State-of-the-art considerations in small cell lung cancer brain metastases

Background Small cell lung cancer (SCLC) frequently leads to development of brain metastases. These unfortunately continue to be associated with short survival. Substantial advances have been made in our understanding of the underlying biology of disease. This understanding on the background of previously evaluated and currently utilized therapeutic treatments can help guide the next steps in investigations into this disease with the potential to influence future treatments. Design A comprehensive review of the literature covering epidemiology, pathophysiology, imaging characteristics, prognosis, and therapeutic management of SCLC brain metastases was performed. Results SCLC brain metastases continue to have a poor prognosis. Both unique aspects of SCLC brain metastases as well as features seen more universally across other solid tumor brain metastases are discussed. Systemic therapeutic studies and radiotherapeutic approaches are reviewed. Conclusions A clearer understanding of SCLC brain metastases will help lay the framework for studies which will hopefully translate into meaningful therapeutic options for these patients.


INTRODUCTION
Important advances in our understanding of brain metastases are underway [1,2]. While there have been many recent reviews of brain metastases [3][4][5][6] as well as more focused reviews on non-small cell lung cancer (NSCLC) brain metastases [7][8][9] there has been a paucity of contemporary reviews focusing on small cell lung cancer (SCLC) brain metastases. A broad overview of SCLC can be found in a recent review [10]. Within this manuscript we will narrow the focus exclusively to SCLC brain metastases. A comprehensive understanding of this cancer's involvement of the brain will set the stage for the next steps in optimizing its management. A number of patients succumb to SCLC in the brain and we have to define better biology and therapeutics.

Epidemiology
Lung cancer metastases to the brain affect more patients than any other solid tumor metastases in the U.S. Due to the higher incidence of non-small cell lung cancer (NSCLC) when compared to SCLC, it comprises a higher percentage of patients with brain metastases. However, SCLC appears to have a higher propensity for the central nervous system (CNS). In patients with non-metastatic lung cancer the risk of brain metastases in SCLC appears to be double that of NSCLC. Factors which influence this risk are being clarified. The incidence of SCLC brain metastases does not appear to be influenced by race. It does appear to be higher, however, in younger patients (< 60) compared to older [11]. It is unclear if it is influenced by gender, with at least one study noting a significantly higher incidence Review Oncotarget 71224 www.impactjournals.com/oncotarget of development of metachronous brain metastases and a shorter brain metastases free interval in men with limited stage disease. [12] Increasing pathologic stage correlates with higher incidence of brain metastases. Therapeutic management, such as the extent of resection (complete vs. incomplete) of the primary tumor influences this as well [13]. Like all solid tumors, SCLC has the potential to spread to the leptomeninges. The 2 year cumulative incidence of leptomeningeal involvement in SCLC patients is ~10%. Many, but not all, of these patients also have concomitant brain metastases [14]. An older large retrospective study from 1969 to 1980 by the National Cancer Institute demonstrated leptomeningeal involvement in 25% of SCLC patients at 3 years. Leptomeningeal involvement most often occurred in the context of extra-CNS relapse of disease. The greatest risk factors for leptomeningeal involvement appear to be other CNS involvement at diagnosis as well as metastases to other distant sites such as liver and bone [15]. These findings, however, may not reflect contemporary incidence rates.

