In vitro characterization of antigen-sources
In this study, two Mlh1-/- cell lines established from spontaneous GIT were used. The drug response of the cell lines 328 and A7450 T1 M1 was determined before and revealed no significant differences towards standard cytostatic drugs [19]. By assessing the basal secretion profile from supernatants, we indeed observed substantial variations. Focusing on cytokines associated with immune stimulation, A7450 cells generally secreted higher levels of GM-CSF, IL1b, and IL-18 (Figure 2A, left panel). While all of these cytokines enhance NK cell activity and foster Th1 cell development, concentration of chemokines responsible for monocyte- and eosinophil-attraction, such as monocyte chemoattractant protein-1(MCP1), MCP3, and Eotaxin was higher in supernatants from 328 cells (Figure 2A, right panel).
Based on these findings, a co-culture system of tumor-Ags-loaded DCs and lymphocytes was initiated. In this setting, additional differences were seen in the T cell phenotype (Figure 2B and C). DC-loaded with A7450 T1 M1 tumor lysate boosted the frequency of CD3+CD8+ T cells, which were activated and additionally positive for IFNg. By contrast, the phenotypes of leukocytes from 328 lysate-loaded DCs changed faintly compared to the control (Figure 2B).
Mutational profile of antigen-sources
The selected genes of the oncoprint are known for the relevance for tumor initiation, progression, apoptosis, and suppressors functions (Figure 3A). Mlh1-/- tumors harbor mutations in Pik3ca, Msh3, Braf, and/or Kras, and Erbb3 [21]. The A7450 T1 M1 cell line harbors nonsense and missense single nucleotide variants (SNVs) in the Wnt signaling pathway regulator Apc gene. Further hotspots in pre-selected clinical relevant genes are occurring in tumor suppressors Arid1a as well as Fhit.
In a direct comparison, alterations are exclusively distributed. The cell line 328 acquired more missense SNVs in the pre-selected gene set, especially in EGFR signaling members as well as in Nf1. The 328 cell line had additional missense and nonsense Pole mutations. Taking the germline MMR-D into account, the increased number of gene mutations in affected tumor cells is conserved in the cell line 328 compared to A7450 T1 M1 (Figure 3A).
In Arid1a, all of the 34 SNVs appear before or after the ARID/BRIGHT DNA binding domain (Figure 3B), regulating cell proliferation, differentiation, and development [27], as well as the SWI/SNF-like complex subunit BAF250/Osa. Every single SNV is exclusive for the corresponding cell line, none are shared, and all of them are missense mutations.
The prevalence and hotspot mutations in Tcerg1 and Wwox exclusively detected in the 328 cell line are shown in Figure 3B. The mutational hotspot in Tcerg1 is Q1040H within the FF6 domain, the only amino acid change in this gene. FF domains play an essential role in binding the phosphorylated C-terminus of the RNA polymerase II. Furthermore, Tcerg1 is involved in regulating the transcriptional elongation and the pre-mRNA splicing [28]. In Wwox, we found three SNVs, which all affect the short-chain of the dehydrogenase/reductase domain.
For the mutational profile as potential Ag-source, the MSI pathway [29,30] and genes associated with MSI status [31] have an impact on survival (based on hazard ratio in the human counterpart) (Table 1). Except for Cope, the theme of exclusive and distinct SNVs within the 328 and A7450 T1 M1 cell lines continues. However, its influence on survival remains elusive and has no impact, since the alterations are yet unknown or silent. Overall, cell line 328 shows a more substantial amount of affected genes associated with overall survival and disease-free survival. These are Chmp5, Dhx32, Gadd45b, and Inadl.
Then, the coding microsatellite (cMS) mutational profile was analyzed comparatively on a panel of putative MSI target genes (Table 2 and [18]). Overall, A7450 T1 M1 cells harbored mutations in half of the markers. The numbers of cMS mutations in 328 cells was lower (37 %) and the genes affected different, highlighting the individual profile even in these molecular closely matched Mlh1-/- cells that harbor the very same germline mutation. Shared mutations were found in seven candidate genes, such as Taf1b, Rfc3, Akt3, and Spen. While these genes are all classified as tumor suppressors, they may have a high likelihood of being causative for this type of tumor. By deciphering the differences between these two samples in more detail, we identified some exclusive mutations in Mlh1-/- A7450 T1 M1 cells whose resulting neo-Ags may have immunogenic potential. The most promising candidates, in this case are Senp6 and Rasal2. Consequently, we analyzed the frequency of spontaneous immune reactivity against the neoepitopes derived from a -1 frameshift mutation in the cMS of these genes. However, in this test, no significant reactivity was detectable (data not shown), making these candidates unlikely to act as tumor rejection Ags.
Prolonged survival in the prophylactic setting
To test the immunogenicity of whole cancer vaccines on a more global level, Mlh1-/- mice received two independent tumor lysates, either harboring high (= 328, 167 mutations/Mb) or moderate (= A7450 T1 M1, 27 mutations/Mb) TMB [21](Figure 1).
