Correlation of LOXL2 expression in non-small cell lung cancer with immunotherapy

Abstract

Lung cancer is the most prevalent and lethal disease globally, with approximately 80% of cases being non-small cell lung cancer (NSCLC). NSCLC is primarily composed of lung squamous cell carcinoma (LUSC) and lung adenocarcinoma (LUAD). Despite chemotherapy currently being the primary treatment for NSCLC, chemotherapy resistance remains a significant challenge for patients. Recent studies have proposed immunotherapy as a promising new avenue for treating NSCLC. The association between the lysyl oxidase-like 2 (LOXL2) gene and NSCLC was explored using multiple online tools and bioinformatics analysis software based on the available datasets from TCGA. The immune microenvironment of the tumor was explored by calculating ImmuneScore, StromalScore, and TumorPurity of LUAD and LUSC and analyzing the infiltration of 22 immune cells in lung cancer tissues. LOXL2-related loads were obtained from the Xena database for LUSC and LUAD patients, and relevant prognostic genes were identified by analyzing survival curves. Functional and pathway enrichment analyses of prognostic, predictive genes were performed using Gene Ontology (GO) and the Kyoto Encyclopedia of Genes and Genomes (KEGG). The expression of LOXL2 in NSCLC was detected by RT-qPCR. LOXL2 may be involved in the progression of LUAD and LUSC and is closely related to the T-lymphocyte subpopulation, T-reg cells. SEMA7A and VEGFC are identified as the genes that interact with LOXL2 and could be used as prognostic signature genes in NSCLC patients. LOXL2 may become a prognostic marker and a new target for immunotherapy.

Introduction

Lung cancer stands as a leading cause of cancer-related mortality worldwide, contributing to an estimated 26% of global cancer-related fatalities [1-3]. Amongst all lung cancer cases, non-small cell lung cancer (NSCLC) accounts for roughly 80%, further classify into lung squamous cell carcinoma (LUSC) and lung adenocarcinoma (LUAD) [4,5]. Decades of intensive genomic and signaling pathway research have revealed that NSCLC constitutes a diverse group of diseases characterized by genetic and cellular heterogeneity [6], primarily encompassing LUSC and LUAD [7]. Notably, early-stage NSCLC often exhibits mild symptoms and clinical signs, posing challenges for early detection. The aggressive nature and rapid progression of NSCLC frequently lead to patients presenting with locally advanced or metastatic disease at diagnosis [8].

Surgical resection is the preferred therapeutic modality for patients diagnosed with early-stage NSCLC, whereas a multimodal approach combining radiotherapy and chemotherapy is typically employed for those presenting with advanced, inoperable disease stages [9]. However, resistance to chemotherapeutic agents such as cisplatin has been a significant challenge, necessitating the development of new approaches to treat NSCLC [10]. In recent years, there has been growing recognition of the benefits of harnessing the host’s immune system to combat malignancy [11,12]. The emergence of immunotherapy has revolutionized the landscape of NSCLC treatment by capitalizing on this concept [13]. The underlying principle of immunotherapy involves the stimulation of the patient’s endogenous T-cells and the subsequent release of cytokines, which facilitate the targeted destruction of tumor cells [14]. At present, the principal immune checkpoints are PD-1 and PD-2 [15]. Prior investigations have demonstrated the promise of neoadjuvant immunotherapy in patients with substantial tumor burden [16]. Furthermore, an intricate interplay between regulatory T-cells (Tregs) and the tumor microenvironment has been well-documented in numerous studies, underscoring the importance of further exploration into novel targets for immunotherapy [17]. It is therefore of interest to explore new targets for the immunotherapy of NSCLC. The expression of LOXL2 is markedly elevated in tumors, contributing to tumor invasion and migration [18]. However, the utility of LOXL2 as a target for immunotherapy in NSCLC remains uncharted territory and merits further investigation.

In this study, we conducted a bioinformatics analysis of the risk factors influencing the prognosis of non-small cell lung cancer (NSCLC) using differential gene expression, survival curves, the tumor immune microenvironment, associations with other molecules, and KEGG and GO analysis. Furthermore, RT-qPCR analysis demonstrated that LOXL2 was highly expressed in NSCLC. The aforementioned results indicate that LOXL2 may serve as a promising target for immunotherapy in NSCLC patients.

