Integrative and Comparative Genomic Analysis and the Immune Microenvironment Features of Lung Cancer Patients with Tuberculosis

Background: Limited information was known because of the low incidence of co-existence with lung cancer and tuberculosis (TB), it remains special challenging populations for clinical management of cancer immunotherapy. Thus, to investigate the difference on tumour immune microenvironment and genomics between patients with LC alone and LC patients with TB is urgently needed. Methods:Tumour specimens were collected from 87 patients who had LC, with or without TB, at two medical centres. Immunohistochemistry was used to evaluate PD-L1 expression and CD3+/CD4+/CD8+ T-cell inltration. Whole-exome sequencing was performed using samples from 19 patients with LC&TB and 21 patients with LC. Results:Relative to patients with LC alone, patients with LC&TB had lower PD-L1 expression and CD4+/CD8+ T-cell inltration (all P< 0.001). A tumour microenvironment with no PD-L1 expression and CD8- T-cell inltration was most common in the LC&TB patients. Genomic alterations analysis revealed an increased mutation frequency among patients with LC and active TB, obsolete/cured TB, or no TB in terms of the TP53 (23.08% vs. 66.67% vs. 76.19%, P = 0.01), while a decreased trend of the number of single-nucleotide variants/insertions/deletions (P< 0.001), tumour mutation burden (P< 0.001), and number of neoantigens (P< 0.001). Patients with LC&TB had a higher frequency of a specic mutation signature (32.99% vs. 11.23%), as well as potential driver mutations involving the complement C1qB chain (C1QB) mutations. Conclusion: The present study revealed signicant differences in the tumour microenvironment and genomic alterations between patients with LC&TB and patients with LC alone.


Background
Lung cancer (LC) remains the major cause of cancer-related death, with an estimated 1.59 million deaths in 2016, which accounts for approximately 20% of all cancer-related deaths worldwide (1).
Epidemiological research regarding LC has revealed that the major risk factors are smoking, air pollution, occupational factors, and tuberculosis (TB), which will provide the basis for future interventions to improve LC management (2)(3)(4).
Mycobacterium tuberculosis (MTB) is the pathogen that causes TB, which is the deadliest infectious disease in the world today (5,6). Infection by MTB affects approximately one-third of the global population, and insu cient treatment or compromised immune defences lead to the development of TB in 5-10% of infected patients (5,6). Most MTB infections have no symptoms, although TB-induced in ammation typically leads to genetic changes over time, which can drive the development of LC (3).
Thus, patients with TB have a signi cantly increased risk of developing LC and related mortality (3). The increased risk of LC is related to the immunosuppressive state caused by MTB infection (7)(8)(9), and coexistence of LC and TB (LC&TB) has been reported in numerous cases and case-control studies (3,10,11).
Immunotherapy has become a mainstream treatment option for various cancers, including LC and melanoma (12)(13)(14)(15)(16)(17), in addition to surgery, radiotherapy, and chemotherapy. Immune checkpoint inhibitors are a type of immunotherapy that activate T-cell responses in the tumour microenvironment and can enhance the immune system's response to the tumour.
Patients with LC&TB are rare and thus often overlooked in clinical trials, with only limited global data regarding immunotherapy in this patient population (18). Thus, it is unclear whether immune checkpoint inhibitors including anti-PD-1/PD-L1 therapy are contraindicated in patients with LC&TB. Furthermore, the T-cell response plays a key role in the development and progression of TB (19), and different immune microenvironments of LC patients with and without TB might explain their potentially different treatment responses. Therefore, the present study aimed to evaluate the tumour microenvironment characteristics of LC patients with and without TB, including their PD-L expression and tumour-in ltrating lymphocytes (TILs). Furthermore, we evaluated differences in somatic mutations, copy number variations, tumour mutation burden (TMB), HLA expression, and neoantigen numbers according to the presence or absence of TB.

