{"paper_id":"83b5c24c-6938-4a9b-952c-ee48f76703af","body_text":"1 \n \nDual-antigen Doggybone TM DNA vaccine induces potent anti -tumor immunity against \nimmunosuppressive oral cancer  \nAuthors \nGrace M.Y. Tan1, Chuan Wang1, Maria M. Meschis1, Josephine Buckingham2, Kue Peng Lim3, \nSok Ching Cheong 3,4,5, Gareth J. Thomas 2, Sungwon Kim 6, Lisa Caproni 6, Helen Horton 4, \nChristian H. Ottensmeier1,7 & Natalia Savelyeva1* \n \n1Head and Neck Centre, Institute of Systems, Molecular and Integrative Biology, University \nof Liverpool, Liverpool, United Kingdom \nE-mail: N.Savelyeva@liverpool.ac.uk \n2Cancer Science, Faculty of Medicine, University of Southampton, Tremona Road, \nSouthampton, United Kingdom \n3Cancer Immunology and Immunotherapy Unit, Cancer Research Malaysia, Selangor, \nMalaysia \n4Department of Oral and Maxillofacial Clinical Sciences, Faculty of Dentistry, University \nMalaya, Kuala Lumpur, Malaysia \n5Oral Cancer Research and Coordinating Centre  (OCRCC), Faculty of Dentistry, University \nMalaya, Kuala Lumpur, Malaysia. \n6Touchlight Genetics Ltd, Hampton, United Kingdom \n7The Clatterbridge Cancer Center, Liverpool, United Kingdom \nKeywords: DNA cancer vaccine, cancer testis antigens , cold tumor, HPV negative head and \nneck squamous cell carcinoma \n \n  \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 2 \nAbstract  \nAnti-PD1 blockade benefits only a subset of patients with head and neck squamous cell \ncarcinoma (HNSCC) , highlighting the need for approaches that  overcome tumor immune \nresistance. Here, using the doggybone DNA (dbDNATM) platform, we developed CaVac OPT, \nan optimized dual-antigen DNA vaccine  targeting MAGED4B and FJX1 , which are  \noverexpressed in most HPV-negative HNSCC and multiple solid tumors. In the MOC-2 oral \ncancer model, CaVac OPT significantly reduced tumor growth and when combined with anti-\nPD1 therapy, further delayed progression  and improved survival. Immune profiling showed \nincreased infiltration of CD4+ and CD8+ T cells, expansion of stem-like Tcf1+ populations, \nwithout increase in regulatory T cells, a reduced M2/M1 macrophage ratio and activation of \ninterferon gamma associated pathways with suppression of tumor-promoting signals. These \nfindings demonstrate that CaVac OPT reprograms the tumor microenvironment, converting \ncold HNSCC into T cell -inflamed responsive tumors. CaVac OPT  represents a promising  \nstrategy for achieving durable control of aggressive, immunotherapy -resistant head and neck \ncancer.  \n \nIntroduction  \nTherapeutic cancer vaccines are increasingly recognized for their potential to enhance clinical \noutcomes when combined  with immune checkpoint inhibitors (ICIs) , particularly in patients \nrefractory to standalone immunotherapy1–3. These vaccines are engineered to elicit tumor-\nspecific T-cell responses, promoting their infiltration into the tumor microenvironment (TME) \nand augmenting the efficacy ICIs. Among emerging platforms, DNA vaccines have shown \nconsiderable promise. For instance, the DNA vaccine EVX-02, administered alongside ICIs, \ninduced robust T -cell responses in all  10 stage III/IV melanoma patients evaluated, with no \ndisease recurrence observed after one year, a notable contrast to the  historical recurrence rates \nof approximately 30% for nivolumab monotherapy3. Similarly, the DNA vaccine  SCIB1 \ntargeting the unmutated melanoma antigen gp100 , achieved an 85% objective response rate \n(ORR) in previously untreated , unresectable stage III/IV patients when combined with \nnivolumab and ipilimumab , with tumor volume reductions ranging from 40 to 100%, by 25 \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 3 \nweeks post - treatment4. In a  phase I/II trial involving advanced hepatocellular carcinoma , \ncombining DNA vaccines targeting neoepitopes with ICI pembrolizumab yielded an ORR \nexceeding 30%, including an 8% complete response (CR)  rate, outperforming the 12-18% \nresponse rates typically  observed with  anti-PD1 monotherapy. Notably, nearly 90% of the \npatients showed neoantigen-specific T -cell responses, underscoring the capacity of DNA \nvaccines to generate broad, durable cytotoxic CD8+ T cell responses5.  \nDNA vaccines offer distinct advantages, including cost effectiveness, safety, stability \nand flexibility in design and construction. These attributes enable the construction of complex \nvaccines incorporating multivalent antigens and immunostimulatory sequences  to optimize T-\ncell induction 6–8. Similarly, DNA vaccines can encode numerous neoepitopes tailored to a \npatient’s mutanome, thereby enhancing the induction of tumor-reactive T cells and priming de \nnovo responses alongside the expansion of pre -existing T-cell epitopes5 Recent studies also \nsuggest that delivery methods, such as in vivo electroporation and needle-free injection systems \n(NFIS), enhance T -cell responses and clinical efficacy9–11, offering practical solutions for \nadministration without complex formulations  in late -stage cancer patients . However, most \ncurrent DNA vaccines rely on plasmid DNA (pDNA) produced in bacteria, requiring on \nantibiotic resistance genes for propagation and necessitating endotoxin removal. While these \nextraneous bacterial sequences may boost immunogenicity, they also carry risks, such as \ntransgene silencing  and the potential transfer of antibiotic resistance to the host’s \nmicrobiota12,13. This provides a rational to explore enzymatically produced DNA vaccine, such \nas doggybone DNA (dbDNATM).  \n \ndbDNATM comprises linear, double-stranded DNA with covalently closed ends , \nsynthesized enzymatically via rolling circle amplification by the Phi29 phage polymerase and \nsubsequent cleavage by  the N15 phage protelomerase.  Unlike pDNA,  dbDNATM is \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 4 \nminimalistic, encoding only the immunogen of interest under a eukaryotic promoter with a \npolyadenylation signal14,15. Our previous work demonstrated that dbDNATM induces cytotoxic \nT-cell and humoral immune responses comparable to p DNA vaccine s16 with emerging \nevidence suggesting it may offer even greater immunogenicity17,18. While dbDNATM has been \ninvestigated across various applications, its potential as a therapeutic cancer vaccine remains \nunexplored, particularly in the context of  overcoming an immunosuppressive tumor \nmicroenvironment (TME). In this study,  we report the optimization of dbDNATM constructs \ntargeting cancer -testis antigens MAGED4B and FJX1  frequently expressed in solid \nmalignancies, incorporating modifications to enhance T-cell responses. We demonstrate that \nvaccine candidate CaVac OPT elicits significant, durable protective anti-tumor T-cell response \nagainst an HPV-negative oral carcinoma model , characterized by a suppressive TME and \nwhich is unresponsive to anti -PD1 ICI. Furthermore, we show that CaVac OPT promotes a \nshift from M2 to M1 macrophages, underscoring its therapeutic potential.  \n \nResults \nOptimizations to the dbDNATM vaccine constructs to enhance specific antigen delivery  \nHere we chose to target cancer -testis antigens MAGED4B and FJX1 due to their prevalent \nexpression across various cancers, including head and neck squamous cell carcinoma \n(HNSCC) 8,19,20, nasopharyngeal carcinoma21, lung cancer and other solid tumors22,23. Initially, \nwe incorporated several regulatory elements that were previously shown to enhance \nimmunogenicity of plasmid-based DNA vaccines24. These included the SV40 enhancer which \nis designed to augment  transcription and promote nuclear translocation , and t ype C CpG \noligodeoxynucleotides ( CpG-C) to activate innate immune signaling via TLR9 25,26 \n(Supplementary Fig 1). Antigen-specific T cell responses were assessed by interferon gamma \n(IFN-γ) ELISPOT assays, with comparisons made to t he unmodified dbDNATM constructs, \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 5 \ntermed Basic 0, which encoded a fusion of each antigen to domain 1 (DOM) of tetanus toxin. \nThe DOM sequence has been shown to enhance induction of CD8+ T cells to cancer antigens, \nresulting in clinical benefits in patients with solid cancers7,27. Despite the incorporation of these \nenhancer elements , none significantly i mprove T cell responses compared to the Basic 0  \nconstruct (Supplementary Fig 1).   \n To further improve the vaccines performance, we next explored immune-enhancing \nelements to improve antigen delivery to dendritic cells and refi ned the antigen design. This \nincluded fusion of immune -targeting molecules such as Fms-related tyrosine kinase 3 ligand \n(Flt3L), or Macrophage Inflammatory Protein -1 alpha (MIP-1a). Both approaches have been \ntested clinically with demonstration of robust T-cell responses resulting in meaningful clinical \nresponses11,28,29. Further variants for FJX1 included engineered conserved consensus antigenic \nsequences (Con) to break tolerance to this highly conserved antigen 30. For MAGED4B, we \nadditionally generated variants with a complete and partial deletion of the MAGE homology \ndomain (HD)  dHD1 and dHD2 respectively , to eliminate epitopes with potential cross -\nreactivity to non-target peptides , thereby reducing the risk of off -target cytotoxicity 31,32. \nVaccine variants were  compared to a fusion free (FF) construct using a prime/ boost \nimmunization. Among MAGED4B  variants, only the dHD1 construct elicited a notabl y \nenhanced IFN-γ response (Figure 1 A). In contrast, for FJX1, the construct fused to Flt3L \ninduced significantly higher T cell responses (Figure 1B). \nWe next compared the top-performing variants (dHD1 for MAGED4B, and fusion to \nFlt3L for FJX1) to the Basic 0 constructs. dHD1 induced the strongest IFN-γ responses (Figure \n1C) while Flt3L-FJX1 elicited significantly higher responses than Basic 0 (Figure 1D). To \nassess potential vaccine interference, we compared co-administration of both MAGED4B and \nFJX1 vaccines at a single site versus opposite flanks and found comparable T cell responses \nwith no evidence of immune interference in vivo (Supplementary Fig 2). The combination of \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 6 \ndHD1 and Flt3L-FJX1, hereafter referred to as CaVac OPT, was selected for further preclinical \nevaluation in an oral carcinoma model due to its superior immunogenicity targeting both  \nMAGED4B and FJX1.  \n \nFIGURE 1. dbDNATM vaccines encoding MAGED4B and FJX1 elicit antigen -specific T \ncell responses.  C57BL/6 mice (9-12 weeks old) were vaccinated intramuscularly (i.m.) with \ndbDNATM vaccine constructs in the anterior tibialis muscle with electroporation. A prime-boost \nregimen was used, with vaccinations on day  one and day  twenty-one. Splenocytes were \nharvested on day thirty post vaccination and stimulated for 40 h with 1µM MAGED4B or FJX1 \noverlapping peptide pools. Data are presented as spot -forming cells per million (SFC x 10 6 \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 7 \ncells), with background (<10 SFC/106) subtracted. (A) Schematic illustration of the dbDNATM \nvaccine construct targeting MAGED4B. Each construct as indicated also includ ed the CMV \npromoter and the SV40 polyA signal. Grey circles represent teRlL sequences, the recognition \nand cleavage sites for protelomerase TelN. MAGED4B were fused to immune -stimulatory \nmolecules MIP-1α, Flt3L or modified by deletion of the MAGE homology do main (dHD1: \namino acids 412-681; dHD2: amino acids 412-500). Vaccine responses were compared to the \nfusion-free (FF) vaccine. (B) FJX1 were fused to immune -stimulatory molecules MIP -1α or \nFlt3L; the Consensus construct (Con) incorporates consensus sequence modifications derived \nfrom FJX1. Vaccine responses were compared to the fusion-free (FF) FJX1 dbDNA vaccines. \nStatistical comparisons were performed using one -way ANOVA with Tukey’s post-hoc test. \nSimplified schematic of selected vaccine variants (Basic 0, dHD1 and Flt3L) of (C) \nMAGED4B and (D) FJX1 vaccine variants. T cell responses to the most immunogenic variants \n(dHD1 for MAGED4B; Flt3L for FJX1) were compared to the  tetanus DOM fragment- fused \nvaccine (Basic 0). Statistical analysis determined by unpaired two tailed t -test. Statistical \nsignificance is indicated as follows: *p<0.05, **p<0.01, ***p<0.001, ns: not significant.  \n \nCaVac OPT controls tumor growth and potentiates anti PD-1 therapy \nMurine oral carcinoma-2 (MOC-2) is an aggressive head and neck cancer model derived from \noral squamous cancerous epithelia33. MOC-2 tumors are characterized by a low infiltration of \nCD8+ T cells, low expression of MHC-I, high levels of FoxP3+ T regulatory (Tregs)  cells and  \nM2 macrophages33,34;  closely resembling the human oral squamous cell carcinoma  (OSCC). \nHere, MOC-2 cells were engineered to stably express the MAGED4B and FJX1 antigens and \nwere used to evaluate the efficacy of combination dbDNATM vaccines in controlling tumor \ngrowth. Mice were challenged with MOC-2 tumor cells and vaccinated on day three post \ninoculation with either Basic 0 constructs of MAGED4B and FJX1 or CaVac OPT  at low \n(4µg/mouse) or high (10µg/mouse) doses. Mice that received the vector control exhibited a \nprogressive increase in tumor burden, whereas both vaccines reduced tumor growth at high \ndose; however, CaVac OPT also induced a significant inhibition of tumor growth at the low \ndose (p<0.0005) (Figure 2A). Given its superior efficacy at both doses, we further assessed the \ntherapeutic efficacy of CaVac OPT  in mice bearing larger MOC-2 tumors. Mice were \nvaccinated with CaVac OPT  on day seven post-tumor inoculation. Compared to controls, \nvaccinated mice displayed a significantly reduced tumor burden (p<0.05) (Figure 2B).  \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 8 \nTo assess the roles of CD4 + and CD8 + T cells in CaVac OPT -mediated anti-tumor \nimmunity, we administered anti-CD4 or anti-CD8 depleting antibody to MOC-2 tumor-bearing \nmice vaccinated with CaVac OPT (Figure 2C) and compared the tumor volumes at the end of \nexperiments. In CaVac OPT-vaccinated mice, depletion of either CD4+ or CD8+ T cell subsets \ncontributed to loss of protection against tumor challenge, resulting in a higher  tumor burden \ncompared to vaccinated mice given an isotype control (Figure 2C). This indicates that both T \ncell subsets contribute to protection against tumor challenge in this model.  \nWe next evaluated the combination of CaVac OPT with anti-PD1 therapy. Mice were \nchallenged with high tumor burden and vaccinated with CaVac OPT on day three and day \ntwenty-four post-tumor challenge. Starting on day eight post-tumor challenge, mice received \neight doses of anti-PD1 or isotype control antibody (Figure 2D). An anti-PD1 alone group was \nincluded for comparison, as MOC-2 tumors are known to be resistant to anti-PD1 treatment35. \nThe CaVac OPT vaccine significantly sensitized the MOC-2 model to ICI blockade, reducing \ntumor burden in mice and acting synergistically with anti-PD1. Kaplan-Meier analysis further \nconfirmed that combination therapy significantly extended the mean survival compared to \nnaïve mice, vaccine  alone or anti -PD1 alone ( Figure 2D ). These findings demonstrate that \ncombining CaVac OPT with anti-PD1 enhances anti -tumor immunity in the ‘cold’ MOC -2 \ntumor model, conferring protection against the disease.  \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 9 \n \nFIGURE 2. Basic 0 and CaVac OPT vaccine  induces superior anti-tumor immunity in \nthe MOC-2 murine oral carcinoma model. \n(A) Schematics of the vaccination schedule: mice were inoculated with 1.6 x 10⁵ MOC-2 tumor \ncells and vaccinated on day  three and day  twenty-four post inoculation. Two vaccine \ncombinations were tested: Basic 0 (Basic 0 MAGED4B, Basic 0 FJX1) and CaVac OPT  \n(dHD1 and Flt3L -FJX1). Each combination was administered at low (4 µg/mice) and high \n(10µg/mice) doses and compared to a vector control group. Significant tumor growth reduction \nwas observed in mice receiv ed high dose Basic 0 (blue, p<0.01) and low dose CaVac  OPT \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 10 \n(purple, p<0.001) compared to control. Data (top) are shown as mean ± SEM (n= 8 per group); \nindividual tumor measurements are displayed below. Statistical analysis determined by \nDunnett’s multiple comparisons test. ( B) Experimental setup: mice were challenged with 2 x \n10⁵ MOC-2 tumor cells and vaccinated on day  seven and day twenty-eight post-inoculation. \nTumor volume was measured every three days until day thirty. Data are presented as mean ± \nSEM (n=8). Statistical significance was determined by the Mann -Whitney test. ( C) Role of \nCD4+ and CD8+ T cells in CaVac OPT-mediated protection. Mice were treated with depleting \nantibodies against CD4+ or CD8+ (500 µg /mice, i.p., every 3 -4 days in five doses)  followed \nvaccination with CaVac OPT. Final tumor volumes on day  thirty-one were compared to mice \nreceived antibody isotype control. Statistical analysis determined by Dunnett’s multiple \ncomparisons test. ( D) Combination of CaVac OPT and anti -PD1 therapy. Mice were \nchallenged with 2 x 10 5 MOC-2 tumor cells and vaccinated with 4 µg CaVac OPT, with anti -\nPD1 antibody or anti-PD1 antibody alone (200 µg /mice, clone RMP1-14, every 3-4 days from \nday five). IgG2a isotype control antibody (200 µg /mice, clone 1 -1) w as used as control. \nKaplan-Meier survival curves of each treatment group (Log-rank test). Statistical significance \nis indicated as follows: *p<0.05, **p<0.01, **3*p<0.001, ns: not significant.  \n \nCaVaC OPT facilitated tumor infiltration by T cells  \nWe next utilized immunohistochemistry (IHC) to characterize T cell infiltration within \nTME. MOC -2 tumors were harvested at two time points following either priming alone \n(harvest at day fifteen) (Figure 3A) or after prime and boost vaccinations (harvest at day thirty \npost-inoculation) (Figure 3 B). IHC analysis at day  fifteen (post prime)  revealed a notable \ninfiltration of CD8+ and CD4+ T cells in the CaVac OPT-treated group, without an increase in \nFoxP3+ Tregs (Figure 3 C). This led to a significant increase in CD8 +/CD4+ T cell ratio  \ncompared to the controls (data not shown), with no significant change in the CD8+/FoxP3+ ratio \n(Figure 3E). By day thirty, there was a significant absolute increase in antigen-specific CD8+ \nand CD4+ T cells, but not FoxP3+ T cells compared to unvaccinated controls (Figure 3D), also \nresulting in a significant increase in the ratio of CD8+/FoxP3+ Treg ratios (Figure 3F).   \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 11 \n \nFIGURE 3. CaVac OPT vaccination enhances anti-tumor immunity via intra-tumoral T \ncell infiltration of MOC-2 tumors. \nC57BL/6 mice were subcutaneously injected with 2.5 x 10⁵ MOC -2 tumor cells and either \ntreated or left untreated. Tumors were excised, fixed in 4% paraformaldehyde and processed \nfor immunohistochemistry (IHC) analysis. ( A–B) Representative IHC images of tumor \ninfiltrating T cells following (A) prime vaccination alone on day three post tumor inoculation \nor (B) prime and boost vaccination administered on day three and day twenty-four post tumor \ninoculation. Tumor sections from control and CaVac OPT-treated mice were stained for CD8⁺, \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 12 \nCD4⁺, and FoxP3⁺ cells. Positive staining is indicated in brown. Scale bar = 20 μm. Statistical \nsignificance is indicated as follows: *p<0.05, **p<0.01***p<0.001, ns: not significant. % of \nCD8⁺, CD4⁺, or FoxP3⁺ for immunohistochemistry staining are presented in scatter plots. Ten \nrandom fields from each section were counted. (C) Dot plots depicting the percentage of CD8⁺, \nCD4⁺, and FoxP3⁺ cells in across control mice (n=5) and vaccinated mice (n=4) following \nprime vaccination. (D) Dot plots depicting the percentage of CD8⁺, CD4⁺, and FoxP3⁺ cells in \ncontrol mice (n=5) and vaccinated mice (n=6) following prime and boost vaccination.  (E) \nViolin plots illustrate the ratios of CD8⁺/FoxP3⁺ and CD4⁺/FoxP3⁺ T cells across control mice \n(n=5) and vaccinated mice (n=4) after prime vaccination.  (F) Violin plots illustrate the ratios \nof CD8⁺/FoxP3⁺ and CD4⁺/FoxP3⁺ T cells across treatment groups after prime and boost \nvaccination. Statistical analysis was performed using the Mann–Whitney test.  \n \nCaVaC OPT vaccination reprograms the transcriptome towards enhanced IFN-γ signaling \nand cytotoxic T cell recruitment  \nTo assess changes in gene expression, we performed bulk RNAseq analysis on CD45+ \ntumor-infiltrating immune cells  to compare the immune landscape between vaccinated and \ncontrol mice post priming (Figure 4A). We analyzed differentially expressed genes (DEGs) \nand associated signaling pathways between the vaccinated and control groups. At day fifteen \npost-vaccination, 23 genes were significantly upregulated and 70 were  significantly \ndownregulated (log2FC>|0.5|, padj<0.1) (Figure 4B, Supplementary table 1).  Notably, genes \nencoding molecules associated with cytolytic T-cell function such as Gzmk, Ifng and Nkg7 were \nmarkedly upregulated (log2FC +3.0, +2.4, +1.6, respectively; padj<0.1) in the vaccinated group \n(Figure 4B). Additionally, expression of the T cell homing receptor Cxcr3 (log2FC +1.46) and \nchemokine Ccl5 (log2FC +1.47) was elevated, indicating enhanced priming and recruitment of \ncytotoxic lymphocytes to the tumor site. Gene Set enrichment analysis (GSEA) using Hallmark \ngene sets identified 15  significant enriched pathways (p<0.005), including IFN-γ response \n(normalized enrichment score [NES] +1.46) pathway and interferon alpha response (NES +1.5) \n(Figure 4C) (Supplementary table 2 ). Further enrichment analyses with Reactome and Gene \nOntology (GO) databases (Supplementary table 3 ) highlighted pathways related to T cell \nreceptor complex activation. KEGG pathway analysis (Supplementary table 4 ) also showed \npositive enrichment in antigen processing and presentation (mmu04612, NES +2.1). Both DEG \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 13 \nand GSEA analyses consistently highlighted enhanced IFN-γ signaling in the vaccinated group \n(Figure 4 D), suggesting infiltration of effector T cells with enhanced IFN-γ signaling and \nantigen processing and presentation with the vaccinated group (Figure 4D).  \n \nFIGURE 4. CaVac OPT vaccination reshapes immune cell composition in cold tumors. \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 14 \nC57BL/6 mice were subcutaneously injected with 2.5 x 10⁵ MOC -2 tumor cells and treated \nwith or without CaVac OPT. On day fifteen post tumor challenge, tumors were harvested and \nCD45⁺ tumor -infiltrating lymphocytes (TILs) were isolated for bulk RNA sequencing. (A) \nPrincipal Component Analysis (PCA) of variance stabilized (VST) bulk RNA sequencing data \nfrom CD45+ TILs in control (n=4) and CaVac OPT vaccinated samples  (n=5). Axes indicate \nthe percentage of variance explained by each principal component. Each dot represents one \nmouse. (B) Volcano plot of differentially expressed genes (DEGs) between control (n = 4) and \nvaccinated (n = 5) mice. Red dots indicate significantly different genes (log₂FC > 0.5, adjusted \np < 0.1). (C) Gene set enrichment analysis (GSEA) annotation of the top 15 enriched pathways \nfrom Hallmark, top 30 Gene Ontology (GO) and top 10 KEGG  databases. (D) GSEA ridge \nplots and heatmaps  of interferon -gamma response and antigen processing and presentation \ngene signatures. Left: Ridge plot and heatmap shows expression profiles of IFN-γ-related genes \n(NES = 1.46); right: antigen processing and presentation (NES = 2.1) in control and vaccinated \ntumors.  \n \nCaVac OPT increases CD8+ memory T cell differentiation and activation \nWe next performed bulk RNAseq analysis on CD45+ TILs at day thirty post-tumor challenge \nto assess longitudinal changes in gene expression (Figure 5A). Differential expression analysis \nidentified 512 significant DEGs (padj<0.1, log2FC>0.5) (Supplementary table 5), including \n211 upregulated and 301 downregulated genes in vaccinated compared to control mice (Figure \n5B). Notably, transcription factor Tcf7 (log2FC +2.55) and Lef1 (log2FC +3.8), both critical \ntranscriptional regulators of memory T cell differentiation , were significantly upregulated. \nUsing the gene set signature-based algorithm ImmucellAI, we observed a significant expansion \nof CD8+ central memory T cells within the T cell population (Figure 5C). Consistent with these \nfindings, IHC analyses of tumor tissues following boost vaccination confirmed a significant \nenrichment of transcription factor 1 (Tcf1) Tcf1+ cells, which exhibited characteristics of T cell \nmemory stem cells (TSCM) (Figure 5D). Together, these results suggest that boost vaccination \npromotes the expansion of  CD8+ memory T cells, supporting their role in  durable anti-tumor \nprotection. \n \nCaVaC OPT vaccine shifts macrophage polarization towards an M1 phenotype in the TME \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 15 \nMacrophages play a critical role in the TME, capable of polarizing toward phenotypes that \neither promote or suppress tumor progression. Heatmap visualization of bulk RNA-seq data \nrevealed a significant ablation in M2-like gene expression in the vaccinated group (Figure 5E, \nSupplementary table 5). Specifically, tumor-associated macrophages (TAMs) from vaccinated \nmice exhibited diminished expression of transcription factor Gata2 (log2FC -1.2, p<0.002) and \ndownregulation of Il10ra, Il13, Il27 (log2FC -1.1, -1.4, -1.6, p<0.001), suggesting that CaVac \nOPT reduces M2-like macrophage signature in TME. Additionally, expression of inhibitory \nreceptors Cd274, Adora2a, and Adora3a was attenuated in vaccinated mice (log2FC -1.1, -1.0, \n-1.2, p<0.001), further indicating diminished immunosuppressi ve activity of M2 -like \nmacrophages. The gene Lgals9, which encodes galectin -9, was significantly upregulated in \ncontrol mice compared to vaccinated mice (log 2FC -0.8, padj<0.1), indicating a higher \nprevalence of pro-tumorigenic macrophages in the absence of vaccination. GSEA corroborated \nthese findings, identifying negative enrichment of pathways related to regulation of myeloid \ncell activation in immune response (NES -2.67), macrophage cytokine production (NES -2.3) \n(Supplementary table 6) and macrophage activation (NES -2.17) in the vaccinated group \n(Figure 5F).  \nTo validate these transcriptomic signatures at the protein level , we performed IHC \nanalysis on tumor tissue from vaccinated mice. Using CD68+ as a pan macrophage marker and \nCD206+ to define the M2- like phenotype36, we observed that  CaVaC OPT vaccination led to \na significant reduction of the CD206+/CD68+ ratio (p=0.024), as shown in Figure 5G. These \nfindings indicate that CaVaC OPT  vaccination effectively shifts macrophage polarization, \nreducing the proportion of M2 -like macrophages within the total macrophage population in \nmice.  \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 16 \n \nFIGURE 5. CaVac OPT vaccination promotes memory T cell differentiation and reduces \nM2-like pro-tumorigenic macrophage in the tumor microenvironment (TME).  C57BL/6 \nmice bearing subcutaneous MOC-2 tumors were treated with or without CaVac OPT.  On day \nthirty post tumor inoculation , tumors were collected and CD45⁺ TILs were isolated for bulk \nRNA sequencing. Experimental setup: mice were challenged with 2.5 x 10⁵ MOC-2 tumor cells \nand vaccinated on day  three and day twenty-four. Tumor volume was measured every three \ndays until day  thirty post tumor challenge . Statistical significance is indicated as follows: \n*p<0.05, ***p<0.001, ns: not significant. (A) PCA of VST bulk RNA sequencing data from \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 17 \nCD45+ TILs in control and CaVac OPT vaccinated samples.  Axes indicate the percentage of \nvariance explained by each principal component. Each dot represents one mouse. (B) Volcano \nplot comparing DEGs between control (n = 4) and vaccinated (n = 4) groups. Red dots denote \nsignificant genes (log₂FC > 0.5, adjusted p < 0.1).  (C) Boxplots illustrating the differences in \ncentral memory T cell (TCM) estimated by ImmuCellAI between control and vaccinated mice. \n(D) IHC staining of Tcf1 + cells in MOC -2⁺ tumors following prime and boost vaccination. \nRepresentative images from control and CaVac OPT -treated groups show staining for Tcf1 + \ncells. Brown indicates positive staining. Scale bar = 20 μm. Dot plots show the percentage of \nTcf1⁺ cells in vaccinated mice (n = 5) compared with those in the control mice (n = 6). ( E) \nHeatmap illustrating expression of macrophage-related DEGs between control and vaccinated \ntumors. (F) GSEA annotation of macrophage activation gene signatures. Ridge plot and \nheatmap shows gene set enriched in macrophage activation (NES -2.23) in control and \nvaccinated tumors. (G) IHC staining of tumor-infiltrating T cells in MOC-2 tumors following \nprime and boost vaccination. Representative images from control and CaVac OPT -treated \ngroups show staining for CD68⁺, CD206⁺. Brown indicates positive staining. Scale bar = \n20 μm. Violin plots illustrate the ratios of CD206⁺/CD68⁺ across treatment groups. Statistical \nanalysis was performed using the Mann–Whitney test.  \n \nDiscussion \nCancer vaccines are emerging as complementary treatments in tumors with limited responses \nto ICIs37. Recent progress in genetic vaccine platforms, particularly mRNA-based \ntechnologies, has reinvigorated interest in DNA -based vaccine approaches. We evaluated a \nnext-generation dbDNATM vaccine targeting the cancer-testis antigens MAGED4B and FJX1. \nAlthough assessed here in a preclinical model of HNSCC, this vaccine is broadly applicable to \nother solid tumors, as these antigens are widely expressed in other solid tumors 8,19,20. These \nantigens represent highly promising targets, enabling the development of effective “off -the-\nshelf” vaccines without the need for complex epitope  discovery or the logistical challenges \nassociated with personalized vaccine production. Moreover, CaVac OPT could serve as a \nbridging immunotherapy, initiating an early immune response while a personalized vaccine is \nbeing developed, thereby potentially enhancing overall clinical benefit. Unlike conventional \nplasmids, the compact size of dbDNA TM (2 kb smaller than a standard pDNA ) may enable \nefficient nuclear transport, explaining why the SV40 enhancer previously shown to enhance \nlarger DNA constructs38,39 did not improve immunogenicity here. Electroporation is likely to \nbypass the endosomal TLR9 signaling, rendering CpG motifs ineffective, while cytoplasmic \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 18 \npattern recognition receptors, such as cGAS -STING or RIG -I-like receptors , may instead \nmediate innate sensing of dbDNATM16.  \nThe vaccine constructs were individually optimized for each antigen, resulting in \ndistinct design sequence s informed by prior clinical efficacy data . To further refine  the \ndbDNATM vaccine, we focused on enhancing antigen presentation through strategic \nmodification to the antigenic sequence and incorporation of fusion partners that have proven \nto be critical for improvement of immunogenicity. Among several fusion partner s tested, the \ntetanus-derived DOM fragment and Flt3L elicited the highest T-cell responses. Unexpectedly, \ncertain fusion partners led to  poor T -cell induction in contrast with previous  published \nreports30,40. Additionally, removal of the MAGE HD  further exposed immunogenic \nMAGED4B epitopes, suggesting that peripheral tolerance to conserved family domains may \nlimit T-cell induction against MAGED4B.  \nCaVac OPT  vaccination converted cold HNSCC tumors into pro-inflammatory \nmicroenvironment, inducing robust CD8+ and CD4+ T cell responses  without an increase in \nTregs. Chemokine induction  (e.g. Ccl5) and homing receptor expression (Cxcr3) likely \nfacilitate efficient T cell trafficking41 into the TME and exert cytotoxic effects through IFN-γ \nand granzyme K, contributing to tumor cell killing and tumor volume reduction. Following \nvaccination, infiltrating CD8+ T cells exhibited stem-like memory (Tcf7+ and Lef1+),  a subset \nassociated with durable responses42 and clinical benefit  from PD1 blockade43,44, potentially \nexplaining the synergy with anti-PD1 therapy. In parallel, the vaccine reduced the proportion \nof immunosuppressive M2-like TAMs, which are abundant in HNSCC, where they contribute \nto tumor progression through immunosuppressive and pro-tumorigenic effects45–47. In HNSCC, \nTAMs are actively recruited to TME and establish direct contact with squamous carcinoma \ncells. Notably, C-C motif chemokine ligand 18 ( CCL18) produced by M2 macrophages has \nbeen shown to promote epithelial-mesenchymal transition ( EMT) and enhance cancer \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 19 \nstemness, thereby driving metastasis 48. Thus, the ability of our vaccine to reducing M2 \npolarization may play a crucial role in negatively enriching EMT pathways and enhancing anti-\ntumor immunity. Together, these findings position CaVac OPT  as a rational approach to \novercome immune exclusion in HNSCC by simultaneously enhancing T cell priming, memory \nformation and myeloid repolarization.  \nIn conclusion, CaVac OPT has demonstrated that the dbDNATM platform represents an \nimmunogenic cancer vaccine capable of eliciting durable, antigen-specific T cell responses. It \neffectively transforms the immunosuppressive TME through the reduction of M2-like CD206+/ \nCD68+ macrophages and results in significant tumor regression. Moreover, it exhibits a \nsynergistic effect when combined with anti -PD1 therapy, underscoring its promise as a \nplatform for future clinical testing. \n \nMethods \nGeneration of dbDNATM vaccines  \nAll dbDNATM vaccines were generated and subcloned into the proTLx -K STTM expression \nplasmid following previously described protocol49. Cancer-tests antigens including Melanoma-\nassociated antigen D4 isoform 1 (MAGED4B, NP_001258991) and human four -jointed box \nprotein 1 (FJX1, NP_055159) were retrieved from the NCBI database. Consensus sequences  \n(Con) were generated by aligning these sequences using the BLOSUM62 algorithm (default \nmatrix) in EMBOSS (European Bioinformatics Institute; EMBOSS Cons Multiple Sequence \nAlignment EMBL-EBI). Additional sequences, including human macrophage inflammatory \nprotein-1 alpha (huMIP-1a), and FMS-like tyrosine kinase 3 ligand (Flt3L), were obtained from \nNCBI and optimized for human codon usage with GenSmartTM Codon Optimization software \n(GenScript). To enhance secretion, the Mus Musculus IgH signal peptide sequence \nMGWSCIIFFLVATATGVHS was incorporated at the N -terminus of each construct. The \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 20 \nDOM fragment (TT865–1120)8 and the target antigens (MAGED4B or FJX1) were linked using \na seven-amino-acid spacer (AAAGPGP), as previously described 8. Variants of MAGED4B \nwith full (dHD 1) or partial homology domain deletions (dHD2) were generated via site -\ndirected mutagenesis. Briefly, plasmid templates contained an expression cassette flanked by \ndual TelN protelomerase recognition binding sites from E. coli phage N15. Templates were \ndenatured with NaOH and quenched in a reaction buffer containing custom pri mer, dNTPs, \nPhi29 polymerase and pyrophosphatase. The reaction mixture was incubated at 30°C for 30 or \n72 h. Concatemeric DNA was cleaved by adding TelN protelomerase. followed by digestion \nwith 200 units/mL of a template -specific restriction enzyme (New E ngland Biolabs, Hitchin, \nUK) and 200 units/mL exonuclease ExoIII (Enzymatics). The digest mixture was purified and \nprecipitated using polyethylene glycol (PEG) 8000. \n \nIn vivo mouse vaccination \nAll mouse -related procedures were approved by the Home Office under project license \nPP6990832. C57BL/6J mice (6 -10 weeks old, 18 to 22 g) were obtained from Charles River \nlaboratories (Kent, UK) and housed in ventilated cages at the University of Liverpool animal \nfacility.  \n \nTumor cell lines \nMouse oral cancer 2 (MOC -2) cells were cultured in IMDM/F12 (2:1) supplemented with 10 \n% foetal bovine serum (FBS; Gibco), 100 U/mL penicillin streptomycin (Gibco), 5 mg/mL \ninsulin (Sigma, I6634), 400 ng/mL hydrocortisone (Sigma, H0135), 5ng/mL epidermal growth \nfactor (EMD Milipore, 01 -107) at 37 °C, 5 % CO 2 in a humidified incubator as previously \ndescribed. To establish a homogenous MAGED4B - and FJX1- expressing cell line, MOC -2 \ncells were transduced with retrovirus particles generated using the pBABE-MAGED4B-FJX1-\npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 21 \npuro vector and HEK293T gag -pol packaging cells (Takara Bio, 631458). The human \nMAGED4B and FJX1 sequences were cloned into pBABE -Puro vector using NotI and XhoI \nrestriction sites, connected via a GSGSG linker for a single in -frame transcript. Transduced \ncells were selected with 4 µg/mL puromycin (Gibco ) and validated by  IHC using anti -\nMAGED4B (Santa Cruz, sc-393059) and anti-FJX1 (Atlas antibodies, HPA059220) antibodies \nrespectively. Cells were confirmed to be free of mycoplasma and rodent pathogens using PCR \nanalysis (Mouse/Rat CLEAR panel, Charles River) and used at low passage for in vivo studies. \n \nIn vivo immunogenicity and efficacy of dbDNATM vaccines  \nTo evaluate antigen-specific immune responses, mice were randomly assigned to experimental \ngroups and vaccinated  intramuscularly (i.m) in both anterior tibialis muscles  using \nelectroporation with a custom TriGrid® pulse generator (Ichor medical systems, San Diego, \nCA). Booster doses were administered 21 days after priming, optimized for eliciting T-cell \nresponses with DNA vaccines.  \nIn tumor-bearing models, mice were subcutaneously (s.c) injected  on day 0 with 1.6-\n2.0 ×105 cells MOC -2 expressing MAGED4B-FJX1+. Between day three and seven-post \ninoculation, mice were randomized to receive one of the following: a control  dbGFP vector, \ncombined dbDNATM of Basic 0 MAGED4B and Basic 0 FJX1, or an optimized formulation \n(CaVac OPT) dbDNATM dHD and Flt3L-FJX1. Dosing varied by experimental conditions, and \nall groups received boosters twenty-one days post-priming. Tumors were measured every three \ndays, and mice were euthanized when tumor volumes exceeded 500mm 3 or if ulceration \nreached humane endpoint thresholds. Tumor volume was calculated as 0.5 × (width2 x length). \nTumor specimens and splenocytes were harvested as required for downstream analysis. \n \nEx vivo IFN-γ ELISPOT  \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 22 \nSplenocytes were isolated via mechanical dissociation and Lymphocyte density gradient \n(Stemcell Technologies). Cells were cultured in RPMI supplemented with FBS, P/S, 2mM L -\nglutamine (Gibco), 1mM sodium pyruvate (Gibco), 2 -Mercaptoethanol (Merck) and 1x non -\nessential amino acids (Gibco). IFN-γ ELISpot assays were performed per BD Biosciences \nprotocols using 2 × 10⁵ splenocytes/well, stimulated with no peptide (background control), 1 \nµM tetanus p30 (Cambridge Research Biochemistry), or overlapping peptide po ols (OPP) of \nMAGED4B or FJX1 (JPT Peptide, Germany). IFN -γ secreting cells were detected using a \nbiotin-conjugated anti-mouse IFN-γ antibody (BD Biosciences, UK), followed by streptavidin-\nALP (Mabtech, UK) and BCIP/NBT substrate (Mabtech, UK). Spot-forming cells (SFCs) were \nquantified with an automated ELISPOT reader (Aelvis GmbH) and expressed as mean SFCs \nper 10⁶ cells16.  \n \nCombination therapy and T cell subset depletion \nTo evaluate synerg y between vaccination and immune checkpoint inhibition , tumor-bearing \nmice were assigned to four groups: (1) control vaccine dbGFP, (2) 4µg CaVac OPT+ IgG2a \nisotype control, (3) 4µg of CaVac OPT + anti-PD1 and (4) anti-PD1 alone. Beginning five days \npost vaccination, 200 µg of anti-PD1 (Clone RMP1 -14) or IgG2a isotype control ( clone 1-1) \nwas administered intraperitoneally (i.p.) every three or four days in four weeks, for a total of \neight doses. For T cell depletion studies, 500 µg of anti-CD4 (clone GK1.5) or anti-CD8 (clone \nYTS 169) mAbs via i.p. injection, while non-depleted controls received 500 µg of anti-IgG2b \nisotype control ( clone I -1034), for a total of five doses. Depletion was validated via flow \ncytometry after the fourth administration. \n \nTumor Dissociation and tumor infiltrating lymphocytes (TILs) Isolation \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 23 \nTumors were harvested and processed using the Mouse Tumor Dissociation Kit (Miltenyi \nBiotec, 130 -096-730) and the gentleMACS Dissociator (Miltenyi  Biotec) following the \nmanufacturer’s protocol. CD45 + TILs were isolated using Mouse CD45 TIL MicroBeads \n(Miltenyi Biotec, 130 -110-618). Biological replicates consisted of tumors harvested from \nindividual mice per treatment group. \n \nRNA Extraction and Sequencing \nRNA samples were extracted from TILs using the RNeasy mini -Kit (QIAGEN). RNA purity, \nconcentration, and integrity were assessed using a NanoDrop spectrophotometer , Qubit 2.0 \nFluorometer, and Agilent 2100 Bioanalyzer respectively. Only samples with RNA integrity \nnumber (RIN ≥ 4) were used for sequencing. Messenger RNA (mRNA) was enriched from \ntotal RNA using poly -T oligo attached magnetic beads (polyA selection). Libraries were \nprepared, pooled, and sequenced on the Illumina NovaSeq 6000 platform (PE150 mode ) \n(Novogene) to a minimum depth of 6 Gb reads per sample. Raw reads were quality trimmed to \nremove adapter sequences, reads containing >10% ambiguous bases (N), or low-quality bases. \nClean reads  were aligned to the Mus musculus  reference genome (mm39, NCBI RefSeq \nGCF_000001635.27) using HISAT2 (version 2.0.5)50.  \n \nGene expression and pathway analysis \nGene c ounts were normalized and analyzed for differential expression analysis using the \nDESeq2 package (v 1.48.0) in R (v 4.5.0) statistical software51. Differentially expressed genes \n(DEGs) were defined by a fold change (FC) ≥ 1.5 and an adjusted P -value (padj) < 0.1. \nPrincipal component analysis (PCA) plot was visualized using the “ggplot2” R package  (v \n3.5.2)52. The clustering analysis of DEGs was carried out using the “tidyheatmaps” package (v \n0.2.1) of R 53. The enrichment of pathways was identified by GSEA method using \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 24 \nclusterProfiler (v 4.16.0) 54. Functional terms were retrieved from the Mouse Signature \nDatabase (MSigDB) gene sets R package (v 10.0.2), including the biological process (BP), \nmolecular function (MF), and cellular component (CC) 55,56. Then, the analysis was performed \nto identify significantly different regulatory pathways using the Kyoto Encyclopedia of Genes \nand Genomes (KEGG), a major public pathway -related database. The results of GSEA were \nvisualised with ggplot2 and enrichplot (v 1.28.0)57.  \n \nImmune cell infiltration profiling \nImmune cell composition within TIL s was inferred using ImmuCellAI-mouse \n(http://bioinfo.life.hust.edu.cn/ImmuCellAI-mouse/)58. Normalized gene expression matri ces \nfrom treatment were uploaded to predict the relative abundance of immune subsets.  \n \nImmunohistochemistry (IHC) \nTumor samples were fixed in PBS with 4% paraformaldehyde before processing into formalin-\nfixed paraffin-embedded (FFPE) blocks. Sections of 3 µm were cut and processed on the Leica \nBond-RX™ automated staining platform for IHC. Following antigen retrieval and blocking \nwith 10% normal goat serum in PBS (Cell Signaling Technology) , tissue  sections were \nincubated with primary antibodies  to CD8α (1:100, D4W2Z, Cell Signaling), CD4 + (1:100, \nD7D2X, Cell Signaling), FoxP3+ (1:200, D6O8R, Cell Signaling), CD68+ (1:600, E3O7V, Cell \nSignaling), CD206 + (1:200, E6T5J, Cell Signaling) , TCF1 /TCF7 (1:50 , C63D9, Cell \nSignaling). Antigen retrieval was performed using the Bond ™ Polymer Refine Detection Kit \nwith either ER1 buffer (10mM sodium citrate buffer) for 20 minutes or ER2 buffer (Tris-EDTA \nbuffer pH 9.0) for 20 minutes (used for FoxP3 staining). Detection was carried out using an \nHRP-conjugated secondary antibody, followed by DAB chromogenic staining. Sections were \ncounterstained with hematoxylin and mounted for microscopic analysis.  \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 25 \n \nQuantitative image analysis \nSlides were scanned at 20× magnification using PhenoImager Fusion (v2.3.1, Akoya \nBiosciences). Images were visualized and annotated using QuPath v0.5.111. For each tumor \nsample, ten regions of interest (ROIs) measuring 500 µm² were selected and analyzed using \nautomated cell detection. Colour deconvolution was applied to separate individual stains, and \nimmune cell populations (CD8+, CD4+, FoxP3+, CD68+, CD206+, Tcf1/7+) were quantified \nusing QuPath’s positive cell detection  algorithm. Quantification results  were exported and \nfurther processed in Microsoft Excel. \n \nStatistical analysis \nStatistical significance was determined using a two -tailed Mann -Whitney test (two -group \ncomparisons), one -way ANOVA (multi -group comparisons), or log -rank Mantel -Cox test \n(survival analysis) in Prism 10.4.2 (GraphPad, CA, USA). P -values ≤ 0.05 were conside red \nsignificant (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). \n \nAcknowledgements   \nThe authors gratefully acknowledge the skilled technical support provided by the Biomedical \nServices Unit at the University of Liverpool . We are also grateful to the team at Research \nHistology at the University of Southampton for their expert assistance with \nimmunohistochemical staining. \n \nAuthor contributions \nC.W. and S.K. conducted experiments related to vaccine design and construction. G.M.Y.T., \nC.W., and M.M. performed in vivo experiments and data acquisition. G.M.Y.T. analyzed the \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 26 \nbulkRNA-seq data. C.W., L.C. and H.H. contributed to design of vaccine candidates. G .J.T \ncontributed to developing methodology . K .P.L., S.C.C., contributed to conceptualization.  \nC.H.O. contributed to conceptualization and data interpretation. N .S. conceived the ideas, \ndesigned the study and supervised the experiments. G.M.Y.T. , M.M. and N.S. wrote the \nmanuscript. Every author contributed to the significant revision of the manuscript for essential \nintellectual content and gave their final approval for the version to be published. \n \nCompeting interests: N.S., C.H.O., G.J.T., S.C.C., C.W. and K.P. are the authors of the patent \ncovering cancer vaccine targeting MAGED4B and FJX1. N .S. received funding from \nTouchlight Genetics.  \n \nData and materials availability: Sequencing data of bulk RNA -seq were deposited at GEO \nsubmission with the accession numbers: GSE310571 \n \n \n \n \n  \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 27 \n \nReferences  \n1. Awad, M. M. et al. Personalized neoantigen vaccine NEO-PV-01 with chemotherapy and \nanti-PD-1 as first-line treatment for non-squamous non-small cell lung cancer. Cancer Cell \n40, 1010-1026.e11 (2022). \n2. Rojas, L. A. et al. Personalized RNA neoantigen vaccines stimulate T cells in pancreatic \ncancer. Nature 618, 144–150 (2023). \n3. Kleine-Kohlbrecher, D. et al. 623 AI-designed personalized neoantigen vaccine, EVX-02, \ninduces robust T-cell responses in melanoma patients. J Immunother Cancer 11, (2023). \n4. Shaw, H. M. et al. A DNA plasmid melanoma cancer vaccine, SCIB1, combined with \nnivolumab + ipilimumab in patients with advanced unresectable melanoma: Efficacy and \nsafety results from the open-label phase 2 SCOPE trial. JCO 42, 9535–9535 (2024). \n5. Yarchoan, M. et al.  Personalized neoantigen vaccine and pembrolizumab in advanced \nhepatocellular carcinoma: a phase 1/2 trial. Nature Medicine 30, 1044–1053 (2024). \n6. Kim, T. J. et al. Clearance of persistent HPV infection and cervical lesion by therapeutic \nDNA vaccine in CIN3 patients. Nat Commun 5, 5317 (2014). \n7. McCann, K. J. et al. Targeting Carcinoembryonic Antigen with DNA Vaccination: On -\nTarget Adverse Events Link with Immunologic and Clinical Outcomes. Clin Cancer Res \n22, 4827–4836 (2016). \n8. Wang, C. et al. DNA Vaccines Targeting Novel Cancer -Associated Antigens Frequently \nExpressed in Head and Neck Cancer Enhance the Efficacy of Checkpoint Inhibitor. Front. \nImmunol. 12, (2021). \n9. Trimble, C. L. et al. Safety, efficacy, and immunogenicity of VGX -3100, a therapeutic \nsynthetic DNA vaccine targeting human papillomavirus 16 and 18 E6 and E7 proteins for \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 28 \ncervical intraepithelial neoplasia 2/3: a randomised, double -blind, placebo -controlled \nphase 2b trial. The Lancet 386, 2078–2088 (2015). \n10. Aggarwal, C. et al.  Immunotherapy Targeting HPV16/18 Generates Potent Immune \nResponses in HPV -Associated Head and Neck Cancer. Clin Cancer Res  25, 110 –124 \n(2019). \n11. Hillemanns, P. et al. Safety and efficacy of the therapeutic DNA -based vaccine VB10.16 \nin combination with atezolizumab in persistent, recurrent or metastatic HPV16 -positive \ncervical cancer: a multicenter, single -arm phase 2a study. J Immunother Cancer  13, \ne010827 (2025). \n12. Chen, Z. Y., He, C. Y., Meuse, L. & Kay, M. A. Silencing of episomal transgene expression \nby plasmid bacterial DNA elements in vivo. Gene Ther 11, 856–864 (2004). \n13. Ledesma-Feliciano, C. et al. Improved DNA Vaccine Delivery with Needle-Free Injection \nSystems. Vaccines 11, (2023). \n14. Walters, A. A. et al. Comparative analysis of enzymatically produced novel linear DNA \nconstructs with plasmids for use as DNA vaccines. Gene Ther 21, 645–652 (2014). \n15. Scott, V. L. et al.  Novel synthetic plasmid and Doggybone TM DNA vaccines induce \nneutralizing antibodies and provide protection from lethal influenza challenge in mice. \nHum Vaccin Immunother 11, 1972–1982 (2015). \n16. Allen, A. et al. Linear doggybone DNA vaccine induces similar immunological responses \nto conventional plasmid DNA independently of immune recognition by TLR9 in a pre -\nclinical model. Cancer Immunology, Immunotherapy : CII 67, 627 (2018). \n17. Cunliffe, R. F. et al. Optimizing a linear ‘Doggybone’ DNA vaccine for influenza virus \nthrough the incorporation of DNA targeting sequences and neuraminidase antigen. \nDiscovery Immunology 3, kyad030 (2024). \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 29 \n18. Stirling, D. C. et al. Modifying non-coding regions of linear DNA vaccines to explore the \ninterplay of expression and inflammation in immunogenicity. Human Vaccines & \nImmunotherapeutics 21, 2430826 (2025). \n19. Abou Kors, T. et al.  Multi-omics analysis of overexpressed tumor -associated proteins: \ngene expression, immunopeptide presentation, and antibody response in oropharyngeal \nsquamous cell carcinoma, with a focus on cancer -testis antigens. Front Immunol  15, \n1408173 (2024). \n20. Chong, C. E. et al. Over-expression of MAGED4B increases cell migration and growth in \noral squamous cell carcinoma and is associated with poor disease outcome. Cancer Lett \n321, 18–26 (2012). \n21. Gan, C. P. et al. Cancer vaccine overcomes immune evasion of nasopharyngeal carcinoma \nby restoring MHC-I through transcriptional regulation of NLRC5. J Transl Med 23, 1414 \n(2025). \n22. Al-Greene, N. T. et al. Four Jointed Box 1 Promotes Angiogenesis and Is Associated with \nPoor Patient Survival in Colorectal Carcinoma. PLoS One 8, e69660 (2013). \n23. Luo, W.