Material and methods

Gene expression data analysis

Inclusion criteria for this study were as follows: (a) pathologic diagnosis of NSCLC (biopsy or surgically resected tissue), (b) no experience of immunotherapy and other radiotherapy, and (c) having complete gene sequencing results. The exclusion criteria were as follows: (a) incomplete information; (b) combination with other tumors. In the TCGA database, GEPIA identifies the top 500 differentially expressed genes (DEGs) in LUAD and LUSC patients, and subsequently identified 11 commonly related genes by intersecting the results. Subsequently, Kaplan-Meier plotter 2 was employed to assess the overall survival (OS) of LUAD and LUSC cancer patients. The samples were divided into two groups according to gene expression levels in order to determine the significance of each gene in predicting the patient’s prognosis. Subsequently, a Kaplan-Meier survival curve was constructed to compare the two groups and contrast the log10HR value (HRs) and P-value. The two risk factors, LOXL2 and TLDC1, with the most unfavorable prognosis in LUAD and LUSC, were identified. A total of 483 cases of lung adenocarcinoma (LUAD) and 486 cases of lung squamous cell carcinoma (LUSC) were included in the analysis from the GEPIA database. The cases were designated as the “tumor group”, and the expression of LOXL2 in this group was compared with that of the “control group”.

Tumor purity and immune cell infiltration analysis

The ESTIMATE algorithm was employed to calculate ImmuneScore, StromalScore, and Tumor Purity for LUAD and LUSC patients, respectively. The proportion of 22 TILs in LUSC and LUAD was determined using the R package CIBERSORT. Furthermore, GEPIA2021 was employed to visualize gene expression profiles in order to explore the differences in immune cell subtypes. The immune cells included in this study were neutrophils, eosinophils, mast cells, and mast cells in different states of activation. The following cell types were identified: dendritic cells activated, dendritic cells quiescent, macrophages M2, T cell follicular helper cells, T cell regulatory (Tregs), T cell γδ, NK cells quiescent, NK cells activated, monocytes, macrophages M0, macrophages M1, T cell CD4 memory activated, T cell CD4 memory quiescent, T cell CD4 naive, T cell CD8, plasma cells, B cell memory, and B cell naive.

Immune prognosis analysis of LOXL2

ssGSEA was used to analyze the immune profile of LOXL2 in LUAD and LUSC, and the tumor tissues of LUAD and LUSC in the GEPIA database were compared to the adjacent tissues in normal lungs. The results demonstrated a significant positive correlation between LUAD and T-reg (rho=0.311, P=5.71e-13) and between LUSC and T-reg (rho=0.494, P<2.2e-16).

Correlated gene analysis

LOXL2-associated genes from LUAD and LUSC patients were obtained using the Xena database, P-values and R-values were calculated, and differentially expressed genes (DEGs) were analyzed. Cross-tabulation analysis was performed on genes that were upregulated in LOXL2-associated genes in LUAD and LUSC patients. The eight cross-over genes associated with OS were extracted by univariate Cox regression analysis of the results and then the median of their characteristic scores. Patients were finally categorized as high and low risk. Patient survival was analyzed by the Kaplan-Meier method. Time-dependent receiver operating characteristic (ROC) curves were used to predict the accuracy and characteristics of clinical characteristics. The R package rms was employed to generate the nomograms.

Genetic interaction analysis

The search tool for interacting genes/proteins (STRING) database (http://cn.string-bd.org) was used to establish a PPI network of LOXL2.

Gene enrichment analysis

GO and KEGG analysis was carried out by using the R package “clusterProfiler”. Gene Ontology (GO) was used for functional enrichment analysis, while the Kyoto Encyclopedia of Genes and Genomes (KEGG) was used for pathway enrichment analysis. The effects of functional enrichment included molecular functions (MF), cellular components (CC), and biological processes (BP). The top 10 pathway enrichment was analyzed.

Drug sensitivity analysis

We used the “oncoPredict” R package to analyze the batch-corrected prioritization of different drugs in the DREAMT database.

Cell culture

Human bronchial epithelial cell line, human non-small cell lung cancer cell line PC9, A549, H1299 were provided by Jiangsu Province Key Laboratory of Geriatrics, and cultured in DMEM (Sigma-Aldrich) supplemented with 100 μg/ml streptomycin, 100 μg/ml penicillin (Gibco), and 10% fetal bovine serum (FBS) (Invitrogen). All cells were maintained in a 5% CO₂ incubator (Thermo Fisher Scientific) at 37°C. Cells in the logarithmic growth phase were used for experiments.

RT-PCR

Total RNA was extracted from the cells using Trizol reagent (Invitrogen). RNA was reverse transcribed to cDNA using Primescript™ RT Master Mix (Vazyme). Quantitative RT-PCR was performed using ChamQTM Universal SYBR QPCR Master Mix (Vazyme) and Steppe One PlusTM Real-Time PCR System (Application Biosystems, Foster City, CA, USA). The primers for LOXL2 and GAPDH (used as an internal reference) were purchased from Genscript. RT-PCR was performed under the following conditions: Fluorescence signal was acquired at 94° for 60 seconds, 95° for 10 seconds and 60° for 30 seconds, and after 40 cycles, 60° were acquired. Target gene expression levels were normalized to GAPDH expression and then calculated using the 2^-ΔΔCt method. The primer sequences are listed in Table 1.

