Antigenic molecular mimicry in viral-mediated protection from cancer: the HIV case

Starting from downloadable data section of Human Protein Atlas database (https://​www.​proteinatlas.​org/​about/​download), Normal Tissue and Pathology dataset were downloaded. They contain expression profiles evaluated by immunohistochemistry on tissue micro array of proteins of different human tissues samples, non tumoral and tumoral respectively.

Normal Tissue dataset is an extremely wide list containing information about the expression of several proteins in different cell types of a specific human tissue. To identify tumor specific proteins, only those proteins who were defined as “not detected” in normal cells from the same tissues were selected. Subsequently, these proteins were searched in the Pathology dataset to select only those detected in “colon, breast, prostate, lung or brain cancer (ca)” tissue samples at “high”, “medium” or “low” expression level. The definition of expression levels was according to Human Protein Atlas database.

In order to predict MHC-class I epitopes, protein sequences were downloaded from UniProt database (https://​www.​uniprot.​org/​) and the entire sequence was analyzed with NetMHCpan4.1 tool (https://​services.​healthtech.​dtu.​dk/​service.​php?​NetMHCpan-4.​1), to predict the binding affinity [15]. Prediction analyses were performed for the 12 most frequent alleles in the world, selecting only peptides considered weak and strong binders, according to default parameters (WB; SB).

Peptides selected as SB according to NetMHCpan 4.1 prediction tool have been submitted to BLAST for a homology search against HIV-1 (taxid: 11,676) within the non-redundant protein sequences database (https://​blast.​ncbi.​nlm.​nih.​gov/​Blast.​cgi). HIV-derived homologous sequences were analyzed to predict their binding affinity and stability to all alleles as described above.

The 3D structure of interaction between peptides and HLA alleles was generated using Pymol software (PyMol Molecular graphics system, version 1.8.6.2) and Molsoft ICM (http://​www.​molsoft.​com/​, version 3.8-7d) software. The PDB format of complex between HLA molecules and reference peptides were downloaded from RCS Protein Data Bank (PDB) website (https://​www.​rcsb.​org). In particular, HLA-A*01:01 (3BO8), HLA-A*02:01 (1AO7), HLA-A*03:01 (3RL1), HLA-A*24:02 (7JYV), HLA-B*07:02 (7LGD), HLA-B*08:01 (7NUI), HLA-B*15:01 (6VB3), HLA-B*39:01 (4O2E), HLA-B*40:02 (5IEK), HLA-B*58:01 (5VWH). The original peptide sequences in the downloaded structures were replaced with individual predicted epitopes using Pymol visualization system and Molsoft ICM software was used to visualize the molecular docking.

All peptides were synthesized at a purity of ≥ 90% (GenScript; NJ, USA). Lyophilized powders were reconstituted in DMSO Solution (CARLO ERBA Reagents S.r.l., Cornaredo; Italy) and diluted in 90% of 1X PBS (HyClone, Thermo Fisher Scientific Inc., US).

Peptide binding affinity to HLA-A*02:01 molecule and BFA decay assays were performed for each candidate peptide. Human TAP-deficient T2 cell line (174xCEM.T2; ATCC CRL 1992™) was purchased from American Type Culture Collection (ATCC; https://​www.​atcc.​org/​) and cultured in Iscove's modified Dulbecco's medium (IMDM; Gibco Life Technologies) containing 25 mM HEPES and 2 mM L-Glut, supplemented with 20% fetal bovine serum (FBS; Capricorn Scientific GmbH), 100 IU/ml penicillin and 100 μg/ml streptomycin (Gibco Life Technologies). Cells were maintained at 37 °C in a humidified incubator with 5% CO₂. Briefly, T2 were seeded at 3.5 × 10⁵ cells per well in 24 well plates and incubated 16 h at 27°C with peptides (final concentrations: 5, 10, 20, 50 and 100 μM) in IMDM serum free medium. The next day, cells have been incubated for additional 2 h at 37 °C. Following incubation, cells were harvested and centrifuged at 200 × g for 5 min. Subsequently, cells were washed twice with phosphate buffered saline (1X PBS; Gibco Life Technologies) and stained with R-PE conjugated anti human HLA-A2 mono-clonal antibody (cat. 343,306; BioLegend), for 30 min at 4 °C, and analyzed with the Attune™ NxT flow cytometer (Thermo Fisher Scientific). Mouse H-2 Kb-specific OVA SIINFEKL peptide was used as negative control and T2 cells without any added peptide were used as a background control. A fluorescence index (FI) was calculated using the following formula: FI = [mean fluorescence intensity (MFI) sample – MFI background]/MFI background, where MFI background represents the value without peptide. A value of FI > 0.5 was set as threshold to indicate peptides with affinity for the HLA-A*02:01 molecule. For the brefeldin-A decay assay, T2 were incubated with peptides (50 μM) as described above, washed and treated with 1X BFA (brefeldin A solution; cat. 420,601; BioLegend) in IMDM serum free medium, for 1 h at 37°C. Cells were harvested every two hours (T0, T2, T4, T6, T8), washed with phosphate buffered saline (1X PBS; Gibco Life Technologies), stained with anti HLA-A*0201 fluorescent monoclonal antibody (cat. 343,306; BioLegend) and analyzed by flow cytometry. The stability of each peptide bound to HLA-A*02:01 was measured as the DC₅₀ value, which was defined as an estimate of the time required for a 50% reduction of the MFI value at time 0. The DC₅₀ value was calculated according to the formula: MFI at indicated time points/MFI at time 0 × 100. All the experiments were performed in triplicate [11, 16].

