Immune Checkpoint Blockade Delays Cancer Development and Extends Survival in DNA Polymerase Mutator Syndromes

Treating high-mutation burden mice with immunotherapy prior to cancer onset significantly improves survival, raising the possibility of utilizing immune checkpoint blockade for cancer prevention, especially in individuals with increased risk.

Abstract

Mutations in the exonuclease domains of the replicative nuclear DNA polymerases POLD1 and POLE are associated with increased cancer incidence, elevated tumor mutation burden (TMB), and enhanced response to immune checkpoint blockade (ICB). Although ICB is approved for treatment of several cancers, not all tumors with elevated TMB respond, highlighting the need for a better understanding of how TMB affects tumor biology and subsequently immunotherapy response. To address this, we generated mice with germline and conditional mutations in the exonuclease domains of Pold1 and Pole. Engineered mice with Pold1 and Pole mutator alleles presented with spontaneous cancers, primarily lymphomas, lung cancer, and intestinal tumors, whereas Pold1 mutant mice also developed tail skin carcinomas. These cancers had highly variable tissue type–dependent increased TMB with mutational signatures associated with POLD1 and POLE mutations found in human cancers. The Pold1 mutant tail tumors displayed increased TMB; however, only a subset of established tumors responded to ICB. Similarly, introducing the mutator alleles into mice with lung cancer driven by mutant Kras and Trp53 deletion did not improve survival, whereas passaging these tumor cells in vitro without immune editing and subsequently implanting them into immunocompetent mice caused tumor rejection in vivo. These results demonstrated the efficiency by which cells with antigenic mutations are eliminated in vivo. Finally, ICB treatment of mutator mice earlier, before observable tumors had developed delayed cancer onset, improved survival and selected for tumors without aneuploidy, suggesting the potential of ICB in high-risk individuals for cancer prevention.

Significance: Treating high-mutation burden mice with immunotherapy prior to cancer onset significantly improves survival, raising the possibility of utilizing immune checkpoint blockade for cancer prevention, especially in individuals with increased risk.

Graphical Abstract

Introduction

Immune checkpoint blockade (ICB) has provided an extraordinary advance in the treatment of multiple cancer types (1). ICB involves the use of antibodies that target T cell coreceptors such as PD-1, PD-L1, and cytotoxic T-lymphocyte–associated protein 4 (2, 3). ICB has been approved for use as a first-line therapy for non–small cell lung cancer (NSCLC), melanoma, and urothelial and colorectal cancers (3). However, only a subset of patients respond well, and there is a lack of clear biomarkers that predict a favorable outcome. Some of the current predictive biomarkers for favorable ICB response involve high tumor mutation burden (TMB), increased PD-L1 and PD-1 expression, and tumor-immune infiltration (4, 5).

Tumors with mutations in proofreading domains of DNA polymerases and mismatch repair (MMR) deficiency give rise to high TMB (6). In theory, high TMB may lead to generation of mutant peptides that are processed and recognized as neoantigens by the immune system. Tumors evolve by upregulating immune checkpoints such as PD-1 and PD-L1 to evade the immune response and are therefore subject to favorable ICB response (7, 8). However, some patients with high TMB still do not respond, and some tumor types such as thyroid cancers with low mutation burden respond well (9). Thus, the question of how TMB impacts tumor biology and subsequently immunotherapy response remains to be resolved. Also, there are various conflicting studies of analysis of PD-L1 expression with IHC that limit its use as a predictive marker across cancer types (10). As ICB has been a transforming cancer treatment, there is a window of opportunity to model high TMB and ICB therapy response in which the new information could be used to modify treatment protocols and utilize them for the right subset of patients.

We and others have shown that ICB therapy can have high response rates in tumors with high TMB in which exceptional responses have been observed in patients with endometrial cancers, NSCLC, and colorectal cancer (11–14). NSCLC tumors with high TMB have also shown to be enriched in POLD1 and POLE exonuclease mutations (13). These exonuclease domain point mutations disable the proofreading function essential for correcting DNA replication errors, resulting in a mutator phenotype (15). These mutations have also been associated with positive responses to ICB in several cancers (12, 13). Tumors with somatic point mutations such as POLE V411L and POLE P286R have been shown to be associated with increased mutation burden or an ultramutator phenotype (12, 16–18). In addition, the increased expression of immune checkpoint genes is also associated with this mutator phenotype, making these tumors a suitable target for immunotherapy. There are other known clinically relevant germline point mutations, e.g., POLE L424V, which have been shown to be associated with predisposition to multiple cancers and an increased mutation burden (19, 20).

Commonly used oncogenic driver mutation models (e.g., Kras-driven lung cancer models) do not present with high TMB levels that are exhibited by human cancers, making them unsuitable for tumor-immune interaction studies (21). Interestingly, in these models, patients with KRAS-mutant NSCLC also show minimal responses to ICB (22, 23). Therefore, there is a need for appropriate model systems that correctly represent the TMB seen in human cancers as mutational patterns, both clonal and subclonal, have shown to impact tumor biology as well as responses to chemotherapy and immunotherapy (24). Given this background, we set out to create genetically engineered mouse models that recapitulate high mutation burdens that are inherent to many human cancers and test whether these models also respond favorably to ICB. To achieve this, we introduced point mutations in the Pold1 and Pole exonuclease domains in C57BL/6J mice.

The Pold1 and Pole replication polymerases are responsible for lagging and leading strand synthesis, respectively (25). Loss of exonuclease activity leads to increased errors during replication, resulting in increased numbers of mutations during each cell division (26). In the current study, we created mice models with germline Pold1 D400A, Pole D272A E274A, Pole L424V, and conditional Pole V411L point mutations. These mice developed a spectrum of spontaneous tumors, mainly thymic and splenic lymphomas and lung and intestinal tumors, consistent with previously published Pold1 and Pole mutant models (27–29). Exome sequencing of these tumors from these mouse models revealed a highly variable but elevated TMB with mutational signatures similar to that seen in POLD1 and POLE mutant human cancers. Treatment of these elevated mutation burden mouse models with ICB caused regression of only some tumors that was insufficient to provide a survival advantage. In addition, we also bred Pold1 and Pole mutant mice with a LSL-Kras G12D/+ p53 flox/flox conditional model for NSCLC, which interestingly did not show a particularly high TMB. However, single-cell passaging of tumor-derived cell lines from these models modestly increased the mutation burden and immunogenicity, leading to tumor-immune rejection. These findings are indicative of the generation of immunogenic mutations with efficient immune editing during tumor evolution in vivo that is bypassed by introducing mutations in vitro. Finally, in contrast to treating mutator mice with ICB once tumors arose, which was ineffective, treating mutator mice with ICB earlier and prior to the appearance of obvious tumors delayed cancer onset, leading to improved survival, suggesting the use of ICB in high-risk settings for cancer prevention.

