1
Scien i ic RepoR s | 6:24828 | DOI: 10.1038/s ep24828
www.na u e.com/scien i ic epo s
P olonged sleep es ic ion induces
changes in pa hways in ol ed
in choles e ol me abolism and
in lamma o y esponses
Vilma Aho1,*, Hanna M. Ollila1,2,3,4,*, E kki K onholm5, Isabel Bondia-Pons6,7, Pasi Soininen8,9,
An i J. Kangas8, Mika Hil o6, Ilkka Seppälä10, Johannes Ke unen2,8,9, Me i Oikonen11,
Emma Rai oha ju10, Tuulia Hyö yläinen6,7, Mika Kähönen12, Jo ma S.A. Viika i13,
Mikko Hä mä14, Mikael Sallinen14,15, Vesa M. Olkkonen16,17, Ha i Alenius18, Ma i Jauhiainen2,
Tiina Paunio2,3, Te ho Leh imäki10, Veikko Salomaa19, Ma ej O ešič6,7, Olli T. Rai aka i11,20,
Mika Ala-Ko pela8,9,21,22 & Ta ja Po kka-Heiskanen1
Sleep loss and insu icien sleep a e isk ac o s o ca diome abolic diseases, bu da a on how
insu icien sleep con ibu es o hese diseases a e sca ce. These ques ions we e add essed using wo
app oaches: an expe imen al, pa ial sleep es ic ion s udy (14 cases and 7 con ol subjec s) wi h
objec i e e i ica ion o sleep amoun , and wo independen epidemiological coho s (al oge he
2739 indi iduals) wi h ques ions o sleep insu iciency. In bo h app oaches, blood ansc ip ome and
se um me abolome we e analysed. Sleep loss dec eased he exp ession o genes encoding choles e ol
anspo e s and inc eased exp ession in pa hways in ol ed in in lamma o y esponses in bo h
pa adigms. Me abolomic analyses e ealed lowe ci cula ing la ge HDL in he popula ion coho s
among subjec s epo ing insu icien sleep, while ci cula ing LDL dec eased in he expe imen al sleep
es ic ion s udy. These indings sugges ha p olonged sleep dep i a ion modi ies in lamma o y and
choles e ol pa hways a he le el o gene exp ession and se um lipop o eins, inducing changes owa d
po en ially highe isk o ca diome abolic diseases.
1Depa men o Physiology, Facul y o Medicine, Uni e si y o Helsinki, Finland. 2Genomics and Bioma ke s uni and
Ins i u e o Molecula Medicine FIMM, Na ional Ins i u e o Heal h and Wel a e, Helsinki, Finland. 3Depa men
o Psychia y, Uni e si y o Helsinki and Helsinki Uni e si y Hospi al, Finland. 4S an o d Uni e si y Cen e o
Sleep Sciences, Palo Al o, CA, USA. 5Depa men o Ch onic Disease P e en ion, Popula ion S udies Uni , Na ional
Ins i u e o Heal h and Wel a e, Tu ku, Finland. 6VTT Technical Resea ch Cen e o Finland, Espoo, Finland. 7S eno
Diabe es Cen e A/S, Gen o e, Denma k. 8Compu a ional Medicine, Ins i u e o Heal h Sciences, Uni e si y o Oulu,
Oulu, Finland. 9NMR Me abolomics Labo a o y, School o Pha macy, Uni e si y o Eas e n Finland, Kuopio, Finland.
10Depa men o Clinical Chemis y, Fimlab Labo a o ies, and Uni e si y o Tampe e, School o Medicine, Tampe e,
Finland. 11Resea ch Cen e o Applied and P e en i e Ca dio ascula Medicine, Uni e si y o Tu ku, Tu ku, Finland.
12Depa men o Clinical Physiology, Uni e si y o Tampe e and Tampe e Uni e si y Hospi al, Tampe e, Finland.
13Depa men o Medicine, Uni e si y o Tu ku, and Di ision o Medicine, Tu ku Uni e si y Hospi al, Tu ku, Finland.
14B ain and Wo k Resea ch Cen e, Finnish Ins i u e o Occupa ional Heal h, Helsinki, Finland. 15Ago a Cen e ,
Uni e si y o Jy äskylä, Jy äskylä, Finland. 16Mine a Founda ion Ins i u e o Medical Resea ch, Helsinki, Finland.
17Ins i u e o Biomedicine, Ana omy, Uni e si y o Helsinki, Finland. 18Uni o Excellence o Immuno oxicology,
Finnish Ins i u e o Occupa ional Heal h, Helsinki, Finland. 19Depa men o Ch onic Disease P e en ion, Na ional
Ins i u e o Heal h and Wel a e, Helsinki, Finland. 20Depa men o Clinical Physiology and Nuclea Medicine,
Tu ku Uni e si y Hospi al, Tu ku, Finland. 21Oulu Uni e si y Hospi al, Oulu, Finland. 22Compu a ional Medicine,
School o Social and Communi y Medicine & Medical Resea ch Council In eg a i e Epidemiology Uni , Uni e si y o
B is ol, B is ol, Uni ed Kingdom. *These au ho s con ibu ed equally o his wo k. Co espondence and eques s o
ma e ials should be add essed o T.P.-H. (email: [email p o ec ed])
Recei ed: 25 Oc obe 2015
Accep ed: 05 Ap il 2016
Published: 22 Ap il 2016
OPEN
www.na u e.com/scien i ic epo s/
2
Scien i ic RepoR s | 6:24828 | DOI: 10.1038/s ep24828
Sho sleep du a ion, complain s o poo sleep quali y and diagnosed sleep p oblems ha e in epidemiological
s udies been associa ed wi h me abolic diso de s, which ela e o low-g ade ch onic in lamma ion, including
ca dio ascula diseases, me abolic synd ome, obesi y, and ype 2 diabe es melli us1–4. Howe e , he unc ional
pa hways and molecules ha media e hese e ec s a e la gely unknown.
