Background The ability to modulate immune-inhibitory pathways using checkpoint blockade antibodies such as αPD-1, αPD-L1, and αCTLA-4 represents a significant breakthrough in cancer therapy in recent years. This has driven interest in identifying small-molecule-immunotherapy combinations to increase the proportion of responses. Murine syngeneic models, which have a functional immune system, represent an essential tool for pre-clinical evaluation of new immunotherapies. However, immune response varies widely between models and the translational relevance of each model is not fully understood, making selection of an appropriate pre-clinical model for drug target validation challenging.
Methods Using flow cytometry, O-link protein analysis, RT-PCR, and RNAseq we have characterized kinetic changes in immune-cell populations over the course of tumor development in commonly used syngeneic models.
Results This longitudinal profiling of syngeneic models enables pharmacodynamic time point selection within each model, dependent on the immune population of interest. Additionally, we have characterized the changes in immune populations in each of these models after treatment with the combination of α-PD-L1 and α-CTLA-4 antibodies, enabling benchmarking to known immune modulating treatments within each model.
Conclusions Taken together, this dataset will provide a framework for characterization and enable the selection of the optimal models for immunotherapy combinations and generate potential biomarkers for clinical evaluation in identifying responders and non-responders to immunotherapy combinations.
The traditional drug development pipeline has relied on testing tumor growth inhibition of human tumor cells in vitro, then testing these molecules in vivo in immunodeficient mice bearing xenografted human tumors. However, this strategy ignores the importance of cross talk between the tumor and other cell types present in the tumor microenvironment (TME), including those of the immune system, that can dramatically impact response to therapy. The ability to modulate immune-inhibitory pathways represents a significant breakthrough in cancer therapy in recent years. Checkpoint blockade antibodies targeting programmed cell-death protein 1 (PD-1), programmed death-ligand 1 (PD-L1), and cytotoxic T-lymphocyte antigen 4 (CTLA-4) have shown great promise in the clinic, causing complete tumor regression and durable responses in a segment of patients. Blockade of the PD-L1/PD1 axis prevents inhibition of T-cell function, while blockade of CTLA-4 induces expansion of tumor reactive T-cells and there is strong interest in identifying small-molecule-immunotherapy combinations to increase the proportion of responses to checkpoint blockade. Identifying the right combinations as well as patients who will respond will rely on building a better understanding of the dynamic interplay between the tumor and the immune system which requires models with a functionally intact immune system. Identification and selection of appropriate in vivo models of immune response requires a better understanding of the dynamic interplay between the tumor and immune system across different models. Syngeneic models represent some of the most established models to investigate immune hypotheses. While several studies have characterized immune populations at single timepoints in syngeneic models, we have sought to characterize kinetic changes in immune populations that occur over time in some of the most commonly used models to better understand the underlying differences in response to immunotherapies.
All animal studies were performed according to UK Home Office and IACUC guidelines. Cell lines CT-26, 4 T1 and MC38 were purchased from ATCC. CT-26 (5 × 10 5 cells/mouse) or MC38 (1 × 10 7 cells/mouse) tumor cells were implanted subcutaneously (s.c.) in the left flank of female Balb/c and C57Bl/6 mice, respectively. 4 T1 (1 × 10 5 cells/mouse) tumor cells were implanted orthotopically in mammary fat pad 8 of female Balb/c mice. For time course (untreated) studies mice were randomized by body weight on the day of tumor implant, the tumors were collected on day 3 (CT-26 and MC38), day 7 (CT-26, MC38, and 4 T1), day 10 (MC38), day 14 (CT-26 and 4 T1) and day 18 (4 T1). For treated CT-26 studies 5 × 10 5 cells/mouse were implanted, and mice were randomized by body weight 2 days post implant. For treated MC38 studies 1 × 10 7 cells/mouse were implanted, and mice were randomized by cage on the day of implant. Mice were injected intraperitoneally with 10 mg/kg of co-formulated α-PD-L1 (mouse IgG1, clone D265A; AstraZeneca) and α-CTLA-4 (mouse IgG1, clone 9D9; AstraZeneca) or the respective isotype controls (αNIP; AstraZeneca) on day 3, 7 and 10 (CT-26) or day 1, 4 and 8 (MC38) post-implant.
At end of study tumor tissues were chopped then transferred into the gentleMACS C Tube containing RPMI. Cells were liberated from tumors for downstream application using a mouse tumor dissociation kit (Miltenyi Biotec) and octodissociator (Miltenyi Biotec) according to manufacturer’s instructions. Cells were stained with a viability marker (Live/Dead Blue, ThermoFisher) according to manufacturer’s instructions and blocked in anti-CD16/CD32 antibody (ThermoFisher). Cells were stained with fluorescence-conjugated antibodies in flow cytometry staining buffer with Brilliant Stain Buffer (BD Biosciences). Intracellular staining was performed using the FoxP3/transcription factor staining buffer set (ThermoFisher). For extracellular-only panels, cells were fixed in fixation buffer (BD) for 15 min prior to reading. Cells were analyzed on a BD fortessa flow cytometer and analyzed using FlowJo software (V.10, Treestar) or Cytobank. Gating strategies are shown in Additional file 2: Table S2.
