ArticlePDF Available

Weighing the predictors: host traits and coinfecting species both explain variation in parasitism of Rock Ptarmigan

Article

Weighing the predictors: host traits and coinfecting species both explain variation in parasitism of Rock Ptarmigan

Abstract and Figures

Testing hypotheses in ecological and evolutionary parasitology can require testing whether host traits or coinfecting parasites explain variation in parasitism by focal species. However, when host traits and coinfecting parasites are considered separately, relations between either and parasitism by focal species can be spurious—a problem that is addressed when both are considered together. We assessed whether abundances of focal parasites related to host age/sex and coinfecting parasites for three endoparasites and nine ectoparasites of Icelandic Rock Ptarmigan (Lagopus muta) collected over 12 yr (2006–2017), and quantified the variation in focal parasitism explained by these predictors. Host traits and coinfecting parasites explained significant variation in abundance of all nine focal parasite species for which models converged, when those models were based on groups of parasites sharing tissue tropism and/or transmission pathways and included year as a random effect. We found a single spurious relation: a host age–sex interaction effect that was removed once concurrent parasitism was considered. When considering focal parasites within groups of coinfecting parasites, we found cases of positive, negative, and lacks of correlations. The amount of variation in focal parasite abundance explained by host traits versus coinfecting parasites depended on the focal parasite and its group. Overall variation explained was both related to the prevalence of the focal parasite, possibly due to underlying parasite aggregation, and similar to variation explained in other models in ecology and evolution. We conclude that host traits and coinfecting parasites often combine to determine infection by focal species. Future studies should also explore the mechanisms underlying parasite–parasite relations and their potential impacts on host demography for this and other study associations, and assess relative effects of host traits and coinfecting parasites on focal parasitism.
Content may be subject to copyright.
DISEASE ECOLOGY
Weighing the predictors: host traits and coinfecting species both
explain variation in parasitism of Rock Ptarmigan
ANDR
EMORRILL ,
1,
O. K. NIELSEN,
2
U. STENKEWITZ,
2,3,4
G. R. P
ALSD
OTTIR,
3
M. R. FORBES,
1
AND K. SK
IRNISSON
3
1
Department of Biology, Carleton University, 1125 Colonel By Drive, Ottawa, Ontario K1S 5B6 Canada
2
Icelandic Institute of Natural History, Urridaholtsstraeti 6-8, Gardabaer IS-212 Iceland
3
Institute for Experimental Pathology, Keldur, University of Iceland, Reykjavik IS-112 Iceland
Citation: Morrill, A.,
O. K. Nielsen, U. Stenkewitz, G. R. P
alsd
ottir, M. R. Forbes, and K. Sk
ırnisson. 2021. Weighing the
predictors: host traits and coinfecting species both explain variation in parasitism of Rock Ptarmigan. Ecosphere
12(8):e03709. 10.1002/ecs2.3709
Abstract. Testing hypotheses in ecological and evolutionary parasitology can require testing whether
host traits or coinfecting parasites explain variation in parasitism by focal species. However, when host
traits and coinfecting parasites are considered separately, relations between either and parasitism by focal
species can be spuriousa problem that is addressed when both are considered together. We assessed
whether abundances of focal parasites related to host age/sex and coinfecting parasites for three endopara-
sites and nine ectoparasites of Icelandic Rock Ptarmigan (Lagopus muta) collected over 12 yr (20062017),
and quantied the variation in focal parasitism explained by these predictors. Host traits and coinfecting
parasites explained signicant variation in abundance of all nine focal parasite species for which models
converged, when those models were based on groups of parasites sharing tissue tropism and/or transmis-
sion pathways and included year as a random effect. We found a single spurious relation: a host agesex
interaction effect that was removed once concurrent parasitism was considered. When considering focal
parasites within groups of coinfecting parasites, we found cases of positive, negative, and lacks of correla-
tions. The amount of variation in focal parasite abundance explained by host traits versus coinfecting para-
sites depended on the focal parasite and its group. Overall variation explained was both related to the
prevalence of the focal parasite, possibly due to underlying parasite aggregation, and similar to variation
explained in other models in ecology and evolution. We conclude that host traits and coinfecting parasites
often combine to determine infection by focal species. Future studies should also explore the mechanisms
underlying parasiteparasite relations and their potential impacts on host demography for this and other
study associations, and assess relative effects of host traits and coinfecting parasites on focal parasitism.
Key words: age-biased parasitism; coinfection; interactions; Lagopus muta; rock ptarmigan; sex-biased parasitism;
spurious relations.
Received 29 January 2021; accepted 12 April 2021. Corresponding Editor: Sarah M. Zohdy.
Copyright: ©2021 The Authors. This is an open access article under the terms of the Creative Commons Attribution
License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.
4
Present address: University of Icelands Research Centre at Snæfellsnes, Hafnarg
otu 3, Stykkish
olmi 340 Iceland.
E-mail: andre_morrill@carleton.ca
INTRODUCTION
Researchers are often interested in the inu-
ence of host traits on exposure or susceptibility
to parasites. Studies have shown that host sex
(Schalk and Forbes 1997, Paquette et al. 2020),
age (Krasnov et al. 2006, Fecchio et al. 2015),
genotype (Little and Ebert 1999), and host condi-
tion (Wu et al. 2019) are all apparent determi-
nants of the level of parasitism experienced by
vwww.esajournals.org 1August 2021 vVolume 12(8) vArticle e03709
the host. Researchers use such information to test
life history and sexual selection theory (Read
1988, Forbes 1993) and hypotheses about popula-
tion and coevolutionary dynamics of parasites
and their hosts (Carius et al. 2001, Stenkewitz
et al. 2016).
Other past and contemporary studies focus on
the inuence of coinfecting parasite species on
the prevalence (Carvalho-Pereira et al. 2019, Wu
et al. 2019), abundance or intensity (Carvalho-
Pereira et al. 2019), and distribution of focal para-
sites in hosts (Jackson et al. 2006). Those
researchers are interested in factors that structure
parasite infracommunities including processes
such as facilitation (Karvonen et al. 2019), compe-
tition (Mideo 2009, Goater et al. 2014), and even
predation (Sousa 1993). Typically, the studies
focusing on effects of host traits on parasitism by
focal species are conducted without testing for
effects of concurrent parasitism; however, recent
studies demonstrate how both coinfecting para-
sites and host traits are important in explaining
natural levels of infection and infestation (Telfer
et al. 2010, Carvalho-Pereira et al. 2019, Wu et al.
2019, Veitch et al. 2020). Research from this com-
bined perspective is still rare and generally does
not quantify the relative contribution of each of
host traits and coinfecting species to explaining
variation in focal parasitism (but see Veitch et al.
2020).