Table 1.

List of primers

Gene	Primer (5’-3’)
LOXL2	F: AACGAGGCGACCCTTGCAGC
	R: GGGTGCGCTTGCGGTAGGTT
GAPDH	F: CTCCTCCACCTTTGACGC
	R: CCACCACCCTGTTGCTGT

Cell viability

The Cell Counting Kit-8 (Vazyme) was performed to detect cell proliferation. We purchased doxorubicin from Beijing Huamei Company and used it to treat 1×10^4 cells in a 96-well plate. The cells were exposed to different concentrations of doxorubicin for 24 hours. 10 μl CCK8 reagent was added into each well according to the instruction of CCK8. Then, absorbance of samples was detected at 450 nm wavelength.

Statistical analysis

Kaplan-Meier plots used HR and P-value or Cox P-value for the log-rank test. Comparisons were made using unpaired and paired t-tests for normally distributed variables or Mann-Whitney U-test and Wilcoxon signed-rank test for non-normally distributed variables. R software was used to perform all statistical analyses. P-value <0.05 was considered significant.

Results

Expression feature of LOXL2

We identified the top 500 DEGs in LUAD and LUSC from the TCGA databases. We crossed them to obtain eleven common DEGs, including PITX3, STAP1, DBP, FAM72B, LOXL2, ERH, ADGRD1, CTD-2626G11.2, RP11-1046B16.3, MYL2, TLDC1 (Figure 1A). Next, we used Kaplan-Meier Plotter and GEPIA database to explore the relationship between this set of genetic changes and the survival rate of LUAD and LUDSC patients (Figures S1, S2). We found six protective factors: STAP1, DBP, ADGRD1, CTD-2626G11.2, 1046B16.3, MYL2 and five risk factors: PITX3, FAM72B, LOXL2, ERH, TLDC1. Among these, we focused on two risk factors LOXL2 and TLDC1, which had high-risk values in LUAD and LUSC (Figure 1B, 1C). Based on the GEPIA dataset, we analyzed LOXL2 and TLDC1 in lung cancer tissues and adjacent tissues. We included lung cancer cases (n=483 for LUAD; n=486 for LUSC) as the “tumor group”. Compared to the “control group”, we found that the expression level of LOXL2 in the “tumor group” was higher (Figure 1D, 1E).

Figure 1.

LOXL2 is highly expressed in non-small cell lung cancer. A. Venn diagrams of the top 500 abnormally expressed genes related to LUAD and LUSC (R package analysis). B. Survival analysis of 11 intersection gene expression levels in LUAD and LUSC. C. LOXL2 and TLDC1 are associated with shorter overall survival. D, E. The level of LOXL2 in LUAD and LUAC cancer tissues is higher than in para cancer. F. Univariate analysis was used to analyze the correlation between LOXL2 expression and clinical prognosis in NSCLC patients.

Furthermore, we verified that LOXL2 was highly expressed in LUAD and LUSC by calculating the p-value, HR value, and 95% cl of the LOXL2 gene in the Oncomine database and TCGA database (P<0.05) (Figure 1F; Table 2).

Table 2.

Univariate analysis of the correlation between LOXL2 expression and clinical features and OS in LUAD and LUSC patients

