Epigenetic reprogramming of Runx3 reinforces CD8 + T-cell function and improves the clinical response to immunotherapy

Multiomics data analysis from a clinical cohort showed that Runx3 is the key mediator of improved clinical response with low-dose DAC-primed anti-PD-1 immunotherapy

We recruited two groups of patients treated with anti-PD-1 vs. anti-PD-1/DAC (DP). DNA methylation EPIC and RNA-seq were performed as shown in the workflow (Fig. 1a). To explore the DNA methylation reprogramming profile of CD8+ T cells in patients, we obtained the DNA methylation profile of CD8 + T cells from anti-PD-1-treated patients versus patients treated with primed DAC (Fig. 1a). Compared to anti-PD-1-treated patients, large-scale demethylation was detected in DAC-primed anti-PD-1-treated patients, and clinical response was evaluated in correlation with treatment (Fig. 1a-c). The number of hypomethylation sites reached 113972, and these genome-wide demethylation changes were mainly distributed on N-shelf and S-shelf (Fig. S1e).

Fig. 1figure 1

DAC can trigger large-scale apparent reprogramming of CD8+ T cells. a Workflow of the experimental design including clinical sample collection and sequencing. b Schematic chart showing that DNA methylation reprogramming is correlated with clinical response and relapse. c Correlation analysis of tumor size and the anti-PD-1/DAC treatment cycles of patients. The patients were treated as indicated. Tumor size increase>50% were considered to indicate progression. d Analysis of genome-wide methylation variations in CD8 + T cells between the two indicated groups. The methylation was screened according to a |Diff beta value >0.1 and P< 0.05. Blue represents hypomethylation sites, and red represents hypermethylation sites. e Violin diagram showing the genome-wide methylation distribution of each patient. Red represents the monotherapy group, and blue represents the combined therapy group. The left panel is the baseline period of C1D0, and the right panel is the end of C2D0 treatment. f Statistical analysis of genome-wide DNA methylation levels. Upper panel: mean value of DNA methylation in the C1D0 and C2D0 periods; lower panel: median value of DNA methylation in the C1D0 and C2D0 periods (two-tailed unpaired t tests. n.s.: not significant, ***P < 0.001, ****P < 0.0001) g Distribution analysis of DNA methylation levels after anti-PD-1 or anti-PD-1/DAC. The DNA methylation value was divided into 20 sections from 0-1. A value < 0.15 was taken as the low methylation level, and a value > 0.85 was taken as the high methylation level. Upper panel: treated with anti-PD-1. Lower panel: treated with anti-PD1/DAC (left panel: before treatment; right panel: after treatment). h IPA pathway enrichment analysis. The input data are DMSs in each period. The size of the circle shows the number of enriched genes in each pathway, and the color depth represents the degree of enrichment

Furthermore, we identified 295 hypomethylated DMSs, while the number of hypermethylated DMSs was 951. Moreover, we found a significantly increased number of hypomethylation sites in the DP combined treatment group (Fig. 1d, Fig. S1b), which suggests that large-scale demethylation of CD8+ T cells occurred after DP treatment.

To further explore the dynamic profile of DNA methylation reprogramming and its correlation with clinical outcome, we compared DMSs at the C1D6 (5 days after DAC treatment) stage vs. the C1D0 (baseline before treatment) stage and the C2D0 (after one cycle of anti-PD-1 treatment) stage vs. C1D6 (5 days after DAC treatment) stage. The results showed that large-scale demethylation occurred after DAC treatment (Fig. S1c). In contrast, dynamic DNA methylation reprogramming was not observed in the PD-1 blockade monotherapy group (Fig. S1d). Furthermore, to confirm DNA methylation reprogramming in individual patients, we quantitatively analyzed the genome-wide methylation levels and differential methylation levels of each patient (Fig. 1e). The results showed that all CD8+ T cells of all patients in the DP group had significant demethylation profiles during C2D0 (after one cycle of anti-PD-1 treatment), which implied that the demethylation caused by DAC showed no individual differences. Furthermore, the methylation level was analyzed with statistical analysis, and significant difference was observed upon treatment with anti-PD-1 vs. anti-PD-1/DAC (Fig. 1f). Analysis of the distribution of DNA methylation showed that significant DNA demethylation occurred in high methylation sites when anti-PD-1/DAC treatment was conducted (Fig. 1g).

