1 Introduction
Precision therapy that aims to treat cancer on the basis of specific gene mutations is available only for a small percentage of cancer cases [
1,
2]. The tumor suppressor p53 is mutated in approximately half of all cancer cases [
3–
5], indicating that p53 mutation-based therapies have the potential to transform the landscape of precision therapy. By 2019, at least 45 p53 mutation-based treatment approaches have been investigated in laboratories, with the small molecules APR-246 (Eprenetapopt) and COTI-2 entering clinical trials [
6]. We recently reported that the approved drug arsenic trioxide (ATO) efficiently restored the tumor-suppressive function to a subtype of p53 mutants [
7–
9], spawning a series of p53-targeting ATO trials registered at ClinicalTrials.gov. However, no p53-targeting drugs or regimens have been approved.
Targeting the signaling dependencies acquired by cancer as a result of p53 mutations, rather than the mutant p53 itself, is an alternative approach for exploiting p53 mutations [
10]. Although the WEE1 inhibitor AZD1775 was proposed to benefit patients with p53-mutated ovarian cancer preferentially [
11], this proposal still lacks clinical evidence. To our knowledge, the treatment of myelodysplastic syndrome (MDS) with decitabine is the only regimen exhibiting statistically significant preference in benefitting patients harboring p53 mutations. Our retrospective study on 196 MDS cases suggested that the p53 mutation is a major adverse prognostic factor for overall survival (OS) in MDS [
12]. Subsequent studies on 109 MDS cases surprisingly discovered that decitabine produced a complete response (CR) rate of 66.7% in the p53-mutated subgroup that was significantly higher than that (21.3%) in the p53-wild-type counterparts [
13,
14]. A mechanistic study revealed that decitabine activated type I interferon signaling to inhibit p53-deficient myeloid malignant cells [
15]. The efficacy of decitabine in treating p53-mutated MDS was reproduced in an independent prospective clinical study on a cohort of 116 patients with MDS and acute myeloid leukemia (AML) [
16]. Yet the series of findings were restricted in myeloid malignancies, whereas more than 99% p53 mutation cases occur in solid tumors.
In the past decade, intense efforts have been made to explore the efficacy of decitabine as a repurposed treatment for patients with solid tumors; however, these studies did not classify patients by p53 status [
17–
23]. Decitabine exhibited elusive efficacies in these studies and has not yet been approved for the treatment of solid tumors. Notably, the positive outcomes of decitabine treatment have been repeatedly reported in ovarian cancer [
19–
21], a special cancer type with a p53 mutation frequency of approximately 95% [
24]. These findings tempt a speculation that similar to those with myeloid malignancies, p53-mutated patients with solid tumors may preferentially benefit from decitabine [
13,
14,
16]. Similar to ovarian cancer, triple negative breast cancer (TNBC) is a cancer type featuring common p53 mutations [
21,
25]. Encouragingly, a preclinical animal study has suggested that decitabine is effective in treating TNBC PDX models [
26]. However, to our knowledge, only clinical trials on decitabine for the treatment of breast cancer, but not TNBC, have been reported. In 2 phase I studies on breast cancer, decitabine, in combination with standard treatment agents, resulted in manageable side effects but exhibited limited efficacy [
17,
27]. The unsatisfactory efficacy of decitabine may be attributed to the lack of p53 mutations in the trialed patients with breast cancer.
On the basis of the above rationale, we performed a retrospective tissue microarray (TMA) study on a TNBC cohort, a prospective DEciTabinE and Carboplatin in TNBC (DETECT, NCT03295552) clinical trial, and a matched mechanistic study on isogenic TNBC cell lines to explore the potential of repurposing decitabine for the treatment of p53-mutated TNBC.
2 Materials and methods
2.1 Patients and TMA
For the retrospective cohort, 132 patients who underwent surgery from January 2009 to December 2013 and were pathologically diagnosed with TNBC at the Comprehensive Breast Health Center of Ruijin Hospital were enrolled. The formalin-fixed paraffin-embedded (FFPE) tissue samples of each patient were obtained. Slides from the original paraffin blocks were stained with hematoxylin and eosin (H&E) and reviewed for representative tumor areas. Two 0.6 mm cores were retrieved from the donor block and transferred to the recipient block. Four-micrometer sections were cut for immunohistochemical analysis.
2.2 Immunohistochemistry and evaluation
TMA slides were stained with antibodies for the N-terminal of p53 (Abcam, DO1, mouse monoclonal, 1:2000 dilution), the C-terminal of p53 (Abcam, E47, rabbit monoclonal, 1:2000 dilution), DNMT1 (Abcam, ab19905, rabbit polyclonal, 1:200 dilution), DNMT3A (Abcam, ab2850, rabbit polyclonal, 1:200 dilution), DNMT3B (Abclonal, A2899, rabbit polyclonal, 1:200 dilution), and 5-methylcytosine (5mC, Abcam, ab10805, mouse monoclonal, 1:200 dilution). Tissue samples with adequate immunoreactivity were used as the positive controls for each antibody. Negative controls were produced through the omission of primary antibodies.
