Gene expression and survival data were obtained from the TCGA portal (http://www.cbioportal.org/public-portal/). Copy numbers of the CD8a, IFNγ, CD33, and FOXP3 genes and survival data were obtained for eight cancer types: breast adenocarcinoma, sarcoma, lung adenocarcinoma, liver cancer, head and neck squamous cell carcinoma, colon carcinoma, ovarian carcinoma, and melanoma.

All gene constructs were generated as previously described [19]. The wild-type IL-12 plasmid DNA (pDNA), CCL22 DNA (pCCL22), and CXCL2 DNA (pCXCL2) were obtained from Genscript (New Jersey, USA), and the ttIL-12 pDNA was described previously [19]. The control pDNA (Ctrl) was wtIL-12 with the IL-12 gene deleted.

The 4 T1 murine breast cancer cell line and LM8 murine osteosarcoma cell line were obtained from ATCC (Manassas, VA). Both cell lines were maintained in Dulbecco modified Eagle medium containing 10% fetal bovine serum (Life Technologies, Carlsbad, CA) at 37 °C and in an atmosphere containing 5% CO₂.

BALB/c and C3H female mice were obtained from The Jackson Laboratory (Bar Harbor, ME), and CB17SC scid ^−/− female mice were purchased from Taconic Farms (Germantown, NY). All were 6 to 8 weeks old upon initiation of the experiments. Detailed information can be found in the supplementary material.

In brief, orthotopic 4 T1 and LM8 tumors were initiated by inoculating 1 × 10⁵ cells in the third mammary fat pads of the BALB/c mice and in the right tibia of C3H mice, respectively. The first pDNA treatments (10 μg; wtIL-12, ttIL-12, or Ctrl) via intramuscular electroporation were performed as described previously [19]; a second identical treatment was administered 10 days later. For metastatic tumor analysis, 4 T1-bearing mice and LM8-bearing mice were euthanized 20 days or 5 days after primary tumor removal, respectively, and lungs, livers, and bones were collected to analyze metastatic status. India ink inflation was performed to determine the level of lung metastasis, and white metastatic nodules were counted using a dissecting microscope.

To overexpress CCL22 and CXCL2 in vivo, 10 μg pCCL22 or pCXCL2 was administered via intramuscular electroporation into 4 T1- and LM8-bearing mice 3 days prior to ttIL-12 treatment as described above. Seven days after the second treatment, orthotopic 4 T1 and LM8 tumors were collected for later analysis.

For the PDX model, patient-derived OS60-SJ osteosarcoma tumor cells (generously provided by Dr. Richard Gorlick, the Pediatric Preclinical Testing Consortium, The University of Texas MD Anderson Cancer Center) were implanted subcutaneously into CB17SC scid ^−/− female mice. When tumors reached 300 mm³ in size, the mice were treated with Ctrl, human wtIL-12, or human ttIL-12 pDNA as already described once per week for 4 weeks. To make these CB17SC scid ^−/− mice immunocompetent, they were injected with 2 × 10⁷ human peripheral blood mononuclear cells (PBMCs) intravenously every 2 weeks along with pDNA treatment.

Formalin-fixed, paraffin-embedded lung and liver tissue sections were stained with hematoxylin and eosin (H&E; Sigma Chemical, St Louis, MO). Frozen tumor sections were fixed with cold acetone, acetone plus chloroform (1:1), and acetone. Tissue sections were blocked with blocking buffer (5% normal horse serum and 1% normal goat serum in phosphate-buffered saline solution) and incubated with rat anti-mouse CD8α antibody (clone YTS105.18; AbD Serotec, Raleigh, NC). For FOXP3 staining, sections were blocked with blocking buffer containing 0.1% Triton X-100 for 40 min and incubated with rat anti-FOXP3 antibody (#12653; Cell Signaling Technology, Danvers, MA) overnight at 4 °C. The next day, tissue sections were blocked and incubated with goat anti-rat horseradish peroxidase secondary antibody (Life Technologies) for 1 h at room temperature, which was followed by DAB staining for 5~10 min at room temperature. Nuclei were then counterstained with hematoxylin (Sigma Chemical), and tumor sections were mounted with ClearMount Mounting Solution (Life Technologies). Slides were visualized under a Nikon eclipse Ti fluorescence microscope (Nikon Instruments, Melville, NY).