Materials and Methods

Mice

All mice maintenance, breeding, and experiments were approved by the Rutgers University Institutional Animal Care and Use Committee. Pold1 D400A mutant mice were created by electroporation of C57BL/6J zygotes with a mixture of 50 ng/µL Cas9 protein (Millipore Sigma), 0.6 pmol/µL each crRNA [spacer sequence CCC​GAG​AGA​TGA​GGT​ATG​GG; NCBI GRCm38.4 Chr7:44540539-44540558(+)] and tracrRNA (Millipore Sigma), and 50 ng/µL ssODN donor (5′TCT​GAA​GAG​AAG​GTA​GCT​GAG​GTA​AAG​ACA​GCC​TGG​TTG​GCC​TTT​GTG​TCC​GTC​CCC​TCC​ACA​GGC​CTG​GGC​CGA​CTT​CAT​CCT​TGC​CAT​GGA​CCC​TGA​CGT​GAT​CAC​CGG​CTA​CAA​CAT​TCA​GAA​CTT​TGc​aCT​CCC​ATA​CCT​CAT​CTC​TCG​GGC​ACA​GGC​CCT​AAA​GGT​GAG​GGA​AGC, lower case is sequence change; Integrated DNA Technologies, Ultramer). Intact electroporated zygotes were transferred on the same day into pseudopregnant mice who were allowed to give birth to potential founders. Four founder mice had the D400A change determined by Sanger sequence analysis. PCR primers to amplify the region are POLDA 5′TTG​CAA​GTG​CGG​AGG​TTG​TCT​TGG and POLDB 5′CTG​TCC​ACA​GCG​ACG​AAT​TTC​CG.

Pole D272A E274A mutant mice were created by electroporation of C57BL/6J zygotes with a mixture of 50 ng/µL Cas9 protein (Millipore Sigma), 0.6 pmol/µL each crRNA [spacer sequence AGG​GAA​TTT​GAG​AGG​CAG​TT; NCBI GRCm38.4 Chr5:110294546-110294565(−)] and tracrRNA (Millipore Sigma), and 50 ng/µL ssODN donor (5′GGC​TGA​ATA​TGG​GTG​TCC​AAA​GGA​TGT​ATC​CAG​ACC​TGA​ATT​CTC​CGT​ATA​TTT​TGT​CCT​TTT​AGG​ACC​CTG​TGG​TTT​TGG​CAT​TTG​cCA​TCG​cGA​CGA​CtA​AAC​TGC​CTC​TCA​AAT​TCC​CTG​ATG​CTG​AGA​CCG​ATC​AGA​T, lower case is sequence change; Integrated DNA Technologies, Ultramer). Intact electroporated zygotes were transferred on the same day into pseudopregnant mice who were allowed to give birth to potential founders. Three founder mice had both D272A and E274A changes determined by Sanger sequence analysis. PCR primers to amplify the region are POLEA 5′TGC​GGT​ACT​GGT​GAG​TGA​ACC​TAG and POLEB 5′TCA​GAA​GGC​AGA​TGC​AGG​AGA​ACC.

Pole L424V mutant mice were created by electroporation of C57BL/6J zygotes with a mixture of 50 ng/µL Cas9 protein (Millipore Sigma), 0.6 pmol/µL each crRNA [spacer sequence TGTGGGCAGTCATAATCTCA; NCBI GRCm39 Chr5:110444899-110444918(+)] and tracrRNA (Millipore Sigma), and 50 ng/µL ssODN donor (5′GTCCTCAGGGTCCAGCTCTACAGGGTCATAGCCAAGTTTGGCCTTGGCAGCTGCtTTaAcATTATGACTGCCCACAGGAAGGTAACTGTCCCTCTTCACCCACCTGGAAAAAACACAGATTC-3′, lower case is sequence change; Integrated DNA Technologies, Ultramer). Intact electroporated zygotes were transferred on the same day into pseudopregnant mice who were allowed to give birth to potential founders. One founder with the L424V was determined by restriction digest with MseI (New England Biolabs) and Sanger sequence analysis. PCR primers to amplify the region are POLEC 5′ ACCTGAGAGTCAAGTGGAGGCTC and POLED 5′ GAAAAGCTAGTACCCATCTCAGGTAAC.

LSL-Pole V411L mutant mice were generated by microinjecting into pronuclei of C57BL/6J (RRID:IMSR_JAX:000664) zygotes a mixture of 50 ng/µL Cas9 protein (Millipore Sigma), 0.6 pmol/µL single-guide RNA [with spacer sequence AGT​GGA​GGC​TCA​AGT​GGC​AT (Millipore Sigma); NCBI GRCm39 Chr5:110444754-110444773(+)], and 10 ng/µL of donor plasmid DNA. The plasmid construct consisted of 4 kb 4×PolyA-stop-cassette (17) flanked with approximately 1.4 Kb of homology arms on each side; however, the 3′ arm contains the modified exon 13 with V411 codon replaced by CTT codon for leucine.

Injected zygotes were transferred on the same day into pseudopregnant mice who were allowed to give birth to potential founders. Five founder mice were found with the cassette properly targeted into the locus as determined by long PCR reactions; three of those founders were confirmed by Sanger sequencing to carry, along with the cassette, the V411L mutation determined by Sanger sequencing analysis. PCR primers to amplify the region encompassing the mutation are P286RC 5′AAT​TCC​GCA​AGC​TAG​CCA​CC and POLEF 5′ AGC​TCT​ACA​GGG​TCA​TAG​CCA​AG.

All mice strains are deposited to the Jackson Laboratory repository with Jax stock no. 040149 - 040152 for Pold1 D400A, Pole D272A E274A, Pole L424V, and LSL-Pole V411L/+ resepctively.

Tamoxifen injections

The Pole V411L/+ allele was induced by intraperitoneal tamoxifen injections by activating Cre in Ubc-CreERT2/+; LSL-PoleV411L/+ mice. Tamoxifen (Sigma-Aldrich, T5648) was dissolved at the concentration of 20 mg/mL in 98% sunflower seed oil and 2% ethanol. Ten µL/g body weight was intraperitoneally injected once per day for four consecutive days.

The mice were allowed to recover from tamoxifen treatment for a week, and samples were taken for genotyping to confirm the Pole V411L/+ allele activation. The PCR primers used to confirm activation were POLEC 5′ ACC​TGA​GAG​TCA​AGT​GGA​GGC​TC and POLED 5′ GAA​AAG​CTA​GTA​CCC​ATC​TCA​GGT​AAC.

Whole-exome sequencing analysis

Fresh-frozen mouse tissues for tumor and corresponding normal samples were collected at fixed or humane endpoints. The fresh-frozen tissues were submitted to Azenta Life Sciences for whole-exome sequencing (WES). Samples were collected for sequencing on the Illumina platform with a 2 × 150 bp configuration. Sequence quality control was conducted using fastQC v0.12.1 (RRID: SCR_014583), and adapter removal and read trimming were performed using Trimmomatic v0.39 (RRID: SCR_011848) in paired-end mode (30). Subsequently, sequences were aligned against the mouse reference GRCm38 using BWA-MEM 0.7.17 (Li H, 2013). Qualimap 2.3 (SCR_001209; ref. 31) was used to examine sequencing alignment and facilitate the quality control of the data in SAM/BAM format. SAMTOOLS (GigaScience; RRID: SCR_002105) was used to sort and index the BAM files. Base recalibration was performed using GATK 4.2.5 (32) with known-site dataset from the Mouse Genomes Project (RRID: SCR_013645; https://www.sanger.ac.uk/data/mouse-genomes-project/). The final BAM files were also marked as duplicated, sorted, and indexed using GATK tools. GATK Mutect2 (RRID: SCR_001876) was used for somatic mutation calling with a panel of normal and tumor–normal paired mode as described previously (Cararo Lopes and colleagues, DOI: https://doi.org/10.1101/2024.05.28.596339; ref. 33), and mutect-filter in GATK Mutect2 was used for an initial filter on the variants. A hard filter was also performed using bcftools v1.11-3 (GigaScience), with tumor AD (allelic depths for the ref and alt alleles in the order listed) ≥3, DP [approximate read depth (reads with MQ = 255 or with bad mates are filtered)] ≥10 for tumor and normal both, and tumor AF (allele fractions of alternate alleles in the tumor)≥0.1. SnpEff 5.2 (RRID: SCR_005191; ref. 34) was used for annotating and predicting the effects of genetic variants on genes and proteins. To remove germline mutation artifact in this pipeline, we filtered recurrent mutations that occurred in more than one sample within a group. Recurrently mutated genes were identified using SnpSift v5.0e (RRID: SCR_015624; ref. 35) and Linux awk/grep commands. Python 3.9.12 (RRID: SCR_024202) and RStudio 2022.12.0+353/R 4.2.1 (RRID: SCR_000432) were used for data visualization.

