16S rRNA Sequencing for the Microbiome, RNA-Seq for Tissue Gene Expression

The 16S rRNA sequencing for tissue microbiome and RNA-seq for host lung tissue gene expression were performed by BGI-TECH (http://www.genomics.cn). Subsequent data processing and analysis were in Supplementary Materials . All sequencing data have been deposited under the BioProject: PRJNA790037 and are publicly available at the location: https://www.ncbi.nlm.nih.gov/sra/PRJNA790037

FLC Microbiome: Low Total Species, Lower Diversity, Less Tend to Change

Importantly, FLC patients had less normal specific OTU compared with sporadic ones (314 vs 353), but they showed a slightly higher cancer-specific OTU (289 vs 266) ( Figure 1H ). Secondly in Figures 2A , FLC indeed had lower α-diversity than the sporadic group in both cancer/normal (mean value: C1<C2, N1<N2, in observed species, Chao, ace). For details, FLC cancer not only had less species richness but also low species evenness, while sporadic cancer microbiome seemed to have relatively higher species richness but also better species evenness (mean value: C1<C2, in Shannon, Simpson); FLC normal also had lower species richness, but it may completely miss some rare species found in sporadic normal, thus seemingly better in species evenness, so the Shannon and Simpson diversity were higher than sporadic normal (mean value: N1>N2, in Shannon, Simpson). In Figure 2E , β-diversity revealed: FLC normal microbiome showed clear overlap with sporadic normal, suggesting similarity; but their cancer microbiome separated quite considerably, reflecting more difference. Further, the FLC microbiome changed relatively less between cancer/normal, while sporadic microbiome changed considerably more from normal to cancer (C1 vs N1 had some clear overlap, and some subjects outside the overlapped area, it meant when N1 developed into cancer form C1, the microbiome changed with a smaller variance. Comparatively, C2 vs N2 had less overlap, it meant when N2 developed into cancer form C2, the microbiome changed with a bigger variance.). Notably, 68% of FLC patients were men, so the features of the FLC microbiome were not mainly caused by gender, even women’s microbiome showed similar features in the next paragraph. Taken together, the FLC microbiome seemed to be smaller, low-diversity, and inactive to change.

Female Microbiome: High Total Species, Low Specific Species, Less Prone to Change; Male Microbiome: Low Total Species, High Specific Species, More Likely to Change

Male/female: our female patients had the minimum normal tissue-specific OTU among all normal subgroups (210 vs 314, 353, 295, 374, 454, 274, 369), while male patients had the highest normal tissue-specific OTU in all subgroups (454, the maximum specific OTU) ( Figures 1H–K ). Women also had much lower specific OTU than men in both normal/cancer tissue, (210 vs 454, 233 vs 321) ( Figure 1J ). Compared with the benign control, women also had much lower normal-specific OTU than men (280 vs 689) ( Figure 1N ). Taken together, in normal lung tissue, women tended to have relatively lower specific species, while men have higher specific species. However in Figure 2C , our women showed generally higher α-diversity than men in both cancer/normal (mean value: C.F>C.M, N.F>N.M, in observed species, Chao, ace, Shannon); additionally, the clearly higher observed-species mean value in females indicating: women’s microbiome might have more total species than that of men, while male microbiome harbored relatively fewer total species. Accordingly, our men seemed to have a more “concentrated” microbiome than the women ( Figure 1F ). In Figure 2G for β-diversity: our female normal microbiome also showed major overlap with male normal, reflecting general similarity; but the female microbiome did not change too much between normal/cancer (N.F vs C.F almost overlapped together); in contrast, our male microbiome changed clearly more from normal to cancer (N.M vs C.M had relatively less overlapped area, also distributed differently), which led to the separation of male/female cancer microbiome (C.M vs C.F). Summarily, our female microbiome was constituted by high total species, but a lot of these microbes might be shared with others, as a result, the women carried lower specific species; it could possibly explain: carrying a wide range of “shared lung microbes”, made female microbiome less prone to change. Notably, 50% of female patients were sporadic, even some features of the female microbiome were similar to FLC, it could not be completely attributed to FLC ratio.