pMHC complexes were generated by combining purified disulfide-stabilized HLA-A*0201 monomer (100 μg/ml) with 100 μM peptide for 30 min. in PBS at room temperature [17]. pMHC complexes were centrifuged for 5 min. at 3300 × g to sediment any aggregated MHC molecules. For each 100 μl pMHC, 9.02 μl (0.2 mg/ml stock, SA-PE (Bio-legend, 405,204), SA-APC (Biolegend, 405,207), SA-PE/Cy7 (Biolegend, 405,206), SA-PE-CF594 (BD, 562,284)) or 18.04 μl (0.1 mg/ml stock, SA-BV421 (Biolegend, 405,226), SA-BV650 (Biolegend, 405,231) streptavidin-conjugate was added and incubated for 30 min. on ice. D-biotin (Sigma-Aldrich) was added at a final concentration of 25 μM to block any free binding sites and multimers were stored at − 20 °C with 5% glycerol and 0.5% BSA. To stain for T cell reactivity, PBMCs from 16 HIV-1 patients and 10 healthy donors (2–5 × 10⁶) were incubated with a pool of pMHC multimers (3 μl/multimer) and dasatinib (50 nM final, LC laboratories, D-3307) for 15 min. at 37 °C. Cells were then stained with antibodies CD3-FITC (1:40, BD, 345,763) and CD8-BV480 (1:100, BD, 566,121) and LIVE/DEAD Fixable Near-IR (1:1000, Invitrogen, L10119) for 30 min. on ice and washed twice in FACS buffer (PBS + 2% FCS). Gating for CD3⁺/CD8⁺ T cells was done on alive cells and binding to pMHCs was assessed by measuring specific fluorescence associated with each individual pMHC. Samples were acquired on a flow cytometer (LSRFortessa, BD) and analyzed by the Attune NxT software v3.1.2 (ThermoFisher scientific).

Comparison between individual data points were performed with the two-sided Student’s t-test and ANOVA, as appropriate. Paired, for the matched samples collected at the Istituto Nazionale Tumori—IRCCS—"Fond G. Pascale", Naples, Italy (INT); unpaired for the samples extracted by the publicly available data (http://​gent2.​appex.​kr/​gent2/​). Normally distributed data were represented as mean ± S.E.M. Two-way ANOVA and Bonferroni post-hoc analysis were used to examine the significance of differences among groups. All P values were two-tailed and considered significant if less than 0.05.

In order to identify proteins specifically overexpressed in breast, prostate and colon cancers and not detected in normal cells, the Normal Tissue dataset (15,170 unique proteins) and the Pathology dataset (15,313 unique proteins) available at the Human Protein Atlas were sequentially interrogated (https://​www.​proteinatlas.​org/​about/​download). The same analysis was performed for the lung and brain cancers, which show increased SIRs in PLWHA [5].

An average of 5,768 proteins showed no expression in normal cells of the five tissues, with a range from 4,814 (breast glandular cells) to 7,019 (lung pneumocytes). In parallel, an average of 4,134 proteins were found to be detected in cancer cells, with a range from 2,640 (lung ca) to 5,650 (colon ca). Of the latter proteins, an average of 213 proteins (5.15%) were found to be at high levels in cancer cells, as defined by the Human Protein Atlas, with a range from 86 (lung ca) to 408 (colon ca). Cancer-specific proteins were selected for each tissue as those not detected in normal cells and highly expressed in cancer cells,. The number of the latter proteins ranged from 1 to 2 for breast, prostate, lung and brain tumors. The only exception was the colon cancer for which 44 cancer-specific proteins were identified (Table 1). The subsequent filter was set for selecting proteins not detected in > 70% of normal cells of any tissue. Indeed, only such proteins should not be affected by immune tolerance impairing the immune system to mount responses against self antigens. According to such additional filter, six cancer-specific proteins in colon cancer (CEACAM8, CDH17, GLOD5, PPM1E, TRIM16, EPCAM) and one in breast cancer (CCR9) were identified. None of the cancer-specific proteins identified in prostate cancer, brain and lung cancer met these parameters (Additional file 1: Fig. S1). Nevertheless, they were selected for the subsequent analyses.