Patient population
The study was approved by the ethics committees of Zhejiang Cancer Hospital and A liated Hangzhou Chest Hospital. This study evaluated data from 87 patients with stage I-IV LC with TB, who were treated between 2013 and 2019 at two Chinese medical centres (Zhejiang Cancer Hospital and A liated Hangzhou Chest Hospital). The eligibility criteria were as follows: LC con rmed via pathological or cytological evaluation, complete clinical and follow-up data, availability of ≥ 5 blank sections from tissue specimens, and no history of malignancy within the previous 5 years. The presence of active TB (ATB) was con rmed based on 3 sputum samples for mycobacterium culture, nucleic acid ampli cation test and/or lung biopsy (if feasible) (20), while the obsolete TB (OTB) was con rmed based on the patient's medical records, the T-cell spot test (TSPOT) or interferon gamma release assay (IGRA), and imaging ndings. Cured TB (CTB) was de ned as "history of active TB and have been cured when cancer diagnosed". Of note, latent TB was not included in our study. The control group contained stage I-IV LC patients from Zhejiang Cancer Hospital. The study was approved by the ethics committee of Zhejiang Cancer Hospital and A liated Hangzhou Chest Hospital.
For genomic analysis, 21 LC&TB patients, treated with surgery, with more than 15 white para n tablets was selected. During 2013-2019, specimens from 192 LC cases had been collected in the hospitals' biobank and we performed propensity score analysis to select 21 paired samples from patients with LC alone. These samples, and normal control tissues, were subjected to whole-exome sequencing (WES) and analysis of copy numbers and somatic mutations. These cancer-speci c genomic alterations were then evaluated for associations with the patients' clinical characteristics. The study owchart is shown in 2.2 DNA extraction, exome sequencing, and data processing After omitting 2 samples from the LC&TB group because of poor quality, DNA was extracted from the remaining 40 formalin-xed para n-embedded (FFPE) samples using the QIAamp FFPE DNA Kit (Qiagen, Frankfurt, Germany). The DNA was then fragmented using a Covaris M220 focused ultrasonicator (Covaris, Woburn, USA) and subjected to sequencing library construction. The DNA for WES was captured using the VariantBaits™ Human All Exon Kit (LC-Bio) according to the manufacturer's recommended protocol. Sequencing was performed using the Illumina NovaSeq™ 6000 system to generate 150-bp paired-end reads, and FASTQ software (version 0.20.0) (21) was used to remove low-quality reads and clean the data. The supplementary materials (available online) show the methods for somatic mutation calling, identi cation of somatic copy number alterations, identi cation of potential driver genes, extraction of mutation signatures, and prediction of HLA types and neoantigens.

Immunohistochemistry
Immunohistochemistry was used to evaluate the expression of PD-L1 and the presence of tumourin ltrating CD3 + T-cells, CD4 + T-cells, and CD8 + T-cells. We prepared FFPE tissue sections (4-µm) on positively charged glass slides, which were stained using primary antibodies targeting CD3 (SP162,  results were judged by two pathologists who were blinded to the patients' clinical characteristics, and disagreements were resolved via discussion. The expressions of PD-L1 and T-cell markers were used to create different subtype groups, and differences between the LC and LC&TB groups were displayed using stacked bar graphs.

Statistical analysis
All data were expressed as mean ± standard deviation or number (percentage), unless otherwise indicated. Categorical variables were compared using Fisher's exact test. The Kruskal-Wallis test was used to compare clinicopathological factors, gene alterations, and TMB. Paired groups were compared using the Wilcoxon test, and comparisons of more than two groups were performed using the Kruskal-Wallis chi-squared test. Linear correlations were evaluated using Pearson's correlation coe cient. Most statistical analyses were performed in the R statistical environment (version 3.5.1 or later) and the "survival" package was used to generate and compare the Kaplan-Meier survival curves. The survival event was de ned as patients died with lung cancer. GraphPad Prism software (version 8.4; GraphPad Software, San Diego, CA, USA) was used to create the stacked bar graphs. Multivariable survival analyses were performed using the Cox proportional hazards model and propensity score matching analysis was performed using SPSS software (version 24; IBM Corp., Armonk, NY, USA). Differences were considered statistically signi cant at two-sided P-values of < 0.05.

Clinicopathological characteristics
The study included 59 patients with LC&TB (33 patients with ATB and 26 patients with O/CTB) and 28 patients with LC alone. The groups' clinicopathological characteristics are summarized in Table 1 and Supplementary Table S1 (available online). There were no signi cant inter-group differences in terms of sex, smoking status, histology, pathological T status, lymph node metastasis, distant metastasis, clinical stage, differentiation degree, venous/lymphatic/perineural invasion, adjuvant radiotherapy, adjuvant chemotherapy, or family history of cancer. Most patients had non-small-cell lung cancer (NSCLC, adenocarcinoma or squamous cell carcinoma), although the LC&TB group included 1 patient with adenosquamous carcinoma, 1 patient with large-cell carcinoma, and 1 patient with small-cell carcinoma. No signi cant differences in EGFR mutation status were observed between the LC&TB and LC groups. All tested samples were negative for ALK mutations, although related testing was not performed for 20% of the patients because of insu cient samples. Table 1 The clinicopathological characteristics of entire patient cohort (divided into two groups  Supplementary Figures S1-4). Moreover, the predominant phenotype in the LC&TB group involved no PD-L1 expression and CD8 − T-cell in ltration (P < 0.001), while the predominant phenotype in the LC group involved PD-L1 expression and CD8 + T-cell in ltration ( Fig. 2A-C).