Smoker/nonsmoker, in Figure 2D : in normal lung, nonsmokers’ microbiome showed overall higher α-diversity than smokers’ (mean value: N.N>N.E, in all 5 indices), suggesting generally better lung microenvironment; but smokers had slightly higher cancer α-diversity than nonsmokers (mean value: C.E>C.N, in observed species, Chao, ace, Simpson). That related to what was found in β-diversity ( Figure 2H ): Even their normal microbiome had visible overlap with each other; smoker microbiome changed more from normal to cancer (N.E vs C.E had much less overlapped area, also distributed differently), which led to wider separation in smoker/nonsmoker cancer microbiome (C.N vs C.E had much less overlapped area, and distributed quite differently); while nonsmokers’ microbiome seemed to change much less from normal to cancer (N.N vs C.N largely overlapped together). Summarily: if normal lung had more total species and higher α-diversity, reflecting a microbiome with better buffering potential, thus the microbiomes were less likely to change from normal to cancer. Comparatively, less total species and lower normal lung α-diversity might lead to the microbiome being less stable and more likely to change.

Another finding: from normal to cancer, a slightly increased “cancer α-diversity” could be observed in four subgroups: sporadic ( Figure 2A , mean value C2>N2 in observed species, Shannon, Simpson), low-IAP ( Figure 2B , mean value C.L>N.L in observed species, Chao, Shannon, Simpson), men ( Figure 2C , mean value C.M>N.M in Chao, Shannon, Simpson) and smokers ( Figure 2D , mean value C.E>N.E in all five indices). Accordingly, these four subgroups also showed relatively larger β-diversity change from normal to cancer ( Figure 2E , C2 vs N2. Figure 2F , C.L vs N.L. Figure 2G , C.M vs N.M. Figure 2H , C.E vs N.E. all showed relatively less overlapped area or different distribution, compared with their counterpart). One assumption was: some cancer microbiome tended to have more species richness and better species evenness; since host “cancer soil” may support certain microbes’ colonization and reproduction; resulting in higher cancer α-diversity. Potentially, the slight increase in cancer-associated species may reflect the elevated chaos in the cancer tissue itself, which carried various somatic mutations. Other groups also reported increased microbiome richness and diversity in cancer tissue (23, 25).

Smoking and IAP Decreased Specific-OTU Biodiversity, Especially in Normal Lung

The influence of smoking and IAP on biodiversity appeared complex. Compared with low-IAP patients, the High-IAP group seemed to have slightly lower cancer α-diversity, but higher normal α-diversity ( Figure 2B , mean value C.H<C.L in observed species, Shannon, Simpson; mean value N.H>N.L in all five indices). Similarly, smokers had a little higher cancer α-diversity, but lower normal α-diversity, when compared with nonsmokers ( Figure 2D , mean value C.E>C.N in observed species, Chao, ace, Simpson; mean value N.E<N.N in all 5 indies). In Figure 2F for β-diversity: high-IAP microbiome changed relatively less between cancer/normal (C.H vs N.H largely overlapped together), and high-IAP normal microbiome overlapped greatly with low-IAP normal, but their cancer microbiomes were clearly different (C.H vs C.L quite less overlapped area); since low-IAP microbiome changed apparently more from normal to cancer (N.L vs C.L very small overlap area). However, considering subgroup-specific OTU made that problem much clearer: smoking and IAP dramatically decreased specific-OTU biodiversity, especially in normal lung tissues. In the normal part, high-IAP specific OTU had a big decrease compared to the low-IAP group ( Figure 1I : 295 vs 374); and smokers specific OTU dropped much more than nonsmokers ( Figure 1K : 274 vs 369). Similarly in the cancer part, IAP only made a slight specific OTU change ( Figure 1I : 273 vs 279), but smoking caused the biggest specific OTU drop ( Figure 1K : 172 vs 379). Additionally, compared with the benign control, the low-IAP group and nonsmokers both showed higher normal specific OTU than their counterparts ( Figure 1M : 500 vs 435, Figure 1O : 504 vs 394). Conclusively, pollution inside the lung microenvironment was noxious to the “host lung ecosystem”. Generally, pollutants from smoking and coal burning could influence the microbiome of oral, lung, and gut, leading to various diseases (26, 27), via directly changing microenvironmental oxygen, pH, and chemical composition; affecting the immune system (28), or promoting colonization of specific taxa through biofilm formation (26, 29).