Dataset	P value	HR	95% CI	Prognostic
GSE3141	<0.0001	3.0597	1.7798-5.2600	Poor	Go
TCGA	0.0011	1.4389	1.1574-1.7890	Poor	Go
GSE11969	0.0012	2.2213	1.3702-3.6009	Poor	Go
Stage l	0.1672	1.7682	0.7876-3.9693
Stage ll	0.516	1.5434	0.4166-5.7188
Stage lll	0.0014	3.4612	1.6128-7.4280	Poor	Go
GSE67639	0.0015	1.4023	1.1375-1.7288	Poor	Go
GSE50081	0.0023	2.1125	1.3056-3.4179	Poor	Go
Stage l	0.0428	1.8640	1.0203-3.4056	Poor	Go
Stage ll	0.1498	1.7766	0.8127-3.8839
GSE30219	0.0077	1.5230	1.1177-2.0753	Poor	Go
GSE5123	0.0251	2.8011	1.1376-6.8970	Poor	Go
Stage l	0.4481	1.7116	0.4269-6.8620
Stage ll	0.2313	2.4212	0.5692-10.3000
Stage lll	0.2918	3.6458	0.3290-40.3971
GSE26939	0.0293	1.8049	1.0613-3.0694	Poor	Go
Stage l	0.1121	1.8641	0.8646-4.0194
Stage lll	0.1187	2.6001	0.7828-8.6362
GSE31210	0.0522	1.9740	0.9935-3.9221
Stage l	0.0015	4.7791	1.8171-12.5698	Poor	Go
Stage ll	0.3536	0.5551	0.1601-1.9252
GSE12428	0.0566	2.9957	0.9695-9.2566		Go
GSE33072	0.0910	1.6524	0.9229-2.9585		Go
GSE68465	0.1093	1.2647	0.9487-1.6860		Go
GSE68571	0.1336	1.8647	0.8261-4.2092		Go
Stage l	0.2186	1.9920	0.6646-5.9708
Stage ll	0.4599	0.5556	0.1169-2.6405
GSE17710	0.2148	0.5670	0.2313-1.3897		Go
Stage l	0.6453	1.2765	0.4514-3.6101
Stage ll	0.0807	0.1560	0.0194-1.2551
GSE37745	0.2954	1.2176	0.8421-1.7604		Go
Stage l	0.3817	1.2318	0.7721-1.9651
GSE37745	0.2954	1.2175	0.8421-1.7604		Go
Stage l	0.3817	1.2176	0.7721-1.9651
Stage ll	0.2517	1.2318	0.7032-3.8342
Stage lll	0.0789	1.6420	0.1594-1.1054
GSE31908_GPL96	0.3379	0.4198	0.4808-8.4458		Go
Stage l	0.4208	2.0151	0.1952-49.9487
GSE31908_GPL97	0.3389	3.1225	0.0442-2.9246		Go
GSE102287	0.3431	0.3697	0.5849-4.6697		Go
GSE41271	0.3752	1.6527	0.8082-1.7588		Go
Stage l	0.0460	1.1923	1.0119-3.6844	Poor
Stage ll	0.2885	1.9308	0.1872-1.6460
Stage lll	0.5018	0.5550	0.4432-1.4893
GSE14814	0.4097	0.8125	0.4067-1.4434		Go
Stage l	0.5010	0.7661	0.2710-1.8932
Stage ll	0.0585	0.7163	0.4185-2.0663
GSE13213	0.5543	0.9299	0.6420-2.2847		Go
Stage lll	0.1295	1.2111	0.7902-6.3195
GSE10245	0.5596	2.2347	0.24472.1422		Go
GSE29013	0.5720	0.7240	0.2015-2.4230		Go
Stage l	0.8679	0.6987	0.0916-7.5108
Stage lll	0.7513	0.8295	0.1581-3.7854
GSE11117	0.6797	0.7736	0.4787-3.0952		Go
Stage lV	0.7480	1.2173	0.3764-3.8970
GSE5843	0.6938	1.2112	0.3011-2.2229		Go
Stage l	0.6080	0.8181	0.2301-2.0696
GSE19188	0.6975	0.6901	0.4482-1.7106		Go
GSE4573	0.7531	0.8756	0.5204-1.6039		Go
Stage l	0.9703	1.0146	0.4719-2.1815
Stage ll	0.2390	1.8968	0.6522-5.5165
Stage lll	0.2249	0.4603	0.1315-1.6114
Poepman	0.9458	1.0227	0.5350-1.9551		Go
combined	<0.0001	1.3787	1.2564-1.5130	Poor

Note: cutoff: upper 25% VS other 75%.

LOXL2-related immune cell infiltration

We performed the immune score, stromal score, and tumor purity determinations of LOXL2 (Figure 2A-F), and the results showed that LOXL2 is negatively correlated with an immune score, positively correlated with a stromal score, and negatively associated with tumor purity. Considering that the tumor immune microenvironment plays a significant role in the development of tumorigenesis, we then used the R package CIBERSORT to determine the ratio of 22 TILs in LUAD and LUSC patients (Figures 2G, 2H, S3A, S3B).

Figure 2.

LOXL2 is associated with tumor microenvironment and tumor mutation burden. A-F. Use box plots to show the immune score, stromal score, and tumor purity of LOXL2 on LUAD and LUSC. G, H. Taking the median TMB value as a cutoff, the relative expressions of 22 tumor-infiltrating immune cells in the low- and high-TMB samples were determined (P<0.05, P<0.01, P<0.001, ns, not significant).

LOXL2 was positively correlated with T-reg

We characterized the immunology profile of LUAD and LUSC samples with low LOXL2 and high LOXL2 expression by ssGSEA. Found that LOXL2 in LUAD and LUSC patients is associated with T-reg. GEPIA, to compare LUAD and LUSC tumor tissues, adjacent tissues, and normal lungs (Figure 3A, 3B), found highly expressed T-reg in tumor tissues. In addition, Spearman Correlation Text confirmed the positive correlation between LUAD (Figure 3C) and LUSC (Figure 3D), and T-reg (LUAD, T-reg rho=0.311, P=5.71e-13; LUSC, T-reg, rho=0.494, P<2.2e-16). All these results indicated that LOXL2 might potentially regulate immune infiltration and the response to immunotherapy.

