Introduction
Approximately 10
14 microbes are believed to be present in the human gastrointestinal tract. This corresponds to the number of cells and a DNA content 1,000 and 10,000 times greater than that in the human body, respectively [
1]. The composition of the gut microbiota can vary during life owing to changes in diet, lifestyle, and habits. However, 90% of the species colonizing the gut microbiota belong to the
Firmicutes and
Bacteroidetes phyla [
2,
3].
The bacteria that form the microbiota play a key role in human health. They are essential for intestinal digestion, prevention of pathogenic bacterial invasion, and regulation of the immune system [
4,
5]. In addition to their physiological roles, the microbiota is actively involved in human diseases [
6,
7], including tumor development and responses to treatments [
8,
9]. Such a role has been mainly attributed to the production of specific metabolites, which may influence the genesis and development of cancer, as well as regulate the innate and adaptive immune responses [
10‐
15].
A different perspective on the role of microbiota in cancer development and evolution is provided by the immunological mechanism based on the “molecular mimicry”. The latter is considered the major mechanism underlying immune disorders [
16]. In particular, gut microbiota dysbiosis has been implicated in the activation of pathogenic T-cell responses, leading to gut-distal autoimmune diseases [
17]. Activation of diabetogenic CD8
+ T cells by molecular mimicry between microbial antigens of the gut microbiota and pancreatic islet autoantigens supports the evidence that cross-reactive CD8
+ T cells can be elicited at the gut level with effects at distant sites [
18]. Similarly, epitopes derived from microbiota (MoAs) may mimic tumor-associated antigens (TAAs) if they share identical or structurally similar amino acid residues at the same position along the epitope sequence. Therefore, the presentation of TAA-like MoAs to the immune system, in the context of MHC class I/II molecules, would elicit CD4
+/CD8
+ T cells cross-reacting with TAAs presented by tumor cells [
19,
20].
Sporadic evidence for homology between MoAs and TAAs, together with T-cell cross-reactivity, has been previously reported [
21‐
24]. Very recently, our group performed an unprecedented extensive analysis and found that sequence homology between TAAs and peptides from microbiota species of the Firmicutes and Bacteroidetes phyla is a frequent finding [
25]. Most MoAs – TAAs paired epitopes share 6–7 identical residues or conservative substitutions along the sequence, with limited impact on the charge of the peptide. Strikingly, three of these pairs had identical sequences. Furthermore, the paired TAAs and MoAs are characterized by highly similar or even identical structural conformations, especially in the core TCR-facing residues with identical planar and dihedral angles. Finally, the areas of interaction with both HLA and TCR mostly match, suggesting that the paired peptides can be recognized by cross-reacting T-cells [
25]. This may strongly influence the fate of tumor progression and provide a novel set of antigens for the development of next-generation anti-cancer therapeutic vaccines [
26].
The present study shows that circulating CD8+ T cells that react with a large array of previously undescribed MoAs can be identified in both HS and CP. In addition, reactivity against TAAs was also observed in healthy individuals, suggesting previous priming by similar MoAs. Interestingly, CD8+ T cells cross-reacting with MAGE-A1 and paired MoAs were identified in three subjects.
Materials and methods
Peptide identification and epitope prediction of TAAs
BLAST homology search and MoAs epitopes prediction
Epitope modelling and molecular docking
The structural conformation of the predicted epitopes bound to HLA was evaluated by modifying the peptide included in the crystallized structure of HLA-A*02:01 deposited in the Protein Data Bank (
https://www.rcsb.org). Briefly, the 1AO7 complex (PDB
https://www.rcsb.org/structure/1AO7), which includes the HTLV-I LLFGYPVYV epitope crystallized with the HLA-A*02:01 molecule, and the α and β chains of TCR and β2-microglobulin were used as templates. The sequence of the peptide bound to MHC was modified and replaced with the selected nonamers using PyMol software (PyMol Molecular graphics system, version 1.8.6.2). The modified structure was then visualized using the Molsoft Mol Browser (version 3.8-7d).
Samples collection
Peripheral blood was obtained from 15 cancer patients (5 hepatocellular carcinoma, 8 lung cancer, and 2 colon cancer with liver metastasis) and 10 healthy subjects. All samples were processed at the National Cancer Institute in Naples under informed consent, as approved by the Institutional Review Board. Fresh human peripheral blood mononuclear cells (PBMCs), isolated by density gradient centrifugation using Ficoll-Hypaque, were cryopreserved at −150 °C in FBS (Gibco, Thermo Fisher Scientific) plus 10% DMSO until analysis.
