Background
Leptomeningeal metastasis (LM) resulting in Neoplastic Meningitis (NM) is a spread of malignant cells to the leptomeninges, the subarachnoid space and dissemination of tumor cells within the cerebrospinal fluid (CSF). NM is a complication of patients with progressive cancer (70%) and prognosis is poor once manifest NM with neurological deficits develops [
1]. Patients with breast cancer, lung cancer, malignant melanoma and hematopoietic neoplasms are affected most frequently [
1,
2].
Recently, the combination of neurological examination, radiographic imaging of the neuroaxis and CSF cytology are used to diagnose NM [
3,
4]. However, all these techniques suffer from limitations and may only detect NM at an advanced stage of malignant CSF invasion. CSF cytology has the highest specificity in diagnosis and detects tumor cells in the CSF in >95%, however the sensitivity of this method is generally less than 50% [
3]. Depending on the quality of the cytospin preparation and the expertise of the clinical professional, the diagnostic accuracy of the CSF analysis is variable [
2,
4]. Early diagnosis of NM as well as therapy monitoring is crucial for the patient outcome and an important prerequisite for disease control, since NM of solid tumors without therapy progresses to death within 4–6 weeks accompanied by severe neurological symptoms [
5,
6]. To overcome the lack of sensitivity, research has focused on biomarkers and characterization of tumor cells in the CSF [
7‐
10]. Proteomic approaches attempted to relate differential CSF protein profiles obtained from mass spectrometry-based studies to neoplastic diseases to identify potential biomarkers [
11,
12]. Nevertheless, it is difficult to attribute disease identification and progression to a single protein. CSF profiling of the biological activity of multiple biomarkers may more accurately reflect the pathophysiological processes than one biomarker, increasing specificity and sensitivity [
13].
The metastatic invasion of tumor cells and dissemination to secondary sites is promoted by deregulated activity of extracellular metalloproteases (MPs), such as matrix metalloproteases (MMPs) and A Disintegrin and Metalloproteases (ADAMs). Both classes of proteases are members the metzincin superfamily, meaning that they utilize a complexed zinc ion in their catalytic site to constitute proteolytic activity. Major roles for MMPs and ADAMs in cancer progression have been reported, in particular, tumor cell invasion, angiogenesis and metastasis [
14,
15]. Cancer promoting protease function is part of a tightly regulated multidirectional network of numerous proteolytic enzymes and their physiological inhibitors, which is modulated context-dependent and involves proteolytic interactions of tumor, stromal and infiltrating immune cells [
16]. The biological activities of MMPs and ADAMs in the context of neoplastic meningitis remain less clear but likely given the dissemination of tumor cells through the Virchow–Robin space into the CSF as outlined (Additional file
1: Fig. S1). The occurrence of extracellular proteases in the CSF has been reported under various pathological conditions of the brain including infection [
17‐
21], inflammation [
22‐
26], neurodegeneration [
27‐
32], ischemia [
33,
34] and neoplasm [
35,
36]. Proteolytic enzymes involved in the disturbance of tissue integrity and the penetration of tumor cells into the subarachnoid space could serve as diagnostic markers for NM [
37,
38]. Thus, the analysis of specific patterns of protease activation in the CSF may reflect whether a condition is benign or malignant, and may be a suitable approach for specific and sensitive detection of NM.
In order to obtain metalloprotease profiles for patients with NM in a pilot study, we utilized a multiplex approach for simultaneous detection of MMP and ADAM activities in the CSF of patients with NM compared to control individuals and tumor patients without NM. The Proteolytic Activity Matrix Analysis (PrAMA) is a combined experimental and mathematical method based on time-lapse fluorescence measurements of a panel of moderately specific FRET-based polypeptides in small volumes of biological samples. The observed cleavage patterns are compared to a standard table of catalytic efficiencies measured from purified mixtures of recombinant metalloproteinases and should reflect changes in specific enzyme activities [
13,
39] in CSF samples from patients affected by NM.
