hour. Next, 1.5 × 10⁶ attIL12-PBMCs (labeled with Calcein Green AM; Thermo Fisher Scientific) and 1.5 × 10⁶ control-PBMCs (labeled with CMTPX red cell tracker; Thermo Fisher Scientific) were mixed at a ratio of 1:1, loaded into a spiral chamber (Abnova), and passed through the slide chip at a flow rate of 1.8 mL per hour using the CytoQuest microfluidics pump (Abnova). The slide chip was then imaged on a fluorescent microscope (Keyence). Cells were counted using Keyence BZ-X700 analysis software.

As previously described (18), fresh tumor lesions were stored in RPMI medium and processed on ice within 30 minutes after surgical removal. The specimens were washed with Hanks balanced salt solution 3 times and minced into 1- to 2-mm pieces. Then, mixed enzymes were added according to the protocol provided with the Miltenyi Biotec Tumor Dissociation Kit (Miltenyi Biotec:130-095-929) and the dissociation was performed in a gentleMACS dissociator (Miltenyi Biotec:130-093-235). After digestion, the samples were filtered through 40-μm sterile strainers and centrifuged at 800 × g for 5 minutes. Subsequently, the supernatants were discarded, and the cell pellets were suspended in 1-mL PBS (HyClone). To remove red blood cells, 2 mL of red blood cell lysis buffer (eBioscience:00-4333-57) was added, and cells were incubated at 25°C for 10 minutes. The solution was then centrifuged at 500 × g for 5 minutes and re-suspended in PBS (18). The samples were stained with trypan blue (Sigma) and cell viability was evaluated under a phase-contrast light microscope (Nikon).

The raw unique molecular identifier (UMI) count data were loaded into the R Seurat (version 4.0.0; ref. 19) package for R downstream analysis. Cells with UMI numbers <200 or with over 25% mitochondrial-derived UMI counts were considered low-quality cells and were filtered out. The count matrix was first log-normalized with the scale factor set at 10,000 total genes per cell. The 6 samples were then integrated into a new matrix using the IntegrateData function with the top 2,000 highly variable genes selected by the Seurant function FindVariableGenes with default parameters. Principal component analysis was performed on the integrated data matrix. The top principal components were selected on the basis of the ElbowPlot function, representing at least 80% of the total variances. The main cell clusters were identified with the FindClusters function offered by Seurat, with resolution set at 0.5. The clusters were then visualized with UMAP (uniform manifold approximation and projection) plots.

For each cluster, the marker genes were identified by comparing the cluster with the other clusters using the FindConservedMarkers function. The Seurat Findallmarker function was performed to identify preferentially expressed genes in clusters and differentially expressed genes between control and treatment cells. The enriched pathways and hallmarks were identified by pre-ranked Gene Set Enrichment Analysis using the gene list ranked by log-transformed P values with signs set to positive/negative for a fold change of >1 or <1, respectively.

As previously described (14), the directly measured outcomes were analyzed using 2-sided Student t tests to compare 2 treatment groups or 1-way ANOVA to compare more than 2 treatment groups. The statistical analyses were conducted using GraphPad Prism 8 software. All data values represent 3 replicates and are shown as median ± SEM. Significance was defined as a P value of <0.05.

The single-cell RNA-sequencing data generated in this study are publicly available in Gene Expression Omnibus at GSE206209.

In previous studies, we reported our discovery of a linear peptide (VNTANST) that can target a VNTANST–IL12 protein to tumor cell-surface vimentin (CSV; refs. 20–23), which is expressed on the surfaces of almost all highly malignant tumor cells (21, 24). Our previous studies confirmed accumulation of IL12 in tumors and inhibition of tumor progression in numerous murine tumor models after gene therapy with this tumor-target IL12 (ttIL12), in which a ttIL12-encoding plasmid DNA was administrated directly into mice (20, 22, 23). However, ttIL12 gene therapy via intramuscular injection is only able to eliminate microscopic tumors and is still able to induce a significant amount of toxic cytokines (25).

In the current study, we used attIL12-armed PBMCs, which express ttIL12 on the surface of cell membranes without releasing out of cells, for tumor targeting and treatment. Briefly, we generated the attIL12 construct by first including a trans-membrane domain before the stop codon of the human IL12p35 subunit to anchor IL12 to the cell surface, and then inserting the tumor-targeted peptide-encoding sequence right before the stop codon of the human IL12p40-encoding subunit (Fig. 1A). This attIL12 was packed into a lentivirus that could transduce PBMCs stimulated with anti-CD3/28 beads and was expressed on the PBMCs’ surfaces as determined by flow cytometry (Fig. 1B, left) and immunofluorescence staining (Fig. 1B, right).