For mutational signature analysis, single-nucleotide variants (SNV) and insertions and deletions (indels) were counted using Linux commands awk, sed, and grep and bcftools v1.11-3. By considering the pyrimidines of the Watson–Crick base pairs, we counted six different possible substitutions: C>A, C>G, C>T, T>A, T>C, and T>G using Linux command awk/grep. To further characterize the mutation signatures, we used Catalog of Somatic Mutations in Cancer (COSMIC) mutation signatures v3.4 (https://cancer.sanger.ac.uk/signatures/; RRID: SCR_002260) and SigProfiler Bioinformatic Tools, SigProfilerExtractor v1.1.23. (RRID: SCR_023121)

De novo germline mutation calling

A cohort-based pipeline employing GATK v4.3.5 HaplotypeCaller with the GRCm38 reference was used for de novo germline mutation calling on individuals. For cohort population integration, GATK VariantRecalibrator and applyVQSR were applied separately under SNP and INDEL modes. Private variants from distinct groups were extracted by utilizing bcftools v1.11-3. Final filters used are cohort DP >250 and AF ≥0.4.

Copy-number variant analysis

The R package CopywriteR (36) v2.22.0 (RRID: SCR_025864) utilizing paired tumor samples with their respective normal controls was used for copy-number variant (CNV) analysis. Segmentation was performed using a 20k bin size based on the reference GRCm38, which divides the genome into regions of consistent copy-number alterations. Subsequently, CNVs were called based on those segmented data using log₂-transformed ratio thresholds.

Adeno-Cre virus infection

LSL-Kras ^G12D/+ , p53^flox/flox (KP), KP-Pold1 D400A/D400A, KP-Pole D272A/D272A, KP-Pole L424V/L424V, and KP-Pole V411L/+ mice were anaesthetized and infected intranasally with adenovirus-expressing Cre recombinase (University of Iowa Adenoviral Core) at 4 × 10⁷ plaque-forming units per mouse when the mice were 8 to 10 weeks old to induce lung tumors.

Tumor-derived cell line generation and single-cell passaging

Lung tumors were isolated from mice infected with adeno-Cre, dissociated, and allowed to grow in RPMI media containing 10% FBS and 1% penicillin–streptomycin solution. For single-cell passaging, the cells were serially diluted in 96-well plates to select single-cell clones. Wells with single colonies were trypsinized, and the collected cells were grown in 6-well plates until confluent. This was considered as one passage, and the process was repeated for 14 generations. The tumor-derived cell lines (TDCL) were authenticated using exome sequencing and tested negative for Mycoplasma contamination when stained with DAPI (4'6-diamidino-2-phenylindole).

Incucyte growth assays

∼20,000 cells were plated in RPMI media supplemented with 10% FBS and 1% penicillin–streptomycin solution in 24-well plates. The cells were allowed to grow for 4 days and monitored for percent confluence using Incucyte live cell imaging and analysis system.

Flow cytometry

Lung tumors from KP models were harvested at 12-week time point and were homogenized in RPMI medium supplemented with 10% FBS (Gibco) using gentleMACS Octo disscociator (Miltenyi Biotec) according to the manufacturer’s protocol. Immune cell collection and cell surface immunostaining was performed with anti-CD3 antibody (1:300 dilution, clone 17A2, 56-0032-82, Invitrogen; RRID: AB_529507). Data were obtained using an LSR-II flow cytometer (BD Biosciences) and analyzed with FlowJo software (Tree Star; RRID: SCR_008520).

IHC

Mouse normal and tumor tissues were fixed in 10% formalin overnight and were stored in 70% ethanol before paraffin embedding. For antibody staining, the paraffin cut tissue sections were deparaffinized with xylene and ethanol, rehydrated, and boiled for 30 minutes in 10 mmol/L citrate buffer (pH 6.0). The sections were blocked with 10% goat serum in PBS for 1 hour at room temperature. Primary antibody incubation (anti-CD3 antibody, Abcam, Ab16669 1:100; RRID: AB_443425) was performed overnight at 4°C. The next day, anti-biotinylated secondary antibody incubation was done for 20 minutes at room temperature followed by 3% hydrogen peroxide for 5 minutes and horseradish peroxidase streptavidin solution (SA-5704, Vector Laboratories; RRID: AB_2336692) for 15 minutes. Next, the slides were washed and developed with DAB (Vector Laboratories) according to the manufacturer’s instructions. The sections were counterstained with hematoxylin, dehydrated, and mounted with Cytoseal mounting media (Thermo Fisher Scientific). The images were taken at 60× using a Nikon Eclipse 80i microscope, and at least 10 images per condition were quantified for percent positive cells.

ICB antibody treatments

Pold1 mice with tail tumors were treated with either 250 µg of control IgG (Bio X Cell #BE0089; RRID: AB_1107769) or anti-PD-1 antibody (RMP-14 clone, Bio X Cell #BE0146; RRID: AB_10949053) every 5 days. Pold1 and Pole mutant mice were treated with either 250 µg of control IgG (Bio X Cell #BE0089; RRID: AB_1107769) or anti-PD-1 antibody (RMP-14 clone, Bio X Cell #BE0146; RRID: AB_10949053) and 200 µg of either control IgG (Bio X Cell #BE0089; RRID: AB_1107769) or anti-PD-1 antibody (29F.1A12 clone, Bio X Cell #BE0273; RRID: AB_2687796) once every 2 weeks for their lifespan.

Statistics

All statistical analyses were performed using GraphPad Prism v9.5.0. software (RRID: SCR_002798). The sample size was chosen in advance depending on the available number of mice for each experiment. Mice of both genders were allocated randomly to experimental groups, and no statistical methods were used to predetermine sample sizes. Healthy 8- to 10-week-old mice with the right genotype were included in the treatment experiments. All experiments were conducted with biological replicates, and no data were excluded. The investigators were not blinded during experiments and outcome assessment.

Data availability

WES data generated in this study are publicly available with the BioProject ID PRJNA1123738. The American Association for Cancer Research (AACR) GENIE data analyzed in this study were obtained from AACR Project GENIE at https://www.aacr.org/professionals/research/aacr-project-genie/.

The Cancer Genome Atlas (TCGA) data analyzed in this study were obtained from cBioPortal for CANCER GENOMICS under TCGA PanCancer Atlas Studies at https://www.cbioportal.org/.

Patients with POLD1 and POLE exonuclease mutations were selected, and detailed clinical information is available upon request. All other raw data are available upon request from the corresponding author.

