GAP43, a novel metastasis promoter in non-small cell lung cancer

Patients and tissues

In total, 70 patients with operable NSCLC were collected from our previously constructed database [17]. Among them, 37 patients developed brain metastases as the first relapse event within 3 years after surgery, and 5 patients underwent brain metastatic tumor resection. Their formalin-fixed and paraffin-embedded (FFPE) tissues were obtained from the Tissue Bank of Zhejiang Cancer Hospital. The follow-up proceeded as previously described [17]. The median follow-up time was 42 (range 5–108) months. During follow-up, 2 of 33 3-year recurrence-free patients relapsed, 30 of 70 patients died of tumor progression, and 1 of 70 patients was lost to follow-up. All patients provided informed consent before surgery, and the study was approved by the Institutional Review Committee of Zhejiang Cancer Hospital.

Luminex assay

Based on previous publications [18–20], we selected 36 genes hypothesized to be associated with brain metastasis. These genes represented three functional categories: brain growth and metabolism, epithelial–mesenchymal transition and cytokines. Six 10-μm-thick paraffin sections were cut from each NSCLC FFPE sample for the Luminex assay. Tissue homogenates were prepared using a QuantiGene Sample Processing Kit for FFPE Tissues (Affymetrix, Santa Clara, CA, USA) according to the manufacturer’s protocol. A QuantiGene plex 2.0 Reagent System (Affymetrix) was used to capture target RNA and amplify the signal according to the manufacturer’s recommended procedure. Finally, the signal was detected using a Luminex 200 System (Luminex, Austin, TX, USA).

Immunohistochemistry (IHC)

Four-micrometre-thick paraffin sections were cut, mounted on slides, deparaffinized in xylene, rehydrated in decreasing ethanol dilutions and incubated in 3% hydrogen peroxide buffer for 30 min. After blocking of endogenous peroxidase activity, the slides were boiled with citrate buffer (pH 6.0) in a pressure cooker for 90 s, followed by blocking of non-specific binding sites with 5% blocking serum for 30 min at room temperature. The sections were then incubated with rabbit anti-GAP43 antibody (Huabio, Hangzhou, China) at a dilution of 1:1200 overnight at 4 °C. Afterwards, 1:2000 diluted goat anti-rabbit secondary antibody (Dako, Glostrup, Denmark) was applied for 30 min. The reaction products were visualized using 3,3′-diaminobenzidine (DAB; Dako), and the sections were lightly counterstained with hematoxylin and mounted. The results were evaluated by two experienced pathologists and photographed.

Cell culture

The human NSCLC cell lines SK-MES-1, NCI-H1650, NCI-H1975, NCI-H2122, A549, NCI-H838, NCI-H460 and NCI-H661 were purchased from American Type Culture Collection (Manassas, VA, USA). All cells were authenticated by short tandem repeat (STR) profiling and were found to be negative for mycoplasma. SK-MES-1 cells were maintained in Eagle’s minimum essential medium (EMEM; Genom, Hangzhou, China) containing 10% fetal bovine serum (FBS, Gibco); all other cell lines were maintained in RPMI 1640 medium (Gibco) containing 10% FBS. All cells were cultured at 37 °C in a 5% CO₂ incubator.

Western blotting

Adherent cells were collected by gentle scraping, washed three times in phosphate-buffered saline (PBS), and lysed in Nonidet P 40 cell lysis buffer supplemented with phosphatase inhibitor cocktail (Cwbio, Beijing, China) and protease inhibitor cocktail (Roche, Mannheim, Germany). A Bradford calorimetric assay (Cwbio) was used to measure the protein concentration of the cell lysates. Thirty micrograms of total protein was separated on a 10% SDS–polyacrylamide gel electrophoresis (SDS-PAGE) gel and then transferred to a polyvinylidene fluoride membrane (Millipore, Bedford, MA, USA). Each membrane was blocked in TBS-Tween-20 (TBS-T) containing 5% non-fat milk at room temperature for 2 h, followed by incubation with anti-GAP43 antibody (1:1000, Huabio, Hangzhou, China) at 4 °C overnight. The next day, the membrane was incubated with an HRP-conjugated secondary antibody (Servicebio, Wuhan, China) at room temperature for 2 h, and the results were detected using an enhanced chemiluminescence (ECL) reagent (Cwbio). Tubulin (1:1000; Antibody Revolution, San Diego, CA, USA) was used as the loading control.