To validate the biological function of attIL12-PBMCs, we monitored IL12 and IFNγ secretion in the supernatant of medium over time by ELISA (Fig. 1C and D). We also transduced the PBMCs with a wild-type IL12(wtIL12)-bearing lentivirus to serve as a positive control. Compared with medium conditioned with wtIL12-PBMCs, the detectable level of secretion of IL12 and IFNγ in medium conditioned with attIL12-transduced PBMCs was 5–10-fold lower (Fig. 1C). To investigate whether these low IFNγ-inducing attIL12-PBMCs were still active in vivo, mice bearing OS1 osteosarcoma PDX were infused with attIL12-PBMCs or control-PBMCs. attIL12-PBMCs notably suppressed tumor growth after 3 infusions at an interval of every 2 weeks (Fig. 1D and E). Most of the tumors from mice treated with attIL12-PBMCs became smaller and maintained their small size.

The significant tumor growth inhibition we observed could have been a coincidence because the PBMCs were derived from an individual donor. To exclude this possibility, PBMCs from different donors were armed with attIL12 (Fig. 2A). The attIL12 transduction process and the validation of attIL12 expression on the PBMCs showed consistent results among different donors, with attIL12 expression on approximately 60% of PBMCs (Figs. 1B and 2A). This showed that attIL12 transduction efficacy was consistent in PBMCs from different donors. Next, we assessed the antitumor efficacy of attIL12-PBMCs in 2 osteosarcoma PDX models, OS2 and OS17. The results demonstrated that attIL12-PBMCs significantly impeded both OS2 and OS17 tumor growth and notably prolonged the survival time of treated mice (Fig. 2B–E). At the end point of experiment, mice of the attIL12-PBMC group still had tumor. To boost the clinical relevance of our experiments, we also tested the efficacy of attIL12-PBMCs in the OSD model of osteosarcoma (OSD) lung metastasis (Fig. 2F). As in the subcutaneous PDX tumor model, attIL12-PBMCs dramatically extended the mice's overall survival time over control PBMC treatment (Fig. 2G). On day 65 after tumor-cell inoculation, we randomly selected 3 mice from each group and detected lung metastasis with hematoxylin and eosin (H&E) staining. We found that the area containing lung metastases in the attIL12-PBMC treatment group was significantly smaller than in the control group (Fig. 2H). This result showed that attIL12-PBMCs also significantly inhibited the progression of OSD lung metastasis. Remarkably, these efficacy results were obtained with PBMCs from 3 independent donors.

Toxicity is the key challenge for clinical application of IL12-based therapies because systemic infusion of IL12 protein has yielded lethal consequences (26). Both wtIL12 gene and wtIL12-armed T-cell (including CAR T-cell) therapies are safer than IL12 protein therapy, but the risk of toxicity remains high (27, 28).

Because we have discovered that attIL12-modified PBMCs did not induce toxic cytokine production in vitro, it was plausible to hypothesize that infusion of attIL12-PBMCs would yield only minor toxicity. If confirmed, this would be a crucial finding for the entire field of IL12-based cell therapy. To test this hypothesis, we measured levels of toxic cytokines in peripheral tissue and conducted histologic examination of major organs on day 5 after infusion of attIL2-, wtIL12- or control-PBMCs (Fig. 3A). As expected, inflammatory cells were identified in the livers of mice treated with wtIL12-PBMCs, but not in the hearts, lungs, kidneys, or brains (Fig. 3B). Of note, inflammatory cells were not found in any organs of mice treated with attIL12- or control-PBMCs (Fig. 3B). Consistent with the accumulation of inflammatory cells in the liver, substantial accumulation of IL12, IFNγ, and TNFα was detected in only the livers of wtIL12-PBMC–treated mice, not in any other organs (Fig. 3C). Likewise, neither attIL12-PBMC nor control-PBMC infusion caused any toxic cytokine induction in the liver or other organs. Consistent with the liver-specific inflammation induced by wtIL12, a high level of IL12 was detected in the serum of wtIL12-PBMC–treated mice (Fig. 3D). In contrast with the findings in the livers and blood, significantly higher levels of IL12, IFNγ, and TNFα were detected in tumors of mice that received either wtIL12- or attIL12-PBMC infusions (Fig. 3E). To test our hypothesis further, we examined liver and kidney function by comparing blood chemistry. The results showed that both the alanine aminotransferase and aspartate aminotransferase levels in the blood were significantly higher in mice treated with wtIL12-PBMCs than in either control- or attIL12-PBMC–treated mice (Fig. 3F). These results suggested that toxic effects in the liver can follow wtIL12-PBMC infusion, and that attIL12-PBMC infusion does not induce liver toxicity. Therefore, attIL12-PBMC infusion within the therapeutic window is safe.

