Background
Immunotherapy with immune-checkpoint inhibitors (ICI) has become a mainstay of treatment for a range of solid cancers, including melanoma, bladder cancer, non-small cell lung cancer and Hodgkin’s lymphoma [
1‐
5]. CTLA-4, PD-1, or PD-L1 are the so far most studied checkpoint molecules and ICI widely applied in the clinic to improve patients’ prognosis. This blockade reactivates exhausted T-cells, prevents T-cell inhibition, and promotes effector T-cell proliferation to stimulate T-cell-mediated tumor cell killing [
6‐
8]. Atezolizumab, Avelumab and Durvalumab are FDA approved as PD-L1 blocking antibodies. Monotherapy results in antitumor immune responses yet have a limited long-term therapeutic efficacy in most cases.
Lessons learned from the last years identified mismatch-repair deficiency (dMMR) as a molecular subtype with high response rates toward ICI. DMMR-driven carcinogenesis emerges sporadically because of MMR gene promoter hypermethylation or as part of defined hereditary tumor syndromes such as Lynch Syndrome and constitutional mismatch-repair deficiency [
9‐
14]. The spectrum of cancer types related to dMMR is complex and includes, among others, gastrointestinal, endometrial and urothelial cancers [
15]. A hallmark of dMMR tumors – irrespective of organ manifestation – is an ultramutated tumor phenotype (= TMB high), leading to a high abundance of frameshifted neo-epitopes on the tumor cells’ surface. This latter feature underlines the tremendous potential for immunological targeting of dMMR cancers [
15‐
17]. Indeed, in 2017, the FDA approved
α-PD-1 ICI Pembrolizumab and Nivolumab for treatment of dMMR cancers agnostic of cancer site [
18], which was extended lately for the first-line treatment of patients with un-resectable or metastatic dMMR colorectal cancer (CRC). Pre-existing Th type1 immune responses and high numbers of tumor-infiltrating CD8
+ T-cell clones (= IFN
γ signature) constitute positive predictive biomarkers [
19]. However, roughly 25% of patients show intrinsic resistance and in most cases initially responding patients gradually develop resistance, highlighting the necessity of improving treatment options [
20‐
23]. As for PD-L1, limited preclinical data exist. PD-L1 expression on tumor-infiltrating lymphocytes is thought to be a potential predictor for patients’ response to
α-PD-1 therapy, but it is not well established for dMMR cancers because of the generally low expression [
24,
25]. A recent phase II study in patients with dMMR metastatic or unresectable CRC revealed antitumor activity of Avelumab monotherapy [
26]. Additional clinical trials are ongoing with different combinations being employed. One of them is based on tumor lysates or specific neoantigen-derived peptides. The former act as “global” vaccines and induce objective responses in some patients. To refine combination approaches preclinically, we employed the Mlh1 knock-out mouse model for dMMR-related diseases. Preceding vaccination approaches yielded prolonged overall survival in the therapeutic and prophylactic setting [
27,
28]. Residual tumor cells showed an upregulation of immune-checkpoint molecules as part of acquired resistance. To counteract vaccination-induced immune escape and improve overall survival, we here applied a murine
α-PD-L1 antibody (clone: 6E11) in combination with repeated vaccination.
Methods
Cell culture & vaccine preparation
Cells were cultured in DMEM medium, supplemented with 10% FCS (fetal calf serum), 6 mM Glutamine, and antibiotics (all from Biochrom, Berlin, Germany). The tumor lysate was prepared from a A7450 tumor allograft as described [
29].