We then performed Ingenuity Pathway Analysis on DMSs, which showed that DMSs were enriched mainly in T-cell exhaustion- and T-cell activation-related pathways, including the T-cell exhaustion signaling pathway, Th1 and Th2 activation pathways, and in the role of NFAT in the regulation of the immune response (Fig. 1h). Furthermore, we also analyzed the C2D0 (after one cycle of anti-PD-1 treatment) vs. C1D0 (baseline before treatment) status in CD8+ T cells in the DP group. The DMSs were also mainly enriched in the T-cell exhaustion signaling pathway, CTLA4 signaling in cytotoxic T lymphocytes, interferon signaling, and interferon signaling (Fig. 1h). The above results showed that DAC treatment may result in the reversal of T-cell exhaustion and augment T-cell activation and infiltration.

To investigate whether DNA methylation reprogramming correlates with gene expression, we performed RNA-Seq and analyzed the expression profiles of CD8 + T cells. The results showed that there were a large number of differentially expressed genes (DEGs) in the two groups of CD8+ T cells. As shown in the volcano map, we found that the expression fold changes of DEGs increased significantly in the C2D0 (after one cycle of anti-PD-1 treatment) period (Fig. 2a). By analyzing the differences in the DP group and comparing C2D0 (after one cycle of anti-PD-1 treatment) versus C1D0 (baseline before treatment), we found downregulated genes such as CCR2, TNFSF14, and TNFSF4 and upregulated genes such as CCL3, TIGIT IFNG and CD69 and cytokines such as CCR3 and CCR5 were upregulated significantly under DAC treatment, and TOX was downregulated under DAC treatment. These data indicate that exhausted T cells can be reversed by DAC and that infiltration-related cytokines are upregulated by DAC but not anti-PD-1. We further performed IPA pathway enrichment analysis (Fig. 2b) and GSEA (Figure S2b) to explore the possible biological significance of these findings. The results showed that these differences were enriched in the T-cell exhaustion pathway, Th1 and Th2 activation pathway and T-cell receptor pathway, which was highly consistent with the enrichment results of DMSs in DNA methylation profiling. Through tSNE analysis, we found that the CD8+ T cells consisted of two distinct groups after anti-PD-1 treatment (Fig. 2c).

Fig. 2figure 2

Expression profile and integrated multiomics analysis in CD8 + T cells identified important signaling pathways in response to DAC treatment. a The expression fold changes of DEGs increased at different stages. Left panel: Expression of differentially expressed genes in the C1D0 period in anti-PD-1-vs. anti-PD-1/DAC-treated patients. Right panel: Expression of differentially expressed genes in the C2D0 period in anti-PD1-vs. anti-PD-1/DAC-treated patients. Red indicates up-regulated genes, and blue indicates down-regulated genes. b IPA pathway enrichment analysis. The input data are the DEGs in each period. The size of the circle shows the number of enriched genes in each pathway, and the depth of the color represents the P-value of enrichment. c Workflow of the experimental design and tSNE analysis of DMSs and DEGs in the C1D0 and C2D0 periods. Blue represents the combined therapy group with DAC and anti-PD-1, and red represents the monotherapy group with anti-PD-1. Upper panel: tSNE analysis of DMSs. Lower panel: tSNE analysis of DEGs. d Intersecting gene analysis of DMSs and DEGs. Orange represent DMSs, and blue represents DEGs. e IGV showed Runx3 methylation levels in different patients at different stages. f Correlation of gene expression and the promoter methylation level of Runx3. Upper panel: Violin diagram showing the statistical analysis of the difference in methylation levels in the Runx3 promoter region. The figure shows the median, upper quartile and lower quartile. Two-tailed unpaired t tests. Lower panel: The expression levels of Runx3 in different periods were analyzed by a line diagram. The x-axis represents the period, and the y-axis represents the FPKM value

Through multiomics joint analysis of DEGs and DMSs, we found 3729 genes that intersected in terms of DNA methylation and expression profile (Fig. 2d). Through pathway analysis, we found that the pathways were mainly enriched in the regulation of T-cell activation, cytokine signaling in the immune system, immune system development, and regulation of the leukocyte cell‒cell adhesion pathway, which suggests that the combination of DP may play a central role in the development, activation and response to cytokines of the immune system.

Due to the consistency between DMSs and DEGs, we conducted a joint analysis of the two omics datasets. We found that Runx3 was the most significantly regulated gene. After treatment with DAC, all CR patients maintained stable demethylation levels on the Runx3 promoter and high expression levels of Runx3 (Fig. 2e). The correlation of DNA methylation and expression was confirmed (Fig. 2f). DNA demethylation was also observed in UPN21; however, the state of demethylation could not be well maintained as in this patient, and the expression level of Runx3 was downregulated back to baseline levels. Then, we found that this patient had disease relapse after 6 months, which highlights the importance of DNA methylation status for clinical outcome. The above case of recurrence showed a clear correlation between DNA methylation reprogramming and clinical outcomes, and epigenetic reprogramming of the Runx3 promoter plays a key role in regulating Runx3 expression during DAC-primed anti-PD-1 treatment.

