1
Scien i ic RepoR s | 6:21323 | DOI: 10.1038/s ep21323
www.na u e.com/scien i ic epo s
Tempo al dynamics o Puumala
han a i us in ec ion in cyclic
popula ions o bank oles
Liina Vou ilainen1,2, E a R. Kallio3, Jukka Niemimaa1, Olli Vapalah i2,4 & Heikki Hen onen1
Unde s anding he dynamics o zoono ic pa hogens in hei ese oi hos popula ions is a p e equisi e
o p edic ing and p e en ing human disease epidemics. The human in ec ion isk o Puumala
han a i us (PUUV) is highes in no he n Eu ope, whe e popula ions o he oden hos (bank ole,
Myodes gla eolus) unde go cyclic luc ua ions. We conduc ed a 7-yea cap u e-ma k- ecap u e s udy
o moni o seasonal and mul iannual pa e ns o he PUUV in ec ion a e in bank ole popula ions
exhibi ing a 3-yea densi y cycle. In ec ed bank oles we e mos abundan in mid-win e mon hs du ing
yea s o inc easing o peak hos densi y. P e alence o PUUV in ec ion in bank oles exhibi ed a egula ,
seasonal pa e n e lec ing he annual popula ion u no e and accumula ion o in ec ions wi hin each
yea coho . In au umn, he PUUV ansmission a e acked inc easing hos abundance, sugges ing
a densi y-dependen ansmission. Howe e , p e alence o PUUV in ec ion was simila du ing he
inc ease and peak yea s o he densi y cycle despi e a wo old di e ence in hos densi y. This may esul
om he high p opo ion o indi iduals ca ying ma e nal an ibodies cons aining ansmission du ing
he cycle peak yea s. Ou excep ionally in ensi e and long- e m da ase p o ides a solid basis on which
o de elop models o p edic he dynamic public heal h h ea posed by PUUV in no he n Eu ope.
Unde s anding he dynamics o zoono ic pa hogens in hei ese oi hos popula ions is an impo an s ep
owa ds p edic ing he isk o zoono ic diseases o humans, such as in ec ions caused by han a i uses ha accoun
o ca. 50 000 cases wo ldwide each yea 1,2. Han a i uses ( amily Bunya i idae) a e ca ied by oden s, so ico-
mo phs, and ba s3. Some o hem cause haemo hagic e e wi h enal synd ome (HFRS; i uses ha bou ed
by Old Wo ld a s and mice, as well as a icolines) o han a i us ca dio pulmona y synd ome (HCPS; i uses
ha bou ed by New Wo ld a s and mice) in humans. In e ms o human heal h, he mos impo an Eu opean
a i us is Puumala han a i us (PUUV) which causes neph opa hia epidemica (NE), a mild o m o HFRS4,5. The
p incipal hos o PUUV is he bank ole (Myodes gla eolus), which is dis ibu ed o e mos o Eu ope6 and is he
dominan oden species in bo eal o es s.
In gene al, he incidence o human han a i us in ec ions acks he abundance o he local oden hos pop-
ula ion7–12. In empe a e Eu ope, bank ole popula ions show seasonal luc ua ions and occasional, i egula
e up ions caused by mas ing, i.e., a hea y seed c op o oak and beech12–14. Mas ing has been linked o human
NE epidemics, likely b ough abou by g ow h o he local ole popula ion12,15,16. In bo eal o es s o no he n
Eu ope, deciduous ees do no unde go mas ing e en s and ole popula ion abundance cycles o 3 o 5 yea s a e
hough o be d i en by specialis p eda o s17–19. Ac oss Eu ope, he incidence o NE is highes in Finland, no h-
e n Sweden and no he n Russia5,20. Al hough bank ole popula ion dynamics ha e p o en use ul o p edic ing
NE epidemics in bo h empe a e and bo eal biomes9–12,21, de ailed analyses o long- e m PUUV ansmission in
ese oi popula ions a e lacking. Fu he mo e, he ela ionship be ween oden abundance and human in ec ion
a e is somewha obscu e; o example, a highe incidence o human NE was obse ed du ing inc easing a he
han peak phases in wo ole cycles in cen al Finland whe e PUUV is endemic9.
Ou objec i e was o ully desc ibe he pa e n o PUUV ansmission in cyclic popula ions o bo eal bank
oles. Mo e speci ically, we aimed o cha ac e ize he seasonal and mul iannual luc ua ions in 1) he abundance
o PUUV-in ec ed bank oles, which is p esumably he main d i e o NE epidemics in humans, 2) he p e alence
o PUUV in ec ion, and 3) he a e o PUUV ansmission (as indica ed by he acquisi ion a e o PUUV-speci ic
1Na u al Resou ces Ins i u e Finland, Van aa, Finland. 2Uni e si y o Helsinki, Depa men o Vi ology, Finland.
3Uni e si y o Jy askyla, Depa men o Biological and En i onmen al Science, Finland. 4Helsinki Uni e si y Hospi al,
Depa men o Vi ology, Finland. Co espondence and eques s o ma e ials should be add essed o L.V. (email: liina.
[email p o ec ed])
Recei ed: 25 June 2015
accep ed: 15 Janua y 2016
Published: 18 Feb ua y 2016
OPEN
www.na u e.com/scien i ic epo s/
2
Scien i ic RepoR s | 6:21323 | DOI: 10.1038/s ep21323
an ibodies), which we expec ed o shed mo e ligh on wi hin-hos ansmission o PUUV. In addi ion, we s udied
whe he he accumula ion o a e o PUUV in ec ions di e ed be ween annual coho s bo n in di e en cycle
phases. Fo hese pu poses, we moni o ed PUUV in ec ions in bank oles h ough nea ly h ee ole cycles using a
cap u e-ma k- ecap u e me hod. Fo he i s ime, PUUV in ec ion dynamics we e also s udied in de ail du ing
he win e when he incidence o human NE is highes in he bo eal zone.
Me hods
E hical s a emen . All handling p ocedu es o wild bank oles ollowed he Finnish Ac on he Use
o Animals o Expe imen al Pu poses (62/2006) and ook place wi h pe mission om he Finnish Animal
Expe imen Boa d (license numbe s HY 45-02, HY 122-03, and HY 54-05). All e o s we e made o minimize
animal su e ing. The species s udied is no p o ec ed in Finland o included in he Red Lis o Finnish Species.