Results

Germline and conditional mouse models with Pold1 and Pole exonuclease mutations have elevated mutation burden and develop spontaneous cancers

Germline Pold1 D400A, Pole D272A E274A, and Pole L424V and conditional Pole V411L/+ point mutations reside in exonuclease domains of Pold1 and Pole (Fig. 1A), and mouse models with these mutations were created using a knock-in CRISPR-Cas9 system. We also created a conditional mouse model for Pole V411L, which is a commonly observed pathogenic somatic mutation associated mainly with endometrial carcinomas in humans (12). Pold1 D400A and Pole D272A E274A were previously published (27, 37), but these models were not studied in the context of immunotherapy response and were unfortunately no longer available. For germline models, heterozygous cohorts of mice were bred to generate homozygous animals. The homozygous breeders were then used to generate experimental mice. For survival analysis, wild-type (WT), heterozygous, and homozygous progenies from the heterozygous breeding were monitored for their lifespan. The progenies from the homozygous animals were not used further for breeding to avoid the accumulation of mutations over generations. The mice were genotyped, and sequencing analysis was done to confirm the presence of point mutations (Supplementary Fig. S1A–S1C). The conditional LSL-Pole V411L/+ mouse model was bred with Ubc-creERT2/+ mice to generate Ubc-creERT2/+; LSL-Pole V411L/+ mice in which the whole-body Pole V411L allele is induced after tamoxifen injection (Supplementary Fig. S1D). Mice were genotyped again to confirm the induction of the Pole V411L allele, and it was verified to be induced throughout the whole body (Supplementary Fig. S1E).

Figure 1.

Pold1 and Pole homozygous mutant mice develop spontaneous tumors and other cancers. A, Schematic representation of mouse Pold1 and Pole genes with exonuclease and polymerase domains with locations of mutations germline Pold1 D400A, Pole D272A E274A, and Pole L424V and conditional Pole V411L with an LSL cassette. Exo, exonuclease. B, Kaplan–Meier survival curve analyses of old-generation Pold1 D400A and new-generation Pold1 D400A_10G mutant mice. The color key represents types of tumors detected. P value was calculated using the log-rank test. C, Survival analyses of germline Pole D272A E274A and Pole L424V and conditional Pole V411L mice. The Pole L424V/+ showed a cancer phenotype like Pole L424V homozygous mutant mice. The survival analysis of Pole V411L/+ is plotted as survival percentage against days elapsed after tamoxifen injections. The color key represents types of tumors detected. P value was calculated using the log-rank test. (The survival analyses for Pold1 D400A and Pole D272A E274A were performed together and compared with wt and have been separated for presentation purposes). A, Created in BioRender. Sawant, A. (2024) https://BioRender.com/l97n652.

Both Pold1 and Pole exonuclease domain mutant mice, with increased error-prone mutagenesis, develop spontaneous malignancies and die of cancer. The majority (∼70%) of the spontaneous tumors seen in these models were thymic and splenic lymphomas, which often infiltrated the lungs and liver followed by intestinal and lung tumors and few cases of pancreatic tumors and sarcomas. Seventy-five percent mice from the first cohort of survival analysis of Pold1 D400A homozygous mice presented with tail skin tumors, which were rarely seen in the Pole models (Fig. 1B; Supplementary Fig. S1F). A change in survival profile and tumor incidence was observed in later generations in Pold1 mice as they were bred consistently for ∼10 generations (Fig. 1B). These Pold1 D400A_10G mice that are approximately 10 generations from the original Pold1 D400A mice had earlier onset of tumors, with majority of these early tumors being thymic lymphomas that reduced survival (Fig. 1B). Comparison of exome sequencing of DNA from normal tissue from Pold1 D400A_10G mice and their ancestral Pold1 mice revealed 113 additional mostly missense mutations and 14 genes with truncating mutations in the germline (Supplementary Table S1A). Three COSMIC cancer-associated genes (Alk, Kmt2d, and Spen) have missense mutations, with Spen having dense missense mutations on the same allele (Supplementary Table S1B). Thus, the accelerated cancer onset may be attributed to genetic anticipation observed in Pold1 D400A mutant mice that were bred over time, which was characterized by acquisition of new cancer-related mutations, reduced latency, and early appearance of thymic lymphomas with successive generations.

The heterozygous Pold1 D400A/+ and Pole D272A/+ mice rarely showed a cancer phenotype, and their lifespan was comparable with WT mice (Fig. 1B; Supplementary Fig. S1G). Pole D272A E274A homozygous mutant mice showed a median survival of 494 days and died predominantly from lymphomas (Fig. 1C). The Pole L424V point mutation has been found both as a germline as well as somatic mutation in patients (19, 38). Both heterozygous and homozygous germline Pole L424V mutant mice died of cancer and had a median survival of 502 and 378 days, respectively (Fig. 1C; Supplementary Fig. S1F). The conditional LSL-Pole V411L/+ mice also succumbed to several spontaneous tumors after tamoxifen injection, mainly lymphomas, intestinal tumors, pancreatic tumors, and some skin tumors, and had a median survival of 346 days (Fig. 1C; Supplementary Fig. S1F). Thus, all models of Pold1 and Pole proofreading mutations showed accelerated tumorigenesis and reduced survival. The three Pole proofreading mutation models showed increased but variable tumorigenic potential as observed by the differential median survival (Fig. 1C).

The mutational landscapes of Pold1 and Pole mutant tumors in mice and humans are similar

To determine the mutation burden, a set of tumors arising in Pold1 and Pole mutant backgrounds were collected and subject to WES along with histology (Supplementary Fig. S2A), as described previously (Cararo Lopes and colleagues, DOI: https://doi.org/10.1101/2024.05.28.596339). WES analysis included three intestinal tumors, four tail tumors, four thymic lymphomas, and one lymphoma from Pold1 D400A homozygous mutant mice; four intestinal tumors, three lung tumors, and two lymphomas from Pole D272A E274A homozygous mutant mice; two intestinal tumors, two lung tumors, one lymphoma, three sarcomas, and one thymic lymphoma from Pole L424V/L424V mice; five intestinal tumors and one sarcoma from Pole L424V/+ mice, and four intestinal tumors, six lymphomas, one thymic lymphoma, and one pancreatic tumor from Pole V411L/+ mice. The total mutation frequency, including SNVs and indels, was increased in Pold1 and Pole mutant mouse tumors (Fig. 2A). In Pold1 D400A mutant mice, tail tumors had the highest mutation burden with more than 300 mutations per Mb of exome, followed by thymic lymphomas and intestinal tumors. Pold1 D400A lung tumors had the lowest mutation burden (Fig. 2A). In Pole D272A E274A mutant mice, intestinal tumors and lymphomas had the highest mutation burdens, followed by lung tumors (Fig. 2A). In Pole L424V/L424V mutant mice, intestinal tumors, lung tumors, and lymphomas all had high mutation burdens followed by sarcomas (Fig. 2A). In Pole L424V/+ mutant mice, intestinal tumors had the highest mutation burden followed by sarcomas. In Pole V411L/+ mice, intestinal tumors, lymphomas, and a pancreatic tumor all had high mutation burden (Fig. 2A). Thus, all four models confirm that these mutator alleles increased TMB. The vast differences in TMB across different tumor types may reflect differences in the number of DNA replication cycles the tumor-initiating cell underwent prior to clonal expansion of tumor growth and establishment of the truncal mutation burden.

Figure 2.

The mutational landscape of Pold1 and Pole mutant tumors was similar in mice and humans. A, Tumor mutation frequency was determined from WES analysis of spontaneous tumors from Pold1 D400A, Pole D272A E274A, Pole L424V, and Pole V411L/+ mice. B, Number of indels vs. SNVs in spontaneous tumors from Pold1 D400A, Pole D272A E274A, Pole L424V, and Pole V411L/+ mice in human Pold1 and Pole mutant tumor samples from TCGA and in human Pold1 and Pole mutant tumor samples from AACR GENIE. C, Comparison of mutation signatures using the six-base substitution of tumor types from Pold1 D400A, Pole D272A E274A, Pole L424V, and Pole V411L/+ mice and in human Pold1 and Pole mutant tumor samples from TCGA.