For the immunohistochemistry (IHC) evaluation of p53, the percentage of stained cells was counted and divided into 3 classifications as follows: absent (0% of tumor cells stained), scattered (≥1% and < 40% of tumor cells stained), and diffused nuclear staining (≥40% of tumor cells stained). Scattered staining was determined as the wild-type pattern, whereas absent and diffused staining was defined as the mutant patterns. The scores for DNMT1, DNMT3A, DNMT3B, and 5mC were determined by combining the proportion of positively stained tumor cells and intensity of staining. The proportion of positively stained tumor cells was graded as follows: 0 (no positively stained tumor cells), 1 (1%–24% positively stained tumor cells), 2 (25%–49% positively stained tumor cells), 3 (50%–74% positively stained tumor cells), and 4 (> 75% positively stained tumor cells). The intensity of staining was recorded on a scale of 0 (no staining), 1 (weak staining), 2 (moderate staining), and 3 (strong staining, brown). A staining index (SI) was calculated as proportion × intensity, and the median SI was analyzed for each stained protein. Slides with SIs more than the median score were defined as having high expression levels, whereas those with SIs less than the median score were classified as having low expression levels.
2.3 Eligibility of participants for the DETECT trial
Female patients with de novo stage IV or recurrent TNBC were enrolled in the prospective DETECT clinical trial. ER, PR, and HER2 statuses were centrally reviewed. Eligible patients should receive no more than 1-line treatment and no platinum analog or demethylation agents for metastatic disease. For patients who received platinum in the adjuvant setting, the disease-free interval (DFI) after treatment should be more than 1 year. Additional eligibility criteria included measurable target lesions in accordance with the Response Evaluation Criteria in Solid Tumors (RECIST) 1.1, Eastern Cooperative Oncology Group (ECOG) performance status ≤ 1, normal organ and bone marrow function, and no active brain metastasis. All patients provided written informed consent, and the study was approved by the institutional review board of Ruijin Hospital and has been registered on Clinicaltrial.gov as NCT03295552.
2.4 Treatment plan of the DETECT trial
The open-label, one-armed phase I/II clinical trial was conducted at the Comprehensive Breast Health Center of Ruijin Hospital. After baseline evaluation, eligible patients received consecutive low doses of decitabine (Decitabine for Injection, Chia Tai Tianqing Pharmaceutical Group Co., Ltd., China) of 7 mg/m2 from days 1 to 5 followed by carboplatin at a dose of area under curve (AUC) 6 at day 6 every 3 weeks for 6 cycles. Tumor responses were evaluated by computed tomography (CT) or magnetic resonance imaging (MRI) in accordance with RECIST 1.1 for every 2 cycles. The patients were treated until objective disease progression, intolerable toxicity, withdrawal of consent, or the end of 6 cycles. Treatment was allowed to be delayed for up to 3 weeks beyond the planned resumption of the next cycle. Sequential therapy was determined in accordance with the decision of physicians.
The primary endpoints were objective response rate (ORR) in accordance with RECIST 1.1. Secondary endpoints included clinical benefit rate (CBR), adverse events (AEs), progression-free survival (PFS), and OS.
AEs were evaluated in accordance with the Common Terminology Criteria for Adverse Events (CTCAE 4.0) per week after regimen application. When AEs of Grade 3 or 4 first occurred, reducing the dose of carboplatin from AUC 6 to AUC 5 was suggested. When secondary severe AEs (SAEs) occurred, the administration time of decitabine was modified from 5 days to 3 days. In total, 2 dose reductions were permitted.
2.5 TP53 sequencing of FFPE tissue
Genomic DNA (gDNA) extraction was performed on 3 10-mm thick unstained scrolls of the FFPE specimen of the patients’ primary tumors by using a GeneRead DNA FFPE kit (Qiagen, 180134, Germany). A total of 100 ng of gDNA was amplified by primers of each exon of p53 via polymerase chain reaction (PCR) with 2× Taq Master Mix (Vazyme, P112-01, China) in a total reaction volume of 30 μL. Next, PCR products were purified by using a Wizard® SV Gel and PCR Clean-Up System (A9282, Promega, America) and then subjected to Sanger sequencing. The primers used are listed in Supplemental Data File 3.
2.6 Cell culture
HEK293T and human TNBC cell lines, including HCC1937 and BT549, were all bought from the Type Culture Collection of the Chinese Academy of Sciences (Chinese Academy of Sciences, Shanghai, China). The cell lines were placed in a humidified incubator at 37 °C with 5% CO2; cultured in RPMI 1640 (Thermo Fisher, 22400105, America) or DMEM (Thermo Fisher, 11995073, America) as suggested by the repository; and supplemented with 10% FBS, 100 U/mL penicillin, and 100 mg/mL streptomycin. All cell lines tested mycoplasma-negative.
2.7 Retrovirus packaging and infection
The 6 p53 mutants detected from patients’ samples (missense: H193R, G266V, R282W, and E286K; nonsense: Y107* and W146*) as well as p53-L25Q/W26S (p5325,26) and p53-L25Q/W26S/F53Q/F54S (p5325,26,53,54) were individually constructed on a retroviral vector (MigR1) with an IRES-driven EGFP reporter by using Fast Site-Directed Mutagenesis Kit (TIANGEN Biotech, KM101, Beijing, China). For retrovirus packaging, 2 × 106 HEK293T cells were seeded in a 10 cm dish for 16 h and then cotransfected with 2 helper plasmids (6 μg of VSV-G and 9 μg of gag-pol) and 16 μg of the mutant p53 plasmid. Transfection was performed by using Hilymax transfection reagents (Dojindo, H357, Japan) in accordance with the manufacturer’s instructions, and the medium was replaced after 4 h. Retrovirus-containing medium was collected at 48 h post-transfection. A total of 2 × 105 HCC1937 cells were infected with the retrovirus by being spun in a 6-well plate at 2000 rpm for 2 h in the presence of 8 μg/mL polybrene (Hanbio, HB-PB-500, Shanghai, China). After 6 h, an equal volume of fresh RPMI1640 medium was added to the original virus-containing medium. A 72 h postinfection, the infected cells (GFP-positive cells) were purified through flow cytometry.
