Patient characteristics
A total of 669 CSF samples consecutively obtained from six groups of patients between 2014 and 2019 were used for the analysis of nano-sized particle proportion. Preliminary analysis showed that the proportions of nano-sized particles changed after treatment in LM patients (Supplementary Fig. 1B). Hence, for this group of patients, 197 post-treatment samples were excluded and only pre-treatment samples were included in the comparative analysis between patients groups. Demographic characteristics of the 472 patients in the proportion analysis are summarized in Table 1 according to patient group. The median age of all patients was 48 years (range, 0.2–90 years). Non-small cell lung cancer (NSCLC) was the most frequent primary cancer type among LM and brain metastasis patients.
Table 1
Clinical characteristics of CSF sample analysis (n = 472)
Groups
|
Total
|
Cancer control
(n = 100)
|
Healthy control
(n = 73)
|
LM
(n = 150)
|
Brain metastasis
(n = 28)
|
Brain tumors
(n = 72)
|
Other disease of CNS
(n = 49)
|
Gender
|
|
|
|
|
|
|
|
Male
|
215 (45%)
|
70 (70%)
|
31 (43%)
|
53 (36%)
|
13 (46%)
|
36 (50%)
|
12 (25%)
|
Female
|
257 (55%)
|
30 (30%)
|
42 (57%)
|
97 (64%)
|
15 (54%)
|
36 (50%)
|
37 (75%)
|
Median age
(range)
|
48
(0.2–90)
|
17
(1.5–90)
|
59
(20–80)
|
52
(2.0–79)
|
55
(7–79)
|
17
(1.9–80)
|
40
(0.2–83)
|
Combined disease
(%)
|
|
Leukemia (62)
Bladder ca. (8)
Bone tumor (7)
Lymphoma (4)
Breast ca. (3)
Colon ca. (3)
Prostate ca. (2)
Cholangioca. (2)
Melanoma (2)
Others (7)
|
Unruptured An. (43)
Moyamoya ds. (12)
Hydrocephalus (5)
ICA stenosis (4)
BPH (2)
Head trauma (2)
Fibromatosis (2)
Others (3)
|
NSCLC (64)
Breast ca. (35)
Glioma (15)
Non-glial BT (10)
SCLC (4)
Stomach ca. (4)
PCNSL (2)
MUO (2)
melanoma (2)
Others (16)
|
NSCLC (16)
Breast ca. (6)
HCC (1)
Melanoma (1)
Ovarian ca. (2)
Sarcoma (1)
SCLC (1)
|
Glioma (19)
Medulloblastoma (15)
Germ cell tumor (9)
Non-glial MBT (6)
Ependymoma (7)
Pituitary adenoma (6)
Other benign BT (6)
Neurinoma (4)
|
MS (25)
ICH (13)
Infection (9)
Other AID (2)
|
Abbreviations: AID, autoimmune disease; An, aneurysm; BPH, benign prostate hypertrophy; BT, brain tumor; HCC, hepatocellular carcinoma; ICA, internal carotid artery; ICH, intracranial hemorrhage; LM, Leptomeningeal metastasis; MBT, malignant brain tumor; MS, multiple sclerosis; MUO, malignancy of unknown origin; NSCLC, non-small cell lung cancer; PCNSL, primary central nervous system lymphoma; SCLC, small cell lung cancer |
aNumbers in parenthesis are vertical proportion. |
Nano-sized particle peaks in CSF observed by Dynamic Light Scattering (DLS)
Relative proportions of nanoparticles in CSF samples by DLS based on intensity particle size distribution (PSD) were obtained using Zetasizer Nano and built-in software. Although the Zetasizer Nano can detect PSD at the millimeter range, we deliberately abandoned peaks outside the nanometer range (> 1,000 nm) on the relative proportion calculation. Majority of the samples (n = 373, 79%) exhibited two peaks of various PSD proportion (Fig. 1(A and B)). However, a single peak pattern ranging widely from 10 nm to 1,000 nm (8%) or the three peaks pattern (13%) showing the 2nd peak between 10 nm and 100 nm and the 3rd peak at 100–1,000 nm range were also observed. For samples that exhibited two peaks, the estimated sizes of CSF particles in the small and the large peaks were 10.5 nm (standard deviation (SD), 4.52) and 176 nm (SD, 179.6), respectively (Fig. 1(C and B)).
The proportions of nanoparticle peaks in CSF varied across individual patients and among patient groups. For samples that exhibited two peaks, the relative proportions of the small and the large peaks differed depending on patient group (Supplementary Fig. 2(A)). Compared with all other patient groups, LM patients exhibited a significantly lower small peak proportion (35% vs. 55%) and higher large peak proportion (64% vs. 44%, p < 0.0001). Upon further analysis, we found that the large peak was associated with a significantly larger particle size in LM patients than in the other patient groups (mean 252 vs. 188 nm, p < 0.0001; Supplementary Fig. 2B).