The cancer-specific proteins selected in each cancer were submitted to the NetMHCpan software to predict 9 aa-long epitopes (nonamers) strong binders (SBs) to HLA alleles covering more than 70% of the World population. The analysis was performed considering all the overlapping nonamers with a one-amino-acid lateral shift covering the entire protein sequence. In order to select the best predicted epitopes, we focused our analysis only on those showing an affinity < 100 nM to HLA molecules, which we have been previously shown to have a 100% concordance with ex vivo binding assay [11, 12, 16].

The number of SBs with such characteristics predicted from the cancer-specific proteins was highly variable among the five cancers. The highest number of SBs (138) was predicted in colon cancers from the 6 cancer-specific proteins, covering all the evaluated HLA alleles; the lowest number [27] was predicted in breast cancer from the single CCR9 cancer-specific protein (Fig. 1A). Regardless the different numbers of SBs predicted for the cancer-specific proteins from the 5 cancers, their average affinity to the HLA alleles was very high (e.g. < 50 nM) (Fig. 1B). The number of SBs specific to each HLA allele was highly variable. The highest number was predicted for HLA-A*02:01 (51), the second and third highest numbers were predicted for HLA-B*15:01 (42) and HLA-B*58:01 (35). On the contrary, the lowest numbers by far were predicted for HLA-A*01:01 [7] and HLA-A*26:01 [10]. The number of predicted SBs for all other HLA alleles ranged between 13 and 30 (Additional file 1: Fig. S2). Furthermore, the average affinity to the HLA alleles was highly variable without reaching a statistical difference. The highest average affinity was observed for HLA-A*02:01 specific epitopes (e.g. 26 nM), the lowest was observed for HLA-B*08:01 specific epitopes (e.g. 52 nM) (Fig. 1C).

In order to verify the expression of the predicted SBs on normal cells, the HLA Ligand Atlas was interrogated. Interestingly, 271 (93.7%) of the 298 SBs predicted in the cancer-specific proteins were found not expressed in the HLA ligandome of normal cells (Table 2). All peptides found at the HLA Ligand Atlas were excluded from subsequent analyses. Moreover, some of such predicted TAAs have been previously described to be presented by cancer cells and/or be target of T cells [18‐22].

The following step was to verify the molecular mimicry between TAAs predicted for the cancer-specific proteins and HIV-1 antigens. Indeed, this might explain a cross-protective anti-cancer immune response elicited by HIV-1 antigens in HIV-1 patients. To this aim, the predicted 298 TAAs were screened in BLAST for homology to HIV-1 peptide sequences. In total, 25 HIV-1 peptides were identified with sequence homology to predicted TAAs and showing high binding affinity to HLA alleles (i.e. < 100 nM) (Table 3). Interestingly, the highest number of HIV-1 peptides were identified for colon cancer (eleven) and breast cancer (six). Five peptides were identified for lung cancer, two for prostate cancer and only one for brain cancer (Fig. 2A). The broadest HLA coverage (6 alleles) was observed for the HIV peptides homologous to colon-specific TAAs, while the narrowest one (1 allele) was observed for the HIV peptides homologous to prostate-specific and brain-specific TAAs. Most importantly, only HIV peptides homologous to colon-, breast- and prostate-specific TAAs showed strong binding to HLA-A alleles (01:01; 02:01; 03:01 and 24:02) which are the most frequent in the world population (Fig. 2 B; Additional file 1: Fig. S3A; Additional file 1: Fig. S4). Considering the individual cancer-specific proteins, the highest number of HIV peptides homologous to TAAs (six) were predicted for the CCR9 (breast ca) and EPCAM (colon ca) proteins, covering the two most frequent HLA-A alleles in the world population (02:01 and 24:02) (Fig. 2 C; Additional file 1: Fig. S3B). The average binding affinity of such HIV peptides to the HLA alleles was very high (18.23 nM for A*02:01; 43,35 nM for A*24:02) (Fig. 2 D).

