Background

Brain radiotherapy plays an important role in the management of brain metastasis [1–3]. Although effective, brain radiotherapy may cause both short and/or long-term toxicity leading to reduction in quality of life[3, 4]. This toxicity, referred as radiotherapy-induced brain injury (RBI) may occur during and/or after treatment and is classified as acute (within days and up to 4 weeks after the end of radiotherapy), early-delayed (1 month to 6 months after radiotherapy), or late-delayed (later than 6 months to several years after treatment) [5–7]. The pathophysiology of RBI is not well understood. However, evidence supports a multifactorial process leading to apoptosis and functional alternations across multiple brain cell types, and vascular changes leading to disruption of the blood-brain-barrier (BBB) [6, 8–11].

Currently, diagnosis of RBI is clinical and radiologic, and may be difficult to distinguish from tumor progression [12–18]. While early onset RBI, is mostly reversible, late appearing RBI, which can present either as radio-necrosis or as periventricular leukoencephalopathy, is mostly irreversible [5, 6, 9, 18–20]. Given the critical importance of accurate diagnosis of RBI for optimal patient management, there is an unmet need for sensitive, specific, and readily available diagnostic tools to monitor RBI development. A validated, reliable set of biomarkers to detect and monitor RBI would be one way of meeting this challenge. In this study we aimed to explore the utility of plasma-based biomarkers for detection and monitoring RBI using a previously published approach [21–26] based on identifying cell-specific methylation patterns in circulating free DNA (cfDNA). Since DNA methylation signature is specific for each cell-type, this approach enables to infer cell death in various settings. Based on this approach we developed a quantitative, cell specific circulating cfDNA markers for brain injury [25, 27]. In contrast to approaches designed to identify circulating tumor DNA [28], this approach does not rely on genetic differences between the host and the cancerous tissue of interest, and hence can detect death of genetically normal cells [29]. As radiotherapy is hypothesized to induce brain cells apoptosis as well as BBB disruption, cell free DNA fragments derived from neurons and glia, collectively termed brain-derived circulating free DNA (BncfDNA), are expected to be released to the circulation, and hence be identified by specific methylation signatures. In this report we describe BncfDNA dynamics throughout radiotherapy administered for the treatment of brain metastases, in the context of RBI.

Methods

Results

bncfDNA levels increase in RBI and decrease following acute RBI resolution

Nine patients suffered from RBI. Clinical presentation is correlated with a marked increase of total bncfDNA before and during RBI in 8 patients, and mild increase in 1 patient (Figs 1A and 1C and S2–S4 and S10). Interestingly, in 3 patients gradual elevation in bncfDNA precedes the clinical presentation of acute RBI (Figs 1A and S2; patients 1, 10, 11), suggesting that bncfDNA elevation precedes the onset of acute RBI before it is clinically evident. Clinical improvement of acute RBI symptoms, documented in 3 patients, is correlated with a decrease in total bncfDNA (Figs 1D and S2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. BncfDNA levels dynamic during RBI and non-RBI.

Total BncfDNA levels measured for each patient (in log10 scale) during follow-up period for patients with RBI [A] and non-RBI [B]. Clinical onset of RBI is represented by a star (acute RBI–blue, early delayed–orange, late delayed–green). Clinical presentation of non-RBI is marked as detailed in the legend (progressive disease–red square, tumor response–brown hexagon; stable disease–purple rhomboid;). In patient 4, 18 with RBI, additional time-points of non-RBI clinical onset are marked in accordance to Fig 1B, as mentioned above. Fold change (log 2 scale) of bncfDNA levels in 9 patients during RBI onset [C], 3 patients during resolution of RBI [D], and 15 patients during non-RBI events [E]. Total bncfDNA marked in purple represent the mean summation of all cell-specific bncfDNA; oligodendrocytes (blue); Neuron (green) and astrocytes (Yellow). In the Y-Axis fold change as calculated: the ratio in bncfDNA levels measured during RBI, RBI resolution or non-RBI and the levels measured in the antecedent time point. BncfDNA: brain-derived circulating DNA SD: stable disease; PD: Progressive disease; TR: Tumor response; URP: Undetermined radiological progression.

https://doi.org/10.1371/journal.pone.0285646.g001

Discussion

In this proof-of-concept study we aimed to determine if bncfDNA can be used to detect and monitor radiation related brain tissue injury among patients with brain metastases. We followed 24 patients before, during and after radiotherapy clinically, radiologically (MRI) and by sequential measurement of plasma derived bncfDNA. We found that bncfDNA elevation characterizes radiation-induced brain injury, both acute, early and late-delayed, as defined by clinical and radiological criteria. In 3 patients, increased levels of bncfDNA preceded the presentation of RBI (acute and early-delayed respectively), suggesting that bncfDNA elevation also reflects asymptomatic RBI and can indicate RBI before it is clinically evident. Thus, we suggest that bncfDNA may serve as a biomarker for early detection of RBI, both symptomatic and asymptomatic. In addition, decreased bncfDNA levels, reflecting cfDNA clearance, characterize resolution of RBI. We also found that changes in bncfDNA are not consistent in non-RBI events. In tumor response and stable disease bncfDNA both rise and drop, while progressive disease is associated with a decrease in bncfDNA levels. Interestingly, increase in bncfDNA levels is common to blood samples collected 45–90 days after brain radiotherapy, either if it is associated with RBI or tumor response. As this time period corresponds to the typical timeframe for early-delayed RBI appearance [5, 6, 30], we suggest that elevation of bncfDNA levels in those patients might reflect asymptomatic RBI [31]. The consistent increase of bncfDNA in RBI and decrease in progressive disease suggests its potential to differentiate between radiotherapy effects and tumor progression. However, additional mechanisms for decreasing levels of bncfDNA during progressive disease might exist.