Figure 3.

LOXL2 was positively correlated with T-reg. A, B. Correlation between LOXL2 expression and T-reg in LUAD and LUSC patients. C, D. Associations between LOXL2 expression and immune subtypes in LUAD and LUSC.

Correlated gene expression

Furthermore, we evaluated the association of fifty-two corresponding bases of LUAD (Figure 4A; Table 3) and ten related genes of LUSC (Figure S4; Table 4) with LOXL2. Based on the intersection of the related genes of LUAD and LUSC and LOXL2, we identified eight common genes (Figure 4B): PDGFRA, SEMA7A, FGF5, CMTM3, UCN2, MANF, PDGFC, VEGFC. Among them, LOXL2, SEWA7A, FGF5, UCN2, and VEGFC were positively correlated with LUAD; while LOXL2, SEMA7A, and VEGFC were positively correlated with LUSC (all P<0.05) (Figure 4C).

Figure 4.

SEMA7A and VEGFC positively correlate with LOXL2 in non-small cell lung cancer. A. Correlation analysis of the differential up-regulation and down-regulation of related genes on LUAD and LUSC. B. Venn diagrams of genes related to LOXL2 in LUAD and LUSC. C. The 2 intersection genes that related to OS were extracted by univariate Cox regression analysis.

Table 3.

Association of fifty-two corresponding bases of LUAD with LOXL2

GENE1	GENE2	P	R
LOXL2	PDGFRA	2.97444E-20	0.36848778
LOXL2	GDF11	2.49057E-14	0.308106677
LOXL2	SEMA4F	2.31871E-18	0.350596022
LOXL2	IL1RN	4.10483E-17	0.385049349
LOXL2	CCL21	9.49702E-19	0.35435584
LOXL2	SEMA7A	2.8837E-58	0.599078545
LOXL2	FGF5	3.26866E-26	0.418530278
LOXL2	CMTM3	8.7196E-62	0.613532508
LOXL2	SECTM1	2.27652E-15	0.319714484
LOXL2	HDGF	3.45809E-45	0.537939759
LOXL2	GDF7	1.51068E-17	-0.342533156
LOXL2	UCN2	1.5078E-16	0.371179367
LOXL2	MANF	2.4182E-42	0.522782038
LOXL2	PDGFC	1.18991E-18	0.398061079
LOXL2	ADM	3.65267E-40	0.510629978
LOXL2	VEGFC	3.21155E-45	0.538106606
LOXL2	STC1	2.41538E-69	0.659673055
LOXL2	GIP	6.12859E-12	0.336351246
LOXL2	TOR2A	3.65868E-16	0.32825584
LOXL2	TGFA	5.19121E-15	0.366262391
LOXL2	AIMP1	1.3148E-23	0.397702146
LOXL2	ESM1	1.50638E-46	0.569273722
LOXL2	GREM1	4.3189E-51	0.567128278
LOXL2	SEMA6B	1.78133E-43	0.554551206
LOXL2	CD320	8.90038E-25	0.407236163
LOXL2	LTBP3	1.33309E-16	0.332859725
LOXL2	BMP1	2.52567E-83	0.688311411
LOXL2	SEMA4C	5.06991E-51	0.566800334
LOXL2	CXCL8	7.83508E-26	0.452239551
LOXL2	APLN	4.08522E-20	0.367221915
LOXL2	ANGPTL7	3.20052E-15	-0.318092326
LOXL2	CCL11	3.29126E-16	0.328742125
LOXL2	SAA1	6.0212E-17	0.336430152
LOXL2	VEGFB	2.92379E-25	0.411089378
LOXL2	CLCF1	5.6556E-24	0.400722429
LOXL2	DEFB103B	1.09719E-15	0.323157274
LOXL2	RABEP2	9.27753E-14	0.301500619
LOXL2	BMP8A	4.29787E-12	0.337930853
LOXL2	CSF1	4.10798E-12	0.338131359
LOXL2	JAG2	5.90689E-31	0.453085117
LOXL2	SEMA4B	3.73141E-58	0.618957825
LOXL2	ANGPTL5	7.44501E-11	-0.324968647
LOXL2	SEMA4D	1.59881E-11	0.332036618
LOXL2	FAM3C	1.75315E-35	0.512438649
LOXL2	GMFB	2.65536E-18	0.350020547
LOXL2	NMB	8.22844E-11	0.324502476
LOXL2	S100A6	8.25088E-21	0.415408511
LOXL2	DEFA1	3.48669E-14	0.306432507
LOXL2	CKLF	1.41157E-25	0.413583014
LOXL2	MIF	2.19098E-20	0.412079957
LOXL2	CCL16	1.58955E-15	-0.321414289
LOXL2	CCL3	1.29679E-10	0.322372806

Table 4.