The animal apping ook place wi h pe mission om he landowne s.
Roden apping and sampling. The s udy ook place a Konne esi in cen al Finland (62°34′ N, 26°24′ E),
whe e PUUV is highly endemic in bank oles9,22–24. In his egion, he g ound is co e ed by snow o an a e age
o 160 days om la e No embe o la e Ap il25. PUUV ansmission dynamics we e s udied in de ail using a
cap u e-ma k- ecap u e (CMR) me hod on a la ge “co e g id”. Addi ionally, appings we e pe o med on 14
smalle “sa elli e g ids” o ex end he geog aphic scale o he da ase . The co e g id included a young s and o bi ch
and willow on d ained soil su ounded by ma u e sp uce-domina ed coni e ous o es wi h he unde s o ey eg-
e a ion domina ed by mosses and dwa sh ubs (e.g., Vaccinium). Sa elli e g ids we e si ua ed in pine, sp uce, and
mixed o es s o di e en ages wi hin 5 km o he co e g id and a leas 750 m om i and each o he . T appings
ook place om Ap il 2002 o May 2009.
The co e g id consis ed o 246 apping s a ions a 15 m in e als, co e ing a o al a ea o 5.8 ha. One Ugglan
Special li e- ap (G ahnab, Sweden) was placed a each apping s a ion unde a shee me al chimney ha p o-
ided shel e o cap u ed animals bu enabled access o aps du ing he snowy pe iod. In all, 60 apping sessions
we e conduc ed on he co e g id. Du ing he snow- ee pe iod, he apping in e al was app oxima ely one
mon h and longe du ing win e s, so ha he in e als las ed om 17 o 166 days (median 37, in e qua ile ange
30–43 days). T aps we e bai ed wi h oa seeds and po a o and se in he e ening. A e se ing, aps we e checked
en imes a ca. 8-hou in e als. Du ing low empe a u es (< – 5 °C) in win e , aps we e checked 3 imes pe
day a 4-hou in e als and le open (non- apping) o e nigh . Newly- apped bank oles we e subcu aneously
agged wi h a ansponde (ID-100A Mic o ansponde , T o an L d, U.K.). On each apping session, oles we e
bled h ough he e o-o bi al sinus and hei body mass, age (based on pelage26), sex, and sexual ma u i y (pe o-
a e and/o lac a ing emales and males wi h sc o al es es ega ded as ma u e) we e eco ded. Animals ecap u ed
du ing a single apping session we e immedia ely eleased.
A o al o 21 apping sessions on he 14 sa elli e g ids we e conduc ed in he beginning (May), middle (July),
and a e he end (Oc obe ) o he bank ole b eeding season. Nine Ugglan Special li e aps bai ed wi h oa seeds
and po a o we e se in a g id o 3*3 aps wi h 15 m in e als o 3 nigh s and checked once pe day. Occupied aps
we e eplaced wi h eshly-bai ed ones and animals we e b ough in o he labo a o y, bled h ough e o-o bi al
sinus, and sac i iced. In Oc obe 2004 and July and Oc obe 2006, li e aps we e eplaced wi h s anda d snap
aps bai ed wi h ye b ead. The sex and b eeding s a us o all animals snap- apped in he sa elli e g ids we e
de e mined be o e he hea was dissec ed and placed in o 200 μ L phospha e-bu e ed saline and s o ed a –20 °C
p io o he analysis o PUUV an ibodies.
De e mina ion o PUUV in ec ion s a us o each animal in each apping session ( ). All blood
(dilu ed 1:10 in phospha e-bu e ed saline) and hea samples we e es ed o he p esence o PUUV-speci ic an i-
bodies using an immuno luo escen an ibody es (IFAT) desc ibed elsewhe e27. Based on IFAT esul s, an in ec-
ion s a us was de e mined o each animal o each apping session ( ). The classi ica ions o in ec ion s a us
a e shown in Table1 (he ea e , uppe case le e s indica e in ec ion s a us deduced om he animal’s se ological
his o y, and lowe case le e s indica e ha he s a us was assigned acco ding o o he c i e ia).
As PUUV causes a ch onic in ec ion in he bank ole28, a PUUV-se oposi i e (P) esul was in e p e ed as an
in ec ed (I) animal. Howe e , young animals cap u ed be ween May and Oc obe may be se oposi i e as a esul
o ma e nal an ibodies (Ma Ab) ecei ed om an in ec ed emale which p o ide empo a y immuni y o up o
80 days om bi h29. The e o e, ini ially se oposi i e young animals ha we e la e cap u ed and de e mined o
be se onega i e we e conside ed as being Ma Ab + (M) when i s cap u ed. De e mining he in ec ion s a us o
young animals ha we e caugh in mo e han one apping session and which always es ed se oposi i e was mo e
complex. Fo hose se oposi i e animals cap u ed in he co e and sa elli e g ids, he numbe genuinely in ec ed
(i) was es ima ed o each apping session by summing he assigned indi idual p obabili ies (es ima e based on
body mass) o being genuinely in ec ed. These p obabili ies we e calcula ed om a s a is ical model (i.e., Ma Ab
model, Supplemen a y Fig. S1; see below) based on 588 cap u es wi h a known se ological his o y in he co e a ea;
i.e., 1) animals ha had acqui ed o /and los an ibodies be ween he wo apping sessions, and 2) se oposi i e
animals ha we e old enough (> 8 mon hs) o exclude he possibili y o a posi i e esul due o Ma Ab. In he
Ma Ab model, in ec ion s a us was examined in ela ion o body mass using gene alized addi i e models (GAM)
wi h a binomial dis ibu ion (0 = Ma Ab +, 1 = genuinely in ec ed) and logi link unc ion (gam unc ion o
gamm4 lib a y30 in he R so wa e package31). Because he g ow h a e o young animals a ies o e he b eeding
season depending on i hey ma u e immedia ely o delay ep oduc ion, sepa a e models we e un o each mon h
o cap u e (Supplemen a y Fig. S1). I an animal was assigned a p obabili y >0.9 o being genuinely in ec ed, i
was conside ed as genuinely in ec ed (p obabili y o being in ec ed = 1) in la e apping sessions. Fo example,
i a summe -bo n bank ole was i s cap u ed as se oposi i e in Augus weighing 19.4 g ams, se oposi i e in
Sep embe weighing 18.4 g ams, and se oposi i e in Oc obe weighing 16.7 g ams, i was alloca ed espec i e
www.na u e.com/scien i ic epo s/
3
Scien i ic RepoR s | 6:21323 | DOI: 10.1038/s ep21323
p obabili ies 0.924, 0.908, and 0.963 o being genuinely in ec ed by he GAM models. Howe e , as he ini ial
p obabili y (0.924) exceeded 0.9, he animal was assigned a p obabili y o 1 o being in ec ed, and in ec ion s a us
II (in ec ed in – 1 and ), o Sep embe and Oc obe .
