Male C57BL/6 mice (5–10 weeks) were purchased from the Jackson Laboratories and were maintained in a pathogen-free environment. Animal studies were pre-approved and carried out in accordance with University of Texas MD Anderson Cancer Center Institutional Animal Care and Use Committee (IACUC) guidelines. Animals were anesthetized with isoflurane for tumor inoculations and blood draws, and euthanized according to IACUC guidelines.

Mouse tonsil epithelial cells expressing HPV-16 E6 and E7 and H-Ras (mEER) were a kind gift from Dr. J. Lee (NantKwest Inc., Culver City, CA). These cells were maintained in complete media as previously described [14], and sub-cultured at 80% confluence the day before tumor induction in mice.

Tumor-infiltrating lymphocytes (TIL) were analyzed by 16-color multi-parametric flow cytometry using the following antibodies: BUV737 anti-CD3 (17A2), Alexa Fluor 700 anti-Granzyme B (GB11), BV605 anti-CD11c (HL3), APC-Cy7 anti-CD11b (M1/70), anti-mouse CD16/32 (2.4G2, mouse Fc-block) from BD Bioscience (San Jose, CA), BV650 anti-CD8 (53–6.7), APC anti-CTLA-4 (UC10-4B9), PerCP-ef710 anti-Lag3 (C9B7W), PE-Cy7 anti-PD-1 (RMP1–30), BV711 anti-PD-L1 (10F.9G2), PE-Cy5 anti-F4/80 (BM8) from Biolegend (San Diego, CA), Alexa Fluor 488 anti-FoxP3 (150D/E4) and e450 anti-Gr-1 (RB6-8C5) from eBioscience (Waltham, MA). The following antibodies for in vivo administration were purchased from BioXcell (West Lebanon, NH) and used at the concentrations shown: α-PD-1 (RMP1–14 at 250 μg per dose), α-CTLA-4 (9H10 at 100 μg per dose) and α-Lag-3 (C9B7W at 200 μg per dose). The STING agonist ML-RR-S2 CDA (ADU-S100) was procured from MedChemExpress (Monmouth Junction, NJ). For fluorescence immunohistochemistry, rabbit monoclonal anti-mouse PD-L1 antibody was purchased from Abcam (Cambridge, MA) and chicken anti-rabbit IgG cross-absorbed antibody Alexa Fluor 594 conjugate from Invitrogen (Carlsbad, CA).

Mice were implanted with 4 × 10⁴ mEER cells in 50 μl PBS into the base of the tongue or 1 × 10⁶ mEER cells in 200 μl PBS subcutaneously in the flank. Mice were monitored closely and euthanized when a necrotic tumor was observed and/or when the mice lost 20% or more of their initial weight. For characterization of TILs, mEER cells were mixed in a 2:1 ratio with Matrigel (BD Biosciences, San Jose, CA) and 1 × 10⁵ cells in 50 μl per animal were implanted in the tongue. For the mEER pseudometastasic setting, 4 × 10⁴ cells for survival experiments or 1 × 10⁵ cells for TIL analysis were implanted in the tongue, and 1 × 10⁶ cells subcutaneously in the flank of each mouse.

Starting between days 5 and 7 after tumor challenge mice received intraperitoneal injections of immune checkpoint therapeutic antibodies or their combination, three times at three-day intervals. Control animals were untreated. For the pseudometastasic model, STING agonist (ML-RR-CDA) was given by intratumoral (i.t.) injection at day 10 when tumors had reached at least 4 mm in diameter, and a second time on day 16. The immune checkpoint antibodies were administered i.p. at days 10, 13, 16 for TIL analysis and an additional dose on day 19 for survival experiments.

For characterization of TIL, mice were euthanized at the days specified in the results section. Tongue and flank tumors were collected and digested as previously described [10]. Purified leukocytes were stained for multi-parametric flow cytometry analysis with a 16-color antibody panel. Cells were blocked with mouse Fc-block, stained with surface markers, fixed and permeabilized with the FoxP3 Fix/Perm Kit (eBioscience, Waltham, MA) followed by staining for intracellular markers. Samples were run in an LSR-II X-20 Fortessa (BD Biosciences, San Jose, CA) at the South Campus Flow Cytometry Core, MD Anderson Cancer Center (Houston, TX) and analyzed using FlowJo version 10 (Flowjo LLC, Ashland, OR). The live/dead fixable aqua dye (Thermo Scientific, Waltham, MA) was used to gate out dead cells and to include only live cells for analysis. Live leukocyte gate was set based on forward and side scatter to include both lymphocytes and larger myeloid cells. Tregs were identified based on CD4⁺Foxp3⁺ expression within the CD3⁺ gate. From CD3⁻ gate, CD11b⁺Gr-1⁺ cells were identified as total MDSC population. The ratio of CD8⁺ T cells to Tregs or MDSC were calculated by dividing the percentage of CD8⁺ T cells with that of CD4⁺Foxp3⁺ or CD11b⁺Gr-1⁺ cells.

