Evaluation of the impact of pre-operative stereotactic radiotherapy on the acute changes in histopathologic and immune marker profiles of brain metastases

The unique acute effects of the large fractional doses that characterize stereotactic radiosurgery (SRS) or radiotherapy (SRT), specifically in terms of antitumor immune cellular processes, vascular damage, tumor necrosis, and apoptosis on brain metastasis have yet to be empirically demonstrated. The objective of this study is to provide the first in-human evaluation of the acute biological effects of SRS/SRT in resected brain metastasis. Tumor samples from patients who underwent dose-escalated preoperative SRT followed by resection with available non-irradiated primary tumor tissues were retrieved from our institutional biorepository. All primary tumors and irradiated metastases were evaluated for the following parameters: tumor necrosis, T-cells, natural killer cells, vessel density, vascular endothelial growth factor, and apoptotic factors. Twenty-two patients with irradiated and resected brain metastases and paired non-irradiated primary tumor samples met inclusion criteria. Patients underwent a median preoperative SRT dose of 18 Gy (Range: 15–20 Gy) in 1 fraction, with 3 patients receiving 27–30 Gy in 3–5 fractions, followed by resection within median interval of 67.8 h (R: 18.25–160.61 h). The rate of necrosis was significantly higher in irradiated brain metastases than non-irradiated primary tumors (p < 0.001). Decreases in all immunomodulatory cell populations were found in irradiated metastases compared to primary tumors: CD3 + (p = 0.003), CD4 + (p = 0.01), and CD8 + (p = 0.01). Pre-operative SRT is associated with acute effects such as increased tumor necrosis and differences in expression of immunomodulatory factors, an effect that does not appear to be time dependent, within the limited intervals explored within the context of this analysis.


Methods
Patients. Patients who underwent pre-operative SRT followed by resection of brain metastasis were queried from an institutional registry (IRB# 1672008). The study was approved by the Miami Cancer Institute Institutional Review Board, and informed consent was obtained from every patient. All patients were treated with a previously established dose-escalated pre-operative SRS paradigm to the gross tumor volume with no additional clinical target volume or planning target volume expansions 16 . Only patients who also had non-irradiated primary tumor tissue samples available for comparative analyses were eligible for this particular study. Patient information, including primary tumor histology, size, volume, and location as well as treatment details, including prescription dose, number of fractions, and dose per fraction were abstracted from the electronic medical record. The start and end timing of SRS/SRT were extracted from the radiation oncology treatment database, and surgical details, including the time of surgery, were collected from the operative reports. Local failure was defined using the Response Assessment in Neuro-Oncology Brain Metastases (RANO-BM) criteria 17 with recurrence identified as enhancing tumor apart from post-surgical changes and confirmed by multi-disciplinary peer review. All the methods adhered to relevant ethical guidelines for handling human data.
Tissue analysis. All non-irradiated primary tumors and paired irradiated brain metastases were evaluated for tumor necrosis using hematoxylin-eosin staining 18 T-cells (CD3 +, CD4 +, CD8 +), natural killer cells (CD56 +), vessel density (CD31 +), vascular endothelial growth factor (VEGF), and apoptotic factors (caspase-3) were evaluated by immunohistochemical (IHC) analyses. Immunomodulatory effects were assessed by determining CD3 + (T-cell receptor) (SP7 antibody), CD4 + (T helper cell) (4B12 antibody), CD8 + (cytotoxic T lymphocyte) (4B11 antibody), and CD56 + (natural killer cell) (123C3.D3 antibody) staining 19 . The number of positive cells per high power field (HPF, 400x) were counted only in the vicinity of the tumor nests, averaging over 10 high-power-fields. If less than 1 positive cell per HPF on average was positive, the score was recorded as negative. Mean vessel density (MVD) was assessed by CD31 staining (JC70A antibody) 20 . The number of cells positive per low-power-field (LPF, 100x) was counted. On average, 10 LPFs were included. VEGF (EP1176Y antibody) was assessed by determining the H-score, as defined by (3 × % of intensely-positive tumor cells) + (2 × % of moderately-positive tumor cells) + (% of weakly-positive tumor cells), regardless of magnification 21 . To assess apoptosis (caspase-3) (polyclonal antibody), an H-score was calculated using the same criteria 22,23 . Statistical analysis. For continuous variables, means and standard deviations (SD) were used to present normally distributed data, with medians and interquartile ranges for non-normal data. For categorical data, sample size and percentages were computed. For all univariate analyses, Welch's t-test was used to compare the paired samples and the Wilcoxon rank sum test was used for non-normally distributed data; a two-sided test was used to detect statistically significant differences. A p-value of < 0.05 was considered statistically significant. To assess the relationship between dose/volume and H&E necrosis, a loess line was fit to the data to ascertain a linear relationship. www.nature.com/scientificreports/ Representative tissue samples demonstrating pairwise comparisons of necrosis and immunomodulatory cell populations from a brain metastasis treated with pre-operative SRS and a non-irradiated NSCLC primary tumor are illustrated in Fig. 1. Tumor necrosis was found to be significantly higher in irradiated brain metastases than non-irradiated primary tumor tissues (mean paired difference: 40, SD: 56, p < 0.001) ( Fig. 2A). There appeared to be no difference in the proportion of tumor necrosis with respect to time interval from SRT to surgery: there was a median of 40% necrosis observed < 24 h after SRT, 65% at 24-48 h, 52% at 48-72 h, and 45% at > 72 h (p = 0.56) (Fig. 2B). Given the wide spectrum of tumor sizes in this study, we also looked at the effect of tumor necrosis in patients treated with SRS using a volume-corrected dose analysis (prescribed dose in Gy/tumor volume in cm 3 ). No linear relationship was observed on the basis of the loess line. In the three patients (14%) who experienced a local failure, two had cerebellar metastases and the other one was occipital. Interestingly, the necrosis score of the brain metastases that demonstrated local failure was 0 for two patients and only a 10% increase compared to the primary tumor in the third patient.

