1
Scien i ic RepoR S | (2019) 9:12838 | h ps://doi.o g/10.1038/s41598-019-49195-w
www.na u e.com/scien i ic epo s
P o iling o lung mic obio a
discloses di e ences in
adenoca cinoma and squamous cell
ca cinoma
Síl ia Gomes1,2, B uno Ca adas1,2, Joana Ca a ina Fe ei a1,2, Pa ícia Isabel Ma ques1,2,
Ca a ina Mon ei o1,2, Ma ia Sucena3, Ca a ina Sousa3, Luís Vaz Rod igues4, Gilbe o Teixei a5,
Paula Pin o6,7, Tiago Ta a es de Ab eu6, C is ina Bá ba a6,7, Júlio Semedo6, Leono Mo a6,
Ana So ia ca alho
8, Rune Ma hiesen
8, Luísa pe ei a1,2,9 & Susana Seixas
1,2
The lung is a complex ecosys em o hos cells and mic obes o en dis up ed in pa hological condi ions.
Al hough bac e ia ha e been hypo hesized as agen s o ca cinogenesis, li le is known abou mic obio a
p o ile o he mos p e alen cance sub ypes: adenoca cinoma (ADC) and squamous cell ca cinoma
(SCC). To cha ac e ize lung cance (LC) mic obio a a i s a sc eening was pe o med h ough a pooled
sequencing app oach o 16S ibosomal RNA gene (V3-V6) using a o al o 103 b onchoal eaola la age
luid samples. Then, iden i ied axa we e used o inspec 1009 cases om The Cance Genome A las and
o anno a e umo unmapped RNAseq eads. Mic obial di e si y was analyzed pe cance sub ype,
his o y o ciga e e smoking and ai low obs uc ion, among o he clinical da a. We show ha LC
mic obio a is en iched in P o eobac e ia and mo e di e se in SCC han ADC, pa icula ly in males and
hea ie smoke s. High equencies o P o eobac e ia we e ound o disc imina e a majo clus e , u he
subdi ided in o well-de ined communi ies’ associa ed wi h ei he ADC o SCC. He e, a SCC subclus e
di e ing om o he cases by a wo se su i al was co ela ed wi h se e al En e obac e iaceae. O e all,
his s udy p o ides i s e idence o a co ela ion be ween lung mic obio a and cance sub ype and o
i s in luence on pa ien li e expec ancy.
Lung Cance (LC) is he mos common and le hal cance wo ldwide wi h a epo ed incidence o 11.6%, a mo -
ali y a e o 18.4% and acco ding o ecen es ima es - 2.1 million new cases1. F om a his ological poin o iew,
non-small cell lung cance (NSCLC) is he mos p e alen ype, which can be u he subdi ided in o wo majo
sub ypes: adenoca cinoma (ADC) and squamous cell ca cinoma (SCC). To da e, di e se en i onmen al ac o s
besides ciga e e smoking, like biomass bu ning, indoo and ou doo pollu an s, a e sugges ed o play a ole in
LC pa hogenesis, as well as in ch onic obs uc i e pulmona y disease (COPD)2. This la e illness, cha ac e ized
as he pe sis ence o ai low limi a ion in esul o b onchi is and/o emphysema, is also ecognized as a c i ical
como bidi y in LC, always associa ed o a wo se p ognosis3. Mo eo e , a mechanis ic link be ween COPD and
LC has been p oposed pa ially due o he indings o ch onic in lamma ion and i s epea ed cycles o inju y and
epai , which in COPD a e known o p omo e umo igenesis and malignan ans o ma ion4.
No ably and simila ly o he model es ablished o Helicobac e pylo i in gas ic cance , Mycobac e ium
ube culosis has been hypo hesized as a LC agen , once i induces in lamma o y ma ke s and causes signi ican
1Ins i u o de In es igação e Ino ação em Saúde, Uni e sidade do Po o (I3S), Po o, Po ugal. 2Ins i u e o Molecula
Pa hology and Immunology o he Uni e si y o Po o (IPATIMUP), Po o, Po ugal. 3Pneumology Depa men ,
Cen o Hospi ala de São João (CHSJ), Po o, Po ugal. 4Depa men o Pneumology, Unidade Local de Saúde da
Gua da (USLG), Gua da, Po ugal. 5Depa men o Pneumology; Cen o Hospi ala do Baixo Vouga (CHBV), A ei o,
Po ugal. 6Unidade de Técnicas In asi as Pneumológicas, Pneumologia II, Hospi al Pulido Valen e, Cen o Hospi ala
Lisboa No e (CHLN), Lisbon, Po ugal. 7Ins i u o de Saúde Ambien al, Faculdade de Medicina da Uni e sidade de
Lisboa, Lisbon, Po ugal. 8Compu a ional and Expe imen al Biology G oup, CEDOC, Faculdade de Ciências Médicas,
Uni e sidade No a de Lisboa, Lisboa, Po ugal. 9Faculdade de Medicina da Uni e sidade do Po o, Po o, Po ugal.
Co espondence and eques s o ma e ials should be add essed o S.S. (email: [email p o ec ed])
Recei ed: 4 Ma ch 2019
Accep ed: 19 Augus 2019
Published: xx xx xxxx
open
2
Scien i ic RepoR S | (2019) 9:12838 | h ps://doi.o g/10.1038/s41598-019-49195-w
www.na u e.com/scien i ic epo s
www.na u e.com/scien i ic epo s/
al e a ions in lung issues5,6. Howe e , his associa ion o ube culosis is ha d o disen angle because o i s
co-occu ence wi h o he isk ac o s such as smoking, emphysema and b onchi is5.
In his a ea o knowledge, he impac o lung mic obio a, o he bac e ia communi ies inhabi ing he lung, has
been mo e ex ensi ely s udied in COPD han in LC. This is likely o esul om COPD pa ien s o en su e ing
om acu e exace ba ions, which a e conside ed o be o in ec ious na u e and caused by bac e ia and/o i us7.
