.
O P E N A C C E S S S O U R C E : Cell Metabolism @ ScienceDirect
Highlights
• Treatment type can induce different kinetics of cell-free DNA (cfDNA) release
• Irradiation-induced necrosis is a key contributor of cfDNA release in preclinical models of cancer
• Cellular senescence is a major determinant of cfDNA kinetics and blocks its release
• Selective elimination of senescent cells through apoptosis increases cfDNA release
Summary
The kinetics of circulating cell-free DNA (cfDNA) release may provide a real-time assessment of induced cell death. However, there is a limited understanding of the underlying biological rationale for cfDNA release following distinct treatments and cell death mechanisms. Here, we uncover a complex interplay between apoptosis, necrosis, and senescence in determining cfDNA release kinetics. Utilizing multiple in vitro and in vivo preclinical models, we show how cfDNA release is modulated through a combination of apoptotic and senescent triggers and inhibitors. Interestingly, we identify treatment-induced senescence as a previously unrecognized determinant of cfDNA kinetics that can counteract its release. Necrosis is the predominant cell death mechanism that consistently contributes to cfDNA release in response to ionizing radiation, and, surprisingly, apoptosis plays a comparatively minor role in some tumors. Based on our results, we propose a model to explain cfDNA release from cells over time, with important implications for future studies.
Introduction
The presence of fragmented cell-free DNA (cfDNA) within peripheral blood has been recognized for more than 70 years (Mandel and Metais, 1948). Since then, investigation of circulating cfDNA in serum and plasma has seen tremendous interest from researchers in many disciplines of medicine, including obstetrics, transplant medicine, and oncology. cfDNA is thought to be released by cells undergoing cell death. Apoptosis has been proposed as the primary mechanism by which cfDNA is released into circulation due to its characteristic fragment length (∼140–180 bp, mirroring the size of nucleosome-bound DNA fragments) (Jahr et al., 2001; Diehl et al., 2008; Lo et al., 2010; Thierry et al., 2010; Mouliere et al., 2011; Schwarzenbach et al., 2011; Diaz and Bardelli, 2014; Jiang et al., 2015). More recently, necrosis has been recognized as another potential source of cfDNA, with the short cfDNA fragments produced secondarily as a result of partial digestion by plasma nucleases (Jahr et al., 2001; Jiang et al., 2015; Jiang and Lo, 2016; Wan et al., 2017; Wang et al., 2017).
In cancer patients, a portion of cfDNA is derived from tumor cells (i.e., circulating tumor DNA [ctDNA]). Because ctDNA is released during tumor cell death, the kinetics of ctDNA may reflect the efficacy of anti-cancer cytotoxic therapies. In addition, due to its short half-life in the bloodstream (Lo et al., 1999; To et al., 2003; Diehl et al., 2008; Jiang et al., 2015; Yao et al., 2016; Muhanna et al., 2017), ctDNA may provide a real-time snapshot of disease burden or treatment response. Yet, to date, few studies have investigated the kinetics of ctDNA or cfDNA release. The first description of cfDNA kinetics came from patients with nasopharynx cancer, in whom ionizing radiation (IR) caused increased levels of ctDNA within 1 week (Lo et al., 2000). More recent studies have begun to suggest that early ctDNA release could reflect tumor cell death and be a harbinger for favorable treatment outcomes in multiple cancer types (Kamat et al., 2006; Rago et al., 2007; Cao et al., 2012; Xi et al., 2016; Husain et al., 2017). However, there continues to be a limited understanding of such cfDNA kinetic patterns and their underlying biology—for instance, how different treatments affect mechanisms of cell death and the timing and magnitude of subsequent cfDNA release. We sought to characterize the kinetics of cfDNA release and to evaluate the biological underpinnings of this process. Our findings reveal a complex interplay of distinct mechanisms of cell death and their influence on cfDNA release.
Results
Early Release of cfDNA Is Mediated by Treatment-Specific Induction of Apoptosis and Necrosis
To evaluate the relationship between cfDNA kinetics and apoptosis, we characterized cfDNA release in a panel of head and neck squamous cell carcinoma (HNSCC) and non-small-cell lung cancer (NSCLC) cell lines following induction of apoptosis. Cells were exposed to a single dose of staurosporine, and their caspase activity and cfDNA levels were measured at various times (6–48 h) after treatment. Upon treatment, apoptosis was rapidly induced in all of the examined cell lines (Figure 1B; Figure S1A), followed by cfDNA release within 48 h of treatment (Figure 1C). Pretreatment of Cal33 and HMS-001 cells with a pan-caspase inhibitor (z-vad-fmk) before treatment with staurosporine lead to a near-complete reduction in caspase activity (Figure S1C) and subsequent partial reduction in cfDNA release (Cal33: −50.6% ± 9.6%, p < 0.001 at 24 h; −52.4% ± 11.3%, p < 0.01 at 48 h; HMS-001: −81.9% ± 13.5%, p < 0.001 at 24 h; −46.4% ± 11.2%, p < 0.001 at 48 h) (Figure 1D). Interestingly, we observed no statistically significant correlation between staurosporine-induced caspase activity and cfDNA release over time, which is in agreement with the partial effect of pan-caspase inhibition on cfDNA release (Figures S1C and S1E; Figure 1D). HMS-001 cells showed a more robust apoptotic response to staurosporine treatment than Cal33 cells (Figures 1B and 1E), with Cal33 cells depicting a larger propensity toward necrosis-induced cell death and cfDNA release (Figures 1E and 1F). In stark contrast to staurosporine treatment, induction of genotoxic stress through IR exposure resulted in little-to-no increase in caspase activity or cfDNA release within 6–48 h (Figures 1G and 1H).