Se onega i e animals we e conside ed as non-in ec ed and he e o e suscep ible (S). In some a e cases, an
adul indi idual was se onega i e despi e being p e iously conside ed genuinely in ec ed. These an ibody esul s
we e in e p e ed as alse nega i es. When an animal was no cap u ed du ing a pa icula apping session ( ),
bu cap u ed wi h an unchanged PUUV an ibody s a us be ween he p eceding ( – 1) and he ollowing ( + 1)
apping sessions, ha same an ibody s a us was also de e mined o apping session .
De e mina ion o PUUV in ec ion s a us in he p e ious apping session ( –1) o each ani-
mal cap u ed in apping session . To calcula e he se ocon e sion a e ( he numbe o indi iduals ha
acqui ed PUUV an ibodies be ween apping sessions – 1 and ou o hose suscep ible a – 1) in he co e g id,
he in ec ion s a us in he p e ious apping session ( – 1) o each animal cap u ed in apping session needed
o be de e mined. The possible p eceding ( – 1) s a us we e: I (In ec ed), S (Suscep ible), M (Ma Ab+ ), and 0
(suckling pups and hus no a pa o he popula ion). Fo animals ha we e cap u ed and es ed o an ibodies
a bo h – 1 and , de e mina ion was s aigh o wa d. Howe e , o indi iduals ha we e i s ime cap u ed a ,
p obabili ies o belonging o he suscep ible g oup a he p e ious apping session we e assigned (see de ails in
Table1).
Fi s ly, young animals likely o ha e been suckling pups and he e o e no pa o he suscep ible popula ion
a –1 we e excluded om he analysis. The exclusion c i e ion was a body mass a ha was lowe han he body
mass o he ligh es animal cap u ed o he second ime in he same mon h o any yea .
Secondly, young indi iduals ha we e se onega i e (?S) a hei i s cap u e ( ) we e di ided in o se onega i e
and he e o e suscep ible (sS) a – 1, and se oposi i e due o ma e nal an ibodies and he e o e non-suscep ible
(mS) a – 1: hei assigned numbe s we e calcula ed using he p opo ion suscep ible a –1 (SS) among he
se onega i e indi iduals a (SS + MS), so ha (sS) = (?S)*[SS/(MS + SS)], and (mS) = (?S) – (sS). In nine ea ly
summe apping sessions, none o he young summe -bo n animals (N = 38) had a known se ological his o y
and he e o e his calcula ion me hod could no be applied. In hese cases, he p opo ion o indi iduals ca ying
Ma Ab a – 1 [mS/(sS + mS)] was de e mined o be he same as he in ec ion p e alence among o e -win e ed
PUUV an ibody/in ec ion s a us Numbe
Se onega i e; suscep ible a 1871
SS: nega i e, cap u ed as nega i e a –11027
MS: nega i e, cap u ed as posi i e a –178
0S: assigned by weigh as no in popula ion a –197
?S: nega i e, no cap u ed bu assigned by weigh as in popula ion a –1669
sS: assigned as suscep ible a –1555.0
mS: assigned as immune due o ma e nal an ibodies a –1114.0
Se oposi i e (P): in ec ed (I/i) o p o ec ed by ma e nal an ibodies (M/m) a 1415
II: posi i e, cap u ed as posi i e a –1 828
SI: posi i e, cap u ed as nega i e a –1 223
?P: posi i e, no cap u ed a –1364
0M: posi i e yea -bo n, la e cap u ed as nega i e; assigned no in popula ion a –147
mM: posi i e yea -bo n, la e cap u ed as nega i e; assigned as immune a –1 31
0m: posi i e yea -bo n, no e-cap u ed as nega i e; assigned no in popula ion a –152
mm: posi i e yea -bo n, no e-cap u ed as nega i e; assigned as immune by weigh 49.6
?i: posi i e adul , o a yea -bo n assigned as in ec ed by weigh 184.4
ii: assigned as in ec ed and se ocon e ed be o e –1118.2
si: assigned as in ec ed and se ocon e ed be ween –1 and 59.1
xi: no da a o assign PUUV s a us o –1 7
To al PUUV in ec ion s a us
Suscep ible (SS + MS + 0 S + ?S) 1871.0
In ec ed (II + SI + ?i) 1235.4
Ma e nal an ibody posi i e (0M + mM + 0m + mm) 179.6
To al N o cap u es 3286
Table 1. The known (uppe case) and assigned (lowe case) PUUV s a us o all bank ole cap u es on he
co e g id. P = s e oposi i e, S/s = suscep ible, M/m = ma e nal an ibody posi i e, I/i = in ec ed, 0 = no in
popula ion, ? = unknown, x = unde e mined. Capi al le e s indica e s a us deduced om se ological his o y,
small le e s he assigned s a us. In PUUV in ec ion/an ibody s a us, he i s cha ac e indica es PUUV s a us
a −1 (p e ious apping session), and second cha ac e he PUUV s a us a (cu en apping session).
Se ocon e sion a es o o al popula ion and yea ly coho s pe apping session we e calcula ed om (SI + si)/
(SS + SI + ss + si), ma ked in bold.
www.na u e.com/scien i ic epo s/
4
Scien i ic RepoR s | 6:21323 | DOI: 10.1038/s ep21323
emales in he p eceding apping session ( – 1); acco ding o ea lie esul s, PUUV in ec ion has no e ec on
b eeding success in ea ly summe among o e -win e ed indi iduals29,32.
