Intra-tumoral injection of caerin 1.1 and 1.9 peptides in vaccinated TC-1 tumour bearing mice with PD-1 blockade modulates macrophage heterogeneity and the activation of CD8+ T cells in the tumour microenvironment

Development of a vaccine formula that alters the tumour-inltrating lymphocytes to be more immune active against a tumour is key to the improvement of clinical responses to immunotherapy. Here, we demonstrate that, in conjunction with E7 antigen specic immunotherapy, and IL-10 and PD-1 blockade, intra-tumoral administration of caerin 1.1 and 1.9 peptides further improves the tumour microenvironment (TME) when compared with injection of a control peptide. We used single cell transcriptomics and mass spectrometry-based proteomics to quantify changes in cellular activity across different cell types within the TME. We show that the injection of caerin 1.1/1.9 increases immune activating macrophages and NK cells, while reducing immunosuppressive macrophages with M2 phenotype, and increased numbers of activated CD8+ T cells with higher populations of memory and effector-memory CD8+ T subsets. Proteomic proling demonstrated activation of Stat1 modulated apoptosis and production of nitric oxide. Further, computational integration of the proteome with the single cell transcriptome was consistent with deactivation of immune suppressive B cell function following caerin 1.1 and 1.9 treatment. quantitative activated CD8+ T cells, mediate proinammatory apoptosis of Stat1 key role interacting with the regulators on multiple pathways. demonstrated that caerin 1.1/1.9 containing treatment may result in improved antitumor Harnessing the novel candidate genes preferentially enriched in the immune active cell populations may allow further exploration of distinct macrophages, T cells and their functions in TC-1 tumour, providing a valuable resource for researchers in the eld. (E) 10 in 1.1/1.9 and (F) Comparison of biological processes enriched in the signicantly elevated proteins in caerin 1.1/1.9 and P3 groups, respectively, with respect to untreated group. (G) The correlation between the gene expressions (in 19 cell populations) of the proteins showing signicant upregulation only in caerin 1.1/1.9 group, and the fold change of these proteins in relative to untreated group. The KEGG pathways enriched in caerin 1.1/1.9 and P3 groups (P<0.05).

Intra-tumoral injection of caerin 1.1 and 1.9 peptides in vaccinated TC-1 tumour bearing mice with PD-1 blockade modulates macrophage heterogeneity and the activation of CD8+ T cells in the tumour microenvironment Cancer therapeutic vaccines aim at eliciting effector T cells, especially tumour antigen speci c CD8 + T cells that target tumour cells, without affecting the normal cells or tissues unaffected as demonstrated in clinical trials, such as those against melanoma 10 . However, vaccine-induced regression of high risk human papillomavirus infection related high-grade CIN lesions 11 , but not established cervical cancers 12 were observed. Immunisation and simultaneous blocking of the cytokine interleukin 10 (IL-10) drastically increases vaccine induced antigen speci c CD8 + T cell responses and improved tumour growth inhibition in a prophylactic setting compared with the same vaccination without IL-10 signalling blockade 13 .
Tumour inhibition was also improved in a therapeutic setting by intraperitoneal administration of anti-IL-10 receptor antibodies 14 .
The immunosuppressive tumour microenvironment (TME) can dampen the function of tumour in ltrating effector T cells 15 . The TME promotes the development of tumour associated macrophages, myeloid derived suppressive cells, B cells, Th2 type T and regulatory T cells. Tumour associated macrophages (TAM) are attracting increasing attention as they play key roles in tumour spread and in response to different therapies 16 . TAM can substantially accelerate the progression of untreated tumours, but also in uence the e cacy of anticancer drugs, including checkpoint blockade immunotherapies 17,18 . Speci cally, TAM can assume opposing phenotypes and functions that are either tumoricidal (e.g., M1like MΦ) or tumour-supportive (e.g., M2-like MΦ).
Host-defence caerin peptides isolated from Australian amphibians, Litoria genus, were observed to be active against cancer cells, such as TC-1 19 , HeLa 20, 21 . Caerin 1.1 and 1.9 inhibit HIV-infected T cells within minutes post-exposure at concentrations non-toxic to T cells, and inhibit the transfer of HIV from dendritic cells (DCs) to T cells 22 . They are cytotoxic to HPV18+ HeLa cells, and HPV 16 early protein E6/E7 transformed TC-1 cells in vitro, and the anti-tumour effects were more profound when caerin 1.1 and 1.9 were used together 19 . Moreover, caerin1.1 and 1.9 were able to inhibit TC-1 growth in vivo when locally injected intratumorally, and the inhibition requires an intact adaptive immune system 13,20 . The signalling of TNFα mediated apoptosis and T-cell receptor was stimulated after HeLa cells were treated with the mixture of caerin 1.1 and 1.9 21 . Activation of TCR pathway observed in proteomic analysis suggested that HeLa cells became more sensitive to T cell mediated killing 21 .