During the mouse model experiment just described, an antibody depleting CD8⁺ T cells (250 μg per mouse; clone 2.43) was administered intraperitoneally to mice twice a week for 3 weeks starting on day 3 after tumor cell inoculation.

Portions of primary tumor and metastatic tissues were digested with collagenase IV/DNase (0.2 mg/mL and 10 μg/mL) at 37 °C for 30 min. Cells were put through 40-μm filters to obtain single-cell suspensions and stained as previously described [21]. Cell surfaces were stained with anti-mouse CD4, anti-mouse CD206, anti-mouse CD11b, anti-mouse F4/80, anti-mouse Gr1.1, anti-mouse CD3, anti-mouse CD45, anti-mouse CD8a, anti-mouse CD28, anti-mouse NKG2D, anti-human CD45, anti-human CD8, anti-human CD33, and anti-human HLA-DR antibodies (BioLegend, San Diego, CA) labeled with fluorescein isothiocyanate, PerCP/Cy5.5, phycoerythrin, or Brilliant Violet 421. Flow cytometry was performed using an Attune flow cytometer (Life Technologies), and the data were analyzed with FlowJo software.

Total RNA was isolated from tumor tissues using TRIzol reagent (Thermo Fisher Scientific, Waltham, MA) according to the manufacturer’s instructions. Equal amounts of RNA were reverse-transcribed into cDNA by using iScript™ gDNA Clear cDNA Synthesis Kit (Bio-Rad, Berkeley, CA). qRT-PCR was performed using the Power SYBR Green Master kit (Thermo Fisher Scientific, Waltham, MA), with a StepOnePlus™ Real-Time PCR System (Thermo Fisher Scientific, Waltham, MA), as previously described [22]. The level of mRNA expression was normalized to that of the housekeeping gene GAPDH. The primer sequences of all mouse genes can be found in supplementary materials.

The IL-12 and IFNγ proteins in mouse tumors were measured in triplicate by commercially available kits (R&D Systems, Minneapolis, MN) by following the manufacturer’s instructions. Mouse CCL22, CXCL9, CXCL10, CCL5, and CXCL2 proteins were assayed by using commercial kits from Biolegend (mouse proinflammatory chemokine mix and match subpanel).

For each reported result, data were pooled from at least two independent experiments. The Student t-test was used to compare results between two treatment groups; one-way ANOVA was used to compare results from more than two treatment groups. The statistical significance of differences in survival curves was determined by log-rank survival analysis. All quantified data are presented as mean ± standard deviation (SD) or as indicated. The Prism software (GraphPad Software, Inc., La Jolla, CA) was used to determine the P values, and P values < 0.05 were considered statistically significant.

Mutation-induced oncogene activation and tumor suppressor gene inactivation have been widely investigated as markers to predict survival duration for patients with cancer. However, no one mutation is known to universally predict OS duration across different tumor types. This is not surprising, because different tumors may utilize different pathways or oncogenes to drive tumor progression and metastasis. However, immune profile infiltration is a common factor across all tumor types. Indeed, some studies have found that a certain type of immune profile was associated with long-term OS [23, 24]. T cell infiltration and frequency of FOXP3⁺ Tregs, MDSCs, and tumor-associated macrophages all have been discussed in the literature as predictors of OS in specific tumor types [25, 26], but none were analyzed across multiple types of tumors. Our first goal for this study was to identify a single immune profile that can universally predict long-term OS across different tumor types. Our second goal was to develop a simple immune therapy that converts tumors from an immune profile associated with short OS to the immune profile associated with longer OS.

