Olaparib increases DNA damage and reduces viability in Brca-deficient cell lines

The negative impact of PARP inhibition on the survival of ovarian cancer cells has been previously established [7, 8]. In this study, ID8 cells with CRISPR-mediated loss of Trp53 alone or in combination with Brca1 or Brca2 deletion were used [21, 22]. To determine the extent of DNA damage and subsequent cell death in these models, we treated these cell lines with olaparib for 24 h and measured viability and γH2AX expression levels (Fig. 1A, B). In both Brca-deficient models, treatment with olaparib resulted in a significant increase in γH2AX staining, which was undetectable in the Brca-proficient control cells (Fig. 1A, left panel). Viability of the ID8 Trp53−/− cells also did not change in response to any dose of olaparib (Fig. 1B). In contrast, cells with double knockout of Trp53−/− and Brca1−/− modestly responded to olaparib toxicity, with the highest dose of olaparib (10 µM) causing a significant 25% reduction in viability. Despite an equivalent amount of DNA damage, the ID8 Trp53−/− Brca2−/− cells were the more olaparib-sensitive cell line, with viability reduced significantly at every dose of olaparib tested. Since olaparib differently affected Brca-null cell viability, the expression of genes associated with the PARP pathway in the ID8 cell lines was compared by querying RNA-seq data previously published in our lab [23]. We discovered that the ID8 Trp53−/− Brca2−/− cells growing in vitro demonstrated a higher expression of both PARP1 and PARP2 compared to the Brca1−/− model (Fig. 1C). Furthermore, expression levels of the DNA repair gene XRCC1 were highest in the Brca2-null model (Fig. 1C). Interestingly, nearly all PARP pathway-associated genes were highest in the Brca-proficient ID8 Trp53−/− model. However, expression of XRCC1, the enzyme responsible for the actual DNA repair was low in relation to all other genes in the Trp53−/− cells (Fig. 1C). Thus, in the ID8 model, mutations in the Brca genes are associated with reduced cell viability and increased susceptibility to DNA damage in response to olaparib treatment.

Fig. 1

In vitro olaparib treatment induces DNA damage, reduces cancer cell viability and upregulates PD-L1. A Immunofluorescence staining images (left) and quantification (right) of γH2AX following a 24-h treatment of ID8 cells with 10 µM olaparib, representative of the extent of DNA damage. Student t test; *p < 0.05. B Histograms show the percent viability of ID8 cancer cells following treatment with 0, 1.0, 5.0 and 10 µM olaparib for 24 h, as assessed by AlamarBlue assays. Viability at each dose was compared to the 0 µM group. Analysis was done using a one-way ANOVA with Tukey’s multiple comparisons. Mean ± SEM (n = 3); *p < 0.05, **p < 0.01. C Heatmap presenting the expression levels of proteins involved in the PARP DNA repair pathway in the ID8 cell lines. RNA-seq data [23] collected from cell lines growing in vitro were analyzed to develop a normal distribution model of the expression of genes of interest. Each box represents the mean value of log-normalized RNA-seq data for 3 biological replicates. D PD-L1 expression in ID8 cells treated with 10 µM olaparib for 24 h, assessed using flow cytometry. Mean ± SEM (n = 3), one-way ANOVA; ***p < 0.0001, ****p < 0.01

Olaparib upregulates the expression of PD-L1

Previously, it was noted that in ovarian cancer cells, the PARP inhibitor niraparib can upregulate the expression of PD-L1, a critical regulator of immune activity [24]. To establish the influence of olaparib treatment on PD-L1 expression in the various ID8 models specifically, we measured its expression after a 24-h treatment. Regardless of the cell genotype, olaparib significantly increased the percentage of cells which expressed PD-L1 on their surface within the treated group, albeit to a proportion less than in the positive control, IFN-γ (Fig. 1D). The results of this experiment are analogous to the data found in the literature and identify PD-L1 as a target for combination therapy in vivo [24].

Olaparib, anti-PD-L1 and their combination differently influence the survival of tumour-bearing mice

To assess the impact of treatments on the survival of tumour-bearing mice, two animal studies were performed. In the first study, mice were injected with one of the three ovarian cancer cell lines: ID8 Trp53−/−, ID8 Trp53−/− Brca1−/− or ID8 Trp53−/− Brca2−/−. Treatment with olaparib did not significantly prolong the survival of the Trp53−/− model, and the median length of survival of the control (48 days) and olaparib groups (52 days) were similar (Fig. 2A). In contrast, treatment significantly improved survival of the mice in the Trp53−/−Brca1−/− and Trp53−/−Brca2−/− models by approximately 33% and 30%, respectively (Fig. 2B, C). The absence of any response in the Trp53−/− model resulted in the exclusion of this group from the next in vivo study.