Se e al p e ious s udies ha e shown ha expe imen al es ic ion o sleep o 4–5 h pe nigh o 1–2 weeks
ac i a es immune esponses5–7, down- egula es gene pa hways o mac omolecula biosyn hesis and me abolic
p ocesses8, and modi ies glucose me abolism by inducing insulin esis ance9–11. Howe e , he epo ed se um
lipid le els, cen al ac o s in he pa hogenesis o a he oscle osis, ha e shown only mild, inconsis en , o no
e ec s12–14, lea ing open he ques ion wha speci ic me abolic modula ions induced by sleep es ic ion could
explain he inc eased associa ion o es ic ed sleep o a he oscle osis, which is cha ac e ised by slow build-up
o lipid plaques in he walls o he a e ies, p omo ed by in lamma o y esponses and dec ease in mac ophage
e e se choles e ol anspo (RCT)15.
The du a ion o sleep es ic ion in p e ious s udies has anged om one o i e days, which in he de elop-
men o ch onic diseases is a sho pe iod. In ying o unde s and he ole o es ic ed sleep as a p edisposing
ac o o such diseases, he key ques ions a e: how do he sho - e m modi ica ions in me abolism and in lam-
ma ion de elop when he du a ion o he sleep es ic ion is p olonged, and a e hese modi ica ions such ha
could explain he associa ion be ween es ic ed sleep and inc eased isk o e.g. a he oscle osis, as e idenced by
he epidemiological s udies?
The assessmen o ci cula ing lipid p o iles using nuclea magne ic esonance (NMR) spec oscopy goes
beyond he ypically measu ed o al lipids, like o al choles e ol and iglyce ides, and allows de ailed cha ac e -
iza ion o many lipop o ein ea u es a he subclass le el, including he size o he pa icles, which a e impo an
in lipid physiology and pa hophysiology16,17. Analysis o ci cula ing lipid molecules based on mass spec ome y
(MS), which gi es an ex ensi e p o ile o indi idual lipid molecules bu does no quan i y lipop o ein- ela ed
measu es, has p e iously been applied in ci cadian esea ch18–21. Recen ly one sho - e m sleep dep i a ion
s udy22 and one pa ial sleep es ic ion s udy23 ha e applied MS-based lipid analyses in se um, while one s udy
used NMR o assess u ine me aboli es in sho - e m sleep dep i a ion24, bu NMR-based lipop o ein subclass
analyses ha e no been used o analyse he e ec s o sleep loss. The assessmen o he ela ionship be ween sleep/
sleep insu iciency and me abolomics p o iles in la ge epidemiological coho s has no , o ou knowledge, been
epo ed be o e.
The sho - e m e ec s o insu icien sleep we e assessed in a highly con olled expe imen whe e a g oup o
olun ee s es ic ed hei sleep o 4 hou s pe nigh du ing 5 days (sleep es ic ion, SR, N = 14 cases and N = 7 con-
ols). The esul s ocusing on he immunological e ec s a he le el o gene exp ession, cy okines, and CRP ha e been
p e iously published5. The pu a i e longe - e m e ec s we e assessed in eal-li e condi ions using wo independen
epidemiological coho s (DILGOM25, N = 518, and Young Finns S udy, YFS26, N = 2221), whe e he insu iciency
o sleep was e alua ed based on sel - epo ed sleep pa ame e s (subjec i e sleep insu iciency, SSI). Genome-wide
ansc ip ome and NMR-based me abolome we e ob ained om all h ee samples, and mass spec ome y-based
lipidome om he SR s udy pa icipan s (see low o he analyses depic ed in Supplemen a y Fig. S1).
Resul s
Subjec i e sleep insu iciency in epidemiological coho s. Pa ial sleep loss was induced expe imen-
ally as 4 h sleep/nigh o 5 nigh s o 14 heal hy young males in he SR s udy (including also 7 con ol subjec s;
o al N = 21; age (mean ± s.d.) 23.2 ± 2.2 y, Supplemen a y Table S1)5. To s udy sleep loss in eal li e condi ions,
subjec i e eeling o insu icien sleep was assessed using ques ionnai e in o ma ion in wo epidemiological sam-
ples. In he DILGOM subsample wi h in o ma ion o subjec i e sleep su iciency and omics da a (N = 472, 46%
men, age (mean ± s.d.) 51.9 ± 13.8 y, Supplemen a y Table S1)25, we used he ques ion ”Do you, in you opinion,
sleep enough?”. The answe op ions we e dicho omised, combining subjec s epo ing o sleep enough “almos
always” (N = 168) o “o en” (N = 218) o a pheno ype o ‘subjec i e su icien sleep’ (noSSI, N = 386). Subjec s
epo ing o “seldom o almos ne e ” (N = 86) sleep enough we e conside ed as ha ing ‘subjec i e sleep insu i-
ciency’ (SSI, N = 86, 18.2%).
In he Ca dio ascula Risk in Young Finns S udy (“Young Finns S udy”, YFS; N = 2221, 55% men, age
(mean ± s.d.) 37.7 ± 5.0 y, Supplemen a y Table S1) eplica ion coho 26, su iciency o sleep was assessed using
wo ques ions: one add essing sel - epo ed sleep du a ion (“How many hou s do you usually sleep pe nigh ?”)
and ano he on sel - epo ed sleep need (“How many hou s o sleep do you need pe day o eel well es ed?”).
Subjec i e sleep du a ion was sub ac ed om subjec i e sleep need, and indi iduals sleeping mo e han an hou
o e hei sleep need (N = 37, 1.7%) we e excluded om u he analyses. Remaining subjec s we e g ouped
in o h ee g oups based on hei le el o SSI: no (o only mild) SSI (sleep need – sleep du a ion = − 1… 0… 1 h;
N = 1825, 82.2%), mode a e SSI (1.5–2 h; N = 304, 13.7%), and hea y SSI (> 2 h; N = 55, 2.5%). Despi e he di -
e ences in he ques ions and age g oups, he o e all p e alence o SSI (mSSI o hSSI), 16.2%, in he YFS sample
was qui e simila o he 18.2% ound in he DILGOM. In acco dance wi h hese esul s, he p e alence o “sleep
deb ”, using closely simila c i e ion as used in he YFS sample, has been ea lie ound o be 20% in 1004 F ench
young adul s27.