Discussion
In addition to the classic 4 "R's of radiation biology" (repair, repopulation, redistribution, and re-oxygenation), the unique effects of radiosurgery commonly confer its designation as the fifth "R" 24 . Yet, published reports evaluating the radiobiologic effects of SRS/SRT are limited to preclinical studies and a few clinical reports in extracranial disease sites, even though treatment of intracranial tumors remains the most common indication for stereotactic treatment. To date, limited data regarding intracranial changes have been published. The present study provides the first in-human comparative evaluation of the acute biologic effects of SRS/SRT employing a comprehensive evaluation of multiple tissue parameters, including tumor necrosis, immunomodulatory cell populations, angiogenesis, apoptosis, and vascular density. We observed a significantly higher proportion of tumor necrosis in irradiated brain metastases compared to non-irradiated primary tumors. Despite the lack of pre-irradiation tumor biopsies (which may be limited in histologic evaluation and be subject to sampling bias) and given the substantially higher rates of tumor necrosis and the histologies evaluated in this study (which are not commonly associated with innate tumor necrosis), we attributed these effects to SRS/SRT. We previously published indirect, but supportive evidence in a small series of 4 patients evaluated with multiple PET-FDG scans. The results showed significant FDG uptake increase within 24 h, which can occur early in the necrosis/inflammation process 15 . We surmise that tumor necrosis occurs soon (around 24 h after treatment) and that this early effect would persist for an extended window (of a few days) during which resection of the brain metastasis typically occurs. This hypothesis is supported by pre-and post-SRT biopsy specimens from ten patients treated with spine SRT and surgical stabilization where tumor necrosis was not observed in any of the pre-SBRT or same-day surgery biopsy specimens, but was demonstrated in 5 of the 6 biopsy specimens obtained a day after surgery 25 . Although most pre-operative SRT series have utilized a 24 h to one week interval to surgery, and ongoing clinical trials range up to 2 weeks (NCT03750227 and NCT03741673), there is no optimal time window suggested by this collective analysis.
Dose and fractionation schedules for treatment of brain metastasis are based on the empirical evidence balancing the rate of local disease control with the risk of radiation necrosis 26 . These decisions are based on imagingdefined definitions without comparative histologic analyses. Pre-operative radiosurgery series typically utilize a 20% dose-reduction strategy, with multiple institutions using doses as low as 12 Gy in 1 fraction 12 . In our series we employed an approach utilizing the definitive doses for intact brain metastases established in cooperative group trials 27 and validated in a recently conducted phase I dose-escalation trial in patients undergoing preoperative SRS and resection 16 . Although systematic dose-escalation SRT experiences with extracranial tumor sites, such as prostate cancer, have demonstrated a dose-dependent increase in tissue effects (albeit with 2-year post-treatment biopsies) which correlate with risk of local failure 28 a similar relationship was not observed in this study. However, our post-SRS/SRT tissue acquisition window was very narrow. The 1-year actuarial local control rate of 95% achieved with higher-than-conventional dosing is promising and merits long-term follow-up and study. Of note, it is interesting that descriptively, two of the three patients with a local failure had 0% necrosis in the resected tumor sample, and the third patient with only a 10% higher necrosis rate than the primary tumor. In our prior PET study, failure to respond with acute FDG uptake elevation was associated with local failure 15 leading us to postulate that the two phenomena could be correlated. Therefore, this study also supports the safety of SRS dose-escalation in this phase 1 trial, which can be further explored in future clinical trials, where it would be very useful to correlate dose with both tissue changes and clinical outcomes.