Mlh1−/− mouse model and in vivo treatment protocol
Ethical statement
The German local authority approved all animal experiments: Landesamt für Landwirtschaft, Lebensmittelsicherheit und Fischerei Mecklenburg‐Vorpommern (7221.3‐1‐026/17; -026/17‐3), under the German animal protection law and the EU Guideline 2010/63/EU. Mice were bred in the animal facility of the University Medical Center in Rostock under specific pathogen‐free conditions. Mlh1 genotyping was done according to [
21]. During their whole lifetime, all animals got enrichment in the form of mouse-igloos (ANT Tierhaltungsbedarf, Buxtehude, Germany), nesting material (shredded tissue paper, Verbandmittel GmbH, Frankenberg, Germany), paper roles (75 × 38 mm, H 0528–151, ssniff‐Spezialdiäten GmbH), and wooden sticks (40 × 16 × 10 mm, Abedd, Vienna, Austria). During the experiment, mice were kept in type III cages (Zoonlab GmbH, Castrop‐Rauxel, Germany) at 12‐h dark:light cycle, the temperature of 21 ± 2 °C, and relative humidity of 60 ± 20% with food (pellets, 10 mm, ssniff‐Spezialdiäten GmbH, Soest, Germany) and tap water ad libitum.
Experimental protocol
Mice with PET/CT proven gastrointestinal tumors (GIT), located in the duodenum, were conducted to therapy using four weekly tumor lysate boosts. Vaccination was sustained (10 mg/kg bw, biweekly, n = 10 mice) until tumors progressed, but for a maximum of 12 times. Treatment with α-PD-L1 (clone 6E11, kindly provided by Genentech, a subsidiary of Roche, South San Francisco, USA, dissolved in PBS) given at 2.5 mg/kg bw intravenously was done once (n = 4 mice) or thrice (n = 10 mice) every second week (q2wx3). Mice receiving the combination of α-PD-L1 were given vaccine first, followed by α-PD-L1 injection. Here again, combinations included single or triple α-PD-L1 applications (n = 10 mice/group; q2wx1 and q2wx3). Control mice were left untreated (n = 10 mice). Reduction of suffering was guaranteed by providing daily prepared soaked pellets, twice-daily monitoring of the health status using a score sheet and by applying humane endpoints (weight loss > 15%, pain/distress, changes in social behavior). All mice were sacrificed before they became moribund to prevent pain and distress. At this time, blood samples, spleens, lymph nodes and GIT were removed for further analyses.
PET/CT imaging
PET/CT imaging scans were performed on a small animal PET/CT scanner (Inveon PET/CT, Siemens Medical Solutions, Knoxville, TN, USA) according to a standard protocol as described before [
30]. Briefly, mice were anesthetized by isoflurane (1–3%, supplemented with oxygen) and received a mean dose of 16.03 ± 1.10 MBq
[18F]FDG intravenously via a microcatheter placed in a tail vein. Static PET scans were acquired using a small animal micro PET/CT scanner (Inveon PET/CT Siemens, Knoxville, TN, USA). The PET image reconstruction method consisted of a 2-dimensional ordered subset expectation maximization algorithm (2D-OSEM) with four iterations and six subsets. Attenuation correction was performed on the basis whole body CT scan and a decay correction for [
18F] was applied. PET images were corrected for random coincidences, dead time and scatter. By marking the entire tumors, starting at the edge and cutting through the whole
[18F]FDG-enriched tumor, volumes and SUVs were determined. This was done by using Inveon Research Workplace 4.2 software.
Immune phenotyping
Blood samples were taken routinely from the retrobulbar venous plexus. Single cell suspensions of spleens and GIT were obtained upon passing them through a cell strainer (100 µm). Samples (2 × 105/Well) were stained with a panel of conjugated monoclonal antibodies (mAb, 1 μg each) followed by lysis of erythrocytes (155 mM NH4Cl (MERCK Millipore, Darmstadt, Germany), 10 mM KHCO3 (MERCK Millipore) and 0.1 mM EDTA (Applichem, Darmstadt, Germany). Negative controls consisted of lymphocytes stained with the appropriate isotypes (Biolegend, San Diego, USA). Cells were washed, resuspended in PBS and analyzed by flow cytometry on a Flow Cytometer (BD FACSVerse™, BD Pharmingen). Data analysis was performed using BD FACSuite software (BD Pharmingen).
