The landscape of tumor mutational burden and its clinical significance in patients with lung cancer: a Multi-omics study with a meta-analysis


 Background: Current research on tumor mutational burden (TMB) has focused on tumor immunotherapy responsiveness, but the role of TMB in non-immunotherapy patients is unclear. The purpose of this study is to explore the effect of TMB on lung cancer patients in order to clarify and expand the clinical significance of TMB in lung cancer.Methods: We download mutation data of lung cancer cases from The Cancer Genome Atlas (TCGA) database to analyze TMB and its composition, and study the relationship between TMB and clinicopathological characteristics of lung cancer patients. We then systematically retrieved and analyzed studies on the relationship between TMB and survival outcomes. The hazard ratio (HR) and its 95% confidence interval (CI) were used as an effective size to assess the survival outcomes. The subgroup analyses based on the pathological type, treatment method, TMB detection method and detection materials were also performed to explore the factors that might affect the interpretation of TMB results.Results: TMB in lung squamous cell carcinoma is lower than those in lung adenocarcinoma. In lung adenocarcinoma, patients with EGFR mutation have lower TMB than patients with EGFR wild-type. The summary analysis found that TMB is a better prognostic factor in small cell lung cancer, and more evident in small cell lung cancer receiving immunotherapy. TMB is a neutral or poor prognostic indicator in non-small cell lung cancer, but a better prognostic factor in non-small cell lung cancer receiving immunotherapy. In patients with lung adenocarcinoma, including those with EGFR mutation and receiving EGFR-targeted therapies, high TMB means worse survival. TMB detected by blood specimens is inconsistent and unstable compared to TMB detected by tissue. The clinical significance of TMB from blood specimens needs further study on extensive sample data.Conclusions: The pooled results indicated that TMB is a good prognostic factor in lung cancer patients receiving immunotherapy. But high TMB is connected with worse survival in non-small cell lung cancer without receiving immunotherapy, especially in lung adenocarcinoma. For lung adenocarcinoma patients with both EGFR mutation and high TMB, how to make a choice between EGFR-targeted therapy and immunotherapy is still a problem that requires further research.


Background
Lung cancer, including non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC), is the most common malignant tumor and the leading cause of tumor-related death in the world [1]. NSCLC, accounting for approximately 85% of all lung cancer incidences, is mainly composed of lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), and lung large cell carcinoma [2]. SCLC, accounting for around 10-15% of all lung cancers, is characterized by neuroendocrine and has a high degree of malignancy [3]. Different pathological types of lung cancer have different biological characteristics and various treatment options [4][5][6]. The clinical application of targeted therapy has led to a dramatic change in the treatment of patients with LUAD [7,8]. Unfortunately, targeted therapy for oncogenic driver mutations in EGFR or ALK fusions that works on LUAD are generally ineffective against LUSC and SCLC [5,9]. Immunotherapy based on immune checkpoint inhibitors (ICIs) provides de nite e cacy for patients with chemotherapy-resistant lung cancer, and is gradually moving towards rst-line treatment for lung cancer [10]. But there are also some patients who have no response to ICI therapy, especially those containing oncogenic driver mutation [11].
With the development of next-generation sequencing technology, whole-exome sequencing (WES) and targeted next-generation sequencing (targeted NGS) have been used in clinical decision-making [12]. The changes in tumor driver genes such as EGFR, ALK, MET, BRAF, ROS1, HER2 can be detected based on NGS, and those results are used to guide the application of targeted drugs [13]. Tumor mutational burden (TMB), as a product of the era of NGS, is also gradually widely used in clinical practice [14]. Under this circumstance, how to precisely understand the meaning of TMB and use it to manage lung cancer is of great practical signi cance.
In addition to PD-L1 expression, microsatellite instability (MSI), and mismatch repair de ciency (MMR), TMB also has emerged as a promising biomarker for response to ICIs therapy in clinical trials [15]. The hypothesis that tumors with more neoantigens may respond better to immunotherapy is based on the positive relationship between tumor-speci c neoantigens and increased immunogenicity [16]. Most current researches focus on the relationship between TMB and the e cacy of immunotherapy in various cancers [17][18][19]. However, the role of TMB in lung cancer patients without receiving immunotherapy has not been fully elucidated. TMB is counted by the sum of non-synonymous mutations that also contain the oncogenic driver mutations, but the additional mutations may constitute potential resistance pathways to targeted therapies [11]. In order to expand the interpretation of the potential clinical signi cance of TMB, we analyzed the composition of TMB in lung cancer, and then reviewed the role of TMB in lung cancer patients with different characteristics.