In this study, TC-1 tumour bearing mice immunised with a HPV16 E7 peptide-based vaccine containing anti-IL-10 receptor antibody and PD-1 blockade were locally injected with a mixture of caerin 1.1 and 1.9 peptides. Tumour-in ltrating lymphocytes (TILs) were isolated for scRNA-seq analysis to reveal the cell types of TILs in TC-1 tumour and the modulation of the TIL landscape by the immunotherapy containing caerin 1.1/1.9. Mass spectrometry-driven quantitative proteomic analysis was performed to investigate the overall effect of the changes in TILs. Our study provides new insights into the heterogeneity of TILs and their functions in TC-1 tumour, including novel markers to de ne immune activating macrophages and CD8 + T cell subpopulations, as well as the molecular mechanisms underlying TME modulation by caerin 1.1/1.9.

Results
Single cell RNA-seq identi ed six different macrophage populations in TC-1 tumour In this study, untreated TC-1 tumour bearing mice and tumour bearing mice immunised with a HPV16 E7 peptide-based vaccine containing anti-IL-10 receptor antibody and PD-1 blockade were locally injected with a mixture of caerin 1.1 and 1.9 peptides (molar ratio 1:1) ("caerin") or a control peptide. The two treatment groups showed signi cantly reduced tumour mass when compared with the untreated group, with a reduction of 69% (caerin, P=0.0011) and 42% (control peptide, P=0.0352). An additional injection of caerin further reduced the tumour weights, though the reduction was insigni cant (P=0.11). Elispot results from spleen and draining lymph nodes (LN) of individual mice (3 mice) indicate that two immunotherapy and peptide treated groups demonstrated similar E7 speci c CD8 + T cells in the spleen and draining lymph nodes (Fig. 1A).
Thirty days after TC-1 tumour challenge, total viable CD45 + leukocytes were isolated from tumours using a tumour in ltrating cell isolation kit ( Fig. 1B and Supplementary Data 1). Gene expression data from cells extracted from the control tumours and the two treated tumours were aligned and projected in a 2dimensional space through t-stochastic neighbour embedding (t-SNE) to allow identi cation of tumour associated immune cell populations and the overlapping patterns associated with the control tumours and the two treated tumour groups ( Fig. 1C and Supplementary Fig. 1 27 were also identi ed as contaminants (also see Supplementary Data 2 and Supplementary File 1).
The macrophage populations were examined in more detail, and expression of the top 20 marker genes of each macrophage population was compared to their expression across all macrophage cell populations, and to the expression of other genes in the same macrophage population (Fig. 1F). Speci c gene expression patterns differentiating these MΦ populations and certain overlaps could be determined. Many of the top 20 marker genes of cluster 1, including Pf4, Arg1, Fabp5, and Mmp12, were associated with M2 MΦ (Fig. 1F; Supplementary Data 2), and similar numbers of these cells were detected in tumours treated with caerin peptides or control peptide. (Supplementary Fig. 2C and 2D). The top three highly expressed genes with signi cance in cluster 2 were H2-Eb1, Cd74 and C1qc (Supplementary Data 2 and Supplementary File 1), con rming a MΦ signature. These genes were found to mark MHCII hi border associated MΦ in mouse brain 26 . The high expressions of several H-2 (MHCII) members, including H2-Eb1, H2-Ab1, H2-Aa, H2-DMb1 and H2-DMa, were also con rmed in cluster 2, thus we considered these as MHCII hi MΦ hereafter. Cluster 3 had a mixed cell phenotype, including proin ammatory MΦ (Cxcl10, Gbp2, and Thbs1), Ly6c hi in ltrating MΦ (Chil3 and Plac8), and dendritic cells (Rsad2, I t1, I t2 and  I 205). Thus, this cluster was labelled as MΦ/DCs. Increase in the size of this cluster in caerin peptide and control peptide treated tumours suggested that analysis of its subpopulations might help to further clarify its function. It was mainly composed of macrophages (Nop16, Gatm and Pf4) and NKT cells (Ntpcr, Mrpl28 and Commd10), and was identi ed as MΦ/NKT. Due to the exclusive high expression of Ear2 in cluster 9, it was assigned as Ear2 hi MΦ (Chil3, Adgre5, Ace, and I tm6) 28 2C). In contrast, the proportions of MHCII hi MΦ were elevated by approximately 5-fold (caerin peptide) and 4-fold (control peptide) treated tumours, and MΦ/DCs and Ear2 hi MΦ were similarly increased in the caerin and control peptide treated tumours.
