Patient characteristics
Clinical characteristics are shown in Table 1. For NM patients with lung cancer, the majority of these patients had lung adenocarcinoma (27/42, 64.3%), while of 55 patients had known primary malignancies and 10 patients (18.2%) had NM as the first clinical manifestation. These patients were followed up for two month or longer.
A total of 62 CSF samples were collected from these 58 NM patients, in which three CSF samples were collected from a single patient, while two CSF samples were collected from other two patients at distinct time points. Furthermore, among the 62 CSF samples, 30 CSF samples were collected from 28 patients who received intrathecal chemotherapy and systemic radiotherapy, chemotherapy, and/or molecule-targeted therapy, 11 CSF samples were obtained from 11 patients who received intrathecal chemotherapy, and 12 CSF samples were obtained from 12 patients who received systematic therapy. The remaining nine CSF samples were collected from nine patients who did not receive any anticancer therapy.
Cancer-associated gene mutations in the 62 CSF specimens, regardless of the origin of the primary cancer and the mutated genes functional enrichment analysis
The 62 CSF samples were all positive for ctDNA and mutations of cancer-associated genes. Specifically, 68 (47.6%) of the 143 cancer-associated genes analyzed in the present study had mutations in at least one NM CSF sample, and 62 (100%) of the NM CSF samples carried at least one mutated gene. The most commonly mutated gene was TP53 (54/62, 87.10%), followed by EGFR (44/62, 70.97%), PTEN (39/62, 62.90%), CDKN2A (32/62, 51.61%), APC (27/62, 43.55%), TET2 (27/62, 43.55%), GNAQ (18/62, 29.03%), NOTCH1 (17/62, 27.42%), VHL (17/62, 27.42%), FLT3 (16/62, 25.81%), PTCH1 (15/62, 24.19%), BRCA2 (13/62, 20.97%), KDR (10/62, 16.13%), KIT (9/62, 14.52%), MLH1 (9/62, 14.52%), ATM (8/62, 12.90%), CBL (8/62, 12.90%), and DNMT3A (7/62, 11.29%) (Figure 1). Furthermore, GO (Gene Ontology) annotation and KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway analyses were used to explore the potential functions of these high frequency mutated genes, and it was found that these high frequency mutated genes were rich in the PI3K-Akt signaling pathway (Figure 2).
Furthermore, the variant allele frequency was divided into five groups: ≥50%, 30%-50%, 10%-30%, 1%-10% and 0.2%-1%, respectively. Then, the variant allele frequency was associated with the detectable tumor cells in the CSF samples, and it was found that when the variant allele frequency was greater than 1%, the tumor cells were extensively detected in the CSF, when compared to CSF samples without detectable tumor cells (P<0.001, Figure 3).
Copy number variations (CNVs)
Data on the CNVs in these CSF ctDNA samples were also obtained, and it was found that high CNVs occurred in 22 of 58 NM patients. Among these 22 patients, the primary tumors were 15 NSCL cancers (13 lung adenocarcinoma, one squamous cell carcinoma and one unspecified), four gastric cancers, and one each of breast cancer, parotid carcinoma, and unknown primary cancer. The deletion of the CDKN2A copy number was the most frequent CNV that occurred in seven CSF ctDNA samples, and six of which were non-small cell lung cancers (6/20, 30%). In the increase in CDK4 copy number that occurred in five lung adenocarcinomas, four of these exhibited an increase in MDM2 copy number. In addition, an increase in MDM2 copy number was also detected from another lung adenocarcinoma patient. Two CSF ctDNA samples had a gain of ERBB2 (HER2) copy number from a parotid carcinoma patient, while an increase in CD44 copy number was identified in three patients, in which each patient has breast cancer, gastric cancer and unknown cancer, respectively. In addition, an increased EGFR copy number occurred in three lung adenocarcinoma patients. Other CNVs of tumor-associated genes were detected in five patients (six positive CSF ctDNA samples) with decreased AR copy numbers, five patients had decreased CD274 copy numbers, three patients each has a decreased PDCD1LG2 copy number, two patients each has an increase in FGFR2, CCNE1, or NKX2-1 copy numbers, respectively, and one patient had increased TIAF1, GAS6, or IL6 copy numbers, or reduced CSNK2A1, JAK2, MED12, or SMAD4 copy numbers.
Association of gene mutations with intrathecal chemotherapy and systemic therapy
The data on these unique mutated genes were summarized for each group of patients, followed by a gene-annotation enrichment analysis. It was found that the ERK1/2 pathway was mostly enriched by the GO and KEGG pathway analysis in patients who received both intrathecal chemotherapy and systemic therapy (Figure 4 and Table 2).