Theoretically, enlargement of brain metastases may result in elevation of bncfDNA, as brain metastases may induce peritumoral tissue injury through mass effect, inflammatory cascade, or other possible mechanisms. In a separate work [27] we identified elevated neuron-, oligodendrocyte- and astrocyte-derived cfDNA in a small subpopulation of patients with pre-radiated brain metastases, while in most patients (26 of 29) brain metastases were not associated with significant elevated levels of bncfDNA compared with healthy individuals and cancer patients without brain metastases. In addition, no correlation is found between bncfDNA levels and other parameters in patients with pre-radiated brain metastases, including age, gender, recent neurosurgical intervention, and previous or recent chemotherapy exposure. While more work is needed to understand the relationship between brain metastases and brain-derived cfDNA, our findings suggest that bncfDNA levels reflect RBI better than they do brain metastases progression. The level of total cfDNA reflects systemic tumor burden, that can increase or decrease independently of brain disease. Thus, compared to the ratio of bncfDNA to total cfDNA, absolute levels of bncfDNA may reflect RBI (S10 Fig). Brain-derived cfDNA levels depend both on the amount of brain cell injury, as well as its ability to enter the plasma. Thus, the pronounced increase in bncfDNA levels as measured in association with RBI in contrast to brain metastases progression may indicate a stronger BBB disruption induced by radiotoxicity [32, 33]. This also suggests the potential of bncfDNA to serve as a biomarker for BBB disruption to guide management in patients with brain metastases. However, additional mechanisms related to bncfDNA clearance, can also contribute to the individual variability of bncfDNA levels in pretreated patients as well as in non-RBI non-progressive tumor status. Thus, although decrease in bncfDNA levels during progressive disease might be related to a lack of brain cell injury in contrast to RBI, it is also might be attributed to differences in other mechanisms related to BBB disruption and/or cfDNA clearance from the plasma. Further studies are required to substantiate these ideas.

As differentiating tumor progression from pseudo-progression and radionecrosis is critically important for patients’ management of primary and secondary brain tumors, several non-invasive strategies have been tested mainly in the context of the primary brain tumor glioblastoma multiforme (GBM), and are mainly based on tumor-derived circulating biomarkers to detect tumor progression [20–31, 33, 34]. However, all previously published methods, are based on the identification of tumor derived DNA, which in the case of brain metastases might be challenging owing to the broad molecular diversity among metastatic malignancies (variation by tumor histology and organ of origin with additional patient specific variation). In contrast to strategies using tumor DNA for detection of tumor progression, we suggest a method for detecting non-tumor DNA as a biomarker for normal tissue injury. As variations in brain-derived cfDNA levels appear to correlate with brain tissue damage, it can be used to detect treatment-related brain damage in the context of brain metastases. Moreover, the clear divergence in bncfDNA changes in RBI (increased levels) and progressive disease (decreased levels), suggests a potential utility of bncfDNA to help distinguish RBI from progressive disease in clinically ambiguous cases.

Given the very short half-life of cfDNA (estimated 15–120 minutes) [34], elevated levels of BncfDNA likely reflect active cell death of glia and neurons during RBI. Thus, in addition to its potential diagnostic utility, measurement of bncfDNA may suggest which brain cell types are being affected by brain radiotherapy at the time of sampling. Our results suggest the involvement of both neuronal cell apoptosis as well as glial cell death in all types of RBI, although in different proportions. Specifically, the most consistent elevation in acute RBI is related to neuron-derived bncfDNA, which might indicate a higher sensitivity to radiation damage of neuronal cells comparing to glial cells. In contrast, elevation of oligodendrocyte-derived cfDNA in different proportion between patients with acute RBI suggests for variable cellular radiosensitivity of oligodendrocytes between individualized patients. However, a more prominent and consistent increase in oligodendrocyte-derived cfDNA associated with early-delayed RBI and late-delayed RBI, supporting current knowledge attributing oligodendrocytes apoptosis an important role in the pathophysiology of these processes [9, 20, 31]. Moreover, as current hypotheses claim astrocyte proliferation and astrogliosis as a fundamental process induced by radiotherapy contributing to brain damage [9], our results indicating elevation of astrocyte-derived bncfDNA associated with RBI, suggest an induction of astrocyte cell death by acute and late radiation effects, as an additional component of RBI pathophysiological mechanism, requiring further investigation.

This pilot study is limited by the low number of patients tested, as well as by a lack of tissue biopsy to confirm diagnosis of tumor progression and RBI. Hence, larger studies will be required to assess specificity and sensitivity of the various glial and neuronal bncfDNA, as well as the pathophysiological and clinical implications. However, there is a reason to believe that measurements of cfDNA methylation-based glial and neuronal cell deaths can be used as a diagnostic assay for early identification and monitoring of RBI, as well as differentiating radiation treatment effects from tumor progression.

Conclusion

By measuring methylation markers of bncfDNA, we can identify neuron and glia cell death associated with RBI, both in symptomatic and asymptomatic patients. This quantitative minimally invasive method could be part of a novel biomarkers for early detection and monitoring RBI, with the potential to distinguish between tumor progression and pseudo-progression. Additional large-scale studies are required to further understand the biological background and clinical outcome of radiotherapy toxicity to the brain as well as to other organs.

Supporting information

Acknowledgments

The authors are grateful to the patients who participated in this study. We thank Ziva Inbar and Aviad Klinger for their great contribution in statistical analysis and graphs production; Dana Sheril Roffe and Aliza Moshayev for assisting with serum collection.