Prognosis
Prognosis for patients with SCLC remains poor. Earlier SCLC-specific analyses, of patients from trials conducted from 1983 to 2005, using the Radiation Therapy Oncology Group (RTOG) recursive partitioning analysis (RPA) classification system revealed a median overall survival (OS) of 4.9 months (range 0.3-40.3 months) in patients with newly diagnosed SCLC brain metastases. Patients with older age, poorer performance status, uncontrolled primary tumor, and/or extracranial metastases had worse outcomes, correlating with the RTOG RPA classes seen across the aggregated histologies of all solid tumor brain metastases [16].
In the more contemporary disease-specific graded prognostic assessment (GPA) system the same prognostic factors remained valid. Additionally, the number of brain metastases influenced outcomes. Whether this is secondary to the biology of oligo-vs. multi-metastatic disease, the imaging eras (CT vs MRI) when the studies were conducted, the treatments utilized in the management of the two conditions, or a combination of these factors is unknown. Median OS for all patients with SCLC brain metastases was seen to be 4.90 months [17] however ranged from 3.0-14.8 depending on these four validated risk factors (KPS, age, number of brain metastases, and extracranial metastases). Further support is lent from smaller studies focused on SCLC brain metastases where performance status, presence of extracranial metastases, and the number of brain metastases have all been associated with OS [18]. Survival in recurrent SCLC is presumably poorer than in the newly diagnosed setting. There is less data to clearly define prognosis in that specific setting, however. In patients with leptomeningeal involvement, survival is limited with OS reported as 1.3-2.4 months [14]. While improvements in survival have been seen across a range of solid tumor histologies with brain metastases, this has unfortunately not been the case in SCLC.

Pathophysiology
SCLC is a tumor which arises from pulmonary neuroendocrine cells as well as other potential candidate cells such as alveolar type 2 cells [10]. It has the potential to metastasize early and extensively. The brain is a common site for its metastases. This is likely due to the appropriate "seeds" arriving and thriving in an optimal "soil" [19]. Our understanding of this phenomenon has grown since Piaget's description. It is still limited by a number of factors, including the infrequent biopsy or resection of SCLC brain metastases when compared to other histologies and the limited number of autopsies performed on cancer patients in the modern era. It is clear that in other solid tumors there is genetic divergence between the primary site and the brain metastases which has been termed "branched evolution" [20]. It is uncertain at this time if this holds true for SCLC brain metastases but has been seen in multiple other primary cancers. We will first review the neuro-anatomic features of SCLC brain metastases before delving into the genomic and proteomic features.

Neuro-anatomic localization
SCLC is more likely to be associated with multiple brain metastases as opposed to single brain metastases. Mapping the distribution of SCLC in the brain may further expand our understanding of the interaction between metastatic cells with their micro-environment. It will also help optimize anatomically targeted treatment modalities such as radiation therapy [21]. Our initial understanding of the neuroanatomic localization of SCLC brain metastases arose from autopsy studies. One early single center review of 15,000 autopsies performed between 1969 and 1984 concluded that SCLC metastases were equally distributed throughout the brain in contrast to other lung cancer subtypes that favor posterior circulation territories [22]. This study was limited by the slice thickness of 1cm provided by routine autopsies, rendering its resolution similar to early CT imaging studies and admittedly overlooking smaller brain metastases. Utilizing MRI much smaller metastases can be delineated and the population data can be aggregated into anatomic frequency maps [21,23,24]. Using this technique, SCLC was found to favor the cerebellum [21,24]. This disproportionate distribution raises both pathophysiologic questions regarding the potential trophic factors leading to increased aggregation and likely more importantly, factors leading to facilitated growth in these locations as well therapeutic questions regarding differential radiation dosing. When investigating the pattern of growth, both well-demarcated www.impactjournals.com/oncotarget lesions and diffusely infiltrating ones have been seen on autopsy studies. The vascular co-option which is well described in melanoma brain metastases has not been noted in association with SCLC [25].
In addition to the brain, SCLC can also involve other CNS structures. These patterns of spread are less well studied. Spinal cord parenchyma can be a sight of metastasis. It is relatively rare in comparison to brain metastases, occurring in only ~2% of SCLC patients [26]. This appears to occur predominantly in the setting of parenchymal brain metastases [14]. Leptomeningeal spread of SCLC can occur metachronously, synchronously, or independent of the diagnosis of brain metastases. When associated with brain metastases, this is somewhat more frequent with posterior fossa metastases [14]. Leptomeningeal involvement can also lead to direct invasion of both the spinal cord and brain parenchyma [15].