Association of ten related genes of LUSC with LOXL2

GENE1	GENE2	P	R
LOXL2	PDGFRA	5.46926E-10	0.32491428
LOXL2	SEMA7A	7.77056E-30	0.457736861
LOXL2	FGF5	1.305E-30	463270539
LOXL2	CMTM3	4.11249E-30	45972208
LOXL2	CCN1	1.83492E-13	306935752
LOXL2	UCN2	4.35921E-13	0.302312857
LOXL2	MANF	1.4595E-11	0.331374895
LOXL2	PDGFC	1.75505E-23	0.408057926
LOXL2	VEGFC	7.30389E-17	0.393951854
LOXL2	BMP3	3.60237E-13	-0.303338747

Predictive significance of SEMA7A and VEGFC in LUAD and LUSC

The common genes SEMA7A and VEGFC related to LUAD, LUSC, and LOXL2 were used as independent factors for Cox survival analysis (Figure 5A-D). SEMA7A (Log-rank P=0.016) and VEGFC (Log-rank P<0.01) were prognostic factors in LUAD. SEMA7A (Log-rank P=0.0017) and VEGFC (Log-rank P=0.003) were predictive factors in LUSC.

Figure 5.

SEMA7A and VEGFC are prognostic factors in LUAD and LUSC. A, B. The Kaplan-Meier plot of SEMA7A in LUAD and LUSC. C, D. The Kaplan-Meier plot of VEGFC in LUAD and LUSC. E. Protein-protein interaction (PPI) network. Molecules with the highest correlation with LOXL2.

Next, we built a PPI network to understand the mode of interaction of LOXL2 (Figure 5E). The PPI network comprised 31 nodes, showing the relationship between LOXL2, SEMA7A, and VEGFC. The results prove that SEMA7A and VEGFC are prognostic factors of LUAD and LUSC.

Enrichment of LOXL2-correlated gene

GO enrichment analysis in terms of biological processes (BP), cellular components (CC) and molecular functions (MF) revealed that the significant regulatory processes of LOXL2 on BP were axonogenesis, axon guidance, and neuronal projection guidance (Figure 6A). For CC, alterations in LOXL2 most clearly controlled processes in the semaphorin receptor complex and the collagen-containing extracellular matrix (Figure 6B). In the results shown for MF, semaphorin receptor activity was the most responsive to regulation by LOXL2, which was most clearly associated with different sites of regulation (Figure 6C).

Figure 6.

The tasks of LOXL2 and the correlations among their functions. A-D. Bar plot of Go and KEGG functional enrichment analyses. BP indicated biological process; CC indicated cellular component; MF indicated molecular function.

KEGG pathway enrichment of the LOXL2 interactive gene showed that axon guidance, focal adhesion, and PI3K-Akt signaling pathway were enriched pathways (Figure 6D).

Therapeutic targets and mechanisms of drugs

Volcano plots showed the priorities of different drugs in the DREAMT database and after batch correction (Figure 7A). Then we analyzed the marketing status of all oncology drugs (Figure 7B). A large proportion of experimental drugs are approved, mainly including dopamine receptor antagonists, cyclooxygenase inhibitors, serotonin receptor antagonists, glucocorticoid receptor antagonists, adrenergic receptor agonist, adrenergic receptor antagonists, and bacterial cell wall synthesis inhibitor (Figure 7C). Further analyses screened nine drugs: cefuroxime benzocaine, benzocaine, cefazolin, methotrexate, tacrolimus, rimantadine, doxorubicin, cefuroxime, guanethidine (Table 5).

Figure 7.

The therapeutic targets and mechanisms of drugs. A. Volcano plots shows the priorities of different medicines found in the DREIMT database. B. The mechanism of action of tumor drugs on the market. C. Classification of the medications for treating tumors according to their approval status.

Table 5.

