Vaccines | Free Full-Text | Local Enrichment with Convergence of Enriched T-Cell Clones Are Hallmarks of Effective Peptide Vaccination against B16 Melanoma
[ad_1]
1. Introduction
Vaccination with mRNA-encoded antigens is a powerful immunological approach that was significantly boosted during the recent COVID-19 pandemic. Prior to the pandemic, it had been under extensive development for various malignancies, including cancer. In work by the group of Ugur Sahin, who was one of the leading developers of Pfizer’s SARS-CoV-2 vaccine [1], initially, this technology had been developed as a cure for B16 melanoma [2,3]. Shortly after studies in mice and just before the pandemic, it had undergone trials as therapeutics for patients with metastatic melanoma and glioblastoma [4,5]. For metastatic melanoma, this trial demonstrated a 75% remission rate for over 2 years.
Several B16 peptide vaccines are also efficient in the induction of mutation-reactive cytokine-secreting T-cells and in tumor control [2,3]. Peptide vaccines are promising therapeutic agents since they are easier to produce and store compared to mRNA vaccines. However, in spite of their good T-cell induction properties, peptide vaccines often exhibit modest tumor control [6,7]. There are multiple factors that skew T-cells’ responses in certain directions, including reaction to the adjuvant, organ-/tissue-/cell-type targeting, peptide length, TCR affinity, pharmacokinetics, and similarity with autoantigens [8,9]. The complexity of such factors hinders the prediction of the integral response and requires testing in each individual case.
To evaluate the mechanism underlying the efficiency of peptide vaccines, we used neoantigen peptide vaccines, described previously for B16 murine melanoma [2]. The top 50 mutations were selected according to their expression, MHC binding, and functional impact. Then, 27mer peptides with each of these mutations were evaluated for immunogenicity in vitro in comparison with their wild-type counterparts. We selected two of these peptides that demonstrated a maximal response to the mutated epitope with no response to the WT variants. Their epitopes are within Tubb3 (MUT20: G402A, peptide p20) and Kif18b (MUT30: K739N, peptide p30). MUT30 appeared to be the most efficient B16 antitumor vaccine in both the peptide and mRNA forms, while data on MUT20’s effect on tumor growth are lacking [2,3].
Here, we compared two B16 peptide neoantigen vaccines, p20 and p30, discovered in [2]. We observed that the induction of T-cell responsiveness to the neoantigens does not correlate with antitumor activity. Via the in vitro restimulation of cells from vaccinated mice with antigen peptides, we showed that both vaccines elicit T-cell reactivity, but more prominently for p20. However, in the p20-vaccinated mice, the T-cells had a much lower tumor reactivity compared with the p30-vaccinated mice, which correlated with effects on tumor growth. By comparing cytotoxic T lymphocyte (CTL) repertoires, we showed that after p30 vaccination, the CTL clonality increased in the tumor-draining lymph nodes (dLNs), and enriched clones had a branch of similar T-cell receptor (TCR) variants. This is contrary to that of p20, which induced none of these features. This observation supports the consensus that clonality and convergence of the antigen response are key factors in effective antitumor immunity.
2. Materials and Methods
2.1. Mice Vaccination and Tumor Model
The experiments were carried out on transgenic C57Bl/6-FoxP3-EGFP mice (kindly provided by Alexander Rudensky, Sloan Kettering Institute, New York, NY, USA). The transgenic mice were bred against a C57Bl/6 genetic background by knocking in the chimeric construct of the eGPF subcloned into the first exon of the FoxP3 gene [10].
The mice were vaccinated with synthetic peptides p20 (FRRKAFLHWYTGEAMDEMEFTEAESNM) and p30 (PSKPSFQEFVDWENVSPELNSTDQPFL), both from Genscript Biotech B.V. (Rijswijk, The Netherlands). 40 mg/mL peptide stock solutions were prepared in DMSO (Sigma-Aldrich Co. LLC, St. Louis, MO, USA) and stored at −20 °C. Vaccinations were administered 21 and 7 days before lymphocyte isolation or tumor inoculation. Each vaccination was performed with 50 µg of the peptide per mouse in PolyI:C or complete Freund adjuvant (CFA) (both from InvivoGen Inc., San Diego, CA, USA). For the CFA injections, the peptides were dissolved in PBS at 500 µg/mL and mixed thoroughly with CFA at a 1:1 ratio. A total of 200 µL of peptide/CFA emulsion was injected subcutaneously (s.c.) into each flank and into the back at the tail’s base on both sides. For the PolyI:C vaccination, a total of 200 µL of PBS solution with peptide at 250 µg/mL and PolyI:C at 250 µg/mL was injected s.c. into both flanks.
