Integrated Transcriptome Analysis Reveals KLK5 and L1CAM Predict Response to Anlotinib in NSCLC at 3rd Line

The oral multi-targeted tyrosine kinase inhibitor (TKI) anlotinib is effective for non-small cell lung cancer (NSCLC) in clinical trials at 3rd line. However, a fraction of patients remains non-responsive, raising the need of how to identify anlotinib-responsive patients. In the present study, we aimed to screen potential biomarkers for anlotinib-responsive stratification via integrated transcriptome analysis. Comparing with the anlotinib-sensitive lung cancer cell NCI-H1975, we found 1,315 genes were differentially expressed in anlotinib-resistant NCI-H1975 cells. Among the enriched angiogenesis-related genes, we observed high expression of KLK5 and L1CAM was mostly associated with poor clinical outcomes in NSCLC patients through Kaplan-Meier survival analysis in a TCGA cohort. Moreover, an independent validation in a cohort of ALTER0303 (NCT02388919) indicated that high serum levels of KLK5 and L1CAM were also associated with poor anlotinib response in NSCLC patients at 3rd line. Lastly, we demonstrated that knockdown of KLK5 and L1CAM increases anlotinib-induced cytotoxicity in anlotinib-resistant NCI-H1975 cells. Collectively, our study suggested serum levels of KLK5 and L1CAM potentially serve as biomarkers for anlotinib-responsive stratification in NSCLC patients at 3rd line.


INTRODUCTION
Biomarkers play an important role in therapies of non-small cell lung cancer (NSCLC). Genomic features, such as gene amplification, point mutations, gene over-expression, and chromosomal translocation, have been identified as biomarkers in NSCLC (1). NSCLC, as the leading cause of cancer mortality worldwide, has greatly benefited from biomarker investigations. Precision therapies have dramatically improved progression free survival (PFS) and overall survival (OS) of NSCLC patients whose tumors harbor positive driver gene mutations, such as EGFR (19 Del and L858R) (2), rearranged ROS1 (3), or translocated ALK (4). Furthermore, immune checkpoint inhibitors have significantly prolonged PFS and OS in specific advanced NSCLC patients, due to the assessment of PD1/PDL1 expression and tumor mutation burden (TMB) (5)(6)(7). Therefore, biomarkers for drug-responsive stratification play crucial roles in NSCLC precision therapy.

Cell Culture
Human NSCLC cell lines NCI-H1975, PC-9, HCC-827, and A549 were obtained from the ATCC: The Global Bioresource Center (https://www.atcc.org/). All cell lines were validated to exclude mycoplasma contamination using a TransDetect PCR Mycoplasma Detection Kit (TransGen, China). The cells were cultured in RPMI 1640 medium (Gibco, USA) supplement with 10% FBS (Gibco, USA), 0.1 mg/ml streptomycin and 100 U/ml penicillin. All cells were incubated at 37 • C and 5% CO 2 in a humidified incubator.

Cell Viability Analysis
In total, 1,500 cells per well were cultured in 96-well plates. After incubating with culture medium overnight, the cells were then exposed to anlotinib for 24 h. CCK8 (Dojindo, Japan) was used to evaluate cell viability according to the manufacturer's protocol. The absorbance was measured at 450 nm using a spectrophotometric plate reader (Bio-Tek, USA). Cell viability was performed according to our previous studies (18,19).
Establishment of an Anlotinib-Resistant NCI-H1975 Cell Line As our previous study described (20), as shown in Figure 1A, 10 7 NCI-H1975 cells were exposed to 100 mg/ml ENU (Sigma, USA) for 24 h. Anlotinib administration was performed to screen anlotinib-resistant NCI-H1975 cells. For the first 5 days, NCI-H1975 cells were exposed to anlotinib (4 µg/ml) and the medium was changed every day. Then, anlotinib (6,8,10, and 12 µg/ml) treatments were performed over the next two months. The resulting cells (approximately 100 cells) showed viability when exposed to anlotinib (12 µg/ml). After approximately 1 month of culture, the anlotinib-resistant NCI-H1975 cells were used in functional assays.