Un il now, s udies aiming o cha ac e ize COPD and LC mic obio a used di e en biological specimens,
including spu um, b onchoal eaola la age luid (BALF) and lung issue. And, whe eas in COPD, samples we e
collec ed du ing dis inc phases o he disease (s able o exace ba ion); in LC, hose we e ga he ed in some
ins ances om bo h umo and non- umo egions. Mo eo e , al hough p e ious wo ks we e mainly based in
he sc eening o 16S ibosomal RNA (16S RNA) gene, hei expe imen al app oaches conce ning sample size,
hype a iable egions co e ed and sequencing echnologies employed a e qui e di e se and consequen ly, hei
indings a e no always consensual8–20.
None heless, mos s udies seem o ag ee in he common co e mic obio a o bo h heal hy and diseased sub-
jec s domina ed by Fi micu es, Bac e iode es, P o eobac e ia and Ac inobac e ia phyla, and by S ep ococcus,
Haemophilus, Veilonella, Pseudomonas and P e o ella gene a8–23. In e es ingly, in COPD, mic obio a ends o
be ela i ely s able o e ime and a ec ed du ing exace ba ions, when i shi s owa d P o eobac e ia manly due
o a Mo axella inc ease and a S ep ococcus educ ion8,22. Con e sely, in LC, a Fi micu es swi ch was sugges ed
in esul o augmen ed S ep ococcus, G anulica ella, Veillonella and Megasphae a p e alence14,16,21,24. C i ical
changes in mic obio a composi ion we e sugges ed o occu along wi h ai way disease p og ession12,17,18,22.
Howe e , he use o ce ain bac e ial axa as po en ial bioma ke s o imp o ed pa ien s a i ica ion o e en as
ea able ai s, is a om being a eali y.
He e, we explo e an associa ion o lung mic obio a wi h cance , while add essing also he impac o COPD
co-mo bidi y. B ie ly, we pe o med a mic obio a p o iling in a small se o Po uguese samples, used la e o
guide he cha ac e iza ion o bac e ial communi ies in an ex ended coho o ADC and SCC cases om The
Cance Genome A las (TCGA) Resea ch Ne wo k25. This allowed us o de ec signi ican di e ences in mic obio a
di e si y o cance sub ypes, as well as o iden i y among SCC a well-de ined communi y (En e obac e iaceae)
connec ed o a wo se pa ien su i al mainly due o non-cance complica ions. Fu he mo e, we iden i ied
Ach omobac e as a g am-nega i e bac e ium linked wi h bo h SCC and COPD.
Ma e ials and Me hods
Sample collec ion. BALF was collec ed om subjec s unde going b onchoscopy o e alua ion o lung dis-
ease a h ee hospi als in Po ugal: Cen o Hospi ala São João (CHSJ), in Po o; Cen o Hospi ala Baixo Vouga
(CHBV), in A ei o; and Hospi al Pulido Valen e - Cen o Hospi ala Lisboa No e (CHLN), in Lisbon. In o med
consen was ob ained o all pa icipan s and sample collec ion o Human Resea ch was app o ed by hospi al
e hical commi ees: Comissão de É ica pa a a Saúde (CES) – CHSJ, Comissão de É ica – CHBV and Comissão de
É ica pa a a Saúde (CES) – CHLN. The s udy was conduc ed in acco dance wi h e hical guidelines and egula-
ions o Human esea ch and wi h Helsinki Decla a ion. Sample collec ion was a ge ed owa d a ec ed lung
segmen s and done by b onchoscope wedging in o subsegmen al lung egions. In his s udy, we used only b on-
choscope wo king channel washes, which we e done wice wi h a minimum olume o 15 mL (0.9% saline solu-
ion). Samples we e hen s a i ied in LC (N = 49) o con ols (N = 54) based in posi i e o nega i e cy ology
esul s (Supplemen a y Table1). Howe e , in a ollow-up analysis ca ied ou up o 2 yea s a e BALF collec ion,
wo cases we e ound o be alse posi i es and ano he ou ini ially classi ied as nega i e, o e ime p og essed o
LC26. The pooled sample s a egy p e en ed he ealloca ion o hese cases o con ols and ice- e sa.
Lung mic obio a 16S RNA sc eening. DNA ex ac ion om BALF (200 µL) was pe o med using
DNA Mini ki (Qiagen) acco ding o manu ac u e ins uc ions o cap u ing bac e ial DNA in body luids.
Two 16S RNA agmen s spanning hype a iable egions V3-V4 and V4-V6 we e ampli ied using uni e sal
p ime s (Supplemen a y Table2). Pooled samples con aining PCR p oduc s (~200 ng/sample) we e gene a ed
o LC (N = 49) and con ols (N = 54) and p ocessed as p e iously desc ibed27. B ie ly, wo lib a ies we e con-
s uc ed acco ding o Ion Xp ess™ Plus F agmen Lib a y Ki p o ocol (Li e Technologies) and an in an Ion
PGM™ Sys em - 316™ chip (Li e Technologies). The ools USEARCH, UCHIME, QIIME and G eengenes we e
used in he analysis o ope a ional axonomic uni s (≥97% nucleo ide sequence iden i y cu -o ) as p e iously
desc ibed27.
TCGA da ase . Raw RNAseq eads om umo s and clinical da a iles co esponding o 515 ADC and 501
SCC cases om TCGA, we e downloaded om Genomic Da a Commons (GDC) Da a Po al (h ps://gdc.cance .
go /). To pe o m a quan i a i e analysis o lung mic obio a we used eads no aligning wi h human e e ence
sequence (unmapped eads) as inpu o QmihR28. This pipeline combines T immoma ic, Bow ie2, and RSEM o
a p obabilis ic in e ence o bac e ial axa abundances28. To ins uc bac e ial sequence su eys, we i s de ined
a mic obio a e e ence panel based in p e ious e idence o lung coloniza ion in heal hy and diseased pa ien s.