In vivo work in mice demonstrated that DAC treatment promoted T-cell infiltration and downregulated T-cell exhaustion

To further investigate the underlying mechanism of the “epigenetic sensitization” role of DAC immunotherapy, we aimed to reproduce the clinical observation in mice and establish an in vivo mouse model. C57BL/6 mice were implanted with MC38 cells and treated with either DAC, anti-PD-1 or anti-PD-1/DAC at the indicated times, simulating the clinical situation of patients (Fig. 3a).

Fig. 3figure 3

DAC downregulated T-cell exhaustion and upregulated T-cell infiltration by demethylating Runx3 and promoting Runx3 expression. a Workflow of the experimental design and analysis of the tumor growth curve using the MC38 mouse model treated with DAC, anti-PD-1 or DAC/anti-PD-1(n= ). b Tumor growth curve of mice treated with DAC, anti-PD-1 or anti-PD-1/DAC. Upperpanel: average tumor growth curves;(two-tailed unpaired t tests, *P <0.05, **P<0.01, ***P<0.001) c The proportions of GranB+, perforin+, TNF-α+, IFN-γ+, Ki67+ and CD8+ T cells were analyzed by flow cytometry. Samples were taken from the blood, spleen, tumor, or lymphocytes of MC38 mice treated with DAC, anti-PD1 or anti-PD-1/DAC as indicated. (n=5, two-tailed unpaired t tests, *P <0.05,**P <0.01***P < 0.001). d Leftpanel: The proportion of Runx3+CD8+ T cells in each group was analyzed by flow cytometry (n=5, two-tailed unpaired t tests, **P<0.01). Right panel: DNA methylation level change on Runx3 promoter in T cells treated with DAC, anti-PD-1 or antiPD-1/DAC. Y axis: Mehylation level of Runx3 (%); X axis: sampes of mice treated with DAC, anti-PD-1 or DAC/anti-PD-1. Triplicate samples were applied for each experiment and the median was shown as horizontal line within the box plots. (p<0.05)

As shown in Fig. 3b and c in MC38 tumor-transplanted mice, we found that DAC combined with anti-PD-1 significantly inhibited tumor growth, promoting the infiltration of TILs. In the MC38 model, anti-PD-1 treatment showed insignificant inhibition of tumor growth, but the combination of DAC and anti-PD-1 significantly inhibited the growth of tumors. We further analyzed the function of CD8+ T cells with flow cytometry. As shown in Fig. 3c, we found that the proportion of proliferative T cells increased significantly (ki67+CD8+ T). The proportions of killing (GranB+CD8+T, proferin+CD8+ T) and secretory cells (IFN-γ+CD8+ T) cells were also significantly improved.

Our results also showed that the DNA methylation status of Runx3 decreased significantly in DAC- and DAC/anti-PD-1-treated mouse CD8+ T cells but not in WT and anti-PD-1-treated mouse CD8+ T cells (Fig. 3d).

To explore the immune status of the peripheral immune system and tumor microenvironment, we examined the number and function of T cells in lymph nodes, spleens, peripheral blood and tumor. The results showed that the numbers of CD3+ T cells in the blood, spleen, and lymph gland were decreased in the combination therapy group. The numbers of CD8+ T cells were decreased in the spleen and blood and increased in the lymph gland in combination therapy. This indicated that the proliferation, killing and secretion of interferon by tumor-infiltrating T cells were all significantly enhanced, indicating the overall recovery of T-cell function in the tumor microenvironment (Figure 3c). Flow cytometry showed that the proportion of Runx3+CD8+ T cells increased more when cells were treated with DAC than when cells were with anti-PD-1 and peaked when cells were treated with both DAC and anti-PD-1 (Figure 3D).

Conditional knockout of Runx3 proved that Runx3 is indispensable for DAC to play the role of “epigenetic sensitizer” for anti-PD-1 resistance

To rule out the antitumor role of Runx3 in cancer cells, we constructed conditional knockout mice to prove the specific function of Runx3 in T cells and immunotherapy.