Freshly harvested tumors were flash frozen in Shandon Cryomatrix embedding resin (Thermo Scientific, Waltham, MA). Cryostat sections (5 μM) were cut and placed on glass slides. The sections were fixed using cold methanol at − 20 °C for 20 min. Blocking of non-specific sites were performed using PBS-based Super block (Thermo Scientific) containing 0.3% Tween-20 for 30 min. The sections were then incubated successively with pre-titrated dilutions of anti-mouse PD-L1 primary antibody (1:250) overnight at 4 °C and chicken anti-rabbit IgG Alexa Fluor 594 conjugate (1:2000) for 1 h at room temperature. The slides were washed between steps using PBS containing 0.1% Tween-20. Sections were mounted using ProLong™ Gold Antifade containing DAPI nuclear counterstain (Molecular Probes, Eugene, OR). Adjacent sections stained with secondary antibody alone were used as staining controls to assess non-specific background. Stained slides were imaged with a fluorescence microscope equipped with digital camera (Olympus USA, Center Valley, PA), and using TRITC (for Alexa Fluor 594) and DAPI filters. Fluorescence photomicrographs obtained from four random regions for each section were analyzed for mean fluorescence intensity for PD-L1 expression using NIH ImageJ software.

Total RNA was extracted from tumor tissue using PureLink RNA mini kit (Thermo Scientific) according to the manufacturer’s instructions. The quality and concentration of RNA was determined using a NanoDrop UV spectrophotometer, and the RNA integrity was verified using Agilent 2100 BioAnalyzer (Palo Alto, CA).

RNA sequencing was performed using lllumina HiSeq 2000 at the Sequencing and Microarray Facility, MD Anderson Cancer Center (Houston, TX). Fastq files were filtered for Phred quality score of 20 and adapter sequences with minimum length of 35 to remove low quality reads using BBduk BBMap (The U.S. Department of Energy Joint Genome Institute, Lawrence Livermore National Laboratory, Walnut Creek, CA). The mRNA-Seq paired-end reads were aligned onto the mouse genome build UCSC mm10 (NCBI 38) and transcript level quantification of counts was performed using Salmon algorithm [15], followed by differential expression analysis based on a negative binomial distribution model using DESeq2 [16].

Total RNA were reverse transcribed to obtain cDNA using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA). Real-time qPCR was performed using target-specific forward and reverse primers and iQ SYBR Green qPCR Supermix using CFX384 Touch real-time PCR detection system (Bio-Rad). Relative quantification was calculated by 2^(−ΔΔCq) method and expressed relative to endogenous control 18S. The following mouse primers pairs were used, PD-L1 (CD274): TGC GGA CTA CAA GCG AAT CAC G (forward), CTC AGC TTC TGG ATA ACC CTC G (reverse); Ciita: ACC TTC GTC AGA CTG GCG TTG A (forward), GCC ATT GTA TCA CTC AAG GAG GC (reverse); Mx1: TTC AAG GAT CAC TCA TAC TTC AGC (forward), GGG AGG TGA GCT CCT CAG T (reverse); Isg15: ACG GTC TTA CCC TTT CCA GTC (forward), CCC CTT TCG TTC CTC ACC AG (reverse); Ifng: AAC TGG CAA AAG GAT GGT (forward), GAC CTC AAA CTT GGC AAT AC (reverse); 18S: CCA TTC GAA CGT CTG CCC TAT (forward), GTC ACC CGT GGT CAC CAT G (reverse).

Blood was collected from anesthetized mice through the retro-orbital plexus at day 15 or day 21 after tumor challenge and analyzed for liver enzymes (AST and ALT) at the Clinical Pathology Laboratory in the Veterinary Medicine and Surgery Department at MD Anderson Cancer Center (Houston, TX).

Mice were imaged at day 19 or day 23 after tumor challenge on the 1 T Bruker ICON at the MD Anderson Cancer Center Small Animal Imaging Facility as previously described [10]. Tumor volume was determined in three dimensions with ImageJ software after defining region of interest of the tumor on all possible sections.