The association of CSF ctDNA concentration with Karnofsky performance status (KPS) score, gene mutation and CSF tumor cells
The CSF ctDNA concentration was not statistically associated with the KPS scores (r = -0.038, P=0.768; Figure 5A) or the number of gene mutations (r = -0.195, P=0.129; Figure 5B). Furthermore, the number of gene mutations was not associated with the KPS score (r = 0.192, P=0.135; Figure 5C). However, it was found that the CSF ctDNA concentration was associated with tumor cells in the CSF, when compared to that without circulating tumor cells (Z=-2.883, P=0.004; Figure 5D).
Cancer-associated gene mutations in the 45 CSF samples obtained from 42 NM patients with lung cancer
In the subgroup analysis, it was found that CSF ctDNA was detected in all 45 CSF samples obtained from 42 lung cancer patients with NM, and gene mutations were also detected in all patients. Specifically, EGFR mutations occurred in 39 of 45 patients (86.67%), followed by TP53 (38/45, 84.44%), PTEN (27/45, 60.00%), TET2 (18/45, 40.00%), APC (17/45, 37.78%), CDKN2A (14/45, 31.11%), GNAQ (14/45, 31.11%), and NOTCH1 (11/45, 24.44%). A number of gene mutations previously reported with lung cancer were identified in CSF with NM, while EGFR, TP53, PTEN, TET2, APC, CDKN2A, GNAQ, NOTCH1, FLT3, VHL, BRCA2, PTCH1, CBL, MLH1, BRAF, NRAS, TSC2, CSF1R, KIT, MAP2K1, MSH2, TSC1, HRAS, IFITM1 and BCL9 mutations were statistically more common in the present cohort of NM, when compared to the lung cancer noted in the COSMIC database (https://cancer.sanger.ac.uk) (Table 3).
Enriched genes and gene pathways in NM patients with lung cancer
Mutated genes in the 42 NM patients with lung cancer were analyzed by GO annotation and KEGG pathway analyses. The top three GO terms were negative regulation of cell proliferation, peptidyl-tyrosine phosphorylation and positive regulation of ERK1 and ERK2 cascade. KEGG pathway analysis found these genes were associated with various biological processes, which included the general signaling pathways underlying the progression of cancer (P=5.21×10-30; q=5.05×10-28), chronic myeloid leukemia (P=7.01×10-20; q=3.40×10-18), endometrial cancer (P=3.82×10-19; q=1.24×10-17), bladder cancer (P=6.39×10-19; q=1.55×10-17), melanoma (P=1.78×10-18; q=3.44×10-17), glioma (P=1.62×10-17; q=2.61×10-16), prostate cancer (P=8.66×10-17; q=1.20×10-15), and non-small cell lung cancer (P=1.59×10-15; q=1.93×10-14), while the related signaling pathways were the ErbB signaling (P=4.92×10-10; q=3.67×10-9), VEGF signaling (P=6.00×10-7; q=3.06×10-6), MAPK signaling (P=1.33×10-6; q=5.88×10-6), p53 signaling (P=4.14×10-6; q=1.61×10-5), and m-TOR signaling (P=1.00×10-5; q=3.33×10-5) pathways (Figure 6). Furthermore, the KEGG pathway analysis revealed that EGFR, TP53, CDKN2A, CDK4, BRAF, NRAS, HRAS, JAK3, KRAS, MAP2K1, MAP2K2, PIK3CA and RB1 were strongly associated with non-small cell lung cancer.
The association of EGFR mutations between lung cancer tissues and NM CSF samples
Next, EGFR mutations were associated between lung cancer tissues and the NM CSF samples available in 10 patients (Table 4). Specifically, CSF samples were collected from N033, N063, N077, N1088, N156, N331, N355 and N1286 during the TKI therapy, while N079 and N090 before the TKI. It was found that there were roughly the same EGFR mutations between lung adenocarcinoma tissues and CSF of nine patients, except for N1088, in which the EGFR mutation was undetectable in the CSF sample.
A representative case
In the present cohort, there was a lung adenocarcinoma patient (#N156) who underwent surgical lung cancer resection, and tumor tissues had an EGFR 19Del mutation detected by NGS. Thus, the patient orally received 125 mg of icotinib three times a day for six months and thereafter. However, the patient had a headache during the icotinib therapy for the primary tumor. The head contrast enhanced MRI showed the linear and strip abnormal enhancement of the cerebellar sulcus (Figure 7A) after the patient’s cancer spread into the leptomeninges, and the CSF cytology examination showed tumor cells in the CSF (Figure 7B). Furthermore, the CSF sample revealed EGFR 19Del and T790M mutations in the CSF ctDNA by NGS technology. Given such a situation, the patient was given 80 mg of AZD9291 once a day to replace the icotinib for 18 months, and the patients overall health condition improved. Furthermore, a complete response was confirmed by the contrast-enhanced brain MRI (Figure 7C), CSF cytology (Figure 7D), and undetectable EGFR mutations in the CSF samples. The patient has been alive for nearly 3 years since the diagnosis of NM.