Exac ly, we conside ed a o al o 567 bac e ial axa acco ding o he da a a ailable in: 1) he Human Mic obiome
P ojec (HMP) – ai ways; 2) specialized li e a u e; and 3) ou own 16S RNA s udy; o which whole genome
sequences could be collec ed om h ps://www.ncbi.nlm.nih.go /genome/mic obes/. Upon quali y con ol, ela-
i e abundances o 112 gene a we e ob ained o 509 ADC and 500 SCC cases. These samples we e hen s a i ied
acco ding o se e al a iables including ances y (Eu opean o A ican), gende , age a diagnosis (≤65 o >65),
ana omic posi ioning (Uppe o Lowe lung), localiza ion in lung pa enchyma (Pe iphe al and Cen al Lung) and
pa hological umo s age (S ages I, II and III + IV). In addi ion, pos -b onchodila o o ced expi a o y olume in
1 second (FEV1) and o ced i al capaci y (FVC) a io we e used o de e mine he p esence (FEV1/FVC < 0.70)
o absence o ai low obs uc ion3. Smoking his o y in pack pe yea s (PPY) was conside ed using a i s 20 PPY
3
Scien i ic RepoR S | (2019) 9:12838 | h ps://doi.o g/10.1038/s41598-019-49195-w
www.na u e.com/scien i ic epo s
www.na u e.com/scien i ic epo s/
subdi ision (da a no shown). Gi en ha mos cases la gely su passed his alue, a naï e 45 PPY spli was used
ins ead based on i s p oximi y o a e age alues (all cases 48; ADC 42 and SCC 53 PPY). Se e al pa ien ollow-up
a iables we e also conside ed in his s udy o e alua e clinical signi icance o collec ed da a. These included i al
(dead o ali e) and cance ( umo ee o wi h umo ) s a us, days o dea h and p ima y he apy ou come.
S a is ical analysis. S a is ical analysis o mic obio a di e si y was pe o med in R s udio (h ps://www.
s udio.com/; e sion 1.1.383) using phyloseq29. Alpha di e si y was e alua ed h ough in e se Simpson and
Shannon indexes. Be a di e si y, which in eg a es phylogene ic ela ionships o bac e ia was calcula ed by
weigh ed UniF ac. Dis ances ma ixes we e used in P incipal coo dina es analysis (PCoA) and in hie a chical
clus e ing (comple e linkage) o TCGA samples. The linea disc iminan analysis (LDA) e ec size (LE Se) algo-
i hm30 was used o de ec axa wi h di e en ial abundances be ween TCGA cases. Su i al analyses and log
ank es s o pai wise compa isons o di e en case se s we e ca ied ou h ough he Coho Compa ison ool
a ailable a GDC Da a Po al.
Resul s
Cha ac e iza ion o lung bac e ial communi ies. In ou pooled sequencing app oach (16S RNA V3–
V6) o cases and con ols using BALF samples (Supplemen a y Table1), we we e able o iden i y a o al o 11
phyla and 54 gene a wi h ela i e equencies abo e 0.1% (Supplemen a y Tables3, 4). The p e ailing phyla in
ou da ase we e P o eobac e ia, Fi micu es, Ac inobac e ia and Bac e oide es (Fig.1a), as i could be expec ed
om p e ious s udies8–22. Some a ia ion in phyla p opo ions we e obse ed be ween pools o P o eobac e ia,
(38.7% in cases s 49.2% in con ols; Z-sco e P = 0.284) and Ac inobac e ia (16.5% in cases s 8.0% in con ols;
Z-sco e P = 0.187). Among gene a, Haemophilus (P o eobac e ia); S ep ococcus and Veillonella, (Fi micu es);
Co ynebac e ium and Ac inomyces (Ac inobac e ia) and P e o ella (Bac e oide es) we e he mos common in he
wo pools. Again, ou sc eening ag eed wi h o me epo s o lung mic obio a8–22, disclosing only non-signi i-
can changes be ween pools in ela i e abundances o Haemophilus (29.5% in cases s 37.5% in con ols; Z-sco e
P = 0. 390) and Co ynebac e ium (8.2% in cases s 1.3% in con ols; Z-sco e P = 0.095). O e all, bac e ial com-
muni ies we e bo h ai ly di e se as indica ed by Shannon index a genus le el (2.69 in cases s 2.53 in con ols).
Figu e 1. Cha ac e iza ion o lung cance (LC) mic obio a. (a) Rela i e abundance a he phylum and genus
le els o Po uguese cases and con ols. This da a co esponds o he DNA pooling o 16S RNA amplicons
(V3-V6) o 49 and 54 indi idual samples, espec i ely. (b) Rela i e abundance o gene a iden i ied among umo
sec ions o 1009 lung cance cases om The Cance Genome A las (TCGA). ADC: adenoca cinoma (N = 509).
SCC: Squamous cell ca cinoma (N = 500).
4
Scien i ic RepoR S | (2019) 9:12838 | h ps://doi.o g/10.1038/s41598-019-49195-w
www.na u e.com/scien i ic epo s
www.na u e.com/scien i ic epo s/
Howe e , o he in e se Simpson, LC cases we e ound o be conside ably mo e di e se han con ols (7.98 s
5.74, espec i ely).
In he TCGA se ies ( umo RNAseq) se e al disc epancies o o me es ima ed p opo ions we e obse ed.