To identify an immune profile associated with prolonged survival, we focused on T cell activation markers and immunosuppressive cell markers in an analysis of TCGA data. Our rationale was that functional T cells in tumors are effective in eliminating tumor cells and enhancing responses to chemotherapy, radiation, or other treatments. Although CD69, CD29, and CD25 are T cell activation markers, these markers are found in many other cells [27–29]. IFNγ is also not T cell specific, but its expression in T cells and other immune cells is a fairly good indicator of survival. In fact, IFNγ^Hi alone is reasonably accurate in predicting OS rate (Fig. 1a,c), but the predicted OS rates are not very high in some types of tumors (Additional file 1: Table S1). For example, the OS rates of patients with IFNγ^Hi lung cancer or melanoma were 23.6.4 and 39.7%, respectively (Additional file 1: Table S1). To refine the profile toward longer OS, we added the CD8 gene; the combination of IFNγ^Hi and CD8^Hi improved accuracy of identification of patients with longer OS, but further inclusion of parameters indicating low expression levels of immunosuppressive cell markers, such as FOXP3 and CD33, even more greatly improved the accuracy of the combination’s prediction of longer OS. For example, including these two immunosuppressive markers in the profile increased the predicted OS rate from 54.1 to 68% in patients with sarcoma (Fig. 1b,d). This observation makes sense because CD33 is a key marker for MDSCs and FOXP3 is a key indicator for Tregs, and both Tregs and MDSCs can shut down T cell activity by secreting unfavorable molecules such as IL-10, transforming growth factor beta (TGF-β), IL-35, and arginase 1 (Arg1) [30, 31]. Of course, inclusion of additional genes may further improve the accuracy of prediction of OS rate, but further expansion of the combination of biomarkers would have hampered utility for OS analysis across tumor types in the TCGA database, because some datasets lack data on RNAs or genes of interest. Expanding the biomarker panel may also limit its clinical application. However, these four genes, at the indicated expression levels, together predict long-term OS in multiple types of tumors (Additional file 1: Table S1). Our results strongly indicate that the CD8a^HiIFNγ^HiCD33^LowFOXP3^Low immune profile is associated with long-term OS across multiple tumor types.Fig1Fig. 1

The CD8^HiIFNγ^HiCD33^LowFOXP3^Low immune profile is correlated with favorable clinical outcome. a,c Kaplan-Meier survival analysis of human sarcoma (a) and breast carcinoma (c) based on IFNγ gene copy number. The overall survival rates were stratified by IFNγ gene expression level: the 25% of patients with the highest IFNγ levels were placed in the IFNγ^Hi group and the 25% with the lowest IFNγ levels were placed in the IFNγ^Low group. b,d Patients with sarcoma (b) and patients with breast cancer d were grouped by IFNγ, CD8, CD33, and FOXP3 gene expression levels: the 25% of patients in the IFNγ^Hi group with the highest CD8a levels were placed in the CD8^HiIFNγ^Hi group, and the 25% of patients in the IFNγ^Low group with the lowest CD8a levels were placed in the CD8^LowIFNγ^Low group. The CD8^HiIFNγ^Hi group and the CD8^LowIFNγ^Low group were then substratified into the CD8^HiIFNγ^HiCD33^LowFOXP3^Low group, the CD8^HiIFNγ^HiCD33^HiFOXP3^Hi group, the CD8^LowIFNγ^LowCD33^LowFOXP3^Low group, and the CD8^LowIFNγ^LowCD33^HiFOXP3^Hi group. Kaplan-Meier analyses were used to determine the association of each group with a favorable survival outcome

The consistent association of the CD8a^HiIFNγ^HiCD33^LowFOXP3^Low immune profile with long-term OS suggests that any treatment that can convert tumors to this immune profile may prolong OS. In our attempts to identify such a treatment, we have investigated several reagents, but all failed. This study describes our investigation of the impact of ttIL-12 on the immune profiles of two different common tumors. ttIL-12 (US patent 9,657,077 B2) was selected because this product targets a universal tumor-specific protein, CSV, and promotes accumulation of IL-12 in tumors [19].

We first assayed the effect of ttIL-12 on tumor progression using two aggressive orthotopic mouse tumor models: an epithelial tumor (4 T1, breast carcinoma) and a mesenchymal tumor (LM8, sarcoma). BALB/c mice bearing a 4 T1 tumor and C3H mice bearing an LM8 tumor were treated with empty (Ctrl), wtIL-12, or ttIL-12 pDNA. As expected, ttIL-12 treatment showed higher potency in suppressing tumor growth than wtIL-12 treatment in both 4 T1 and LM8 models (Additional file 1: Figure S1).