Fig. 2

Syngeneic models of ovarian cancer differentially respond to treatment based on their genotype. 5 × 106 ID8 Trp53-/- cells with or without Brca1 or Brca2 deficiency were injected IP to make syngeneic tumour models that were treated with olaparib, anti-PD-L1 or both. Olaparib was given by 18 daily IP injections at a dose of 50 mg/kg/day; the anti-PD-L1 was administered as five daily IP injections of 200 µg anti-PD-L1 followed by one 100 µg injection every four days for a total of 11 doses. The combination group received both drugs. Mice were euthanized at humane endpoint. A-C Kaplan–Meier survival plots showing the response of all three ID8 models to olaparib monotherapy relative to vehicle control. D-E Kaplan–Meier survival plots showing the response of Brca1- and Brca2-deficient ID8 models to olaparib, anti-PD-L1 and a combination of both drugs. Log-rank (Mantel-Cox) test. ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001

As we had detected enhanced expression of PD-L1 in response to in vitro olaparib therapy, in the second animal study, mice harbouring ID8 Trp53−/− Brca1−/− or ID8 Trp53−/− Brca2−/− tumours were treated with either olaparib or anti-PD-L1 monotherapies, or a combination of both drugs, using the treatment regimen summarized in Supplementary Figure S 1. In the Trp53−/− Brca1−/− model, the anti-PD-L1 monotherapy nearly doubled the median survival of tumour-bearing mice (74 days) compared to the isotype control (38 days), making it the most beneficial overall treatment in terms of survival (Fig. 2D). The survival of mice given olaparib monotherapy (50 days) and combination therapy (52 days) were very similar and were both significantly longer than the isotype control (Fig. 2D). In the Trp53−/− Brca2−/− model, the anti-PD-L1 monotherapy did not improve survival (Fig. 2E). However, the olaparib monotherapy did significantly improve the survival of mice harbouring Brca2-deficient tumours (68 vs. 52 days; Fig. 2E). Similar to the Brca1-deficient group, the outcomes from the combination therapy were not different from the survival after olaparib monotherapy (Fig. 2E). Overall, the olaparib and combination therapy similarly improved the survival of both models, but no synergy was observed. However, the anti-PD-L1 monotherapy demonstrated efficacy specifically in prolonging the survival of the Trp53-/- Brca1-/- mice, while its effects on other treatment groups were not significant.

Olaparib, anti-PD-L1 and their combination transform the immune composition of the Brca1- and Brca2-deficient tumour microenvironments

The effects of PARPi treatment on the composition of the ovarian TME have not been studied extensively. As such, we set out to characterize the immune composition of various innate and adaptive immune cell populations within the PARPi-treated animals 36 h after the end of therapy. We collected ascites fluid, or performed a peritoneal wash, as well as spleen and analyzed the immune cell populations by flow cytometry (Supplementary Fig. S2). We also characterized anti-PD-L1 as a monotherapy and in combination with olaparib. This addition provided us with a more sophisticated and complete image of the TME composition in response to novel therapeutics. Analysis of the immune composition of the spleen from tumour-bearing mice was also performed to determine the influence of these therapeutics on the systemic immune system (Supplementary Fig. S 3-S 4).

Heatmaps summarize the characterization of the peritoneal microenvironment of the ID8 Trp53−/− Brca1−/− (Fig. 3A) and Trp53−/− Brca2−/− (Fig. 3B) models. All three treatments uniquely transformed the Trp53−/− Brca1−/− TME (Fig. 3A). The administration of monotherapies and their combination was associated with higher percentages of multiple different cell types as seen in the positive z-scores in the heatmap rows which represent treatment groups (Fig. 3A). The cell frequency changes which are significantly different from isotype control and/or the combination therapy are shown in Fig. 3C-L and included some notable effects. The combination therapy resulted in an 8% increase in overall T cell frequency compared to the isotype control (Fig. 3C). The olaparib monotherapy increased the total frequency of CD8 T cells and activated (CD44 expressing) CD4 T cells by approximately 5%, thereby doubling this population (Fig. 3E, F). Exceptionally, the combination therapy significantly reduced the frequency of activated, CD44 + CD4 + T cells to near zero levels compared to the monotherapies (Fig. 3F). Similarly, the combination therapy resulted in a > 90% reduction in activated, CD44 + CD8 + T cells compared to all other groups, reducing the percentage of these cells to near zero levels (Fig. 3G). Olaparib alone also reduced the CD4/CD8 T cell ratio compared to all groups by approximately 50% (Fig. 3H).