To identify the mechanism by which attIL12-PBMCs fail to induce systemic toxicity, we tested the migration and accumulation of CSV-targeted attIL12-PBMCs. We hypothesized that attIL12-PBMCs not only abolished toxic cytokine induction but also promptly accumulated in tumors, thereby avoiding systemic toxic effects. To test this hypothesis in vitro, we developed an assay using a vimentin-coated microfluidic chip to show specific capture of attIL12-PBMCs but not the control-PBMCs; attIL12 contains a vimentin-binding ligand peptide (Fig. 1A). In brief, equal number of control-PBMCs (red CMTPX labeled) and attIL12-PBMCs (green calcein AM labeled) were mixed and loaded onto the vimentin-coated microfluidic chip, and the chip was washed under an automatic microfluidic flow. After binding, unbound cells were washed away, and bound cells were scanned with an automated fluorescence microscope to determine which cell types were bound to vimentin on the chip (Fig. 4A and B). We found that most of the vimentin-bound cells were attIL12-PBMCs (green), and only rarely control-PBMCs (red). This result showed that attIL12-PBMCs could bind vimentin protein on the chip. Next, instead of coating the chip with vimentin, we coated the chip with osteosarcoma tumor cell lines with different percentages of CSV-positive cells. As shown in Fig. 4C, CSV expression in osteosarcoma cell lines ranged from 49.9% to 19.4%. The binding activity was higher in tumor cells with higher CSV expression (Fig. 4D).

In light of these in vitro data, we also tested tumor-targeted migration of attIL12-PBMCs in vivo. We collected tumors from PDX-bearing mice treated with attIL12-PBMCs or control-PBMCs and quantified the infiltrated PBMCs using IHC and flow cytometry. In line with in vitro findings, the population of CD45⁺ cells was significantly higher in tumors from attIL12-PBMC–treated mice than in those from either wtIL12-PBMC– or control-PBMC–treated mice (Fig. 4E and F). Significantly, the IHC staining of several organs revealed that PBMCs were only found in tumors, not in any other organs, in attIL12-PBMC–treated mice. However, wtIL12-PBMCs accumulated in the liver. Thus, we found that attIL12 promotes the tumor-targeted migration of modified PBMCs both in vitro and in vivo. This property may serve as the key mechanism for avoiding toxicity following attIL12-PBMC cell infusion.

To uncover the specific mechanism by which attIL12-PBMCs mediated PDX tumor growth inhibition, we first measured tumor-cell apoptosis in attIL12-PBMC–treated and control-PBMC–treated PDX-bearing mice. Results of TUNEL assays showed that the percentage of apoptotic cells was very low, at 3%–5%, in both the attIL12-PBMC–treated and control-PBMC–treated groups; no significant differences were detected between the groups (Supplementary Fig. S1A and S1B). This result suggested that apoptosis may not be the primary mechanism by which attIL12-PBMC treatment inhibits PDX tumor growth.

To validate this in vivo observation, we also performed in vitro CTL assays to confirm that tumor-cell killing is not the primary mechanism by which CAR T cells inhibit tumor growth (24, 29, 30). To assess the CSV-targeted cell-killing activity of attIL12-PBMCs, we cocultured the CSV high-expressing metastatic osteosarcoma cell line OSD with attIL12-PBMCs or control-PBMCs at a 1:5 ratio (tumor cells to PBMCs) for 48 hours. After coculture, we determined the number of PBMCs and tumor cells by flow cytometry (Fig. 5A). attIL12-PBMCs did not kill osteosarcoma cells more effectively than did control-PBMCs. This result confirmed the in vivo observation.

The lack of direct cell killing prompted us to look for other mechanisms for the tumor shrinkage and stabilization observed following attIL12-PBMC infusion. Unexpectedly, H&E staining revealed apparent osteogenic differentiation in 3 independent osteosarcoma PDX tumor models following attIL12-PBMC infusion (Fig. 5B). This result suggested that attIL12-PBMCs may influence the terminal differentiation of osteoblasts.

