The Management and Biology of Metastatic Cancers to the Brain

Introduction

Brain metastases are a serious consequence of many solid tumors, which can result in significant morbidity and mortality. Approximately 10–20% of cancer patients develop brain metastases, and while the actual incidence of brain metastases is somewhat difficult to quantify, it has been estimated that there were between 21,000 and 43,000 cases of brain metastases in the United States in 2010 ( ). The most common primary tumors that metastasize to the brain, in order of incidence, are non-small cell lung cancer (NSCLC), small cell lung cancer (SCLC), melanoma, breast cancer, and kidney cancer. Less commonly, gastrointestinal and prostate cancers can spread to the brain as well. In a study published in 2004, the prevalence of brain metastases for all primary sites combined was 9.6%; the greatest of them being lung cancer (19.9%), followed by melanoma (6.9%), renal cancer (6.5%), breast cancer (5.1%), and colorectal cancer (1.8%) ( ). The majority of brain metastases develop through a hematogenous route from the primary tumor site via the arterial circulation. Probably due to the progressive narrowing of the blood vessels that leads to the trapping of embolic tumor cells, the most common sites of metastases to the brain are the gray-white junction and terminal watershed areas at zones between major intracranial arteries ( ). Approximately two out of three of these brain lesions are symptomatic, and can pose significant problems in terms of management. In this chapter, we will review the function of the blood–brain-barrier (BBB) under normal conditions, and then examine how the BBB becomes disrupted because of brain metastases. Finally, we will analyze some of the potential treatment paradigms for brain metastases, with additional discussions of two seminal studies; each supporting the hypothesis that chemotherapeutic treatment of brain metastases can be based upon the characteristics of the primary tumor.

In most cases, brain metastases are detected after the primary tumor has been diagnosed; uncommonly, a patient may present with symptoms from a brain metastasis first and upon work-up for the brain lesion, the site of origin is revealed. Brain metastases are frequently symptomatic, but can vary in the manner in which they present. Common manifestations include headaches, focal weakness or paralysis, visual field disturbance, cognitive disturbance, altered mental status, ataxia, and seizures. Seizure is the presenting symptom in about 10% of all cases of brain metastases, and is more common in patients with multiple metastases or when melanoma is the primary tumor. Presentation with neurological symptoms may also be due to hemorrhage, particularly in the case of metastases from a primary renal cell carcinoma or melanoma.

The prognosis for patients with brain metastases is generally poor, with the median overall survival (OS) in the range of 3–9 months. Two well-established prognostic assessments for these patients are currently available; they are the Recursive Partitioning Analysis (RPA) by , and the Graded Prognostic Assessment (GPA) from . The RPA divides patients into three stratified classes for the purpose of sorting patients into clinical trials by predicted outcome so as to compare patients who have similar disease. The classes are based on the following three positive outcome predictive factors: Age <65, Karnofsky Performance Score (KPS) 70 or higher, and controlled primary tumor with no evidence of extracranial metastases. Median OS in patients in Class 1 was 7.1 months, while patients in Class 3 (KPS <70) had an OS of 2.3 months ( ). The GPA system is another prognostic index that separates patients into four groups with assigned scores based on age, KPS, known extracranial metastases, and number of metastases ( ).

BBB Structure

The BBB is a highly specialized structure created by the interaction between astrocytes and vascular endothelial cells. The foot processes of the astrocytes surround over 90% of small blood vessels, forming tight junctions with minimal fenestrations and minimal pinocytotic vesicle activity. Fortifying the BBB is a collagen-rich basal membrane that surrounds the brain capillary endothelial cells. The separation of the blood and brain is further enhanced by the absence of the lymphatic system from the central nervous system (CNS) preventing macromolecules from entering the brain by passive transport, thereby creating an immunologically distinct space. There is also significant regulation of the BBB microenvironment by active transport modulators (active efflux pumps), such as P-glycoprotein (P-gp) and multidrug resistance-associated proteins (MRP), as shown in Fig. 2.1 , which will be discussed later. It has been shown by Kroll and colleagues that an intact BBB prevents passage of ionized, water-soluble molecules of >150 Daltons, and repels negatively charged ionic compounds due to its negatively charged luminal surface ( ). Most natural product chemotherapy agents (NPCAs) are charged, have molecular weights from 600 to 1400 Daltons, and account for about half of all traditional chemotherapies ( ). There are also chemotherapies that are much larger molecules, as they are bound to albumin, which is approximately 66,000 Daltons and are negatively charged. The great majority of chemotherapy drugs are >150 Daltons, charged and bound to albumin in the serum (50–90% of the drug is bound to albumin). Notably, topotecan and irinotecan (topoisomerase I inhibitors) are only about 30–35% bound to albumin, leaving a quantity of free drug to cross the BBB.