The types of mutations in the mutator mouse tumors were examined further and compared with those in humans. The relative number of indels to SNVs is higher for tumors with Pold1 mutations than Pole mutations in both mouse and human tumors from TCGA and AACR GENIE (17 Pold1 mutant and 427 Pole mutant tumor samples; Fig. 2B). Similar results were observed in normal aging tissues of Pold1 and Pole mutation carriers in humans (20). These results are consistent with the role of Pold1 in lagging strand DNA synthesis and generation of mutant Okazaki fragments that may be incorrectly processed and lead to an increased percentage of indels. Next, the six-substitution signature analysis of TCGA and AACR GENIE cohorts confirmed similarities between the mouse and human tumor mutation signatures (Fig. 2C). The COSMIC single-base substitution (SBS) mutational signatures also showed similarity between the mouse and human tumors carrying Pold1 and Pole mutations (Supplementary Fig. S2B and S2C). Both the Pold1 and Pole mutant tumors from mouse and human have SBS1 and SBS5, which are clock-like signatures, as well as SBS15, which is associated with DNA MMR defects. In addition, Pole mutant tumors also have SBS10a and SBS10b mutational signatures that were previously associated with Pole mutations, whereas Pold1 mutant tumors have SBS23 and SBS54 mutational signatures. There are differences in mutational signatures between the mice and human tumors, and they may be attributed to different tumor types compared between the mice (intestinal, lymphoma, tail, and lung) and human tumors (endometrial, breast, colorectal, cervical, esophagogastric, and glioblastoma). The Pole models also showed a conspicuous presence of the SBS28 mutation signature in all tumors. The etiology of this signature is currently unknown (Supplementary Fig. S2B). Thus, the mutational patterns of the polymerase mutator syndromes have similarities across mice and humans.

Examination of the cancer-associated genes most recurrently mutated in the Pold1D400A mutant mice across all tumor types revealed that they were Pten (8/12 tumors) and Brca2, Med12, and Notch1 (7/12 tumors; Supplementary Table S2A). For tail tumors, due to the very high mutation burden, many cancer-associated genes were mutated, including Med12, Cntrl, Bub1b, Brca2, Notch1, Csmd3, Naca, Ranbp2, Fat1, Nbea, Ncor1, Spen, Fat4, Kmt2c, and Kmt2d in all four tail tumors. For thymic lymphomas, the most recurrently mutated genes are Pten, Brca2, and Smarca4 (3/4 tumors), consistent with loss of Pten being an early mutation selected in mouse thymic lymphoma development (39). For intestinal tumors, only Pim1 was found recurrently mutated in 2/3 tumors (Supplementary Table S2A). Across all tumor types, many of the mutations in tumor suppressors (Brca2, Csmd3, Fat1, Ncor1, Spen, Kmt2c, Kmt2d, and Smarca4) are likely passenger mutations because they are heterozygous (allele frequency less than 0.5; Supplementary Table S2B). The most recurrently mutated cancer-associated gene in Pole mutant mice was Lrp1b (25/36 tumors). For intestinal tumors, the most frequently mutated genes were Lrp1b (11/15), Fat4 (11/ 15), Muc16 (10/15), Fas (10/15), and Nbea (10/15). For lung tumors, the most frequently mutated genes were Fat4, Csmd3, Pde4dip, Robo2, Ncor1, Med12, Grin2a, Fbxw7, Msh6, and Arhgef10l (2/5 tumors). For lymphomas, the most frequently mutated genes were Lrp1b (8/9), Csmd3 (6/9), and Fat3 (6/9). For sarcomas, the most frequently mutated genes were Lrp1b, Muc16, Jak1, and Pten (2/4 tumors; Supplementary Table S3A). It is intriguing that Lrp1b being a larger gene was found to be frequently mutated in Pole mutant tumors (25/36 tumors) but not in tumors from Pold1 mutant mice (4/12 tumors; Supplementary Tables S2 and S3A). However, Lrp1b, a purported tumor suppressor, along with tumor suppressors Ncor1, Grin2a, Fbxw7, and Msh6 are likely passenger mutations because they are heterozygous (allele frequency less than 0.5; Supplementary Table S3B). Thus, genomic analysis demonstrated that Pold1 and Pole mutant tumors evolve by accumulating mutations in cancer-associated genes.

Partial response to acute ICB in Pold1 tumors

Next, we studied the response of the Pold1 D400A mutant mouse tumors to acute ICB as these mice develop tumors that are visible and that have a very high TMB at ∼1 year of age. The Pold1 mutant mice with apparent tail skin tumors were treated with either control IgG or anti-PD-1 antibody every 5 days, and tumor size was monitored (Fig. 3A; Supplementary Fig. S3). The red arrows indicate tumors that are shrinking, and the green arrows indicate tumors that have either increased in size or remained stable (Fig. 3A). ICB has shown exceptional responses in some high-TMB cancers. Anti-PD-1–mediated inhibition leads to reduced immune suppression and activation and proliferation of T cells (40). Anti-PD-1 treatment of Pold1 D400A homozygous mutant mice showed a reduction in tumor size compared with control IgG in a subset of tumors, whereas others continued to grow or remained stable (Fig. 3A and B). Tumors with a consistent decrease in size during the treatment period were considered responding tumors. Anti-PD-1 antibody treatment increased the percentage of responding tumors from 2.7% to 32.5% (Fig. 3C). Next, we checked for immune infiltration in the tumor sections from these mice. IHC analysis showed increased CD3 staining in anti-PD-1 antibody–treated tumor sections (mean 6.4% positive cells) compared with the control antibody–treated group (mean 1.7% positive cells), consistent with inhibition of PD-1 signaling and increased T-cell proliferation (Fig. 3D). However, when survival analysis was examined, there was not a significant difference between control and anti-PD-1 antibody-treated groups, in which all the mice succumbed to other cancers, mainly splenic lymphomas, detected when the mice were sacrificed at humane endpoints (Fig. 3E). It is possible that the lymphomas from Pold1 mice were unresponsive to ICB because of their lower mutation burden compared with tail tumors. Alternatively, ICB may be ineffective when only a subset of tumors are responsive independent of TMB and/or when tumor burden is high along with elevation of intrinsic resistance mechanisms. These results suggested that acute ICB treatment when started after the appearance of tumors in the mutator mice increased responsiveness 10-fold, but as this represented only 1/3 of the tumors, it was insufficient to affect overall survival.

Figure 3.

Acute ICB treatment in Pold1 mutant mice tail tumors showed a favorable response in some tumors. A, Gross images of mice tails with tumors treated with either control IgG or anti-PD-1 antibody. Green arrows, tumors that remained stable or increased in size; red arrows, tumors that regressed with treatment. B, The tumor size measurements taken and calculated every 5 days over the course of treatment for control IgG (n = 12 tumors) and anti-PD-1 antibody treatment (n = 17 tumors) groups of n = 5 mice each. C, Percentage of responding tumors was calculated in control IgG–treated (n = 5) and anti-PD-1 antibody–treated (n = 5) mice. *, P < 0.05 was calculated with the Student t test. D, Representative images of IHC for CD3 in tumor sections treated with either control IgG or anti-PD-1 antibody. The percent positive cells are plotted for tumor sections of control IgG (n = 10) and anti-PD-1 antibody treatment (n = 14) groups. ***, P < 0.001 is calculated with the Student t test. Scale bar, 30 μm. E, Survival analyses of Pold1 homozygous mutant mice with visible tail tumors treated with either control IgG or anti-PD-1 antibody. The light gray rectangle represents the treatment window in which treatment was started as the tumors appeared and continued for every 5 days for their lifetime. P > 0.05 was considered nonsignificant and calculated with the log-rank test. The color key represents types of tumors detected. RX, dose given.