Thi dly, o indi iduals ha we e i s cap u ed as se oposi i e, and om which he Ma Ab and 0 indi iduals
a – 1 had been ex ac ed, i.e., (?i) = (?P– 0M– mM– 0m– mm), he numbe ha se ocon e ed be ween – 1 and
(si) was es ima ed using he p opo ion ha se ocon e ed be ween – 1 and (SI) among he known posi-
i e indi iduals (SI + II) so ha (si) = (?i)*[SI/(SI + II)]. Especially du ing he b eeding season, animals bo n
in di e en yea s (i.e., old o e -win e ed and young summe -bo n oles) coexis ed on he s udy g id. Gi en
ha di e en -age animals likely all in o he ou in ec ion classes (0, M, S, I) on di e gen p opo ions, all he
abo e-men ioned assignmen s we e made pe yea coho , i.e., animals bo n in he same summe .
S a is ical analyses. Da a se s. The popula ion-le el dynamics o (a) he abundance o in ec ed ani-
mals, (b) he p e alence o in ec ion, and (c) he pe capi a se ocon e sion a e we e examined using wo
popula ion-le el da ase s (da ase s “P e alence o PUUV pe apping session” and “Se ocon e sion a e pe ap-
ping session”, a ailable om he D yad Digi al Reposi o y: h p://dx.doi.o g/10.5061/d yad.g8140 33). In addi ion
o popula ion-le el dynamics, we examined he e ec s o age and cycle phase on he accumula ion o in ec ions
and he se ocon e sion a e in yea ly coho s o bank oles using wo coho -le el da ase s (da ase s “PUUV p e -
alence pe yea coho ” and “Se ocon e sion a e pe yea coho ”, a ailable om he D yad Digi al Reposi o y:
h p://dx.doi.o g/10.5061/d yad.g8140 33).
The seasonal and mul iannual a ia ion in he abundance o in ec ed indi iduals was s udied using a da ase
(“P e alence o PUUV pe apping session”33) o apping indices (indi iduals/100 ap nigh s) o PUUV-in ec ed
bank oles om all apping sessions conduc ed on he co e g id (60 apping sessions, 246 aps) and he 14 sa -
elli e g ids (21 apping sessions, 126 aps). The same da ase was used o s udy he empo al a ia ion in PUUV
in ec ion p e alence. Fo he sa elli e g ids, in ec ion p e alence was calcula ed as he numbe o indi iduals
known o be (o assigned as) in ec ed pe he o al numbe es ed, and o he co e g id as he minimum numbe
o in ec ed pe he minimum popula ion size (i.e., including animals ha we e no cap u ed a bu a leas once
be o e and once a e ).
The da ase used o s udy empo al a ia ion in se ocon e sion a e (“Se ocon e sion a e pe ap-
ping session”33) consis ed o he pe capi a se ocon e sion a e calcula ed o each apping session on
he co e g id by di iding he o al numbe o se ocon e ed by he o al days o exposu e, i.e., (SI + si)/
[(SS + SI + ss + si)* in e al], whe e (SI + si) was he numbe o animals known o be o sco ed as se o-
con e ed be ween – 1 and , (SS + SI + ss + si) was he numbe o animals known o be o sco ed as
suscep ible on – 1, and “in e al” was he numbe o days be ween – 1 and . Indi iduals wi h Ma Ab
we e excluded om he suscep ible popula ion. Acco ding o popula ion abundances obse ed he e
and in an ea lie s udy9, he apping sessions we e di ided in o wo ca ego ical a iables: h ee phases
o he densi y cycle ( a iable “cycle phase”: “inc ease”, “peak”, and “low” phases; onse o phase on
1s June) and h ee ole densi y cycles ( a iable “cycle ID”: 2001− 2003, 2004− 2006, and 2007− 2009; onse o
cycle on June 1s o he inc ease phase). June was selec ed as he onse o a biological yea , as he i s summe -bo n
indi iduals usually en e ed he popula ion in ha mon h. Also a dicho omous a iable “b eeding season” was
included in he da a se , indica ing whe he he ime in e al be ween – 1 and p edominan ly belonged o he
yea ly b eeding season o no .
In addi ion o s udying in ec ion p e alence and se ocon e sion a e in he whole popula ion, we examined
he e ec s o age and cycle phase on he accumula ion o in ec ions and he se ocon e sion a e in yea ly coho s
o bank oles. Thus, in ec ion p e alences (“PUUV p e alence pe yea coho ”33) and se ocon e sion a es
(“Se ocon e sion a e pe yea coho ”33) we e calcula ed o each bi h yea coho and o each apping session
hey we e p esen on he co e g id. Age was de e mined as he numbe o mon hs a e 1s May o he yea o
bi h, i.e., he ea lies mon h when indi iduals o he yea -bo n coho could be expec ed o en e he popula ion
(al hough hey we e no obse ed in ou apping da a). Bi h yea s we e di ided in o “inc ease”, “peak”, and “low”
phases as abo e.