DNA-barcoded pMHC-multimer library preparation
All peptides were synthesized with a purity of ≥ 90% (GenScript, Piscataway, NJ, USA). The lyophilized powders were reconstituted according to the manufacturer’s instructions.DNA barcoded multimer libraries for selected peptides were generated as previously described by Bentzen et al. [
27]. Briefly, individual peptide–MHC (pMHC) complexes were generated by incubating for 1 h with 200 μM of each peptide and 100 μg/mL of HLA-A*02:01 MHC molecules using direct peptide loading [
28]. The pMHC monomers were then coupled to a phycoerythrin (PE)- for TAAs peptides, or allophycocyanin (APC)- for MoAs-derived peptides, conjugated dextran backbone DNA barcode-labelled. Unique DNA-barcoded multimers were used to detect pMHC-specific T cells.
Staining of antigen-specific T cells with DNA-barcoded pMHC multimers
PBMC from both cohorts were thawed and washed twice in RPMI1640 medium (Fischer Scientific 72400047) and 10% fetal bovine serum (FBS, Fischer Scientific 16140071). cells were then washed once in barcode cytometry buffer (BCB; PBS + 0.5% BSA + 100 mg/mL herring DNA + 2 mM EDTA) and incubated with DNA barcoded pMHC multimers for 15 min at 37 °C, followed by incubation at 4 °C for 30 min with CD8-BV480 (BD 566121) and dump channel antibodies CD4-FITC (BD 345768), CD14-FITC (BD 345784), CD19-FITC (BD 345776), CD40-FITC (Serotech MCA1590F), CD16-FITC (BD 335035), and a dead cell marker (LIVE/DEAD Fixable Near-IR, Invitrogen 2451278). The cells were washed twice with BCB, fixed in 1% paraformaldehyde (PFA), washed twice more, and resuspended in BCB. Cells were then acquired on a flow cytometer (AriaFusion, BD Biosciences); APC-pMHC multimer and double-positive PE/APC-pMHC multimer-binding CD8 + T cells were separately sorted (Suppl. Fig.
1). Sorted cells were centrifuged for 10 min at 5000 × g and the cell pellet stored at -20 °C.
DNA-barcode sequence analysis
DNA barcodes from the isolated cells, as well as from an aliquot of the original multimer pool (10,000 × final dilution in the PCR reaction; used as a baseline) were amplified using a Taq PCR Master Mix kit (QIAGEN 201443) and 3 µL of forward and reverse primer (LGC Biosearch Technologies). Purified products (QIAquick PCR Purification Kit) were sequenced using PrimBio (PA, USA). DNA barcode sequencing data were processed using Barracoda software package2 (
https://services.healthtech.dtu.dk/service.php?Barracoda-1.8). This tool identifies the DNA barcodes used in an experiment, assigns a sample ID and pMHC specificity to each barcode, calculates the number of reads and clonally reduced reads for each pMHC-associated DNA barcode, and includes statistical data processing. Fold change (FC) in read counts mapped to a given sample relative to the mean read counts mapped to triplicate baseline samples was estimated using normalization factors determined by the trimmed mean of M-values method. P-values were calculated by comparing each experiment individually to the mean baseline sample reads using a negative binomial distribution, with a fixed dispersion parameter set to 0.1. False discovery rates (FDRs) were estimated using the Benjamini–Hochberg method described by Bentzen et al. [
27]. At least 1/1,000 reads associated with a given DNA barcode relative to the total number of DNA barcode reads in that given sample were set as the threshold to avoid false-positive detection of T cell responses. DNA barcodes with FDR < 0.1% (corresponding to
p < 0.001) and Log2FC > 2 over the baseline values for the total pMHC library were considered significant and true T cell responses. The T cell frequency for each significantly enriched barcode was calculated from the percentage read count of the barcode relative to the percentage of CD8+ multimer+ T cells. A non-HLA-matching healthy donor sample was included as a negative control, and any T cell recognition determined in this sample was removed from the full dataset to exclude potential non-specific pMHC binding to T cells.