Runx3 flox mice were generated by a CRISPR/Cas9-based approach. Briefly, two sgRNAs were designed with the CRISPR design tool (http://www.sanger.ac.uk/) to target either the upstream or downstream region of the transcript NM_019732.2 exon 4 of mouse Runx3. Through subsequent hybridization and genotype identification with Lck-cre mice, Runx3fl/fl;Lck-Cre mice were finally constructed. The Runx3fl/fl mice showed no difference in weight or development. No autoimmune disease was observed. Then, both Runx3fl/fl and Runx3fl/fl;Lck-Cre mice were divided into three groups: the control, anti-PD-1-treated, and anti-PD-1/DAC-treated group (Fig. 4a).

Fig. 4figure 4

The epigenetic sensitization effect of immunotherapy was eliminated in Runx3fl/fl;Lck-Cre mice. a Workflow of the construction of Runx3 conditional knockout mice and analysis of the tumor growth curve in Runx3fl/fl and Runx3fl/fl;Lck-Cre mice treated with anti-PD-1(n=5, two-tailed unpaired t tests, **P< 0.01). Upper panel: average tumor growth curves; lower panel: individual tumor growth curves. b Analysis of the tumor growth curve in Runx3fl/fl and Runx3fl/fl;Lck-Cre mice treated with anti-PD-1or DAC+ anti-PD-1(n=5,two-tailed unpaired t tests, **P<0.01). Upper panel: average tumor growth curves; lower panel: individual tumor growth curves. c tSNE analysis of immune cell subsets in the CD8+Tils of Runx3fl/fl and Runx3fl/fl;Lck-Cre mice treated with anti-PD-1. Upper panel: tSNE data showing the overall distribution of each subgroup. Lower panel: Histogram showing the absolute numbers of cells of various subtypes (cell number/104 CD45+ cells). d tSNE analysis of immune cell subsets in the CD8+ Tils of Runx3fl/fl and Runx3fl/fl;Lck-Cre mice treated with DAC+anti- PD-1. Upper panel: tSNE data showing the overall distribution of each subgroup. Lower panel: Histogram showing the absolute numbers of cells of various subtypes (cell number/104 CD45+ cells)

The antitumor immunity was decreased in Runx3fl/fl;Lck-Cre mice (Fig. 4a, b). There were significant differences between the anti-PD-1 group and DAC/anti-PD-1 group in control Runx3fl/fl mice (Fig. 4c, d). In addition, there was no significant difference in tumor growth curve between mice in the DAC-primed anti-PD-1 and anti-PD-1 group in Runx3fl/fl;Lck-Cre mice. No significant differences in tumor volume and mass were observed, which indicated that the effect of DAC was eliminated after conditional knockout of the Runx3 gene.

To further investigate the role of Runx3 in CD8 + T cells, we applied single-cell flow mass spectrometry analysis (CyTOF) and found that the proportion of CD8+ tumor infiltrating lymphocytes (TILs), changed significantly. Comprehensive analysis suggested that Runx3 increased the proportion of Teff and TRM cells and interfered with the balance of immune cell subtypes (Fig. 4c, d).

Runx3 plays a critical role in T-cell infiltration and effector and memory T-cell differentiation and functions to attenuate T-cell exhaustion

Tumor immunotherapy consists of multiple steps of the T-cell functional response. First, T cells differentiate into effector T cells and then memory T cells to exert antitumor functions. Second, it is essential for T cells to infiltrate into tumors to kill tumor cells. A lack of immune cells in the tumor microenvironment is an important reason for the low response. Third, T cells are often in a state of exhaustion or dysfunction [17,18,19,20,21], and the exhaustion of T cells affects the PD-1 antibody response rate as well [28]. To elucidate the specific role of Runx3, we performed mass cytometry (CyTOF) to compare the T-cell function of conditional knockout mice and control mice. We employed 42 markers to cover cytokines, T-cell exhaustion markers, T-cell proliferation and T-cell killing ability. Since we observed a significant increase in CCR after DAC treatment and we found that DAC promoted T-cell infiltration in our mouse model, several CCRs were included among these 42 markers.

First, we observed significant downregulation of CD8+ T cells and effector T cells in peripheral blood, spleen, tumor tissue of mice with Runx3 deficiency (Fig. 5a, b). Furthermore, we found that Runx3 deficiency significantly downregulated CCR3 and CCR5, consequently impairing T-cell infiltration (Fig. 5c). Deficiency of Runx3 also impaired T-cell function by affecting T-cell differentiation and T-cell exhaustion (Fig. 5d). We observed increased levels of Lag3, Tim3 and CTAL4 but decreased levels of IFNγ, TNFα and IL-2. Interestingly, we did not find increased levels of PD-1 in Runx3-deficient mice. Previous work has shown that PD-1 is expressed only after T cells are activated, while Runx3 deficiency significantly downregulates effector T cells and memory T cells, which in turn might balance the expression level of PD-1 to increase exhausted T cells.