Introducing Pold1 and Pole proofreading mutations in autochthonous mouse models for lung cancer only moderately increased the TMB

Genetically engineered mouse models of NSCLC exhibit a low TMB having bypassed the normal mutagenic mechanisms that normally lead to oncogenic mutations and that generate the truncal TMB making them unsuitable for immunotherapy studies (21). In addition, POLD1 and POLE exonuclease domain mutations have been associated with hypermutated phenotype in human NSCLC tumor samples (13). We asked whether introducing Pold1 and Pole mutator alleles in these models would increase TMB and consequently modify these models to be better suitable for studying immunotherapy responses. We utilized the KP model for NSCLC (41) in which the oncogenic Kras^G12D/+ allele is activated and the Trp53 tumor-suppressor gene is deleted, mediated by intranasal delivery of adeno-Cre virus. This results in sporadic induction of oncogenic driver mutations in the lung and spontaneous lung tumor development. We bred KP mice with germline homozygous Pold1 D400A, Pole D272A E274A, and Pole L424V and conditional LSL-Pole V411L/+ mice and aimed to determine the differences in low-TMB mice (KP model) and potentially high-TMB Pol mutator KP mice (Fig. 4A). At the time of Cre-mediated tumor initiation the Pold1 D400A, Pole D272A E274A, and Pole L424V mutations were germline, whereas the conditional Pole V411L/+ allele was induced concurrently with the activation of Kras^G12D/+ and deletion of Trp53 alleles.

Figure 4.

Pold1 and Pole exonuclease mutations in autochthonous NSCLC mouse models moderately increased the TMB. A, Schematic representation of the experiment in which mice were nasally infected with adeno-Cre, resulting in lung tumor formation. The lung tumors were then harvested to generate TDCLs. B, Kaplan–Meier survival curve analyses of KP vs. KP-Pole D272A E274A, KP vs. KP-Pold1 D400A, KP vs. KP-PoleL424V, and KP vs. KP-PoleV411L/+ mice infected intranasally with adeno-Cre. P values were determined using the log-rank test. C, Total mutation frequency from WES of lung tumors from KP, KP-Pold1 D400A, KP-Pole D272A E274A, KP-PoleL424V, and KP-PoleV411L/+ mice at 12-week time point. D, Schematic representation of single-cell passaged LP and HP TDCLs and subcutaneous injection in C57BL/6J and NSG mice. E, Paired analysis of total mutation burden per Mb of exome in LP and HP of KP and KP-Pole D272A E274A cell lines. F, Tumor growth analysis of KP-Pole D272A E274A-low (red) and -high (sky blue) passage cell lines in immunocompetent C57BL/6J and immunocompromised NSG mice. P values were determined using the unpaired Student T test. A, Created in BioRender. Sawant, A. (2024) https://BioRender.com/w75h211; D, Created in BioRender. Sawant, A. (2024) https://BioRender.com/w50c770.

Introduction of Pold1 and Pole proofreading mutations in KP mice did not affect the survival of these mice (Fig. 4B), and no change in lung wet weights and immune infiltration was observed when the mice were euthanized at 12-week time point (Supplementary Fig. S4A and S4B). These data suggested that Pold1 and Pole mutator alleles did not affect tumorigenesis in KP mice. Exome sequencing analysis of lung tumors from KP and KP-Pold1 D400A, KP-Pole D272A E274A, KP-Pole L424V, and KP-Pole V411L/+ mice revealed only moderately increased number of mutations and did not signify a high TMB phenotype that is presented by human NSCLC cases, which have >10 mutations/Mb (Fig. 4C). Whereas the TMB is significantly higher in KP-Pold1 and KP-Pole mice lung tumors compared with KP mice with WT Pol alleles, the increased TMB was modest and much smaller than the spontaneous tumors that developed in the Pold1 and Pole mutator mice models. It is worth noting that spontaneous tumors mostly developed when the mice were 1 to 1.5 years old, whereas tumors progressed in 3 to 4 months in conditional KP models.

Based on the low TMB in the KP tumors with the mutator alleles, we questioned whether lack of high TMB might be the result of immune editing and elimination of immunogenic mutations occurring at a rate sufficient to permit tumorigenesis unobstructed. To address this, we created TDCLs from lung tumors of KP-Pole D272A E274A mice (Fig. 4D). To increase the mutation burden, we planned to passage the cells as single-cell clones for up to 10 to 20 generations. We tested the cells at p14, which had increased the mutation burden from an average of 53 exonic mutations/Mb in low-passage (LP) cells to an average of 70 mutations/Mb in high-passage (HP) cells (Fig. 4E). Expectedly, the passaging of KP cells did not accumulate an increased number of mutations (Fig. 4E). The LP and HP cell lines grew at similar rates in vitro (Supplementary Fig. S4C). The karyotype and copy-number profile of both KP and KP-Pole D272A E274A cell lines also changed from LP to HP cells (Supplementary Fig. S4D). The LP and HP KP-Pole D272A E274A cell lines were then subcutaneously transplanted in immunocompetent (C57BL/6J) and immunocompromised NOD/SCID gamma (NSG) mice. We observed that LP KP-Pole D272A E274A TDCLs readily formed tumors in immunocompetent mice, but the HP KP-Pole D272A E274A TDCLs showed delayed or no tumor growth in immunocompetent mice (Fig. 4F). In contrast, both LP and HP TDCLs readily formed tumors in immunocompromised NOD/SCID gamma mice (Fig. 4F). These results suggest that passaging of KP-Pole D272A E274A TDCLs in vitro in which immune editing did not occur was sufficient to increase the immunogenicity for tumor rejection by the immune system. Altogether, these results suggested that introducing Pold1 and Pole mutator alleles in autochthonous mouse models of lung cancer only moderately increased the mutation burden. However, single-cell passaged KP-Pole D272AE274A TDCLs were able to accumulate additional mutations, resulting in increased immunogenicity and tumor rejection, suggesting mechanisms of immunoediting are at play in vivo.

Prophylactic ICB delays cancer onset and improves survival of mice with Pold1 and Pole proofreading mutations