Gene exp ession in lipid pa hways. Pa hway analysis o di e en ially exp essed genes. Gene exp es-
sion p o iles we e analysed om pe iphe al blood mononuclea cells (PBMC) in he SR s udy (N = 9 cases,
N = 4 con ols)5 and whole blood in he DILGOM coho (N = 472) using mic oa ays. Lipid- ela ed pa hways
we e en iched among ansc ip s down- egula ed a e 5 nigh s o expe imen al SR5. In he epidemiological
DILGOM sample, linea eg ession was used o co ela e RNA exp ession wi h SSI, adjus ing o age and gende .
T ansc ip s om 725 genes (2% o he o al 35,420 ansc ip s analysed) had lowe exp ession among subjec s
wi h SSI (N = 86) compa ed o subjec s wi h no sleep complain s (N = 386) (poin wise P < 0.05). En ichmen
www.na u e.com/scien i ic epo s/
3
Scien i ic RepoR s | 6:24828 | DOI: 10.1038/s ep24828
o biological p ocesses among hese genes was analysed using DAVID Func ional Anno a ion Clus e ing28. Fi e
Gene On ology (GO) clus e s had en ichmen sco es > 1.3 ( e e ing o geome ical mean o he P alues o he
pa hways < 0.05) in subjec s wi h SSI (Supplemen a y Table S2).
The GO clus e 5 (“Lipid clus e ”, P = 0.045, Supplemen a y Table S2 and Supplemen a y Fig. S2) included
4/5 o he op pa hways ha we e en iched among down- egula ed genes in he SR s udy (pe mu ed P < 0.001)
epo ed p e iously (Table1)5. The Lipid clus e emained signi ican (P < 0.05) also a e including BMI as a
co a ia e in he explana o y model.
The common pa hways in he expe imen al SR s udy and he DILGOM sample we e: Choles e ol anspo ,
S e ol anspo (bo h P = 0.048), Choles e ol homeos asis and S e ol homeos asis (bo h wi h a bo de line sig-
ni icance o P = 0.052) (Table1). The genes con ibu ing o hese GO pa hways in he DILGOM sample we e
ATP-binding casse e, sub- amily G, membe 1 (ABCG1), ca eolin 1 (CAV1), Niemann-Pick disease, ype C1
(NPC1), and Niemann-Pick disease, ype C1, gene-like 1 (NPC1L1), while in he expe imen al SR s udy hey
we e ABCA1 and NPC1. All genes and pa hways o he Lipid clus e a e shown in Supplemen a y Fig. S2.
Gene exp ession eplica ion. Gene exp ession was measu ed in he YFS eplica ion sample (N = 1407)
om whole blood using Illumina mic oa ays o e alua e whe he he exp ession o he genes ound in he
down- egula ed pa hways in DILGOM (ABCG1, CAV1, NPC1, and NPC1L1) was lowe in subjec s wi h SSI
also in his sample. The ABCG1 inding eplica ed (P < 0.05) in his coho , suppo ing he supp essi e e ec o
subjec i e sleep loss on his choles e ol anspo e .
Se um lipids and lipop o eins. NMR me abolomics. Nex , we examined whe he he ansc ip ional
changes we e e lec ed in he se um lipid and lipop o ein p o iles using NMR-based me abolomics analyses.
This high- h oughpu me hod p o ides concen a ion in o ma ion o o e 200 me abolic measu es, including
di e en sized VLDL, IDL, LDL, and HDL pa icles, a ious a y acids, amino acids, and small molecule ene gy
me aboli es29.
Concen a ion o lipop o ein pa icles and hei componen s. In he SR s udy (N = 14 cases, 6 con ols), he
numbe o small, medium, and la ge LDL pa icles (P a e co ec ion < 0.01) as well as small VLDL pa icles
(P < 0.05) was dec eased a e SR compa ed o BL, while he e we e no changes in he numbe o small and
medium size HDL pa icles (Fig.1A, Supplemen a y Table S3). La ge HDL pa icles showed a end o inc ease
du ing SR (P be o e co ec ion o mul iple es ing < 0.05, no signi ican a e co ec ion). These changes we e
also e lec ed by changes in LDL/VLDL s uc u al p o ein apoB-100. ApoB-100 le els dec eased (P < 0.005)
whe eas apoA-I (majo s uc u al p o ein in HDL) le els did no change.
In he DILGOM sample (N = 414), he numbe o se um la ge HDL pa icles was lowe among indi iduals
wi h SSI (poin wise P < 0.05). SSI had an independen associa ion o la ge HDL concen a ion also a e adding
BMI as a co a ia e in he model (P < 0.05). The e was no signi ican di e ence in he le els o any o he LDL o
VLDL subclasses, al hough he e was a consis en end o inc ease in bo h LDL and VLDL pa icles o all sizes
(Fig.1B, Supplemen a y Table S3).
The lowe numbe o la ge HDL pa icles in subjec s wi h SSI eplica ed (P < 0.005) in he YFS sample
(N = 2077) (Fig.2). In his sample, numbe o XL HDL was also lowe (P < 0.005). No di e ences we e obse ed
in small o medium HDLs. Fo he epidemiological samples, age and gende we e adjus ed o in he linea eg es-
sion model. The dec ease in la ge HDL was independen ly associa ed wi h SSI (P < 0.01) also when p obable
sel - epo ed obs uc i e sleep apnoea (OSA) was included in he model.
Mass spec ome ic measu emen s. In he SR s udy (N = 14 cases, N = 7 con ols), molecula lipids we e u he
analysed wi h MS-based lipidomics. Al oge he 20 lipids we e inc eased a e SR in he sleep-dep i ed cases
Gene On ology Pa hway Expe imen al SR Epidemiological SSI
GO ID Name Genes N P alue Pe mu ed P
Con ibu ing
genes P alue
Con ibu ing
genes
GO:0032365 in acellula lipid anspo 9 1.71E-05 0.001 ABCA1, CPT1B – –
GO:0030301 choles e ol anspo 8 1.79E-04 0.001 ABCA1, NPC1 0.048 ABCG1, CAV1,
NPC1, NPC1L1
GO:0015918 s e ol anspo 8 1.79E-04 0.001 ABCA1, NPC1 0.048 ABCG1, CAV1,
NPC1, NPC1L1
GO:0042632 choles e ol homeos asis 8 1.79E-04 0.001 ABCA1, NPC1 0.052 ABCG1, CAV1,
NPC1, NPC1L1
GO:0055092 s e ol homeos asis 8 1.79E-04 0.001 ABCA1, NPC1 0.052 ABCG1, CAV1,
NPC1, NPC1L1
Table 1. Lipid pa hways down- egula ed in expe imen al sleep es ic ion (SR) and epidemiological
subjec i e sleep insu iciency (SSI). 4/5 o he op Gene On ology (GO) Biological P ocesses ha we e ound
en iched among down- egula ed ansc ip s in he expe imen al SR (pe mu ed P < 0.001)5 we e also iden i ied
among he genes wi h lowe exp ession in DILGOM subjec s wi h SSI. These pa hways we e in ol ed
in (chole)s e ol anspo and homeos asis, and con ibu ed o he “Lipid clus e ” (Clus e 5, P = 0.045,
Supplemen a y Fig. S2). Down- egula ion o he NPC1 gene was sha ed in bo h samples.