Tumors were generated by a subcutaneous (s.c.) injection of 5 × 104 B16F0 cancer cells in 300 μL of PBS into the left flank. B16F0 melanoma cells were obtained from the ATCC, expanded, and stored aliquoted at −150 °C. Before the inoculation, the cells were thawed and grown for 2 weeks in DMEM medium (Paneco Ltd., Moscow, Russia) supplemented with 10% fetal bovine serum (FBS, Gibco BRL, Grand Island, NY, USA), 0.06% L-glutamine, 50 units/mL penicillin, and 50 μg/mL streptomycin. The cells were incubated at 37 °C and 5% CO2 and passaged 2–3 times per week. Right before the injection, the cells were detached by trypsin, counted, and resuspended at a final concentration of 106 cells in 6 mL of PBS.
2.2. Tissue Processing and In Vitro Restimulation
For evaluation of immune memory, the mice were sacrificed at day 0 with isoflurane (Esteve Pharmaceuticals S.R.L., Milan, Italy), and inguinal lymph nodes were isolated. For in vitro restimulation assays, the lymph nodes were minced several times and placed in a 6-well plate in full RPMI medium with glutamine (Life Technologies Ltd., Paisley, UK) supplemented with 10% defined FBS (USA origin, Sigma-Aldrich Co. LLC, St. Louis, MO, USA), 50 units/mL penicillin, and 50 μg/mL streptomycin. After 4 h of outgrowth, the lymphocytes were collected, passed through a 100 μm cell strainer, washed two times with PBS, and stained with 0.5 μg/mL TagIt-Violet (BioLegend Inc., San Diego, CA, USA) for 10 min at 37 °C. After staining with TagIt-Violet, the cells were washed, resuspended in full RPMI medium, and seeded in a 96-well plate in six wells per mouse at 4–5 × 105 cells/well. Either 40 μg/mL of p20 peptide or p30 peptide or an equal amount of DMSO were supplemented in two wells for each mouse. After 7 days of culture, the cells were collected by triple washing with Versene solution (Paneco Ltd., Moscow, Russia), washed, and stained for flow cytometric analysis.
For the study of immunity after tumor challenge, the mice were sacrificed on day 16 of tumor growth. The lymph nodes were isolated and thoroughly minced, passed through a 100 μm cell strainer, washed twice, and processed for analysis by flow cytometry and sorting.
2.3. Flow Cytometric Analysis
For staining of cultured cells, the pellets were resuspended in 50 μL RPMI medium with the addition of 1 μL of antibodies per sample: CD4-V450 or CD4-BV650 (both from BD Biosciences, San Jose, CA, USA), CD3-APC or IA/IE-Alexa649 (BioLegend Inc., San Diego, CA, USA), CD69-BV510 (BD Biosciences), CD25-BV605 (BioLegend), and CD8-APC/Cy7 (BD Biosciences). Freshly isolated cells were first stained with biotin-labeled B220/CD45R antibodies (1 μL in 50 μL of RPMI per sample) for 1 h on ice, then washed and stained with the panel of labeled antibodies supplemented with 0.5 μL of Steptavidin-PE (Jackson Immunoreseach Europe Ltd., Ely, UK). The cells were stained for 1.5–2 h in ice, diluted with 200 μL of RPMI, and analyzed/sorted without wash. Flow cytometry and cell sorting were performed on a FACSAriaIII cell sorter (BD Biosciences) equipped with 405 nm, 488 nm, 561 nm, and 633 nm lasers using a 70 μm nozzle. For RNA isolation and repertoire analysis, the cells were sorted directly in 100 μL of RLT lysis and storage buffer (Qiagen GmbH, Hilden, Germany).
The T-cell subset composition and the expression of activation markers were obtained from flow cytometric data using FlowJo v.10 software (BD Biosciences). Percentages of T-cell subsets were analyzed and depicted with GraphPad Prism v. 8 software (GraphPad Software Inc., La Jolla, CA, USA). The data were presented as individual points with mean values. A two-way ANOVA with Šídák’s multiple comparison tests was used for statistical inference.