Prophylactic vaccination yielded significantly prolonged cancer-free survival in Mlh1-/- A7450 T1 M1-treated mice. Median survival time was 37 weeks, whereas it was only 22 weeks in control mice (p<0.001). The Mlh1-/- 328 vaccine had a minor impact on survival, reaching a median survival of 25 weeks (Figure 4A). The tumor spectrum observed in this study largely covers the distribution seen in Mlh1-/- mice. Two thirds of Mlh1-/- A7450 T1 M1-treated mice developed GIT or generalized lymphomas in the spleen; remaining mice developed lymphomas in the thymus (1 case), skin malignancies (1 case) or died spontaneously (2 cases). Mice receiving the Mlh1-/- 328 tumor lysate showed a comparable tumor spectrum. Here, 70 % suffered from GIT or generalized lymphomas in the spleen, one mouse developed a thymic lymphoma, and two mice died because of unknown malignancy (suspected lymphomagenesis).
The survival benefit of mice vaccinated with the Mlh1-/- A7450 T1 M1 lysate was reflected by immunological changes in the peripheral blood. While T cell numbers only gradually increased, we observed elevated levels of circulating NK cells (Figure 4B).
Then, the reactivity of peripheral blood leukocytes was assessed upon co-incubation with different target cells by IFNg-ELISpot assay (Figure 5A). Autologous Mlh1-/- tumor targets triggered IFNγ secretion of lymphocytes from vaccinated mice. The highest reactivity was seen between days 56 and 84 and mainly against target cells that were used for vaccination. We even observed differences between the two vaccines; A7450 T1 M1 cells evoked IFNg secretion more effectively from lymphocytes than 328 cells (p<0.01). In line with the increased number of NK cells upon A7450 T1 M1 vaccination, leukocytes from vaccinated mice reacted against NK target cells YAC-1 (p<0.01).
Tumor microenvironment
Next, the tumor microenvironment was studied in detail to explore the quantity and quality of leukocyte infiltrates. Prophylactic vaccination leveraged the microenvironment (Figure 5B). The Mlh1-/- A7450 T1 M1 vaccine largely prevented infiltration of CD11b+/Gr1+ myeloid-derived suppressor cells and F4/80+ tumor-associated macrophages (TAMs). While these cell types were barely detectable, we observed high numbers of infiltrating CD11c+ DCs as well as CD8+ cytotoxic T cells (CTL). By contrast, the Mlh1-/- 328 lysate triggered MDSC infiltration in the tumor, CTL were occasionally found. PD1 expression was not altered by any vaccination and, thus, expression levels were highly comparable with control tumors.
Therapeutic vaccination
Then, we moved to the therapeutic approach (Figure 1). The survival benefit of mice treated with a lysate from the Mlh1-/- A7450 cells compared to the 328 lysate was shown before [21] and (Figure 6A). Here, the median overall survival was 11 weeks. By contrast, the 328 lysate failed to improve outcome, which was slightly longer than in untreated control mice (5 vs. 4 weeks). To see whether treatment can be improved by adding low-dose chemotherapy, the vaccination protocol was extended by gemcitabine given one day before treatment initiation, followed by two monthly injections (Figure 1 and Figure 6A). With this combined chemo-vaccine, survival was prolonged in mice treated with cell-line derived tumor lysates 328 (9 weeks; p<0.05 vs. control). With regard to the A7450 T1 M1 + chemo group, there was a trend towards longer progression-free survival, yet this did not reach statistical significance (hazard ratio: 0.9).
Accompanying PET/CT imaging largely reflected the survival data (Figure 6B). 328-vaccinated tumors progressed, with no gross changes compared to untreated controls. Gemcitabine in conjunction with the lysate yielded stable disease. The same was true for tumors treated with the A7450 T1 M1 lysate, showing virtually no progression during the first weeks of treatment. Here again, combined chemo-immunotherapy improved tumor growth control. Individual tumors even tended to shrink (Figure 6B). Still, the antitumoral stimulus provided by the combined chemo-vaccine was not strong enough to induce long-term regression, and tumors finally progressed.
Immunological changes upon vaccination
Therapeutic vaccination altered splenic immune cell composition. Spleens from 328-vaccinated mice tended to have reduced amounts of CTL (Figure 6C). Levels of CD11b+Gr1+ MDSC as well as CD69+ activated T cells remained similar to controls. Gemcitabine had no impact on immune cell distribution at all. Spleens from mice receiving the chemo-vaccine combinations had similar phenotypes as those treated with the 328 vaccine alone.
In contrast, the immune phenotype of spleens from A7450 T1 M1-vaccinated mice positively changed with significantly lower numbers of MDSCs but higher levels of activated CD69+ T cells (Figure 6C). This effect was even independent of gemcitabine and thus related to the vaccine itself. Accompanying functional ELISpot analysis confirmed these findings with high reactivity against autologous target cells A7450 T1 M1 (Figure 6D). Leukocytes from mice treated with the chemo-vaccine combination tended to have higher reactivity against NK cell targets YAC-1 compared to those getting the A7450 T1 M1 monotherapy. In line with the results from the prophylactic setting, there was no cross-reactivity against other Mlh1-/- tumor targets, i.e. 328 and 1351.
We finally examined whether alterations were evident on tumor resection specimens in situ. Generally, A7450 T1 M1 vaccinated tumors were more infiltrated than 328-treated tumors (Figure 7). By delineating mice that had no response from those achieving stable disease in PET/CT, we indeed found clear differences in the tumor microenvironment. Tumors of short-term survivors (328) were highly infiltrated with TAMs and had higher numbers of LAG-3- and PD-L1-expressing lymphocytes (Figure 7). Granulocytes were rarely detectable. Resection specimens from long-term survivors harbored few TAMs, virtually no MDSCs or LAG-3+ lymphocytes. Hence, these data nicely reflect the in vivo response.