www.na u e.com/scien i ic epo s/
4
Scien i ic RepoR s | 6:24828 | DOI: 10.1038/s ep24828
as compa ed o he con ols (Supplemen a y Table S4). The inc eased lipids comp ised mos ly polyunsa u a ed
phospha idyle hanolamines (PE) and phospha idylcholines (PC) (P < 0.05; Supplemen a y Table S4).
Ne wo k analysis on expe imen al SR. The indings om he SR s udy sugges ed ha choles e ol ans-
po had declined and in lamma ion inc eased. An impo an ansc ip ional egula o o he RCT is he nuclea
li e X ecep o (LXR) signalling, which p omo es RCT and dec eases in lamma ion30–32. We hus hypo he-
sized ha dec ease in LXR ac i i y could be media ing he e ec s o sleep es ic ion on he immune sys em and
me abolism (Fig.3, Table2). Al e na i ely, he ac i i y o lipid ans e p o eins could be changed. To s udy hese
hypo heses and he associa ions be ween he obse ed changes, we selec ed ele an immunological, me abolic,
and sleep a iables measu ed in he expe imen al SR s udy (lis ed in Supplemen a y Table S5), including he
pa ame e s whe e changes had been de ec ed also in he epidemiological coho s, and pe o med a dependency
ne wo k analysis (Supplemen a y Fig. S3).
Immunological pa ame e s. In lamma ion supp esses LXR ac i i y ia oll-like ecep o s (TLR)33–35. We ha e
ea lie shown ha he gene coding o TLR4 was up- egula ed in ou 5 nigh s’ SR p o ocol5. Also p oin lamma-
o y cy okines in e leukin 1β (IL-1b) and umou nec osis ac o α (TNF-a) ha e been shown o supp ess LXR
ac i i y36 (Fig.3), and hese cy okines ha e been consis en ly shown o inc ease in expe imen al sleep es ic-
ion7,37. We obse ed highe exp ession o he gene encoding o TNF-a in ee-li ing indi iduals wi h SSI in he
DILGOM sample (poin wise P < 0.05), and genes encoding o bo h IL-1b and TNF-a in he Young Finns epli-
ca ion sample (P < 0.005 and P < 0.05, espec i ely) (Figs2 and 3). In addi ion o he cy okines, up- egula ion o
in lamma ion- ela ed genes encoding o TLR4 (Fig.2), MyD88 (an essen ial signal ansduce in he IL-1 and
TLR pa hways), inducible p os aglandin endope oxide syn hase (cyclooxygenase 2, PTGS2), and Fas cell su ace
dea h ecep o (FAS) – obse ed in he expe imen al SR s udy – eplica ed in he YFS sample (Fig.3). Inc eased
in lamma ion could be u he augmen ed h ough dec eased LXR ac i i y30.
Dependency ne wo k analysis. In o de o dis inguish di ec and indi ec in e ac ions o hese immune, me a-
bolic and sleep a iables, we u ilised undi ec ed Gaussian g aphical model whe e he a iables a e connec ed i
and only i hei pa ial co ela ion is signi ican ly non-ze o (using FDR mul iple es ing o he selec ion o edges)
o isualise he dependencies as a ne wo k38.
The lipids ha we e inc eased a e SR associa ed o se e al immunological pa ame e s (Supplemen a y Fig. S3;
P < 0.05 o all associa ions shown in he igu e). The co ela ion analysis e ealed signi ican posi i e associa ions
wi h PE(38:3), PE(38:5e), PE(36:2e), ChoE(16:1) and B-cells. PE(38:3) was u he nega i ely co ela ed wi h
TNF-a, which was posi i ely co ela ed wi h in e leukins 18 and 1 and in e e on-γ (IFNG). LXRA associa ed
posi i ely wi h B cells and TLR8, and nega i ely wi h ABCA1, NRIP1 and LXRB. TLR4 had a s ong nega i e
associa ion wi h NPC1 and ABCA1 (Supplemen a y Fig. S3), as would be expec ed i LXR ac i i y was dec eased.
In he con ol subjec s (Supplemen a y Fig. S4), he associa ions be ween he a iables we e ewe and weake
han in he expe imen al g oup. Pa icula ly, sca ce associa ions be ween he lipids and he immunological a i-
ables we e obse ed. As he ne wo ks a e baseline-co ec ed, hey show he changes occu ing be ween baseline
and sleep es ic ion imepoin . As he con ol g oup was no subjec ed o any ea men (besides s aying in he
labo a o y) be ween hese imepoin s, i was expec ed ha no majo changes would be de ec ed.