Cell Apoptosis Analysis
In total, 5 × 10 5 cells per well of NCI-H1975 or anlotinibresistant NCI-H1975 were cultured in six-well plates for 24 h. Then, the cells were exposed to anlotinib for 24 h. To assess the apoptosis rate, an Annexin V-FITC/PI Apoptosis kit (Zoman Biotechnology Co., Ltd, China) was used to determine the phosphatidyl serine and membrane integrity of each cell. Briefly, the anlotinib-treated and anlotinib-untreated cells were stained with annexin V-FITC and PI simultaneously and then detected by flow cytometry (BD LSRFortessa, USA). The ratio of early apoptosis and total apoptosis were analyzed by FlowJo 7.6 (BD, USA).

Cell Invasion Analysis
Cell invasion was evaluated by transwell assay. One day before the experiment, all cells were incubated in RPMI 1640 (Gibco, USA) for starvation. 5 × 10 4 NCI-H1975 cells or anlotinib-resistant NCI-H1975 cells per well were then seeded on the top pre-coated chamber in 100 µl RPMI 1640. Five hundred microliter RPMI 1640 containing 15% FBS was added into the lower chamber. After 24 h of incubation, the non-invasive cells were cleaned, and the invasive cells were fixed with 4% PFA for 30 min. The invasive cells were stained with 0.1% crystal violet (Sigma, USA), and photographed using fluorescence microscopy (Nikon, Japan).

RNA-seq Library
The preparation of RNA-seq library was performed according to our previous studies (18,21,22). Briefly, NCI-H1975 cells or anlotinib-resistant NCI-H1975 cells were cultured in 10 cm dishes. Then, 1 ml Trizol reagent (Life Technologies, Inc., USA) was used to lyse the cell samples, followed by total RNA isolation using standard procedures. An Oligotex mRNA Mini Kit (Qiagen, Germany) was used to purify mRNA. Approximately 100 ng mRNA of each sample was used for reverse-transcription, followed by end repair, ligation, using NEBNext Ultra Directional RNA Library Prep Kit (NEB, USA) and PCR amplification (12 cycles) using Q5 High-Fidelity DNA Polymerase (NEB, USA). Lastly, the PCR products were subjected to Illumina sequencing by Next 500 (Illumina, USA). All raw data were deposited at EMBL database under accession number E-MTAB-5997 and E-MTAB-7068.

Bioinformatics Analysis
Raw sequencing data were mapped to a reference genome (hg38) by Tophat. Cufflinks was used to determine the differential transcription pattern. Kilo-base of per million reads mapped (RPKM) was used to define gene expression level. To screen significant differential genes, we filtered the genes whose gene expression levels were no more than a 2-fold change. Log 2 (Fold Change) > 1 represented at least 2-fold up-regulation, and log 2 (Fold Change) < −1 represented at least 2-fold down-regulation. Gene ontology (GO) analysis and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis were performed using a public bioinformatics resource platform (DAVID, https://david. ncifcrf.gov/) by uploading the differential gene lists.

Quantitative Real-Time PCR
Total RNA extraction and reverse transcription reactions were performed according to our previous studies (18,21,22). Briefly, the mRNA levels of the genes of interest were detected by quantitative real-time PCR (RT-qPCR) using ABI step one plus (Applied Biosystems, USA). GAPDH was used as a control gene for normalization. The relative levels of mRNA were calculated as 2 Ct . All primer sequences used for RT-qPCR are listed in Table S1.

Transcriptome Analysis of the TCGA Cohort
RNA-seq data and clinical data for NSCLC patients [including lung adenocarcinoma (LUAD) and lung squamous carcinoma (LUSC)] were downloaded from the TCGA portal (https:// cancergenome.nih.gov/). The RNA-seq data of normal controls were excluded based on TCGA barcode principle (https://wiki. nci.nih.gov/display/TCGA/TCGA$\pm$barcode). After filtering the unqualified samples, 503 LUAD patients and 494 LUSC patients were used for survival analysis. The method of raw data collection was described by the Cancer Genome Atlas Research Network. The correlation analysis of RPKM values and overall survival was performed by R package (survival, version. 2.41-3). Best cutoff value was determined using the "Ward method." Briefly, to determine the P-value, we detected the correlations between each mRNA level and OS. The cutoff value was defined as the lowest P-value.