The confirmation of predicted binding of peptides to HLA alleles could be experimentally confirmed only for the A*0201. The TAP-deficient T2 cell line was loaded with the individual peptides from paired TAA and HIV peptides. In particular, the VVAGIVVLV TAA derived from the colon ca-associated EPCAM protein was compared with the VVAGIIALV and VIAGIVALV epitopes derived from the HIV-1 vpu protein. The analysis confirmed that each peptide bound the HLA-A*02:01 molecule inducing a dose-dependent increase in the HLA surface expression on T2 cells over the background.

The results confirmed the comparable binding to HLA-A*02:01 of the three peptides as suggested by the predicted affinity values (Fig. 3A). Likewise, the peptide—MHC dissociation kinetics showed that the 50% dissociation value (Thalf) was reached 4 h after the peptide loading (T4), for the EPCAM TAA, and 6 – 8 h after the peptide loading (T8) for the two HIV peptides (Fig. 3B and C). Such results were representative of affinity and stability of the other HLA-A*02:01 associate peptides (data not shown).

HIV-1 is characterized by an extensive intra- and inter-host genome variability with generation of molecular “quasi-species” [23]. Therefore, the biological significance of the described homology between TAAs and HIV peptides is relevant only if the amino acid sequences are conserved across viral isolates. In order to assess such a sequence conservation, about 5000 HIV-1 sequences deposited in Blast were aligned to generate sequence logos. Indeed, logos provide the consensus sequence of all aligned peptides sequences as well as the relative frequency of amino acids at every position of the peptides. The results showed that the HIV peptides with the highest homology to TAAs identified in breast (CCR9), colon (EPCAM, GLOD5, CEACAM8) as well as prostate (NDUFS2) cancer and linked to most frequent HLA alleles (01:01; 02:01 and 24:02), are highly conserved across HIV isolates (Additional file 1: Fig. S5). In particular, HIV peptides RLRDLLFLV and NYTEIIHTL (homologous to breast CCR9) and VVAGIIALV (homologous to colon EPCAM) show 8 of 9 amino acids conserved in the consensus sequence. The HIV peptide SSDNSSRPY (homologous to prostate NDUFS2) shows identity in 7 of 9 amino acids compared to the consensus sequence.

Additional HIV peptides homologous to TAAs from breast and colon cancer and linked to less frequent HLA alleles show high conservation across HIV isolates. Finally, similar results are observed for the HIV peptide IPRRTTQRI (homologous to brain DPYSL2) showing identity in 7 of 9 amino acids compared to the consensus sequence. On the contrary, HIV peptides homologous to lung TAAs show poor identity (3 or 5 of 9 amino acids) to the consensus sequence (Additional file 1: Fig. S6).

In order to verify that predicted paired TAAs and HIV-1 epitopes, including the HIV-1 consensus sequences, share contact residues with both the HLA molecule and the TCR α and β chains, epitope modelling and molecular docking were performed.

Epitopes crystallized with HLA alleles showing sequence homology with TAAs or HIV-1 peptides described in the present study were not found and examples of epitopes were used as templates to conduct the analyses. Crystallized pMHC complexes including also the TCR chains were available for the HLA-A*02:01, HLA-A*01:01, B*07:02, B*08:01.

As predictable from the sequence homology, the HIV peptides homologous to TAAs derived from breast, colon and prostate cancers showed highly similar, if not identical, conformation and contact points with HLA molecule and TCR chains as compared to corresponding TAAs. In detail, the breast CCR9 CYTIIIHTL and the HIV- 1 NYTEIIHTL show a conservative Cys to Asn and a non-conservative Ile to Glu substitutions in p1 and p4, respectively. The colon EPCAM VVAGIVVLV and the HIV- 1 VVAGIIALV show conservative Ile to Val and Val to Ala substitutions in p6 and p7, respectively. The prostate NDFUS2 VSDGSSRPY and the HIV-1 SSDNSSRPY show non-conservative Val to Ser and Gly to Asn substitutions in p1 and p4, respectively. They all do not induce significant conformational changes in the structure of paired epitopes. Moreover, the consensus sequences highly conserved across HIV isolates showed identical structure to individual HIV-1 peptides. In detail, the CCR9-like NYTEIIHTL peptide and the consensus NYTEIIYTL show a single non-conservative His to Tyr substitution in p7. The EPCAM-like VVAGIIALV peptide and the consensus VVAAIIALV show a single conservative Gly to Ala substitution in p4. The NDFUS2-like SSDNSSRPY peptide and the consensus SSNSSSRPY show a non-conservative Asp to Asn and a conservative Asn to Ser substitutions in p3 and p4, respectively (Fig. 4).