There are two main reasons why treating both
host traits and coinfecting parasites together is
important. First, host traits and coinfecting para-
sites might each individually account for signi-
cant (and similar or different amounts of)
variation in measures of parasitism by focal para-
sites. Second, patterns in parasitism ascribed to
host traits might better be explained by consider-
ation of coinfecting parasites or vice versa (the
problem of spurious relations). For example, a
parasite that shows male biases in parasitism
may appear correlated with another parasite also
showing male biases in parasitism because both
parasites tend to be present on males and absent
from females. Thus, inferences about one para-
site species facilitating the other may be in error.
Additionally, the extent to which samples
include both lightly parasitized females and
more heavily parasitized males will likely dictate
the likelihood of nding pairwise correlations
between the focal parasites. In other words,
spurious relations in this scenario may be a func-
tion of both abundance of parasites and the
extent to which both sexes are included in sam-
ples.
It is arguably very important to consider both
host traits and coinfecting parasites together not
only to achieve a fuller understanding of the con-
tribution of the host and parasites to emerging
patterns of parasitism, but also to avoid making
spurious claims about the inuence of either, or
at least weigh the available evidence in support
of either as a determinant of parasitism. This
approach is the focus of the present study.
In addressing our research questions, it is
important to note that, in nature, the majority of
parasite species are multi-host parasites and the
vast majority of host individuals (and popula-
tions) are coinfected by multiple parasite species,
although tests are often based on single-host,
single-parasite species associations (Rigaud et al.
2010). Expected relations between parasitism and
(traits of) a host species might be determined
more or less (or conversely not at all) by selection
imposed by other exploiting parasite species or
other host species being exploited. Herein, we
address the extent to which researchers can
determine the contributions of host traits and
parasitism by other species to explaining varia-
tion in focal parasitism of a single host species.
This also begs the question as to how much vari-
ation in parasitism remains unexplained.
The species associations we study are Icelandic
Rock Ptarmigan (Lagopus muta, hereafter ptarmi-
gan) known to be infected by 17 species of para-
site, of which this study focuses on 12 (three
endoparasitic species: two nematodes and a coc-
cidian; and nine ectoparasitic species: ve mites,
three lice, and one y; Stenkewitz 2017). Details
of the chosen parasite species including their tax-
onomy, life cycle and life history, and degree of
specialization are provided elsewhere, as are the
rationales for their choice (Skirnisson et al. 2012,
Stenkewitz et al. 2016). Briey, the chosen para-
site species were well represented parasitic fauna
that could be identied to species (i.e., they were
not exceedingly rare parasites, those for which
ptarmigan could be considered accidental hosts,
or parasites that were insufciently described
taxonomicallyfor those parasites, see Nielsen
et al. 2020). One recent study has shown that host
traits (age and/or sex and/or their interaction)
vwww.esajournals.org 2August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
accounted for signicant variation in parasitism
by 11 of the 12 species, while controlling for the
random effect of year of study (Nielsen et al.
2020).
In this study, we address the extent to which
coinfecting parasite species explain variation in
addition to, or in lieu of, variation in parasitism
explained by host traits. This approach requires
rst that we identify candidate parasite species
to include in expanded models that also include
host traits. Including all 11 other parasite species
as predictors of variation in parasitism by singu-
lar focal species is not particularly parsimonious
nor tractable.
We decided on three groups of parasite species
based on known aspects of the parasitesbiology
(namely, tissue tropism and transmission mode
or route). We were primarily interested in which
species might use similar habitats on the bird or
share pathways of infection. As such, our
approach implicitly focuses on more direct asso-
ciations between coinfecting parasites (e.g., com-
petition for shared host resources), rather than
more indirect interactions mediated by the hosts
immune system. Hypothetical indirect interac-
tions may be important drivers of focal parasite
abundances, even when those interactions occur
between parasite species that do not share simi-
lar (tropism/transmission) biology (Poulin 2001).
Our analyses will be sensitive to potential indi-
rect interactions between members within the
same parasite species groupings, though these
hypothetical interactions are less tractable than
more direct parasiteparasite interactions and
thus will feature less in our interpretations of
results. It is also important to note that more
direct positive associations between parasites
arising from shared intermediate hosts or shared
transmission modes are not necessarily pairwise
interactionsper se, as the presence/abundance
of one parasite is not the cause of increased abun-
dance in the other.
We ask the following questions: Do abun-
dances of non-focal species within the same
groups explain signicant variation in the abun-
dance of focal parasites? And does inclusion of
such species in expanded models change any
inferences made about the inuence of host traits
on a focal parasite speciesabundance (the prob-
lem of spurious relations; cf. Nielsen et al. 2020)?
We herein document that inclusion of non-focal
parasites often explains variation in abundance
of focal parasite species and that this variation is
typically in addition to, and not in lieu of, varia-
tion in parasite abundance explained by host
traits. We test the relative inuence of host traits
and abundance of other parasite species on
explaining variation in parasitism by focal spe-
cies and compare this with variation left unex-
plained.
MATERIALS AND METHODS
Study site, bird collection, and sampling of
parasites
Our data set included samples of ca. 100
ptarmigan for each of 12 yr (20062017). This
data set had annual measurements of abundance
(or their proxies) of 12 parasites, which were all
identied to species and sampled frequently
enough (>1% of individuals infected) such that
ptarmigan would not be considered accidental
hosts. An additional coccidian, Eimeria rjupa,was
omitted from tests because of its extremely high
aggregation in samples, which resulted in statis-
tical models including it failing to converge.
We provide a brief summary of the study site
and methods of bird collection and processing of
parasites (details in Skirnisson et al. 2012, Stenke-
witz 2017). 1140 ptarmigan were shot in autumns
of 20062017 at upland sites near Lake M
yvatn,
in northeast Iceland (65°370N, 17°000W), during
the rst week of October (Sk
ırnisson and Nielsen
2019). 721 juvenile birds (ca. three months old;
50.1% females) and 419 adult birds (one+years
old; 35.8% females) were collected. As stated in
Skirnisson et al. (2012), ectoparasites were
collected through a combination of ltered vacu-
uming and direct removal of parasites. Endopar-
asitic helminths were extracted from the host
small intestine and ceca, while coccidian oocysts
were quantied from fecal material following the
modied McMaster procedure (see Skirnisson
et al. 2012). No blood parasites have been
reported for ptarmigan in Iceland (Skirnisson
et al. 2012).
Grouping sampled parasites based on tissue
tropism and transmission biology
Combining all possible coinfecting parasites
as predictors in analyses of focal parasite abun-
dances was not practical: The generalized linear
vwww.esajournals.org 3August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
mixed-effects models (GLMMs; see Statistical
methods and modeling approaches) failed to
converge whenever all parasites were included
with host traits in explaining infection levels of
focal parasites. Therefore, we focused our analy-
ses on three groups of parasites (detailed below).