Figure 1. Early Release of cfDNA Is Mediated by Treatment-Specific Induction of Apoptosis
(A) Experimental schema for (B), ©, (G), and (H). Human papillomavirus (HPV)-positive and -negative HNSCC cell lines and NSCLC cell lines are indicated in each panel.
(B and C) Cells were treated with 0.5 μM staurosporine, and caspase activity (B) and cfDNA release (D) were measured. (B) Data represent fold change in caspase activity from control cells, as measured by luminescent caspase activation assay at indicated times in (A). © Data represent fold change in cfDNA release from control cells, as measured by qPCR assay using hLINE-1 primers at indicated times in (A).
(D) Following pretreatment with z-vad-fmk, Cal33 and HMS-001 cells were treated with 0.5 μM staurosporine, and cfDNA release was measured. Shown is the fold change in cfDNA release post-treatment to control cells. Data represent mean values ± SEM.
(E) The effect of 0.5 μM staurosporine and 10 μM ionomycin on proliferation (% confluence) in Cal33 and HMS-001 cells monitored by the IncuCyte live-imaging system. Grey dashed line indicates the time of treatment initiation.
(F) Fold change in cfDNA release post-ionomycin treatment to control cells in Cal33 and HMS-001 cells.
(G and H) Cells were treated with 8 Gy IR, and caspase activity (G) and cfDNA release (H) were measured. Data represent fold change in caspase activity (G) and cfDNA release from control cells (H), as described above (B and C). (B, C, and F–H) Data represent mean values ± SD; representative experiment is shown (n = 2–3). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; ns, not significant; unpaired Student’s t test (D).
See also Figure S1.
Delayed Release of cfDNA after Genotoxic Stress
As IR did not trigger apoptosis or cfDNA release from HNSCC and NSCLC cells within 48 h of exposure, we next examined later time points (72–144 h) to measure delayed effects (Figure 2A). All of the cell lines showed an increase in caspase activity between 72–144 h post-IR (Figure 2B), which was corroborated by immunofluorescence staining for cleaved-Poly (ADP-ribose) polymerase (PARP) in HMS-001 cells treated with IR (Figure S1A). Treated cells exhibited variable magnitudes and profiles of cfDNA release, with a subset displaying peak release between 72–96 h post-IR (Figure 2C). IR-induced apoptosis and cfDNA release over time were not strongly correlated (Figure 2D), and caspase inhibition had a variable impact on IR-induced delayed cfDNA release (Cal33: −30.5% ± 6.6%, p < 0.01 at 96 h; −4.17% ± 6.4% p > 0.05 at 144 h; HMS-001: −42.6% ± 11.8%, p < 0.001 at 96 h; −40.9% ± 12.8%, p < 0.001 at 144 h) (Figure 2E; Figure S1D).
Figure 2. Delayed Release of cfDNA following Genotoxic Stress In Vitro
(A) Experimental schema for (B) and ©.
(B and C) Cells were treated with 8 Gy IR, and caspase activity (B) and cfDNA release © were measured. Data represent fold change in caspase activity (B) and cfDNA release © from control cells at indicated times in (A). Data represent mean values ± SD; representative experiment is shown (n = 2–3).
(D) Correlation between fold change in caspase activity and cfDNA release over time (72–144 h) post-8 Gy IR. (Pearson R value shown, p > 0.05). The area-under-the curve (AUC) was calculated by summing the fold change in caspase activity or cfDNA release between 72–144 h. Data represent mean values ± SD from three independent experiments.
(E) The effect of pan-caspase inactivation (z-vad-fmk) in Cal33 and HMS-001 cells treated with 8 Gy IR on cfDNA release. Shown is the fold change in cfDNA release to control cells. Data represent mean values ± SEM. ∗p < 0.05, ∗∗p < 0.01. ∗∗∗p < 0.001; ns, not significant; unpaired Student’s t test (E).
See also Figures S1–S3.