Procartaplex cytokine assay
Cytokine levels in plasma samples were determined according to the manufacturer’s instructions of the Procartaplex™ multiplex immunoassay (Thermo Fisher Scientific, Schwerte, Germany). Measurement as well as cytokine quantification was performed on a Bioplex 2000 (Bio-Rad Laboratories GmbH, Munich, Germany) in combination with the BioPlex Manager Software. Absolute plasma cytokine and chemokine level are presented [ng/ml].
Fragment length analysis of cMS target genes
A panel of non-coding and coding MS marker was analyzed as described before [
31]. MSI is defined by mono- and/or bialellic band shifts usually characterized by deletions (indicated with minus symbol + number).
Nanostring targeted gene expression profiling
The T cell–inflamed tumor microenvironment was analyzed by targeted gene expression profiling of tumor RNA from fresh frozen or Tissue-Tek® embedded treatment and control samples (n = 3 samples/group). Total RNA was isolated using the RNeasy Mini Kit according to the manufacturers’ instruction (Qiagen, Hilden, Germany). Total RNA concentrations were measured using the NanoDrop ND1000 (Thermo Fisher Scientific). Gene expression analysis was conducted on the NanoString nCounter gene expression platform (NanoString Technologies, Seattle, WA) applying the PanCancer IO 360™ Panel. This panel enables digital profiling of 770 genes that shape the tumor-immune interface and allows for characterization of pathways relevant in immune response and escape. Quality control, normalization and data analysis was done by applying the nSolver™ Analysis Software 4.0 including nCounter Advanced Analysis (version 2.0.115). Data are presented as Heatmap and log10 (p value) as well as log2 fold change.
Immunofluorescence
Cryostat sections of 4 μm were air-dried and fixed in cold pure methanol for 8 min. Unspecific binding sites were blocked in 2% BSA (Roth) for 2 h followed by incubation with 1 μg of the following FITC- and PE-labeled mAbs: CD4, CD8α, CD11b, Gr1 (Immunotools, Friesoythe, Germany), CD11c, CD104, LAG-3, PD-1, F4/80 and PD-L1 (Biolegend). Sections were washed, embedded in Roti Mount Flour Care DAPI (Roth, Karlsruhe)and target proteins visualized on a confocal laser scanning microscope (LSM780, Zeiss, Jena, Germany) using 20× objectives.
IFN-γ ELISpot
2.5 × 103 targets/well (2 GIT cell lines: Mlh1−/− A7450, Mlh1−/− 328, 1 lymphoma cell line: Mlh1−/− 1351, and YAC-1 cells) were seeded in IFNγ–specific mAb (Mabtech, 3321–3)–coated, 96-well microtiter plates. Peripheral blood leukocytes (5 × 104/Well) or splenocytes (1 × 104/well) from vaccinated and control mice were added in triplicates and co-cultured overnight. Bound antibody (Mabtech, 3321–6) was visualized by BCIP/NBT (KPL, Gaithersburg, Maryland, USA); spots were counted using an ELISpot reader. Presented are the numbers of IFNγ–secreting cells per 10,000 effector cells corrected for background levels counted in the absence of target cells, which was always ≤ 5 spots/well. Target cells without effector cells showed no background level.
Statistics
All values are expressed as mean ± SD. After proving the assumption of normality (Kolmogorov–Smirnov test), differences between vaccinated and control mice were determined using the unpaired Student’s t test or one-way ANOVA (Bonferroni or Dunnett’s multiple comparison). Kaplan–Meier survival analysis was done by applying the log rank (Mantel Cox) test. Statistical analyses were performed using GraphPad Prism 5 (San Diego, CA). The criterion for significance was set to p < 0.05.
Discussion
In this study, we describe a strategy to combine active tumor vaccination with an ICI in a clinically relevant dMMR mouse model [
32]. DMMR is associated with high tumor mutational burden [
33‐
35] and thus harbors a tentatively high likelihood of being susceptible to immunotherapy.