We next sought to unravel the phenotype and functions behind the speci c gene expression patterns of each macrophage. The expressions of key lineage-associated genes of M2 MΦ were compared in parallel with their expressions in other macrophages (Fig. 2D). Several marker genes appeared exclusive to M2 MΦ, such as Mmp12, Arg1, Mmp13 and Slc6a8, for which expression is associated with tumour angiogenesis and invasiveness 32,33,34 . The role of Arg1 in immunosuppression has been described elsewhere 35 . Some marker genes of M2 MΦ were also highly expressed in TAMs, suggesting potential similarity between these two macrophages in terms of cellular function.
The distribution of M2 MΦ in untreated tumours and tumours injected with caerin peptides was compared in Fig MHCII hi MΦ were signi cantly increased in control and caerin peptide treated tumours, more signi cantly with caerin treatment. The expression of selected marker genes across six macrophage populations were displayed in Fig. 3A, which suggests Lira5, Cxcl9, Dnase1l3 and Cd300e as the signatures exclusive to MHCII hi MΦ. Cadm1, Cxcl9 and Cd300e expression was increased in macrophages, and Dnase1l3 has been reported as a signature of CD141 + CLEC9A + DCs 36 . In addition, Clec12a was recently found to highly expressed in myeloid cells including macrophages and DC subsets 37 . Thus, the MHCII hi MΦ cluster displayed a characteristic of mixing phenotypes and its subpopulations were further investigated.
A total of ve subpopulations of MHCII hi MΦ were identi ed and projected in a 2-dimensional tSNE space, with the subpopulation 0, 1 and 2 possessing the highest cell numbers (Fig. 3B). Subpopulation 3 was present in control and caerin peptide treated tumours in similar numbers, while subpopulation 4 was negligible in untreated tumours (Supplementary Data 3). The normalised cell numbers of ve subpopulations with different treatments were compared in Fig. 3C, and caerin peptide treatment stimulated much higher number of subpopulations 0, 1 and 2 when compared to control peptide treatment (group C), with fold changes of 4.22, 8.14 and 3.89 relative to the untreated group. The expressions (Log2 FC) of the top 10 marker genes of each subpopulation were compared across all ve subpopulations (Fig. 3D) The correlations amongst these subpopulations were evaluated based on the upregulated genes ( Fig.   3E). C1_MHCII hi -IFIT and C3_MHCII hi -DCs were correlated to a much higher degree, compared to the connections amongst other three clusters. Genes that were shared between C1_MHCII hi -IFIT and C3_MHCII hi -DCs revealed biological processes mutually exerted by these two subpopulations, including metabolism of lipids and lipoproteins, G-protein signalling and FCGR activation. Reactome pathways based on marker genes that were unique to each subpopulation were analysed ( Supplementary Fig. 5).
C0_MHCII hi -CXCL2 showed an enrichment in caspase-mediated cleavage of cytoskeletal proteins, immune system, apoptotic cleavage of cellular proteins and apoptotic execution phase. The signalling of interferon, interferon gamma and cytokine in immune system were detected in C1_MHCII hi -IFIT with signi cance. The pathways found in C2_MHCII hi -ResMΦ were less signi cant (high P-values) than those in other subpopulations, such as GPCR ligand binding, chemokine receptors bind chemokines and collagen formation, which were less relevant to activating immune response. C3_MHCII hi -DCs showed enrichment in haemostasis, GPVI-mediated activation cascade and adaptive immune system. Many cell cycle related pathways were found enriched in C4_MHCII hi -PROG (Supplementary Data 3).
The populations of Ear2 hi MΦ were remarkably elevated in control peptide and particularly in caerin peptide treated tumours ( Supplementary Fig. 6A). The distribution of cells expressing selected proin ammatory marker genes appeared aligned well with Ear2 hi MΦ in caerin peptide compared to untreated and control peptide treated tumours. Signi cant upregulation of Ear2, Ace, Adgre4, Serpinb2 and Prtn3 in Ear2 hi MΦ was identi ed in the caerin treated tumours compared to untreated tumours ( Supplementary Fig. 6B). A High degree of gene expression concordance c was present amongst Ear2 hi MΦs, yet distinct biological processes were found in subgroups Ear2 hi MΦ ( Supplementary Fig. 6C).