Verification of exosomes in CSF
Based on previous studies [6, 7, 10], we assumed that EVs in CSF account for large nanoparticles observed in our CSF samples. To understand the type and origin of EVs, the expression level of several cellular markers was detected in concentrated CSF (Fig. 1(E)). CSF samples were free with cellular proteins such as GM130 and cytochrome c, indicating the EVs markers present in CSF were not originated with cells contaminating CSF samples. Instead, CD81 and CD63, markers of microvesicles, were clearly detected. We then isolated intact EVs from CSF samples to examine their morphology and sizes. Transmission electron microscopic analysis of EVs isolated from CSFs revealed that EVs in CSF exhibited typical shapes of membranous nanovesicles secreted from mammalian cells and their sizes were highly heterogeneous ranging from 50 to 200 nanometers in diameter (Fig. 1(F)).
As an indirect method, we observed that the use of a commercially available exosome purification kit that bases on the size exclusion chromatography technology and eliminates protein or nucleic acid from samples (Exo-spin™) nearly abolished the small peak and increased the proportion of the large peak in DLS measurement (Supplementary Fig. 3A and B).
Differences in EV concentration and size in CSF measured by Nanoparticle Tracking Analysis (NTA)
We measured the nanoparticles (presumed to be ‘extracellular vesicles (EVs from here) concentration and size in samples of unaltered CSF (neither diluted nor concentrated) after two-step centrifugation from different patient groups by NTA using NanoSight NS300 at the same camera condition as described in the Methods (Fig. 2(A)).
The measured EVs concentration of 472 samples was a mean of 3.46 × 108 particles/ml (SD, 0.29 × 108) and different among patients groups (Fig. 2(B)). Healthy control patients (HC) exhibited the lowest EV concentration, with a mean of 2.22 × 108/ml (± 159 × 108). Cancer control (CC) and brain metastasis (BM) patients exhibited mean EV concentrations of 2.68 × 108/ml (± 6.41 × 108) and 2.49 × 108/ml (± 3.15 × 108), respectively. LM patients exhibited the highest mean EV concentrations of 7.15 × 108/ml (± 9.15 × 108), which was a significantly higher than all other groups (p < 0.0001) except other CNS disease group. Patients with other CNS diseases (OD) showed the widest distribution of EV concentrations. The OD group consisted of patients with autoimmune disease (n = 25 out of 27 patients with multiple sclerosis), intracranial hemorrhage (n = 13), or CNS infection (n = 9). Patients with autoimmune disease exhibited a mean EV concentration of 1.31 × 108/ml (± 1.33 × 108), whereas patients with intracranial hemorrhage or CNS infection exhibited higher mean EV concentrations of 15.32 × 108/ml (± 24.15 × 108) and 14.86 × 108/ml (± 27.58 × 108), respectively (Supplementary Fig. 4A).
We also analyzed the size of EV measured by NTA according to patients groups. The mean EV size of LM group was significantly larger than other groups (Fig. 2C and supplementary table 1, ANOVA, p < 0.001). The mean EV size of the other CNS disease group was also significantly smaller than other groups (p < 0.001). In detail, the subgroup of intracranial hemorrhage and autoimmune disease (171 ± 36 nm and 175 ± 22 nm, respectively) drove the EV size of OD groups into the smallest (Supplementary Fig. 4B).
To compare relative EV size distribution between groups, empirical cumulative distribution function (ECDF) plots were generated. LM groups showed the largest distribution as depicted by ECDF plot (Fig. 2(D)), followed by OD group. As depicted on the ECDF plot, the heterogeneity of EV size depend on increase of large size EV in LM patients, whereas that of OD group came from the relatively small sized EVs. Also, we analyzed the proportion of EV size with an interval of 50 nm from 0 to 300 nm, which was tentative limitation of EV size range (Fig. 2(E)). The LM group showed significantly higher proportion of EV size more than 150 nm compared to BT and control groups (p < 0.05, Kruskal-Wallis test, Dunn's Multiple Comparison method). Meanwhile, the proportion of 50–150 nm sized EV was higher in BT and control groups than that of LM (p < 0.05).
We compared these values by NTA to those measured by DLS. In DLS measurement, we assumed the large peak to be EV-sized and calculated a mean of the large peak size by intensity PSD to be a mean EV size. The discrepancy of the mean values between DLS and NTA was various among patients groups (supplementary table 1).
Verification of non-vesicular particles among NTA measured EV in CSF
Although protein level is relatively low in CSF compared to serum, it has been known that various lipoproteins could be observed at EV size range in human biofluids [23]. To identify protein aggregates in CSF nanoparticles, we treated CSF with proteinase K (2 ng/mL) incubated for 60 minutes at 37 °C. All samples showed the shift-to-left pattern of peaks above 250–300 nm, but the EV concentration after proteinase is varied (Supplementary Fig. 5(A, B)). Fourteen (64%) samples showed decreased EV concentration, whereas 6 samples (27%) revealed increased and 2 samples remain without discernible change (< 20%). HC and BM showed a mean EV concentration ratio (proteinase K treated/untreated) 0.98 (± 0.32) and 0.97 (± 0.70), respectively. However, LM groups revealed significantly decreased EV concentration ratio of 0.69 (± 0.28) (Supplementary Fig. 5C, paired t-test, p < 0.05).