Figu e 1. Changes in lipop o ein pa icles in (A) expe imen al sleep es ic ion (SR) and (B) epidemiological
subjec i e sleep insu iciency (SSI). Concen a ion di e ences o di e en sized e y low densi y (VLDL),
in e media e densi y (IDL), low densi y (LDL), and high densi y (HDL) lipop o ein pa icles. (A) Expe imen al
SR compa ed o baseline (BL, no malised o 1) (*P < 0.05, pai ed es ; N = 14). (B) DILGOM subjec s wi h SSI
compa ed o subjec s wi hou SSI (noSSI, no malised o 1) (*poin wise P < 0.05, linea modelling adjus ing o
sex and age; N = 414).
www.na u e.com/scien i ic epo s/
5
Scien i ic RepoR s | 6:24828 | DOI: 10.1038/s ep24828
Lipid ans e p o ein and enzyme ac i i ies in SR. Lipid ans e p o eins comp ise he wo key eg-
ula o s o se um lipid balance be ween lipop o eins. Howe e , we did no ind signi ican changes in he ac i i y
o choles e ol es e ans e p o ein (CETP) o phospholipid ans e p o ein (PLTP) in he expe imen al SR
s udy. Fu he mo e, no changes we e obse ed in he ac i i y o leci hin-choles e ol acyl ans e ase (LCAT), he
enzyme con e ing HDL unes e i ied choles e ol o choles e yl es e , o he a he op o ec i e enzyme pa aoxonase
1 (PON1) linked o an ioxida i e p ope ies o HDL pa icles.
Discussion
The main inding o he p esen s udy was ha es ic ion o sleep ei he expe imen ally o in na u al li ing
condi ions modi ied lipop o ein me abolism and immune esponses. The changes we e de ec ed a he le el o
ansc ip ome as well as in he ci cula ing lipid p o ile. The ansc ip ional changes - he down- egula ion o
e e se choles e ol anspo - ela ed gene pa hways - we e simila in he ela i ely sho e m exposu e o he
expe imen al sleep es ic ion and in he epidemiological coho s among subjec s epo ing insu icien sleep.
In e es ingly, he lipop o ein p o ile showed dec eased LDL in expe imen al SR, esembling LDL changes in
acu e-phase esponse39, bu in epidemiological SSI la ge HDL was dec eased. Dec ease in HDL has been ega ded
as one impo an isk ac o o ca dio ascula diseases40.
The classical iew on he ela ionship be ween se um lipid le els and isk o ca dio ascula e en s, o mula ed
based on he indings o he F amingham s udy41 and since con i med by many epidemiological s udies, s a es
ha se um high LDL and low HDL choles e ol le el is a isk combina ion o ca dio ascula e en s42. These
Figu e 2. Replica ion in he Young Finns S udy. La ge (L) and ex a-la ge (XL) HDL dec eased wi h
inc easing le el o subjec i e sleep insu iciency (SSI) (P < 0.005; N = 2077), whe eas he exp ession o
in e leukin 1 β (IL1B) and oll-like ecep o 4 (TLR4) genes was highe in subjec s wi h SSI (P < 0.005 and
P < 0.05, espec i ely; N = 1407). NoSSI = no o only mild SSI, mSSI = mode a e SSI, hSSI = hea y SSI.
HDL g aphs ep esen mean ± s.e.m. concen a ions in se um. Gene exp ession is shown ela i e o he
mean exp ession in he noSSI g oup ( ela i e mean ± s.e.m.). E ec o SSI on HDL concen a ions and gene
exp ession was modelled wi h linea eg ession adjus ing o age and sex.

www.na u e.com/scien i ic epo s/
6
Scien i ic RepoR s | 6:24828 | DOI: 10.1038/s ep24828
Figu e 3. Summa y o he indings om he expe imen al and he epidemiological s udies. (A) P oposed
model o explain he indings. 1 Sleep loss ac i a es in lamma o y esponses h ough oll-like ecep o s
(TLR)5 2supp essing li e X ecep o (LXR) ac i i y33–35. 3 Dec eased LXR ac i i y leads o dec eased
e e se choles e ol anspo (RCT) and syn hesis o a y acids (FA) and iglyce ides (TG), and inc eased
immunological ac i a ion30–32. (Red a ows showing inc ease, g een a ows dec ease.) (B) The igu e
summa ises ou indings om ansc ip omics and NMR me abolomics in expe imen al (E) sleep es ic ion
(SR; N = 21), and in he DILGOM epidemiological coho (D; N = 518) and Young Finns S udy eplica ion
coho (Y; N = 2221) subjec s wi h subjec i e sleep insu iciency (SSI). Numbe s 1 and 3 e e o he loca ions
in he model p oposed in Fig.3A. (a) Up- egula ion o TLR and o he in lamma o y genes/gene pa hways in SR
epo ed in5. Pa hway analysis o up- egula ed genes in subjec s wi h SSI in DILGOM con i med up- egula ion
o B-cell ac i a ion, lymphocy e ac i a ion, and immune sys em de elopmen (P < 0.05) also a epidemiological
le el. Indi idual genes showed only non-signi ican ends o highe exp ession in SSI. (b) TLR4 and se e al
www.na u e.com/scien i ic epo s/
7
Scien i ic RepoR s | 6:24828 | DOI: 10.1038/s ep24828
indings ha e encou aged e o s o de elop ea men s wi h he aim o ei he o lowe LDL le els (e.g. s a ins) o
inc ease HDL le els (e.g. CETP inhibi o s and niacin). Pa icula ly he la e app oach has been a disappoin men :
pha macological inc ease o se um HDL le els has no a ec ed he isk o ca dio ascula diseases42, indica ing
ha he me e se um HDL choles e ol concen a ion is no a su icien me ics o explain i s epidemiologically
e i ied ca diop o ec i e e ec . HDL can p o ec agains a he oscle osis by mul iple mechanisms, including
he abili y o e lux choles e ol om endo helial mac ophages (mac ophage e e se choles e ol anspo )43, o
h ough an i-in lamma o y44, an ioxida i e45 and an iapop o ic46 pa hways. A newly disco e ed egula ion by
mic oRNAs adds o he complexi y o he ask o disco e mechanisms ha explain he ela ionship be ween HDL
and ca dio ascula disease (CVD) isks47. Mo eo e , modula ions in HDL pa icle composi ion can ans o m i
o dys unc ional and, h ough his modula ion, inc ease he isk o CVD40.