Detection of Serum KLK5 and L1CAM Levels
Twenty-eight peripheral blood samples from patients with refractory advanced NSCLC (time since prior anlotinib treatment: 2 weeks; Registered No. NCT02388919) were provided by Chia-tai Tianqing Pharmaceutical Co Ltd, Jiangsu Province, China. These samples were selected from 294 anlotinib clinical trial participants randomly. The participants received anlotinib as 3rd line or after 3rd line therapy. For each cycle of medication, the patients received anlotinib (12 mg/day) for 2 consecutive weeks and then discontinued for 1 week. Anlotinib treatment was terminated at disease progression or if intolerable toxicity occurred. The patients with stable disease or partial response lasting 80 days were defined as responders while those patients with disease progression ≤80 days were defined as non-responders. The patients harboring any driver mutations (detected by standard methods afforded by participant hospitals), such as EGFR, ROS1, and ALK, were defined as positive. All enrolled patients were followed up every 8 weeks until death. The ELISA kit for KLK5 detection was purchased from Abcam. Serum L1CAM levels were determined using the DRG Diagnostics ELISA kit (Marburg, Germany). All experimental procedures were performed according to the manufacturer's protocols.

Specificity and Sensitively Analysis
For the TCGA cohort of NSCLC patients, the receiver operator characteristic (ROC) curve for predicting OS was generated by the cutoff value of the mRNA level using GraphPad Prism (GraphPad software, version 5, USA). For anlotinib response prediction, the ROC curves for predicting PFS and OS were generated by the cutoff value of the serum protein level.

Statistical Analysis
There were at least three biological replicates, excluding RNAseq analysis, for each sample. PFS and OS were summarized as median values and were analyzed using the Kaplan-Meier method. The Mantel-Cox test was used to perform Meier survival analysis in GraphPad Prism 5. Log-rank test, two-tailed Student's t-test, or one-way ANOVA with post-hoc Bonferroni correction were used to examine the raw data. Differences were considered significant at * P < 0.05, * * P < 0.01, and * * * P < 0.001.

Anlotinib-Induced Cytotoxicity Disappeared in Anlotinib-Resistant NCI-H1975 Cells
To verify the anti-tumor effects of anlotinib, we administered anlotinib to NSCLC cell lines (including NCI-H1975, PC-9, HCC827, and A549). After exposure to anlotinib (8 µg/ml) for 24 h, the cell viabilities of those NSCLC cell lines decreased to different degrees ( Figure 1B). Among the various lines of NSCLC cells, the NCI-H1975 cells underwent the most cytotoxicity.
To investigate the effect of anlotinib resistance, we established anlotinib-resistant NCI-H1975 cells in vitro (see methods). We treated NCI-H1975 cells and anlotinib-resistant NCI-H1975 cells with anlotinib simultaneously and then examined the cell viability, cell apoptosis, and cell invasion activity. Under the anlotinib (6 µg/ml) stress, the viability of NCI-H1975 cells decreased remarkably, and the similar phenomenon was not observed in anlotinib-resistant NCI-H1975 cells ( Figure 1C). Furthermore, after exposure to anlotinib (4 µg/ml) for 24 h, the apoptosis rate of NCI-H1975 cells increased significantly, while the apoptosis rate of anlotinib-resistant NCI-H1975 cells almost remained unchanged (Figures 1D-G). Consistent with the above results, the invasive ability of anlotinib-resistant NCI-H1975 cells was virtually unaffected, although cells were also exposed to anlotinib (2 µg/ml) for 24 h (Figures 1H-J). These results suggest that anlotinib resistance in NCI-H1975 cells might be attributed to activation/inactivation of tumor survival-related biological processes or signaling pathways.