T cell staining with pMHC tetramers
Specific matched peptides (TAA/microbiota) with a T cell response detected using DNA-barcode labelled multimers were selected to generate combinatorial fluorescently labelled pMHC tetramers [
29,
30]. Single-fluorochrome pMHC tetramers were produced by conjugating individual pMHC complexes generated as described above to a library of fluorophore-labelled streptavidin (SA) molecules, including PE(Biolegend 405204), APC (Biolegend 405243), PE-CF594 (BD 562284), PECy7 (Biolegend 405206), BV421 (BD563259), and BV650 (BD 563855). pMHC molecules were incubated with their respective SA-conjugated fluorochromes for 30 min at 4 °C, followed by incubation with D-biotin (Sigma) (25 μM final concentration) for 20 min at 4 °C. pMHC tetramers for each specificity were generated in two colors and mixed at a 1:1 ratio before staining the cells.
PBMCs were thawed and washed with R10 + 10% fetal FCS. Cells were incubated with desatinib (50 nM final concentration) and 1 μL of pooled pMHC multimers per specificity for 15 min at 37 °C in 80 a total volume. cells were then mixed with 20 μL antibody staining solution containing CD8-BV480 (BD B566121) (final dilution 1/50), dump channel antibodies (CD4-FITC (BD 345768; final dilution 1/80), CD14-FITC (BD 345784; final dilution 1/32), CD19-FITC (BD 345776; final dilution 1/16), CD40-FITC (Serotech MCA1590F; final dilution 1/40), CD16- FITC (BD 335035; final dilution 1/64)), and a dead cell marker (LIVE/DEAD Fixable Near-IR (Invitrogen L34976; final dilution 1/1000)) and incubated for 30 min at 4 °C. Cells were washed twice in FACS buffer (PBS + 2% FCS) and acquired on an LSRFortessa flow cytometer (BD Biosciences).
In vitro pre-immunization
To confirm the presence of cross-reacting CD8+ Tcells and their increase after a re-stimulation, PBMCs were cultured in presence of SSX2-BACT2 and SSX2-BACT3 peptides. Cells were plated at a density of 2 × 106cells/mL in 3 mL of complete medium in a 6 well plate and stimulated with peptides at a final concentration of 10 uM in presence of 10 U/mL of IL-2 (Sigma) and 25 µL/mL of ImmunoCult™ Human CD3/CD28 T Cell Activator (StemCell technologies). After 5 days, cells were harvested, centrifuged at 1200 rpm for 5 min and stained with single-fluorochrome pMHC tetramers, generated as described above, and incubated with desatinib (50 nM final concentration) and 1 μL of pooled pMHC multimers per specificity (SSX2-PE; BACT2/BACT3-FITC) for 15 min at 37 °C in 80 a total volume. Cells were then mixed with CD8 PE-Cy7 (Life Technologies) and CD3 superbright 436 (Invitrogen) and incubated for 30 min at 4 °C. Cells were washed twice in FACS buffer (PBS + 2% FCS) and acquired on an AttuneNxT flow cytometer (LifeTechnologies).
Interferon-gamma detection
PBMCs from three healthy HLA-A02:01 positive subjects were cultured in RPMI 1640 (Gibco) supplemented with 2 mM L-Glut (HyClone), 10% human serum (Sigma-Aldrich), 100 IU/ml penicillin and 100 μg/ml streptomycin (Capricorn). Cells were maintained at 37 °C in a humidified incubator with 5% CO2. PBMCs were seeded at 2.5 × 106 cells/ml in 3 ml in a 6 well plate and cultured in presence of IL-2 (Sigma) at a final concentration of 10 U/mL and 25 µL/mL of ImmunoCult™ Human CD3/CD28 T Cell Activator (StemCell technologies). Following 3 days incubation, the interferon-gamma (IFN-γ) production was evaluated through the IFN-γ Secretion Assay –Cell Enrichment and Detection Kit (Miltenyi Biotec). Briefly, cells were harvested, centrifuged and incubated 4 h at 37 °C with SSX2, SSX2-BACT2 and SSX-BACT3 peptides at a final concentration of 10 uM. Unstimulated and PHA stimulated PBMCs were used, respectively, as negative and positive controls. Subsequently, cells were washed and stained with IFN-γ Catch Reagent, incubated 45 min at 37 °C, centrifuged and labelled with IFN-γ Detection Antibody (PE), CD8 PE-Cy7 (Life Technologies) and CD3 super bright 436 (Invitrogen). After 15 min incubation on ice, cell were washed, resuspended in 500uL of cold buffer and analysed by flow cytometry (AttuneNxT-LifeTechnologies).