Parasites with greater likelihood of experiencing
direct associations (e.g., through direct competi-
tion for shared host resources, transmission via a
shared route) were grouped together, which
would also facilitate the forming of explanatory
hypotheses. We recognize that indirect interac-
tions, for example, those moderated by the
immune system, may drive or hide correlations
between parasites that are or are not in the same
group, and that those associations could be
missed or misattributed using our current
approach. However, we note that preliminary
tests of all 60 possible pairwise parasiteparasite
correlations for which models would converge
identied only two potentially signicant associ-
ations outside of our proposed within-group
tests (see Discussion). This highlights the rele-
vance of our groupings of parasites in capturing
probable associations between species.
For the rst group, we considered the two
nematode species together with the remaining
coccidian (the three endoparasitic species). It is
important to note that the nematode Tri-
chostrongylus tenuis is directly transmitted, as is
the coccidian Eimeria muta, whereas the nema-
tode Capillaria caudinata is indirectly transmitted
by ingestion of earthworms (Stenkewitz 2017).
Furthermore, T. tenuis is found in the ceca, C.
caudinata in the ceca and small intestine, and
the coccidian E.muta in the ceca, and small and
large intestine (Stenkewitz 2017; i.e., all three at
least potentially share some location of infection
[the ceca]).
The second group included four species of
ectoparasites: two species of skin mites (Metami-
crolichus islandicus and Myialges borealisthe lat-
ter speciestissue tropism also includes
plumage), a louse y(Ornithomya chloropus), and
a skin and plumage louse species (Amyrsidea
lagopi). The mite species are known or thought to
be vectored by the louse y; M. borealis are
known to attach to O. chloropus, while Metami-
crolichus species can be phoretically transported
by louse ies (Mironov et al. 2005, Mironov et al.
2010). O. chloropus is largely included in this
group due to its relationship with these skin
mitestransmission. In comparison, the louse A.
lagopi is directly transmitted, but it is included
initially in this group given that it shares tissue
tropism with the skin mites. However, because
A. lagopi shares direct transmission and some tis-
sue tropism (plumage) with parasites in the third
group, we also consider an additional set of anal-
yses where the groups are revised to exclude A.
lagopi from the second group and include it with
the third group of parasites.
The third group included the ve remaining
ectoparasitic species: two louse species and three
mite species whose habitat was quills, vanes,
down, or plumage more generally as tissue gen-
eralists (lice: Goniodes lagopi and Lagopoecus af-
nis; mites: Mironovia lagopus,Tetraolichus lagopi,
and Strelkoviacarus holoaspis). All ve species are
directly transmitted.
Statistical methods and modeling approaches
We approached the statistical modeling of the
same for each analysis involving groups of para-
sites. We used a series of GLMMs (negative bino-
mial response with log link; function glmer.nb
from the R package lme4 (Bates et al. 2015))
where the abundance of each parasite species (in
a group) was treated as a response variable
potentially explained by abundance of other par-
asite species (within that group), host age or sex
or their interaction as xed effects, and year of
sampling as a random effect. Thus, if there were
nparasites in a group, this would potentially
result in nglobal GLMMs and their competing
submodels, providing models converged. Com-
peting models were compared using an informa-
tion theoretical approach and ranked based on
Akaikes information criterion corrected for small
sample sizes (AIC
c
); models with a DAIC
c
<2as
compared to the lowest AIC
c
(i.e., best) model
are listed in summary tables (Burnham and
Anderson 2002).
Most models converged without issue. How-
ever, models failed to converge where abun-
dance of E. muta was considered as the response
variable or when it was considered a predictor
variable with C. caudinata as the response para-
site. In comparison, the models did converge
with T. tenuis as the response variable providing
that abundance of E. muta was log
10
(x+1)-
transformed. This meant that only one of three
vwww.esajournals.org 4August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
global models was evaluated for the rst group
of parasites, that is, the T. tenuis response model.
For the second group of four ectoparasites
including the louse y, all four global models
converged and they and their submodels could
be compared with respect to overall t and with
respect to which factors in best-tting models
explained signicant variation in a focal para-
sitesabundance. For the last group of ve
directly transmitted feather parasites, the global
model with M. lagopus as the focal (response)
parasite failed to converge and was thus
excluded from further consideration; all other
global models in this group were t without
issue.
We addressed whether parasitehost trait rela-
tions for focal parasites found in a previous
study (Nielsen et al. 2020) were robust to inclu-
sion of other parasites as predictors by compar-
ing nine global models for focal parasites and
their associated submodels. Furthermore, we
could decompose models to address whether
host trait patterns for a focal parasite species
depended on the presence or absence of a partic-
ular predictor parasite. We also could test
whether predictor parasites accounted for addi-
tional variation in a focal parasitesabundance
beyond that explained by host traits and in ways
that made sense (e.g., negative association if
sharing tissue tropism, and positive association if
sharing pathways of infection). Where suitable,
these parasiteparasite associations are depicted
with response diagrams that summarize whether
any signicant parasiteparasite correlations are
consistently positive or negative across models,
and whether their signicance depends on which
parasite speciesabundance was used as the
response variable. Along with these response
diagrams, we report the model-estimated differ-
ence in mean abundance of focal parasites
between uninfected hosts and those infected with
the overall mean abundance of a predictor para-
site of interest.
Finally, to quantify the proportions of
observed variance in abundance of parasites
explained by the GLMMs, we estimated the total
(conditional) R
2
of the best-tting models for
each focal parasite, as well as the R
2
attributable
exclusively to the xed effects (i.e., to host traits
and coinfecting parasites; the marginal R
2
) fol-
lowing the methods described by Nakagawa
et al. (2017). The total variance referenced in
these methods (the denominators of the R
2
equa-
tions) incorporate residual variance using model
family and link function-specic derivations, in
addition to incorporating random effect variance
and xed effect variance. For all measures of R
2
,
we applied the r.squaredGLMM function in the
MuMIn package (Barton 2019), using the tri-
gamma function to estimate residual variance as
this is the most accurate for models using a log
link function (Nakagawa et al. 2017). We addi-
tionally estimated the variance explained indi-
vidually by the parasite and host trait xed
effects using a form of commonality analysis
whereby a version of the best-tting model with
parasite-related xed effects removed is also
evaluated for marginal R
2
, and this is subtracted
from the full-model marginal R
2
(Ray-Mukherjee
et al. 2014).
All analyses were performed using R (version
3.6.1; R Core Team 2019). Parasite abundance
refers to the number of infecting parasites of a
given species on an individual host, while mean
abundance refers to the mean number of infect-
ing parasites in a sample including both infected
and uninfected individuals (Reiczigel et al. 2019).
Mean intensity refers to the average number of
infecting parasites in a sample including only
infected hosts. Prevalence is the proportion of
hosts in a sample, which were infected with a
given parasite (Reiczigel et al. 2019). Condence
intervals around mean abundance and mean
intensity are bias-corrected and accelerated (BCa)
95% condence intervals, while parasite preva-
lences are provided with Clopper-Pearson 95%
condence intervals (R
ozsa et al. 2000, Reiczigel
et al. 2019). Where applicable, statistical signi-
cance is determined at a level of a=0.05.