To account for the affect of cfDNA degradation on the observed differences in cfDNA release profiles, we subjected conditioned media to various perturbations and serial analyses. We found that cfDNA degradation over time was consistent across cell lines and that incubation of conditioned media at 4°C or addition of EDTA blocked cfDNA degradation (Figure S2).
IR-induced cfDNA release was also independent of HNSCC cell line clonogenic survival and human papillomavirus (HPV) status (Figures S3A–S3D). Interestingly, the NSCLC cell lines exhibited a trend toward decreased cfDNA release over time compared to the HNSCC cell lines, which has been previously described in lung adenocarcinoma patients (Abbosh et al., 2017; Figures S3E–S3G).
Delayed cfDNA Release and Cell Death Mechanisms In Vivo
To further characterize the mechanisms underlying delayed cfDNA release kinetics, we established 5 xenograft mouse models and serially measured cfDNA levels following a single dose of IR to the tumor. In 3 of the 5 xenograft models, peak release in tumor-derived cfDNA release was observed 96–144 h post-IR (Figure 3A). Interestingly, in one HNSCC (Vu147T) and one NSCLC (HCC-827) xenograft model, we consistently observed very low to undetectable levels of tumor-derived cfDNA, either at baseline or following IR (Figures S4A and S4B). Notably, Vu147T and HCC-827 were the lone examples of both an HNSCC and NSCLC cell line, respectively, without cfDNA release in vitro following IR (Figure 2C).
Figure 3. The Effect of Genotoxic Stress on cfDNA Release in Xenograft Mouse Models
(A) The effect of 20 Gy IR on cfDNA release from HNSCC (n = 2) and NSCLC (n = 1) xenograft tumors. Data shown are fold change in cfDNA density from Pre-IR (0 h) at indicated times in experimental schematic (top).
(B) Immunohistochemistry of endpoint HNSCC and NSCLC xenograft tumors for cleaved caspase-3 and necrosis. Representative images from three HNSCC cell line xenograft tumors (HMS-001, Cal33, and Vu147T) and two NSCLC cell line xenograft tumors (HCC-827 and PC-9). Quantification of images between control (0 Gy) and irradiated (20 Gy) endpoint tumors. The number of mice per condition were as follows: HMS-001 (n = 8), Cal33 (n = 9), Vu147T (n = 7), HCC-827 (n = 5) and PC-9 (n = 5–6). Data represent mean values ± SEM. Scale bar, 50 μm. ∗p < 0.05, ∗∗p < 0.01. ∗∗∗p < 0.001; ns, not significant; unpaired Student’s t test (B) and repeated-measures ANOVA followed by Dunnett’s multiple comparison test (A).
See also Figure S4.
Although most irradiated xenograft models exhibited increased apoptosis levels, only a small proportion of tumor cells were positive for apoptotic markers (0%–9.5%) (Figure 3B; Figure S1B). In 2 of the 3 models that displayed delayed cfDNA release, there was a notable increase in necrosis post-IR (mean absolute difference of 6%–12%, p < 0.05 for both). Moreover, the third model (PC-9) showed the highest overall percentage of necrotic area within the xenograft tumors (Figure 3B). Consistent with their behavior in culture, Vu147T xenografts displayed the lowest overall and induced levels of apoptotic markers (Figure 3B). More interestingly, the two xenografts with undetectable cfDNA release showed little-to-no necrosis (Figure 3B). Taken together, our findings suggest that necrosis and (to a lesser extent) apoptosis appeared to be the dominant mechanisms contributing to the observed IR-induced release of cfDNA. Thus, our in vivo models recapitulate the in vitro findings of cfDNA release and reveal potentially generalizable underlying biological determinants of delayed cfDNA release following genotoxic stress.
Cellular Senescence Blocks cfDNA Release
To evaluate the contribution of other IR-induced mechanisms on cfDNA release, we next examined a marker of growth arrest and cellular senescence (p21), the other key cell responses to IR (Maier et al., 2016). IR exposure resulted in a reduction in p21 levels in the HMS-001, Cal33, and PC-9 xenograft tumors, but not in the Vu147T and HCC-827 xenografts (Figure 4A). More interestingly, HCC-827 showed increased IR-induced p21 expression. Additional candidate markers were examined in a subset of the xenografted tumors for the effects of intratumoral vessel density, hypoxic tumor microenvironment, and cell proliferation on cfDNA levels (Figures S4C and S4D). There was no difference in vessel density or levels of hypoxia between treatment groups. In the HMS-001 and Cal33 xenograft models, IR caused a marked drop in Ki67 staining, an indication of reduced proliferation. In contrast, Vu147T xenografts showed no significant difference in Ki67 staining between the irradiated and control tumors. Overall, these results suggest that IR-induced cellular senescence is negatively correlated with cfDNA release.