These free, uncharged camptothecin-derived drugs may partially explain why the topoisomerase I inhibitors, such as topotecan and irinotecan, are useful in gliomas and metastatic tumors to the brain. Certain antimetabolites, including 5-Fluorouracil (5-FU), cytarabine (Ara-C), and gemcitabine, are able to cross the BBB because they are not charged and are relatively small (less than 200 Daltons). Because gadolinium ions are 500–950 Daltons, gadolinium-enhanced Magnetic Resonance Imaging (MRI) is a potential way to evaluate for BBB interruption ( ). However, a limitation to this method of determining the integrity of the BBB is that it only takes into account molecular size, not charge. Further, have shown that the amount of “leakage” detected on MRI may not directly correlate with the amount of drug that actually reaches the brain parenchyma.

The BBB in primary CNS cancers (gliomas, etc.) will only be briefly mentioned to serve as a comparison to the BBB in the case of metastases to the brain, as the focus of this chapter is on metastatic lesions to the brain. It is important to note that it has been shown that the BBB is still relatively intact in primary CNS solid tumors, certainly more so than in metastatic brain tumors. This may account for somewhat limited response rates of primary CNS cancers to many different types of chemotherapy, especially if the drug is large (>150 Daltons) and charged, and may explain why synthetic antimetabolite drugs such as Ara-C, methotrexate, 6-MP, and gemcitabine are efficacious in regimens for CNS lymphoma, because they are small and uncharged. In addition, primary CNS lymphomas have high Ki-67’s, which are indicative of rapid growth rates. In general, the class of antimetabolites mentioned in the previous sentence is active mainly in G ₁ S, these drugs are nonpolar, and have molecular weights <150 Daltons. These characteristics make them advantageous to use in rapidly growing CNS lymphomas.

Molecular Changes in the BBB with Metastatic Lesions

In , the development of metastases in the brain was described by the “seed and soil” hypothesis first described by Paget; he postulated that tumor cells (the “seed”) have a specific predilection for certain organs (the “soil”). In a multistep process, tumor cells interact with the brain microenvironment through the secretion of a variety of molecules implicated in cell invasion, such as matrix metalloproteinases (MMP), heparanase (HPSE) and hyaluronidase, in order to cross the endothelium by inducing breakdown of the extracellular matrix ( ). Heparan sulfate proteoglycans are a major constituent of the endothelial cell layer that lines the luminal surface of blood vessels. HPSE is an enzyme that cleaves polymeric heparin sulfate molecules into shorter oligosaccharide chains; therefore, it is an important molecule in the process of penetration through the endothelial basement membrane. It has been shown that HPSE also releases angiogenic factors in the tumor microenvironment and thus can facilitate tumor cell invasion, vascularization, and decreased survival ( ). Similarly, have shown that levels of hyaluronidase are higher in human brain metastases than in their primary extracranial tumors, indicating that this might be an important aspect of the brain metastatic phenotype, in which the BBB in metastatic brain tumors is less intact than it is in primary gliomas. In the early stages of deposition of brain metastases, it appears that the BBB is still somewhat intact. In mouse xenografts, an intact BBB was seen with brain metastases up to 2 mm in diameter ( ). Over time, however, disruption of the foot processes occurs, leading to the breakdown of astrocyte and endothelial cell binding, which increases permeability of the BBB. Animal studies with sodium fluorescein, a marker of BBB permeability, have shown more physical disruption of the interactions between astrocytes and the BBB endothelial cells associated with metastatic lesions than with primary brain tumors. Thus, the BBB in metastatic brain tumors is more disrupted and less organized than the BBB in primary gliomas, which is an important factor when deciding which chemotherapy drug to use therapeutically in that particular patient.