Wi h an o e all equency o 59.4%, P o eobac e ia su passed by a he emaining phyla Ac inobac e ia,
Fi micu es, and Bac e oide es (23.4%, 12.0% and 4.4% espec i ely; Supplemen a y Table5). Con e sely, op gen-
e a comp ised Pseudomonas, Acine obac e , Sphingomonas, Me hylobac e ium and En e obac e (P o eobac e ia);
P opionibac e ium, Co ynebac e ium, and Mic ococcus (Ac inobac e ia); and S ep ococcus and S aphylococcus
(Fi micu es); all wi h a e age p e alence abo e 3%. Taxa p e iously iden i ied as abundan in 16S RNA
pooled sequencing we e con i med o be p esen in TCGA bu a a iable equencies, anging om 3.7% o
Co ynebac e ium o 0.3% o Haemophilus (Fig.1b; Supplemen a y Table6). Conce ning mic obio a di e si y,
TCGA cases showed simila alues o ou BALF samples o Shannon index (2.86 ± 0.43) and highe s a is ics o
in e se Simpson (11.54 ± 5.06).
Di e en ia ion o ADC and SCC sub ypes. The a ailabili y o clinical pa ame e s o TCGA cases
allowed an in-dep h analysis o possible ac o s a ec ing lung mic obio a. Aside om some a iabili y in gene a
abundance pe cance sub ype (Fig.1b), we ound ha SCC ends o show highe di e si y han ADC as indica ed
by in e se Simpson (Fig.2a). This di e ence seems o be co ela ed wi h Eu opean ances y, male gende , hea y
smoking (PPY >45) and olde ages a he ime o diagnosis (>65 yea s) (Fig.2b). Ye , we could no de ec any
e ec on mic obio a di e si y when conside ing umo localiza ion, uppe o lowe lung and cen al o pe iphe al
pa enchyma (Fig.2b).
Figu e 2. Alpha di e si y o lung mic obio a om lung cance (LC) cases o The Cance Genome A las (TCGA).
(a) In e se Simpson and Shannon indexes o LC cases g ouped by his ological sub ype. (b) In e se Simpson
index o LC sub ypes g ouped acco ding o di e en clinical a iables a ailable a TCGA da abase (ances y,
gende , age a diagnosis, smoking his o y, lung egion and lung pa enchyma). Welch’s - es was used o access
s a is ical signi icance o pai wise compa isons (*P- alue < 0.05; **P- alue < 0.01, ***P- alue < 0.001). ADC:
adenoca cinoma. SCC: Squamous cell ca cinoma.
5
Scien i ic RepoR S | (2019) 9:12838 | h ps://doi.o g/10.1038/s41598-019-49195-w
www.na u e.com/scien i ic epo s
www.na u e.com/scien i ic epo s/
We also used TCGA da ase o es he hypo hesis o a loss o mic obio a di e si y along wi h disease p og es-
sion. Ne e heless, no di e ences we e obse ed ac oss pai wise compa isons o cance s ages (I, II and III + IV)
o COPD p esence and absence. S a i ica ion by cance sub ype did no al e he esul s (Supplemen a y Fig.1).
To e alua e he simila i y o mic obio a p o iles weigh ed UniF ac dis ances we e calcula ed. As plo -
ed in PCoA (Fig.3), LC communi ies a e qui e a iable ac oss samples and o e lap be ween ADC and SCC.
None heless, some indi idual p o iles appea o clus e and o be co ela ed wi h ei he ADC (uppe igh quad-
an ) o SCC (lowe igh quad an ). No o he a iable was ound o agg ega e g oups o cases ( esul s no shown).
Then, o gain a be e insigh in o LC mic obio a p o iles we pe o med a hie a chical clus e ing o TCGA
samples, ollowed by a g aphic display o hei lung communi ies a wo axonomic anks - phylum and genus
(Fig.4a–d). In he phylum analysis some he e ogenei y among cases could be al eady wi nessed, as indica ed
by uppe ee clus e s and subclus e s. Highe abundances o P o eobac e ia we e connec ed wi h a i s clus e
(p_C1) and a second one (p_C2) could be di ided in o h ee majo subclus e s. Basically, hese di e ged in he
ela i e p opo ions o common phyla: p_C2s1 was domina ed by Ac inobac e ia; p_C2s2 had balanced equen-
cies o P o eobac e ia, Ac inobac e ia, Fi micu es and Bac e iode es and p_C2s3 was Fi micu es en iched.
The analysis a he genus le el depic ed a la ge complexi y o lung mic obio a, whe e cases o en clus e ed in o
small g oups showing long e minal b anches. No ably, inside p_C1 wo clus e s con as ed wi h he emaining
ee by hei sho e e minal b anches (p_C1/g_C1 and p_C1/g_C2; Fig.4a). Whe eas p_C1/g_C1 could be cha -
ac e ized by a communi y composed by p e alen gene a such as Sphingomonas, B e undimonas, Acine obac e
and Me hylobac e ium; p_C1/g_C2 could be de ined by En e obac e , Mo ganella, Kluy e a and Capnocy ophaga.
In e es ingly, p_C1/g_C1 con ained only ADC cases (N = 32), all o hem loca ed in he uppe igh quad an
o PCoA plo (Supplemen a y Fig.2). In con as , p_C1/g_C2 included essen ially SCC cases (89 in 94), his
u n co esponding o he plo lowe igh quad an (Supplemen a y Fig.2). In he p_C2s2 a single clus e
eme ged as less he e ogeneous (p_C2s2/g_C1), in his ins ance, his could be co ela ed wi h high equencies o
P opionibac e ium and mos ly linked o ADC cases (32 in 42).