RESULTS
Parasitological measures
Detailed descriptions of prevalences and levels
of infection by the 12 parasite species are pro-
vided in Nielsen et al. (2020). All parasite species
were reported in all 12 yr, except for T. tenuis,
which was reported in nine years only, and over-
all prevalence ranged from a low of 0.06 (0.05
0.08 95% CI) for M. lagopus to a high of 0.92
(0.910.94) for T. lagopi. Mean abundances ran-
ged from a low of below ve parasites per host
vwww.esajournals.org 5August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
(T. tenuis, M. islandicus, M. borealis, M. lagopus,
L.afnis, A. lagopi, and O. chloropus) to between
ve and ten parasites per host (G. lagopi and C.
caudinata) to between 15 and 30 parasites per
host (T. lagopi and S.holoapsis)to>150 cysts per
host (E.muta).
Group 1: endoparasites
Abundance of the nematode T. tenuis as response
variable.As mentioned, only models having T.
tenuis as the response variable and E. muta and
C. caudinata as predictors converged. Summa-
rizing the competing best models (Table 1)
showed that host age explained variation in
abundance of T. tenuis with adults having more
worms than juveniles; those patterns agree with
Nielsen et al. 2020. Thus, inclusion of two poten-
tial predictor parasites did not change host age
T. tenuis relations. The three best competing
models reported a signicant negative effect of
C. caudinata on T. tenuis, which share the hosts
ceca, but no signicant effect of the coccidian, E.
muta, on abundance of T. tenuis (Fig. 1a;
Appendix S1: Fig. S1). Compared with a host
without C. caudinata infection, the best-tting
model predicts that given the overall mean abun-
dance of C. caudinata (7.75; 95% CI 6.369.66),
the average T. tenuis abundance would decrease
by an estimated 12.8% in all years and age cate-
gories (Appendix S1: Fig. S1).
Group 2: skin ectoparasites and louse fly vector of
mites
Abundance of O. chloropus as response variable.
Previously, the louse y, O. chloropus,showeda
signicant juvenile bias and age:sex interaction
(juvenile bias specic to females) when controlling
only for year of sampling as a random factor
(Nielsen et al. 2020). Hereafter, we are referring
specically to previous results of Nielsen et al.
(2020) whenever current results are compared
to previously established patterns with respect
to host traits. These same effects of host traits
were seen in three of four best models including
three potential predictor parasites; age, but not
the age:sex interaction, was also signicant in
the fourth best model (Table 2). Thus, host trait
effects are robust in explaining abundance of O.
chloropus. Furthermore, M. islandicus had a con-
sistent negative relationship with O. chloropus;
the other two parasite species did not account
for signicant variation in abundance of the
louse y (Fig. 1b). Given an infection of M.
islandicus equal to its overall mean abundance
(3.17; 95% CI 2.544.06), the best-tting model
predicts, relative to an uninfected host, a 6.20%
decrease in O. chloropus mean abundance across
all years, age, and sex categories (Appendix S1:
Fig. S2).
Abundance of the mite M. islandicus as response
variable.For the most part, the previously
reported effects of host traits on M. islandicus did
not depend on inclusion of parasites as predic-
tors in expanded models. There were still a sig-
nicant juvenile bias in the three best models,
and a signicant female bias in the overall best-
tting model (Table 3). In the three best models,
M. islandicus was inversely associated with both
O. chloropus and A. lagopi and positively associ-
ated with M. borealis. Thus, parasites explained
additional variation in abundance of M. islandi-
cus. Compared with uninfected hosts, the best-
tting M. islandicus model predicted that, given
infections of either O. chloropus or A. lagopi equal
Table 1. Best-tting (DAIC
c
<2) generalized linear mixed-effects models relating Trichostrongylus tenuis abun-
dance to Capillaria caudinata, log-transformed Eimeria muta and Rock Ptarmigan age and/or sex, with year as a
random factor.
Model
Model estimates
df logLik AIC
c
DAIC
c
C. caudinata E. muta Age Sex Age:Sex
10.018* 1.135** 5580.496 1171.045 0
20.017* 1.114** 0.236 6 580.303 1172.681 1.636
30.018* 0.028 1.107** 6580.466 1173.006 1.961
Notes: Empty cells representing model predictors indicate absences of that predictor from the respective model (a blank row
of predictors would indicate the null model). Estimates in bold indicate statistical signicance of the estimate within the model.
Positive estimates for age indicate higher mean abundance in juveniles, while positive estimates for sex indicate higher mean
abundance in males.
*P<0.05; ** P<0.01.
vwww.esajournals.org 6August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
to their overall averages (0.76 and 1.80, respec-
tively; 95% CIs 0.690.84 and 1.442.24), mean
abundance of M. islandicus should decrease by
15.7% or 7.53% (Appendix S1: Fig. S3). Given a
M. borealis infection equal to its overall average
(0.56; 95% CI 0.460.69), predicted average M.
islandicus abundance increases by 55.9%
(Appendix S1: Fig. S3).
Abundance of the mite M. borealis as response
variable.The previously reported juvenile bias
in abundance of M. borealis is maintained in the
three best models (Table 4), despite the fact that
two of the parasite species also account for signif-
icant variation in abundance of M. borealis in all
three models. More specically, abundance of M.
borealis is positively related to abundance of M.
islandicus and inversely related to the abundance
of A. lagopi. This inverse relationship holds even
after accounting for variation in M. borealis attri-
butable to M. islandicus, which itself is inversely
associated with A. lagopi (see Abundance of the
mite M. islandicus as response variable). In com-
parison, O. chloropus does not signicantly relate
to M. borealis in any of the three best models
(Table 4) despite being consistently inversely
related to M. islandicus, the latter being consis-
tently positively associated with M. borealis. The
best-tting model predicts that, compared with
an uninfected host, having an infection of M.
islandicus equal to its overall average (3.17; 95%
CI 2.544.06) is associated with a 33.3% increase
in M. borealis abundance (Appendix S1: Fig. S4).
An infection of A. lagopi equal to its overall aver-
age (1.80; 95% CI 1.442.24) relates to a relative
predicted decrease of 9.18% in M. borealis abun-
dance (Appendix S1: Fig. S4).
Abundance of the louse A. lagopi as response
variable.There is no reason to expect that pat-
terns of juvenile bias and female bias specicto
adults in abundance of A. lagopi were spurious.
Both age and sex and their interaction remained
signicant in all four best models that included
relevant coinfecting parasites (Table 5). In agree-
ment with analyses above, M. islandicus is consis-
tently signicantly negatively correlated with A.
lagopi, whereas O. chloropus is not a signicant
predictor of A. lagopi abundance. M. borealis is
only present as a non-signicant predictor in the
best-tting model of A. lagopi abundance in
apparent contrast to patterns above, where abun-
dance of M. borealis is the response variable.
Fig. 1. Correlations between coinfecting parasites as
indicated by generalized linear mixed-effects models tak-
ing host traits (age and sex) and year into account.