As acute ICB treatment of Pold1 D400A homozygous mutant mice with overt cancer showed only a partial response in a subset of tumors and no survival advantage, we tested whether an earlier intervention, in which tumor numbers, heterogeneity, and burden are reduced, could improve response. ICB was started when the mice reached adult age at 8 to 10 weeks old and prior to the appearance of overt cancer, and their survival and tumor profiles were monitored. Two clones of anti-PD-1 antibodies, RMP-14 and 29F.1A12 clones, were used for biweekly treatments in all Pold1 and Pole proofreading mutation mouse models. The mouse PD-1 cDNA was used as an immunogen to produce RMP-14, whereas a recombinant PD-1–IgG fusion protein was used to produce 29F.1A12. In vitro binding assays demonstrated that the 29F.1A12 clone had increased avidity compared with the RMP-14 clone (42). The treatment frequency for the cancer prevention strategy was reduced to biweekly treatments compared with the acute treatment of existing tumors (treatment every 5 days) in anticipation that less dosing may be effective with reduced tumor burden, heterogeneity, and progression along with avoiding potential toxicities. The anti-PD-1 RMP-14 treatment in Pold1 D400A homozygous mutant mice improved survival compared with the control IgG–treated group (P = 0.0099 Grehan–Breslow–Wilcoxon test; P = 0.0413; log-rank test; Fig. 5A). The median survival for the control IgG and anti-PD-1–treated groups showed improvement from 133 to 342 days (Supplementary Fig. S5A). The cause of death was malignancies, mainly lymphomas, which were seen earlier, and intestinal tumors and tail tumors, which were seen as the mice got older (Supplementary Fig. S5A). Next, WES of the tumors that appeared later was performed to check for any mutations that are associated with immunotherapy resistance mechanisms. Mutational analysis of the control IgG and anti-PD-1 antibody–treated groups of tumors showed increased TMB for tail tumors but decreased TMB for intestinal and thymic lymphomas in the treated group (Fig. 5B). The most recurrently mutated COSMIC cancer gene in anti-PD-1–treated tumors is Gnas (4/6 tumors) whereas in control antibody–treated tumors is Pten (3/6 tumors; Supplementary Tables S4 and S5), which have been previously implicated in modulating ICB response (43, 44). Also, all six of the anti-PD-1–treated tumors were diploid, whereas the control antibody–treated tumors showed a trend of increased amount of CNVs compared with anti-PD-1–treated tumors, and 2/6 tumors displayed high level of aneuploidy (Fig. 5C and D). Recurrent CNVs exclusive to anti-PD-1–treated tumors were identified. (Supplementary Table S6). Of note, the Apol7c locus is significantly amplified in three anti-PD-1–treated tumors but not in any control antibody–treated tumors and has been found to play a role in antigen cross-presentation by dendritic cells (45). These results suggest that anti-PD-1 treatment may have stronger selection for clones without aneuploidy and fewer numbers of CNVs than in control antibody–treated tumors. The same experiment was performed with the 29F1.A12 clone of the anti-PD-1 antibody; however, a significant improvement in survival as was seen with the RMP-14 clone was not observed (Fig. 5A; Supplementary Fig. S5B). These findings suggest that ICB can have improved efficacy when used early, prior to the obvious appearance of tumors. Why one ICB clone would have superior efficacy than another has been reported, but the mechanism is unclear (42).

Figure 5.

Prophylactic ICB treatment delayed cancer onset in Pold1 and Pole homozygous mutant mice. A, Survival analyses of 8- to 10-week-old Pold1 D400A homozygous mutant mice treated with either control IgG or anti-PD-1 antibody (RMP-14 clone; P = 0.0099 Grehan–Breslow–Wilcoxon test) and with their respective control IgG antibodies (29F.1A12 clone; P = nonsignificant). The color key represents types of tumors detected. The light gray rectangle represents the treatment window in which treatment was started when the mice reached adult age of 8- to 10-week-old and continued every 2 weeks for their lifespan. B, Total mutation frequency from WES of control antibody–treated and anti-PD-1 antibody–treated (RMP-14 clone) Pold1 D400A tumors. C, The quantification of CNVs in control antibody–treated (n = 6) and anti-PD-1 antibody–treated tumors (n = 6). D, CNV profiles of tumors from control IgG–treated and anti-PD-1 antibody (RMP-14)-treated mice. The samples with high level of aneuploidy are highlighted with red rectangles. mad, median absolute deviation. E, Survival analyses of 8- to 10-week-old Pole L424V homozygous mutant mice treated with either control IgG or anti-PD-1 antibody (RMP-14 clone; P = nonsignificant) and with their respective control IgG antibodies (29F.1A12 clone; P = 0.0264 Grehan–Breslow–Wilcoxon test). The color key represents types of tumors detected. Ab, antibody; RX, dose given.

Next, we performed similar experiments with these two clones of ICB antibodies in the Pole L424V mouse model. This germline mouse model is an ideal test group for these preventive studies as this exact mutation is associated with a predisposition to multiple cancers, including polyposis and colorectal cancers in humans (46). We used homozygous mutant mice at 8 to 10 weeks old age and continued biweekly treatments with either control IgG or anti-PD-1 antibody clones (Fig. 5E). In contrast to the Pold1 D400A mouse model, we observed that anti-PD-1 antibody treatment with the 29F.1A12 clone in Pole L424V mutant mice initially delayed cancer onset and improved survival (P = 0.0264 Grehan–Breslow–Wilcoxon test) compared with the RMP-14 clone–treated cohort (Fig. 5B). The mice died of mostly lymphomas and intestinal tumors in these experiments (Supplementary Fig. S5C and S5D). These findings support the superior efficacy of earlier intervention with ICB in cancers produced by polymerase mutator syndromes. They also demonstrate that the elevated TMB produced by polymerase mutator syndromes confers response to ICB. These mouse models provide a foundation for investigating the mechanisms controlling response to ICB and for probing the means to improve effectiveness through combinatorial approaches.

Discussion

In this report, we established mouse models with homozygous point mutations in exonuclease domains of the replicative DNA polymerases Pold1 and Pole, two of which (Pole L424V and V411L) are also seen in humans. These mutations disable the proofreading functions of the polymerase and lead to increased numbers of somatic mutations with each cell division. Pold1 and Pole mutant mice developed spontaneous tumors, with different tissue specificity, suggesting that Pold1 and Pole mutations influenced tissue-specific differential tumor development, as previously observed (27). This may be due to the specific mutational signature and its propensity to mutate oncogenic drivers in a particular tissue and the number of DNA replication cycles of the tissue stem cells or progenitors required to generate these mutations.

Patients with Pole mutations have been reported to be prone to colorectal and endometrial cancers (46). Introducing Pole exonuclease domain mutations in C57BL/6J mice caused them to mainly develop lymphomas and some intestinal and lung tumors shown in this study and others (29, 47). The types of tumors formed seem to be influenced by the mouse background as Pole mutations leading to endometrial tumors with high TMB have been reported (16). However, the mutation signatures reported in mouse models with endometrial tumors did not completely recapitulate genomic profile of human cancers with Pole proofreading mutations (16). The survival analyses and tumor incidence data in the current study showed appearance of multiple tumors in Pole and Pold1 mutant mice, suggesting accumulation of mutations, leading to high TMB and possible acquisition of new driver and passenger mutations (Supplementary Tables S2 and S3). Six-substitution analysis of the tumors from all four of our models indicated that the mutation signatures in mouse tumors were comparable with those in Pold1 and Pole mutant human cancers. In addition, in the Pold1 D400A mouse model, tail skin tumors showed increased numbers of mutations compared with lymphomas and intestinal tumors, suggesting tissue-specific differences in the rates of mutation accumulation or clonality. As mutation frequency depends on the number of rounds of DNA replication, the rates of stem cell divisions and number of cell divisions required to generate a tumor-initiating clone may be responsible for differential mutation burdens detected in tumors originating from different tissues (48).

Pold1 D400A spontaneous tumors showed an increased percentage of indels compared with the Pole mutant spontaneous tumors, which is consistent with the previous data from normal and neoplastic tissue samples from patients with germline POLD1 and POLE mutations (20). These results can be explained by the Pole being involved in the DNA replication of the leading strand whereas Pold1 is involved with the lagging strand in which DNA mismatches induced by mutant POLD1 may impair effective ligation of Okazaki fragments (49). This increased number of indels over SNPs may also contribute toward increasing immunogenicity in Pold1 D400A mice and partial favorable response to immunotherapy (50). Whether the increased percentage of indels in Pold1 mutant mice compared with Pole mice also contribute toward differences in neoepitopes generated and consequently the favorable clone of the anti-PD-1 antibody that worked remains to be understood.