To further explore the molecular changes induced by attIL12-PBMCs in osteosarcoma tumor cells, we used single-cell RNA sequencing to analyze differences in gene expression. After initial quality control assessment and doublet removal, we captured between 3,589 and 5,074 cells from the tumor samples, which included 3 OS1 tumor samples in the control-PBMC infusion group and 3 OS1 tumor samples in the attIL12-PBMC infusion group. In all samples, more than 90% of tumor cells were viable (Supplementary Fig. S1C and S1D). Unbiased clustering of the cells with UMAP analyses (18) identified 16 main gene profile clusters in parallel (Fig. 5C and D). Using canonical markers, we identified chondroblastic cells, endothelial cells, myeloid cells, pericytes, osteoclasts, and 12 types of osteoblasts. We called out these 12 types of osteoblasts on the basis of expression of 2 early osteogenic differentiation markers, alkaline phosphatase (ALPL) and integrin binding sialoprotein (IBSP), and 2 late osteogenic differentiation markers, osteopontin (SPP1) and osteocalcin (BGLAP; refs. 31, 32). As shown in Fig. 5E, only 2 clusters differed significantly between the attIL12-PBMC–treated and control groups, clusters 7 and 12. The percentage of cells in cluster 7 in the attIL12-PBMC samples was 9.6% whereas it was 3.3% in the control samples. The percentage of cluster 12 cells in the attIL12 group was 2.3%, whereas it was 0.9% in the control group. Cluster 7 was the cluster with the largest difference between the attIL12-PBMC and control groups.

To provide an additional line of evidence that an increase in the cluster 7 cell population was associated with osteogenic differentiation in tumors from attIL12-PBMC–treated mice, we performed a volcano plot analysis of the differentially expressed genes between cluster 7 and other clusters. The results confirmed induction of osteogenic differentiation genes in cluster 7 (Fig. 5F). This finding demonstrated that cluster 7 was the most differentiated osteoblast cluster. We therefore speculated that programming cells toward a bone-like, terminally differentiated cluster 7 phenotype may be one of the molecular mechanisms by which attIL12-PBMC infusion slows tumor growth. To further validate this working hypothesis, we determined the expression of the markers of late osteogenic differentiation BGLAP and SPP1 (31, 32). Both were notably increased in OS1, OS2, and OS17 PDX tumors following attIL12-PBMC infusion (Fig. 5G–I). These results support the notion that attIL12-PBMC infusion induced terminal differentiation of osteosarcoma cells.

To understand how attIL12-PBMCs promoted the cluster7 osteoblast phenotype in vivo, we assessed the results of our analysis of Kyoto Encyclopedia of Genes and Genomes (KEGG) data, which identified an association between expression of bone differentiation genes and protein levels of IFNγ (Fig. 6A). To validate this expected increase in IFNγ level after attIL12-PBMC infusion, we counted IFNγ and CD45 double-positive cells from tumors of both control-PBMC– and attIL12-PBMC–treated mice using flow cytometry. As expected, attIL12-PBMCs dramatically enhanced the proportion of IFNγ and CD45 double-positive cells in both OS1 and OS2 tumor models (Fig. 6B and C). These data implied that attIL12-PBMC treatment substantially increased IFNγ protein secretion in tumors. To determine whether attIL12-PBMC treatment directly induced terminal differentiation of tumor cells, we cocultured OS1 or OS2 tumor cells with attIL12-PBMCs or control-PBMCs in the presence or absence of a neutralizing anti-IFNγ antibody for 6 days. Indeed, attIL12-PBMC coculture induced a significantly higher percentage of SPP1 and BGLAP double-positive OS1 and OS2 tumor cells compared with coculture with control-PBMCs, and tumor cells cocultured with attIL12-PBMCs showed evidence of induction of terminal differentiation (Fig. 6D and E). More importantly, neutralizing IFNγ completely prevented this terminal differentiation.

To avoid the toxic effects associated with wtIL12 while boosting the tumor-specific antitumor immune response driven by IL12, we discovered ttIL12, an IL12 targeting CSV on tumor cells, which displayed an effective inhibitory effect on a variety of small tumors (>5 mm in diameter). To eliminate larger tumors than ttIL12 can, we generated this second-generation of ttIL12–attIL12, to modify expanded T cells. This attIL12 T-cell therapy has showed antitumor efficacy in both immune-competent (mouse tumors) and PDX (patient tumors) models without causing toxic effects (14). Like other types of T-cell therapy, this attIL12 T-cell therapy requires the high cost T-cell expansion, which also increases the risk of contamination and can fail in some patients. To avoid these challenges, we investigated attIL12-PBMCs for treating tumors. We found that infusion of attIL12-PBMCs inhibited osteosarcoma PDX growth and osteosarcoma lung metastasis (Fig. 2). Interestingly, this inhibition was associated not with direct cell killing or apoptosis induction but with osteogenic differentiation. In other words, attIL12-PBMC treatment transforms tumor cells into bone-like terminally differentiated cells, which slows tumor growth.

