Through segmented linear regression analysis, the low and high-TMB thresholds in LUAD were 25 mutations and 667 mutations. In other words, there were less than 25 mutations in the UL-TMB samples while the UH-TMB samples had over 667 mutations in their whole exome sequencing regions. For LUSC, the break points were 60 and 478 mutations, respectively (Supp. Fig. S1). This study presented a comprehensive comparison of the immune features of UL- and UH-TMB lung cancers. First of all, we analyzed the distribution of mutations on genes in different groups. It was easy to understand that the important cancer-related genes such as TP53 and KRAS were both highly mutated in the UL and UH samples. However, there was a difference in the ranking of top mutated genes. Besides, the spectrum of mutations in LUAD also differed from that in LUSC. As shown in the Figure 1, for LUAD, in the UL-TMB group, mutations occurred in no more than 20% of the samples, with the median as merely 8%. As to the UH group, the top 10 mutated genes were found in over 73% of the samples and there was a higher degree of consistency in the patterns of mutation spectra. Among the top 10 mutated genes in the UL group, the known proto-oncogenes included EGFR and KRAS and the known tumor suppressor genes included TP53 and KMT2D. The SCN8A, GRIK5, and FAT3 genes in the UL-TMB LUAD group were highly mutated compared to the corresponding mutation rates in the UH group, and the differences were statistically significant.
Likewise, mutations were found in no more than 30% of the UL-TMB LUSC samples (see Figure S2) whereas about 90% of the UH group underwent mutations. However, compared to the UH group, no gene had a significant increased mutation rate in the UL group.
2. Driver mutations and driver genes.
There are studies showing that about 10 mutations can lead to the formation of tumors. So, is there a higher proportion of driver mutations in low-TMB tumors? To find out the answer, the algorithm Oncodrive-MUT was used in this study to predict the driving force of carcinogenic mutations. With extensive tumor genome data taken into consideration, the algorithm predicted the carcinogenicity of the mutations based on 6,792 samples of 28 cancer types and massive samples from healthy donors. This is combined with the features that describe the functional consequences of genetic mutations. As a result, the mutations were classified as driver and passenger mutations. In this study, for LUAD, driver mutations accounted for 4.8% of all 454 mutations in the UL-TMB group. These driver mutations were derived from the following genes: EGFR, GNAS, TP53, KRAS, RPS6KA3, HCFC1, TSC1, BRAF, SYNCRIP, MED23, FAT1, SRGAP3, SETD2, ZFP36L2, CAD, COL1A1, TP53, and ATR. In the UH-TMB LUAD group, 2.1% of the 47,908 mutations were driver mutations, which originated from a total of 422 genes. The UL group had a significantly higher proportion of driver mutations. In terms of the LUAD samples, the following genes were associated with the driver mutations that occurred in the UL group, including lRPS6KA3, TSC1, SYNCRIP, ZFP36L2 and ATR. TSC1 is a tumor suppressor in the mTOR signaling pathway, which is inactivated by mutation or deletion in a diverse range of cancers. Germline and somatic TSC1 mutation is ais feature of the disease tuberous sclerosis complex. ATR is a tumor suppressor gene involved in DNA damage repair, which is mutated in various cancer types. These UL specific driver genes interfered with the pathway such as TOR signaling pathway which control cell growth and proliferation, cell aging, and DNA replication signaling pathways (see the Figure 2A and Table S1).
For LUSC, driver mutations in the UL-TMB samples were derived from the TP53, BRCA1, CDC73, HERC2, HGF, MAP2K1, MGA, MYD88, NF1, NFE2L2, PIK3C2B, POT1, RB1, SPTAN1, TAOK2, TJP1, TNPO1, and TRAF3, among which MAP2K1, MYD88, PIK3C2B, SPTAN1, TNPO1, and TRAF3 were driver genes only found in the UL group. Particularly, MAP2K1 is an oncogene and intracellular kinase that is mutated at low frequencies in a variety of cancer types, including melanoma, colorectal and lung cancers; MYD88, an oncogene and adaptor protein, is frequently altered in hematologic malignancies including Waldenström’s macroglobulinemia; TRAF3 is a tumor suppressor, signaling molecule, and E3 ligase, which is recurrently mutated and deleted in B cell lymphoma and multiple myelomas. These genes are involved in the pattern recognition receptor signaling pathway, innate immune response-activating signal transduction, toll-like receptor signaling pathway, and regulation of innate immune response, as shown in the Figure 2B and Table S1.
