(Q1-Q3) sample availability varied – In the treatment group up to 43 samples could be analyzed
To prepare the intended immunological investigation of the Audencel clinical trial, we started with mapping the availability of patients and samples for research. While the concomitant clinical paper by Buchroithner et al. [2] had to follow stringent regulatory criteria (e.g. age) for formal efficacy assessment and could analyse 34 vaccinated patients, for the experimental immunology research described here, it was possible to analyse 43 patients (with available samples) that were vaccinated in the course of the clinical trial.
An overview of all samples processed successfully is given in Additional file 1: Table S1. For the four intended blood-based research methods and the two intended tumor tissue-based research methods, sample availability varied considerably. The highest number of blood-based samples (43 prior to Audencel treatment, 34 during Audencel treatment cycles, 7 prior to control treatment) was reached for flow cytometry and qRT-PCR of immune cell markers. A lower number of blood-based samples was reached for ELISPOT (32 prior to Audencel, 22 during Audencel, 4 prior to control) and CBA (36 prior to Audencel, 26 during Audencel, 4 prior to control).
For tumor-based methods, more control samples were available but at the same time fewer treatment samples: For TCR sequencing we arrived at 23 samples prior to Audencel and 15 prior to control. For IHC it was 11 prior to Audencel and 14 prior to control. Tumor specimens were only seized prior to the respective treatment but not at later time points.
Different sample sources were a cause for the variability of samples measured across immunological methods. Additionally, technical limitations (e.g. amount of material needed for a test) restricted full usage of available samples.
(Q1.A) Pre-vaccination blood: CD8+ cells and ELISPOT response correlated with OS under Audencel
As the first actual research step, we wanted to elucidate a possible impact of pre-existing immune system differences across patients on clinical outcome. Thus, we studied immunovariables before DC immunotherapy (Additional file 1: Table S2) and related them with outcome parameters. In that investigation, we began with measuring blood-based variables that characterize the state of the immune system phenotypically and functionally. ELISPOT and CBA assessed the potency of anti-tumor immune reactions while flow cytometry and qRT-PCR registered populations and polarizations of immune cells in the blood (details see Methods). With the help of these techniques we found several associations: pre-vaccination levels of peripheral blood CD8+ T cells, ELISPOT GranzB production, ELISPOT IFNγ production, blood monocytes, and Th1-related blood transcription factors were associated positively with OS. Pre-vaccination Treg levels in the blood were associated negatively with OS.
These findings are based on the following evidence: The percentage of CD8+ T cells in the blood of Audencel patients significantly correlated with OS (Pearson correlation, p = 0.005, Fig. 2b). Patients with “high” levels of CD8+ cells (above the median) already before immunotherapy lived significantly longer under Audencel than patients with pre-therapy CD8+ levels below the median (Kaplan-Meier analysis, p = 0.018, Fig. 2c).
Similarly, patients with pre-existing immunity to autologous tumor antigens lived longer under Audencel (Fig. 3): GranzB production in tumor antigen-specific ELISPOT assays correlated significantly with OS (Pearson correlation, p = 0.007, Fig. 3b). The patient group with GranzB production above the median also lived significantly longer (Kaplan-Meier analysis, p = 0.006, Fig. 3c). For ELISPOT IFNγ, analogous observations were made for progression-free survival (PFS, Pearson correlation: p = 0.040, Kaplan-Meier analysis: p = 0.003). In terms of OS, for ELISPOT IFNγ a significant correlation was registered (Pearson correlation, p = 0.037) but for the separation of survival curves only a trend without reaching significance could be seen (Kaplan-Meier analysis, p = 0.615). Further, the higher the pre-existing blood monocyte count, the longer was OS under Audencel (Pearson correlation, p = 0.005, Additional file 1: Figure S1A). Again, also survival curves were separated significantly (Kaplan-Meier analysis, p = 0.028, Additional file 1: Figure S1B). Regulatory T cells (Tregs), on the other hand, were inversely correlated with OS: the lower the pre-vaccination levels of Tregs, the longer the survival (Pearson correlation, p = 0.0001). Treg-separated Kaplan-Meier curves showed a trend but no significance (p = 0.528). The relative fraction of Th1-related transcription factors (Tbet+IFNγ) correlated positively with OS (Pearson correlation, p = 0.020), but in Kaplan-Meier analysis only a trend was seen (p = 0.241).