Flow cytometry analysis
All flow cytometry data were analyzed using the FlowJo data analysis software (version 10.8.1; FlowJo LLC). For antigen-specific T cell identification using combinatorial pMHC tetramer staining, we gated on single, live, CD3+, CD8+ lymphocytes and selected cells positive in two tetramer colors and negative in the remaining colors. For the IFN-γ detection, cells were gated as single, CD3+, CD8+ lymphocytes and double positive to CD8+ and IFN-γ.
Cytotoxicity assay
Cytotoxic T Lymphocytes (CTLs) were generated from HLA-A*02:01 normal donor peripheral blood mononuclear cells (PBMC). PBMCs from four healthy HLA-A02:01 positive subjects were cultured in RPMI 1640 (Gibco) supplemented with 2 mM L-Glut (HyClone), 10% human serum (Sigma-Aldrich), 100 IU/ml penicillin and 100 μg/ml streptomycin (Capricorn). Cells were seeded at 2 × 106 cells/ml in 3 ml in a 6 well plate in presence of IL-2 (Sigma) at a final concentration of 10 U/mL and 25 µL/mL of ImmunoCult™ Human CD3/CD28 T Cell Activator (StemCell technologies). PBMCs were stimulated with 10ug of SSX2-BACT3 peptide each 3 days for 5 times, cells without peptide were used as baseline control.
For cytotoxicity assay, T2 cells (174 × CEM.T2 CRL-1992-ATCC) were loaded with SSX2; SSX2-BACT2 and SSX2-BACT3 peptides at a concentration of 50 uM, incubated O/N at 27 °C, 2 h at 37 °C and with 1X Brefeldin A for 1 h and co-cultured with stimulated PBMCs for 5 h in a Target: Effector (T:E) ratio of 1:5. Specific cytotoxic activity was evaluated with Cell-mediated Cytotoxicity Assay kit (Immunochemistry Technologies).
Data processing and statistical analysis
T cell recognition data determined by DNA-barcoded pMHC multimer analysis and all peptides with negative enrichment were set to LogFC equal to zero. GraphPad Prism6 was used to generate box plots, and related statistical analysis was used to visualize the flow cytometry data. For statistical analysis, data were assumed to have a non-Gaussian distribution, and non-parametric tests were used.
Discussion
A high number of microorganism-derived antigens (MoAs) showing sequence and conformational homology with tumor-associated antigens (TAAs) have been recently reported, and their implication in eliciting cross-reacting anti-cancer T cells has been proposed [
25,
35,
36]. In the present study, we aimed to confirm that MoAs predicted from extracellular bacteria that form the microbiota are recognized by CD8
+ T cells. Consequently, the presentation of such MoAs in the context of MHC class I molecules is reasonable. Moreover, the cross-reactivity of CD8+ T cells against MoAs and homologous TAAs was investigated in patients with HS and tumors.
The selection of homologous MoAs and TAAs was based on previous observations by our group, and for each TAA, 3–5 corresponding MoAs derived from
Firmicutes and
Bacteroidetes phyla were chosen [
25]. All selected TAAs belong to the cancer testis (CT) subgroup. In particular, antigens belonging to the Melanoma Antigen Gene family (MAGE-A1, MAGE-A3, MAGE-A3/12, MAGE-A10, MAGE-C1, MAGE-C2) and SSX2 are found to be broadly expressed in many tumor types [
37,
38].
When compared to the corresponding TAA, each MoA showed a similar, if not higher, predicted affinity to the HLA molecule, despite 1–2 amino acid differences. Nevertheless, the consensus sequence derived from all selected MoAs was identical to the corresponding TAA. Conformation analyses revealed highly overlapping structures between homologous TAAs and MoAs, with indistinguishable contact areas with both HLA molecules and TCR α and β chains. Overall, this strongly suggests the induction of CD8+ T cells cross-reacting with TAAs and MoAs. A few exceptions to this general observation have been found, especially in the residues interacting with TCR α and β chains, when the substituting amino acid residue in the MoA was of a different chemical/structural group.