Genomic and gene-expression profile
Studies of human surgical tissue from SCLC brain metastases have demonstrated significant upregulation and downregulation of genes (Table 1). These genes can be broadly categorized as related to angiogenesis, cell-cell adhesion, immune activity, and survival/proliferation/ differentiation pathways. A number of angiogenesis related genes including ANGPT4, PDGFRB, COL4A2, and VEGFA are all upregulated. Some such as ANGPT4 and PDGFRB appear to be uniquely upregulated (50-fold) in SCLC brain metastases when compared to NSCLC brain metastases. Others are profoundly upregulated across numerous histologies pointing towards a potentially more universal role in brain metastases [27]. The nature of these potential roles is yet to be fully defined. Their differential dysregulation between tumor types, however, points towards a histology-specific angiogenic profile within brain metastases. In rat models of SCLC brain metastases inhibition of the sulfonylurea receptor 1 (SUR1) with glyburide has led to decreased intracranial vascular permeability and improvement in cerebral edema [28]. These findings may represent potential avenues for therapeutic investigation.
Cell-cell and cell-extracellular matrix adhesion also appears to be important in SCLC brain metastases with numerous upregulated genes including CEACAM1, PECAM1, HSPG2 and CD44 all upregulated [24]. Additional aberrancies including loss of heterozygosity of the E-cadherin gene CDH1 have been noted as well [29]. Analyses of post-mortem tissue from untreated patients demonstrate SCLC brain metastases have a higher incidence (83%) of expression of the chemokine receptor CX3CR1 when compared to metastases to other organs such as liver (28%), locoregional lymph nodes (14%), and adrenal glands (0%) [30]. Normal lung neuroepithelial cells exhibit an inherent ability to traverse from one location to another via a process dubbed slithering [31]. This can lead to postulation regarding the predisposition of SCLC for early metastasis as well as which chemokines and adhesions molecules play a critical role in SCLC metastasis to the brain. One chemokine of interest, CXCR4, is expressed in SCLC and may play an important role in the mechanism of metastasis to the brain. This is since the brain contains a rich resource of the CXCR4 ligand CXCL12 (also referred to as stromal derived growth factor) [32,33].
As with many other malignancies, the role of tumor, microenvironment, and the immune system holds substantial interest and has sparked hope for potential therapeutic targets. The underpinnings are still incompletely understood. SCLC brain metastases have been associated with a high amount of astroglial reactivity when compared to NSCLC adenocarcinomas and squamous cell brain metastases [34]. It has been demonstrated in mouse models of adenocarcinoma that tumor cell-astrocyte gap junctions composed of connexin 43 serve as conduits for transfer of cGAMP into astrocytes leading to astrocytic production of IFNα and TNFα. These paracrine signals activate growth and chemoresistance signals [35]. It is uncertain if this also hold true in SCLC. Downregulation of the immunosuppresive TGFβ1 gene, but upregulation of the immunosuppressive IFNβ1 gene have been described in SCLC brain metastases. Upregulation of the pro-inflammatory TNF gene has also been noted [26]. It is difficult to parcel out what is cause and what is consequence. The majority (93.8%) of SCLC brain metastases specimens have demonstrated the presence of tumor-infiltrating lymphocytes (TIL). All of the TIL+ samples have shown CD3+ TILs and the majority have also exhibited CD8+ TILs. CD45RO+ memory TILs were seen in about half of the brain metastases. Their presence was associated with improved median OS (11 months vs. 5 months, p = 0.007). Programmed death-1 (PD-1) expression has been seen in only a limited (3.1%) subset of TILs in SCLC brain metastases. However, programmed death ligand-1 (PD-L1) has been detected in a higher percentage of TILs (25%) and tumor infiltrating macrophages (28.1%). PD-L1 expression was also seen in > 5% of tumor cells in over 1/3 of the brain metastases samples [36]. In extra-CNS SCLC increased PD-1 and PD-L1 expression in SCLC tumor cells, possibly mediated via deregulation of KIT and DNA methyltransferase 1 (DNMT1), correlated with cisplatin resistance. It is hypothesized that chronic platinum exposure leads to the upregulation of PD1/PD-L1 in the tumor cells [37]. It is uncertain whether this has an impact on brain metastases. The large number of genetic alterations in SCLC [38] may make it a particularly attractive target for immunotherapies.
SCLC can also arise in the context of previous targeted therapies for non-small cell lung cancers. Sequist, et al. initially described the "transformation" of EGFR mutated adenocarcinomas of the lung upon treatment with tyrosine kinase inhibitors. There was a propensity for these Oncotarget 71226 www.impactjournals.com/oncotarget SCLC to also metastasize. It is possible that there is clonal selection with aggressive phenotype of metastasis [39].