2.8 RNA interference
The 4 siRNA oligoes for DNMT1 were purchased from DharmaconTM. A total of 6 × 105 HCC1937 cells were seeded on a 6 cm dish overnight. Subsequently, the mixture of the 4 siRNA oligoes was transfected into the HCC1937 cell line by using DharmaconTM Basic siRNA Resuspension (GE HealthCare, USA) with a final concentration of 25 nmol/L. The transfection medium was replaced with the complete culture medium after 24 h. The designed experiment was then conducted.
2.9 Immunoblotting and luciferase reporter assay
Immunoblotting and luciferase reporter assay were performed as reported previously [
9].
2.10 Quantitative real-time PCR
The total RNA of TNBC cell lines was isolated by using Spin Column Animal Total RNA Purification Kit (Sangon Biotech, B518651, Shanghai, China), whereas the total RNA of FFPE tissue samples was extracted with a RNeasy FFPE kit (Qiagen, 73504, German). A total of 1 μg of total RNA was reversely transcribed into cDNA with HiScript III RT SuperMix for qPCR (Vazyme, R323-01, Nanjing, China) in a total volume of 20 μL. Real-time PCR was performed sequentially by using ChamQ SYBR qPCR Master Mix (Vazyme, R311-02, Nanjing, China) on ViiA™ 7 Real-Time PCR System (Applied Biosystems, America) under the following conditions: 10 min at 95 °C, followed by 40 cycles of 95 °C for 15 s, and 60 °C for 45 s. The primers used are listed in Supplemental Data File 4. The expression of each gene was measured in triplicate and normalized relative to that of reference genes.
2.11 Cell proliferation assay
A total of 1500–2000 cells were seeded in 96-well plates and treated with different combinations of drugs. Decitabine was administered every 24 h for 5 consecutive days, whereas platinum was delivered 1 day after decitabine. Cell viability was determined with Cell Counting Kit-8 (Meilunbio, MA0218, Dalian, China) 48 h after treatment. Briefly, 10 μL of CCK-8 solution was added to 100 μL of medium in each well and incubated at 37 °C for 2 h. Absorbance was then detected at 450 nm by using a microplate reader (Thermo Scientific, America).
2.12 Global DNA methylation determination
Blood samples for peripheral blood mononuclear cells (PBMCs) were obtained at study entry, the first day of every treatment cycle, and progression. PBMC DNA was extracted by using Blood Genomic DNA Mini Kit (CWBIO, CW2087M, Taizhou, China). CT conversion was conducted through bisulfite conversion with EZ DNA Methylation–Golde8n Kit (ZYMO Research, D5006, Irvine, USA). The PCR of the repetitive element LINE-1 was then performed with a forward primer (5′-TTGAGTTGTGGTGGGTTTTATTTAG-3′) and reverse primer (5′-TCATCTCACTAAAAAATACCAAACA-3′). The PCR product was further digested with the Hinf I enzyme (Thermo Fisher Scientific, ER0802, MA, USA) at 37 °C for at least 1.5 h and then separated by electrophoresis on 1%–2% agarose gels.
2.13 RNA sequencing and analysis
Total RNA was isolated by using a total RNA purification kit (Sangon Biotech, B518651, Shanghai, China). Sequencing libraries were constructed from 1 μg of purified mRNA by using Illumina TruSeq RNA Sample Prep Kit (FC-122-1001, America). Libraries were pooled and 150-bp paired-end reads were sequenced on an Illumina HiSeq Xten platform. For bioinformatics analyses, raw sequence reads were initially processed by using FastQC (Babraham Institute, Cambridge, UK) for quality control. Next, adapter sequences and poor-quality reads were removed with Cutadapt. Quality-filtered reads were then mapped to the human genome by using STAR software, and only uniquely mapped reads were retained. Genes with confident sequencing signals (FPKM > 0.1 at baseline) were taken for further analysis. Differential gene expression analysis was conducted with GFOLD (V1.1.4) by Linux. Genes with fold change ≥ 2 or fold change ≤ 0.5 and Gfold absolute value > 0.3 were identified as differentially expressed genes (DEGs). Functional analysis was conducted with DAVID database and Enrichr, and P < 0.05 was considered to indicate significance. The protein–protein interaction (PPI) network of the DEGs was analyzed first by using STRING (v11.0), and a graph was generated with Cytoscape (v.3.9.0). Immune-related genes (IRGs) were retrieved from ImmPort database.
2.14 TCGA and METABRIC data mining
The clinical, mutational, and transcriptomic profiles of the TCGA and METABRIC breast cancer cohort were downloaded from cBioPortal for Cancer Genomics. In the TCGA cohort, the TNBC subgroup was filtered on the basis of the rule of negative ER, PR, and HER2 expression and included 156 patients with TNBC. In the METABRIC cohort, TNBC was defined as “ER-/HER2-” by 3-Gene classifier subtype. This cohort included 267 patients with TNBC. Immune cell infiltration scores in the TCGA TNBC cohort were calculated by ssGSEA (GSVA package) in R (x64 v4.1.0).