Transcriptome Analysis Revealed Anlotinib Resistance in NCI-H1975 Cells Attributed to the Expressions of Angiogenesis-Related Genes
To understand the underlying molecular mechanism of anlotinib resistance in NCI-H1975 cells, we next performed transcriptome profiling analysis on both NCI-H1975 cells and anlotinibresistant NCI-H1975 cells. The analysis flowchart was shown in Figure 2A. In total, 14,312 differentially expressed genes were found. After excluding inactive genes (fold change ≤ 2), 595 up-regulated genes and 720 down-regulated genes were obtained for subsequent analysis (Figure 2B). Compared with wild type (sensitive cell line), a considerable fraction of genes is differentially expressed in anlotinib-resistant NCI-H1975 cells (Figure 2C). GO and KEGG analysis indicated that the up-regulated genes and down-regulated genes are enriched in multiple biological processes (extracellular matrix organization/disassembly, angiogenesis, cell adhesion, and so on) or signaling pathways (ECM-receptor interaction, antigen processing and presentation, viral carcinogenesis, and so on), suggesting that the modulation of these enriched genes may play an important role in the process of anlotinib resistance ( Figure 2D, Table S2). Further analysis suggested that modulation of angiogenesis-related genes (including ANGPTL4, FN1, HSPG2, SRPX2, KLK5, L1CAM, Prr22, FOXJ1, IL24, and TRIM54) potentially contributes to anlotinib resistance (Figures 2D,E, Tables S3, S4), as anlotinib is a multi-targeted anti-angiogenesis drug for cancer therapy (8-10, 12, 23).

High mRNA Levels of KLK5 and L1CAM Are Associated With Poor Clinical Outcomes in NSCLC Patients in the TCGA Cohort
To understand the clinical significances of the angiogenesisrelated genes identified above, we performed survival analysis on NSCLC patients from the TCGA cohort. Kaplan-Meier survival analysis indicated that high mRNA levels of ANGPTL4, FN1, HSPG2, and SRPX2 are associated with poor clinical outcome significantly ( Figure S1). However, we also found that downregulation of Prr22, FOXJ1, IL24, and TRIM54 is also correlated with poor clinical outcome in NSCLC patients (including LUAD and LUSC) ( Figure S2). Moreover, our Kaplan-Meier survival analysis showed that high mRNA levels of KLK5 and L1CAM are most significantly associated with poor clinical outcome of NSCLC patients (including LUAD and LUSC) in the TCGA cohort (Figures 2E, 3, Figures S3A-F). Collectively, these results indicated that the activation of KLK5 and L1CAM most likely to result in poor clinical outcome in NSCLC patients and the anlotinib resistance in NCI-H1975 cells.

Serum Levels of KLK5 and L1CAM Predict Response to Anlotinib in NSCLC Patients
To determine whether serum levels of KLK5 and L1CAM potentially serve as biomarkers for anlotinib-responsive stratification in NSCLC patients at 3rd line, we detected the serum KLK5 and L1CAM levels at baseline in 28 refractory advanced NSCLC patients enrolled in an anlotinib clinical trial (NCT02388919), and then performed response analyses. Previous study has revealed that serum levels of L1CAM could be used as an unfavorable prognostic marker in NSCLC patients (24). However, the implications of KLK5 levels vary in different cancers (25)(26)(27)(28). Our raw data including the clinical information and levels of KLK5 and L1CAM were shown in Figure 4A.   (Figures 4D,E). The sensitivity and specificity analysis also confirmed that serum KLK5 and L1CAM levels at baseline had preferable predictive value for anlotinib response (Figure S3).