2.4. TCR Library Preparation and Repertoire Sequencing
Total RNA was extracted from RLT solutions using the HiPure Total RNA Kit (Magen Biotechnology Co., Ltd., Guangzhou, China). Next, TCRβ cDNA was obtained from 20–40 ng of eluted RNA by 30 min amplification at 42 °C with primer for TRBV constant region and QScribe III revertase (Gene-quest LLC, Moscow, Russia). The resulting cDNA samples (20 μL) were purified with MagPure A4 magnetic beads (Magen Biotechnology) at a 1:1.5 sample:beads ratio (v:v) and resuspended in 20 μL of 10 mM TE buffer (10 mM Tris/HCL, 1 mM EDTA, pH 8.5). Half of the cDNA was used for multiplex PCR (30 cycles) with forward TRBV-specific primers [11] and TRBC-specific reverse primers. All primers were synthesized by Evrogen (Moscow, Russia). Amplicons were purified using MagPure A4 beads at a 1:1.6 ratio, washed twice with ethanol, resuspended in 50 μL of TE buffer, quantified, and used at 100 ng for library preparation. The libraries were prepared with the MGIeasy FS Library Prep Set (MGI Tech Co., Ltd., Shenzhen, China) according to the manufacturer protocol, starting at the “end repair and A-tailing” stage. The circularized libraries were converted to nanoballs and sequenced with the DNBSEQ-G400 sequencer (MGI) using paired-end 125 + 125 nt reads and about 25 reads per sorted cell. TCR repertoires were extracted from FASTQ reads using MiXCR v.4.5.0. software [12].
2.5. TCR Repertoire Analysis
The repertoires were processed using VDJTools v.1.2.1 [13]. For CTL repertoires, only clones with a frequency above 2.5 × 10−5 were taken, with a rescaling of the frequencies of the remaining clonotypes to give 100% in total. This frequency threshold was chosen so that the total TCR reads divided by the number of reads in the smallest clone (supposed to be the product of a single cell) were below the number of cells in the sample. Diversity statistics and V-segment usage were calculated with the VDJTools.
To identify vaccine-specific clones, all samples for each vaccine were pooled, and clonotypes with cumulative frequencies above 5 × 10−5 were considered further. Unique clonesets were obtained by excluding overlapped clonotypes present in both vaccines. Such overlapped clonotypes constituted 2.1% and 2% within the p20 and p30 groups, respectively.
For the construction of clusters of convergent clonotypes, we used the ALICE algorithm with a random TCRβ background generated with the OLGA model [14,15]. The algorithm was applied to pooled clonesets for each vaccine group using the TCRgraper R library with default parameters for mouse TCRs (https://github.com/KseniaMIPT/tcrgrapher, accessed on 7 March 2024). The clonotype was classified as ALICE hit if it had significantly increased numbers of neighbors (adjusted Benjamini-Hochberg p value below 5 × 10−5). The resulting ALICE clonesets were clustered with the repseq Python library (https://github.com/mmjmike/repseq, accessed on 7 March 2024) either with or without linking to V and J segments. The clusters were exported, visualized, and analyzed with Cytoscape 3.10.1 (https://cytoscape.org/, accessed on 7 March 2024). Consensus amino acid sequences for the clusters were generated using the online WebLogo tool v.2.8.2 (https://weblogo.berkeley.edu/, accessed on 7 March 2024).
Overlapping of the current clonesets with known antigen-specific TCRs was performed with an R script. Each antigen-specific CDR3 was aligned with the current repertoires, allowing a single amino acid mismatch and/or one indel and exact match to the TRBV of a specific TCR. Alignment was performed by the Biostrings package from the Bioconductor project (https://www.bioconductor.org/, accessed on 7 March 2024).
4. Discussion
Immunosuppressive responses to peptide vaccines are not uncommon and may be due to various reasons, including the induction of Treg, Th2, or Th17 cells or the exhaustion of reactive T-cell clones [8,25,32,33,34]. Skewing towards an immunosuppressive response may be due to either adjuvant [35], antigen itself [9], or overstimulation of reactive cells [27,36]. In our study, we initially tried to minimize immunosuppressive response by using CFA that had been shown to favor Tfh response over Tfr as opposed to IFA [33,34,35,36,37,38,39]. CFA was combined with a PolyI:C adjuvant that was shown to provide good activity for antitumor vaccination in similar experimental settings [3,38]. This vaccination scheme was efficient for p30, and thus, it is unlikely that the adjuvants were responsible for the failure of p20 vaccination.
Antigen loss is another possible mechanism for tumor escape. The p20 antigen is likely to be expressed only in a subset of B16F0 cells, which is one of the possible reasons for p20 vaccine failure. However, this escape mechanism would delay tumor growth, which was not observed in our study. Another possibility is that the Tubb3 WT variant is more aggressive, but in this case, it would replace the mutant form during culturing and/or grafting, which is also not the case. However, the possibility of a tumor eliminating targeted mutations argues for the use of vaccinations with antigens that also elicit immune reactivity for wild-type forms of the antigen.