Common tumor treatment drugs on the market

Drug_name	Drug_pubchm_id	Summary	FDR	tau	Drug specificity_score	Drug_source_db	Drug_source_name	Drug_status	Drug_moa	Drug_target_gene_names	Drug_target_gene_ids
Cefuroxime	5361202	Cefuroxime boosts case type compared to reference type	0.001652893	99.98135372		LINCS	BRD-K63641886	APPROVED	Bacterial cell wall synthesis inhibitor
Calcifediol		Calcifediol boosts case type compared to reference type	0.001788909	99.98135372	0.55427037	LINCS	BRD-K77175907	APPROVED	Vitamin D receptor agonist	VDR	7421
Benzocaine	2337	Benzocaine boosts case type compared to reference type	0.001841621	99.98135372		LINCS	BRD-K75466013	APPROVED	Sodium channel blocker	SCN10A	6336
Cefazolin		Cefazolin boosts case type compared to reference type	0.001851852	99.98135372		LINCS		APPROVED	Bacterial cell wall synthesis inhibitor	PON1	5444
Mesoridazine	4078	Mesoridazine boosts case type compared to reference type	0.001901141	99.98135372	0.6460514	LINCS	BRD-A14395271	APPROVED	Dopamine receptor antagonist	HTR2A, DRD2	3356, 1813
Tacrolimus	445643	Tacrolimus boosts case type compared to reference type	0.002028398	99.98135372	0.61548928	LINCS	BRD-K65261396	APPROVED	Calcineurin inhibitor	FKBP1A	2280
Rimantadine	5071	Rimantadine boosts case type compared to reference type	0.002325581	99.98135372		LINCS		APPROVED	Antiviral, RNA synthesis inhibitor
Doxorubicin	31703	Doxorubicin boosts case type compared to reference type	0.002421308	99.98135372	0.69624245	LINCS	BRD-A52530684	APPROVED	Topoisomerase inhibitor	TOP2A	7153
Cefuroxime	5361202	Cefuroxime boosts case type compared to reference type	0.001623377	99.96270744		LINCS		APPROVED	Bacterial cell wall synthesis inhibitor
Guanethidine	3518	Guanethidine boosts case type compared to reference type	0.001766784	99.96270744		LINCS	BRD-M18219129	APPROVED	Adrenergic inhibitor	SLC6A2	6530

Drug sensitivity experiments of doxorubicin on LOXL2 expression profile

To further substantiate the predicted drug response to LOXL2 inhibition, we selected doxorubicin, the agent displaying the strongest correlation with our target gene, for subsequent experimental validation. We initially assessed the expression levels of LOXL2 across three commonly used non-small cell lung cancer (NSCLC) cell lines: PC9, A549, and H1299. These expression profiles were quantified via quantitative PCR (qPCR) analysis. As illustrated in the Figure, the H1299 cell line, exhibiting robust LOXL2 expression, and the PC9 cell line, characterized by relatively low LOXL2 expression, were chosen for our drug sensitivity studies (Figure 8A). We then employed a Cell Counting Kit-8 (CCK-8) assay to determine the half-maximal inhibitory concentration (IC₅₀) of doxorubicin at various dose gradients in both the H1299 and PC9 cell lines (Figure 8B, 8C). The IC₅₀ of H1299 was determined to be 2.248 µg/ml. Furthermore, the IC₅₀ of PC9 was 6.026 µg/ml (P<0.05), indicating that the high LOXL2-expressing cell line H1299 exhibited greater sensitivity to doxorubicin than the low LOXL2-expressing cell line PC9.

Figure 8.

Drug sensitivity experiments of doxorubicin on LOXL2 expression profile. A. The expression of LOXL2 mRNA in NSCLC assessed using RT-qPCR (n=12, ***P<0.001). B, C. Cell Counting Kit-8 assay was employed to ascertain the half maximal inhibitory concentration (IC₅₀) of doxorubicin at varying concentration gradients on the H1299 cell line, exhibiting robust LOXL2 expression and PC9 cell line, characterized by relatively low LOXL2 expression (n=3, ***P<0.001).

Discussion

Many patients with non-small cell lung cancer are diagnosed at an advanced stage, and those diagnosed in an early stage often relapse and develop metastatic lesions, despite recent advances in treatment [19]. Tumor immunotherapy has developed rapidly and has attracted increasing attention due to its effectiveness [20-22]. In light of this, our study aims to discover new targets for immunotherapy and predictive indicators for NSCLC patients. We began by the intersecting 500 differentially expressed genes (DEGs) found in both lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC), then perform survival curve analysis on the 11 genes shared between the two groups. Additionally, we compared the expression levels of these 11 genes in para-cancerous and lung cancerous tissues, ultimately identifying LOXL2 as a promising biomarker.

The lysyl oxidase (LOX) family, currently comprising LOXL1, LOXL2, LOXL3, and LOXL4, has been demonstrated to enhance extracellular matrix stability by cross-linking elastin and collagen in the outer matrix [23,24]. Recent studies have demonstrated that LOX also facilitates tumor cell migration and invasion through a number of mechanisms, including the promotion of epithelial-mesenchymal transition (EMT) [25], the activation of the p-FAK/p-paxillin/YAP signaling pathway [26] and the involvement in the formation of a pre-metastatic microenvironment [27]. This has been demonstrated to be a crucial factor in the process of tumorigenesis and metastasis [26,28]. It is anticipated that this will prove to be a promising target for tumor therapy. Among these, lysyl oxidase-like 2 (LOXL2) has been identified as a gene that upregulation promotes tumor infiltration and metastasis [29]. The aberrant expression of LOXL2 in various tumors has been associated with several adverse outcomes, including epithelial-mesenchymal transition (EMT), metastasis, poor prognosis, chemo-radiotherapy resistance, and tumor progression [29,30]. Previous study demonstrated that LOX and LOXL2 are directly regulated by the miR-00/ZEB1 axis, and that LOXL2 might serve as a new target for lung cancer therapy [31]. Subsequently, a number of studies demonstrated a correlation between elevated LOXL2 expression and reduced overall survival, as well as worsening of clinicopathological features of tumors [32]. Furthermore, our findings indicate that LOXL2 is highly expressed in non-small cell lung cancer and is associated with a poor prognosis. Consequently, it is postulated that LOXL2 may represent a novel therapeutic target for the treatment of NSCLC.