In he LE SE analysis o ADC and SCC cases a o al o 37 gene a we e de ec ed o display con as ing co -
ela ions be ween LC sub ypes (Fig.5a). P ecisely, o ADC he gene a wi h highe LDA sco es (>3.5) and
ex eme P- alues (P < 5 × 10−8) we e Acine obac e , P opionibac e ium, Phenylobac e ium, B e undimonas and
S aphylococcus. On he o he hand, o SCC he gene a i ing such equi emen s we e En e obac e , Se a ia,
Kluy e a, Mo ganella, Ach omobac e , Capnocy ophaga and Klebsiella (Supplemen a y Table7). In e es ingly,
mos o hese bac e ia could be co ela ed wi h p e iously iden i ied clus e s - p_C1/g_C1, p_C1/g_C2 and p_
C2s2/g_C1. A simila app oach was used o add ess a possible con ibu ion o bac e ia in o COPD, as a common
co-mo bidi y o bo h ADC and SCC (Fig.5b). Among he 12 axa iden i ied Ach omobac e was he one mos
s ongly co ela ed wi h ai low obs uc ion (LDA sco es >3.5 and P- alues ≤0.010; Supplemen a y Table7).
Bac e ial communi ies as p ognos ic bioma ke s. To in es iga e i ou indings could ha e clinical
po en ial, especially in a be e s a i ica ion o LC cases, we compa ed he su i al cu es o p e iously iden i-
ied clus e s. In a i s s ep, no signi ican di e ences we e de ec ed be ween p_C1 (P o eobac e ia domina ed)
and p_C2s2 (in e media e abundances o common phyla), no e en when sepa a ed by ADC and SCC. Ye , in he
global compa ison, and among SCC cases p_C1 clus e appea s o be associa ed wi h a slowe decay o su i al
a es (P = 0.076 and P = 0.089, espec i ely; Fig.6). Se e al analyses we e pe o med also in ADC, o p_C1/g_C1
(Acine obac e /B e undimonas communi y), p_C2s2/g_C1 (P opionibac e ium communi y) and o he cases, bu
all ailed o each compelling esul s possibly due o hei low sample sizes. On he o he hand, among SCC
he p_C1/g_C2 clus e (En e obac e communi y) was ound o depa u e om he emaining p_C1 cases wi h
a wo se su i al (P = 0.011), close o he one obse ed in p_C2s2 clus e . S ill, he s onges di e gence in SCC
su i al a es was obse ed o non-p_C1/g_C2 (P o eobac e ia domina ed wi hou En e obac e communi y)
and p_C2s2 (P = 0.006; Fig.6). In e es ingly, p_C1/g_C2 was he clus e associa ed wi h he highes mo ali y a e
du ing ollow-up (app oxima ely 5000 days’ maximum o SCC and 7500 days o ADC), and he one co ela ed
wi h an inc eased numbe o dea hs in umo ee pa ien s (Table1). Con e sely, non-p_C1/g_C2 was shown
Figu e 3. Be a di e si y o lung mic obio a om lung cance (LC) cases o The Cance Genome A las (TCGA).
The P incipal Coo dina es Analysis (PCoA) plo was gene a ed using weigh ed UniF ac dis ances. ADC:
adenoca cinoma. SCC: Squamous cell ca cinoma.

6
Scien i ic RepoR S | (2019) 9:12838 | h ps://doi.o g/10.1038/s41598-019-49195-w
www.na u e.com/scien i ic epo s
www.na u e.com/scien i ic epo s/
Figu e 4. Bac e ial communi ies o lung cance (LC) cases om The Cance Genome A las (TCGA). (a)
Hie a chical clus e ing o LC cases buil using weigh ed UniF ac dis ances and comple e linkage me hod. Majo
clus e s iden i ied a e indica ed in he ee. (b) Schema ic ep esen a ion o LC sub ype and gende a iables
a ailable o all samples. (c) Phyla ela i e abundance pe each sample. (d) Gene a ela i e abundance pe each
sample. Less equen axa a e g ouped in a single ca ego y and labeled as “O he s”. ADC: adenoca cinoma.
SCC: Squamous cell ca cinoma.
Figu e 5. Mic obial di e en ia ion o The Cance Genome A las (TCGA) cases acco ding o disease s a us
and linea disc iminan analysis (LDA) e ec size (LE Se) algo i hm. (a) lung cance sub ype. (b) COPD co-
mo bidi y p esence o absence. ADC: adenoca cinoma (N = 509). SCC: Squamous cell ca cinoma (N = 500).
W/o COPD: cases wi hou COPD (N = 164); W/COPD: cases wi h COPD (N = 110).
7
Scien i ic RepoR S | (2019) 9:12838 | h ps://doi.o g/10.1038/s41598-019-49195-w
www.na u e.com/scien i ic epo s
www.na u e.com/scien i ic epo s/
o display a educed mo ali y e en when he disease p og essed nega i ely a e p ima y he apy and a pe iod
o comple e emission (Table1). The small size o p_C1/g_C1 and p_C2s2/g_C1 clus e s p e en ed in ADC an
accu a e e alua ion o he impac o hese well-de ined communi ies in pa ien ou come (Table1).
Discussion
In his s udy, we pe o med a cha ac e iza ion o LC mic obio a using wo dis inc da ase s and me hodologi-
cal app oaches: a pooled sequencing o 16S RNA in BALF samples om Po uguese cases and con ols; and a
su eying o bac e ial RNAseq eads made a ailable h ough TCGA, o which umo sec ions o hund eds o
pa ien s we e collec ed. The main ad an age o he i s app oach was o p o ide a p elimina y and aw o e iew
o lung mic obio a a e y low cos . Howe e , his pooled app oach has se e al limi a ions s a ing by i s inabil-
i y o add ess in e -indi idual a iabili y and o accu a ely pinpoin bac e ial communi ies o each indi idual.
Ano he weakness is ela ed o sample he e ogenei y, which con ains se e al cases lacking a comple e his ological
classi ica ion and con ols ha include manly subjec s wi h o he pa hologies. A las , su eying 16S RNA can
be conside ed also a sho coming, once i is expec ed o in oduce some asce ainmen bias in axa iden i ica-
ion. This ca ea is a ibu ed, on one hand, o he di e en ial annealing a ini ies o uni e sal p ime s used in
16S RNA ampli ica ion, and on he o he , o he dis inc esol ing powe o co e ed hype a iable egions27,31.