On the contrary, HIV peptides homologous to TAAs derived from brain and lung cancers did not show such high conformational similarities. In particular, the brain DPYSL2 IPRRTTQRI and the HIV-1 IPRRIRQRI show two non-conservative Thr to Ile and Thr to Arg substitutions in p5 and p6, respectively. The lung CNIH4 LLNLPVATW and the HIV-1 HCNLSVATW show three non-conservative Leu to His, Leu to Cys and Pro to Ser substitutions in p1, p2 and p5, respectively. They all induce significant conformational changes in the structure of paired epitopes. Moreover, the HIV consensus sequences showed structures significantly different from the individual HIV-1 peptides. In detail, the DPYSL2-like IPRRIRQRI peptide and the consensus IPRRLRQRV show two conservative Ile to Leu and Ile to Val substitutions in p5 and p9, respectively. The CNIH4-like HCNLSVATW peptide and the consensus SCNTSVITQ show three non- conservative His to Ser; Leu to Thr; Trp to Gln substitutions in p1, p4 and p9, respectively; and a conservative Ala to Ile substitution in p7 (Fig. 4). The Los Alamos HIV immunological database was interrogated for experimentally validated HIV epitopes showing sequence homology to those paired to TAAs. Indeed, validated epitopes with such characteristics were found for the breast ca CCR9 CYTIIIHTL epitope (NYTDIIYRL) [24], the brain ca DPYSL2 IPRRTTQRI epitope (IPRRIRQGL) [25] and the colon ca EPCAM VVAGIVVLV epitope (VVAAIIAIV) [26] (Fig. 4). In particular, the latter validated HIV epitope shows a relevant structural and conformational homology to the EPCAM TAA, confirming the biological relevance of the predictive findings. All other HIV peptides homologous to TAAs showed limited and variable similarity in conformation and contact points with HLA molecule and TCR chains as compared to corresponding TAAs (Additional file 1: Fig. S7–S9).

In order to confirm the antigenicity of the predicted peptides and the cross-reactivity with the paired TAAs and HIV-1 peptides, we performed an analysis to detect reactive CD8⁺ specific T cells in PBMCs from HIV patients and healthy subjects (Additional file 1: Fig. S10).

An “ex vivo” expansion of HIV-specific T cells was induced culturing PBMCs with HIV peptides for 7 days. Subsequently, cells were incubated with tetramers loaded with the corresponding TAAs and analysed at cytofluorimeter.

Given the overwhelming prevalence of HLA-A*02:01 and 24:02 alleles in the cohort of enrolled HIV patients (16/22, 73%) (Additional file 2: Table S1), the analysis was performed with the paired HIV/TAA peptides from the EPCAM colon ca (HLA-A*02:01) and from the CCR9 breast ca (HLA-A*24:01) proteins.

Strikingly, results show that HIV-positive subjects cross-react against HLA-A matched TAAs at baseline, even without any prior “ex vivo” expansion, while HIV-negative subjects do not (Fig. 5A and C). Such cross-reactivity is further increased only in HIV-positive subjects after the “ex vivo” expansion with the paired HIV peptide (Fig. 5B and D). The difference between the HIV-positive and -negative groups reaches the statistical significance for the EPCAM colon cancer HLA-A*02:01 paired peptides. The limited number of HLA-A*24:02 HIV-positive subjects does not provide the sufficient statistical power for the trend observed with CCR9 breast cancer paired peptides. The increased cross-reactivity against the TAA in HIV-positive subjects, prior and after the “ex vivo” expansion with HIV peptides, shows a clear trend that does not reach the statistical significance (Fig. 5E–L). No cross-reactivity was observed when T cells from HIV patients were reacted with HLA-matched not-homologous epitope (data not shown).

We and others have previously shown that the molecular mimicry between viral and tumor antigens may have a significant impact in controlling tumor growth and improving the clinical outcome in cancer patients. More recently, we have shown that TAAs may show sequence homology as well as structural similarities with viral peptides and cross-reacting CD8⁺ T cells can be identified to drive the fate of cancer development and progression. In particular, such homologies must be found at the residues facing the T cell receptor (TCR). Therefore, a previous viral infection may represent a natural “anti-cancer vaccination”.

This is the first study showing a molecular mimicry between HIV antigens and TAAs identified in breast, prostate and colon cancers. T cells cross-reacting with paired antigens has been demonstrated.

Therefore, it is highly reasonable that memory CD8⁺ T cells elicited during the HIV infection may play a key role in controlling development and progression of such cancers in the PLWHA lifetime. This represents the first demonstration ever that a viral infection may induce natural “preventive” anti-cancer memory T cells, with highly relevant implications beyond the HIV infection.