Hypoxia also plays an important role in the pathogenesis of metastatic lesions to the brain. have shown that approximately 70% of metastatic brain tumors that are greater than 0.5 mm in diameter have central necrosis and an ischemic environment. Ischemia leads to the breakdown of the BBB by increasing endothelial pinocytosis, opening tight junctions, and damaging endothelial cells ( ). Hypoxia-inducible factor 1 (HIF-1) is upregulated under hypoxic conditions, and HIF-1 induces overexpression of vascular endothelial growth factor (VEGF) in tumors ( ). The presence of VEGF, which stimulates neoangiogenesis, also makes blood vessels more permeable, and demonstrated in breast cancer cells that VEGF modulates transendothelial migration and brain endothelial cell permeability. In order to induce angiogenesis, VEGF must create interendothelial cell gaps and fenestrations in the neovascular endothelium, causing subsequent degeneration of the plasma membrane and loss of the integrity of the BBB. Furthermore, newly created vessels (neoangiogenesis) lack the properties of those normal, nontumor related blood vessel endothelial cells in the same area of the brain and can lead to increased leakage through the BBB ( ). Thus, blood vessels formed in the environment of metastatic brain tumors >2 mm are “leaky” and do not have an intact BBB as in normal brain blood vessels. It has been posited that blocking early VEGF expression may limit BBB breakdown; there are known agents that block VEGF expression, such as bevacizumab, the anti-VEGF antibody, that are in clinical use today.

Currently, the reason for the predilection of some tumors to metastasize to the brain more than others is not completely understood. One theory is that the brain and the organs that form types of cancer that easily metastasize to the brain have the same embryonic neuroectodermal origin; for instance, SCLC and melanoma are derived from tissues that are of embryonic neuroectodermal origin, and they have a high propensity to metastasize to the brain. Another theory has been reported by , and posited that the neovasculature in the metastatic brain lesion expresses characteristics of the blood vessels of the primary extracranial tumor. Interestingly, the fenestrations seen in the blood vessel endothelium in renal cell carcinoma metastases to the brain are similar to those seen in the primary tumor. One can hypothesize that the tumor cells secrete or induce molecules that alter the architecture or structural character of the BBB so that it is similar to its tumor of origin.

The Role of P-Glycoprotein

P-gp (also known as ABCB1) is a product of the MDR1 gene and it is an adenosine triphosphate (ATP)-dependent efflux pump in many organs/tissues that detoxifies the blood and it is located on the luminal surface of BBB endothelial vessel cells and in organs that are vital to detoxification ( Fig. 2.1 ). It plays an important role in the detoxification process of xenotoxins and is expressed in tissues such as the liver, kidney, placenta, colon, BBB, and small intestine. In the capillary endothelial cells of the CNS-BBB, this protein is highly expressed and pumps out many xenotoxins (by hydrolysis of ATP) from the endothelial cells into the capillaries and peripheral circulation. Notably, xenotoxins comprise the great majority of the NPCAs in use today, and they are derived from plants, bacteria, protozoa, and fungi. Half of our chemotherapy drugs are derived from natural sources: from land and marine-based organisms. In general, the majority of NPCAs are not useful in primary brain tumors because they cannot cross the BBB, even if it is only partially intact. Exceptions to this are the topoisomerase I/II inhibitors, such as irinotecan (CPT-11) and etoposide (VP-16), which do cross over the BBB, probably because they are not highly bound to albumin (about 70% is free drug). This also explains why small lipophilic (nonpolar) synthetic drugs, such as the class of lipophilic alkylators/methylators (temozolomide (TMZ), CCNU, BCNU, Thio-TEPA) are useful in primary brain tumors where the BBB is partially intact and the P-gp expression is high. Lipophilic alkylators/methylators, like TMZ, can cross the BBB into primary brain tumors at 40% of their plasma concentration. It has been shown in murine knockout models that deletion of P-gp leads to a two- to threefold higher concentration of NPCAs in the brain parenchyma compared with wild-type P-gp mice ( ). In addition, demonstrated that P-gp expression in the neovasculature of gliomas is different than that of metastatic brain tumors. In a study of adult surgical brain tumor specimens (29 gliomas and 6 metastatic brain lesions), the endothelial cells of newly formed capillaries in nearly 90% of primary gliomas stained highly positive, by immunohistochemical staining (IHC), for P-gp, whereas only 50% of metastatic tumors stained highly positive for P-gp. The metastatic brain tumors with high levels of P-gp by IHC were tumors which intrinsically have high P-gp levels in their primary tumors and organs such as colon, kidney, liver, and pancreas. In addition, the expression of P-gp in the neovasculature of the metastatic brain tumor was similar to the P-gp levels found in the primary, extracranial, tumor vasculature. For instance, primary melanomas (which usually have low P-gp levels) and their related brain metastases had similar levels of P-gp in their neovasculature; the level was much lower than would be expected in the normal BBB. Alternatively, the P-gp level in renal cell or hepatocellular carcinoma metastases (normally high P-gp levels) to the brain was similar to that found in the primary tumor organs. Given the finding that the level of P-gp expression is lower or higher in the neovasculature of brain metastases, based upon the expression of P-gp in the primary tumors, one can hypothesize that this can significantly affect the ability of P-gp to block entry of NPCAs into the brain. Of note, this effect is caused more from the expression of P-gp in the neovascular endothelial cells, rather than the expression in the tumor itself. P-gp is one of the primary efflux pumps not only for the great majority of NPCA’s, but for many targeted agents as well including axitinib, dasatinib, gefitinib, erlotinib, imatinib, sorafenib, sunitinib, everolimus, and lapatinib ( ). This finding is unexpected since many of the targeted agents are of lower molecular weights and some are uncharged.