Arrows move from predictor parasite to response and
indicate relationships where the predictor was present
and signicant within best-tting models. The placement
(distance) of the points in the network diagrams relative
to one another does not have any biological signicance.
Each panel represents a grouping of parasite species dis-
tinguished by transmission and tissue tropism: (a) the
endoparasites Capillaria caudinata,Trichostrongylus
tenuis,andEimeria muta; (b) the skin ectoparasitic group
with associated ectoparasitic y(Ornithomya chloropus,
Metamicrolichus islandicus,Myialges borealis,andAmyr-
sidea lagopi); and (c) the plumage ectoparasites Goniodes
lagopi,Lagopoecus afnis,Mironovia lagopus,Strelkoviacarus
holoaspis,andTetraolichus lagopi.
vwww.esajournals.org 7August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
Increasing M. islandicus infection abundance
from zero to the overall average (3.17; 95% CI
2.544.06) relates to a predicted increase of 28.8%
in mean abundance of A. lagopi, according to the
best-tting model (Appendix S1: Fig. S5).
Summary and importance of associations among
skin ectoparasites and y vector.The skin ectopar-
asites and y vector showed three recurring pair-
wise associations, and one seemingly less
consistent relation (Fig. 1b). M. islandicus was
consistently inversely related to O. chloropus and
A. lagopi and consistently positively related to M.
borealis. These associations were strong but did
not undermine any of the host trait relations in
explaining abundance, as previously docu-
mented for these four species.
Group 3: directly transmitted plumage
ectoparasites
Abundance of the mite M. lagopus as response
variable.Analyses with abundance of the quill
mite M. lagopus as the response variable resulted
in non-converging models, likely due to the rar-
ity of this species (mean overall prevalence of
~6%), and given the number of predictor vari-
ables in models. This species is not considered
further as a response parasite.
Abundance of the mite T. lagopi as response
variable.Inclusion of four other ectoparasitic
species in a global model accounting for varia-
tion in abundance of the mite T. lagopi did not
change the inuence of host traits on the mites
abundancethat is, there were still an adult
bias, a female bias, and a signicant age:sex
interaction where mean abundance increased
among females and decreased among males
with age. Parasiteparasite associations were
fairly straightforward: G.lagopi and S. holoaspis
are consistent, signicant positive estimators of
T. lagopi abundance, whereas M. lagopus was not
present as a correlate in the two best models of
the three and L. afnis was only present as a non-
signicant predictor in the best-tting (and third
best-tting) model (Table 6). The best-tting
model predicted 8.09% and 4.04% increases in
average T. lagopi abundance given infections of
G. lagopi and S. holoaspis equal to their overall
averages (9.07 and 16.86, respectively; 95% CIs
Table 2. Best-tting (DAIC
c
<2) generalized linear mixed-effects models relating Ornithomya chloropus abun-
dance to Amyrsidea lagopi,Myialges borealis,Metamicrolichus islandicus, and Rock Ptarmigan age and/or sex, with
year as a random factor.
Model
Model estimates
df logLik AIC
c
DAIC
c
A. lagopi M. borealis M. islandicus Age Sex Age:Sex
10.02** 0.594*** 0.232 0.471* 71325.25 2664.603 0
20.03 0.016* 0.602*** 0.233 0.471* 81324.96 2666.043 1.44
30.005 0.02** 0.601*** 0.229 0.46* 81325.03 2666.183 1.581
40.02** 0.331** 51328.15 2666.349 1.746
Note: Meaning of empty cells, estimates in bold, and positive and negative estimates for age and sex as in Table 1.
*P<0.05; ** P<0.01; *** P<0.001.
Table 3. Best-tting (DAIC
c
<2) generalized linear mixed-effects models relating Metamicrolichus islandicus abun-
dance to Amyrsidea lagopi,Myialges borealis,Ornithomya chloropus, and Rock Ptarmigan age and/or sex, with
year as a random factor.
Model
Model estimates
df logLik AIC
c
DAIC
c
A. lagopi M. borealis O. chloropus Age Sex Age:Sex
10.044** 0.799*** 0.224** 0.694* 0.724* 0.599 9 1309.74 2637.634 0
20.042* 0.801*** 0.244*** 0.999*** 0.289 8 1310.77 2637.669 0.035
30.045** 0.799*** 0.256*** 1.009*** 71312 2638.102 0.468
Note: Meaning of empty cells, estimates in bold, and positive and negative estimates for age and sex as in Table 1.
*P<0.05; ** P<0.01; *** P<0.001.
vwww.esajournals.org 8August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
8.359.94 and 14.3921.18), as compared to unin-
fected hosts (Appendix S1: Fig. S6).
Abundance of the mite S. holoaspis as response
variable.Previously, researchers reported juve-
nile and male biases in S. holoaspis, although the
sex bias was only present in the rst best-tting
model. When four other parasites are included in
the global model, the pattern of juvenile and
male host biases are still signicant and now con-
sistent in the two best models (Table 7). In agree-
ment with T. lagopi-response models above, S.
holoaspis abundance is positively correlated with
T. lagopi abundance. In addition, G. lagopi and L.
afnis correlate signicantly negatively and posi-
tively (respectively) with S. holoaspis.M. lagopus
is never a present and signicant predictor of S.
holoaspis. Here again, additional variation in
abundance of a focal species is explained by
inclusion of other parasite species. Infection
levels equal to the overall mean abundances of T.
lagopi,G. lagopi, and L. afnis (26.30, 9.07, and
3.26, respectively; 95% CIs 24.4928.60, 8.359.94,
and 2.953.62) related to a 37.8% increase, a
13.5% decrease, and a 14.5% increase in pre-
dicted mean abundance of S. holoaspis when
compared to uninfected hosts, according to the
best-tting model (Appendix S1: Fig. S7).
Abundance of the louse G. lagopi as response
variable.G. lagopi previously demonstrated a
female bias, juvenile bias, and an age:sex interac-
tion whereby females had signicantly higher
mean abundance only for adults. Those patterns
are maintained in the best-tting models where
other parasites are included (Table 8). In agree-
ment with analyses above, G. lagopi positively
correlated with T. lagopi and negatively correlated
with S. holoaspis. In addition, G. lagopi (like S.
holoaspis) correlates positively with the louse L.
afnis.M. lagopus is either absent from the models
or a non-signicant predictor (Table 8). Increases
in infection from no parasites to the overall mean
abundances of T. lagopi,S. holoaspis,andL. afnis
(26.30, 16.86, and 3.26, respectively; 95% CIs
24.4928.60, 14.3921.18, and 2.953.62) result in
predicted changes of 9.92%, 2.94%, and 14.6%
in G. lagopi mean abundance according to the
best-tting model (Appendix S1: Fig. S8).