Summing up, IFNγ, Th1-factors and Tregs were associated with OS in Pearson correlation testing – GranzB, monocytes and CD8+ cells also reached significance when used for Kaplan-Meier curves. For further pre-vaccination blood results see Additional file 1: Table S2: E.g. T cells, B cells, NK cells, granulocytes and sub-populations of them were not associated with survival when measured pre-vaccination. For control patients, none of the variables showed a significant association with survival in Kaplan-Meier analyses (data not shown) – the sample size for control patients was, however, considerably low (see above).
(Q1.B) Pre-vaccination tumor: T cells were associated with a non-significant trend towards longer OS under Audencel
As the next step, we extended our investigation of pre-vaccination parameters from the blood to the tumor. Hence, while the previous analyses focused on peripheral blood immune cells, we here studied tumor-resident immune cells. This time, sufficient material was available for both, the treatment (TCR: n = 23; IHC: n = 11) and the control group (TCR: n = 15; IHC: n = 14).
First, we assessed the repertoire of T cell receptors in GBM tissue via TCR sequencing. We observed that GBM tissue showed a more heterogeneous but also narrower TCR repertoire than blood samples (data not shown). Variables measuring TCR diversity in the tumor (Gini index, clonality, clonal evenness, entropy) were not associated with clinical outcome.
Another TCR sequencing-based analysis looked at the impact of general T cell abundance in the tumor. Given that blood data (Fig. 2) had indicated T cell levels might affect survival, we assessed whether this was also reflected in the sequencing data.
Therefore, we used the number of productive TCR reads as a proxy for T cell abundance. When selecting patients with productive reads (normalized to total reads) above the median, Audencel-treated patients showed a trend towards longer OS than control patients with the same feature but without reaching significance (p = 0.061, Additional file 1: Figure S2A). The abundance of (CD8+) T cells in the blood did not correlate significantly with T cell abundance in the tumor (p = 0.898, Additional file 1: Figure S2B).
Furthermore, we used IHC to explore immunological markers in the tumor. We observed that the overall amount of CD8+ cytotoxic T cells in the tumor correlated positively with PFS (Pearson correlation, p < 0.001, not shown). Similarly, the overall level of CD45RO+ memory T cells was associated with PFS (Pearson correlation, p = 0.017, not shown). Also, the relative amount of microvasculature at the tumor margin (as measured via CD31+ endothelial cells) was related to survival (PFS, Pearson correlation, p = 0.046, not shown). However, none of these markers led to a significant separation of survival curves in the Kaplan-Meier analysis (Additional file 1: Table S2). And none of the ICH markers showed an association with OS of Audencel-treated patients. In the control group, neither PFS nor OS were influenced by IHC markers.
(Q2) Vaccination effects on blood: Audencel stimulated Th1-related functional immunovariables in a dose-dependent manner
Subsequently, we aimed at studying if Audencel might have effects on the immune system. For that, we registered blood variable levels after each round of DC vaccination and plotted their respective dynamics. We found that IFNγ in ELISPOT assays correlated significantly with the number of vaccines given (Pearson correlation, p = 0.038, Fig. 4a). The same holds true for Tbet mRNA levels (Pearson correlation, p = 0.006, Fig. 4a) in blood cells (PBMCs). Also, a combined measure of cytotoxic immune responses (mRNA of Th1 transcription factors Tbet and IFNγ) significantly increased upon Audencel administration in a dose-dependent manner (Pearson correlation, p = 0.006, Fig. 4a) – while blood IFNγ mRNA levels alone declined (Pearson correlation, p = 0.003, not shown). Moreover, the ELISPOT production capacity of Interleukin-2 (IL-2, Pearson correlation, p = 0.001, Fig. 5a) was enhanced with every vaccination and equally the Interleukin-17 (IL-17) production capacity (Pearson correlation, p = 0.002, not shown).
When looking at immunovariables with a decrease upon vaccination, we found that the blood Treg polarization declined with every cycle of Audencel treatment (Pearson correlation, p = 0.036, Fig. 4b). Also, Treg cells in the blood declined with every vaccination (Pearson correlation, p = 0.034, Fig. 4b). Similarly, overall CD4+ cells as well as subsets of them decreased (Additional file 1: Table S3).