There are several known P-gp inhibitors. Notable drugs that are used in other clinical situations that also inhibit P-gp include tamoxifen, verapamil, cyclosporine A, and the cardiac antiarrythmic/antihypertensive group called dihydropyridines. Elacridar is an inhibitor of both P-gp and ABCG2 (breast cancer resistance protein), and it was shown by Lagas and colleagues that this agent is almost as effective in increasing levels of sorafenib in the brain as a double knockout of Abcb1;Abcg2 (P-gp) ( ). This study is important in that it shows the relative importance of ABCG2 transporter in addition to P-gp alone in pumping out many useful drugs and xenotoxins from the vascular endothelial brain cells. Likewise, found that Elacridar increases brain to plasma ratios of gefitinib in mice despite the presence of intact tight junctions, indicating that gefitinib distribution to the brain is restricted due to active efflux by these pumps. More recently, CBT-1, which is well tolerated and is administered by mouth, has been shown to inhibit P-gp, though more prospective trials are needed to confirm its therapeutic efficacy ( ). Tamoxifen penetrates well into brain metastases and therapeutic concentrations can be achieved in both brain tumors and serum ( ). However, when treated with tamoxifen alone, breast cancer patients who have disease that is only metastatic to the brain rarely achieve long-term remissions of their brain metastases without chemotherapy ( ). Based on the theory that tamoxifen might inhibit P-gp, we carried out a laboratory/randomized clinical study of paclitaxel (a NPCA drug) and high dose tamoxifen to assess the ability of tamoxifen to increase paclitaxel intratumoral accumulation in metastatic brain tumors and primary gliomas ( ). We tested 27 patients who were aged 18–80; they all had recurrence of their histologically documented primary brain glioma, or an initial metastatic brain lesion and their neurosurgeons deemed surgical resection as the next step for treatment. There were two arms in this randomized study; all patients were given 175 mg/m ² paclitaxel infused within a 4 h period just prior to their neurosurgical procedure, or they received the same dose of paclitaxel plus 5 days of tamoxifen (a loading dose followed by 160 mg/m ² twice a day, given orally) starting 5 days prior to the neurosurgical procedure. It was considered safe for these patients to delay their surgery by 5 days, as deemed by the neurosurgeons, and they signed informed consent for this IRB-approved protocol. We analyzed plasma and tissue samples from these patients for paclitaxel levels by high performance liquid chromatography (HPLC). Tumor tissue was resected from the center and periphery of the brain tumor (metastatic or primary glioma) and at the interface with normal brain tissue. In this study the metastatic brain tumors originated from NSCLC, SCLC, and renal cell carcinomas ( ).