Abundance of the louse L. afnis as response
variable.L. afnis previously showed a female
bias that only became signicant for adults,
Table 4. Best-tting (DAIC
c
<2) generalized linear mixed-effects models relating Myialges borealis abundance to
Amyrsidea lagopi,Metamicrolichus islandicus,Ornithomya chloropus, and Rock Ptarmigan age and/or sex, with
year as a random factor.
Model
Model estimates
df logLik AIC
c
DAIC
c
A. lagopi M. islandicus O. chloropus Age Sex Age:Sex
10.054*** 0.091*** 1.107*** 6785.537 1583.149 0
20.055* 0.088*** 0.088 1.124*** 7784.749 1583.597 0.448
30.054* 0.091*** 1.113*** 0.045 7 785.505 1585.109 1.961
Note: Meaning of empty cells, estimates in bold, and positive and negative estimates for age and sex as in Table 1.
*P<0.05; *** P<0.001.
Table 5. Best-tting (DAIC
c
<2) generalized linear mixed-effects models relating Amyrsidea lagopi abundance to
Myialges borealis,Metamicrolichus islandicus,Ornithomya chloropus, and Rock Ptarmigan age and/or sex, with
year as a random factor.
Model
Model estimates
df logLik AIC
c
DAIC
c
M. borealis M. islandicus O. chloropus Age Sex Age:Sex
1 0.15 0.107*** 1.346** 2.056*** 2.465*** 81067.68 2151.491 0
20.067*** 1.36** 2.051*** 2.425*** 71069.11 2152.328 0.837
3 0.152 0.11*** 0.109 1.426*** 2.01*** 2.415*** 91067.1 2152.366 0.875
40.069*** 0.106 1.435*** 2.007*** 2.377*** 81068.57 2153.26 1.769
Note: Meaning of empty cells, estimates in bold, and positive and negative estimates for age and sex as in Table 1.
** P <0.01; *** P<0.001.
vwww.esajournals.org 9August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
which were generally less infected than juveniles.
Interestingly, this interaction was lost or not sig-
nicant in the four best models that included
other parasites (Table 9). This lone case of loss of
a signicant host trait effect for a parasite species
requires further analyses; it might be due to the
strong positive relationship between L. afnis
and G. lagopi. We see a potentially signicant
female bias in L. afnis that is specic to adults
when considering only those hosts infected by G.
lagopi (Fig. 2). Note that the 95% CIs between the
sexes overlap much less for adult hosts infected
with G. lagopi than for uninfected adult hosts
note also that the number of hosts infected with
G. lagopi is much larger than the number of hosts
uninfected with that parasite (Fig. 2).
L. afnis also demonstrated consistent and sig-
nicant positive correlations with T. lagopi in the
four best models (Table 9), despite T. lagopi not
demonstrating signicant positive correlations
with L. afnis when modeled as the response.
There is close agreement here with the S.
holoaspis- and G. lagopi-focused models above:
Both parasites demonstrate positive correlations
with L. afnis. Furthermore, abundance of M.
lagopus is not a signicant predictor of L. afnis.
In other words, a majority of parasites in this
group do account for signicant variation in
abundance of the focal parasite. The best-tting
model predicted that, compared with uninfected
hosts, infection loads equal to the overall mean
abundances for G. lagopi,S. holoaspis, and T.
lagopi (9.07, 16.86, 26.30, respectively; 95% CIs
8.359.94, 14.3921.18, and 24.4928.60) would
be associated with 21.4%, 3.19%, and 11.1%
increases in relative L. afnis mean abundance
(Appendix S1: Fig. S9).
Summary and importance of associations among
plumage ectoparasites.Signicant parasitepara-
site associations for plumage ectoparasites are
visualized in Fig. 1c. M. lagopus could not be
modeled as the focal parasite and as a predictor
parasite showed no signicant associations with
any of the four other focal parasite species. In
comparison, abundance of each of the other four
focal species was explained by consideration of
at least two (T.lagopi) and most often three (L.
afnis,G. lagopi, and S. holoaspis) other parasite
species in models. Only the pair G. lagopi and S.
holoaspis showed negative correlations; the
remaining ve pairs of species showed signi-
cant positive correlations. The strength of the
positive associations between L. afnis and G.
lagopi might have been sufcient to render
Table 6. Best-tting (DAIC
c
<2) generalized linear mixed-effects models relating Tetraolichus lagopi abundance to
Goniodes lagopi,Lagopoecus afnis,Mironovia lagopus,Strelkoviacarus holoaspis, and Rock Ptarmigan age and/or
sex, with year as a random factor.
Model
Model estimates
df logLik AIC
c
DAIC
c
G. lagopi L. afnis M. lagopus S. holoaspis Age Sex Age:Sex
10.009*** 0.01 0.002*** 0.367*** 1.324*** 1.034*** 94281.71 8581.601 0
20.01*** 0.003*** 0.342*** 1.334*** 1.035*** 84283.31 8582.755 1.154
30.009*** 0.01 0.029 0.002*** 0.356*** 1.313*** 1.024*** 10 4281.51 8583.235 1.634
Note: Meaning of empty cells, estimates in bold, and positive and negative estimates for age and sex as in Table 1.
*** P <0.001.
Table 7. Best-tting (DAIC
c
<2) generalized linear mixed-effects models relating Strelkoviacarus holoaspis abun-
dance to Goniodes lagopi,Lagopoecus afnis,Mironovia lagopus,Tetraolichus lagopi, and Rock Ptarmigan age and/
or sex, with year as a random factor.
Model
Model estimates
df logLik AIC
c
DAIC
c
G. lagopi L. afnis M. lagopus T. lagopi Age Sex Age:Sex
10.016* 0.042* 0.012*** 1.271*** 0.552** 82793.81 5603.766 0
20.016* 0.042* 0.075 0.012*** 1.261*** 0.544** 92793.65 5605.478 1.713
Note: Meaning of empty cells, estimates in bold, and positive and negative estimates for age and sex as in Table 1.
*P<0.05; ** P<0.01; *** P<0.001.
vwww.esajournals.org 10 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
spurious the previous interaction between age
and sex in explaining abundance of L. afnis.
Inclusion of A. lagopi with group 3 parasites
instead of group 2 parasites
Removing A. lagopi from the analyses involv-
ing the group of ectoparasites including the louse
y did not signicantly change any inferences
concerning the remaining predictors, even in
cases where A. lagopi had previously correlated
signicantly with the focal species. In other
words, the patterns initially observed and
described above for this parasite group with
respect to host traits and other parasites were not
dependent on inclusion of A. lagopi.
Adding A. lagopi to the third group of directly
transmitted plumage ectoparasites had no effect
on the majority of the patterns described above
with respect to host traits and other parasites.
The updated analyses did identify a negative A.
lagopiG. lagopi association. In the three best-
tting A. lagopi-focused models, G. lagopi was pre-
sent as a consistent, negative predictor
(Appendix S1: Table S1; no other parasites were
present as predictors in the overall best-tting
model). Intriguingly, when A. lagopi was intro-
duced into models with G. lagopi abundance as
the response, A. lagopi was present in the rst
and third best-tting models, but never as a sig-
nicant predictor (Appendix S1: Table S2). Thus,
the amblycerid shows two candidate correlations
when considered with other parasites in the sec-
ond group, and one possible candidate correla-
tion when considered with the third group of
parasites.