Fig. 3

Brca deficiencies cause differential changes in the tumour microenvironment in response to therapy. A-B Heatmaps presenting the relative abundance of immune cell types in the (A) ID8 Trp53−/− Brca1−/− and (B) Trp53−/− Brca2−/− tumour microenvironments. Peritoneal washes (n = 5 mice per group) were collected from tumour-bearing mice that had been treated with olaparib, anti-PD-L1 or their combination, and analyzed by flow cytometry 36 h after the last treatment. For each cell type, the value was calculated as the percent of all leukocytes. The percentages were normalized, and each box represents the mean of the 5 biological replicates. C-V. Histograms showing the immune populations that are significantly different from isotype control. The significant changes in cell populations are divided based on genotype: (C-L) for ID8 Trp53−/− Brca1−/− and (M-V) for the Trp53−/− Brca2.−/− tumour-bearing mice. Each dot represents one biological replicate. Mean values with SEM are shown. ISO = isotype control group and OLA = olaparib treated group. One-way ANOVA with Tukey’s multiple comparison test; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

The therapies also uniquely influenced the composition of NK cell populations. The olaparib monotherapy reduced the overall NK cell frequency by ~ 50% (Fig. 3I) while the combination therapy significantly enhanced CD25 + NK populations (Fig. 3J). The monotherapies both increased the frequency of CD44 + NK cells, while the combination therapy reduced this population to nearly zero (Fig. 3K). Lastly, the anti-PD-L1 monotherapy doubled the population of LAG3 positive NK cells (Fig. 3L).

In contrast to the Brca1-null TME, the treatment of Brca2-null tumour-bearing mice resulted in fewer significant changes in the immune cell populations found in the TME, as shown in the heatmap in Fig. 3B. The TME derived from Brca2-null group seem to exhibit a higher prevalence of negative z-scores, especially in response to olaparib alone (Fig. 3B). While the vast majority of immune cell populations appear to be lowest in the TME of the olaparib group, some cell types are present at higher proportions, namely those expressing LAG3, PD-1 and TIGIT, all of which are markers of exhaustion and/or inhibition (Fig. 3B). Statistical analysis of the various cell types revealed a small number of significant differences between the isotype control and treatment groups (Fig. 3N-V). The percentage of LAG3 + NK cells was reduced to about half by the combination treatment, relative to the isotype control (Fig. 3V). The anti-PD-L1 monotherapy reduced the CD4 + T cell population by ~ 50% (Fig. 3N). Overall, responses to therapy varies substantially between these models. The immune cell composition of the Brca2-null TME revealed much fewer modifications than those seen in the Brca1-null model, suggesting the absence or poor efficacy of the anti-PD-L1 monotherapy in the Brca2-null model may be due to a lack of antitumoral immune stimulation in the TME.

Surprisingly, in both the Brca1- and Brca2-deficient tumour models used in this study, the in vivo treatment with anti-PD-L1, whether as a monotherapy or in combination with olaparib, resulted in a systemic increase in nearly all the studied immune cells presenting surface PD-L1. Both the expression levels of this marker and the percentage of cells expressing it were amplified significantly (Fig. 4A-U). This systemic increase was one of the few changes that were consistently changed in all tissues analyzed in this study. Taken together, in vivo treatment with anti-PD-L1 results in upregulation of PD-L1 in the ovarian tumour microenvironment.

Fig. 4

In vivo treatment with anti-PD-L1 results in a compensatory upregulation of this marker in both Brca-deficient tumour models. Peritoneal washes (n = 5 mice per group) were collected from tumour-bearing mice that had been treated with olaparib, anti-PD-L1 or their combination, and analyzed by flow cytometry 36 h after the last treatment. Only the immune populations which are significantly different from isotype control or the combination therapy are shown. The immune populations were divided based on genotype: (A-J) for ID8 Trp53−/− Brca1−/− and (K-U) for Trp53−/− Brca2.−/− tumours. Each dot represents one biological replicate. Mean values with SEM are shown. ISO = isotype control group and OLA = olaparib treated group. One-way ANOVA with Tukey’s multiple comparison test; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

In vivo treatments differentially influence the cytokine composition of the ascites fluids in the Brca-deficient models

The knowledge of the effects of PARPi and monoclonal antibody therapies on cytokine production in the ovarian TME is minimal. As such, we set out to determine the effects of these treatments on the concentrations of 13 cytokines and chemokines. The cytokine arrays conducted on ascites fluid collected at humane endpoint revealed several changes, as shown in the heatmap in Fig. 5A. Of those, changes that were significant in either the Brca1 or Brca2-null models are individually presented in Fig. 5B-G.