All statistics were calculated using GraphPad Prism version 6. Statistical significance was determined using one-way or two-way ANOVA along with post-hoc correction to test differences between multiple groups or Student’s t-test to compare two groups. Mantel-Cox log rank test was used to compare survival curves. P values less than 0.05 were considered significant.

We compared anti-PD-1 responsiveness of mice bearing mEER tumors on the flank to those in the tongue. Tumor bearing mice were treated on days 5, 8 and 11 with α-PD-1 antibody and their survival was monitored. Consistent with our earlier report [11], none of the mice with flank-implanted tumors responded to α-PD-1 therapy while 54% of mice with tongue-implanted tumors exhibited sustained tumor regression with a significant survival advantage (Fig. 1a). The immune correlates for the protective efficacy of α-PD-1 therapy in the tongue tumors included a higher frequency of CD8⁺ T cells, specifically those with cytotoxic potential as evidenced by expression of Granzyme B (CTL). These enhanced T cell frequencies combined with overall pro-inflammatory modulation of the tumor microenvironment also gave rise to elevated ratios of CTL relative to both Tregs and MDSC (Fig. 1b).Fig1Fig. 1

Differential α-PD1 responsiveness of mEER tumors implanted in the flank and tongue. Separate groups of mice were injected with mEER tumor cells in the tongue (4 × 10⁴) or in the flank (1 × 10⁶), and treated with α-PD1 antibodies at days 5, 8 and 11. The percent survival of mice in the different groups is shown (a). Mantel Cox-test was performed to determine the significance of survival for each of the treatment groups relative to respective untreated group ****p < 0.00005. Results represent pooled data from multiple experiments (n = 10–18 mice/group). b At day 15 after tumor implantation mice in the different groups were sacrificed and the TIL were analyzed by flow cytometry to determine the frequencies of Granzyme B expressing functional CD8⁺ T cell populations, CD4⁺Foxp3⁺ Tregs, CD11b⁺Gr-1⁺ MDSC as well as ratios of functional Granzyme B expressing CD8⁺ T cells to Treg and to MDSC

In addition to the differential PD-1 expression in tongue- and flank-implanted mEER tumors, CD8⁺ T cells from tongue tumors showed higher levels of additional immune checkpoint inhibitory molecules, CTLA-4 and Lag3 (Additional file 1: Figure S1). We therefore, tested whether combination therapy to block either of these inhibitory receptors would enhance the efficacy of α-PD-1 therapy of tongue tumors. For this, we treated mice with tongue-implanted tumors at days 5, 8 and 11 with α-PD-1 alone or in combination with α-CTLA-4 or α-Lag3. The majority of mice treated with the combination of α-PD-1 and α-CTLA-4 exhibited tumor free survival through day 80 of follow up, while all of the mice in the control untreated group exhibited high tumor burden (Fig. 3a). In contrast, survival rates for mice treated with the combination of α-PD-1 plus α-Lag3 were not significantly different from those of mice treated with α-PD-1 alone. Monotherapy with α-CTLA-4 resulted in a survival advantage similar to that seen with α-PD-1, while targeting Lag3 alone was relatively ineffective (Fig. 3a). The MRI data of head and neck regions of mice collected at day 19 clearly showed significantly reduced tumor size in mice treated with the combination of α-PD-1 and α-CTLA-4, relative to treatment with either antibody alone or untreated mice (Fig. 3b and c), further supporting the positive survival outcome. These data demonstrate that efficacy of α-PD-1 therapy in the tongue implanted mEER tumors can be significantly enhanced by supplementing with immune checkpoint blockade targeting CTLA-4.Fig3Fig. 3