2.15 Statistics
Statistical analyses were performed on raw data with GraphPad Prism software v8.0. At least 3 separate experiments were performed for each measurement. Continuous variables were compared by t-test or Mann–Whitney U test when abnormally distributed. Chi-square test was conducted for categorical variables. Correlation was calculated by Pearson correlation analysis. Survival analyses were performed with the Kaplan–Meier estimator in univariate setting and Cox proportional hazards model in the multivariate model.
3 Results
3.1 DNMT1 overexpression was associated with p53 mutation and poor prognosis in TNBC
DNA methyltransferases (DNMTs) are the predominant targets of decitabine in cells [
28,
29] and patients [
30]. Interestingly, the expression levels of DNMTs have been reported to be correlated with sensitivity to decitabine treatment in cell and mouse models [
31]. Decitabine was also observed to exhibit high efficacy in treating p53-mutated MDS [
13,
14,
16]. We hypothesize that p53 mutations may potentially result in the overexpression of DNMTs in TNBC. A total of 132 patients with stage I–III TNBC who underwent surgery at the Comprehensive Breast Health Center of Ruijin Hospital from January 2009 to December 2013 were retrospectively included to validate the above hypothesis. FFPE specimens were used to construct a TMA. The baseline characteristics and detailed patient information are listed in Table S1 (seen in the Supplementary Material section) and Supplemental Data File 1, respectively. IHC staining for p53, the 3 main DNMTs (DNMT1, DNMT3A, and DNMT3B), and 5mC were performed, and the clinicopathological and prognostic profiles of the cohort were analyzed (Fig.1). We first tested the IHC effectiveness of a batch of commercially available antibodies by using sections prepared from a cancer cell line with negative and positive controls. The effective antibodies were then selected for TMA IHC (Figs. S1A–E, selected antibodies labeled in blue). We classified the p53 IHC staining patterns into 3 categories in accordance with the positive staining percentage to predict p53 mutations [
32–
34]: absent (0% positive), scattered (1%–40%), and diffused (41%–100%) staining (Fig. S1F). Consistent with previous reports [
32–
34], the absent and diffused staining patterns effectively predicted p53 mutations in the 20 samples with available
TP53 Sanger sequencing results (Supplemental Data File 1, Fig. S1G, prediction sensitivity and accuracy of 92.9% (13/14) and 90% (18/20), respectively, by DO1 and of 100% (14/14) and 80% (16/20), respectively, by E47). Thus, in accordance with the reliable DO1 staining, 103 (78.0%) specimens were predicted to harbor p53 mutations (Fig.1). This finding was in line with
TP53 mutation rates of 68% and 78% exhibited by the TNBC cohorts in the METABRIC and TCGA databases, respectively [
35,
36].
The correlation between DNMT expression and p53 mutation status was then calculated. Compared with the predicted p53 wild-type tumors, tumors with p53 mutations tended to have higher DNMT1 expression (P = 0.037) but comparable DNMT3A, DNMT3B, and 5mC levels (Fig.1). The representative IHC staining patterns of p53, DNMTs, and 5mC are shown in Fig.1 and S1H. In terms of clinicopathological features, increased DNMT1 expression was associated with increased lymphovascular invasion (LVI) and proliferative Ki67 index (Fig. S1I, both P < 0.05). Tumors with IHC patterns indicative of mutant p53 also had a high average proliferative Ki67 index (Fig. S1I).
The associations of DNMT expression and p53 mutation status with patient prognosis were next analyzed. A total of 21 (15.9%) deaths and 23 (17.4%) recurrences were observed in the cohort with a median follow-up of 85.0 months (Fig.1; Supplementary Data File 1). DNMT expression, 5mC, p53 status, and clinicopathological features were subjected to Cox regression analyses to predict recurrence-free survival (RFS) and OS. Univariate analysis revealed that DNMT1 overexpression, together with the well-known negative prognostic markers LVI, tumor size, and lymph node metastases, was significantly correlated with poor RFS and OS (Fig. S1J). The RFS and OS curves for DNMT1 and other biomarkers are shown in Fig.1 and S1K. After adjustment of clinicopathological characteristics in accordance with the multivariate model, we found that DNMT1 overexpression was still significantly associated with inferior RFS (HR = 4.14, 95% CI, 1.56–11.02; P = 0.004) and OS (HR = 4.51, 95% CI, 1.56–13.06; P = 0.006) (Fig.1).
We further investigated the association between prognosis and DNMT1 overexpression in p53 mutant and wild-type subcohorts. In the subcohort of patients with p53 mutants, high DNMT1 expression was significantly associated with poor RFS (P = 0.003) and OS (P = 0.001) (Fig.1). Multivariate analysis confirmed that DNMT1 overexpression was an independent risk factor for RFS and OS in TNBC with mutant p53 (Fig.1). In the subcohort containing 29 patients with wild-type p53, DNMT1 overexpression was associated with slightly worsened PFS and OS; however, this association lacked statistical significance (Fig. S1L, both P > 0.05). The detailed data used in the prognosis analysis of the mutant and wild-type p53 subcohorts are shown in Fig. S1M.
By screening a set of molecular and clinicopathological markers, we report for the first time that DNMT1 overexpression is an independent risk factor in TNBC.