Candida e models. The popula ion-le el dynamics o (a) he abundance o in ec ed animals, (b) he p e alence
o in ec ion, and (c) he pe capi a se ocon e sion a e (da ase s “P e alence o PUUV pe apping session” and
“Se ocon e sion a e pe apping session”33) we e examined wi hin wo ime- ames: a biological yea and a
h ee-yea popula ion densi y cycle. In yea ly coho s o bank oles (da ase s “PUUV p e alence pe yea coho ”
and “Se ocon e sion a e pe yea coho ”33), he accumula ion o PUUV in ec ions and he empo al a ia ion
in se ocon e sion a e we e s udied in a ime- ame o 16 mon hs, he assumed maximum li e span o mos bank
oles34. Gene alized addi i e models (GAM, gamm4 package30 o R so wa e31) we e applied in all analyses ha
included con inuous empo al a iables (i.e. “cycle mon h”, “mon h”, and “coho age”). GAMs a e nonpa ame ic
eg ession models o en used o cha ac e izing associa ions whe e he shape o he ela ionship be ween he
p edic o s and he esponse is no known a p io i – o ins ance, in empo al a ia ion35. Gene alized addi i e
mixed models (GAMMs, unc ion “gamm4”), i.e. GAMs including andom e ec s, we e used o popula ion-le el
analyses on he abundance o in ec ed indi iduals and in ec ion p e alence (da ase “P e alence o PUUV pe
apping session”33), since he da ase included obse a ions om se e al si es (co e and 14 sa elli es). Models
wi hou andom e ec s ( unc ion “gam”) we e used o o he analyses. The abundance o in ec ed indi iduals was
analysed using a log link unc ion wi h Poisson e o dis ibu ions. The apping e o (numbe o ap nigh s)
was used as an o se a iable. The p e alence o in ec ion and se ocon e sion a e (p opo ional ou comes) we e
analysed using a logi link unc ion wi h binomial e o dis ibu ions, weigh ed by he o al numbe o indi idu-
als and exposu e ime (N suscep ible indi iduals*in e al be ween – 1 and ), espec i ely. O e dispe sion was
accoun ed o by including an obse a ion-le el andom e ec in mixed models, and a dispe sion pa ame e in
models wi hou andom e ec s.
www.na u e.com/scien i ic epo s/
5
Scien i ic RepoR s | 6:21323 | DOI: 10.1038/s ep21323
Fo he popula ion-le el dynamics (i.e., abundance o in ec ed animals, in ec ion p e alence and he pe capi a
se ocon e sion a e in he o al popula ion), a se o candida e models was examined o each ime- ame (see
Table2). The “wi hin-cycle” candida e model se included a ull model consis ing o ac o ial a iable “cycle
ID” and sepa a e smoo h e ms o mon hs elapsed (i.e. “cycle mon h”) om he cycle onse (1s o June) o each
“cycle ID”. The ull model o he “wi hin-yea ” model se included he ac o ial a iable “cycle phase” and sepa a e
smoo h e ms o ime elapsed om 1s June (“mon h”) o each “cycle phase” le el (Table2). All submodels nes ed
wi hin he ull models we e included in he candida e model se s.
Seasonal a ia ion in se ocon e sion a e was u he s udied using he dicho omous a iable “b eeding sea-
son” in da ase “Se ocon e sion a e pe apping session”33 as a empo al p edic o ins ead o a con inuous ime
a iable. The ull model o he candida e model se (“Va ia ion in ela ion o b eeding season”, Table2) was a gen-
e alized linea model wi h binomial e o dis ibu ions and a logi link unc ion, whe e he in e ac ion be ween
cycle phase and b eeding season was se as an explana o y a iable.
In yea ly coho s o bank oles, he accumula ion o PUUV in ec ions and he empo al a ia ion in se ocon-
e sion a e we e s udied wi h a candida e model se ha included a ull model consis ing o he ac o ial a iable
“Bi h yea cycle phase” and sepa a e smoo h e ms o coho age ( om 1s May) o each cycle phase, and all
submodels nes ed wi hin he ull model (model se “Va ia ion in ela ion o coho age and bi h yea cycle phase”,
Table2).
All candida e model se s we e anked acco ding o Akaike in o ma ion c i e ia adjus ed o sample size
(AICc36, in MuMIn package37 o R so wa e31). In model se s whe e dispe sion pa ame e s we e used o accoun
o o e dispe sion, quasi-AICc (QAICc) was used o model anking38. In each candida e model se , he mos
pa simonious model wi hin 2 AICc/QAICc uni s om he lowes AICc/QAICc sco e was conside ed he bes
suppo ed by he da a and subsequen ly used o s a is ical in e ence. Fo mixed model se s, an op imal andom
e ec s s uc u e was chosen o each ull model in a simila ashion p io o he selec ion o ixed e ec s, so ha
models wi h andom a iance a ibu able o a) apping si e b) apping si e o each ac o le el (“cycle ID” o
“cycle phase”), and c) obse a ion ( o accoun o o e dispe sion), and combina ions a + b and a + c we e anked
acco ding o AICc sco es. In all analyses, P alues below 0.05 we e conside ed s a is ically signi ican .
Resul s
Du ing he 7-yea s udy, bank ole abundance exhibi ed a dis inc h ee-yea cycle in which he popula ion
inc eased, peaked and hen c ashed (Fig.1a). Du ing he inc ease and peak phases, he highes abundances we e
Candida e models
Dependen a iable
Abundance o
in ec ed
In ec ion
p e alence
Se ocon e sion
a e
Wi hin-cycle dynamics
cycle ID + s(cycle mon h by cycle ID) 626.4 543.3 169.1
cycle ID + s(c ycle mon h) 646.9 533.2 186.0
s(cycle mon h) 762.9 530.1 184.3
cycle ID 643.9 953.7 193.9
in e cep only 761.0 951.2 196.5
Wi hin-yea dynamics
cycle phase + s(mon h by cycle phase) 578.1 507.4 134.6
cycle phase + s(mon h) 600.1 490.5 134.1
s(mon h) 610.3 497.2 136.5
cycle phase 1374.8 549.0 132.1
in e cep only 1381.2 550.9 135.5
Va ia ion in ela ion o b eeding season (dicho omous)
cycle phase * b eeding season 130.4
cycle phase + b eeding season 127.9
b eeding season 130.4
cycle phase 130.8
in e cep only 134.1
Va ia ion in ela ion o coho age and bi h yea cycle phase
cycle phase + s(coho age by cycle phase) 183.3 156.1
cycle phase + s(coho age) 182.2 158.3
s(coho age) 205.3 160.9
cycle phase 519.4 158.7
in e cep only 642.2 161.9
Table 2. AICc (Akaike in o ma ion c i e ion adjus ed o sample size) sco es o all candida e gene alized
addi i e models (GAMs). s(…) deno e GAM smoo h e ms. The AICc sco es o he bes -suppo ed models a e
w i en in bold.

www.na u e.com/scien i ic epo s/
6
Scien i ic RepoR s | 6:21323 | DOI: 10.1038/s ep21323
eached a he end o he b eeding season (Augus -Oc obe ) while abundance emained low du ing he same
pe iod in low phases.