Combining he knowledge o he epidemiologically e i ied inc eased isk o CVD associa ed wi h he low
HDL/high LDL lipid p o ile and sho /insu icien sleep, we expec ed o measu e such lipid p o iles in SR and in
pe sons wi h SSI. This p edic ion did no p o e qui e co ec . In SR we measu ed, using he NMR me abolom-
ics, dec eased le els o LDL while we ound no signi ican changes in HDL le els. In he SSI subjec s we ound
dec eased HDL le els bu no signi ican di e ences in LDL le els. The one week expe imen al SR may ha e been
oo sho o a ec he numbe o HDL pa icles, while in he epidemiological samples, he numbe o la ge HDL
pa icles was lowe among hose who epo ed insu icien sleep. Since he na u e o condi ions ha induce insu -
icien sleep is o en pe sis en 48–51, we a gue ha he SSI subjec s epo ing insu icien sleep had been exposed o
insu icien sleep o a longe pe iod han hose in he i e day expe imen al sleep es ic ion s udy. We p opose
ha he lipid p o ile is modula ed in he cou se o exposu e o insu icien sleep om low LDL le els a ea ly phase
o sleep insu iciency o low HDL le els upon longe exposu e. Howe e , in a c oss-sec ional s udy we canno
p o ide di ec e idence on he du a ion o SSI o hose who epo ed i . The low LDL concen a ion in he SR
may ha e been induced by he in lamma ion, pa icula ly by he ac i a ion o he acu e phase esponse, induced
by he sleep es ic ion5.
While some p e ious gene exp ession s udies ha e indica ed ha choles e ol/lipid me abolism is egula ed by
sleep-wake cycle and sleep es ic ion could modi y i 8,52,53, de ailed cha ac e iza ion o lipid p o iles has emained
sca ce. I has been shown ha se um iglyce ide le els dec eased ollowing es ic ion o sleep o 4 hou s on 5
in lamma ion- ela ed genes had highe exp ession among subjec s wi h SSI in he eplica ion sample Young
Finns S udy. (c) Inc ease in p oin lamma o y cy okines IL-1b and TNF-a assessed by in i o s imula ion o
whi e blood cells and epo ed in7. (d) Highe exp ession o genes encoding o p oin lamma o y cy okines a
epidemiological le el in subjec s wi h SSI. (e) Down- egula ion o genes/gene pa hways o e e se choles e ol
anspo assessed wi h ansc ip omics and epo ed in he p esen publica ion. Concen a ions o la ge HDL
in se um measu ed using NMR me abolomics. ( ) Ace yl-CoA ca boxylase (ACC), he a e-limi ing enzyme
o FA syn hesis, was down- egula ed in SR and DILGOM SSI. No majo di e ences we e obse ed in he
exp ession o o he FA and TG syn hesis genes in SR o SSI. (g) FA and TG measu ed wi h NMR me abolomics
in expe imen al SR and DILGOM. Pai ed es s used o compa ing SR o BL, and linea eg ession used
o modelling he e ec o SSI on gene exp ession o lipid concen a ion, adjus ing o age and sex. See
abb e ia ions in Table2.
Abb e ia ion De ini ion
TLR4, TLR2 oll-like ecep o s 4 and 2
NF-kB nuclea ac o kappa B
MyD88 myeloid di e en ia ion p ima y esponse 88
PTGS2 p os aglandin syn hase 2 = inducible c yclooxygenase
iNOS inducible ni ic oxide syn hase
IL6 in e leukin 6
FAS Fas cell su ace dea h ecep o
IL-1b in e leukin 1 β
TNF-a umou nec osis ac o α
ABCA1, ABCG1 ATP-binding casse e (ABC) anspo e s A1 and G1
NPC1 Niemann-Pick disease 1
ARL7 ADP- ibosyla ion ac o -like 7
PLTP phospholipid ans e p o ein
L HDL la ge high densi y lipop o ein pa icles
SREBF1 s e ol egula o y elemen binding ansc ip ion ac o 1
FASN a y acid syn hase
ACC ace yl-CoA ca boxylase
FA a y acids
TG iglyce ides
Table 2. Va iables in Fig.3B.
www.na u e.com/scien i ic epo s/
8
Scien i ic RepoR s | 6:24828 | DOI: 10.1038/s ep24828
consecu i e nigh s12, and also ha myelope oxidase-modi ied LDL pa icles inc eased unde simila condi ions13.
In a ecen s udy, no changes in blood o al choles e ol, LDL, HDL, o iglyce ides we e ound a e sleep es ic-
ion o 4 hou s on 5 nigh s14. Two ecen s udies ha e measu ed lipid p o iles in human blood samples using mass
spec ome y analysis a e pola and non-pola ex ac ion o he samples a e a sho e m (24 h) o al sleep
es ic ion22 and a pa ial sleep es ic ion o i e days23. In bo h s udies, he majo i y o he al e ed, iden i ied
species, consis ed o lipid species (including phospholipids, sphingolipids, acylca ni ides and phospha idylcho-
lines). In he acu e SR s udy22, only inc eases in lipid le els we e obse ed a e he dep i a ion pe iod, while
in he pa ial SR s udy23 also declines we e eco ded. In he p esen SR s udy, all signi ican changes om BL o
SR, measu ed using he mass spec ome y echnique, we e inc eases, mainly in phospha idyl e hanolamides,
phopha idyl cholines, iglyce ides and choles e ol es e s. In spi e o some disc epancies in he esul s, possi-
bly explained by di e en ex ac ion and o he me hods and/o di e ences in he expe imen al a angemen s,
hese h ee independen s udies poin ou ha es ic ion o sleep, ei he acu ely o pa ially o a longe pe iod,
signi ican ly a ec s se um lipid le els, as measu ed using mass spec ome y. In he p esen s udy, also NMR
spec oscopy measu emen was able o ack changes in lipid p o iles unde condi ions o es ic ed sleep, u he
con i ming ha his condi ion modi ies lipid me abolism.