Con e sely, in he second app oach, we could bene i om a la ge coho o ADC and SCC cases, o which
de ec ed RNAseq eads a e mo e likely o ep esen an accu a e composi ion o lung mic obio a. The unique
disad an age o his s a egy is ha in o de o maximize e iciency in mapping bac e ial eads, we p o ided a
da abase o e e ence genomes28. This was buil using axa iden i ied in ou 16S RNA su ey, combined wi h
HMP da a and published elsewhe e8–22.
O e all, P o eobac e ia eme ged as he p edominan LC phylum, a end cap u ed also in a la ge sample
o cance pa ien s o which non-malignan issue sec ions we e collec ed17. Un il now, inc eased equencies
o P o eobac e ia we e mos ly co ela ed wi h as hma, COPD exace ba ions and ad anced COPD s ages8,9,32,33.
Howe e , gi en cu en indings a P o eobac e ia en ichmen could also be a ea u e o cance ous lungs. In-dep h
su eys unco e ed dis inc i e sco es o P o eobac e ia de ining wo majo clus e s: a i s one uly domina ed by
P o eobac e ia (p_C1); and ano he one displaying in e media e equencies o P o eobac e ia, Ac inobac e ia,
Fi micu es and Bac e oide es (p_C2s2). Al oge he , hese esul s a e sugges i e o subs uc u e in lung mic obio a
ha as a as we could in es iga e is no co ela ed wi h cance sub ype, o any o he e alua ed clinical a iable.
Also, in a global pe spec i e, Pseudomonas, S ep ococcus, S aphylococus, Veillonella and Mo axella we e iden-
i ied among he op ank bac e ia o cance cases i ing he so-called lung co e mic obiome8–22. Ne e heless,
indi idual dis ibu ions showed a di e en scena io, in which cases a e gene ally qui e di e gen in hei mic o-
bial composi ion. The excep ions o his ule a e h ee speci ic communi ies displaying ema kable links o
LC sub ypes: B e undimonas/Acine obac e (p_C1/g_C1) o ADC; En e obac e (p_C1/g_C2) o SCC, and
P opionibac e ium (p_C2/g_C1) o ADC.
Mos imp essi ely, p_C1/g_C2 clus e could be ela ed wi h an o e all poo su i al i compa ing SCC cases
wi hin p_C1 g oup. Fu he mo e, p_C1/g_C2 was cha ac e ized by se e al En e obac e iaceae (En e obac e ,
Mo ganella, Se a ia, Klebesiela and Kluy e a), a axon wi h ecognized pa hogenic po en ial causing ai way
Figu e 6. Su i al plo s o The Cance Genome A las (TCGA) cases o ganized acco ding o di e en mic obial
clus e s iden i ied and lung cance sub ypes. ADC: adenoca cinoma. SCC: Squamous cell ca cinoma.
8
Scien i ic RepoR S | (2019) 9:12838 | h ps://doi.o g/10.1038/s41598-019-49195-w
www.na u e.com/scien i ic epo s
www.na u e.com/scien i ic epo s/
in ec ions in COPD34, colonizing b onchi o LC pa ien s35 and unde lying nosocomial in ec ions wi h esis ance
o an imic obial molecules36. This clus e comp ised also Ach omobac e , ano he mul id ug esis an mic obe
p e iously ound in ai way in ec ions o cys ic ib osis pa ien s and among subjec s wi h solid malignancies37.
Mo eo e , as g am-nega i e bac e ia, En e obac e iaceae and Ach omobac e syn he ize lipopolysaccha ides
capable o s imula ing hos in lamma o y esponses. In his espec , En e obac e iaceae o e g ow h has been
desc ibed as a key e en o gu dysbiosis in obesi y, C ohn’s disease and colo ec al cance 37,38. In as hma and
cys ic ib osis he e is also a g owing body o e idence o a nega i e e ec o En e obac e iaceae39,40. The e o e,
a con ibu ion o p_C1/g_C2 communi y o an enhanced p o-in lamma o y cance mic oen i onmen seems
like a plausible hypo hesis, once i is epo ed o os e umo igenesis and p omo e lung cells malignan ans o -
ma ion4,41. This a ional is suppo ed by epo s o a wo se LC p ognosis when pa ien b onchi a e colonized by
En e obac e iaceae35.
None heless, p_C1/g_C2 clus e was also linked wi h an inc eased mo ali y in absence o any umo , which
sugges s an inc eased isk o his g oup o o he non-cance complica ions. Indeed, se e al s udies al eady
epo ed di e se pulmona y in ec ions, sep icemia and enhanced dea h a es a e cance esec ion and chemo-
he apy in LC subjec s ca ying po en ial pa hogenic mic oo ganisms in hei ai ways35,42,43. Al hough we we e
unable o igo ously add ess he impac o p_C1/g_C2 communi y in he heal h s a us o a g oup o indi iduals
p obably debili a ed by ad anced age, co-mo bidi ies (e.g. COPD, ca dio ascula disease, diabe es, e c.) and inclu-
si ely cance ea men , ou indings ad oca e o a di e en ia ed medical in e en ion in hese pa ien s, namely
in he selec ion o an imic obial he apies.