3.2 Response and AEs in the DETECT trial
Although decitabine has been trialed on a set of solid tumors, its monotreatment or combination treatment has not been trialed on TNBC. Given that DNMT1 overexpression is associated with p53 mutation and behaves as an independent risk factor in TNBC, we designed a prospective phase I/II DETECT trial to explore the efficacy and safety of a combination of decitabine with the standard TNBC therapeutic carboplatin in patients with stage IV TNBC. No more than 1 first-line treatment at the metastatic stage was permitted in this trial. The enrolled patients were treated with consecutive low doses of decitabine per day (7 mg/m2) for 5 days then received a single application of carboplatin at a dose of AUC 6 at day 6 for every 21 days. After either the completion 6 cycles of treatment or observation of objective disease progression, the regimen was changed in accordance with the decision of physicians (Fig.2).
A total of 12 evaluable women with relapsed or newly diagnosed stage IV TNBC were enrolled from August 2017 to August 2019. Their baseline characteristics are shown in Tab.1. The median breast cancer-free interval was 11 months (range, 4–50 months). Five (42%) patients had lung metastases as the initial site and 5 (42%) had lymph node relapse when first diagnosed with recurrent disease. Nine (75%) patients received no prior treatment for metastatic TNBC, and 3 (25%) received first-line treatment when enrolled.
Regarding treatment efficacy, 5 out of 12 (42%) patients showed partial response (PR) following treatment with the combination of decitabine and carboplatin, 2 out of 12 (16.7%) patients had stable disease (SD), and 5 out of 12 (42%) had progressive disease (PD), indicating that this novel combination treatment has an ORR of 42% and a CBR of 58% in TNBC (Fig.2–2D). Among the trialed patients, #10 and #6 had excellent objective responses, showing shrinkages of 76.8% and 68.7% in the target lesions after 6 cycles of treatment (Fig.2 and S2A). In summary, the ORR observed in the DETECT study is more promising than the ORR of reported carboplatin-based therapies (42% vs. 17%–32%) for similar patients with stage IV TNBC [
37–
41].
As of October 2020, the follow-up period ranged from 4.2 months to 40.0 months (Fig.2). All patients developed disease progression and 8 out of 12 died of breast cancer (Fig.2). The median PFS was 3.8 months, whereas the median OS was 15.5 months (Fig.2 and 2G). The responders had significantly longer median PFS than the nonresponders but nevertheless had a comparable median OS (Fig. S2B and S2C). Thus, the combination of decitabine and carboplatin delivered a similar median PFS (3.8 vs. 2.1–5.5 months) and, encouragingly, a more favorable median OS (15.5 vs. 10.4–12.0 months) than the reported carboplatin-based therapies for patients with highly refractory stage IV TNBC [
37–
41].
AEs and treatment duration were next analyzed. The patients received a median of 3 cycles of treatment, wherein the patients with PR and SD all received ≥3 treatment cycles (Fig.2). The observed hematological and nonhematological AEs during treatment are shown in Fig.2 and Tab.2. As expected, AEs were predominantly associated with hematological toxicities, including neutropenia (100%; 41.7% and 16.7% with grade 3 and 4 neutropenia, respectively), leukopenia (100%), and thrombocytopenia (91.7%). For thrombocytopenia, 4 (33.3%) patients showed grade 3 adverse effects, whereas 1 (8.3%) patient showed a grade 4 decrease in platelet count. These hematological AEs mostly alleviated to grades 1 to 2 after the administration of granulocyte colony-stimulating factor (G-CSF) (Fig. S2D). Thus, the protocol was amended such that since February 2018, prophylactic G-CSF was routinely applied at 48 h after the transfusion of carboplatin. Additionally, we observed elevations in the serum levels of liver enzymes, including alanine aminotransferase and aspartate transaminase in the same 4 (33.3%) patients who received the highest number of treatment cycles (Fig.2 and S2E), indicating cumulative hepatotoxicity during treatment. Mild gastrointestinal side effects, including nausea, vomiting, and constipation, were also commonly observed (Fig.2).
3.3 Correlation between p53 mutations and responses in the DETECT trial
Decitabine acts as a DNA demethylating agent by degrading DNMTs [
28–
30]. We first analyzed the relationship between the global DNA demethylation extent of PBMCs and patient response to determine the potential predictive biomarker of decitabine treatment in TNBC. In patients #10 (PR), #11 (PD), and #12 (SD), the global methylation levels all clearly decreased after the first cycle of decitabine treatment (Fig.3, left panel, cycle 1 day 0 vs. cycle 1 day 6). The patient with the best response (#10) did not exhibit the greatest DNA demethylation extent among the 3 patients (Fig.3, right panel). Thus, the extent of DNA demethylation may be not an excellent predictive biomarker in the DETECT trial.
Despite the various clinical trials reported, the observed efficacies of decitabine-based regimens in solid tumors were elusive and their correlation with p53 status was not investigated. We thus determined whether p53 mutation status could predict patient response in the DETECT trial as observed in myeloid malignancies [
13,
14,
16]. Sanger sequencing was performed on
TP53 in 9 patients with available FFPE specimens of primary tumors. Six p53 somatic mutations—Y107*, W146*, H193R, G266V, R282W, and E286K—were detected and found to be all located in the DNA-binding domain (DBD) of p53 (Fig.3 and S3A). Strikingly, all of the 3 responsive patients were from the mutant p53 group (ORR 3/6) and none were from the wild-type p53 group (ORR 0/3) (Fig.3–3E). In addition, the mutant p53 subgroup had better PFS and OS (Fig.3, 3F, and 3G; median PFS 4.6 vs. 3.0 months,
P = 0.202; median OS 16.0 vs. 4.0 months,
P = 0.360) than the other subgroup. However, the differences in ORR and survival did not reach statistical significance due to the small size of the enrolled cohort. Hematological AEs in the 2 subgroups were also comparable (Fig. S3B). Compared with the other subgroup, the mutant p53 subgroup had a higher incidence of liver enzyme elevation and constipation presumably due to their longer treatment durations (Fig. S3B).