Methods
Recruitment of patient cohorts and collection of cerebrospinal fluid (CSF)
This pilot study obtained ethical approval from the local Ethics Committee (Marburg University; File No. 101/15). Informed consent was provided from patients to use their biological specimens and clinicopathological data for research purposes. No selection criteria were applied, i.e. all available patients with suspected diagnosis of NM and with brain metastases undergoing routine lumbar puncture were recruited into the study. CSF samples in the control group were collected during therapeutic lumbar punctures from patients with normal pressure hydrocephalus (3/12) or idiopathic intracranial hypertension (9/12). Patients with normal CSF cytology and blood–brain-barrier (BBB) physiology without clinical evidence for neuro-infectious, neuro-inflammatory, neuro-degenerative or neoplastic brain pathologies were considered as controls. All CSF samples were collected at the Department of Neurology, Philipps University Marburg, Germany. Lumbar puncture to collect CSF was performed by medical staff according to clinical guidelines. Upon collection, CSF served primarily for diagnostic purposes, remaining CSF was placed on ice and centrifuged at 1000g to remove cells. Clarified CSF was aliquoted, snap-frozen in liquid nitrogen and stored at −80 °C for further analysis. Clinicopathological features were documented pseudonymized, patients were grouped in control individuals (crtl), patients with neoplastic meningitis (NM) and patients with brain metastases without neoplastic meningitis (w/o NM) according to neurological examination, contrast-enhanced MRI of the brain and neuroaxis and CSF cytology for clinical diagnosis.
Determination of MMP/ADAM activities in CSF
Cell-free cerebrospinal fluids were tested for MMP/ADAM activity by using the Proteolytic Activity Matrix Analysis (PrAMA) technique developed by Miller et al. using FRET-based polypeptide substrates PEPDab005, PEPDab010, PEPDab008, PEPDab013 and PEPDab014 (BioZyme Inc, Apex, NC), which vary in their specificities towards different ADAM family members and MMPs. PrAMA analysis was performed as described earlier [
13,
39]. Briefly, for time-lapse fluorimetry, a final substrate concentration of 10 μM (diluted from 5 mM stock in DMSO) in 50 μl of activity buffer (1 μM ZnCl
2, 20 mM Tris–HCl pH 8.0, 10 mM CaCl
2, 150 mM NaCl, 6 × 10
−4% Brij-35) was incubated with 50 μl of CSF using 96-well microtiter white opaque plates, each sample was run in technical triplicates. Samples containing sufficient volumes were included for inhibitor studies and repetitive measurements. To some samples, the broad-range MMP/ADAM inhibitor batimastat (Tocris Bioscience, Bio-Techne, Wiesbaden, Germany) was added at a concentration of 1 μM dissolved in DMSO. Fluorescence units versus time were monitored with a Fluostar BMG Optima using excitation and emission wavelengths of 485 and 530 nM at 37 °C, respectively. A non-linear model was used for curve fitting as described previously [
13], the signal of a negative control was subtracted (FRET-substrate only) and time-lapse fluorimetry data were normalized to a positive control (0.01% Trypsin). Specific protease activities were inferred with PrAMA by comparing the pattern of substrate cleavage rates for each sample to a matrix of known substrate specificities for ADAM8, ADAM17, MMP-2 and MMP-9 that was determined using purified enzymes [
13]. All calculations and statistical evaluation of data was conducted using Matlab (2014b, MathWorks, Natick, MA).
Statistical analysis
The increase in fluorescence resulting from substrate proteolysis was tracked every 5 min for 4 h. For interpretation of time-lapse fluorimetry data, a non-linear curve-fitting model that accounted for substrate depletion and photobleaching decay served to determine cleavage rates. Cleavage rates are all presented in heat maps averaged over technical triplicates, clear outliers were excluded using Dixon’s Q-Test with a 90% threshold. PrAMA inference was performed as described previously with 30% sampling error and threshold σ
T = 1.4 [
13]. Based on normal distribution of values as tested by the David test at the significance level p = 90%, statistical significance was determined using a two-tailed unpaired Student’s
t test to compare two sample groups. To compare more than two experimental groups, Analysis of Variance (ANOVA) was used. Values are denoted as not significant (ns, p ≥ 0.05), significant * (p ≤ 0.05), highly significant ** (p ≤ 0.01), or very highly significant *** (p ≤ 0.001).
To group the generated datasets, observed average cleavage rates were hierarchically biclustered mean-centered and variance-normalized by row, using Euclidean distance, average linkage and optimal leaf order. Clear patterns emerging from cluster analysis are indicated by dendrograms flanking the array.
To investigate the relationship between CSF cell count, blood–brain-barrier impairment and protease activities, Pearson’s correlation coefficients (r) were calculated between these parameters, respectively. Statistical significance (p) was determined as described in the section above.