-Z. et al. MAGED4B is a Poor Prognostic Marker of Stomach Adenocarcinoma \nand a Potential Therapeutic Target for Stomach Adenocarcinoma Tumorigenesis. Int J Gen \nMed 16, 1681–1693 (2023). \n24. Dean, D. A., Dean, B. S., Muller, S. & Smith, L. C. Sequence Requirements for Plasmid \nNuclear Import. Experimental Cell Research 253, 713–722 (1999). \n25. Coban, C. et al. Toll-like receptor 9 mediates innate immune activation by the malaria \npigment hemozoin. J Exp Med 201, 19–25 (2005). \n26. Yu, Y.-Z. et al. Three types of human CpG motifs differentially modulate and augment \nimmunogenicity of nonviral and viral replicon DNA vaccines as built -in adjuvants. Eur J \nImmunol 43, 228–239 (2013). \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 30 \n27. Low, L. et al. DNA vaccination with electroporation induces increased antibody responses \nin patients with prostate cancer. Hum Gene Ther 20, 1269–1278 (2009). \n28. Bhardwaj, N. et al. Flt3 ligand augments immune responses to anti -DEC-205-NY-ESO-1 \nvaccine through expansion of dendritic cell subsets. Nat Cancer 1, 1204–1217 (2020). \n29. Pulendran, B. et al.  Flt3-Ligand and Granulocyte Colony -Stimulating Factor Mobilize \nDistinct Human Dendritic Cell Subsets In Vivo1. The Journal of Immunology  165, 566–\n572 (2000). \n30. Duperret, E. K., Yan, J. & Weiner, D. B. Designing consensus immunogens to break \ntolerance to self-antigens for cancer therapy. Oncotarget 9, 35513–35514 (2018). \n31. Cameron, B. J. et al. Identification of a Titin -Derived HLA-A1–Presented Peptide as a \nCross-Reactive Target for Engineered MAGE A3–Directed T Cells. Science Translational \nMedicine 5, 197ra103-197ra103 (2013). \n32. Eakins, R. A. et al. Chemical complementarity between tumor resident, T -cell receptor \nCDR3s and MAGEA3/6 correlates with increased melanoma survival: Potential relevance \nto MAGE vaccine auto-reactivity. Molecular Immunology 150, 58–66 (2022). \n33. Kono, M., Saito, S., Egloff, A. M., Allen, C. T. & Uppaluri, R. The mouse oral carcinoma \n(MOC) model: A 10-year retrospective on model development and head and neck cancer \ninvestigations. Oral Oncology 132, 106012 (2022). \n34. Judd, N. P., Allen, C. T., Winkler, A. E. & Uppaluri, R. Comparative analysis of tumor -\ninfiltrating lymphocytes in a syngeneic mouse model of oral cancer. Otolaryngol Head \nNeck Surg 147, 493–500 (2012). \n35. Zolkind, P. et al.  Cancer immunogenomic approach to neoantigen discovery in a \ncheckpoint blockade responsive murine model of oral cavity squamous cell carcinoma. \nOncotarget 9, 4109 (2017). \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 31 \n36. Jablonski, K. A. et al. Novel Markers to Delineate Murine M1 and M2 Macrophages. PLoS \nOne 10, e0145342 (2015). \n37. Nutsch, K. et al. TIGIT and PD-L1 co-blockade promotes clonal expansion of multipotent, \nnon-exhausted antitumor T cells by facilitating co -stimulation. Nat Cancer 5, 1834–1851 \n(2024). \n38. Fujisawa, J., Seiki, M., Sato, M. & Yoshida, M. A transcriptional enhancer sequence of \nHTLV-I is responsible for trans-activation mediated by p40 chi HTLV-I. EMBO J 5, 713–\n718 (1986). \n39. Barouch, D. H. et al.  A Human T -Cell Leukemia Virus Type 1 Regulatory Element \nEnhances the Immunogenicity of Human Immunodeficiency Virus Type 1 DNA Vaccines \nin Mice and Nonhuman Primates. J Virol 79, 8828–8834 (2005). \n40. Braathen, R. et al. The Magnitude and IgG Subclass of Antibodies Elicited by Targeted \nDNA Vaccines Are Influenced by Specificity for APC Surface Molecules. \nImmunohorizons 2, 38–53 (2018). \n41. Kuo, N. et al. Chemokine-chemokine receptor networks in conventional type I dendritic \ncells: an opportunity to prime and boost anticancer immunity. The Journal of \nPharmacology and Experimental Therapeutics 392, 103725 (2025). \n42. Siddiqui, I. et al.  Intratumoral Tcf1+PD -1+CD8+ T Cells with Stem -like Properties \nPromote Tumor Control in Response to Vaccination and Checkpoint Blockade \nImmunotherapy. Immunity 50, 195-211.e10 (2019). \n43. Sade-Feldman, M. et al. Defining T Cell States Associated with Response to Checkpoint \nImmunotherapy in Melanoma. Cell 175, 998-1013.e20 (2018). \n44. Miller, B. C. et al. Subsets of exhausted CD8+ T cells differentially mediate tumor control \nand respond to checkpoint blockade. Nat Immunol 20, 326–336 (2019). \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 32 \n45. Haque, A. S. M. R. et al. CD206+ tumor-associated macrophages promote proliferation \nand invasion in oral squamous cell carcinoma via EGF production. Sci Rep  9, 14611 \n(2019). \n46. Alves, A. et al. Tumor microenvironment and Oral Squamous Cell Carcinoma: A crosstalk \nbetween the inflammatory state and tumor cell migration. Oral Oncology  112, 105038 \n(2021). \n47. Huang, W., Jiang, M., Lin, Y., Qi, Y. & Li, B. Crosstalk between cancer cells and \nmacrophages promotes OSCC cell migration and invasion through a CXCL1/EGF positive \nfeedback loop. Discov Onc 15, 145 (2024). \n48. She, L. et al.  Tumor-associated macrophages derived CCL18 promotes metastasis in \nsquamous cell carcinoma of the head and neck. Cancer Cell International 18, 120 (2018). \n49. Karda, R. et al. Production of lentiviral vectors using novel, enzymatically produced, linear \nDNA. Gene Ther 26, 86–92 (2019). \n50. Mortazavi, A., Williams, B. A., McCue, K., Schaeffer, L. & Wold, B. Mapping and \nquantifying mammalian transcriptomes by RNA-Seq. Nat Methods 5, 621–628 (2008). \n51. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion \nfor RNA-seq data with DESeq2. Genome Biol 15, 550 (2014). \n52. Wickham, H. Ggplot2. (Springer International Publishing, Cham, 2016). doi:10.1007/978-\n3-319-24277-4. \n53. Mangiola, S. & Papenfuss, A. T. tidyHeatmap: an R package for modular heatmap \nproduction based on tidy principles. Journal of Open Source Software 5, 2472 (2020). \n54. clusterProfiler: an R Package for Comparing Biological Themes Among Gene Clusters - \nPMC. https://pmc.ncbi.nlm.nih.gov/articles/PMC3339379/. \n55. Castanza, A. S. et al.  Extending support for mouse data in the Molecular Signatures \nDatabase (MSigDB). Nat Methods 20, 1619–1620 (2023). \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint \n\n 33 \n56. Subramanian, A. et al. Gene set enrichment analysis: A knowledge -based approach for \ninterpreting genome-wide expression profiles. Proceedings of the National Academy of \nSciences 102, 15545–15550 (2005). \n57. Guangchuang, Y. enrichplot: Visualization of Functional Enrichment Result. \nhttps://doi.org/doi:10.18129/B9.bioc.enrichplot doi:doi:10.18129/B9.bioc.enrichplot. \n58. Miao, Y. -R. et al.  ImmuCellAI-mouse: a tool for comprehensive prediction of mouse \nimmune cell abundance and immune microenvironment depiction. Bioinformatics 38, 785–\n791 (2022). \n \n \n \npreprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. \nThe copyright holder for thisthis version posted January 23, 2026. ; https://doi.org/10.64898/2026.01.21.700768doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}