S ill, he o e all su i al o p_C1/g_C2 does no di e om p_C2s2 clus e , which ad ances Ac inobac e ia,
Fi micu es and/o Bac e iode es as addi ional isk ac o s in SCC possibly h ough simila mechanisms o cance
p og ession. On he con a y, non-p_C1/g_C2 g oup appea s o somehow ole a e new umo e en s, which leads
Cance sub ype Follow-up a iables Bac e ial communi ies P- alues
Adenoca cinoma (ADC)
1: p_C1_g_C1 2: p_C1_wo_g_C1 3: p_C2s2/g_C1 4: p_C2s2_wo_g_C1
Mo ali y
To al Dea hs 0.289
(11/38) 0.333
(36/108) 0.355
(11/31) 0.379
(108/285)
1 s 2: P = 0.6896; 1 s 3:
P = 0.6106; 1 s 4: P = 0.3709;
2 s 3: P = 0.8321; 2 s 4:
P = 0.4147; 3 s 4: P = 0.8476.
Days o dea h
Mean ± SD 725.7 ± 927.0 1072.7 ± 744.8 947.5 ± 779.5 699.4 ± 652.3
Median 434 880 737 553.5
Tumo ee
Numbe o Dea hs 0.056
(1/18) 0.133
(6/45)d0
(0/15) 0.095
(13/137)
1 s 2: P = 0.6621; 1 s 3: P = 1;
1 s 4: P = 1; 2 s 3: P = 0.3214;
2 s 4: P = 0.0338; 3 s 4:
P = 0.3655.
Wi h umo
Numbe o Dea hs 0.500
(4/8) 0.563
(18/32) 0.625
(5/8) 0.674
(58/86)
1 s 2: P = 1; 1 s 3: P = 1; 1 s 4:
P = 0.4377; 2 s 3: P = 1; 2 s 4:
P = 0.2848; 3 s 4: P = 1.
P ima y he apy ou come
Comple e emission 0.828
(24/29) 0.646
(51/79) 0.958
(23/24) 0.784
(171/218)
1 s 2: P = 0.0983; 1 s 3:
P = 0.2044; 1 s 4:
P = 0.8086; 2 s 3: P = 0.0018;
2 s 4: P = 0.0227; 3 s 4:
P = 0.0555.
Squamous Cell Ca cinoma
(SCC)
1: p_C1/g_C2 2: p_C1_wo_g_C2 3: p_C2s2
Mo ali y
To al Dea hs 0.621
(54/87) 0.336
(45/134) 0.441
(97/220) 1 s 2: P < 0.0001; 1 s 3:
P = 0.0053; 2 s 3: P = 0.0575
Days o dea h
Mean ± SD 921.4 ± 1144.5 868.2 ± 695.179 808.5 ± 823.4
Median 494.5 645.0 506.0
Tumo ee
Numbe o Dea hs 0.421
(16/38) 0.132
(9/68) 0.181
(20/110) 1 s 2: P = 0.0015; 1 s 3:
P = 0.0045; 2 s 3: P = 0.4131
Wi h umo
Numbe o Dea hs 0.941
(16/17) 0.625
(15/24) 0.829
(34/41) 1 s 2: P = 0.0281; 1 s 3:
P = 0.4151; 2 s 3 P = 0.0798
P ima y he apy ou come
Comple e emission 0.795
(31/39) 0.913
(84/92) 0.828
(125/151) 1 s 2: P = 0.0795; 1 s 3:
P = 0.6425; 2 s 3: P = 0.0853
Table 1. Pa ien ollow-up da a ac oss di e en bac e ial communi ies. Signi ican p- alues o Fishe ’s exac es
(p < 0.05) a e shown in bold.
9
Scien i ic RepoR S | (2019) 9:12838 | h ps://doi.o g/10.1038/s41598-019-49195-w
www.na u e.com/scien i ic epo s
www.na u e.com/scien i ic epo s/
o a decoupling o i s su i al cu e om he apid decline o p_C1/g_C2, and om he con inuous dec ease o
p_C2s2.
Less ema kable indings we e ob ained o B e undimonas/Acine obac e (p_C1/g_C1) communi y, which
did no di e ge om o he cases conce ning ADC ou comes. Howe e , i compa ed wi h SCC p_C1/g_C2 and
p_C2s2 g oups hose disclosed a end o ex ended su i al o li e ime p ognosis.
We also p o ide addi ional pieces o in o ma ion o a bu den o mic obio a in pulmona y disease. In he
TCGA coho , cases showing COPD co-mo bidi y we e linked o Ach omobac e , hus highligh ing a nega i e
e ec o his axon in SCC and in ai low obs uc ion. Howe e , COPD was no associa ed o any iden i ied bac-
e ial communi y in pa icula , no o augmen ed p e alence o Mo axella and Haemophilus gene a as equen ly
obse ed among hese pa ien s8,12,22. No iceably, only 27% o cases we e p esen ed wi h lung unc ion es s, and o
hose wi h COPD (11%) hei majo i y was classi ied as mild o mode a e cases (FEV > 50%; GOLD 1-2 s ages).
These ea u es could explain he simila alpha di e si y sco es ob ained o pa ien s wi h o wi hou COPD, once
p e ious epo s o a mic obio a loss we e cen e ed in ad anced cases22 ( o opposi e esul s19,40). Mo e s iking
is he lack o di e en ia ion ac oss LC s ages i aking in o accoun o me epo s o a loss o di e si y be ween
umo and non- umo samples17,18. None heless, he e is no p oo so a o a g adual decline o mic obio a wi h
cance p og ession. In con a y, p e ious wo ks unco e ed highe alpha di e si y alues in ad anced cases (IIIB
and IV) han in ea lie disease s ages18. In ou s udy, SCC cases we e in a e age mo e di e se han ADC, a esul
ha can be ela ed o a hea ie smoking load o hese pa ien s since ciga e e and ai pollu an s we e ound o
posi i ely a ec ai way mic obio a ichness17,23. Howe e , o he unknown ac o s mus play a ole in LC bac e ial
coloniza ion o explain he signi ican di e ences obse ed be ween SCC and ADC in hea y smoke s. Mo eo e
and consis en ly wi h a ecen s udy ca ied ou in mild o mode a e COPD12, we also epo ed simila di e si y
le els ac oss dis inc lung egions (b onchial and pe iphe al lung), con adic ing ea lie indings o a mic obio a
di e en ia ion in dispa a e lung ana omical egions10.