The esul s om he gene exp ession pa hways we e mo e consis en han hose o he lipid p o iles. The cho-
les e ol pa hways ha we e down- egula ed in SR and SSI included membe s o he ATP-binding casse e amilies
A and G (ABCA1 and ABCG1). ABCA1 and ABCG1 acili a e he e lux o ee choles e ol and phospholipids
om mac ophage- oam cells o HDL pa icles. O all choles e ol ca ied by HDL, he p opo ion o igina ing
om pe iphe al mac ophages is low compa ed o li e (75%) and in es ine (20%) p oduced HDL. Howe e , his
ou e o emo al o choles e ol (mac ophage RCT) has been ega ded as an impo an (bu no only) componen
in e aining choles e ol balance in a e ial endo helium43. Since ABCA1 and ABCG1 wo k in andem o acili-
a e choles e ol emo al om he mac ophages54, he educ ion in hei exp ession would implica e a educed
po en ial o hese monocy e-de i ed mac ophages o eg ess choles e ol o HDL accep o s55. ABCA1 de iciency
in Tangie disease leads o e y low o absen HDL le els and la ge VLDL pa icles56. Since he ela ionship
be ween choles e ol anspo e exp ession in whi e blood cells and se um HDL concen a ion is a om simple,
i is di icul o e alua e how he obse ed exp ession changes we e e lec ed o he low o choles e ol om mac-
ophages o li e and inally eces (RCT), pa icula ly as he ac i i ies o he enzymes/ anspo e s modula ing
HDL pool (CETP, PLTP, LCAT, PON1) measu ed in he SR s udy we e no a ec ed by. Howe e , i can be no ed
ha he obse ed exp ession changes in he choles e ol anspo - ela ed genes a e compa ible wi h hose in
down- egula ion o LXR ac i i y39,57 (Fig.3).
Ca eolin-1 (CAV1) is ound in ca eolae, choles e ol- ich memb ane lipid o ma ions ha a e in ol ed in
choles e ol a ic and homeos asis58 and egula ion o in lamma ion59 among o he unc ions. CAV1 has been
p oposed o pa icipa e in he egula ion o choles e ol e lux o HDL60,61, possibly in in e ac ion wi h ABCG161.
Some o he s udies ha e no ound an e ec 62,63, and he o e all e ec o CAV1 on choles e ol e lux is complex61.
The obse ed dec ease in CAV1 exp ession may add o he iew ha e e se choles e ol anspo om mac-
ophages is comp omised by insu icien sleep.
P e ious s udies indica e ha he immune sys em is ac i a ed du ing expe imen al sleep es ic ion, and in
pe sons who in epidemiological s udies epo sho sleep du a ion ( e iewed in64,65). We ha e ea lie shown ha
cy okine (in e leukins IL-1b, IL-6 and IL-17) elease as esponse o in i o immunological challenge is inc eased
a e sleep es ic ion, as well as se um acu e phase p o ein CRP7. A gene exp ession le el, he ac i a ion included
inc eases in exp ession o oll-like ecep o s, NF-kB signalling pa hway and in e leukin-8 p oduc ion pa hways5.
Many acu e condi ions, including su gical auma, myoca dial in a c ion and in lamma ion, induce he sys-
emic acu e phase eac ion (APR), wi h acu e phase p o ein p oduc ion in he li e as an in eg al pa o he hos
de ence esponse. In addi ion o changes in p o ein syn hesis, also lipid me abolism unde goes signi ican mod-
i ica ions in APR66,67. Me abolic modi ica ions aim a op imiza ion o he de ence, and would e u n o baseline
le el a e he need o de ence has disappea ed68. Howe e , i he in lamma ion-sus aining condi ion con inues,
as p esumably in he epidemiological SSI g oups o he p esen s udy, he me abolic modi ica ions may de elop
u he and become ch onic39,68. Mo eo e , he modi ica ions di e acco ding o he na u e o he e en ha
a oused he eac ion, e.g. su gical auma and in ec ion66,68. Dec eases in HDL and LDL choles e ol, as well as
RCT, belong o he me abolic adap a ions ound in APR du ing in lamma o y a ack67. While cy okines (includ-
ing IL-1b, IL-6 and TNF-a) a e cen al egula o s o he APR, ecen esea ch sugges s ha an impo an pa o
he me abolic egula ion in APR is channelled ia nuclea ecep o s, including LXR, PPAR and LHR-139. Thus
bo h he down- egula ion o RCT pa hway and up- egula ion o he in lamma o y pa hways could esul om a
dec ease o LXR ac i i y (Fig.3).
Me hodological issues. To e alua e sleep in he epidemiological samples, we used subjec i e e alua ion
o sleep insu iciency, ins ead o sleep du a ion. While sleep du a ion as a measu e o sleep has many ad an-
ages, in e alua ion o sleep insu iciency i has one conside able disad an age: i canno di e en ia e sleep
insu iciency-inducing sho sleep om na u al sho sleep69.
In epidemiological s udies sleep du a ion and sel - epo ed sleep insu iciency ha e shown independen e ec s
on ca dio ascula ou comes (including hype ension, choles e ol le el and ca diac e en s)70 and pe o mance71.
Thus, al hough he co ela ion o sho sleep du a ion and subjec i e sleep insu iciency is usually high (in he
p esen s udy, P = 2.5E-10 and β = − 0.5 h in DILGOM, and P = 4.9E-93 and β = − 0.5 h in YFS, Supplemen a y
Fig. S5), hese a iables may ep esen pa ly o e lapping, bu sepa a e physiological en i ies69,70. In epidemiolog-
ical s udies, a ques ion add essing sleep need and/o subjec i e eeling o sleep insu iciency may help o sepa a e
na u al sho sleepe s om hose ha don’ ge enough sleep.
While he pa icipan s o he expe imen al s udy we e ca e ully sc eened no o ha e any (sleep) diso de s, he
pa icipan s in he epidemiological s udies epo ed di e en medical condi ions, including sleep apnoea, which
www.na u e.com/scien i ic epo s/
9
Scien i ic RepoR s | 6:24828 | DOI: 10.1038/s ep24828
is known o be associa ed o bo h insu icien sleep and ca dio ascula diseases. The e alua ion o he po en ial
e ec s o OSA on ou esul s e ealed ha al hough he subjec s epo ing symp oms o OSA epo ed also mo e
sleep insu iciency, OSA did no explain he di e ence in HDL, as SSI was independen ly associa ed o lowe le el
o la ge HDL.
The main egula ion o me abolism akes place in o he issues (e.g. li e , panc eas) han blood cells, which
we e used in he gene exp ession analyses. Thus i is unclea o which ex en he esul s e lec changes in he
ac ual me abolic p ocesses. An addi ional sou ce o a ia ion lies in he di e ences in age and gende be ween
he popula ions s udied. The e ec s o hese ac o s ha e been add essed by including hem in he linea models
used o s a is ical analyses.