As observed in the retrospective TMA cohort, DNMT1 overexpression predicted poor prognosis. We thus determined DNMT1 protein levels in the 9 available tumor samples, which were all collected before treatment, by IHC staining (Fig. S3C). Of the p53-mutated breast lesions, 5/6 exhibited high DNMT1 expression, whereas only 0/3 of the p53-wild-type tumors exhibited high DNMT1 expression (Fig.3). This finding is consistent with the correlation between p53 mutation and DNMT1 overexpression observed in the TMA cohort. Interestingly, 3 out of the 5 patients with high DNMT1 expression experienced PR, whereas none of the patients with low DNMT1 expression experienced PR (Fig.3 and 3I). Regarding survival outcomes, patients with high DNMT1 expression had longer PFS than those with low DNMT1 expression. This difference just reached statistical significance (Fig.3, P = 0.048). OS did not significantly differ between patients with high and low DNMT1 expression levels (Fig.3). The limited patient number (n = 12) and available tumor biopsy samples (n = 9) in DETECT may have caused the lack of strong statistical significance.
In summary, the combination of decitabine and carboplatin produced an encouraging ORR of 42% (5/12) in patients with refractory metastatic stage IV TNBC. This response is possibly associated with p53 mutation status and DNMT1 overexpression.
3.4 Decitabine preferentially inhibited isogenic TNBC cell lines harboring DETECT-derived p53 mutations
We next dissected the effects of the 6 DETECT-derived p53 mutations on p53 function. Among the 6 mutations, W146* and Y107* have greatly truncated p53 amino acid sequences (Fig.3) that are expected to abolish p53 function completely. The remaining 4 mutations caused only 1 amino acid substitution in the p53 sequence (H193R, G266V, R282W, and E286K). From a structural viewpoint, in contrast to the classic hotspot mutations in the exposed DNA-contacting residues Arg248 and Arg273, these 4 mutations affected buried residues (low solvent-accessible surface area) that are expected to play an important role in maintaining the structural stability of p53 (Fig.4). These 4 p53 mutations may thus act as structural mutations to inactivate p53 [
9]. We confirmed that in the classic luciferase reporter assay on the representative p53 target
CDKN1A [
42], these 4 p53 mutants were indeed heavily defective in transcriptional activity, which is the core function of p53 (Fig.4).
DNMT1, DNMT3A, and DNMT3B are reported to be the cellular targets of decitabine in many solid tumors [
43,
44]. In the human TNBC cell lines HCC1937 and BT549, decitabine decreased DNMT1 most efficiently among the cellular targets of decitabine (Fig.4 and S4C). It also decreased DNMT3A with relatively high efficiency. However, it presented very limited effects on DNMT3B in both cell lines. Thus, DNMT1 is a predominant target of decitabine in TNBC as reported in other solid tumors. We next determined the effect of DETECT-derived p53 mutations on DNMT1 expression in TNBC cells. We constructed 8 isogenic TNBC cell lines by infecting p53-null HCC1937 TNBC cells with a retrovirus with empty vectors or vectors expressing wild-type p53 or 1 of the 6 DETECT-derived p53 mutants (Fig. S4D). Theoretically, any difference observed among these 8 lines could be attributed to p53 mutation status because of the identical genetic background of the cell lines. Interestingly, the p53-deficient lines (vector control and p53-mutated lines) had higher mRNA and protein levels of DNMT1 than those expressing wild-type p53 (Fig.4 and 4E, respectively). This result was consistent with the findings in the TMA cohort (Fig.1). As a control, these p53 mutants failed to upregulate p21 encoded by
CDKN1A (Fig.4). In the METABRIC breast cancer cohort,
DNMT1 expression was significantly higher in p53 mutation-prevalent TNBC than in non-TNBC (Fig. S4A). In the TNBC subcohort of METABRIC, the expression of
DNMT1 was significantly higher in the patients with p53 mutations (Fig. S4B,
P < 0.001; p53 target
CDKN1A and
ACTIN served as the control), recapitulating the findings in our TNBC TMA cohort (Fig.1) and the isogenic TNBC cell lines (Fig.4).
p53 is a well-known transcription factor. We next investigated whether
DNMT1 expression is regulated by p53 at the transcriptional level. In the luciferase report assay on H1299 cells (a p53 null cell line), wild-type p53 exhibited clear transcriptional repression activity on the
DNMT1 promoter (Fig.4,
P < 0.001; transactivation activity on the
CDKN1A promoter as the control). We constructed 2 p53 mutants (p53-L25Q/W26S and p53-L25Q/W26S/F53Q/F54S) with mutations in the p53 transactivation domain to further confirm that the regulation of
DNMT1 is dependent on p53 transcriptional activity. Although these 2 mutants retained intact DNA binding domains, they had severely defective transcriptional regulation activity on various p53 targets [
45]. As expected, the mRNA expression of
DNMT1 in cells harboring these 2 mutants had largely recovered relative to that in the cells harboring the wild-type p53 (Fig.4,
P < 0.001; expression of
CDKN1A as the control). Consequently, p53-L25Q/W26S and p53-L25Q/W26S/F53Q/F54S failed to downregulate DNMT1 at the protein level (Fig.4). These findings together indicate that wild-type p53 possibly downregulates DNMT1 at transcriptional level.