Genomic alterations
To reduce the effects of selection bias, propensity score matching analysis was performed to create 21 pairs of patients with LC&TB or LC alone, who were matched according to sex, age, smoking status, clinical stage, and pathological type (adenocarcinoma or non-adenocarcinoma) (Fig. 1). The samples from these patients were subjected to WES, although 2 samples from the LC&TB group were excluded because of poor DNA quality. The paired patients' clinicopathological characteristics are shown in Table 2. Figure 3 shows the somatic mutations, clinical features, potential driver genes, 30 most commonly mutated genes, and genes that are related to the e cacy of immune checkpoint inhibitor treatment. The most common alterations in the LC&TB group were mutations in ZNF208 (74%), FLG (68%), MUC17 (68%), ZNF729 (63%), and HRNR (58%) (Supplementary Figure S2A). The most common alterations in the LC group were mutations in TP53 (76%), TTN (67%), RYR2 (43%), KMT2D (38%), and MUC4 (38%). The differential analysis results for somatic mutations in the LC&TB and LC groups are listed in Supplementary Table S4. There were no signi cant differences between the LC&TB and LC groups in terms of somatic mutations in EGFR, PIK3CA, KRAS, and BRAF, as well as CDKN2A copy number variations (CNVs) (all P > 0.05), (Supplementary Figure S6). However, a signi cant difference in the average number of single-nucleotide variants/insertions/deletions was observed between the LC and LC&TB groups (P = 0.002), as well as between the LC&ATB (n = 1,131.08), LC&O/CTB (n = 1,143.83), and LC (n = 633.95) groups (P < 0.001) (Fig. 4A). The most common base substitution was C > T in all three groups, with frequencies of 34.4% in the LC&ATB group, 35.41% in the LC&O/CTB group, and 35.27% in the LC group, although the differences were not statistically signi cant (all P > 0.05) (Fig. 4B). The average number of CNVs per sample were 41.84 in the LC&ATB group, 38.67 in the LC&O/CTB group, and 36.67 in the LC group, although the differences were not statistically signi cant (Fig. 4C). A heatmap of the gain/loss CNVs is shown in Supplementary Figure S7A and a heatmap of the 50 most commonly affected genes in the LC&ATB and LC groups is shown in Supplementary Figure S7B. There were also no signi cant differences in the numbers of CNVs in the CD274 and CDKN2A genes (Supplementary Figure  S6H). The results of the GO and KEGG pathway analyses are shown in Supplementary Figure S7C-D, which revealed that the top functional clusters in the LC&TB group involved cancer-related pathways.
Mutation signatures were rst described in 2013 (23) as groups of gene mutations that are related to malignant processes in tumour cells. We performed mutation signature analyses for each group, which revealed mutation signature 4 in the LC&TB and LC groups. Cluster analysis also revealed that the proportion of samples with mutation signature 1 was higher in the LC&TB group than in the LC group (32.99% vs. 11.23%) (Supplementary Figure S8A-C). Mutually exclusive and co-occurring gene pairs are shown in Supplementary Figure S8D. 3.5 TMB analysis, HLA analysis, and neoantigen prediction The LC&ATB and LC&O/CTB groups were characterized by a high average TMB (18.65 mutations/Mb and 18.86 mutations/Mb), which was noticeably higher than the average TMB in the LC group (10.45 mutations/Mb) (Fig. 5A). Furthermore, a high TMB was signi cantly associated with a mutation signature that suggested exposure to cigarette smoke in the LC group (P = 0.036) and the LC&TB group (P = 0.025) (Fig. 5C). Interestingly, a lower TMB was observed in LC&TB patients with a low TP53 mutation frequency (P = 0.027, Fig. 5D). The average numbers of neoantigens were 1,146.08 in the LC&ATB group, 888.33 in the LC&O/CTB group, and 395.29 in the LC group (P < 0.001, Fig. 5E), and the number of nonsynonymous mutations was correlated with the number of neoantigens (r = 0.74, P < 0.01, Fig. 5G). Expression of PD-L1 was not signi cantly correlated with the TMB among all patients with WES results (r = 0.71, P = 0.258, Fig. 5H). The three most common HLA types were identi ed in each group, and the results revealed heterozygous genotypes for all HLA class I subtypes (A, B, and C) in the LC&ATB, LC&O/CTB, and LC groups (P > 0.05, Supplementary Table S5).
We also found several potential driver genes in patients with LC&TB (Supplementary Materials).