Reverse transcription-quantitative polymerase chain reaction (RT-qPCR)

Total RNA was extracted from cells using the TRIzol method. The RNA concentration was measured using a microvolume spectrophotometer, and 500 nanograms of RNA was reverse transcribed using a PrimeScript™ RT Reagent Kit (Takara, Dalian, China). Quantitative PCR (qPCR) was carried out as previously described [21]. The primer sequences were as follows: TTCTTGGTGTTGTTATGGCAAG (GAP43 forward), GAGGAAAGTGGACTCCCACAG (GAP43 reverse), GAAGGTGAAGGTCGGAGTC (GAPDH forward), and GAAGATGGTGATGGGATTTC (GAPDH reverse). The expression levels of GAP43 were analyzed using the 2(−Delta Delta C(T)) method [22] and adjusted according to the expression levels of the housekeeping gene GAPDH.

GAP43 knockdown and overexpression

For transient knockdown, 4 specific siRNAs targeting GAP43 and a negative control siRNA were designed and inserted into a pRNAT-U6.1/Neo vector. For transient overexpression, GAP43 cDNA was subcloned into a p3 × FLAG-CMV-10 vector, and an empty p3 × FLAG-CMV-10 vector was used as the control. Lipofectamine 3000 Reagent (Invitrogen, Carlsbad, CA, USA) was used to carry out the transfections, and after 48 h, the cells were harvested to validate the alterations in GAP43 levels or for subsequent experiments.

For stable GAP43 knockdown, a lentiviral vector was applied. Briefly, the GAP43 knockdown lentiviral vector pHBLV-CMV-shGAP43-3flag-EF1-Luc-T2A-Puro and a negative control lentiviral vector were purchased from Hanbio Biotechnology Co., Ltd., Shanghai, China. 293T cells were co-transfected with lentiviral vectors and the helper plasmids pSPAX2 and pMD2G. After 48 h of transfection, the supernatant was collected and purified by centrifugation followed by ultracentrifugation. NCI-H661 cells at 50% confluence were infected with GAP43 knockdown or negative control lentiviral constructs in 8 µg/ml polybrene. Stably transfected cells were selected using 1 µg/ml puromycin for 2 weeks and maintained in 0.5 µg/ml puromycin. GAP43 knockdown was confirmed by western blotting.

Wound healing assay

The wound healing assay was conducted using a 2-Well Culture-Insert (Ibidi, Munich, Germany). Cells were seeded into the wells, and after removal of the insert, a 500-μm cell-free gap was formed. Multiple time-lapse images of the gap were taken at a magnification of 100×.

Transwell migration and invasion assays

Twenty-four-well, 8-μm pore size transwell inserts (Corning, Oneonta, NY, USA) were used. For invasion assays, the inserts were coated with Matrigel (BD, San Jose, CA, USA) before the cells were added. A total of 500 μl of complete medium was added to the lower chamber, and 200 μl of cell suspension (3.0 × 10⁴ NCI-H661 cells for migration, 2.5 × 10⁴ NCI-H661 cells for invasion, 2.0 × 10⁴ NCI-H1650 cells for migration and invasion) in RPMI 1640 supplemented with 1% FBS was added to the upper chamber. After incubation for 24 h, the cells in the upper chamber were removed by gentle wiping using cotton buds, and the cells on the underside of the inserts were fixed with methanol, stained with 0.1% crystal violet, photographed and counted.

In vivo metastasis assay

BALB/c (nu/nu) male nude mice 4–6 weeks of age purchased from Shanghai Slac Laboratory Animal Co. Ltd. were used, each group with 10 mice. Two hundred microliters of PBS containing 1.0 × 10⁶ cells was injected into the left ventricle of each mouse. Metastatic lesions were monitored every week using bioluminescence imaging (BLI). Briefly, the mice were anaesthetized and injected intraperitoneally with 150 µg of d-luciferin (Yeasen, Shanghai, China) per gram of weight. 10 min later, the bioluminescence was imaged using an IVIS imaging system (Caliper Life Sciences, Alameda, CA, USA) and analyzed using Living Image Software 4.3.1.

Immunofluorescence

Adherent cells at 80% confluence on an 8-well glass slide (Millipore) were fixed in 4% paraformaldehyde for 15 min and permeabilized with 0.1% Triton X-100 for 5 min. Next, 100 µl of phalloidin conjugate working solution (Abcam, Cambridge, MA, USA) was added to each well of the slide and incubated with the cells for 60 min at room temperature away from light. The cell nuclei were stained with DAPI (Biosharp, Hefei, China) for 10 min at room temperature away from light, and fluoroshield mounting medium (Abcam) was used to seal the slide. Fluorescence was photographed using a confocal microscope.