Microbes Enriched or Dropped in Lung Cancer Tissue

Among all subgroups, some genera increased universally or frequently in cancer tissue; including Staphylococcus, Capnocytophaga, Lachnoanaerobaculum, Fusobacterium, Oligella, Rubellimicrobium, Marinococcus Sphingomonas, and Sphingopyxis. Firstly, Staphylococcus is commonly found in the environment, and is normal human flora (30); Capnocytophaga, Lachnoanaerobaculum, and Fusobacterium are normal mouth flora; species from these four genera are known to cause opportunistic infection, especially in immune-compromised hosts (30–33). Moreover, Fusobacterium nucleatum is reported to be prevalent in colorectal cancer (33). Oligella ureolytica was identified as a sporadic colonizer in the respiratory tract of cystic fibrosis patients (34), and also reported in bloodstream infection of lung cancer patients (35). In contrast, Rubellimicrobium are mainly isolated from soil and are also found in the air (36); Marinococcus usually lives in high-salinity environments like the sea, salt lakes, and salt mines (37); its accumulation in cancer tissue might be due to salt body fluid effusion. Intriguingly, Sphingomonas and Sphingopyxis are soil microbes; both are famous for their degradation of PAHs, pesticide, and organometallic compounds (38–41). Therefore, enrichment of these microbes in subjects exposed to IAP indicated: PAHs and other pollutants accumulated in patients’ lungs provided a substrate for the microbes, and possibly support their growth.

On the other hand, some genera decreased or depleted in cancer tissue, presenting a different trend: we found Comamonas often decreased in cancer; while species of Comamonas could cause opportunistic infection but are also known to degrade PAHs and are tolerant to heavy metals (42, 43). Additionally, Peptococcus were not detected in cancer tissue but only found in normal tissue; but previous work suggested Peptococcus is also an opportunistic pathogen (44). Taken together, since both were not likely to be probiotics, their abundance drop in cancer tissue might be explained as being out-competed by other microbes. Finally, some genera were only enriched in certain groups, different genera and species were listed in Tables S28–S39 .

Microbes Showed Special Features in Familial Lung Cancer

There were microbes highlighted in the FLC group ( Table S28 ). From normal tissue to cancer, Capnocytophaga, Fusobacterium rose dramatically in FLC cancer (about 5X and 42X); meanwhile, Sphingomonas, Sphingopyxis both increased apparently in FLC cancer (about 1.3X and 1.4X); reflecting that they were more likely to accumulate in FLC cancer tissue. Additionally, Gemmata was only found in the FLC population from both cancer/normal; it mainly lives in soil, but has also been detected on human skin, so it could be an opportunistic pathogen (45). Rhodococcus decreased in FLC but increased in the sporadic group; some could cause opportunistic infection (46), while its species are also used to degrade environmental pollutants (47).

For further investigation, permutational multivariate analysis of variance (PERMANOVA) was used to process the combined normal/cancer tissue data, and the significantly different microbes were listed in Table S29 . Interestingly, the genera Staphylococcus, Rubellimicrobium, Oligella, Comamonas, and Sphingomonas were the same as detected in separate cancer/normal data; while Alteromonadales and Acetobacteraceae were newly found. Only staphylococcus was less abundant in the FLC group, the others were all elevated in the FLC population, especially PAHs-degrading Comamonas and Sphingomonas.

Microbes Elevated in Patients Exposed to Indoor Air Pollution

Like the FLC population, from normal to cancer, Fusobacterium increased quite dramatically in high-IAP cancer than low-IAP (about 24X vs 4X). Notably, some genera only enriched in IAP-related groups ( Table S30 ). Firstly, Selenomonas, Thermomonas rose in cancer for both IAP groups, and the two genera were much more abundant in high-IAP normal lung. Selenomonas was reported in the oral microbiome of systemic diseases (48); Thermomonas mainly isolated from soil/sediment, which may have pollutants-degradation potential (49, 50); suggesting they were likely to be opportunistic pathogen flourished in high IAP environment. Secondly, we found Acidovorax, Butyricicoccus decreased in cancer for both IAP groups; while others reported Acidovorax were enriched in smokers’ SCC (23); Butyricicoccus mainly in the gut microbiome, and were reported in colorectal cancer (51). Thirdly, Actinomyces (often in human infection (52)) and Rhodococcus dropped in high-IAP cancer, but rose in Low-IAP cancer; these changes reflected the dynamic complexity of the lung cancer microbiome.

Microbes Different Between Men/Women and Smokers/Nonsmokers

Notably, Comamonas were 1.4~2 times higher in nonsmokers in both tissue types; from normal to cancer, Sphingomonas dropped in both men and smokers’ cancer but increased in women and nonsmokers’ cancer, so its abundance was 2~3 times higher in women and nonsmokers’ cancer, compared to their counterparts respectively ( Tables S31, 32 ). Taken together, these PAHs-degrading microbes tended to accumulate in female or nonsmoker groups. In addition, Capnocytophaga was nearly 9 times more abundant in female normal tissue than males’. Vagococcus was only found in the gender group, it rose in cancer of both genders, and was higher in female cancer. Actually, this genus is often found in animals and is rarely reported in human infection (53).