Data acquisition and preprocessing
First, we download the somatic mutation data of patients with LUAD and LUSC respectively from "simple nucleotide variation" category in The Cancer Genome Atlas (TCGA) database (http://portal.gdc.cancer.gov/). The "Masked Somatic Mutation" data processed by VarScan2 software was used for further TMB analysis. Then, the transcriptome data with HTSeq-FPKM format of patients with LUAD and LUSC were download from "transcriptome pro ling" in TCGA database. Moreover, the clinical information of LUAD and LUSC patients, including age, gender, TNM stages, and survival status were all obtained from the TCGA database. Next step, all transcriptome data and clinical data were extracted from the corresponding single les and merged. without TMB-related survival data, and 4) studies in non-lung cancer. For multiple articles reporting a duplicate population, only the most complete or recent one was included. Two reviewers independently performed the selection process of eligible studies, and differences were resolved through discussion.

Data Extraction And Quality Assessment
Two reviewers independently collected the following data using a predesigned form: rst author, publication time, study period, country, sample size (low-and high-TMB), age (low-and high-TMB), gender (low-and high-TMB), tumor histology (low-and high-TMB), tumor stage (low-and high-TMB), driver mutation, treatment, sample source, TMB detection method, cutoff value, follow-up time, and primary endpoint. Disputes in this process were handled through discussion. The hazard ratio (HR) and 95% con dence intervals (CI) of the high TMB group compared to the low TMB group for OS and progression-free survival (PFS) were primarily collected. Disease-free survival (DFS) was considered as PFS for further analysis due to the similar clinical signi cance of DFS and PFS. For studies in which the HR and 95% CI were not provided explicitly, we used Tierney's methods to extract survival data from the Kaplan-Meier curve or the original data [22]. If the above item could not be obtained in the original study, this item was marked as "not acquired (NA)". Two reviewers independently assessed the quality of included cohort studies based on the Newcastle-Ottawa Scale (NOS) [23]. The NOS evaluated the quality of a survey with a scale ranged from 0 to 9. Studies with a score of 6 or high were deemed as high-quality studies.

Statistical Analysis
The HR and its 95% CI were used as effect size (ES) to assess the role of TMB in the survival of lung cancer patients. Heterogeneity across the studies was evaluated using I-squared statistics [24]. A randomeffect model was utilized if I 2 greater than 50%; Otherwise, a xed-effect model was chosen. To detect the different prognostic effects of TMB in different pathological types of lung cancer, various treatments, different methods of TMB detection and diverse sampling, we performed the subgroup analyses based on different scenarios. To con rm the robustness of the pooled results, sensitivity analyses by eliminating a single study at a time were adopted. Potential publication bias was assessed by Begg's funnel plot and Egger's test. The statistical analyses were conducted using the software STATA version 12.0 (Stata Inc, TX, USA). All the tests were two-sided, and a P value less than 0.05 was considered to be signi cant.

Relationship Between Tmb And Clinicopathological Characteristics
The clinicopathological characteristics of LUAD and LUSC patients were summarized in Table 1. In the LUAD cohort, the age of the low TMB group was higher than that of the high TMB group (P = 0.003).
There were more women in the low TMB group and more men in the high TMB group (P = 0.014). In the LUSC cohort, the age of the low TMB group was also higher than that of the high TMB group, but the statistical signi cance was not signi cant (P = 0.061). The proportion of stage I patients is higher in the low TMB group, and the ratio of stage III-IV patients in the high TMB group is higher (P = 0.004). Kaplan-Meier analysis with log-rank test showed that TMB did not signi cantly affect OS in LUAD (  To explore the possible e cacy of ICI immunotherapy in EGFR mutation-driven lung cancer patients, we examined the relationship between EGFR mutation status and two predictive indicators commonly used in immunotherapy, TMB value and PD-L1 expression. Patients with EGFR mutations in LUAD had signi cantly lower TMB values than those with EGFR wild type (

Search results and study characteristics
The literature selection process was shown in Fig. 7. After removing duplicates, a total of 24229 records were retrieved by the search strategy above. By viewing titles and abstracts, 24115 irrelevant articles were excluded. The remaining 114 articles were read in full-text. Finally, 30 articles were included in this analysis.
Thirty eligible articles included 37 independent cohort studies. The main characteristics of the 37 studies were shown in Table 2. Thirty-one studies report the role of TMB in NSCLC and six studies indicate the role in SCLC. Based on the NOS score, these 37 studies have a quality of 6 to 9, with an average score of 7.2.