Fig. 5

Cytokine/chemokine concentrations in ascites reflect BRCA-dependent differences in response to treatment. Ascites fluids were collected at humane endpoint (n = 4 per group). Cytokines were quantified using the LEGENDplex Mouse Cytokine Release Syndrome flow-based assay. A Heatmap showing the normalized concentrations of cytokines in the ID8 Trp53−/− Brca1−/− and ID8 Trp53−/− Brca2−/− models. Each box represents the mean of 4 replicates. B-G Histograms indicating the concentrations that are significantly different from isotype control or the combination in at least one model. The histograms are divided based on tumour genotype: (B-D) shows data from the ID8 Trp53−/− Brca1−/− ascites and (E–G) the Trp53−/− Brca2.−/− ascites fluid. Each dot represents the ascites supernatant from one biological replicate. Mean values with SD are shown. Analysis was done using a one-way ANOVA with Tukey’s multiple comparison; *p < 0.05, **p < 0.01, ***p < 0.001

As presented in the heatmap, the anti-PD-L1 monotherapy resulted in a general reduction in cytokine concentrations in the ID8 Trp53−/− Brca1−/− model compared to isotype control (Fig. 5A). Olaparib monotherapy, on the other hand, resulted in cytokine concentrations that were often higher than the isotype control (Fig. 5A). Changes in cytokine production in the ID8 Trp53−/− Brca2−/− model in response to anti-PD-L1 were akin to those in the Brca1−/− model, but with more substantial reductions (Fig. 5A). Unlike the Trp53−/− Brca1−/− model, olaparib had a general suppressive effect on the levels of several cytokines in the Brca2−/− model (Fig. 5A). The combination therapy had similar effects, leading to both increases and decreases in cytokine production that followed the same trends as the olaparib monotherapy (Fig. 5A).

It is notable that some changes in cytokine abundance were consistent across both models, while others were only present in one model (Fig. 5B-G). Compared to all other cytokines in the Brca1-null TME, IFN-a production was significantly higher with olaparib monotherapy (Fig. 5B). Remarkably, treatment with anti-PD-L1 on its own or in combination with olaparib drastically reduced IFN-a concentrations to near undetectable levels (Fig. 5B). These differences were also seen in the Brca2-null model (Fig. 5E). Expression of granulocyte–macrophage colony-stimulating factor (GM-CSF) closely resembled that of IFN-a, as anti-PD-L1 treatment completely diminished GM-CSF production in both models (Fig. 5C, F), however these changes were only significant in the Brca1-null group (Fig. 5C).

A synergistic effect of both treatments resulted in the highest VEGF concentration after combination therapy of the Brca2-null model (Fig. 5G), as the concentrations of VEGF were more than 4 times greater than the isotype control. Conversely, there were no changes VEGF concentrations of the Brca1-null TMEs (Fig. 5D). In summary, the impact of therapy on cytokine concentrations within the ovarian TME was found to vary between treatments, with anti-PD-L1 exhibiting a unique suppressive effect. In contrast, cytokine concentrations in the olaparib-treated group were generally similar to or higher than the isotype control, albeit not always significantly different. Furthermore, akin to the variations in the cellular composition of the TME, the alterations in the cytokine profiles were found to differ greatly between the Brca1- and Brca2-null models.

Depletion of Brca1 and Brca2 in ovarian cancer cells differentially influences the expression of genes involved in the TME composition

As the TME analysis from the two models revealed differences in various cell populations and cytokines, we analyzed previously published bulk RNA sequencing data [23] from both Brca-deficient IP tumour models to further investigate the differences between them. The analysis revealed notable differences in the transcript levels of several cell type markers in Brca-null TMEs. Expression of CD45, the pan-leukocyte marker, was higher in Trp53−/− Brca1−/− tumours compared to both the Trp53−/− Brca2−/− and the Brca-proficient Trp53−/− models (Fig. 6A). Similarly, expression levels for various immune cell markers such as CD3, CD4, and CD8 were highest in the Brca1-null tumours. Expression of genes associated with immune cell function also varied between the three models. For instance, markers of both leukocyte activation (CD44) and exhaustion (LAG3) were also highest in the Brca1-null tumours (Fig. 6A). Aside from cell type and functional markers, the expression of genes encoding cytokines differed between the models as well. For pro-inflammatory cytokines such as TNF-a and interleukin-1 alpha (IL-1a), expression levels were greatest in the ID8 Trp53−/− Brca1−/− tumours, while the anti-inflammatory cytokine IL-10 was expressed at the lowest level. However, it must be noted that expression levels of IL-11, a potent anti-inflammatory cytokine, were also higher in the Brca1-null model (Fig. 6A).