Inclusion of a co-stimulatory signal in second-generation CARs has dramatically enhanced the antitumor efficacy of CAR T-cell therapy. Nonetheless, in clinical trials these improvements have not to success in treating large, heterogeneous solid tumors (33–36). To eliminate these tumors, researchers have included cytokines such as IL12 or double co-stimulatory signals in the newest generation of CAR T-cell therapies (37–40). Whether these new CART vectors will improve therapeutic efficacy requires further clinical investigation, but it is clear that the addition of cytokines such as wtIL12 and other co-stimulatory signals increases the risk of toxic effects. Studies have also investigated IL12-modified T cells, dendritic cells, and tumor cells, but in general, wtIL12-modified cell therapy induce toxic levels of cytokines in peripheral tissues, causing acute toxicity (28, 41, 42). Other research showed that the unique homing and surveillance properties of myeloid cells enable the accumulation of genetically engineered IL12-armed myeloid cells in pre-metastatic sites with cytokine levels only temporarily elevated in peripheral tissues (43). In contrast, attIL12-based cell modification diminished even this lowered toxicity risk because the membrane-anchored IL12 does not induce any of the cytokines in peripheral tissues that cause cytokine release syndrome. The toxicity study shown here demonstrated that attIL12-PBMCs were safer than wtIL12-PBMCs not only because attIL12-PBMCs inhibited cytokine release but also because they did not accumulate in the peripheral blood and healthy organs (Figs. 3C–E and 4E and F).

In addition, attIL12-PBMCs bound CSV on tumor cells to boost prompt tumor-specific distribution. In previous studies, we discovered that CSV is a universal marker exclusively expressed on the surface of highly malignant tumor cells, such as sarcoma, breast cancer, and colon cancer cells (21, 24). More importantly, the VNTANST peptide specifically binds CSV on the surface of cancer cells but not healthy cells (44). We also demonstrated that IL12 genetically modified with VNTANST (ttIL12) binds to CSV on tumor cells and delivers IL12 to tumor sites (20). Those prior studies were essential to the design of this safe and effective attIL12 vector for PBMC modification.

The concept of differentiation therapy arose from the observation that cytokines and hormones can promote cell differentiation, thereby irreversibly changing the phenotype of cancer cells (45). It is believed that this terminal differentiation in cancer cells leads to tumor remission (31). The rapid clearance of acute promyelocytic leukemia in response to retinoic acid was a hallmark success of differentiation therapy (46, 47). Interestingly, our results showed that attIL12-PBMCs significantly enhanced osteogenic differentiation in osteosarcoma PDX (Fig. 5A and B). Our single-cell sequencing data support the notion that attIL12-PBMCs significantly boosted terminal osteogenic differentiation (indicated by an increase in the cluster 7 cell population; Fig. 5E). In addition, KEGG analysis of genes that were differentially expressed between cluster 7 and other clusters indicated that genes differentially expressed in osteoblasts are associated with induction of IFNγ and cytokine pathways. In vitro, IFNγ production induced by attIL12-PBMC treatment was the key factor inducing terminal differentiation of primary osteosarcoma cells (Fig. 6D and E). Research on differentiation therapy in osteosarcoma is still relatively rare, so this study opens up new directions and perspectives for future osteosarcoma therapy research.

Limitations of our study should be noted when interpreting and extrapolating our data. attIL12-PBMCs should be tested in other solid cancers. In addition, the detailed mechanism by which attIL12-PBMCs promote terminal differentiation of osteosarcoma cells into bone-like cells needs to be further revealed.

In summary, IL12-based attIL12-PBMC therapy is safe and effective against osteosarcoma. Our goal is to move this treatment into a clinical trial. Owing to the convenience of the attIL12-PBMC production process, we believe it will be feasible.

S. Li reports a patent for 16/339,691 pending. No disclosures were reported by the other authors.

Q. Yang: Resources, data curation, validation, investigation, writing–original draft. J. Hu: Supervision, investigation, methodology. Z. Jia: Investigation, methodology. Q. Wang: Software, methodology. J. Wang: Software, methodology. L.H. Dao: Investigation, methodology. W. Zhang: Validation, investigation. S. Zhang: Investigation, methodology. X. Xia: Formal analysis, methodology. R. Gorlick: Resources, supervision. S. Li: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing.

This study was supported by the National Institutes of Health through grant RO1CA120895. Thanks for the technical support of single cell sequencing from advanced technology genomics core facility of University of Texas MD Anderson Cancer Center which was supported by the CA016672 grant and the NIH 1S10OD024977-01 award.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.