Sleep loss is o en accompanied wi h ci cadian misalignmen , also known o cause changes in he egula ion o
he immune sys em and me abolism and associa e o ca dio ascula diseases72,73. In ou expe imen al SR s udy,
he modes delay (16 min) in ci cadian hy hm, measu ed in he mo ning sali a y co isol peak, is unlikely o
signi ican ly modula e he esul s, bu he e ec canno be en i ely uled ou .
In eal li e, people equen ly expe ience consecu i e cycles o sleep es ic ion and eco e y sleep (e.g., shi
wo ke s, people wo king ex ended days). O en eco e y sleep be ween wo pe iods o sleep es ic ion emains
incomple e and induces a ca y-o e e ec 74. In he SR expe imen , pa icipan s we e subjec ed o only one cycle.
F om his pe spec i e, he esul s o he SR s udy may unde es ima e wha ac ually happens o he choles e ol
me abolism o sleep- es ic ed indi iduals in eal li e.
We p opose ha sleep es ic ion a ouses a hos de ence esponse ha sha es ea u es wi h APR a oused by
auma and in ec ion. These esponses modula e lipid me abolism, possibly h ough LXRs, and in he long un
may con ibu e o he inc eased isk o ca dio ascula diseases.
Conclusion
Sleep es ic ion-induced dec ease in exp ession o genes in pa hways ela ed o e e se choles e ol anspo
om mac ophages, in combina ion wi h in lamma o y ac i a ion, may a leas pa ly explain he inc eased isk
o ca dio ascula diseases es ablished in epidemiological s udies o pe sons wi h sho /insu icien sleep.
Me hods
Samples. Expe imen al sleep es ic ion. An expe imen al schedule simula ing a wo king week wi h
es ic ed sleep was execu ed a he B ain and Wo k Resea ch Cen e o he Finnish Ins i u e o Occupa ional
Heal h (FIOH). We ha e ea lie epo ed changes obse ed in glucose me abolism75, cy okines, whi e blood
cell subpopula ions, and C- eac i e p o ein7, leukocy e gene exp ession5, as well as cogni i e pe o mance76
om his SR expe imen . S udy pa icipan s we e men (N = 21; age 19–29 yea s; mean ± s.d. 23.2 ± 2.2 yea s;
Supplemen a y Table S1) wi h egula sleep wake cycle ha was checked by w is -wo n ac ig aphy and sleep
dia ies du ing 1–2 weeks p io o he s udy. In he sleep labo a o y, he pa icipan s we e andomly alloca ed o
he expe imen al g oup (“cases”; N = 14) o he con ol g oup (“con ols”, N = 8) (Supplemen a y Fig. S1). Cases
spen he i s wo nigh s 8 h/nigh in bed (baseline, BL; om 23:00 h o 07:00 h), ollowed by i e nigh s o 4 h/
nigh in bed (sleep es ic ion, SR; om 03:00 h o 07:00 h). Con ols spen 8 h in bed e e y nigh . One con ol
subjec was excluded a e he EEG analysis as he had been sleeping less han 6 h/nigh du ing he expe imen .
To al sleep du a ion in he cases (N = 14) dec eased om 7 h 22 min ± 20 min in BL o 3 h 54 min ± 5 min in SR,
while i emained ela i ely cons an , 7 h 19 min ± 16 min o 7 h 26 min ± 17 min a he same imepoin s, in he
con ols (N = 7).
Du ing waking, main ac i i ies o he pa icipan s included aining and es ing o memo y and mo o asks
simula ing o ice wo k asks. These es s we e conduc ed du ing he day a 10:00–12:30 and 14:30–15:40, and
du ing he sleep es ic ion a 00:30–01:40. In he es oom he illumina ion anged om 150 o 400 lux and in
he li ing oom om 350 o 600 lux. Physical exe cise o lea ing he sleep labo a o y was no allowed du ing he
expe imen . Polyg aphy was measu ed con inuously and EEG sco ed acco ding o he Rech scha en-Kales man-
ual75,77. EEG da a om he p e ious nigh be o e mo ning blood sample- aking o imepoin s BL and SR we e
included in he dependency ne wo k analysis.
P es udy mean (± s.d.) body mass index (BMI) was 23.2 (± 2.4) o con ols and 23.5 (± 2.6) o cases.
S anda dised meals based on he Finnish nu i ion ecommenda ions o 18–30 yea s old no mal-weigh ed men
wi h low ac i i y78 we e ea en a ixed imes h oughou he expe imen : b eak as a 08:00 h (600 kcal), lunch a
12:30 h (800 kcal), dinne a 18:30 h (700 kcal); snacks a 15:30 h (300 kcal) and 21:30 h (200 kcal). In addi ion,
cases a e a piece o ui (apple o o ange) a 00:30 h (50 kcal) which did no exceed he es ima ed inc ease o
ene gy expendi u e du ing he SR79. No ca eine, alcohol, o obacco was allowed du ing he expe imen . Blood
samples we e collec ed a BL and SR imepoin s a 7.30 h a e as ing o e nigh .
Sali a co isol was assessed 10 imes pe day du ing he expe imen , and only a modes , 16 min on a e age,
delay in he ci cadian phase measu ed by he mo ning peak o co isol ( om mean ± s.d. 07:39 ± 0:14 in BL o
07:55 ± 0:11 in SR) was obse ed in he sleep-dep i ed cases75. No changes in he o e all co isol concen a ion
in SR compa ed o BL was obse ed7.
The DILGOM coho . The Die a y Li es yle and Gene ic de e minan s o Obesi y and Me abolic Synd ome
(DILGOM) s udy was o iginally pe o med as an ex ension o he FINRISK 2007 s udy (N = 7993; age 25–74
yea s; om i e geog aphical a eas in Finland). The na ional, c oss-sec ional FINRISK su eys ha e been ca ied
ou e e y 5 yea s since 1972 o assess he isk ac o s o ch onic diseases (e.g. CVD, diabe es, obesi y, cance ) and
heal h beha iou in he wo king age popula ion in Finland. The aim o he DILGOM s udy (N = 5024) was o
cha ac e ise isk ac o s o me abolic and ca dio ascula diseases in he Finnish popula ion bo h a he epide-
miological and a he gene ic le el25. In addi ion o ques ionnai e da a on heal h and li es yle, a blood sample was