F-actin/G-actin ratio in vivo assay

Proteins were extracted and analyzed using an F-actin/G-actin in vivo assay kit (BK037; Cytoskeleton, Denver, CO, USA) based on the manufacturer’s instructions. Briefly, cells were lysed in a detergent-based buffer that dissolved G-actin but not F-actin. A centrifugation step pelleted F-actin while leaving G-actin in the supernatant. Actin in the pellet and supernatant was quantitated by western blotting. The grayscale value of each band was measured using ImageJ software.

Rac1 activation assay

A Rac1 activation assay kit (STA-405; Cell Biolabs, San Diego, CA, USA) was applied to prepare protein according to the manufacturer’s instructions. Briefly, protein was extracted from cells and divided into two equal parts. One part was used to quantitate total Rac1. The other part was incubated with the p21-binding domain (PBD) of p21-activated protein kinase (PAK) Agarose beads to specifically bind and pull-down active Rac1. Next, total Rac1 and active Rac1 were detected by western blotting.

Statistical analysis

The expression index of each gene from the Luminex assay was averaged in two groups according to the median and then analyzed as categorical variables. The Cox proportional hazards regression model was applied to evaluate the brain metastasis-prognostic value of each gene. The Kaplan–Meier estimator was used to generate brain metastasis-free survival curves, and a log-rank test was used to determine significant differences. A web application called KM plotter (http://kmplot.com/analysis/index.php?p=service&cancer=lung) was employed to obtain the progression-free survival and overall survival curves in public database of lung cancer. Correlations between GAP43 expression and clinicopathological factors were examined using a Chi square test and Mann–Whitney U test. Every experiment was repeated at least 3 times. The results of qPCR, western blotting and cell migration and invasion assays were analyzed using Student’s t-test, and the BLI results were analyzed with a Mann–Whitney U test. IBM SPSS Statistics 24.0 software was used for data processing. A P value < 0.05 was considered statistically significant.

Patient characteristics

Genes related to brain metastasis

The mRNA expression levels of 36 candidate genes were measured with a Luminex assay, and a Cox regression model was applied to estimate the prognostic value of each gene. As presented in Table 2, univariate Cox regression analysis indicated that GAP43 and PMP2 were risk genes for NSCLC brain metastasis (crude HR: 2.12; 95% CI 1.09–4.13; P = 0.027 for GAP43; crude HR: 2.12; 95% CI 1.05–4.30; P = 0.036 for PMP2). Additionally, ST6GALNAC5 (P = 0.071), SNAI1 (P = 0.066), CDH2 (P = 0.073) and FABP7 (P = 0.066) showed borderline significant predictive values. Furthermore, multivariate Cox regression analysis involving genes whose P values were less than 0.1 in univariate analyses and clinicopathologic factors demonstrated that GAP43, SNAI1 and grade were independent prognostic factors for NSCLC brain metastasis (Table 2). The risk of developing brain metastases for NSCLC patients with high GAP43 expression levels was 3.29-fold higher than that for patients with low GAP43 levels (95% CI 1.55–7.00; P = 0.002). Compared with NSCLC patients with low SNAI1 expression levels, those with high expression levels had a 1.98-fold higher risk of developing brain metastases (95% CI 1.00–3.95; P = 0.051). Patients with poorly differentiated tumors more easily developed brain metastases than those with well-differentiated tumors (adjusted HR: 2.08; 95% CI 1.00–4.34; P = 0.051). Furthermore, Kaplan–Meier survival curves and the results from a public database (http://kmplot.com) indicated that patients with high GAP43 levels showed worse progression-free and overall survival (Fig. 1a–d).