It is interesting to note that introducing Pold1 and Pole proofreading mutations in the KP NSCLC mouse model only moderately increased the mutation burden and did not result in more immunogenic lung tumors. However, single-cell passaging of TDCLs from these tumors showed increased immunogenicity in which the HP cells were rejected in immunocompetent but not immunodeficient hosts. These results point to the possibility of immune editing in vivo in which there is loss of neoantigens due to immunoselection in which the cells continue to accumulate mutations in the absence of the immune system in vitro (51). Furthermore, the presence of germline Pold1 D400A, Pole D272A; E274A, and Pole L424V mutations may develop tolerogenic immune mechanisms that suppress immune response to these “self-antigens” present since birth. The conditional LSL-Pole V411L/+ allele when induced concurrently with Kras activation and p53 deletion only moderately increased the TMB, suggesting that the accelerated tumor growth rates and loss of p53 and subsequent loss of antigen presentation may further support immunosuppressive mechanisms (52). A recent study showed that MMR deficiency, when introduced simultaneously with oncogenic mutations, did not greatly increase clonal TMB or provide any therapeutic benefit with ICB and suggested that increased intratumor heterogeneity is responsible for poor neoantigen expression and reduced immune infiltration (53). However, the timing of MMR defect may also be critical to increase TMB, as the MMR defect may need to occur well before accumulation/introduction of oncogenic mutations to allow increase of truncal mutation burden and thus result in high truncal mutation burden after oncogenic transformation in the subsequent cancers. Low neoantigen expression was also shown to be responsible for T-cell dysfunction and immune evasion in a colorectal cancer model (54). These new data suggest that there might be factors other than TMB alone that modulate the efficacy of ICB. It is still remarkable how efficiently mutator tumors overcome the immune response against them through their ability to acquire advantageous mutation as well as how little impact the occurrence of immunogenic mutations matters once tumors progress.

Some recent studies have shown that in vivo or transplant mouse models with high TMB did not respond favorably to ICB (29, 53) whereas other studies showed a robust response (16, 55). There are considerable differences, however, in the types of models used, types of tumors formed, timeline of tumor development, and the dosage frequency of ICB antibodies. It is worth noting that we observed a similar average of 100 to 250 mutations/Mb of exome in our models that is comparable with previous reports (17, 29). We acknowledge that smaller sample sizes are one of the limitations of our study. We also noticed differences in clones of anti-PD-1 antibodies used. The antibodies used in this study have been shown to have differences in avidity in vitro and may also have differences in the epitopes recognized (42). An intriguing finding from our study is that tumors from control antibody–treated mice showed increased CNVs and aneuploidy, whereas the anti-PD-1 antibody–treated tumors were all diploid. Recent studies have shown that highly aneuploid tumors are associated with markers of immune evasion and respond poorly to immunotherapy (56, 57). Whether there is immune selection pressure and loss of heterogenous populations of aneuploid cells in the presence of continuous anti-PD-1 treatment and whether epigenetic silencing of genes involved in antigen presentation occurs remain to be determined.

With this study, we conclude that exonuclease domain point mutations in Pold1 and Pole generated spontaneous tumors with high mutation burden and recapitulated genomic profiles of human cancers in which ICB response was improved by treating the mutator mice earlier. Thus, ICB is more effective with less tumor burden and earlier in tumor evolution, suggesting a prevention approach for use of ICB for individuals with a known cancer risk. Even though this is a viable perspective, safety of using prophylactic ICB needs further consideration as ICB is known to cause immune-related adverse events in patients. These models will also be beneficial for testing future immune oncology drugs and combination therapies as tumors developed spontaneously with increased accumulation of mutations over time and mimicked genomic features of their human counterparts.

Supplementary Material

Supplementary Figure 1

Supplementary Figure 1 presents sequencing and mutation confirmation, survival analysis and tumor incidence profile.

Supplementary Figure 2

The figure describes the tumor histology, mutation signature analyses from mouse models and patient samples from AACR Genie project.

Supplementary Figure 3

The figure presents the representative images of Pold1 mutant tail tumors treated with immunotherapy.

Supplementary Figure 4

Supplementary Figure 4 presents lung wet weights, immune infiltration, in vitro growth and CNV profiles for TDCLs.

Supplementary Figure 5

The figure presents the tumor incidence and mean survival plots for the Pold1 D400A and Pole L424V mice treated with immunotherapy before observable tumors appeared.

Supplementary Table 1A

Supplementary Table 1A presents mutations exclusive to Pold1 D400A_10G mice compared to ancestral Pold1 D400A mice.

Supplementary Table 1B

Supplementary Table 1B presents cancer associated genes in Pold1 D400A_10G tumor samples.

Supplementary Table 2A

Supplementary Table 2A presents recurrently mutated genes in Pold1 D400A mutant tumor samples.

Supplementary Table 2B

Supplementary Table 2B presents recurrently mutated genes from tumors of Pold1 D400A mice at allele frequency less than 0.5

Supplementary Table 3A

Supplementary Table 3A presents recurrently mutated genes in Pole mutant mouse models

Supplementary Table 3B

Supplementary Table 3B presents recurrently mutated genes in Pole mutant tumor samples at allele frequency less than 0.5

Supplementary Table 4

Supplementary Table 4 presents recurrently mutated cancer genes in anti-PD1 treated Pold1 D400A mutant mice.

Supplementary Table 5

Supplementary Table 5 presents recurrently mutated cancer genes in control antibody treated Pold1 D400A mutant mice.

Supplementary Table 6

Supplementary Table 6 presents recurrent CNVs exclusive to anti-PD1 treated Pold1 D400A mutant tumors.

Authors’ Disclosures

C.S. Hinrichs reports grants from the NCI during the conduct of the study and grants and personal fees from Neogene Therapeutics, personal fees from Capstan Therapeutics, GlaxoSmithKline, Vir Biotechnology, and PACT Pharma, equity from Scarlet TCR, grants from T-Cure Biosciences, and grants (in-kind) from Iovance Biotherapeutics outside the submitted work; in addition, C.S. Hinrichs has multiple patents for immunotherapy technologies, including membrane-tethered IL15 and IL21, pending, issued, licensed, and with royalties paid from multiple companies. J.D. Rabinowitz reports grants from Rutgers CINJ and Ludwig Cancer Institute during the conduct of the study and personal fees and other support from Bantam Pharmaceuticals, Rafael Pharmaceuticals, Empress Therapeutics, Farber Partners, and Raze Therapeutics, other support from Fargo Biotechnologies, and nonfinancial support from Princeton University–PKU Shenzhen collaboration outside the submitted work. S. Ganesan reports grants from the NIH/NCI during the conduct of the study and personal fees and other support from Merck, grants from Gandeeva and M2GEN, and personal fees from KayoTheraa, EMD Serano, Lunit, and Foghorn outside the submitted work. No disclosures were reported by the other authors.

Authors’ Contributions

A. Sawant: Data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. F. Shi: Data curation, formal analysis, methodology, writing–review and editing. E. Cararo Lopes: Investigation, methodology, writing–review and editing. Z. Hu: Investigation, methodology. S. Abdelfattah: Investigation, methodology. J. Baul: Investigation, methodology. J.R. Powers: Investigation, methodology. C.S. Hinrichs: Resources, writing–review and editing. J.D. Rabinowitz: Resources, writing–review and editing. C.S. Chan: Resources, data curation, software, formal analysis, supervision, validation, investigation, methodology, writing–review and editing. E.C. Lattime: Conceptualization, supervision, funding acquisition, writing–review and editing. S. Ganesan: Conceptualization, resources, supervision, funding acquisition, visualization, writing–review and editing. E.P. White: Conceptualization, resources, supervision, funding acquisition, writing–review and editing.