Ear2 hi MΦ showed the suppression of many biological processes, such as transferase activity, phosphorylation, and cellular protein metabolism in untreated tumours, when compared to the activation of metabolic processes in control peptide treated tumours. Caerin treated tumours had activated cellular structure remodelling and immune response genes including those suggesting myeloid cell differentiation.
More CD8 + T cells in ltrate to TC-1 tumour following vaccination and PD-1 blockade, and CD8 + T cells are more activated in caerin 1.1/1.9 treatment group B In control peptide and caerin peptide treated tumours, the population of CD8 + T cells in ltrating to TC-1 tumour was was 3.73% (caerin) and 4.58% (control peptide) of the total CD45 + cells compared to only 0.27% in untreated tumours (Supplementary Data 1). With peptide treatment, CD8 + T cells were relatively separated from other three T cell populations on the tSNE graph ( Fig. 4A), indicating a possible variation in function. Also, all T cell types showed a more than 10-fold increase between untreated tumours to 40% (caerin) and 37% (control peptide) (Fig. 4B). Analysis of the gene expression pattern in the four T cell populations ( Fig. 4C.) showed that expression of most of these genes was higher in CD8 T cells than that in any other T cell type (see Supplementary Fig. 5

for the expression of marker genes in other T cells),
including genes that enhance the activation of CD8 + T cells, such as Ucp2, Fth1, Apoe, Fcer1g and Calm3.
Since pDCs present antigens (Ag) and induce immunogenic T cell responses through differentiation of cytotoxic CD8 + T cells and effector CD4 + T cells 48,49 , we compared the gene expression of signature genes of CD8 + T cells across four types of T cells and pDCs (Fig. 4D). It shows that Cd8a, Klrc1 and Lag3 were almost exclusive to CD8 + T cells, while comparable expression of Cxcr6 was observed in CD4 -CD8 -T cells, and lower expression of Cd8a, Cd8b1 and Lag3 was observed in pDCs. Ribosome was determined to be the most enriched KEGG pathway, followed by T cell receptor signalling and natural killer cell mediated cytotoxicity (Fig. 4E). The top 25 enriched biological processes in CD4 + CD8 + , CD8 + , CD4 + CD25 + and CD4 -CD8 -T cells were compared in Fig. 4F. Translation was found to be the most enriched process in the T cell subsets except CD4 + CD25 + , and was subsequently excluded to highlight the difference amongst other enriched processes (Supplementary Data 5). Since these cells share similar T cell lineage development, overlaps of certain biological processes were observed, such as T cell differentiation, T cell receptor signalling, and innate immune response was observed as expected. However, T cell relevant processes were more enriched in CD8+ T cells suggested by lower P-values compared to other three subtypes. Furthermore, there were a set of processes only enriched in CD8 + T cells, such as positive regulation of histone deacetylation, regulation of translational initiation, chromosome organisation, activation of cysteine-type endopeptidase activity involved in apoptotic process and regulation of cytokine production ( Fig. 4F and Supplementary Data 5).
The expression of marker genes, including Cd8a, Cd8b1, Tox, Lag3, Ifng, Nkg7, Nrgn, Gldc, Prf1, Abcb9, Nrn1 and Rgs16, were compared amongst untreated, caerin, and control peptide treated tumours (Fig.  5A), where signi cant upregulations of these genes induced by peptide treatment were observed, except Nrn1 for the control peptide treatment. In addition, Cd8a, Cd8b1, Tox, Ifng, Prf1 and Rgs16 were signi cantly elevated by caerin when compared to control peptide, suggesting that CD8 + T cells were more activated by the caerin peptide. The subpopulations of CD8 + T cells were further investigated to reveal the changes of heterogeneity due to peptide treatment, and ve subpopulations were identi ed ( Fig. 5B; Supplementary Fig. 8). We found that the signatures representing naïve T cells, including Sell, Lef1 and Tcf7, had higher expression in untreated tumour (Fig. 5C). Peptide treatment caused elevation of signatures for exhausted T cells, such as Tigit, Lag3, Tox and Pdcd1, while the effector T cells were stimulated by two treatments, suggested by the upregulation of Gzmb and Prf1. The expression of these genes in the ve CD8 subpopulations was also compared to reveal their possible functions. The average expression of the top 20 markers genes of ve subpopulations were compared in Supplementary Fig. 8E, where the marker genes of subpopulation 2 and 3 were more exclusive.