Cell exhaustion and/or apoptosis may be another reason for unresponsiveness to tumor vaccines [36,38,40]. It was shown that CTLs are depleted at sites of antigen persistence after their administration in IFA [36]. In our experiments, we used either a short-lived formulation with water-soluble PolyI:C only or a combination of CFA for the first immunization, followed by a boost with PolyI:C. With the first scheme, we demonstrated tumor tolerance, indicating that it was developed without prolonged antigen persistence at the tumor site. With the second scheme, we compared p20 with p30 and have shown that their differences are not adjuvant-dependent. This does not exclude vaccination-dependent exhaustion or elimination but indicates that it is peptide-specific.
Cells from p20-vaccinated mice have better responsiveness to the antigen in vitro compared to p30. However, in vitro conditions discard many factors that are present in vivo, including those that may shape the immune response and impose anergy on specific clones. Metabolic regulation of Th/Treg balance, disregard of in vivo cytokine milieu and tissue microenvironment, nutrient availability, different compositions, and properties of antigen-presenting cells are among these factors. Several factors contribute to the development of tolerance in the case of p20 compared to p30. (1) Better in vitro responsiveness for p20 indicates a stronger immune response that is more likely to lead to exhaustion and tolerance. (2) For the short 10mer p20 and p30 peptides designed to bind MHC I, the predicted affinity of the p20/MHC complex was 10-times higher than for p30/MHC [41]. This agrees with the stronger response of p20 that we have seen in vitro. Prolonged exposure and better MHC binding can cause tolerance [36,42]. (3) The unmutated form of p20 antigen, tubulin beta 3 class III, is expressed predominantly in nervous tissues, while the parental p30 antigen, Kinesin family member 18B (KIF18B), is present mainly in lymphoid tissues (www.proteinatlas.org, accessed on 7 March 2024). This may lead to preformed immune memory and tolerance for unmutated p30 antigen in vitro and in vivo, with cross-reactivity to p30. This possible natural tolerance restrains p30 reactivity in vitro for LN-isolated cells by KIF18B-specific Tregs, but it is unleashed in tumors due to the absence of these Tregs.
We also observed significantly enlarged lymph nodes in peptide-vaccinated mice, while in adjuvant-only group LNs, they were also enlarged but much less. However, the most abundant were bystander cells characterized by the absence of dLN enrichment or that were found in databases as public or virus-specific TCRs. These clones may have been expanded due to adjuvant stimulation of viral-specific clones and because of the cross-reactivity of antiviral and antitumor clones [43,44].
As in a number of previous studies, we have also encountered difficulties in the detection of antigen-specific clones among high and noisy backgrounds [19,30,45,46,47]. However, the repertoire of CD8+ cells appeared to be more clonal than the repertoire of CD4+ cells [48]. Here, we dissected specific responses using several approaches. We used enrichment in dLN over ndLN as an indication of tumor reactivity. In conjunction with the analysis of TCR clustering, this appeared to be the efficient metric that highlighted differences between p20 and p30 vaccines. For validation of this pipeline, we searched consensus CDR3 sequences in available TCR data. For several dLN-enriched clones, we have found relevant variants that were either induced by anti-CTLA4 immunotherapy or allogeneic skin grafts [19,20]. This confirmed the potential antitumor reactivity of identified sequences in p30-vaccinated mice and validated our identification pipeline.
It should be noted that not all dLN-enriched clones have convergent variants; for example, the allograft skin graft-specific clone CASSDRVEQYF. Such non-convergent clones are highly enriched in dLN after p30 vaccination since clustered clones were skewed towards ndLN. Nevertheless, such individual clones are not convergent, and each individual clone accounts for less than 0.8%; if taken together, they may constitute up to 20% (according to the disbalance in the ALICE clusters). This indicates that either clustering algorithms are not efficient enough in the identification of such clones or vaccination following a tumor challenge induces a non-convergent response. The latter possibility is unlikely because it has been shown that vaccination- and tumor-induced responses are usually polyclonal with shared unrelated reactivities [45,49,50]. These reactivities stem from public clonotypes or some persistence infection- or microbiota-specific clonotypes that are usually polyclonal.
The principles of sharing or spreading immune responses are still not clear, but they appear to be an important component of an effective therapy. Both vaccines were shown to induce CD4+ T-cells primarily [2,3]. Here, we revealed that both CD4 and CD8 responses were activated for both vaccines. Moreover, CTL clonotypes with presumed tumor reactivity in the case of the p30 vaccine are quite diverse (Table 1 and Table S1), making them unlikely to target a single antigen. Such diverse specific clones may appear because of antigen spreading—the phenomenon when an antigen-presenting cell is licensed by the Th cell for activation of other T-cells, reactive to other foreign antigens presented by this cell [51]. Antigen spreading and bystander activation might be important components of an effective antitumor vaccine. The deficiency of these components in vitro does not allow for full therapeutic activity in an in vitro restimulation assay.
[ad_2]