Fig. 6

Deficiencies in Brca1 and Brca2 differently influence the expression of immune-related factors in the ovarian tumour microenvironment. A Heatmap showing the relative expression of immune-related genes in intraperitoneal tumours. Each box represents the mean value of log-normalized bulk RNA-seq data from 4–6 biological replicates. Gene names were replaced by the corresponding protein name. B Pathway analyses for three gene sets that demonstrate the activities of the NF-kB, angiogenic and PD-L1 pathways. The activity of the pathway was approximated by measuring the expression of 12–53 genes which are known to be downstream for the activity of each protein of interest. C Immunohistochemical staining images (left) and quantification of CD31+ cells (brown) in Brca-deficient tumours collected at the humane endpoint (n = 4) from the mice in the isotype control group. Quantifications are based on the percentage area of CD31+ cells in relation to all nucleated cells. D Histograms presenting the concentrations of VEGF in the ascites fluid collected from mice harbouring Brca1- and Brca2-deficient tumours. Cytokines were quantified using the LEGENDplex Mouse Cytokine Release Syndrome flow-based assay. E Histograms presenting the percentage of PD-L1+ cells in untreated ID8 Trp53−/− Brca1−/− and ID8 Trp53−/− Brca2.−/− cell lines. Expression levels were quantified using flow cytometry. Mean ± SEM are shown for all histograms. Analysis was done using the student t test; *p < 0.05, ***p < 0.001

Given the large number of differences detected in the immune compositions of the Brca-deficient TMEs, we investigated further the expression of a key inflammatory mediator, nuclear factor kappa-light-chain-enhancer of activated B cells (NF-kB) [25]. This was achieved by measuring the expression levels of a set of 53 genes that are known to be downstream of this transcription factor. NF-kB, in addition to being a primary driver of inflammation, is a key downstream constituent of one of the major driving forces in TME composition: cyclic GMP-AMP synthase-stimulator of interferon genes (cGAS-STING) signaling [26]. Therefore, analysis of NF-kB signaling could be used as an indicator of a potential change in cGAS-STING activity. Examination of this data revealed that NF-kB activity was highest in the Trp53−/− Brca1−/− model, modest in the Trp53−/− Brca2−/− tumours and lowest in the Trp53−/− group (Fig. 6B). Similar pathway analyses were done for 12 genes associated with angiogenesis, such as Vegf, Vegfr1, Cd31, and Hif1a. Genes associated with angiogenesis were generally higher in the Trp53−/− Brca1−/− tumours (Fig. 6B). These results were confirmed at the protein level using immunohistochemistry and cytokine arrays. Immunohistochemical staining for CD31 revealed that Trp53−/− Brca1−/− tumours have greater levels of angiogenesis compared Trp53−/− Brca2−/− tumours (Fig. 6C). Likewise, results from the ascites cytokine array revealed that concentrations of VEGF are significantly higher in the ascites of the Trp53−/− Brca1−/− model compared to the Trp53−/− Brca2−/− derived samples (Fig. 6D).

Lastly, a pathway analysis of 13 genes downstream of PD-L1 signaling was performed to measure PD-L1 activity in these models. The expression levels of all the genes analyzed for this pathway were once again consistently higher in the Trp53−/− Brca1−/− tumour samples (Fig. 6B). Results of flow cytometry staining support these findings, as the baseline levels of PD-L1 positive cells were found to be significantly higher in the Brca1-deficient cancer cells in vitro when compared to Trp53−/− Brca2−/− cells (Fig. 6E). Taken together, the RNA sequencing data with validations at the protein level by flow cytometry and immunohistochemistry provide strong evidence that, compared to the Trp53−/− Brca2−/− model, the ID8 Trp53−/− Brca1−/− tumours have a more “inflamed” TME, with increased composition of leukocytes, as well as higher expression of inflammatory cytokines and angiogenesis associated genes.