Taken together, Audencel treatment seems to functionally skew the immune system towards Th1 reactions in a dose-dependent manner, resulting in an additional stimulation with every additional vaccination.
Immunovariables such as ELISPOT GranzB, CD3+ cells, CD8+ cells, B cells, NK cells, monocytes or granulocytes did not react to Audencel vaccination (Additional file 1: Table S3).
(Q3) Post-vaccination blood: ELISPOT IFNγ and CD8+ cells were associated with clinical outcome
Given that Audencel apparently altered the immune system, we next wanted to study if changes upon Audencel application were directly associated with survival. However, for none of the variables identified in (Q2), the relative strength of the variable response upon vaccination correlated with PFS or OS (not shown).
Subsequently, we looked at the absolute immunovariable levels post vaccination because we assumed that the overall effect of vaccination might have an influence on outcome – independent of the relative change. To take the different number of vaccinations given across patients and potential time kinetics into account, we used all available post-vaccination time points (from all available patients; n=patients).
As a result, we noticed that several variables measured after vaccination were indeed connected with clinical outcome: Post-vaccination ELISPOT IFNγ production significantly correlated with OS (Pearson correlation, p = 0.022, Fig. 5a) and could significantly separate survival curves (Kaplan-Meier analysis, p = 0.003, Fig. 5a). Similarly, post-vaccination CD8+ cell abundance in the blood correlated with OS (Pearson correlation, p = 0.026, Fig. 5b) and separated survival curves (Kaplan-Meier analysis, p < 0.001, Fig. 5b). Monocyte levels post vaccination showed the same association with OS (Pearson correlation: p < 0.001, Kaplan-Meier analysis: p = 0.008, not shown). Interestingly, this was also true for activated NK cells (Pearson correlation: p = 0.042, Kaplan-Meier analysis: p = 0.024, not shown). In an additional analysis that looked at the data from yet another angle (assuming the post-vaccination average of all time points as the relevant overall post-vaccination level), we registered that CD8+B7H1+ cells and CD4+B7H1+ cells correlated significantly with OS (Pearson correlation, p<0.001 for both) but could not separate survival curves (Kaplan-Meier analysis, CD8+B7H1+ p=0.219, CD4+B7H1+ p=0.085). All further post-vaccination results see Additional file 1: Table S4.
(Q1 + Q2 + Q3) Integration: Patients with “high” immune-capabilities showed better outcome under Audencel
Finally, we studied whether patients with generally “high” anti-tumor immune-capabilities were more likely to benefit from DC vaccination. Our assumption was that the single variables we had found previously could be condensed to one overall illustrative measure. Thus, we integrated prior insights into one parameter via a scoring system. To allow potential future usage as a clinical biomarker, we exclusively used pre-vaccination variables from the blood for that score. Up to 9 points were awarded for individual immunovariables and added up: “high” anti-tumor immune-capabilities were arbitrarily defined as 5–9 points, “low” immune-capabilities arbitrarily as 1–4 points. 1 point each was awarded for high levels (above the median) of Th1 indicators, IFNγ, GranzB, CD8+ cells and monocytes; 0 points were given for high levels of Tregs or low levels of all the other variables mentioned – again relative to the median (Additional file 1: Figure S3). Consequently, 12 of the treatment patients were classified as having “high” and 31 as having “low” capabilities.
As a result, we could observe that patients with “high” anti-tumor immune-capabilities had a significantly better outcome in terms of PFS (Kaplan-Meier analysis, p < 0.001, Fig. 6a) as well as OS (Kaplan-Meier analysis, p = 0.014, Fig. 6b). In the control group, only data from 7 patients were available for this analysis: no association with survival was registered for “high” immune-capabilities (p = 0.695, Additional file 1: Figure S4).
Also, we checked whether patients with “high” immune-capabilities before vaccination were the ones that showed altered levels of immunovariables after vaccination. This was not the case. We found no significant difference in all relevant variables after vaccination between patients with “high” or “low” immune system-capabilities before vaccination (Additional file 1: Figure S5). Patients with “high” variable levels after vaccination were not necessarily the same patients that had “high” immune system-capabilities before vaccination.