Radiology CT and MRI
Clinical management of SCLC brain metastases begins with their symptomatic or incidental detection with imaging. The clinical role of oncologic imaging is to identify, prognosticate, assist treatment plan formulation, and monitor disease status. Conventionally used techniques yield such information by detecting changes in anatomy. The 3rd edition of the American College of Chest Physicians' Guidelines for SCLC Diagnosis and Treatment recommends contrast-enhanced brain CT or MRI as the mainstay for staging SCLC without differentiating between the two modalities [40,41]. In contrast, the National Comprehensive Cancer Network 2017 Clinical Practice Guidelines express preference for MRI over CT for staging in the brain [42]. Generally, its higher resolution makes MRI superior to CT. According to an early study comparing CT vs. MRI for detection of SCLC BMs, CT missed 85% of metastases in the posterior fossa and 97% of all metastases smaller than 5 mm [43]. Similar conclusions were reached comparing the prevalence of brain metastases in patients with newly diagnosed SCLC in the "CT era" (roughly 1980 to 1991) versus the "MRI era" (1991 studied until 2004) [44,45]. Whereas, the MRI detected prevalence was 24% with almost half of these consisting of asymptomatic patients, CT detected only 10% with a bias for symptomatic lesions. These differences in sensitivity need to be taken into account when evaluating outcome data. For example, the number of brain metastases (single vs. multiple) is a prognostic factor when detected via the more sensitive MRI but is not prognostic when utilizing CT which may under-detect metastases creating the false impression of a single metastasis [44].
While the above-mentioned routine techniques visualize tissue structure, advanced imaging techniques intend to assess tissue properties, such as water or blood flow, or metabolism. Diffusion-weighted imaging of the brain, more specifically apparent diffusion coefficient (ADC) mapping has been well studied in SCLC. Low ADC values are generally associated with high cellularity and anaplasticity in tumors, features closely associated with SCLC brain metastases [46,47]. While not all studies could distinguish primary tumor types based on MRI ADC values [48], SCLC appears to have the lowest mean and median ADC differentiating it from melanoma, breast cancer, and  [49]. There may also be a higher intrapatient ADC variability in SCLC brain metastases compared to NSCLC. A comprehensive analysis by Zakaria et al. corroborated the low ADC in SCLC brain metastases, although this feature was shared with melanoma in their analysis [47]. Perfusion-weighted MRI (PWI) and MR spectroscopy (MRS), useful techniques for monitoring progression and aiding differentiation of radiation necrosis vs. tumor recurrence have not been investigated in depth in SCLC.

Nuclear imaging
Positron emission tomography (PET) uses radiotracers to visualize metabolic activity in vivo. The 2-deoxy-2[ 18 F] fluoro-D-glucose (FDG) is the only PET radiotracer at the present that is widely used in a clinical setting and is one of the mainstays for the extracranial staging of SCLC [40,41]. However, FDG-PET can only detect 45% of SCLC brain metastases seen on MRI or even CT, and is incapable of revealing most lesions not seen on these standard modalities [50,51]. More so than the low spatial resolution, the cause of this low sensitivity seen with FDG-PET is high physiologic glucose metabolism in the brain masking metastases that may be hypometabolic relative to the cortex or basal ganglia. This is particularly unfavorable for SCLC where 2 out of 3 brain metastases may go undetected, whereas this proportion is only 20% in NSCLC [52].
Other radiotracers have also been evaluated for assessment of brain metastases. One such tracer is radioactive octreotide, intended to harness the neuroendocrine properties of SCLC. While some brain metastases not evident on MRI were detected via the octreotide scans, its low sensitivity did not validate it as a clinical tracer [53,54]. Mono amino acid radiotracers such as L- methyl-L-tryptophan (AMT) may provide superior contrast to FDG given their lower physiologic uptake in the brain. They have proven useful in the imaging of primary brain tumors [55][56][57]. Two MET-PET studies with a substantial proportion of lung cancer (> 70% in each) demonstrated the ability to distinguish recurrent metastatic brain tumors from radiation necrosis with a sensitivity and specificity of 75% or above, yet included only two SCLC brain metastases whereas the other did not even specify such [58,59]. Finally, utilizing kinetic uptake analysis on AMT-PET, a normalized net tryptophan transport could distinguish brain metastases from lung versus breast cancer with an accuracy of 88%, whereas tracer accumulation enabled distinction from glioblastoma with 81% accuracy. However, low sample size did not allow testing of differences between NSCLC and SCLC here either [60].