Consequently, we calculated the immune, matrix, and tumor purity of LUAD and LUSC and confirmed that LOXL2 was associated with the immune response of NSCLC. The relationship between LOXL2 and T-reg was identified through the analysis of the degree of immune infiltration and the proportion of immune T cells. A positive correlation was observed between LOXL2 and T-reg. Moreover, regulatory T cells (Tregs) have been shown to promote immune suppression in malignant tumors by suppressing the immune response to cancer cells [33]. Treg cells plays a pivotal role in maintaining peripheral tolerance in vivo through the active suppression of self-reactive T-cell activation and expansion. This helps to prevent autoimmune diseases and restrain chronic inflammatory conditions [34]. In patients with early-stage NSCLC, an increased number of circulating and tumor-infiltrating regulatory Tregs are associated with a poorer prognosis and a higher risk of recurrence [35]. In light of these findings, LOXL2 may serve as a promising novel marker for the treatment of LUAD and LUSC, as well as for prognostication.

To further confirmed the prognostic indicative role of LOXL2 in patients with non-small cell lung cancer, we searched for the related molecules of LOXL2 and performed multi-factor Cox survival analysis to identify two prognostic factors SMEATA and VEGFC. SEMA7A, a glycosylphosphatidylinositol-anchored (GPI-anchored) glycoprotein on the plasma membrane. Recent research has shown that FUT8-mediated aberrant N-glycosylation of SEMA7A promotes head and neck squamous cell carcinoma progression [36]. It is a possible therapeutic target for patients with EGFR-TKI-resistant lung adenocarcinoma [37]. VEGFC, a member of the vascular endothelial growth factor/platelet-derived growth factor family, promotes endothelial cell proliferation and angiogenesis. VEGF family consists of seven members, including VEGF-A, VEGF-B, VEGF-C, VEGF-D, placental growth factor (PlGF), and non-human genome encoded VEGF-E and svVEGF [38]. VEGFC promotes tumor growth and metastasis through lymphangiogenesis and lymphatic metastasis, which is mediated by VEGFR-3 [39]. Blocking this pathway leads to apoptosis of lymphatic endothelial cells and disruption of the lymphatic network. Thus, VEGFC is involved in lymphatic metastasis of tumor, which is a feature of poor tumor prognosis [40,41]. These two prognostic factors are further evidence for the prognostic role of LOXL2 in NSCLC.

The “oncoPredict” is an R package for predicting drug responses. It integrates three approaches to (1) correct for overall drug sensitivity for drug-specific biomarker discovery, (2) predict a patient’s clinical drug response, and (3) correlate these predictions with clinical features for in vivo drug biomarker discovery. This new “oncoPredict” R software package can be applied to a variety of in vitro and in vivo drug and biomarker discovery settings [42]. We use the “oncoPredict” R package to analyze the batch-corrected prioritization of different drugs in the DREAMT database. Nine drugs are tested: cefuroxime benzocaine, benzocaine, cefazolin, methotrexate, tacrolimus, rimantadine, doxorubicin, cefuroxime and guanethidine. In order to investigate the relationship between LOXL2 expression and drug sensitivity, we proceeded to identify the compounds with the strongest correlation with LOXL2 levels, which were then subjected to subsequent experiments. The results demonstrated that the H1299 cell line, which exhibited higher LOXL2 expression, exhibited greater sensitivity to doxorubicin compared to the PC9 cell line, which exhibited lower LOXL2 expression. These findings illustrate the potential role of LOXL2 in predicting drug response and emphasize the importance of considering gene expression levels when selecting compounds for therapeutic intervention.

This study has some limitations. Our study is a retrospective, not a prospective analysis, and we analyze the role of LOXL2 in NSCLC by bioinformatics analysis without exploring the mechanism. In conclusion, our results indicate a prognostic role of LOXL2 in NSCLC patients. The analysis of the tumor immune microenvironment suggests the possibility of LOXL2 as a new target for immunotherapy of NSCLC. The results of molecular interactions reveal that SEMA7A and VEGFC may be prognostic factors in NSCLC.

Disclosure of conflict of interest

None.