Discussions
Preclinical studies have indicated that immune homeostasis in TB is regulated via immune checkpoint pathways, such as the PD-1/PD-L1 axis. For example, in animal models, the PD-1 pathway plays a key role in controlling excessive in ammation after MTB infection and regulating the resulting immune response (24). Furthermore, the PD-1/PD-L1 pathway suppresses the accumulation of CD4 + T-cells and IFN-γ production, which is an essential part of the immune response to TB. However, treatment that targets PD-1/PD-L1 may cause CD4 + T-cells to overproduce IFN-γ, which can aggravate TB or cause TB recurrence (24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35). Another study has indicated that a TB antigen can inhibit the Th1 immune response and promote LC metastasis via the PD-1/PD-L1 signalling pathway (36). However, there are limited data regarding the immune microenvironment in patients with LC and previous/current TB, which highlights the need for additional information regarding PD-L1 expression and TILs in the tumour and surrounding microenvironment. Our ndings revealed that, relative to the LC group, the LC&TB group had signi cantly decreased PD-L1 expression and less CD3 + /CD4 + /CD8 + T-cell in ltration, which suggests that these patients have immunologically cold tumours in a nonin ammatory microenvironment (Fig. 6A). To the best of our knowledge, this is the rst comprehensive analysis of the tumour microenvironment landscape in LC&TB and LC.
Previous studies have suggested that PD-L1 expression, TP53 mutation frequency, TMB (37), and HLA molecules may be biomarkers for predicting the response of LC to anti-PD-1/PD-L1 therapy (38-40), although these relationships remain controversial. In our study, the LC&TB group had a markedly higher TMB than the LC group, and the LC&TB group also had a signi cantly lower TP53 mutation frequency. The present study revealed that the LC&TB and LC groups only had heterozygous HLA I genotype, and thus we did not perform any additional analyses. Interestingly, we also observed that the number of nonsynonymous mutations was correlated with the number of neoantigens, which is consistent with previous research (41). Moreover, the TMB was signi cantly associated with smoking history among patients with LC&TB and LC alone, which also agrees with previously reported results (42). These results suggest that there are signi cant differences in terms of genomic alterations and mutation signatures between the LC&TB and LC groups.
A few studies have attempted to identify potential driver genes in patients with LC who had a history of TB (43,44). Adenocarcinoma was the main pathological type of LC among our patients with LC&TB, which is consistent with previously reported results (44). Although, patients with lung adenocarcinoma who have a history of lung scarring or TB are more likely to develop EGFR mutations, relative to patients with conventional lung cancer (44). Nevertheless, the present study failed to detect a signi cant difference between the LC&TB and LC groups in terms of the EGFR mutation frequencies.
Notably, there were potential gene mutations differences between ATB and O/CTB in present study. For instance, The C1QB gene encodes the B-chain polypeptide of complement subcomponent C1q. The complement pathway is an important part of the immune system, and complement-mediated bacteriolysis and cytolysis are important mechanisms in the response to infection by pathogenic microorganisms. We observed that it was mutated at a higher rate among patients with LC&TB (5/19 cases) than among patients with LC alone (0/21 cases). In addition, bioinformatics analysis predicted that C1QB c.274_311del would have a substantial effect on protein expression. Furthermore, the C1QB gene may in uence macrophage and dendritic cell in ltration of NSCLC. Therefore, C1QB may be a driver gene that leads to the developing of LC in patients with lung scarring or TB. In addition, KRAS mutation differences are highly signi cant between ATB and O/CTB (P = 0.003 in Supplementary Figure S6D), but when combined as a single group, there is no difference between LT&TB and LC alone. However, due to the limited sample size of these two groups, it is di cult to draw a clear conclusion. Further studies with large sample size and mechanism investigation are urgently needed.
Because a nonin ammatory microenvironment we have observed in LC&TB and the possibility of reactivation/exacerbation of TB by anti-PD-1/PD-L1 therapy, which suggests that there is a need for new approaches to treating patients with LC&TB. We propose three potential strategies for treating patients with LC&TB (Fig. 6B)