Using a murine
α-PD-L1 antibody, monotherapy itself marginally improved outcome after single application. By increasing the number of injections, overall survival of Mlh1
−/− mice extended to a degree comparable to the vaccine monotherapy. The latter was prepared from a whole tumor lysate with proven antitumor activity from previous studies [
27,
36]. Hence, both treatments prolonged mice’ survival suffering from highly aggressive Mlh1
−/−-driven GIT. Given the fact that Mlh1
−/− tumors, despite their high TMB, do not have a high IFN
γ signature and are not targetable by ICI per se, the improved outcome after
α-PD-L1 monotherapy is intriguing. It is therefore unlikely that mice’ outcome after targeting the PD-L1 axis is better if
α-PD-L1 antibodies are applied more often or over a longer time. Rather targeting both MHC-I and II restricted tumor epitopes—with whole tumor lysates—in combination with PD-L1 blockade seems necessary to affect growth of poorly immunogenic and thus ICI refractory, immunologically cold/warm tumors, as recently shown for triple-negative breast cancer [
37]. So far, we can only speculate on the survival benefit of mice treated with the
α-PD-L1 antibody in monotherapy. In a very recent study on dMMR gastric cancer, CD68
+CD163
− M1-like macrophages were identified as prerequisites for efficient PD-L1/PD-1 blockade because of specific chemokine receptor expression likely activating CTL [
38]. The
α-PD-L1 antibody itself may have also induced immune-independent apoptosis and autophagy in Mlh1
−/− cells. In addition to our RNA expression data, showing signaling pathway alteration, increased release of reactive oxygen species and cytochrome-c was found in atezolizumab-treated osteosarcoma cells, ultimately leading to mitochondrial-related apoptosis [
39].
Another interesting finding of our study was the upregulation of angiogenesis pathways under
α-PD-L1 monotherapy, adding further credence for combined checkpoint-angiogenesis inhibition, currently tested in clinical trials [
40,
41].
However, dosing schedules and accurate timing of each combination partner remain undefined for combined vaccine and ICI strategies. Here, we performed alternating treatment starting with vaccine first. The rationale is based on our previous observations in which repetitive vaccine monotherapy provoked upregulation of immune-checkpoint-molecules on residual tumors [
27]. To counteract therapy-induced upregulation, we here applied
α-PD-L1 therapy during vaccination. This combined treatment yielded complete remission in 30% of mice, finally resulting in significantly improved overall survival. Although complete remission was not achieved in all mice, we would like to stress the point that tumor burden massively reduced in the combination likely because of inducing a T cell–inflamed tumor microenvironment. Other studies reported superior effects when checkpoint-inhibition was given after cessation of the vaccine [
42]. Still, the significantly prolonged overall survival of MLH1
−/− mice achieved in this study argues in favor of concomitant application. By applying dual immune-checkpoint blockade (such as
α- or
α-LAG-3) one may expect even better and long-term tumor growth control.
Most previous trials focused on
α-PD-1 antibodies to increase antitumoral effects of vaccine-induced immunity [
43‐
45]. Rare preclinical data exist on vaccine-
α-PD-L1 combinations. A recent study described prolonged survival and increased tumor cell apoptosis in a hepatocellular carcinoma model treated with a combined DC vaccine and
α-PD-L1 inhibitor [
46], supported by findings from Ji et al., reporting reactivation of neoantigen-specific CTL by combined
α-PD-L1 peptide vaccination [
47]. Likewise, Sun et al. found enhanced tumor-antigen-specific immunity upon combined vaccine-PD-L1-blockade [
48]. By reversing the immunosuppressive status of the micromilieu, PD-L1 is indeed a promising target. Here, we also identified a shaped tumor microenvironment accompanied by peripheral immune activation. By performing a detailed and longitudinal analysis, we found decreased numbers of circulating MDSC and T cell exhaustion markers after combined treatment. Accompanying in-depth gene expression analysis of residual tumors identified increased numbers of total TIL, mainly being cytotoxic T and B cells. Vice versa, levels of exhausted CD8
+ T cells, tumor-associated macrophages and neutrophils reduced in the combination group. Neutrophils are a group of tumor-associated cells which, in conjunction with MDSC, play a major role during cancer development and progression. Their specific location within the tumor (i.e., intra-, peritumoral or stromal) has prognostic relevance [
49]. Abundance of tumor-associated neutrophils may even correlate with local TGFβ expression; in fact, TGFβ blocking improves outcome in preclinical cancer models [
49]. In support of this, TGF-signaling was downregulated here upon combination and likely facilitated conquering primary resistance to checkpoint inhibition [
50]. Though not analyzed in detail here, reduced TGFβ signaling may have also exerted a tumor-intrinsic effect finally blocking the EMT-like transition and preventing Mlh1
−/−-driven tumor progression [
50]. Additional common pathways with prognostic relevance that were altered by the vaccine-
α-PD-L1 combination include PI3K/Akt and Wnt-signaling as well as genes responsible for angiogenesis, matrix remodeling and metastasis. By contrast, genes belonging to the JAK/STAT signaling were upregulated, indicative for enhanced immune-related crosstalk to eradicate Mlh1
−/− tumor cells via IFN-
γ [
51]. These cumulative data nicely explain the improved overall survival in mice treated with the combined vaccine-
α-PD-L1 approach.