Knockdown of KLK5 or L1CAM Increases the Sensitivity of NCI-H1975 Cells and Anlotinib-Resistant NCI-H1975 Cells to Anlotinib
To further investigate the roles of KLK5 and L1CAM in anlotinib resistance, we performed RNA interference assays to evaluate anlotinib-induced cytotoxicity in anlotinibresistant NCI-H1975 cells. When anlotinib was administered, knockdown of KLK5 or L1CAM significantly decreased the cell viabilities of anlotinib-resistant NCI-H1975 cells ( Figure 5A).
Meanwhile, anlotinib-induced apoptosis increased significantly, with combined knockdown of KLK5 or L1CAM (Figures 5B,C). Consistent with these results, the invasive ability of anlotinib-resistant NCI-H1975 cells decreased remarkably, after anlotinib administration and knockdown of KLK5 or L1CAM were performed simultaneously (Figures 5D,E). These data indicated that anlotinib-induced cytotoxicity was partially recovered in anlotinib-resistant NCI-H1975 cells after KLK5 or L1CAM knockdown.
Drug resistance is inevitable in the last stage of all anti-tumor drug-related therapeutic regimes (29). Cancer cells can acquire resistance to the anti-tumor drugs by various mechanisms, including over-expression or mutation of the drug target, activation of pro-survival pathways, and eliminative induction of cell death (30). For example, studies have demonstrated the mechanisms of acquired resistance to 1st generation TKIs in NSCLC patients with a positive EGFR mutation, including EGFR T790M mutation, MET amplification, HER-2 mutation, HGF over-expression, etc. (31). In other words, NSCLC patients are not suitable for 1st generation TKI therapy when primary tumors harbor resistant mutations or over-expression. These genomic alterations have been used as biomarkers for antitumor drug-responsive stratification (32,33). Similarly, acquired resistance to anlotinib has been observed in our clinical trials (8,23). Here, we established anlotinib-resistant NCI-H1975 cells and then demonstrated the resistant effects in vitro. Investigation of anlotinib-resistant NCI-H1975 cells may contribute to screening for biomarkers for anlotinib-responsive stratification in NSCLC patients at 3rd line.
Biomarkers play an important role in precision therapy for NSCLC patients. According to gene mutation types, tumor driver gene-derived inhibitors (including EGFR inhibitor, ROS1 inhibitor, and ALK inhibitor) have been screened and used for stratifying treatments in NSCLC patients (2)(3)(4). Furthermore, positive PD1/PD-L1 expression and TMB will be used as biomarkers for guiding treatment with immune checkpoint inhibitors in advanced NSCLC patients (5-7). Next generation sequencing (NGS) provided the platform for screening the above biomarkers (6,7). Our transcriptome analysis suggested that up-regulation of angiogenesis-related genes contributed to anlotinib resistance. Kaplan-Meier survival analysis in the TCGA cohort indicated that the NSCLC patients harboring high mRNA levels of angiogenesis-related genes (including ANGPTL4, FN1, HSPG2, and SRPX2) have poorer prognosis, suggesting that those patients may be unsuitable for anlotinib therapy.
KLK5 and L1CAM play important roles in cancer progression (including cell proliferation, migration, angiogenesis, invasion, and metastasis) (34,35), and their expression levels are associated with prognosis. KLK5 not only regulates KRT19 expression to increase the malignancy of ovarian cancer cells strongly (36), but also induces miRNA-mediated anti-oncogenic pathways in breast cancer (37). However, KLK5 plays different roles in different cancers (38,39). The analysis of correlation between KLK5 expression and prognosis indicated that higher KLK5 mRNA level could sever as indicator for predicting unfavorable prognosis in ovarian cancer patients (25,26) and breast cancer patients (28) and sever as indicator for predicting favorable prognosis in prostate cancer patients (27) and testicular cancer patients (39). L1CAM has been characterized as an important pro-angiogenesis molecular via regulating metalloproteinase expression (40). More important, higher serum L1CAM levels have been described as an unfavorable prognostic marker in NSCLC patients (24). Our data indicated that knockdown of KLK5 or L1CAM contributes to increased anlotinibinduced cytotoxicity upon anlotinib-resistant NCI-H1975 cells. Furthermore, our results indicated that up-regulated mRNA levels of KLK5 and L1CAM are simultaneously associated with anlotinib resistance in NCI-H1975 cells and poor prognosis in NSCLC patients. Although the two cohorts (TCGA and ALTER0303) there may be differences in the population profile, but, here we found that low serum levels of KLK5 and L1CAM at baseline are favorable biomarkers for anlotinib-responsive stratification in NSCLC patients (ALTER0303 cohort) at 3rd line.
Collectively, our integrated transcriptome analysis revealed that high mRNA levels of KLK5 and L1CAM are candidate biomarkers for predicting OS in NSCLC patients. High serum KLK5 and L1CAM levels are potentially associated with poor anlotinib response in NSCLC at 3rd line. Knockdown of KLK5 and L1CAM contributes to increasing sensitivity to anlotinib upon anlotinib-resistant NCI-H1975 cells. Collectively, this study suggested serum levels of KLK5 and L1CAM have the potential for clinical application for anlotinibresponsive stratification.

DATA AVAILABILITY
The datasets generated for this study can be found in the EMBL database under accession number E-MTAB-5997 and E-MTAB-7068.

AUTHOR CONTRIBUTIONS
Experiments were conceived and designed by BH, XZ, and JL. Cell assays were performed by JL, QS, BZ, JQ, SW, YL, and LZ. Bioinformatics analysis and statistical analysis were performed by LZ, JW, and JL. The manuscript was written by JL and revised by HW, XZ, and BH.