Meta-analysis Results
A total of 22 articles with 28 studies evaluated OS, and 22 articles with 25 studies evaluated PFS. The HRs and 95% CI were pooled using the random-effects models throughout the analysis. On the whole, high-TMB group lung cancer had better PFS compared to low-TMB group (Fig. 8B, P = 0.001). However, no signi cant difference in OS was found between the high TMB group and the low TMB group (Fig. 8A, P = 0.830).
The above total lung cancer patients include different types of lung cancer, different treatment methods, different TMB detection methods, and different test samples. To explore the impact of these factors on results, we conducted a series of subgroup analyses. In a subgroup analysis based on whether to receive ICIs, we found that among patients receiving ICIs, the high TMB group had better OS (Fig. 9A, P < 0.001) and PFS (Fig. 10A, P < 0.001). However, in the non-ICI subgroup, the OS in the high TMB group was worse compared with the low TMB group (Fig. 9A, P = 0.014), and there was no signi cant difference in PFS between the two groups ( Fig. 9A, P = 0.464). In the SCLC subgroup, the OS (Fig. 9B, P = 0.052) and PFS ( Fig. 10B, P < 0.001) in the high TMB group are better than those in the low TMB group, but in NSCLC subgroup, there is no signi cant difference between the high and low TMB groups ( Fig. 9B and Fig. 10B).
NSCLC was further divided into LUAD and LUSC. The results showed that in the LUAD subgroup, the OS ( Fig. 9C, P < 0.001) and PFS (Fig. 10C, P = 0.051) of the high TMB group were worse than the low TMB group. However, in the LUSC subgroup, there was no signi cant difference between the high and low TMB groups ( Fig. 9C and Fig. 10C). Considering that immunotherapy has a greater impact on the clinical prognosis of TMB, we combined the tumor pathological type and ICIs for subgroup analysis. Results showed that in the NSCLC subgroup receiving ICIs, high TMB was associated with better OS (Fig. 9D In the subgroup analysis based on the detection method, we found that targeted NGS and WES did not signi cantly affect the indicator role of TMB in OS (Fig. 11A). However, different detection methods had a greater impact on the prognosis of TMB in PFS. In the targeted NGS subgroup, high TMB predicted higher PFS (Fig. 12A, P < 0.001). In the WES subgroup, TMB does not show an indicator effect (Fig. 12A). In the subgroup analysis of detection methods combined with immunotherapy, whether using the targeted NGS or WES methods, in the subgroup receiving ICI, high TMB tended to predict a better prognosis. Still in the subgroup not receiving ICI, high TMB tended to have a poor prognosis ( Fig. 11B and Fig. 12B). Then we performed a subgroup analysis based on different test samples. In the subgroup tested using blood samples, the high TMB group had worse OS than the low TMB group (Fig. 11C, P = 0.027). In the subgroup detected using tumor tissue samples, the high TMB group had better PFS than the low TMB group (Fig. 12C, P = 0.001). Subgroup analysis was further performed by combining test samples and immunotherapy. The results showed that high-TMB had a better prognosis in the subgroup that used tissue detection and received ICIs (OS, Fig. 11D, P < 0.001; PFS, Fig. 12D, P < 0.001). However, in the subgroup that provided tissue testing but did not receive ICI, high TMB tended to have a worse prognosis (OS, Fig. 11D, P = 0.045).
Machael O n et al. studied EGFR exon19del or L858R mutant LUAD treated with EGFR-TKIs [11]. By using EGFR wild-type LUAD as a control group, they found that the TMB value of EGFR-mutated lung adenocarcinoma was lower than that of EGFR wild-type lung adenocarcinoma patients, and TMB was a poor prognostic factor in metastatic EGFR mutant LUAD treated with EGFR-TKIs [11]. Yanhui Chen et al.
enrolled into two cohorts of LUAD and LUSC, and found that the TMB value in LUAD was lower than that in LUSC [26]. There were only two patients with EGFR mutations in the LUSC group, accounting for 4%. In the LUAD group, there were 58 patients with EGFR mutations, accounting for 43%. In EGFR wild-type LUAD, TMB did not correlate with survival outcomes (P = 0.484). Still, in mixed EGFR mutant and wildtype patients, TMB was closer to a worse prognostic factor (P = 0.062), which means that TMB tended to negatively related to survival in patients with EGFR mutant LUAD [26].