T cell binding screening was based on a panel of DNA-barcoded peptide-major histocompatibility complex (pMHC) multimers (HLA-A*02:01), including all seven selected TAAs and 53 homologous MoAs, together with 64 peptides derived from common viruses. Unstimulated CD8+ T cells from subjects in both experimental groups showed a high level of reactivity to MoAs, which was significantly higher in CP than in HS (95.54% vs. 85.18%). In contrast, reactivity to TAAs and cross-reactivity to MoAs and TAAs were significantly higher in HS than in CP (10.75% vs. 2.19%; 4.07% vs. 2.28%, respectively). Such unexpected observations could be reasonably explained by the priming of HS by MoAs, eliciting a T cell response that cross-reacts with the corresponding TAAs.
The evaluation of specific MoAs bound by CD8
+ T cells showed unique patterns when sorted by single staining (T cells binding only MoAs) or double staining (T cells binding MoAs and TAAs). Single-stained (SS) T cells were found to bind MoAs homologous to MAGE-C2 and MAGE-A3/12 TAAs, in particular C2-FIRM3 (20/25 samples) and A3/12-BACT1 (12/25). Double-stained (DS) T cells, instead, were found to bind MoAs homologous to MAGE-A1 and SSX2 TAAs, in particular A1-FIRM4 (6/25 samples) and SSX2-BACT2 (7/25). The C2-FIRM3 ALKDVEEPV peptide is derived from the AMP-binding proteins of Eubacterium sp. and Clostridia bacterium. The A3/12-BACT1 FLWGSIALV peptide is derived from the cation-translocating P-type ATPase of the Bacteroidales genus. The A1-FIRM4 KVLEYIIKI peptide is derived from the ATP-binding protein of the genus Tissierellales. Finally, the SSX2-BACT2 KAYEKIFYV peptide was derived from an alpha/beta hydrolase of the Bacteroidaceae genus. None of these peptides were present in the Immune Epitope Database and Analysis Resource (iedb.org), representing the newly identified microbiota-derived MHC class I-associated epitopes. The striking consistent T cell reactivity for the C2-FIRM3 and A3/12-BACT1 peptides is likely explained by the presence of Eubacterium sp. and Bacteroidales bacterium in the universal microbiota phylogenetic core, independent of lifestyle and country of origin [
39]. In particular, the Eubacterium spp. populations in the gut has been shown to be positively correlated with the Mediterranean diet [
40].
Double-stained (DS) T cells were found to bind essentially MAGE-A1 TAA (8/25), and three subjects (H-010, T-001, and T-006) showed T cells binding both MAGE-A1 and the corresponding MoAs. Such a result may have a significant impact on a large spectrum of cancer subtypes. Indeed, MAGE-A1 is overexpressed in a significant percentage (≥ 20%, on average) of different tumor types, including colon [
41], melanoma [
42], and lung [
43], as well as in a low percentage (~ 10%) of breast [
44] and liver cancers [
45]. Similarly, the anecdotal observed T-cell cross-reactivity against SSX2 (T-004) or MAGE-C1 (H-004) and their homologous MoAs may be highly relevant. Indeed, SSX2 and MAGE-C1 are overexpressed in various cancers [
39,
46]. We further showed that circulating T cells primed by MoAs were recalled and expanded by an in vitro immunization protocol. Such T cells reacted against TAA in a tetramer-staining analysis producing a relevant increased levels of IFN-γ. Furthermore, PBMCs ex vivo activated with the SSX2-BACT3 peptide showed a comparable cytotoxic activity against TAP-deficient T2 cells loaded with either the same peptide or the homologous SSX2 TAA or SSX2-BACT2 peptides. These results provided the conclusive proof that, indeed, T cells activated by a MoA cross-react with an homologous TAA, exerting a cytotoxic killing activity on target cells expressing the TAA.
Overall, T cell cross-reactivity against TAAs elicited by homologous MoAs may represent a potent immunological shield against a broad spectrum of cancers that can prevent tumor growth in healthy subjects or improve clinical prognosis in cancer patients. To this end, it is unfortunate that the three CP showing cross-reactive T cells in the present study were lost to follow-up, and information about clinical progression was not available.
The functional analysis in a preclinical model will definitely demonstrate the anti-tumor effect of the described cross-reactive T cells.
In conclusion, the data described provide the first large report of several MoAs, some of which have not been reported before, homologous to TAAs recognized by T cells, and cross-reactivity was observed in both HS and CP. Further studies on larger numbers of HS and CP patients will provide validation with a high potential impact on cancer immunotherapy. Indeed, non-self MoAs would become a key tool for developing preventive/therapeutic “multi-cancer” vaccine strategies with much stronger immunogenicity compared to the corresponding self-TAAs.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.