Microbiome Varied in Age, Blood Type, Anatomy Site, Histology Type, and TNM Stage

In order to better understand the microbe variation among other characteristics the data was further analyzed, which covered: age, blood type, anatomy site, histology type, and TNM stage ( Figure 3 ). The combined normal/cancer tissue data from one patient were used to generally represent the microbiome of the individual, and the significantly different microbes were listed in Tables S33-S39 . These microbes were mainly found in the environment, and many of them were opportunistic pathogens.

Open in a separate window

Figure 3

PLSDA results for age, anatomy site, blood type, histology, and TNM stage. PLSDA analysis for patients’ age (A), anatomy site (B), blood type (C), histology type (D), and TNM stage: T stage (E), N stage (F), M stage (G), Tumor stage (H). The combined normal/cancer tissue data from one patient were used to generally represent the microbiome of the individual.

Initially, we found wider genera were more likely to accumulate in patients older than 50 years, while younger patients (<50 years) shared a smaller variety ( Table S33 and Figure 3A ). For example, opportunistic pathogen Actinomyces, Corynebacterium, Deinococcus, Fusobacterium, Mycoplasma were all clearly higher in the older group, in which Deinococcus species were known for their resistance to radiation and oxidative stress (54). On the other hand, Acetobacter, Acinetobacter, Clostridium, Lactococcus, Oscillospira were more abundant in the younger group, in which Acetobacter and Clostridium function in wastewater treatment, helping degrade pollutants (55); gut microbe Oscillospira was described as a future probiotic (56). Other studies also supported age as a parameter helping to shape the lung microbiome (57).

Interestingly, patients’ microbiome could also be grouped by their tumor anatomy site ( Figure 3B and Table S34 ); it meant: the left/right lung, the upper/lower lobe were like different “terrain”; that was different habitat for their own microbe population. For example, most different genera were more abundant in the right lung, with many higher in the lower lobes. That could be explained by the structure of the organ: the right lung has wider and shorter bronchus, making inhaled particles/aerosol easier to enter the right side; and the lower lobes were more likely to accumulate bigger particles carrying microbes. Actually, considerable regional variation can be found in one lung, because numerous factors would impact bacterial growth: oxygen tension, pH, temperature, blood perfusion, alveolar ventilation, epithelial cell structure, deposition of inhaled particles, plus concentration and behavior of immune cells (58–60). Besides, another PAHs-degrading Novosphingobium (61) was also detected but seemed to prefer the upper lobes. In addition, we noted patients’ microbiome varied among different blood types ( Figure 3C , Table S35 ), which consisted with others work: blood type could contribute to shaping human microbiome, for certain bacteria could use host glycans as receptors to adhere to the host, and some liberate host blood glycans as a nutrient (62). Histology type is one major character of cancer, and also inhabited with different microbes ( Figure 3D , Table S36 ); we found 5 different genera and all were higher in adenocarcinoma. Consistently, other studies also found different microbiomes between lung AD and SCC (23).

Not surprisingly, patients’ microbiome also differed in the TNM stage ( Figures 3E–H , Tables S37–S39 ). Firstly, most opportunistic pathogens rose in the T3_4 group ( Figure 3E , Table S37 ), only Massilia and probiotic Lactobacillus were more abundant in the T1_2 group. Secondly, similar was seen in the N group; from N0 to N1-2 ( Figure 3F , Table S38 ), increased genera mainly cause opportunistic infection. Only Blastomonas, Bradyrhizobium, Kaistobacter were higher in the N0 group; among them, Bradyrhizobium, and Kaistobacter were reported to play important roles in pollutant degradation/detoxication, including PAHs and heavy metals (63, 64). Even there were only two M1 patients, their microbes’ composition still drifted outside the M0 population ( Figure 3G ). Finally, opportunistic pathogens increased in stage III_IV ( Figure 3H , Table S39 ); but Blautia, Cloacibacterium, and probiotic Lactobacillus were higher in stage I_II patients. In which gut microbe Blautia was reported as potential probiotics (65), and Cloacibacterium were able to detoxify heavy metals by producing extracellular polymers (66). In conclusion, for the earlier stage, we found relatively higher probiotics and pollutants-detoxication genera; while in the later stage opportunistic pathogens rose in abundance and variety. Similarly, other studies also noted differential abundance between stage I–IIIA and IIIB–IV, with later-stage having more enriched oral commensals, such as Haemophilus, Fusobacterium, Gemella, etc. (67).