A o al o 1 108 bank oles we e cap u ed on he co e g id, o which 615 (56%) we e cap u ed in a leas wo
apping sessions. A i s cap u e, 337 indi iduals we e PUUV se oposi i e, 766 we e se onega i e, and he an i-
body s a us could no be de e mined o i e animals. The median numbe o cap u es pe animal du ing sepa a e
apping sessions was 2 (in e qua ile ange 1− 4), and he maximum numbe o apping sessions he same indi-
idual was cap u ed was 14. In o al, 3 286 cap u es (i.e., indi iduals known o be ali e a a gi en apping session
bu o which an in ec ion s a us could no be de e mined) we e ob ained: 1 415 (43%) es ing PUUV se oposi i e
and 1 871 (57%) se onega i e (Table1). A o al o 1 152 bank oles we e caugh on he sa elli e g ids, o which 324
(28%) we e PUUV se oposi i e, 790 (69%) se onega i e and 38 (3%) could no be es ed.
O he 766 ini ially-se onega i e oles caugh on he co e g id, 499 we e ecap u ed a leas once du ing ollow-
ing apping sessions, yielding a o al o 1 240 ecap u es whe e an animal’s in ec ion s a us was SS o SI (Table1).
Among hese ecap u es o suscep ible animals, 223 se ocon e sions (18%) we e eco ded, o which 35% ook
place be ween he i s and second cap u e, 31% be ween he second and hi d cap u e, and he emaining 34%
be ween he hi d and ou een h cap u e. O he 337 ini ially-se oposi i e animals, 236 we e i s cap u ed in
hei bi h yea be ween May and Oc obe and po en ially, due o hei young age, ca ied Ma Ab. O hese 236,
126 we e ecap u ed a leas once and 78 (62%) became se onega i e and we e hus conside ed o ha e been
ca ying Ma Ab. Fo he 200 cap u es o he emaining 158 ini ially-se oposi i e i s -yea animals, he assigned
p obabili ies (calcula ed o each cap u e om he Ma Ab model based on body mass, see Supplemen a y Fig. S1)
o being in ec ed summed o 98.4, meaning ha a o al o 200 − 98.4 = 101.6 cap u es (51%, 0m + mm in Table1)
we e conside ed Ma Ab posi i e. Ma Ab+ animals we e mo e common in peak phase coho s han hose bo n
in o he cycle phases and accoun ed o app oxima ely 25–35% o se oposi i e animals in July o peak phases
(Fig.1b). The numbe o suscep ible, in ec ed, and Ma Ab+ animals on he co e g id we e simila o pooled num-
be s o animals caugh on he sa elli e g ids (Fig.1b).
Tempo al pa e ns in PUUV in ec ion. Abundance o in ec ed oles. On he imescale o a h ee-yea
densi y cycle, he dynamics in he abundance o PUUV-in ec ed bank oles on he co e and sa elli e g ids we e
bes explained by a GAMM ha included he ac o ial a iable “cycle ID” and sepa a e smoo h e ms o ime
elapsed om he cycle onse o each cycle (“cycle mon h”, Tables2 and 3). Acco ding o he model, in ec ion pa -
e ns du ing densi y cycles 2004–2006 and 2007–2009 we e simila , when he abundance o in ec ed oles eached
high peaks du ing bo h inc ease- and peak-phase win e s (Fig.2a, Table3). On he co e g id du ing inc ease
phases, he peak abundance o in ec ed indi iduals was 8.3 (95% con idence in e al [CI]: 6.1–10.6) and 9.2 (95%
CI: 5.0– 13.5) in ec ed indi iduals/100 ap nigh s du ing he 2004–2006 and 2007–2009 cycles, espec i ely, and
occu ed in Feb ua y. In peak phases, 7.8 (95% CI: 6.2–9.5) and 10.0 (95% CI: 5.5–14.5) in ec ed indi iduals/100
ap nigh s we e a ained in la e Oc obe and ea ly Decembe in 2004–2006 and 2007–2009 cycles, espec i ely.
Du ing he inc ease phases o hese cycles, he peak abundance o in ec ed oles ollowed he peak in o al abun-
dance wi h a lag o app oxima ely 3.5 o 4.5 mon hs, whe eas du ing peak phases, he lag was only 2 mon hs
Figu e 1. The numbe o bank oles on he co e (lines) and sa elli e g ids (c osses), (a) in o al and (b)
acco ding o PUUV in ec ion s a us. Values o he 14 sa elli e g ids a e pooled. Shaded a eas indica e phases
(Jun 1s o May 31s ) o inc easing (ligh g ey), peak (da k g ey) and low (whi e) ole densi y. In a, e ical
lines indica e apping sessions on he co e g id. In b, colou s indica e in ec ion s a us: suscep ible o PUUV
in ec ion (blue), in ec ed wi h PUUV ( ed), and Ma Ab+ (g een).
www.na u e.com/scien i ic epo s/
7
Scien i ic RepoR s | 6:21323 | DOI: 10.1038/s ep21323
(Figs1a and 2a). The 2001–2003 cycle (da a om Ap il 2002 o June 2003) di e ed om he o he wo cycles in
ha he abundance o in ec ed oles emained much lowe and peaked du ing he summe o he peak phase wi h
3.1 (95% CI: 1.6–4.7) indi iduals/100 ap nigh s, hen declined owa ds win e (Fig.2a).