Discussion
Since abnormalities on MRI or in CSF require a substantial extent of disease in the case of leptomeningeal metastasis, it is likely that treatment delay due to the insensitivity of current methodologies contributes to the poor prognosis in NM. On the molecular level, CSF is a heterogeneous mixture of a low number of leukocytes, debris, small amounts of soluble protein, and tumor cells in the case of NM. The cellular components in the CSF are an important source of proteins determining the course of the disease relevance. Small molecules deriving from the brain or the blood reach the CSF compartment via diffusion. Conventional analytic approaches such as ELISA or Western Blot are often not suitable to detect low-abundance protein biomarkers in the CSF, which is characterized by a very low concentration of proteins. Compositional mass spectrometry-based proteomic studies of the CSF help to gain insight into dynamic disease-related changes, but generally are too complex for clinical routine [
42]. Here we attempted to overcome this diagnostic constraint by providing a sensitive method of detecting increased proteolytic activities in small volumes of CSF samples only in NM patients with a reproducible and significant difference to samples from patients with brain metastasis and control patients.
MMP2, MMP-9, ADAM8, and ADAM17 are increased in CSF samples from NM patients. For a single case of a clinically stable patient, that is with no signs of impairment based on cytology, MRI and neurological examination, we observed almost similar activity values with the exception of MMP-2, which slightly increased in the follow-up sample. Moreover, in another patient treated intrathecally with the antimetabolite cytarabine, the increase of MMP-2 and all other proteases examined could be reverted, suggesting that the therapeutic progress is reflected by a decrease in MMP and ADAM activities as observed here.
Although we can clearly distinguish CSF samples from NM patients in contrast to control individuals and patients with brain metastases, the result of the PrAMA analysis has some limitations: firstly PrAMA inference suggests a major contribution of distinct proteases to the observed metalloprotease activities, as they can be partially inhibited by BB-94 that works on MMP-2, MMP-9, ADAM8, ADAM17; secondly, based on activities not inhibited by BB-94, we postulate protease activities other than metalloproteases present in CSF samples.
With respect to the identity of the metalloproteases, we have attempted to confirm their presence by ELISA. Only in a very severe case of NM, a 53-year old patient with progressive breast cancer with severe neurologic deficits and diagnosis based on CSF cytology, we were able to detect MMP-9 and TNF-RI (as ADAM17 substrate) by ELISA (Additional file
1: Fig. S4). From this CSF sample, activities determined by PrAMA were up to 10-times higher than in all other NM cases. According to Dixon’s Q-Test, this sample was defined as outlier and excluded from cluster analysis. Clusters of atypical, malignant cells dominated the cytogram and the blood–brain-barrier function was strongly impaired as determined by an albumin quotient of 80.4. Although multimodal treatment was initiated immediately, NM in this patient progressed to death within 3 months.
Although our results provide pilot data for the high impact of CSF profiling in disease at present, they underscore the value of multiplexing, i.e. with respect to a microfluidic platform [
43] and to a series of other diseases related to CSF changes. Future studies in larger patient cohorts are required to validate our preliminary results. Mechanistically, invasion of malignant cells into the leptomeninges and the CSF is a multistep process that requires differential expression of migration- and invasion-related proteins. On a time scale, disease-related changes in the composition of the CSF are expected to precede transmigration of cells into the CSF. Most likely, the proteolytic activity in the CSF may increase before pathological signs can be detected by MRI or CSF cytology. The case of a 62 year-old patient with minimal CSF involvement from breast cancer surviving more than two years, which is described before, underlines the clinical experience that early detection and subsequent treatment of malignant cells in the CSF is crucial for long-term control of neoplastic meningitis. Very early detection of subclinical CSF invasion may therefore be crucial to improve the otherwise frustrating results of NM treatment.
Clinical data known to date implicate that conventional therapies are not sufficient to achieve long-term control or remission in cases of neoplastic meningitis (Strik, Proemmel 2010). Molecularly targeted therapies that aim at preventing tumor cells from invasion into the CSF space may offer more efficacy. At present, several pharmacological inhibitors of MPs are already available and ADAM-inhibitors are being developed. After a more detailed analysis of the pathophysiological mechanisms, pharmacological inhibition of MPs and ADAM proteins may be efficient tools to prevent or treat malignant CSF-invasion.
In this respect, our results from a multiplexed assay provide a novel perspective into MMP/ADAM contribution to the pathophysiology of NM. More importantly, the simple profiling of protease activities in CSF samples from patients with medical indications seems to reflect individual context-dependent protease function in invasive diseases and should be explored in a larger cohort of NM patients but also in other clinical applications including infectious diseases in the CNS.
Authors’ contributions
The study was conceived by CC, HS, and JWB, experiments were designed by CC and JWB, CC and KD performed the experiments, CC and MAM analyzed and interpreted the data, the final manuscript was prepared by CC, HS, MAM, DAL and JWB, the whole study was supervised by JWB. All authors read and approved the final manuscript.