To ou knowledge his wo k ep esen s he la ges sc u iny o LC mic obio a. Sho ly, we unco e ed a p e-
dominance o P o eobac e ia among cance ous lungs a ea u e sha ed wi h o he ai way diso de s. Howe e ,
P o eobac e ia abundance is no uni e sal and a he dic a es a mic obio a subs uc u e independen ly o LC
sub ype, COPD co-mo bidi y, smoking his o y, age o lung egion. In SCC, we ound e idence o a di e en ial
e ec o bac e ial communi ies in pa ien su i al, pa icula ly when s a i ied in o an En e obac e iaceae clus e .
Gi en ha his axon has documen ed complica ions in pulmona y illnesses, we p oposed a con ibu ion o his
clus e o an in lamma o y cance mic oen i onmen , as well as, o o he pos -ope a i e and/o he apeu ic
non-cance complica ions. Finally, we belie e ha he disco e y o such well-de ined communi ies may shed
ligh in o bac e ia as p omising LC bioma ke s o pa ien s a i ica ion and as u u e p ognos ic ools o e en as
he apeu ic a ge s.
Da a A ailabili y
The da a used in his s udy is included in his manusc ip and in supplemen a y ma e ial iles. Addi ional iles
used o gene a e da a analysis a e a ailable om he co esponding au ho on easonable eques .
Re e ences
1. B ay, F. e al. Global cance s a is ics 2018: GLOBOCAN es ima es o incidence and mo ali y wo ldwide o 36 cance s in 185
coun ies. CA Cance J Clin 68, 394–424, h ps://doi.o g/10.3322/caac.21492 (2018).
2. Meine s, S., Eickelbe g, O. & Konigsho , M. Hallma ks o he ageing lung. Eu Respi J 45, 807–827, h ps://doi.
o g/10.1183/09031936.00186914 (2015).
3. Vogelmeie , C. F. e al. Global S a egy o he Diagnosis, Managemen , and P e en ion o Ch onic Obs uc i e Lung Disease 2017
Repo : GOLD Execu i e Summa y. Eu Respi J 49, h ps://doi.o g/10.1183/13993003.00214-2017 (2017).
4. Ve maelen, K. & B usselle, G. Exposing a deadly alliance: no el insigh s in o he biological links be ween COPD and lung cance .
Pulmona y pha macology & he apeu ics 26, 544–554, h ps://doi.o g/10.1016/j.pup .2013.05.003 (2013).
5. Liang, H. Y. e al. Fac s and ic ion o he ela ionship be ween p eexis ing ube culosis and lung cance isk: a sys ema ic e iew.
In e na ional jou nal o cance . Jou nal in e na ional du cance 125, 2936–2944, h ps://doi.o g/10.1002/ijc.24636 (2009).
6. Mao, Q. e al. In e play be ween he lung mic obiome and lung cance . Cance Le 415, 40–48, h ps://doi.o g/10.1016/j.
canle .2017.11.036 (2018).
7. Acu e Exace ba ions o Pulmona y Diseases. (2017).
8. Wang, Z. e al. Lung mic obiome dynamics in COPD exace ba ions. Eu Respi J 47, 1082–1092, h ps://doi.
o g/10.1183/13993003.01406-2015 (2016).
9. Milla es, L. e al. Func ional Me agenomics o he B onchial Mic obiome in COPD. PLoS One 10, e0144448, h ps://doi.o g/10.1371/
jou nal.pone.0144448 (2015).
10. E b-Downwa d, J. R. e al. Analysis o he lung mic obiome in he “heal hy” smoke and in COPD. PLoS One 6, e16384, h ps://doi.
o g/10.1371/jou nal.pone.0016384 (2011).
11. Eina sson, G. G. e al. Communi y dynamics and he lowe ai way mic obio a in s able ch onic obs uc i e pulmona y disease,
smoke s and heal hy non-smoke s. Tho ax 71, 795–803, h ps://doi.o g/10.1136/ ho axjnl-2015-207235 (2016).
12. P agman, A. A. e al. The lung issue mic obio a o mild and mode a e ch onic obs uc i e pulmona y disease. Mic obiome 6, 7,
h ps://doi.o g/10.1186/s40168-017-0381-4 (2018).
13. Came on, S. J. e al. Me agenomic Sequencing o he Ch onic Obs uc i e Pulmona y Disease Uppe B onchial T ac Mic obiome
Re eals Func ional Changes Associa ed wi h Disease Se e i y. PLoS One 11, e0149095, h ps://doi.o g/10.1371/jou nal.
pone.0149095 (2016).
14. Came on, S. J. S. e al. A pilo s udy using me agenomic sequencing o he spu um mic obiome sugges s po en ial bac e ial
bioma ke s o lung cance . PLoS One 12, e0177062, h ps://doi.o g/10.1371/jou nal.pone.0177062 (2017).
15. Lee, S. H. e al. Cha ac e iza ion o mic obiome in b onchoal eola la age luid o pa ien s wi h lung cance compa ing wi h benign
mass like lesions. Lung Cance 102, 89–95, h ps://doi.o g/10.1016/j.lungcan.2016.10.016 (2016).
16. Hosgood, H. D. 3 d e al. The po en ial ole o lung mic obio a in lung cance a ibu ed o household coal bu ning exposu es.
En i onmen al and molecula mu agenesis 55, 643–651, h ps://doi.o g/10.1002/em.21878 (2014).
17. Yu, G. e al. Cha ac e izing human lung issue mic obio a and i s ela ionship o epidemiological and clinical ea u es. Genome Biol
17, 163, h ps://doi.o g/10.1186/s13059-016-1021-1 (2016).