Wi hin a ime ame o one yea , he abundance o in ec ed animals was bes explained by a model ha
included he ac o ial a iable “cycle phase” and sepa a e smoo h e ms o empo al a ia ion (“mon h”) in each
cycle phase (Tables2 and 3). Du ing inc ease phases, a seasonal pa e n could be seen wi h he abundance o
in ec ed animals inc easing om 0.3 (95% CI: 0.1–0.8) in June o 4.3 (95% CI: 2.1–6.5) in ea ly Feb ua y, and
he ea e declining o 3.1 (95% CI: 1.7–4.5) indi iduals/100 ap nigh s by he end o May (Fig.2b). Du ing peak
phases, he model sugges ed a a he s able le el o in ec ion om June o la e No embe (4.6–5.0 in ec ed indi-
iduals/100 ap nigh s, 95% CI: 2.1–7.2) he ea e declining o 1.1 indi iduals (95% CI: 0.4–1.7) by he end o
Dependen a iable/sou ce o a ia ion Wi hin-cycle dynamics Wi hin-yea dynamics
Abundance o In ec ed
Pa ame ic coe icien s Es ima e (SE) zP alue Es ima e (SE) zP alue
in e cep − 4.15 (0.24) − 17.1 < 0.001 − 3.85 (0.15) − 24.9 < 0.001
cycle ID 2004–2006 − 0.02 (0.25) − 0.1 0.95
2007–2009 − 0.06 (0.71) − 0.1 0.938
cycle phase peak 0.40 (0.17) 2.4 0.016
low − 2.67 (0.36) − 7.5 < 0.001
Smoo h e ms Ed ( d ) χ2P alue Ed ( d ) χ2P alue
cycle mon h by cycle ID 2001–2003 1 (1) 55 < 0.001
2004–2006 7.08 (7.08) 190 < 0.001
2007–2009 5.70 (5.70) 124 < 0.001
mon h by cycle phase inc ease 2.08 (2.08) 13.6 0.001
peak 2.08 (2.08) 22.7 < 0.001
low 1.95 (1.95) 4.4 0.105
Random e ec s σ2sd σ2sd
si e wi hin cycle 2001–2003 0.64 0.80
wi hin cycle 2004–2006 0.31 0.56
wi hin cycle 2007–2009 1.34 1.16
obse a ion 0.70 0.84
si e 0.09 0.31
In ec ion p e alence
Pa ame ic coe icien s Es ima e (SE) zP alue Es ima e (SE) zP alue
in e cep − 1.11 (0.15) − 7.3 < 0.001 − 0.85 (0.22) − 3.8 < 0.001
cycle phase peak 0.19 (0.30) 0.6 0.517
low − 0.77 (0.36) − 2.1 0.032
Smoo h e ms Ed ( d ) χ2P alue
cycle mon h/mon h 6.82 (6.82) 41.7 < 0.001 3.07 (3.07) 31.1 < 0.001
Random e ec s σ2sd σ2sd
si e wi hin cycle 2001–2003 0.75 0.87
wi hin cycle 2004–2006 0.98 0.99
wi hin cycle 2007–2009 1.42 1.19
obse a ion 0.43 0.65
Se ocon e sion a e
Pa ame ic coe icien s Es ima e (SE) zP alue Es ima e (SE) zP alue
in e cep − 6.06 (1.28) − 4.7 < 0.001 − 5.31 (0.14) − 37.4 < 0.001
cycle ID 2004–2006 0.20 (1.30) 0.2 0.88
2007–2009 1.27 (1.85) 0.7 0.497
cycle phase peak − 0.27 (0.19) − 1.4 0.161
low − 2.19 (1.41) − 1.5 0.128
Smoo h e ms Ed ( d ) χ2P alue
cycle mon h by cycle ID 2001–2003 1.43 (1.76) 3.8 0.036
2004–2006 2.12 (2.71) 2 0.129
2007–2009 4.85 (5.53) 3.5 0.007
Table 3. Pa ame e es ima es o he bes -suppo ed models analysing he wi hin-cycle and wi hin-yea
dynamics in he abundance o PUUV in ec ed bank oles, in ec ion p e alence and se ocon e sion a e.
Ed = Es ima ed deg ees o eedom; d = esidual deg ees o eedom; σ2 = he a iance a ibu able o andom
e ec . sd = s anda d de ia ion o σ2.
www.na u e.com/scien i ic epo s/
8
Scien i ic RepoR s | 6:21323 | DOI: 10.1038/s ep21323
May. Du ing low phases, he p edic ed abundance o in ec ed animals was nea ly cons an and emained below
one indi idual/100 ap nigh s (Fig.2b).
In ec ion p e alence. The p e alence o PUUV in ec ion on he co e and sa elli e g ids showed a clea seasonal
luc ua ion bo h wi hin cycles (Fig.2c) and wi hin yea s (Fig.2d). Models wi h dis inc empo al pa e ns o
sepa a e cycles o sepa a e cycle phases we e no suppo ed (Table2). The bes suppo ed wi hin-cycle model
only included he empo al componen , i.e., a smoo h e m o he ime elapsed om cycle onse , and he bes
wi hin-yea model included he seasonal componen , i.e., a smoo h e m o ime elapsed since he onse o he
biological yea and sepa a e in e cep s o cycle phases (Table3). Acco ding o he wi hin-cycle model, he PUUV
p e alence peaked in mid-Ma ch o he inc ease phase (p e alence 69.3%, 95% CI: 49.9–88.7%), and a second,
bu lowe peak ollowed in mid-Ma ch o he peak phase (p e alence 52.7%, 95% CI: 30.7–74.7%). Wi hin a cycle,
he lowes in ec ion p e alences occu ed in ea ly Sep embe du ing bo h inc easing and peak phases (PUUV
Figu e 2. Va ia ion in he (a,b) abundance o PUUV-in ec ed bank oles/100 ap nigh s, (c,d) PUUV in ec ion
p e alence, and (e, ) se ocon e sion a e wi hin he ime ame o (a,c,e) a ole densi y cycle and (b,d, ) a yea .
Lines ep esen p edic ed alues o he co e g id om he bes -suppo ed models ((a) cycle ID + s[c ycle mon h
by cycle ID]; (b) cycle phase + s[mon h by cycle phase]; (c) s[cycle mon h]; (d) cycle phase + s[mon h]; (e) cycle
ID + s[cycle mon h by cycle ID]; ( ) cycle phase; s[…] deno e GAM smoo h e ms). Shaded a eas indica e he
95% con idence in e als o pa ame e es ima es. “C” and “S” in a− e deno e obse ed da a in co e and pooled
sa elli e g ids, espec i ely. Numbe s 1− 8 in deno e biological yea s 2001− 2008. The cha ac e sizes indica e
he numbe o animals (c,d) and o al days o exposu e (e, ). G een, ed, and blue colou s indica e di e en
cycles in (a,c,e) and di e en cycle phases in (b,d, ).