Surgery
Due to the frequent multi-metastatic picture in SCLC surgery does not typically have a role in the management of this disease. In the three randomized trials evaluating the role of surgical resection of single brain metastases patients with SCLC were excluded, limiting our understanding of the role of surgical management in this patient population [61][62][63]. To our knowledge there are no studies specifically evaluating the role of craniotomy and surgical resection of SCLC brain metastases. Due to the propensity of multiple brain metastases and the relative radiosensitivity of these tumors the likelihood of a comprehensive study evaluating surgical resection of SCLC brain metastases is unlikely in the foreseeable future.

Radiotherapy
Because brain metastases are a frequent problem in patients with SCLC [64], and the burden of brain metastases can impact on quality and length of survival, several prospective trials have examined the use of prophylactic cranial irradiation (PCI) in SCLC patients who present without brain metastases. This is performed for limited disease and the most recent data for extensive disease does not favor PCI. These trials have consistently shown that PCI leads to a reduction in the incidence of brain metastases and a prolongation in survival in patients with limited disease with initial response [65,66]. In spite of the compelling evidence supporting the use of PCI, a recent study demonstrated that 40% of SCLC patients for whom a survival benefit from PCI has been shown, do not receive PCI due to concerns of cognitive toxicity on the part of both patient and physician [67]. A recent trial of PCI for limited-stage SCLC demonstrated a 62% (95% confidence interval 50-74%) rate of cognitive toxicity [68]. In another study, PCI was associated with a higher rate of decline in patient-reported cognitive function [69]. Observations of cognitive toxicity from PCI for SCLC appear similar to those seen after PCI for locally advanced non-small cell lung cancer and after whole-brain radiotherapy for brain metastases [70][71][72], and demonstrate a differential sensitivity of memoryrelated cognitive domains to cranial irradiation. Building upon extensive preclinical and clinical data supporting the memory-specificity and radiosensitivity of the hippocampal neural stem cell compartment [73,74], an ongoing phase III trial (NRG CC003) seeks to determine whether conformal avoidance of the hippocampal dentate gyrus using intensity-modulated radiotherapy during PCI can prevent cognitive toxicity, while still providing the intracranial control benefit of PCI [75].
In SCLC patients who present with brain metastases but have never been treated with PCI, therapeutic wholebrain radiotherapy is the standard of care. Similar to absent role for surgery in SCLC brain metastases, given their multi-metastatic propensity, radiosurgery is contraindicated for PCI-naïve SCLC patients presenting with brain metastases. However, radiosurgery can be an www.impactjournals.com/oncotarget effective therapeutic option for recurrent or progressive brain metastases after prior PCI or WBRT and can help obviate the need as associated neurotoxicity of repeat WBRT [76].