Frozen tumors were homogenized using liquid nitrogen and a mortar and pestle to create a powder and 10 mg of tissue was used for RNA isolation by carrying out a Qiazol extraction followed by RNA extraction using the RNeasy Plus Mini Kit with a DNase digestion using the RNase-free DNase Kit (Qiagen) on the Qiacube HT (Qiagen) according to manufacturer’s instructions. RNA concentration was measured using the NanoDrop ND8000 (NanoDrop). Reverse transcription was performed using 50 ng of RNA with a Reverse Transcription kit and cDNA was then pre-amplified (14 cycles) using a pool of TaqMan primers (listed in Additional file 3: Table S3), following the manufacturer’s instructions (Life Technologies). Sample and assay preparation of the 96.96 Fluidigm Dynamic arrays were carried out according to the manufacturer’s instructions. Data were collected and analyzed using Fluidigm Real-Time PCR Analysis 2.1.1 software. dCt was calculated by taking Ct – Average Ct housekeeping genes. An average dCt for all vehicle controls was calculated and (dCt – dCt (average. vehicle)) was used to calculate negative ddCt. 2^negativeddCt was used to calculate Fold Change. P values were calculated by performing a student’s t-test on the negative ddCt values in JMP Software and p< 0.05 was considered significant. Data were plotted using Spotfire 6.5.3 software or GraphPad Prism (V7). Gene set variation analysis (GSVA) scoring was performed using an in house R script using genes defined in Rooney et al.
For RNA sequencing, total RNA was extracted using the RNeasy 96 Qiacube HT Kit (Qiagen), quality validated using nanodrop and Quantit RNA Assay Kit (Thermo Fisher), and submitted for TrueSeq Stranded mRNA library preparation, following the manufacturer’s instructions (Illumina). Resulting libraries were sequenced on the HiSeq4000 System, generating on average ~24Million mapped reads. The python toolkit bcbio 1.0.8 was used to quality control and analyze the sequencing data. In brief, the sequencing reads were aligned using hisat2 2.1.0 for quality control purposes and a QC report was generated using multiqc. Quantification of expression of the transcripts was performed directly against the mouse mm10 Ensembl transcriptome using Salmon 0.9.1 without alignment, or adapter trimming. The R package tximport was used to create a gene by sample count table. Subsequently, the DESeq2 R package (version 1.16.1) was used to normalize for library size and perform differential expression analysis.
Genes with an average count of less than 1 per sample were removed. Pathway analysis was performed with IPA QIAGEN Inc., utilizing fold changes and FDR corrected p-values obtained by DESeq2. A customized support vector regression (SVR) model was developed in-house based on the CIBERSORT algorithm to achieve immune cell deconvolution. In brief, this machine learning approach infers the cell type composition of a given tissue sample by hypothesizing a linear relationship between the mixed gene expression profile in the tissue and the expression profile of isolated immune cells provided as reference. Here, we utilized a signature matrix optimized for mouse leukocyte deconvolution to determine the relative proportions of 25 murine immune cell types in the RNA.
Tumor proteins were lysed in RIPA buffer and diluted to 1 ng/μl before using the Olink mouse exploratory panel (O-link) according to the manufacturer’s instructions. In brief, pairs of oligonucleotide-labeled antibody probes bind to their targeted protein. The oligonucleotides hybridize in a pair-wise manner when brought in close proximity. The addition of a DNA polymerase leads to proximity-dependent DNA polymerization, generating a unique PCR target sequence, which is subsequently detected using a Fluidigm Biomark microfluidic real-time PCR instrument (Fluidigm). The quantification cycle (Cq) values from a DNA extension control are subtracted from the measured cq value and an interplate correction factor applied to yield a normalized protein expression value (NPX), which is log2-transformed.
Error bars relate to SEM unless indicated in figure legends. Appropriate statistical testing was performed using JMP Software, GraphPad Prism (V7) or an in-house R tool. Statistical significance is indicated as follows: * p< 0.05, ** p< 0.01, *** p< 0.001, **** p ≤ 0.0001.
In order to better understand how some of the most commonly utilized syngeneic models respond to checkpoint inhibition, we chose CT-26, MC38, and 4 T1 models for characterization after treatment with a clinically relevant combination of α-mPD-L1 + α-mCTLA-4, which has been shown to induce anti-tumor immune responses in syngeneic models. After tumor implantation, mice were dosed twice a week with a combination of α-PDL-1 + α-CTLA-4 or isotype controls for 2 weeks and tumor growth and survival were measured. In the context of these experiments, the CT-26 model showed the most robust response to checkpoint inhibition with 10/12 animals showing reduced tumor growth or stasis leading to an enhanced survival. In our hands, the MC38 tumor model showed a more varied response to the same checkpoint inhibition therapy, with delayed tumor growth, but only 1/12 mice showing complete response to therapy. However, despite only a modest reduction in tumor growth, checkpoint inhibition enhanced survival in this model. In contrast to the efficacy observed in CT-26 and MC38 after checkpoint inhibition, the 4 T1 tumor model showed no difference in tumor growth and no enhanced survival benefit in response to checkpoint inhibition. All three models expressed PD-L1 in both the myeloid and tumor (CD45-) compartments. Given this variation in response across these three models observed in our lab and others, we sought to further characterize the kinetics of immune cell infiltration into the tumor microenvironment over the time course of tumorigenesis in these models as a means to better understand the possible reasons underlying differences in response.
Fig. 1 Impact of α-mPD-L1+ α -mCTLA-4 treatment on tumor growth in syngeneic models. Line graphs show tumor volumes from (a) Balb/c CT-26 tumor bearing mice treated with isotype control or (b) anti-mPD-L1 + anti-mCTLA-4 combination treatment; (c) C57Bl/6 MC38 tumor bearing mice treated with Isotype Control or (d) anti-mPD-L1 + anti-mCTLA-4 combination treatment; (e) Balb/c 4T1 tumor bearing mice treated with isotype control or (f) anti-mPD-L1 + anti-mCTLA-4 combination treatment. Vertical dotted lines