www.na u e.com/scien i ic epo s/
9
Scien i ic RepoR s | 6:21323 | DOI: 10.1038/s ep21323
p e alences 12.4% [95% CI: 4.1–20.7%] and 17.8% [95% CI: 9.1–26.5%], espec i ely), and in ea ly Decembe
du ing low phase (p e alence 4.2%, 95% CI: 0–10.2%; Fig.2c). The wi hin-yea model p edic ed a yea ly p e -
alence peak in mid-Ma ch, when 59% (95% CI: 39–83%), 64% (95% CI: 44–86%), and 40% (95% CI: 20–65%)
o bank oles we e in ec ed du ing inc ease, peak and low phases o he cycle, espec i ely (Fig.2d). The yea ly
bo om p e alence was p edic ed o occu in mid-Sep embe , when 15% (95% CI: 6–23%), 17% (95% CI: 8–26%),
and 7% (95% CI: 2–13%) o animals we e in ec ed du ing inc ease, peak and low phases, espec i ely. In he
wi hin-yea model, he e was no signi ican di e ence be ween inc ease and peak phase p e alences, bu du ing
low phases he p e alence was signi ican ly lowe han du ing inc ease phases (Table3).
Pe capi a se ocon e sion a e. F om he wi hin-cycle model se examining he pe capi a se ocon e sion a e,
he bes -suppo ed model included sepa a e smoo h e ms o ime o di e en cycles (Tables2 and 3), and he
pa e ns o se ocon e sion a e showed no simila i y be ween cycles (Fig.2e). Acco ding o model p edic ions
o cycle 2007− 2009, he se ocon e sion a e inc eased om July o Janua y in inc ease phase 2007 and om
Augus o Oc obe in peak phase 2008. In he wi hin-yea model se , he bes -suppo ed model included sepa-
a e in e cep s o cycle phases and no wi hin-yea empo al a ia ion, al hough he e was no signi ican di e -
ence be ween cycle phases (Tables2 and 3, Fig.2 ). Al hough wi hin-yea models wi h empo al componen s
we e no well suppo ed (Table2), an inc easing end in se ocon e sion a e was obse ed om Augus o
Oc obe in inc ease phases 2004 and 2007, and peak phases 2005 and 2008 (Fig.2 ). This inc ease was cap u ed in
wi hin-cycle model p edic ions o yea s 2007 and 2008 (cycle 2007− 2009 in Fig.2e). In he model se whe e he
con inuous ime a iables we e eplaced wi h he dicho omous a iable “b eeding season”, he model “b eeding
season + cycle phase” was bes suppo ed (Table2), so ha he se ocon e sion a e was signi ican ly highe ou -
side han du ing he b eeding season (coe icien o non-b eeding season = 0.39, SE = 0.18, z = 2.1, P = 0.036).
PUUV in ec ion p e alence and se ocon e sion a e in ela ion o coho age and cycle phase a
bi h. O he model se examining in ec ion p e alence in ela ion o age and cycle phase o he bi h yea , he
bes -suppo ed model included a common smoo h e m o coho age bu sepa a e in e cep s o each bi h yea
cycle phase (Tables2 and 4). The g adual inc ease o in ec ion p e alence wi h age was simila in inc ease- and
peak-phase coho s (Fig.3a): in bo h coho s, nea ly all su i ing indi iduals we e PUUV in ec ed by he end o
hei second summe (inc ease phases: 91.4%, 95% CI: 83.4–99.4%; peak phases: 89.1%, 95% CI: 78.8–99.4%).
In coho s bo n du ing low phases, in ec ion p e alence was signi ican ly lowe (Table4), eaching only 27.4%
(95% CI: 0.0–65.5%) by he end o hei second summe . O e all, he in ec ion p e alence inc eased s eadily wi h
coho age, al hough a pe iod o s eepe inc ease was obse ed be ween Sep embe and Janua y. The o al numbe
o animals eached i s peak be ween Augus and Oc obe in e e y yea coho (Fig.3a, dashed lines). Howe e ,
he peak abundances o inc ease-phase coho s (2004 and 2007: Fig.3a dashed ed lines) we e 40–50% lowe han
hose o peak-phase coho s (2005 and 2008: Fig.3a dashed blue lines).
Se ocon e sion a e be ween wo apping sessions in a yea coho was bes explained by a model ha
included sepa a e smoo h e ms o age o coho s bo n in di e en cycle phases (Tables2 and 4). The empo-
al pa e ns o p edic ed se ocon e sion a es we e dissimila (Fig.3b): se ocon e sion a e showed a s eady
inc ease om Sep embe o Decembe in inc ease-phase coho s, esul ing in a s eep inc ease in in ec ion p e -
alence (Fig.3a). F om Decembe o Ma ch, he a e declined and inc eased he ea e coinciding wi h he onse
o he b eeding season and con inuing in o summe , al hough wi h wide con idence in e als. In peak-phase
coho s, no empo al end was seen in se ocon e sion a e excep o a sligh , s eady decline (Fig.3b). Howe e ,
Dependen a iable/sou ce o a ia ion Pa ame e es ima es
In ec ion p e alence
Pa ame ic coe icien s Es ima e (SE) zP alue
in e cep − 0.16 (0.13) − 1.2 0.217
bi h yea cycle phase peak − 0.27 (0.15) − 1.8 0.08
low − 3.41 (0.93) − 3.7 < 0.001
Smoo h e ms Ed ( d ) χ2P alue
coho age 3.89 (4.83) 48.5 < 0.001
Se ocon e sion a e
Pa ame ic coe icien s Es ima e (SE) zP alue
in e cep − 5.53 (0.25) − 22.2 < 0.001
bi h yea cycle phase peak − 0.11 (0.30) − 0.4 0.716
low − 2.12 (1.70) − 1.2 0.219
Smoo h e ms Ed ( d ) χ2P alue
coho age by bi h yea cycle phase inc ease 3.62 (4.49) 1.8 0.125
peak 1.00 (1.00) 0.1 0.714
low 1.00 (1.00) 0.9 0.358
Table 4. Pa ame e es ima es o he bes -suppo ed models analysing he PUUV in ec ion p e alence and
se ocon e sion a e du ing he li espan o a yea ly bank ole coho . Ed = Es ima ed deg ees o eedom;
d = Residual deg ees o eedom.