Systemic therapies
While radiation therapy is the cornerstone of the treatment of SCLC brain metastases, the role of systemic therapies has long been under investigation ( Table 2). While at this time there is no data to support the routine use of systemic therapies in newly diagnosed SCLC brain metastases [77,78], this therapeutic approach has the potential to change our management paradigm in the not-to-distant future. A number of studies over the past few decades have evaluated systemic therapies in patients with both newly diagnosed and recurrent SCLC brain metastases [79][80][81][82][83][84][85][86][87][88][89][90][91]. All of the studies have utilized traditional cytotoxic chemotherapies and not targeted therapies as has been seen with other solid tumor brain metastases. These trials looked at chemotherapy alone or in conjunction with WBRT. Some studies have also included histologies other than SCLC. Direct comparison between these studies is difficult for a number of reasons which include the different eras in which they were conducted, different endpoints, and different study designs. To complicate the picture further numerous trials of systemic therapies for extensive-stage SCLC allowed for the presence of brain metastases, but not all have evaluated CNS-specific endpoints.
The agents which have been investigated in this setting all have some degree of activity in extra-CNS SCLC. They fall into the broad categories of topoisomerase inhibitors (etoposide, teniposide, doxorubicin, topotecan, irinotecan), platinum agents (cisplatin, carboplatin), vinca alkaloids (vincristine, vinorelbine), and alkylating agents (cyclophosphamide, temozolomide). The role of immunotherapy (such as pembrolizumab or nivolumab with ipilumimab) is beginning to emerge for SCLC [92,93]. However, immunotherapy has not been systematically studied in SCLC CNS metastases. CNS responses have been seen across a range of treatment regimens, supporting potential efficacy in the target organ with no clear signal for superiority of one regimen over another. However, despite CNS responses in a substantial number of patients, OS remains dismal across numerous studies.
The radiographic responses to systemic therapies are, however, heartening. They force us to push against Oncotarget 71229 www.impactjournals.com/oncotarget the dogma that most systemically delivered agents are excluded from the nervous system. This needs to be considered on both a histology-specific and therapyspecific basis. Our current understanding of the ability of specific therapeutic agents to cross into the healthy and the diseased CNS is limited. This is in large part due to the practical limitation of performing pharmacokinetic (PK) studies on CNS tissue in living humans.

CONCLUSIONS
The suboptimal outcomes which continue to be seen in patients with SCLC brain metastases warrant the need for further investigation. The robust radiographic response rates provide clear evidence for the biologic activity of our current treatment modalities. Their limited effect, however, on improving survival support the need for additional advances. Ongoing efforts to limit the toxicity of radiotherapy may prove beneficial in this patient population as well as in other solid tumors. The ability of systemically administered treatments, including those traditionally thought to have limited CNS penetration, holds out even greater hope. Conventional cytotoxic chemotherapy, targeted therapies/immunotherapies, and/ or a combination of these have the potential to improve OS in this recalcitrant disease.
While there currently are no reliable prognostic or predictive molecular biomarkers in SCLC as our understanding of the disease evolves this will hopefully change. The study of SCLC brain metastases will continue to be limited by the paucity of tissue samples for this type of research. However, it should be noted that the neuropathologic work as well as imaging and therapeutic studies described above are beginning to shed light on how we may best address the problem of SCLC brain metastases. Refining our existing therapeutic modalities such as WBRT to limit their toxicity will be an impactful advance. In addition to the promise of hippocampal avoidance, further tailoring of the radiation fields informed by neuroimaging studies could improve efficacy while decreasing toxicity. Systemic therapies actively targeting SCLC brain metastases are also of interest. Potential targets include components of the angiogenic pathway such as SUR1 or a host of potential immunotherapeutic targets. Parceling out the answers to CNS-specific questions from large therapeutic trials which include patients with brain metastases may provide us with important insights and adequate safety and efficacy signals to justify moving forward with brain metastasesspecific trials. Finally, the prevention of brain metastases in the SCLC patient population would be an important advance. The targeting of chemokines and adhesion molecules may play a role in achieving this goal. While improvements in survival for patients with SCLC brain metastases have been limited, the groundwork for important advances is present.

CONFLICTS OF INTEREST
Dr. Lukas has served on advisory boards for Abbvie, AstraZeneca, Novocure, and Ziopharm and has received honoraria for an internal lecture/presentation for Abbvie and support for meeting travel from Roche. Dr Gondi has served on advisory board for INSYS Biotherapeutics and received speakers honoraria from Novocure. Dr. Kumthekar has received honoraria for serving as a consultant for Abbvie and Angiochem. All other authors have nothing to disclose.