We then evaluated the effect of ttIL-12 on immune profile in tumors. CD8⁺ T cells, Tregs, and MDSCs were assayed using antibodies against CD8a, FOXP3, and CD11b/Gr1, respectively. To our surprise, ttIL-12 treatment not only increased numbers of CD8⁺ T cells but also decreased numbers of FOXP3⁺ Tregs and MDSCs in 4 T1 tumors compared with both Ctrl treatment and wtIL-12 treatment (Fig. 2). ttIL-12 treatment also decreased M2 macrophages in tumors, as determined by using antibodies against F4/80/CD11b/CD206 (Additional file 1: Figure S2). Similar results were obtained in the LM8 model (Additional file 1: Figure S3). These results suggest that ttIL-12 has extra independent biological function on top of the known IL-12 function, which is to block immune suppression. We also measured the tumor levels of IL-12 and IFNγ proteins, which can overcome the suppressive tumor microenvironment and shift the balance to favor CD8⁺ T cell antitumor immunity. Our data show that tumors treated with ttIL-12 accumulated more IL-12 and expressed higher levels of IFNγ than those treated with wtIL-12 (Fig. 2e; P = 0.0024). Notably, the ratio of CD8⁺ T cells to Tregs in ttIL-12–treated tumors was significantly higher than in wtIL-12–treated tumors (Fig. 2c; P = 0.0092), which shows that ttIL-12 has higher potency than wtIL-12 therapy in inducing the immune profile associated with longer OS.Fig2Fig. 2

ttIL-12 enhanced infiltration of CD8⁺ T cells into tumors, decreased infiltration of Tregs and MDSCs, and increased tumor levels of IFNγ. (a) Treatment scheme. BALB/c mice were inoculated with 4 T1 breast cancer cells by intra-mammary fat pad injection. CH3 mice were inoculated with LM8 sarcoma cells by intratibial injection. Mice from both groups were treated with empty control plasmid DNA (pDNA; pCtrl), wild-type IL-12 pDNA (pwtIL-12), or tumor-targeted IL-12 pDNA (pttIL-12; n = 5~8 mice per treatment group). Primary tumors were removed surgically as indicated. b Frozen sections from 4 T1 primary tumors were stained with CD8 antibody (upper panels) or FOXP3 antibody (lower panels). FOXP3 is a marker of T regulatory cells (Tregs). Representative sections are shown. Scale bars, 100 μm. Absolute numbers of CD8⁺ cells and FOXP3⁺ cells were determined by microscopy. c Calculated CD8⁺/Treg ratios. d Representative flow cytometry plots of percentages of CD11b⁺Gr1⁺ cells (myeloid-derived suppressor cells [MDSCs]); absolute numbers of MDSCs per 10⁵ total tumor cells were calculated from the flow cytometry data. e Levels of the IL-12 and IFNγ proteins in primary 4 T1 tumors were measured by ELISA. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

To investigate the potential impact of the combination of ttIL-12 treatment and surgery on metastasis, 4 T1 or LM8 tumor–bearing mice treated with ttIL-12, wtIL-12, or Ctrl underwent surgical removal of the primary orthotopic tumor after the pDNA treatment, and the metastatic tumor nodules in their lungs and livers were counted 20 days (4 T1) or 5 days (LM8) later. Numbers of metastatic nodules in the lungs were lower in the 4 T1 mice treated with ttIL-12 plus surgery (ttIL-12 + S) than in those treated with wtIL-12 plus surgery (wtIL-12 + S), as validated via H&E staining in lungs and livers. Numbers of lung metastatic nodules were dramatically lower in 4 T1-bearing mice treated with ttIL-12 + S than in those treated with control plasmid plus surgery (Ctrl+S), while wtIL-12 + S slightly decreased lung metastatic nodules in both 4 T1 and LM8 tumor–bearing mice (Fig. 3a, c). Moreover, ttIL-12 + S decreased liver metastatic nodules in the LM8 model compared with wtIL-12 + S (Fig. 3b; P = 0.0044). Together, these data show that ttIL-12 + S had greater anti-metastatic effect than wtIL-12 + S in both sarcoma and breast carcinoma models.Fig3Fig. 3

ttIL-12 induced a greater anti-metastatic effect than wtIL-12. Tumor-bearing mice were treated with control pDNA (pCtrl), wild-type IL-12 pDNA (pwtIL-12), or tumor-targeted IL-12 pDNA (pttIL-12). a Lungs were removed from 4 T1 tumor–bearing mice and were sectioned and stained with hematoxylin and eosin (H&E) to detect metastatic nodules. Representative sections are shown. Scale bars, 500 μm. b Livers and (c) lungs were removed from LM8 tumor–bearing mice, sectioned, and stained with H&E to detect metastatic nodules. Representative sections are shown. Scale bars, 500 μm for H&E and low-magnification immunohistochemistry. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