Variance explained by fixed and random effects
The total variance in focal parasite abundance
explained by each of the best-tting models is
visualized in Fig. 3, with total R
2
partitioned into
proportions attributable to host trait xed effects
(age and sex), coinfecting parasite xed effects,
and the random effect (year). Total R
2
of the
models ranged from 1.25% to 33.8% (median =
4.65%), with the majority of focal species (six of
nine) demonstrating total R
2
values less than
10%. In most cases (six of nine), the majority of
the explained variance was due to the xed effect
predictors rather than the random effect year. Of
the variance explained by the xed effects,
Table 8. Best-tting (DAIC
c
<2) generalized linear mixed-effects models relating Goniodes lagopi abundance to
Lagopoecus afnis,Mironovia lagopus,Strelkoviacarus holoaspis, Tetraolichus lagopi, and Rock Ptarmigan age and/or
sex, with year as a random factor.
Model
Model estimates
df logLik AIC
c
DAIC
c
L. afnis M. lagopus S. holoaspis T. lagopi Age Sex Age:Sex
10.042*** 0.002* 0.004** 1.062*** 0.644*** 0.659*** 93177.63 6373.438 0
20.042*** 0.057 0.002* 0.004** 1.047*** 0.658*** 0.671*** 10 3177.28 6374.761 1.322
Note: Meaning of empty cells, estimates in bold, and positive and negative estimates for age and sex as in Table 1.
*P<0.05; ** P<0.01; *** P<0.001.
Table 9. Best-tting (DAIC
c
<2) generalized linear mixed-effects models relating Lagopoecus afnis abundance to
Goniodes lagopi,Mironovia lagopus,Strelkoviacarus holoaspis, Tetraolichus lagopi, and Rock Ptarmigan age and/or
sex, with year as a random factor.
Model
Model estimates
df logLik AIC
c
DAIC
c
G. lagopi M. lagopus S. holoaspis T. lagopi Age Sex Age:Sex
10.021*** 0.002* 0.004* 1.176*** 0.216* 82234.29 4484.725 0
20.021*** 0.002* 0.004* 1.028*** 0.433* 0.286 9 2233.35 4484.87 0.145
30.021*** 0.057 0.002* 0.004* 1.192*** 0.211* 92233.74 4485.661 0.935
40.021*** 0.047 0.002* 0.003* 1.055*** 0.409* 0.26 10 2232.98 4486.162 1.436
Note: Meaning of empty cells, estimates in bold, and positive and negative estimates for age and sex as in Table 1.
*P<0.05; *** P<0.001.
vwww.esajournals.org 11 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
whether the majority related to host trait factors
or coinfecting parasites depended on the parasite
group being considered: for all but one (A. lagopi)
of the ectoparasites in group 2, more variance in
focal parasite abundance was explained by coin-
fecting parasites, whereas more variance was
explained by host traits than by coinfecting para-
sites for the third group of ectoparasites. For the
endoparasitic T. tenuis, more variance in abun-
dance was explained by coinfecting parasites
(E. muta and C. caudinata) than host age or sex,
though the vast majority of the total (small; 1.3%)
R
2
was due to the random effect. While the quan-
tied total R
2
values of the models indicate that
the majority of variance in parasite abundance
remains unexplained, a strong, positive relation-
ship between overall focal parasite prevalence
and total model R
2
is important (Fig. 4; positive
correlation R
2
=0.82, P<0.001). This result sug-
gests that the proportion of hosts infected with
the focal parasite places a limit on total variance
in a focal parasites abundance explained by the
relevant model.
DISCUSSION
This study was undertaken to determine
whether coinfecting parasites accounted for varia-
tion in abundance of focal parasites in addition
to, or in lieu of, variation in levels of focal para-
sitism explained by host traits. Following from
this, we were interested in whether previously
documented host traitfocal parasitism relations
(Nielsen et al. 2020) might have been spurious
and best explained by considering infection by
coinfecting parasites. As mentioned, we also were
interested in whether any signicant relations
between coinfecting parasite species made sense
based on their inclusion in groups of parasite spe-
cies sharing tissue tropism and/or pathways of
infection. Finding that coinfecting parasites
explained additional variation in abundance of
n = 39
=3
n
=
136
n = 39
n
=
29
n =
n
= 85
n
=
30
7
n
=
3
1
6
Sex
Femal e
Male
Not infected by G. lagopi Infected by G. lagopi
JA JA
0
2
4
6
8
Host age
L. affins mean abundance (±95% CI)
Fig. 2. Mean abundance of Lagopoecus afnis parasitizing Rock Ptarmigan (Lagopus muta) either infected or
uninfected by Goniodes lagopi. Classes of ptarmigan are further differentiated by their age (juvenile or adult) and
sex, demonstrating potential age:sex interactions. The 95% condence intervals are bias-corrected and accelerated
condence intervals.
vwww.esajournals.org 12 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
focal parasites in ways that largely made sense,
we next asked whether the magnitude of varia-
tion in levels of focal parasitism explained by
coinfecting parasites exceeded that explained by
host traits. Finally, we discuss whether overall
levels of variation explained related to attributes
of focal parasites (e.g., inclusion in particular
groups and prevalence). We deal with each gen-
eral objective and specicndings relating to
these general objectives, in turn.
Coinfecting parasites explain additional variation
in focal parasite abundance
We found that coinfecting parasites often
accounted for signicant additional variation in
parasitism of Rock Ptarmigan by focal species in
expanded models and that previously docu-
mented hosttrait effects for these focal parasites
largely appear real. We chose to control for unex-
plained variation among years by including year
as random factor in models. Co-occurring para-
sites showed positive, negative, or no correla-
tions across the three groups (endoparasites, skin
ectoparasites with the louse y, and directly
transmitted plumage ectoparasites).