Publication Bias And Sensitivity Analysis
According to Begg's and Egger's test, the publication bias of OS (Begg's P = 0.767; Egger's P = 0.765) and PFS (Begg's P = 0.528; Egger's P = 0.591) was not signi cant. By omitting each study one by one, no single study with a signi cant impact on the combined results was found, which means that the metaanalysis results were reliable.

Discussion
With the development of sequencing technology, TMB has become an essential clinical indicator [28,29].
Currently, TMB is mainly studied as a predictive biomarker for ICIs treatment response [15,30]. Recent studies have concluded that in patients with higher TMB, the survival bene t of patients receiving ICIs is better than that of patients receiving chemotherapy alone, but in patients with low TMB, the survival bene t of ICIs is not statistically signi cant [17,18]. In this study, we did not group patients according to TMB value, but grouped according to lung cancer types, treatment methods, detection methods, test samples. We found that high TMB in the immunotherapy group was an excellent prognostic indicator, which corresponded to high TMB as a marker of immunotherapy responsiveness. But in NSCLC patients who receive chemo-radiotherapy or targeted therapy but not ICIs treatment, high TMB is associated with a worse prognosis, which suggests that TMB itself may be a poor prognostic indicator in NSCLC.
According to the type of lung cancer, TMB is a good prognostic indicator in SCLC, but a poor prognosis indicator in NSCLC. Immunotherapy enhances the role of TMB in SCLC to indicate better prognosis, but reverses the role of TMB in NSCLC, making TMB from a poor prognostic indicator to a better prognostic indicator. The bene cial effect of high TMB on long-term survival of NSCLC and SCLC patients treated with ICIs validates the therapeutic value of immunotherapy for patients with high TMB [31,32]. In LUAD, including EGFR mutated and EGFR-TKIs treated LUAD, high TMB is a poor prognostic indicator.
Fortunately, it was found in this study that TMB values are usually lower in LUAD patients, especially those with EGFR mutations. However, for LUAD patients with EGFR mutations and high TMB, whether targeted therapy is adequate and whether additional or alternative immunotherapy is needed should be further studied [33].
In terms of different detection methods, the detection results of WES and targeted NGS are consistent, and have no signi cant impact on the clinical effect evaluation of TMB. However, the TMB value detected by WES is higher than that detected by targeted NGS, which may be because the WES method contains more mutation sites. In clinical applications, targeted NGS is obviously more suitable for clinical work than WES because it saves testing resources and time [34]. At present, different types and quantities of gene panels are used for targeted NGS in different institutions, which results in various TMB benchmark values detected by various institutions. Exploring gene panels suitable for speci c tumors in speci c populations will help promote the better clinical application of TMB [35].
Compared with circulating tumor DNA (ctDNA) testing, tissue testing is still the current mainstream method [36]. Blood testing has the advantage of convenient material collection, but the testing technology needs further investigation [37]. In this study, a total of three articles with four studies [26,38,39]

Conclusion
In this study, we found that TMB is an excellent prognostic indicator in lung cancer patients receiving immunotherapy. Still, it is a poor prognostic indicator in patients with lung cancer receiving traditional chemo-radiotherapy or targeted therapy. EGFR mutation status in patients with lung adenocarcinoma is associated with lower TMB, but some patients have both EGFR mutation and high TMB status. How to choose immunotherapy or targeted therapy for lung adenocarcinoma patients with both EGFR mutation and high TMB needs to be further studied in well-designed randomized controlled clinical trials.

Declarations
Ethics approval and consent to participate Not applicable. This is a secondary analysis of gene sequencing data and literature.    Kaplan-Meier survival curves with log-rank tests of high TMB and low TMB groups in LUAD (A) and LUSC (B).

Figure 5
Relationship between EGFR mutation status and TMB value in LUAD (A) and LUSC (B).

Figure 6
Relationship between EGFR mutation status and PD-L1 expression in LUAD (A) and LUSC (B).

Figure 7
Flow chart of searching the relevant studies included in this meta-analysis.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.