The 5 days of high dose tamoxifen was well tolerated, without any toxicity above grade 2, and there were no complications (i.e., healing of surgical scar) from the interventions. Pharmacokinetic (PK) studies for serum levels of paclitaxel and tamoxifen showed no difference in the levels of paclitaxel in the two groups. There was also no difference in serum/blood tamoxifen levels between metastatic brain tumor and primary glioma patients. After performing a multivariate, linear regression analysis, adjusting for tumor type and other factors, we found that median intracellular paclitaxel concentrations increased significantly from tamoxifen. Paclitaxel concentrations were 3.70 and 2.46 higher in the periphery and tumor-brain interface of metastatic brain lesions, respectively, than in gliomas ( P = 0.01) ( Fig. 2.2 ) ( ). The median paclitaxel concentrations were 2.68-fold higher in the center of the tumor ( P = 0.03) ( Fig. 2.2 ). However, the important effect of tamoxifen was that the paclitaxel concentrations were 3.70 and 2.46-fold higher in the periphery and tumor-normal brain interface of metastatic brain lesions as compared with primary brain tumors. These examinations were derived using HPLC analysis of tissues from the center and periphery of patient’s brain tumors. The area of the tumor that is growing in both metastatic and primary brain tumors is the tumor periphery and tumor-normal brain interface. These are the important areas to assess because the tumor center is mainly necrotic tissue and is not actively growing. Thus, tamoxifen was able to significantly increase paclitaxel traversement into metastatic brain tumors more so than primary gliomas and it occurred in the growing periphery (3.70 fold) and normal brain-tumor interface (2.46 fold). This study also showed that baseline penetration and accumulation of paclitaxel alone was higher in metastatic brain tumors as compared to gliomas, which have more intact BBB-Pgp ( ).

We also examined the effects of high dose tamoxifen upon the concentration of paclitaxel in the cerebrospinal fluid (CSF) of brain tumor patients. In a prospective, IRB-approved study with informed consent, we randomized 10 patients with either primary or metastatic brain tumors to paclitaxel alone (175 mg/m2/IV) or a course of tamoxifen (loading dose followed by 160 mg/m2 PO BID on days 1–5 prior to surgery for their tumor and 3 h just before surgery) followed by paclitaxel (175 mg/m2/IV). We obtained CSF and plasma levels of paclitaxel and performed PK studies for tamoxifen and paclitaxel by HPLC ( ). Unexpectedly, we found a trend toward lower paclitaxel concentrations in CSF when paclitaxel was given together with tamoxifen. For those who received paclitaxel alone, we detected a 2.4-fold greater mean CSF paclitaxel concentration and a 3.7-fold higher median CSF to plasma paclitaxel ratio for those who received paclitaxel alone, as compared to those who received both drugs ( P = 0.02). This finding was consistent with the reported finding that P-gp in the endothelial cells of the choroid plexus pumps NPCA’s in an opposite direction and concentrates drugs inside the CSF and out of the brain ( ). Therefore, it is surmised that agents that inhibit P-gp (such as tamoxifen), may increase efflux of NPCA out of the blood-CSF barrier into the peripheral circulation and, paradoxically, lower the CSF concentrations of these agents. This finding was unexpected, but it implies that P-gp in the brain effluxes NPCA out of the tumor and brain into the CSF where it is partitioned into the bloodstream away from the brain and tumor metastases. The choroid plexus is very high in P-gp content and here the NPCA’s are effluxed into the bloodstream in order to protect the tumor and brain. Adding tamoxifen to paclitaxel, in this situation, causes more NPCA to be effluxed out of the CSF and into the peripheral blood circulation resulting in less paclitaxel in the CSF ( ). The function of P-gp in the choroid plexus is to protect the CSF-BBB-brain from xenotoxins like paclitaxel. Thus, use of P-gp inhibitors for treated carcinomatous meningitis may actually be hindered and paradoxically decrease the efficacy and concentration of the NPCA anticancer drug in the CSF circulation.