Change in EV concentration in LM patients after intraventricular chemotherapy
Among LM patients, 41 were enrolled in a prospective clinical trial (http://cris.nih.go.kr, Identifier: KCT0000082) of ventriculolumbar perfusion (VLP) chemotherapy with methotrexate [24]. Briefly, 24 mg methotrexate premixed with artificial CSF was continuously infused to the lateral ventricles, and lumbar drainage was used to drain the CSF at the same infusion rate by hydrostatic pressure for 3 consecutive days. Abide by the protocol, these patients had matched pre-treatment (day 0) and post-treatment (day 4) CSF samples, and we analyzed changes in EV concentration after intraventricular chemotherapy (Supplementary Fig. 6A). We found that EV concentration decreased in 25 patients (61%), did not change (< 20%) in five patients (12%), and increased in 11 patients (27%) (Supplementary Fig. 6B). We evaluated the relationship between treatment-induced change in EV concentration and overall survival. The median overall survival (OS) of patients with increased EV concentration was 342 days (95% confidence interval (CI), 172–512) and it was significantly prolonged to compared the patients with ‘no change (< 20% of EV concentration)’ (median 216 days, 95% CI, 117–315) and the patients with decreased EV concentration (median 119 days, 95% CI, 90–148) (p = 0.037, Fig. 3(A)). Next, to eliminate from different primary cancer of various prognosis, we did subgroup analysis of 33 patients with the same primary cancer of NSCLC, adenocarcinoma. Median OS was 342 days (95% CI, 165–519) in patients with an increased EV concentration (n = 10) but it was only 170 days (95% CI, 97–242) in patients with ‘no change’ (n = 4) and 104 days (95% CI, 76–132) in patients with a decreased EV concentration (n = 19). The difference of OS was more significant in this subgroup analysis compared with that in all patients (p < 0.001, Fig. 3(B)). Taken together, we can assume that the change of CSF EV concentration (nanoparticle concentration measured by NTA) might be a prognostic biomarker in the LM patients who received the VLP with methotrexate treatment.
Analysis of exosome surface markers among NTA measured EV in patients with LM receiving intraventricular chemotherapy
As we tentatively defined NTA measured nanoparticles to EVs in this study without exosome extraction process, we were inevitably included non-exosomal particles among EVs. Thus, we verified exosome concentration change using surface markers in selected (remained CSF is enough for further study) patients among those with increased and decreased EV concentration of the above survival analysis [24].
Beads bearing each of CD9/ CD63/ CD81 antibody capturing exosome in CSF were measured their mean fluorescence intensity (MFI) in patients with increased (n = 10) and decreased (n = 19) EV concentrations (Fig. 4(A and B)). MFI values of each sample were converted into a ratio of post-treatment to pre-treatment paired samples. In the decreased group, the EV markers were significantly reduced after intraventricular chemotherapy (ratio 0.64, p < 0.001), whereas increased groups showed no significant exosome concentration change (ratio 1.13).
We also verified these exosome changes further using ExoView Tetraspanin ChipTM in a limited number of CSF samples from the EV-increased and EV-decreased groups (n = 3 from each, Fig. 4(C and D)). A triplicate of each sample revealed that each exosome markers (anit-CD9/ CD63/ CD81) after the intraventricular chemotherapy were significantly decreased after the intraventricular chemotherapy (ratio of 0.58/ 0.53/ 0.47) in EV decreased group. Whereas the number of exosome markers were not significantly changed in EV increased group.
Change in miRNA expression in patients with LM receiving intraventricular chemotherapy
As changes in EV concentration after the treatment were related to patients’ OS, we next measured expression of a well-known onco-miRNA, miR-21, in NSCLC (adenocarcinoma). Among those above 33 patients with NSCLC, fourteen samples were available for this further analysis. The miR-21 expression measured by droplet digital polymerase chain reaction (ddPCR) (Supplementary Table 2) was normalized by EV concentration measured by NTA (Fig. 5). The miR-21 expression declined (> 0.2 fold change) in 4 out of 5 patients with an increased post-treatment EV concentration (mean fold change, 0.33; SD, 0.45) and was elevated in 5 of 6 patients with EV-decreased group (mean fold change, 17.90; SD, 30.7). Among the EV concentration ‘no change’ patients, one patient showed declined miR-21 expression and other 2 patients revealed no change of miR-21 expression (< 0.2 fold change) (mean fold change 0.64; SD 0.55). Together, no patients showed elevated miR-21 expression among 8 patients with ‘no change’ or EV-increased groups in contrast to 5/6 patients in EV-decreased group revealed elevated miR-21 expression. Thus, the probability of elevated miR-21 expression after the treatment was significantly higher in the EV-decreased CSF samples compared with the EV-increased or ‘no change’ ones (Fisher’s exact test, p < 0.005).