The rst cluster C0_CD8-CCL5 cells characterised by marker genes Ccl5 50 , Cd3e 51 , Cxcr6 52 and Gzmk 53 , were considered as memory T cells. Most of the top 20 highly expressed genes in the second cluster were various ribosomal proteins, such as Rpl32, Rpl26, Rpl23 and Rpl28. It has been reported that translation is upregulated during effector CD8 + T cell expansion 54 . In addition, Tnfrsf9 55 and Prf1 56 appeared to highly express in this subpopulation. Thus, these were likely effector CD8 + T cells and were named C1_CD8-TNFRSF9. The third cluster, C2_CD8-CDCA5, was characterised by signi cant upregulation of Cdca5, Cdc6 and Ccna2 (Supplementary Data 5), commonly associated with dividing T cells 57 . Additionally, several histones and regulators, including Tmsb10 and Ptma, were among those genes with highest expression in C2_CD8-CDCA5. The fourth cluster possessed more than 1,500 genes with signi cant upregulation ( Supplementary Fig. 8) and was characterised by Fth1, Cd74, and I tm3. The relevance of Fth1 to CD8 + effector T cell response was reported, which revealed that it played an immunomodulatory role in cytokine signalling, adaptive immunity, and cell death 58  We then projected CD8 + T cells onto the two-dimensional state-space de ned by Monocle3 for sample similarity and pseudotime analysis, to obtain the information inferring lineage trajectories from expression data (Fig. 5C). Most cells from each subpopulation aggregated based on expression similarities, and different clusters formed into a relative process in pseudotime that began with C2_CD8-CDCA5 (dividing CD8 + T cells), then developed in separate directions, with one direction developing to C3_CD8-FTH1 cells (effector-memory CD8 + T cells). It appeared that C0_CD8-CCL5 (memory CD8 + T cells), C1_CD8-TNFRSF9 (effector CD8 + T cells) and C4_CD8-MS4A4B (naïve CD8 + T cells) started to emerge at approximately similar pseudotime on the other direction, gradually overlapping on three branches along the pseudotime trajectory, two of which also included certain amount of C2_CD8-CDCA5 cells, indicating functional divergence of this subpopulation. On these two branches, C2_CD8-CDCA5 aggregated with C0_CD8-CCL5 and C1_CD8-TNFRSF9, which suggested close correlation between regulatory, effector and memory CD8 + T cells, and different functions might be executed. C4_CD8-MS4A4B was diversely present together with C0_CD8-CCL5 and C1_CD8-TNFRSF9 along the pseudotime, especially at the middle area (Fig. 5C), implying their close association.
Seven states were thus identi ed based on pseudotime analysis (Fig. 5D), where cells in transitional state 2 and state 5 exclusively corresponded to C0_CD8-CCL5 and C2_CD8-CDCA5, respectively. Most cells of state 1, 3 and 7 were C0_CD8-CCL5, C1_CD8-TNFRSF9 and C2_CD8-CDCA5. A transitional state 4 was identi ed, which consisted of all clusters except C3_CD8-FTH1. The predicted developmental trajectory was also con rmed by the marker genes with similar expression pattern, which hierarchically clustered these markers along the pseudotime in each state ( Supplementary Fig. 8) The genes signi cantly differentiating the branches were also analysed, with expression variation of top 10 genes along the pseudotime trajectory compared amongst different groups ( Fig. 5F to 5H). Most of these genes were expressed around pseudotime zero in the untreated tumours but had a signi cantly prolonged expression with control or caerin peptide treatment. During the rst transition, the genes highly associated with immune system, and their expression, declined at an early stage in states 5 and 6 possibly due to low cell numbers, then increased sharply onwards pseudotime 25, where more cells expressing these genes were observed in caerin treated tumours. There was also a slow increase of expression of Apoe, C1qb, Cd74, H2-Aa, H2-Ab1and H2-Eb1 along the pseudotime on the branch involving state 1, 2, 3, 4, 5 and 7, which also correlated with higher cell number stimulated by caerin (Fig. 5F). Most of genes such as Rps11, 15a, 36, 24 and 26 during the branching displayed in Fig. 5G appeared downregulated along the pseudotime in state 1, 2, 4 and 5, indicating the deactivation of translation, which was also the case in state 1, 2, 4, 5 and 7 for the third trajectory separation (Fig. 5H). In addition, the expression trend of these genes in state 3, 4 and 5 aligned well with the cell distribution in caerin treated tumours.