While most pronounced effects were in fact seen in the combination therapy and thus interpretable as synergistic, the monotherapy itself modulated the tumor microenvironment. Anti-PD-L1 treatment-induced genes relevant for autophagy and downregulated NF-κB-signaling, which is in line with data from a recent trial on triple-negative breast cancer cells, treated with Atezolizumab [
52]. Upon vaccination, matrix remodeling/metastasis-related genes and genes of the Wnt-and TGF-signaling were downregulated and a direct indicator of successful reversal of intrinsic resistance. Indeed, tumor-intrinsic
β-catenin activation prevents T cell priming and infiltration into the tumor microenvironment and results in resistance to anti-PD-L1/anti-CTLA-4 therapy [
53]. Vice versa, Wnt-pathway suppression restores DC infiltration, a phenomenon seen here upon therapy characterized by elevated levels of tumor-infiltrating CD11c
+ DC that confirm successful therapy-related downregulation of the Wnt-pathway.
Inter-individual differences throughout the treatment groups reflect the different overall survival times of mice. Here, short-term survivors had low TIL scores and vice versa. Teasing out what are the (patient-) individual baseline differences is the challenge for the next wave of pre- and clinical trials with immunotherapy to refine treatment on the long run.
Another interesting finding was the altered molecular profile in typical cMS marker upon treatment. One may speculate that treatment successfully eliminated single mutated clones, whereas other emerged under the immune-selective pressure. We identified somatic cMS mutations in NKtr1 and Kcnma1 in all treatment groups that were infrequent in control tumors. By contrast, somatic mutations in Spen, Apc, and Casc3 were no longer detectable. Notably, residual tumors from the combination therapy harbored the lowest mutation frequencies in Akt3, Clock, Il1F9, and Rfc3, especially compared with α-PD-L1 treatment (= 100% mutation frequency).
Among others, question remains why some tumors regressed, while others finally progressed. Sustained tumor IFN signaling induces PD-L1 expression on tumor and immune cells and is considered a acquired resistance mechanism [
54]. However, this only partially explains the different in vivo response. Reports from human dMMR CRC describe contradictory PD-L1 abundance on tumor-infiltrating lymphocytes or tumor cells [
55,
56]. Its role to mediate immune escape is undebatable and results from a phase II study already confirmed antitumor activity of Avelumab with manageable toxicity in most, but clearly not all patients with previously treated dMMR mCRC and recurrent/persistent endometrial cancer [
26,
57]. Heterogeneity among tumors, such as the varying TMB, different genomic variations (in cMS), Indoleamine 2,3-Dioxygenase 1-based immune escape, and the activated Wnt/
β-catenin signaling may provide an explanation for the difference seen here. Understanding how Mlh1
−/− tumor and immune cells react to our treatments holds promise for novel immune-modulating strategies and will hopefully help to guide the way for clinical vaccine-based immune-checkpoint regimens.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.