Fig. 5figure 5

Runx3 deletion hampers CCRs expression and tumor infiltration of CD8+ T cells. a Determination of immune cell subsets in the peripheral blood and spleen of Runx3fl/fl and Runx3fl/fl;Lck-Cre (Runx3CKO)mice by tSNE analysis after mass cytometry. Left panel: tSNE data showing the overall distribution of each subgroup. Right panel: The histogram shows the absolute numbers of cells of various subtypes (cell number/104 CD45+ cells). b CD8+ T cells distribution in the tumor tissue of Runx3fl/fl and Runx3CKO mice by tSNE analysis. c tSNE plots showing Runx3 and CCRs expression of T cells after anti-PD-1/DAC treatment. The plots represented CD45+ immune cells in mice tumors, and the circle indicated CD8+ T cell population. d tSNE plots showing expression of marker genes of T cells after anti-PD-1/DAC treatment. The plots represented CD45+ immune cells in mice tumors, and the circle indicated CD8+ T cell population. e Schematic illustration showing that demethylation of Runx3 by DAC promoted CCRs expression and T-cell infiltration

In control Runx3fl/fl mice, we found that DAC combined with PD-1 antibody promoted the infiltration of T cells and the secretory ability of TILs (IFN, TNF-α). The proliferation ability (Ki67) and killing ability were significantly increased (GranB, Perforin, etc.), and this promoting effect was significantly improved after using DAC. For Runx3-knockout mice, the secretion, proliferation and killing ability of T cells were inhibited, suggesting that Runx3 may be the key mediator of the epigenetic sensitization function of DAC (Fig. 5c, e).

Runx3 predicts the anti-PD-1 immunotherapy responses in a spectrum of tumor types

We deemed that it would be interesting to investigate whether Runx3 levels can predict the immune response to anti-PD-1. Immunotherapy has transformed the treatment landscape for a variety of tumors and has demonstrated durable response rates in some refractory tumors, yet unresponsiveness and severe immune-related side effects have been reported in some treated patients. Therefore, biomarkers are urgently needed to screen people who can benefit from immunotherapy.

From a previous clinical study, our results showed that there was a strong correlation between the Runx3 expression and the clinical response (Fig. 6a). The receiver operating characteristic (ROC) curve, which is a useful graphical tool for assessing the predictive performance of a biomarker (Fig. 6a), indicated that a biomarker panel distinguished two groups: the responsive and non-responsive groups. ROC curves are universally used standards to evaluate biomarkers. We drew a ROC curve of Runx3 with the clinical response and found that the Runx3 level could crucially predict the clinical response. We show in Fig. 6b that Runx3 levels were associated with effector T-cell levels and memory T levels, not overall CD8+ T cell levels. Therefore, Runx3 contributes to the clinical ICB response by influencing the functional differentiation but not the overall level of CD8+ T cells. Noteworthy, when we treated patients with DAC, although the Runx3 level increased, the overall level of CD8+ T cells did not change, which indicates that this effect is not caused by changes in the number of T cells (Figure S5). The above data showed that number of functional T cells, but not overall level of T cells plays important role for the clinical response.

Fig. 6figure 6

Runx3 is a key molecular marker of the clinical response to immunotherapy. a Violin diagram showing the expression levels of Runx3, CD28, CD226, FasL and STAT4 in T cells in responders and nonresponders. The figure shows the median, upper quartile and lower quartile. Two-tailed unpaired t tests. b Correlation analysis between effector T cells, memory T cells and Runx3. The x-axis represents Runx3 expression in T cells, and the y-axis represents the abundance of memory T cells or effector T cells. c. Kaplan‒Meier survival curves between high and low expression of Runx3 in T cells and prognoses in different cancer types

In addition to our clinical cohort, we explored the TISIDB database and analyzed the correlations of Runx3 levels and clinical prognoses of anti-PD-1 therapy. We found that Runx3 levels correlated well with clinical prognosis and survival rate. High expression of Runx3 was related to a favorable prognosis of anti-PD-1 regimen in patients with colorectal cancer, breast cancer and lymphatic cancer. This suggests that Runx3 is an important regulatory factor in anti-PD-1 immunotherapy and a potential biomarker for prognosis prediction. Taken together, these data show that Runx3 is not only a key mediator of DAC and CD8+ T cell function but also a potential biomarker for the clinical immune response.

留言 (0)

沒有登入
gif