TMT10plex labelling quantitative proteomics revealed higher immune response induced by the injection of caerin 1.1/1.9 To validate our scRNA-seq data and capture treatment-dependent alterations in protein content for the TC-1 tumour, we performed quantitative proteomic analysis of tumours using the TMT labelling method (details of protein quantitation and annotations are in Table S6). The pairwise comparison showed that signi cantly more proteins were regulated with caerin treatment when compared to control peptide (Supplementary Fig. 9A and 9B). The hierarchical clustering of quanti ed proteins implied consistency between biological triplicates of each group. A total of 238 proteins were uniquely upregulated in with caerin treatment, while the upregulation of 51 proteins were observed with both caerin and control peptide treatments ( Supplementary Fig. 9C).
The  File 2). A correlation was observed between proteins signi cantly upregulated only with the injection of caerin and cell populations identi ed in scRNA-seq (Fig. 6G). Of those proteins showing a fold change greater than 2, many were closely correlated with normalised expressions of genes in the populations of monocytes, MΦ and DCs, such as Iigp1, Gbp2, Irf5 and Parvg. There were a few proteins more closely associated with their gene expression in T cell and NK populations, including Satb1, Spn, Dok2 and Hip1r. Marker gene Gzmc appeared exclusive to NK cells, and protein upregulation was only considered signi cant with injection of caerin. Stat1 was detected as an upregulated gene in nearly all cell populations except B cells and was elevated with caerin treatment in the proteomic analysis, suggesting that Stat1 was largely regulated by caerin.
The KEGG pathways enriched (P-value<0.05) in upregulated proteins were compared for the different treatment in Fig. 6H, and more pathways were signi cantly identi ed with the caerin treatment, including apoptosis, natural killer cell mediated cytotoxicity, necroptosis, the signalling of nod-like receptor (NLR), TNF, chemokine, NF-Kappa B, RIG-I like receptor and toll-like receptor and several disease-related pathways. Amongst these KEGG pathways, NLR signalling was determined as the most enriched pathway (P-value=1.55E-10), supported by the signi cantly increased concentrations of Gbp2, Gbp5, Nlrc4, Ccl2, Tlr4 and so forth by caerin treatment; the genes of many of these proteins were detected as signatures for MΦ/DCs by scRNA-seq. Notably, the antigen processing and presenting KEGG pathway was less signi cantly changed with caerin treatment (P-value=6.0E-4) compared to control peptide treatment (P-value=1.0E-11), in accordance with the observation that the population of B cells was remarkably reduced by the injection of caerin peptides.

Discussion
Here, we demonstrated that intra-tumour injection of caerin 1.1/1.9 peptides further modulated the TME in TC-1 tumour bearing mice immunised with a HPV16 E7 peptide based therapeutic vaccine coupled with PD-1 blockade, when compared with intra-tumour injection of a control peptide. The scRNA-seq of CD45 + tumour in ltrating cells and proteomic analysis of tumour uncovered that caerin 1.1/1.9 changed heterogeneity and function of tumour in ltrating leukocytes, especially macrophage populations (largely elevated the populations of Arg -, MHCII hi and Ear2 hi MΦs), reducing pro-tumorigenic B cells and inducing more active CD8 + T cells, which thus modulated the TME to a pro-in ammatory environment that may favour tumour rejection.
The numbers of tumour in ltrating CD8 + T cells are similar between the treatment groups (Fig.4), but signi cantly more than those of untreated group. Although the antigen speci c CD8 + T cells in the spleen and draining lymph nodes are similar in the two treatments, suggesting that caerin peptides do not in uence the generation of vaccine induced CD8 + T cells (Fig. 1A), the CD8 + T cells in caerin 1.1/1.9 group were more activated than those isolated from control group, with Ifng, Pfr1, Rgs16 highly expressed in the former. We are currently investigating whether the improved TME by caerin1.1/1.9 treatment can be translated to increased survival time in this group, together with therapeutic immunisation and PD-1 blockade.
Most of our knowledge of TAMs comes from histological examinations and in vitro pro ling using ow cytometry 60  and Ebf1 70 . This was also re ected in proteomic analysis that antigen processing and presentation pathway was remarkably less enriched compared to that in control group, and the biological processes associated with MHCI and MHCII was insigni cantly modulated by caerin 1.1/1.9 with respect to untreated or control groups (Fig. 6). Meanwhile, proteasome pathway was largely inhibited, due to the downregulations of Psmb5, Psmb6, Psmb7, Psmd9 and H2afz, in TC-1 tumour treated by caerin 1.1/1.9 as suggested by proteomic analysis (Supplementary Data 6). A recent study found that the inhibition of proteasome caused B cells to become unable to deplete actin at the centrosome and renders them incapable of separating the centrosome from the nucleus, thus cell polarity and organisation was impaired 71 . In addition, proteomic analysis also found that proteoglycans (PGs) in cancer pathway was inhibited by caerin 1.1/1.9-containing treatment (Supplementary Data 6), which might also negatively affect B cell development 72 .