Mutations were classified as either early or late based on their clonal status. It was found that in the UL TMB LUAD and LUSC groups, mutations largely occurred following subcloning or changes in chromosome number. In other words, most mutations occurred at the later stages of tumors; as to the UH groups, the opposite was the case (Figure S6).
3. Pathways dis-regulation related to mutation burden.
Generally, in addition to driver genes, there are plenty of infrequently mutated genes that may contribute to tumor biology by affecting the expression levels of relevant genes. Hence, it is interesting to analyze pathways and networks involved and illuminate the biological processes involved. In this study, Gene Set Variation analysis (GSVA) was employed in the comparison between the pathways regulation of the UL- and UH-TMB groups. See the figure below.
As to LUAD, it was found that in the UH-TMB group, the biological pathways with upregulated expression were mainly associated with the cell cycle, base excision repair, homologous recombination, DNA replication, and mismatch repair pathways. Interestingly, in the UL-TMB group, upregulation of expression occurred in the primary bile acid biosynthesis, arachidonic acid metabolism, drug metabolism cytochrome 50, complement and coagulation cascades, and calcium signaling pathways (see Figure 3). For the LUSC samples, expression of the Eicosanoid ligand-binding receptors, the CaM pathway, and the PLC-gamma1 signaling pathway were upregulated in the UL group (Figure S3).
The tumor map was drawn using an interactive plotting system to visualize, the expression data and explore the molecular similarities of the cancer samples. By labeling TMB groups’ information of each sample, it was noted that the UL-TMB samples frequently fell between normal tissue and the major tumor subtypes. The tumor map of the LUAD and LUSC samples is shown shown as Figure S4.
Cancer progression involves the gradual loss of a differentiated phenotype and acquisition of progenitor and stem-cell-like features. This study used a one-class logistic regression (OCLR) machine-learning algorithm to analyze the stemness features of the samples, which were associated with oncogenic dedifferentiation. The results showed that the stemness of the UH group was significantly higher than that of the UL group (Wilcoxon rank-sum test P < 0.05) for both LUAD and LUSC (Figure 4 A is for LUAD, B is for LUSC).
5. Immunophenotype analysis.
The immunophenoscore (IPS) was calculated on an arbitrary 0–10 scale based on the sum of the weighted average Z score of the four categories. In terms of the overall IPS, there was no difference between the two groups. However, when the IPS was decomposed into 4 categories, the differences between the UL- and UH-TMB groups were statistically significant in antigen processing by major histocompatibility complex (MHC) proteins, effector cells (ECs), and suppressor cells (SCs) for LUAD. As to checkpoints (CPs) or immunomodulators, the UL group’s score appeared to be higher than the UH group’s. See the Figure 5B, 5C and 5D.
Although the differences between the UL- and UH-TMB groups lacked statistical significance for LUSC; they exhibited similar trends as the differences between the UL- and UH-TMB LUAD groups. Specifically, the UL group had higher scores for antigen presentation mediated by MHC and immunomodulators, indicating that the immune status remained normal. Also, it was found that the UL group’s scores for ECs and SCs were significantly lower than those of the UH group’s, indicating immunosuppression and inactivation of all immune cells in the samples.
The findings above were proved in the analysis of 22 immune cell type fractions using the CIBERSORT method. However, some special cells showed the opposite. For example, the proportion of dendritic cells was significantly higher compared to other activated cells while there were much less T cells follicular helpers in the UL group. T cells CD4 memory resting, dendritic cells resting, and mast cells resting were also highly infiltrated in the UL group (Figure S5).
6. Mutation signature analysis.
Results from the mutation signature analysis were rather interesting (see the Figure 6). Comparing the UL groups of the LUAD and LUSC samples, their mutation spectra shared great similarities, which were dominated by signature 1 and 3. As to the UH groups, the mutation spectra were basically covered by signature 4. Signature 1 was present in all cancer types and in most cancer samples as a result of the endogenous mutational process initiated by the spontaneous deamination of 5-methylcytosine. Signature 3 was associated with failure of DNA double-strand break-repair by homologous recombination. Signature 4 was found in head and neck cancer, liver cancer, lung adenocarcinoma, lung squamous carcinoma, small cell lung carcinoma, and esophageal cancer, which was associated with smoking, with its profile similar to the mutational pattern present in the experimental systems exposed to tobacco carcinogens. It was likely to be caused by tobacco mutagens. Signature 29 was seen in cancers associated with tobacco chewing and appeared to be different from Signature 4.