We next determined the effect of DETECT-derived p53 mutations on the response of TNBC cells to decitabine treatment. Notably, in the cell growth assay, the p53-mutated lines were significantly more sensitive to decitabine treatment than the cell line expressing wild-type p53 at different treatment concentrations (Fig.4, 0.1–10 μmol/L). This result potentially explains the better response observed in the patients with p53 mutations in the DETECT trial (Fig.3 and 3E). By contrast, p53 mutations did not significantly confer platinum treatment sensitivity to the isogenic TNBC cell lines (data not shown). Moreover, the combination of the mechanistically distinct decitabine and cisplatin predominantly showed an additive effect (combination index close to 1) on inhibiting the cell growth of p53-deficient TNBC cell lines (Fig. S4E).
We next performed rescue experiments by knocking down DNMT1 and determining whether the enhanced sensitivity of p53 mutant TNBC cells to decitabine can be abrogated (sequences of DNMT1 siRNA seen in Supplementary Data File 5). DNMT1 was efficiently knocked down in the 2 p53 mutant TNBC cell lines HCC1937 and BT549 (Fig.4). After treatment with a concentration gradient of decitabine, cell viability was determined in these 2 cell lines before and after DNMT1 knockdown. Consequently, DNMT1 knockdown dramatically decreased decitabine sensitivity in both cell lines (Fig.4, IC50 increased from < 1 μmol/L to > 10 μmol/L in HCC1937 cells; IC50 increased from approximately 1 μmol/L to > 10 μmol/L in BT549 cells). Thus, the enhanced susceptibility of TNBC cells with p53 mutations to decitabine can be reduced by DNMT1 knockdown.
3.5 Decitabine potently induced IRF7-mediated immune responses in p53-deficient TNBC
We next investigated the mechanism underlying decitabine-induced cytotoxicity in p53-deficient TNBC cells. In the DETECT trial, decitabine delivered the best responses in tumors harboring p53 G266V and R282W mutations (Fig.3–3D). R282W is 1 of the 6 hotspot mutations of p53 [
46], whereas G266V is rare in cancers. Indeed, R282W had significantly higher prevalence than G266V in the TCGA pancancer cohort and MSK-IMPACT breast cancer cohort (Fig. S5A and S5B). In addition, the expression levels of
DNMT1 and
CDKN1A were comparable between patients p53-R282W and p53-G266V in the TCGA pancancer cohort (Fig. S5A). Given that the mRNA and protein levels of DNMT1 in HCC1937 cells harboring p53-R282W and p53-G266V were comparable (Fig.4 and 4E) and these 2 isogenic cell lines exhibited comparable treatment sensitivity to decitabine (Fig.4; IC
50 values were both approximately 1 μmol/L), we chose the prevalent R282W mutation in the following study.
RNA-seq was performed to explore the change in the transcriptome of HCC1937 cells expressing p53-R282W upon decitabine treatment. We identified 1003 differentially expressed genes (DEGs), among which 870 were upregulated (FPKM fold change > 2, Gfold > 0.3) and 133 were downregulated (FPKM fold change < 0.5, Gfold < −0.3) upon decitabine treatment (Fig.5; Supplementary Data File 2). Interestingly, the 870 upregulated DEGs were highly enriched in immune response-related terms in the Gene Ontology (GO) biological process analysis. The top 2 enriched terms included leukocyte migration and immune system process (Fig.5, labeled with asterisks). Functional analyses based on REACTOME and UP-KEYWORDs also suggested strong associations between the upregulated 870 DEGs and immune response-related terms (Fig.5 and S5C, labeled with asterisks). As expected, the DNA methylation-related terms “epigenetic regulation of gene expression (GO:0040029)” , “DNA methylation (R-HSA-5334118)” , and “epigenetic regulation of gene expression (R-HSA-212165)” were enriched in cells upon decitabine treatment (Fig.5 and 5C; details seen in Fig.5). This finding is consistent with the well-established role of decitabine in DNA demethylation, providing further support for the reliability of our RNA-seq study.
We next focused on the genes involved in immune responses. The myeloid malignant cell line Thp-1 harboring the identical p53-R282W mutant was included for comparison because decitabine exhibited clinical efficacy in treating p53-mutated myeloid malignancies [
13,
14,
16]. In our RNA-seq results, 613 genes were recorded as IRGs in the ImmPort database [
47] (Supplementary Data File 2). Overall, decitabine upregulated the set of 613 IRGs with higher potency in HCC1937 cells than in Thp-1 cells (Fig.5). Moreover, decitabine showed differential preferences in upregulating the IRGs in the 2 cell lines, showing preference for
IRF7,
TNFRSF10D,
TPM2,
CCL26, and
GALR2 in HCC1937 and
IL2RG and
S100P in Thp-1 (Fig.5, labeled with orange and blue, respectively). Consequently, the transactivation profiles of these IRGs in the 2 cell lines were correlated with each other to a limited extent (Fig.5, r = 0.119,
P = 0.003). Notably, IRF7 was one of the most representative genes because it was upregulated by decitabine by up to 16-fold in HCC1937 but by only 1.4-fold in Thp-1 (Fig.5 and 5F). Protein–protein interaction (PPI) network analysis suggested that the 174 decitabine-upregulated IRGs (Gfold > 0.3) in HCC1937 clustered in a PPI network centered around IRF7 (Fig.5, brown dots). The genes in this cluster were enriched in the Toll-like and RIG-I-like receptor signaling pathways in KEGG enrichment (Fig. S5D), indicating the activation of innate immune responses. By contrast, no significant cluster was revealed for the 33 decitabine-upregulated IRGs in Thp-1 cells (Fig. S5E).