To identify the mechanism that may account for the enhanced anti-metastatic effect of ttIL-12 + S, we analyzed the metastatic nodules in lungs and livers of LM8 tumor–bearing mice for infiltrated CD8⁺ T cells, MDSCs, and Tregs using flow cytometry and immunohistochemical staining. Mice treated with ttIL-12 + S showed higher frequencies of CD8⁺ T cells (P < 0.0001) and lower frequencies of MDSCs (P < 0.0001) and Tregs (P < 0.001) in lung metastatic nodules than mice treated with wtIL-12 + S (Fig. 4, Additional file 1: Figure S4), which was consistent with ttIL-12’s role in inducing the CD8a^HiIFNγ^HiCD33^LowFOXP3^Low immune profile in primary orthotopic tumors. Similar results were found in the liver metastatic nodules of LM8 tumor–bearing mice (Additional file 1: Figure S5).Fig4Fig. 4

ttIL-12 increased CD8⁺ T cell infiltration and decreased MDSC and Treg infiltration into lung metastatic nodules. a,b Mice bearing LM8 tumors were treated with empty plasmid DNA (pCtrl), wild-type IL-12 plasmid DNA (pwtIL-12), or tumor-targeted IL-12 plasmid DNA (pttIL-12). At the end of the treatment period, their lungs were removed and subjected to flow cytometry to determine percentages of (a) CD8⁺CD45⁺ cells (CD8⁺ T cells) and (b) CD11b⁺Gr1⁺ cells (myeloid-derived suppressor cells [MDSCs]) in metastatic nodules, and absolute numbers of CD8⁺ T cells and MDSCs were calculated from the flow cytometry data. c Sections of paraffin-embedded lungs from LM8 tumor–bearing mice were labeled with FOXP3, and absolute numbers of FOXP3⁺ cells in metastases were determined by microscopy. Representative images are shown. Scale bars, 25 μm. Pooled data from two independent experiments with n = 3 mice per treatment group. Means ± standard error of the mean are shown. **P < 0.01, ***P < 0.001, ****P < 0.0001

To understand the mechanism underlying the impact of ttIL-12 treatment on immune profile change in tumors, we used RT-PCR to analyze the tumors for several chemokines and cytokines known to participate in the migration and recruitment of CD8⁺ T cells, Tregs, and MDSCs. Although both ttIL-12 and wtIL-12 significantly increased levels of CCL5, which recruits CD8⁺ T cells into tumors, ttIL-12 treatment was more effective than wtIL-12 treatment in increasing tumor levels of IFNγ and decreasing tumor levels of CCL22 and CXCL2 (Additional file 1: Figure S6B), chemokines that recruit Tregs and MDSCs, respectively. To validate these mRNA-based gene expression results, an ELISA assay was performed. This assay confirmed that ttIL-12 treatment significantly increased the quantities of IFNγ protein (Fig. 2e) and reduced the quantities of CCL22 and CXCL2 proteins in tumors, compared with wtIL-12 treatment (Fig. 5a, Additional file 1: Figure S6C). Moreover, measurement of levels of CXCL9 and CXCL10 proteins, which are CD8⁺ T cell attractants, showed that ttIL-12 increased CXCL9 levels in both primary tumor and lung metastatic nodules in the 4 T1 and LM8 models, while ttIL-12 increased CXCL10 levels only in lung metastatic tumors in the 4 T1 model and liver metastatic tumors in the LM8 model (Fig. 5a, Additional file 1: Figure S6C). Together, these findings indicate that ttIL-12 increases CD8⁺ T cell frequency in tumors and is associated with increases in expression of IFNγ and CXCL9 and that ttIL-12 decreases frequencies of Tregs and MDSCs and is associated with reductions in expression of CCL22 and CXCL2 in the tumor microenvironment.Fig5Fig. 5