Previously, we found that caerin 1.1/1.9 attracted NK cells to the TC-1 tumour 20 . Though the numbers were relatively low (less than 100) in the current study, NK cells were elevated signi cantly in caerin 1.1/1.9 group (Supplementary File 2). The overlapping distribution of its marker genes with CD4 + CD8 + and CD8 + T cells suggested certain similarity in their cellular function. We found that Cd56 bright NK cells were largely activated only in caerin 1.1/1.9 group. Cd56 bright NK cells were reported to have a regulatory role in early immune response due to the capability of producing different cytokines and shaping adaptive response 73 . Notably, scRNAseq determined Gzmc as the marker gene with the second highest upregulation and the largest mean expression, and there were more NK cells showing higher expression of Gzmc stimulated by caerin 1.1/1.9. This was consistent with the observation uncovered by the proteomic analysis that the content of Gzmc was increased and natural killer cell mediated cytotoxicity pathways activated signi cantly only in the tumour tissue of caerin 1.1/1.9 group, suggesting a higher cytotoxicity might be induced to cause more e cient cell death in TC-1 tumour, which was re ected as the enrichment of apoptosis and intrinsic pathway for apoptosis only in this group (Supplementary Fig.   14F). It has been found that Gzmc can support CTL-mediated killing via the granule exocytosis pathway during late primary alloimmune responses 74 , potentially related to the mutual function shared between NK and two T cell populations. A previous study uncovered the activation of caspase-independent cell death with a mitochondrial phenotype by Gzmc 75 , while our scRNA-seq analysis revealed the pathway, activation and myristolyation of BID and translocation to mitochondria, the second most enriched process in caerin 1.1/1.9 group.
NLR signalling activation was indicated by both scRNA-seq and proteomic analysis. Nlrc4 acts as a downstream transcriptional target of p53 and shows potential anti-tumorigenic functions 76 . The upregulation of Nlrp3 was shown to induce Stat1 phosphorylation through IFNγ, thus promote an antitumorigenic environment 77 . The proteomic analysis identi ed Stat1 only upregulated in caerin 1.1/1.9 group with signi cance, and it appeared as the node with the highest degree of interactions with other upregulated proteins. Stat1 was found to act as an essential mediator of the antitumor response by inhibiting MDSC accumulation and promoting T-cell mediated immune responses in murine head and neck squamous cell carcinoma 78 . NF-kB pathway was signi cantly activated with the presence of caerin 1.1/1.9 (Fig. 6G), which might function synergistically with Stat1 to induce more iNOS and Il12 79 , thus triggering recruitment of NK cells and CTLs. This was consistent with the observation that the NK population and the activation of CD8 + T cells were largely promoted, and the much higher upregulation of Il12b was detected in migDCs by scRNA-seq in caerin 1.1/1.9 group. These NK cells and CD8 + T cells joined the C0_MΦ/DC-CCL5 to elicit apoptosis in TC-1 tumour, which was highly activated in caerin 1.1/1.9 group exclusively suggested by proteomic analysis. In addition, the co-activation of Stat1 and Rela (signi cantly upregulated in neutrophils and migDCs) potentially triggered the expression of iNOS, consequently producing nitric oxide that could contribute to tumour elimination 80  Diego, CA) after extensive washing in wash buffer for 2 hours at room temperature Antigen. speci c IFN-γ secreting cells were detected by sequential exposure of the plate to avidin-horseradish peroxidase (Sigma-Aldrich) and DAB (Sigma-Aldrich). The plate was washed, allowed to air dry overnight, foci of staining were counted by ELISPOT reader system (CTL, Germany).
Isolation of tumour in ltrating CD45+ cells TC-1 tumour from untreated, vaccinated and PD-1 blockaded in conjunction with the injection of caerin 1.1/1.9 (molar ratio 1:1, at 39 µg/mL) or control peptide P3 mice were pooled (3 tumours/group), cut into 2×2mm pieced, digested by adding 2.35 mL of RPMI 1640, 100 µL of Enzyme D, 50 µL of Enzyme R, and 12.5 µL of Enzyme A into a gentleMACS C Tube, followed by disassociation using gentle MACS Dissociator from Miltenyi (Gladbach, Germany). After removal of dead cell and cell debris, the remaining cells were labelled with CD45 microbeads (130-110-618) and passed through LS column. The purity and viability of the CD45+ cells were more than 80% con rmed by ow cytometry and trypan blue staining.