We explored the transcriptome profiles of 156 TNBC samples in the TCGA database to predict the responses of the set of immune cells upon IRF7 upregulation in TNBC. Immune cell infiltration scores for each sample were analyzed by ssGSEA, and their correlation with IRF7 mRNA expression were determined. Consequently, the infiltration levels of effective T cells, including type 1 helper, activated CD8, and effective memory CD8 T cells, were most significantly correlated with IRF7 expression in TNBC (Fig.5, all r ≥ 0.50, all P≤ 1.76E−11). The profiles of the correlation between immune cell infiltration and IRF7 expression are shown in Fig.5 and S5F.
IRF7 expression and T cell infiltration in the patients in the DETECT trial were thus determined. Given that most patients with stage IV TNBC with distant metastasis lesions were difficult to resample at the time of disease progression or medication change in routine clinical practice, we only managed to obtain post-treatment cancer tissue from the patient with the best response in the DETECT trial (Fig.2 and 2E, patient #10, harboring the p53 G266V mutation and showing a target lesion shrinkage of 76.8% after finishing 6 cycles of treatment). Given the satisfactory response of this patient, extensive radical mastectomy was performed after all decitabine application to obtain post-treatment cancer tissue. IHC staining showed that IRF7 in the cytoplasm and nuclei of the dissected tumor tissue upon treatment was dramatically upregulated relative to that in the core needle biopsy sample of metastatic lesions at diagnosis (Fig.5). CD8 was also clearly upregulated in resident intratumoral T cells after treatment (Fig.5), prompting the recruitment and infiltration of cytotoxic T cell. In summary, decitabine potently induced innate immune response, which featured striking IRF7 upregulation and T cell activation, in p53-mutated TNBC.
4 Discussion
Here, we performed a set of integrative studies to explore the potential of repurposing decitabine for the treatment of solid-tumor TNBC. Our retrospective TMA study discovered a key correlation between DNMT1 overexpression and p53 mutation and, for the first time, revealed that DNMT1 is a valuable predictive marker for poor survival in TNBC (Fig.6, left panel), validating the subsequent prospective DETECT trial on the use of decitabine, a DNMT1-degrading agent that was reported to benefit patients with myeloid malignancy and p53 mutations preferentially [
13,
14,
16]. The DETECT trial on patients with metastatic stage IV TNBC reported an ORR of 42% (5/12), which is more encouraging than the previously reported ORRs (17%–32%) of carboplatin-based regimens for this highly refractory disease [
37–
41]. Our DETECT study further identified a preferential benefit in p53-mutated TNBC (ORR 3/6 vs. 0/3; median PFS 4.6 vs. 3.0 months; median OS 16.0 vs. 4.0 months) (Fig.6, middle panel), recapitulating the clinical efficacy of decitabine in myeloid malignancies reported by us and others [
13,
14,
16]. Finally, our matched mechanistic study on the isogenic TNBC cell model revealed that decitabine triggered an innate immune response featuring striking IRF7 upregulation and T cell activation (Fig.6, right panel).
Notably, the decitabine-induced immune response is more profound in the studied p53-mutated TNBC cell line HCC1937 than in the p53-mutated myeloid malignancy cell line Thp-1. This profound immune response exhibits strong preference in the upregulation of genes involved in innate immune response, particularly
IRF7. The mechanism of how decitabine treatment potently upregulates
IRF7, a master of innate immune response [
48] and a regulator of adaptive immunity [
49], remains unknown.
IRF7 expression was previously reported to be heavily repressed by promoter hypermethylation in solid tumors, including fibrosarcoma and liver cancer [
50,
51]. Moreover,
IRF7, but not
IRF5, was specifically silenced by promoter hypermethylation in immortal fibroblasts derived from patients with p53-deficient Li–Fraumeni syndrome and could be reactivated by decitabine pretreatment, leading to efficient cell death through senescence [
52]. Thus, in p53-deficient TNBC, decitabine is likely to upregulate
IRF7 by directly demethylating its promoter. Investigating whether the striking upregulation of
IRF7 by decitabine is common in other types of solid tumors harboring p53 mutations is interesting.
The DETECT study is limited by its small cohort size, which accounted for the failure to reach statistical significance despite the higher response rate and better survival observed in the p53-mutated subgroup than in other subgroups. Considering the manageable AEs observed in the DETECT study, an extension trial for this regimen is desired to consolidate the current promising findings. In the envisioned extension trial, p53 mutation and DNMT1 overexpression could be used as predictive biomarkers before treatment, whereas IRF7 upregulation could be applied as a prognostic biomarker during treatment.
In summary, the current studies on TNBC, together with previous works on myeloid malignancies, indicate the possibly extensive dependence of decitabine on p53 mutations. Given that p53 has a mutation frequency of approximately 50% in cancer, decitabine-based regimens may potentially broaden the population of patients with cancer that benefits from precision oncology.
PDF
(3834KB)