ttIL-12 led to tumor conversion to the CD8^HiIFNy^HiCD33^LowFOXP3^Low immune profile through enriching CXCL9 and decreasing CXCL2 and CCL22 in tumors. 4 T1 tumor–bearing mice were treated with control pDNA (pCtrl), wild-type IL-12 pDNA (pwtIL-12), tumor-targeted IL-12 pDNA (pttIL-12), or ttIL-12 plus CXCL2 pDNA (pttIL-12 + pCSCL2) or CCL22 pDNA (pttIL-12 + pCCL22). Seven days after the second injection of CXCL2 or CCL22 plasmid DNA, primary tumors and lungs were surgically removed. a Levels of CCL5, CXCL2, CXCL9, CXCL10, and CCL22 proteins were determined by ELISA in the primary tumors (upper panels) and lung metastatic nodules (lower panels). b Samples of tumors were subjected to flow cytometry to determine percentages of CD8⁺ T cells, NKG2D⁺CD8⁺ T cells (non-exhausted), and myeloid-derived suppressor cells (MDSCs). Proportions of CD8⁺ T cells, NKG2D⁺CD8⁺ T cells, and MDSCs in primary tumors were calculated from flow cytometry data. c Primary tumor sections were stained with FOXP3 and numbers of FOXP3⁺ (T regulatory) cells determined by microscopy. Means ± standard error of the mean are shown. NS, not significant; *P < 0.05, **P < 0.01

To verify that inhibition of metastasis and induction of the CD8a^HiIFNγ^HiCD33^LowFOXP3^Low immune profile by ttIL-12 treatment prolongs OS, we monitored the survival of LM8 and 4 T1 tumor–bearing mice treated with ttIL-12, wtIL-12, or Ctrl in combination with surgery. While both wtIL-12 + S and ttIL-12 + S treatments prolonged survival in the 4 T1 model compared with Ctrl+S treatment (P < 0.0001; n = 8~14), ttIL-12 + S treatment led to significantly longer OS than wtIL-12 + S treatment (Fig. 6a; P = 0.0312). In the LM8 model, mice treated with wtIL-12 + S had curtailed survival compared to those treated with Ctrl+S; this is attributed at least partly to toxicity, because the host mouse strain C3H is very sensitive to IL-12 treatment. However, the C3H mice treated with ttIL-12 + S showed no evidence of toxicity. The reduced toxicity of ttIL-12 treatment may have been due to lower levels of IL-12 in the serum than in mice treated with wtIL-12 (Additional file 1: Figure S6A). Moreover, ttIL-12 + S significantly extended the survival of mice compared with both Ctrl+S and wtIL-12 + S in these tumor-bearing IL-12–sensitive mice (Fig. 6b) (P = 0.0086 and P = 0.0125, respectively; n = 8~9). To further prolong survival, some LM8 tumor–bearing mice were given ttIL-12 treatment both before and after surgical removal of the primary tumor instead of before surgery only. The extra ttIL-12 treatment after surgery led to a long-term OS rate of 60% (Fig. 6c), showing that ttIL-12 treatment both before and after surgery is more beneficial than ttIL-12 treatment before surgery only.Fig6Fig. 6

ttIL-12 prolonged survival compared to wtIL-12 in two independent mouse tumor models and a human tumor model. Overall survival of tumor-bearing mice treated with empty plasmid DNA (pDNA; pCtrl), wild-type IL-12 pDNA (pwtIL-12), tumor-targeted IL-12 pDNA (pttIL-12), or pttIL-12 plus a CD8-depleting antibody (pttIL-12 + αCD8a) was determined by using the Kaplan-Meier analysis. Treatment strategy was as described in Fig. 2a. a Overall survival of 4 T1 tumor–bearing mice by treatment group. b Overall survival of LM8 tumor–bearing mice. c Overall survival of LM8 tumor–bearing mice treated with an extra administration of pDNA 15 days after primary tumor removal. (D) Overall survival of OS60-SJ osteosarcoma patient-derived xenograft (PDX) tumor-bearing mice overall survival. The dark arrows represent pDNA treatment, the blue arrows represent primary tumor removal, and the red arrows represent PBMC injections. Data represent cumulative results from at least two independent experiments. *P < 0.05, **P < 0.01***P < 0.001, ****P < 0.0001