The effect of SRS/SRT on the immunomodulatory cell populations was an interesting finding of this study. In comparison to primary tumors, we observed a decrease in populations of CD3 +, CD4 +, and CD8 + cells in irradiated and resected brain metastases. Although we did not observe a dependency of these factors on dose or time interval from SRT to surgery, this finding is tempered by the small sample size and short study window. The minimal fractional dose of 7-8 Gy known to induce T-cell infiltration, primarily via CD8 + cells 29 , was exceeded in all but one patient in this series (treated at 6 Gy per fraction). However, very high fractional doses, such as those used in this series exceeding 15-18 Gy in 1 fraction, may actually induce T-regulatory cell activity and downregulate the immunomodulatory effects of radiotherapy 30 . Moreover, although necrosis is commonly associated with inflammation, the finding of lower immunomodulatory cell populations in irradiated metastases may be due to the short interval from SRT to surgery, prior to immunomodulatory cell infiltration 31 . In addition www.nature.com/scientificreports/ to these radiotherapy-related effects, we observed no association between corticosteroid use/dose and changes in immunomodulatory cell populations. As an increasing proportion of patients receive immunotherapeutics for treatment of their metastatic disease, an in-depth analysis of pre-operative radiosurgery specimens to systematically study changes as a result of dose and fractionation may further inform clinical practice, especially if there is any association between change in cell populations and the efficacy of these agents. A few studies in a variety of extracranial disease sites have also evaluated the tissue effects of SRS/SRT. A phase I trial of pre-operative SRT for prostate cancer (25 Gy in 5 fractions), followed by radical prostatectomy two weeks later, demonstrated a lack of change in apoptosis in the resected tissue samples but did demonstrate reduced cell proliferation measured by p21 WAF activation 32 . A phase I trial of pre-operative partial breast irradiation (15, 18, and 21 Gy in 1 fraction), followed by lumpectomy within 10 days, not only revealed gene expression changes in post-irradiated tumor samples compared to pre-irradiation biopsies, but also demonstrated dose-response effects in parameters related to immunity and inflammation 33 . A study of pre-operative SRS (20 Gy in 1 fraction) for breast cancer, followed by breast-conserving surgery three months later, demonstrated a median residual tumor cellularity of only 3% in 8/10 patients 34 . A phase II trial of pre-operative SRT (54-60 Gy in 3-8 fractions) for patients with early-stage NSCLC, followed by lobectomy or sublobar resection, demonstrated a pathologic complete response rate of 60% via H&E staining but did not report other tissue parameters, such as necrosis or apoptosis 35 . Unlike these series, pre-operative SRS/SRT in our study was not associated with changes in other tissue parameters including vessel density, apoptotic factors, or VEGF. However, it is important to note that these other studies compared pre-and post-radiotherapy biopsy specimens of the treated tumors, not irradiated metastases compared to non-irradiated primary tumors. Other factors that may account for these differences include the short window between SRS/SRT and surgery in our series compared to the interval between radiotherapy and surgery in extracranial sites 25,36 , differences in tumor histologies represented, and differences in the response to radiotherapy in intracranial sites compared to extracranial sites given the inherent differences in the tumor microenvironment 37 .
This study has several limitations. First, although eight different primary tumor histologies were represented in this study, the majority of patients (41%) had NSCLC. As tumor microenvironments are related to primary tumor histology and molecular profile, this prevents extrapolation of the results of this study to other histologies not well represented as all histologies are distinct and underrepresented in this limited series. For example, certain histologies, such as renal cell carcinoma, melanoma, and thyroid cancer have innate tumor hemorrhage and necrosis, which would be difficult to differentiate from the effects of radiotherapy 38 . Further, although the majority of our patients had NSCLC, even within this entity, there are several molecular subtypes, and hence one cannot generalize these results to all tumor types and subtypes. Moreover, given that the tumor histology and molecular profile ultimately dictate systemic therapy, the impact of this on the tumor microenvironment could not be evaluated in a study with this small sample size. Second, we compared irradiated brain metastases to nonirradiated primary tumor samples, given the lack of ability to obtain pre-radiotherapy biopsies. Differences in the tissue samples between these sites are known to exist, although these are typically related to genomic drivers 39 or receptor subtypes 40 . Third, although a number of tissue parameters were evaluated, the sensitivity of different methods of analysis, for example differential gene expression via RNA-sequencing 41 , may yield additional insight above that detected via IHC analyses alone. Fourth, the overall sample size of this study is small and therefore limits definitive conclusions regarding the effect of SRS on brain metastases. Further, this limits the power of the study to evaluate the impact of dose, tumor histology, systemic therapy, and timing of SRS on these tissue parameters. Therefore, current studies in which patients are randomized to pre-operative versus post-operative SRS (i.e. NCT03750227 and NCT03741673) and future studies (i.e. recently activated NRG BN012) could provide valuable opportunities to further elucidate this question through collection of both primary tumors and extracranial metastases in addition to intracranial metastases.

Conclusion
In this series, dose-escalated pre-operative SRS/SRT was associated with favorable rates of tumor control, and as compared to the primary non-irradiated tumor, increased tumor necrosis and a reduction in multiple immunomodulatory cell populations. Differences in immunomodulatory factors may be consequential to multiple factors, including corticosteroid use and the immunosuppressive effect of high-dose SRS/SRT. Understanding this complex interplay in a larger sample size is critical for a better understanding of the impact of SRS/SRT in the brain. www.nature.com/scientificreports/