GEM generation and barcoding
The Single Cell 3' Protocol upgrades short read sequencers to deliver a scalable micro uidic platform for

Genome Alignment
Cell Ranger (http://support.10xgenomics.com/single-cell/software/overview/welcome) uses an aligner called STAR (https://github.com/alexdobin/STAR), which performs splicing-aware alignment of reads to the genome. Cell Ranger then uses the transcript annotation GTF to bucket the reads into exonic, intronic, and intergenic, and by whether the reads align (con dently) to the genome. A read is exonic if at least 50% of it intersects an exon, intronic if it is non-exonic and intersects an intron, and intergenic otherwise.
Cell Ranger further aligns exonic reads to annotated transcripts, looking for compatibility. A read that is compatible with the exons of an annotated transcript, and aligned to the same strand, is considered mapped to the transcriptome. If the read is compatible with a single gene annotation, it is considered uniquely (con dently) mapped to the transcriptome. Only reads that are con dently mapped to the transcriptome are used for UMI counting.

Calling Cell Barcodes
Cell Ranger takes as input the expected number of recovered cells, N. Let m be a robust estimate of the maximum total UMI counts, taken as the 99th percentile of the top N barcodes by total UMI counts. All barcodes whose total UMI counts exceed m/10 are called as cells.

Depth normalisation
When combining data from multiple libraries, the read depth between libraries were equalised before merging, to reduce the batch effect introduced by sequencing. Subsample reads from higher-depth libraries until they all have an equal number of total reads per cell.

Expression QC
Seurat was used to explore QC metrics and lter cells with the following ltration.
Based on the number of genes identi ed, total UMI number and the ratio of the mitochondria gene expression in one cell, high quality cells were ltered to be included in the following subtype clustering ( Supplementary Fig. 1). With the quality control lters, 4,648 cells (untreated tumour), 6,523 cells (caerin 1.1/1.9 treatment) and 6,409 cells (the negative control P3 treatment) were included in the analysis (Supplementary Data 1).
Normalising the data After removing unwanted cells from the dataset, the next step is to normalize the data. By default, we employ a global-scaling normalization method "LogNormalise" that normalizes the gene expression measurements for each cell by the total expression, multiplies this by a scale factor (10,000 by default), and log-transforms the result. The formula is shown as follows: A gene expression level = log 10 (1 + (UMI A ÷ UMI Total) × 10000).

Clustering cells
Seurat implements a graph-based clustering approach. Distances between the cells are calculated based on previously identi ed PCs. Seurat approach was heavily inspired by recent manuscripts which applied graph-based clustering approaches to scRNA-seq data -SNN-Cliq 81 and CyTOF data-PhenoGraph 82 . To cluster the cells, we apply modularity optimisation techniques -SLM 83 , to iteratively group cells together, with the goal of optimizing the standard modularity function.

Differentially expressed genes analysis
We used likelihood-ratio test to nd differential expression for a single cluster, compared to all other cells. We identi ed differentially expressed genes as following criteria: 1) P-value ≤ 0.01.
2) Log FC ≥ 0.360674. LogFC means log fold-change of the average expression between the two groups.
3) The percentage of cells where the gene is detected in speci c cluster > 25%.

Marker genes analysis
We further selected the top 20 genes as the marker genes according to the result of differentially expression genes. Then the expression distribution of each marker gene was demonstrated by using heat map and bubble diagram. The putative biological identity of each cluster was assigned by using a murine gene expression atlas 23

Protein-protein interaction (PPI) analysis
Interactions among signi cantly regulated proteins were predicted using STRING 85 . STRING provides a critical assessment and integration of protein-protein interactions from multiple resources, including direct (physical) as well as indirect (functional) associations. A spring model to generate the network images. All resources were selected to generate the network and 'con dence' was used as the meaning of network edges and the required interaction score of 0.700 was selected for all PPI, to highlight the most con dent interactions. Neither the 1 st nor 2 nd shell of the PPI was included in this study. Protein without any interaction with other proteins was excluded from displaying in the network.
Biological process and pathway analysis The enrichment of biological processes, WikiPathways 86 , KEGG pathways 87