Efficacy of α-PD-1 therapy of tongue-implanted mEER tumors is enhanced by combination treatment with α-CTLA-4 but not α-Lag3. Mice were challenged with mEER tumor cells (4 × 10⁴) in the tongue and treated with antibodies targeting individual checkpoint receptors PD-1, CTLA-4, or Lag3 or using combinations of α-PD-1 and α-CTLA-4 or α-PD-1 and α-Lag-3 antibodies. The percentages of mice surviving in the different groups are shown (a). Statistical significance was calculated using Log-rank (Mantel-Cox) test. The significant difference for each treatment group compared to untreated control group is indicated by colored stars and between groups is shown on the legend; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Tongue tumor volume was measured by MRI (T2-weighted sagittal image) at day 19 after tumor implantation and representative data is shown for one mouse in each group (b) along with group means ± SD (n = 4–16 mice/group) (c). **p < 0.01, ****p < 0.0001, one-way ANOVA. Flow cytometry analyses of TIL isolated at day 15 from tongue tumor-bearing mice subjected to different treatments showing frequencies of total CD8⁺ T cells, Granzyme B expressing CD8⁺ T cells (d), CD4⁺FoxP3⁺ Treg, CD11b⁺Gr-1⁺ MDSC (e) as well as ratios of GrnzB⁺CD8⁺ T cells to Treg and to MDSC (f). Data shown are mean + SD from two experiments (except for anti-Lag3 group) with individual data points representing pooled TILs of 2–3 tumors. Statistical significance was calculated using one-way ANOVA with Turkey post-hoc test; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Even though α-PD-1 monotherapy was ineffective in treating mice with flank-implanted mEER tumors (Fig. 1a), supplementing α-PD-1 therapy with α-CTLA-4 resulted in regression of 40% of sub-cutaneous mEER and a significant survival advantage (Additional file 1: Figure S3). Since type I and type II interferons (IFNs) are known inducers of PD-L1 expression, which is recognized as a biomarker for α-PD-1 responsiveness on a variety of tumor cells [20, 21], we performed RNA-seq analysis and identified that an IFN pathway signature (both type I and type II) was activated at a significantly higher level in tongue-implanted mEER tumors relative to those on the flank (Additional file 1: Figure S4A and B). This is consistent with the PD-1/PD-L1 expression patterns (Fig. 2) as well as the relatively superior responsiveness of tongue tumors to α-PD-1 therapy (Fig. 1). Based on this information, we reasoned that treatment with type I and/or type II IFNs would improve the α-PD-1 therapy sensitivity of mEER tumors by modulating the expression of PD-1/PDL-1. Additionally, since the cytosolic nucleic acid sensor, Stimulator of Interferon Induced Genes (STING) activates IFN secretion [22], and intratumoral administration of cyclic dinucleotide (CDN) STING agonists such as ML-RR-CDA (ADU-S100) has been shown to activate both IFN-α/β and IFN-γ signaling pathways [23, 24], we tested whether stimulation of the STING pathway would induce expression of PD-1/PD-L1 to promote responsiveness to α-PD-1 therapy. We first performed in vitro stimulation of mEER tumor cells with IFN-α, IFN-γ or ML-RR-CDA, and observed increased PD-L1 expression in response to these treatments (Additional file 1: Figure S4C and D). We and others have shown previously that intratumoral delivery of STING agonist is effective in inducing local as well as systemic anti-tumor immune responses [25, 26]. Therefore, we investigated intratumoral STING agonist treatment as a strategy to reverse the non-responsiveness of flank-implanted mEER tumors to α-PD-1 therapy concurrent with maintaining or improving the antitumor efficacy of α-PD-1 therapy in tongue tumors.

For these studies, we adopted a pseudometastasic model where mice were implanted with mEER tumors in the tongue as well as on the flank. Different groups of mice were treated with injection of STING agonist into the flank tumors as a monotherapy or in combination with systemic α-PD-1 and/or α-CTLA-4 treatment (Fig. 4a). We observed that intratumoral injection of STING agonist induced complete regression of flank tumors when combined with α-PD-1, or α-CTLA-4 or both together in the majority of mice (Fig. 4b). Importantly, the majority of the mice receiving the combination of intratumoral STING agonist and both systemic α-PD-1 and α-CTLA-4 therapy exhibited a significant survival advantage and clearance of both the flank (Fig. 4c) and tongue tumors (Additional file 1: Figure S5).Fig4Fig. 4

Abscopal anti-tumor efficacy of intratumoral STING activation in combination with systemic checkpoint antibodies. Mice were inoculated with mEER tumor cells both in the flank (1 × 10⁶) and tongue (4 × 10⁴) and treated with intratumoral (i.t.) administration of STING agonist (ML-RR CDA) on days 10 and 16 along with or without immunotherapy employing individual or combinations of α-PD-1 and α-CTLA-4 antibodies on days 10, 13, 16, and 19 (a). Growth of flank-implanted tumors over time for individual mice in different treatment groups is expressed in terms of tumor area (mm²) in (b). The data is pooled from three separate experiments and the total number of mice in each group is noted. The survival curves for mice in different treatment groups are shown in (c). Statistical significance for differences in survival of mice in different combination treatment groups relative to untreated control group were calculated using Log-rank (Mantel-Cox) test; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001