Factorsa	HR	95% CI	P value
Univariate analysis
BMI1	0.58	0.29–1.16	0.124
POU5F1	0.75	0.38–1.47	0.397
KLF4	1.11	0.57–2.18	0.757
SMO	0.86	0.43–1.68	0.650
JAG1	0.85	0.43–1.67	0.634
SPARC	1.44	0.73–2.84	0.292
ST6GALNAC5	1.90	0.95–3.81	0.071
SNAI1	1.89	0.96–3.74	0.066
TNF	1.23	0.63–2.41	0.545
CXCR5	0.77	0.39–1.51	0.448
GFAP	1.00	0.51–1.96	0.995
DYNC1LI2	1.13	0.58–2.21	0.729
MMP3	1.21	0.62–2.37	0.579
CDH2	1.87	0.94–3.71	0.073
EGFR	0.65	0.33–1.28	0.212
VCAN	1.50	0.76–2.97	0.247
HGF	0.76	0.39–1.50	0.427
CTNNB1	0.76	0.39–1.50	0.428
CXCR4	1.46	0.74–2.88	0.275
KIF3A	1.47	0.74–2.89	0.269
GPM6A	0.70	0.36–1.38	0.307
HBEGF	0.95	0.48–1.86	0.869
CXCL13	0.99	0.51–1.95	0.985
TGFB2	0.92	0.47–1.80	0.803
PLP1	0.81	0.39–1.69	0.568
TWIST1	1.45	0.74–2.86	0.280
PTGS2	0.87	0.44–1.71	0.679
BPTF	0.77	0.39–1.52	0.443
FABP7	1.92	0.96–3.83	0.066
GAP43	2.12	1.09–4.13	0.027
PMP2	2.12	1.05–4.30	0.036
UGT8	1.23	0.63–2.41	0.549
GPM6B	1.00	0.51–1.96	0.990
FAM107A	0.62	0.31–1.23	0.169
IL8	1.27	0.65–2.49	0.492
MMP9	1.16	0.59–2.27	0.668
Multivariate analysisb
GAP43	3.29	1.55–7.00	0.002
SNAI1	1.98	1.00–3.95	0.051
Grade	2.08	1.00–4.34	0.051

HR hazard ratio, CI confidence interval

Italic values were statistically significant (P < 0.05)

^aFactors were analyzed as categorical variables, and low-expression groups or well-differentiated groups were referred

^bGenes whose P values were < 0.1 in univariate analyses and clinicopathologic factors were involved in the multivariate Cox regression analysis, and the Forward LR (based on partial maximum likelihood estimation) was used to identify independent risk factors

GAP43 expression in brain metastatic tissues and NSCLC cell lines

To further investigate the role of GAP43 in NSCLC, 5 FFPE NSCLC tissues with paired brain metastatic tissues and a panel of human NSCLC cell lines were collected. Immunohistochemistry showed that 3/5 brain metastatic tissues had a higher level of GAP43 expression than the corresponding primary NSCLC tissues (Fig. 1e). Additionally, as illustrated in Fig. 1f, GAP43 expression was high in NCI-H661 cells, comparatively lower in SK-MES-1 and NCI-H460 cells and absent in NCI-H1650, NCI-H1975, NCI-H2122, A549 and NCI-H838 cells.

GAP43 promoted cell migration and invasion in vitro

To study the effect of GAP43 on cell migration and invasion in vitro, NCI-H661 and NCI-H1650 cells were chosen to perform transient GAP43 knockdown and overexpression, respectively. The suppression of GAP43 in NCI-H661 cells and overexpression of GAP43 in NCI-H1650 cells were confirmed at both the mRNA and protein levels (Figs. 2a, b, 3a, b). As revealed in a wound-healing assay (Fig. 2c), GAP43 knockdown in NCI-H661 cells significantly impaired cell lateral migration ability. Additionally, Transwell assays demonstrated that GAP43-silenced NCI-H661 cells migrated and invaded less efficiently than control cells (Fig. 2d, e). Similarly, GAP43 significantly heightened the migratory and invasive capacities of NCI-H1650 cells (Fig. 3c–e).

GAP43 depletion inhibited metastasis in vivo

To test and verify the function of GAP43 in vivo, NCI-H661 cells were stably transfected with luciferase-expressing GAP43 knockdown lentiviral vectors (Fig. 4a) and injected into the left ventricle of nude mice. Markedly, both brain and bone metastases decreased after GAP43 was silenced in NCI-H661 cells (Fig. 4b, c).

GAP43 knockdown triggered F-actin depolymerization

To further explore the molecular mechanisms, we investigated epithelial-mesenchymal transition (EMT) markers, including E-cadherin, N-cadherin, fibronectin and vimentin, applied a Human Phospho-Kinase Array and Human XL Cytokine Antibody Array from R&D, Minneapolis, MN, USA. However, no significant difference was found related to tumor migration (data no shown).

To examine alterations in cytoskeletal organization, F-actin in stable GAP43-knockdown NCI-H661 cells was stained with phalloidin, and a marked structural change was observed due to decreased filamentous F-actin after GAP43 knockdown (Fig. 4d). Furthermore, an F-actin/G-actin ratio assay showed similar results, specifically, the amount of F-actin was reduced with GAP43 depletion (Fig. 4e). Next, a Rac1 activation assay was performed and showed that active Rac1 levels were decreased and total Rac1 stayed invariant after GAP43 was silenced (Fig. 4f).