We recognize that limiting our analyses to rela-
tionships between parasites sharing tissue trop-
isms and transmission modesthat is, the three
parasite groupingsprevents any detection of
possible indirect interactions between infecting
species with more contrasting ecologies. Relation-
ships between disparate parasites have been
Fig. 3. Quantities of variance in observed focal parasite abundances explained by the best-tting generalized
linear mixed-effects models, measured using estimated R
2
. Total R
2
is partitioned into the proportion of variance
explained by the random factor (year) and the proportion explained by xed effects, the latter then further parti-
tioned into the proportion of variance explained by host trait vs. coinfecting parasite predictors. The three groups
of parasite species ordered from left to right are distinguished by transmission and tissue tropism: the endopara-
sites Capillaria caudinata,Trichostrongylus tenuis, and Eimeria muta; the skin ectoparasitic group with associated
ectoparasitic y (Ectoparasites 1; Ornithomya chloropus,Metamicrolichus islandicus,Myialges borealis, and Amyrsidea
lagopi); and the plumage ectoparasites (Ectoparasites 2; Goniodes lagopi,Lagopoecus afnis,Mironovia lagopus,
Strelkoviacarus holoaspis, and Tetraolichus lagopi). Parasite species are ordered along the x-axis within groups
according to increasing total R
2
. Note that the y-axis ranges from zero to slightly less than 0.4.
vwww.esajournals.org 13 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
documented in other host systems; for example,
in free-ranging African buffalo (Syncerus caffer),
gastrointestinal nematode (macroparasite) preva-
lence negatively correlated with that of a
mycobacterium (microparasite), with evidence
strongly suggestive of immune-mediated interac-
tions (Jolles et al. 2008, Ezenwa et al. 2010). In a
system much more similar to that of our present
study, willow ptarmigan (Lagopus lagopus)endo-
and ectoparasites were found to positively corre-
late, when analyzing the sums across parasite
species within certain taxonomic or trophic
groups (Holmstad et al. 2008). Indirect
interactions could inuence Icelandic Rock
Ptarmigan parasites and explain additional varia-
tion in observed focal abundances. We note, how-
ever, that in preliminary basic analyses of all
potential pairwise abundance correlations for 12
species, only two consistent correlations arose out-
side of those tested in our more thorough within-
group models. T. lagopi correlated positively with
both M.borealis and M. islandicus. No correlations
were observed between parasites from group 2 or
3 with any of the more disparate group 1 para-
sites. Therefore, while indirect, immune-mediated
relationships between infecting parasites may
arise, we suspect that the majority of variation in
experienced focal parasite abundances explained
by other parasites infecting these hosts is con-
strained to associations with species within the
same groupings. The structure and sparsity of our
data prevent us from including all coinfecting par-
asites into focal parasite models (given that we
required successful convergence of our GLMMs).
However, our groupings based on tissue tropism
and transmission mode allow us to address hypo-
thetical biological drivers of the documented rela-
tionships. We consider each of the groups
parasites in turn and the respective possible mech-
anisms driving the correlations, acknowledging
that more indirect, immune-mediated interactions
may still be at play and that the emergence of sig-
nicant correlations does not necessarily indicate
direct interactions between species.
With respect to the endoparasites, T. tenuis
abundance decreased signicantly with increas-
ing infection of the other nematode, C. caudin-
ata, a pattern that could have arisen from direct
interspecic competitive interactions, as both
inhabit the ceca (Stenkewitz 2017; see also Mideo
2009 for examples of parasite competitive inter-
actions). Alternatively, the negative association
could result from increased mortality of dually
infected birds: Abundance of C. caudinata is
associated with higher overwintering morality
(Stenkewitz 2017). Indirect, immune-mediated
negative interactions also cannot be ruled out
(Poulin 2001). Despite partially overlapping in
niche, E. muta and T. tenuis showed no associa-
tion. This lack of any association may be due to
the high aggregation of E. muta resulting in a
reduced probability of single hosts experiencing
high infection intensities of both parasite species
simultaneously (Morrill et al. 2017).
0.0
0.1
0.2
0.3
0.00 0.25 0.50 0.75 1.00
Prevalence (±95% CI)
Total R2
Parasite group
Endoparasites
Ectoparasites 1
Ectoparasites 2
Fig. 4. Estimates of the total variance explained (R
2
)
by best-tting generalized linear mixed-effects models
of focal parasite abundances generally increasing with
overall prevalence of the focal parasite. Measures of
prevalence are provided with Clopper-Pearson 95% con-
dence intervals. Each point represents a single focal
parasite species. The three groups of parasite species are
distinguished by transmission and tissue tropism: the
endoparasites Capillaria caudinata,Trichostrongylus
tenuis,andEimeria muta; the skin ectoparasitic group
with associated ectoparasitic y (Ectoparasites 1;
Ornithomya chloropus,Metamicrolichus islandicus,Myialges
borealis,andAmyrsidea lagopi); and the plumage ectopar-
asites (Ectoparasites 2; Goniodes lagopi,Lagopoecus afnis,
Mironovia lagopus,Strelkoviacarus holoaspis,andTetraoli-
chus lagopi).
vwww.esajournals.org 14 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
With respect to the skin ectoparasites and the
louse y, the correlations between species were
predominantly negative. The exceptional posi-
tive association between M. borealis and M.
islandicus may have arisen because both species
are thought to be transmitted by the louse y, O.
chloropus. However, the presence of pre-imaginal
stage M. borealis on Rock Ptarmigan has never
been detected (our study only assessed infection
by M. borealis adults; Mironov et al. 2010). The
negative relationship between M. islandicus and
O. chloropus is consistent with observations that
the louse y avoids hosts heavily infected by M.
islandicus (Stenkewitz 2017). The recurring nega-
tive correlations between A. lagopi and M. islandi-
cus, as well as the negative A. lagopiM. borealis
relationship observed for M. borealis-response mod-
els, might be due to keratinization of the skin
following parasitism by the astigmatan mite pre-
venting effective parasitism by the louse, or due to
more direct competitive interspecicinteractionson
the hosts skin. Intriguingly, A. lagopi infection corre-
lated with reduced fecundity, whereas M. islandicus
infection was associated with reduced adult
mortality of Rock Ptarmigan (Stenkewitz et al. 2016).
Dual costs for hosts with both also might explain
negative associations between those species.
The third group characterized by directly
transmitted plumage ectoparasites demonstrated
mostly positive correlations between species
pairs, with only one negative correlation. These
positive relationships might be due to the species
sharing transmission pathways (Stenkewitz
2017), but because their niche is sufciently var-
ied (i.e., quills, vanes, down, general plumage),
they possibly avoid strong, competitive interac-
tions. Importantly, parasitessuch as the astig-
matan T. lagopi may in fact be paraphages, whose
infestation is benecial to hosts and whereby
cumulative costs with other parasites are not
expected (Do~
na et al. 2019). The strong negative
correlations between S. holoapsis and G. lagopi
may result from competitive interactions
between the two species given their partially
shared use of niche (i.e., host plumage). Such an
interaction, like others presented, would require
experimental verication. The absence of M. lago-
pus as a signicant predictor of any focal parasite
abundance may be due to its relatively low
prevalence (~6%) and mean intensity (<3 mites
per infected host on average), which additionally
means that it is less likely to be directly transmit-
ted between birds. Alternatively, as M. lagopus
mites specically infect within the quills, it may
not directly interact with other coinfecting
ectoparasites (Sk
ırnisson and Nielsen 2019).
Documented host traitfocal parasite relations
are mostly not spurious
In initiating this study, we thought of parasite
parasite interactions as potentially indirect ways
to produce apparent age or sex biases in para-
sitism. Coinfection is the norm in parasitehost
interactions (Rigaud et al. 2010). Here, infestation
by one parasite could result in facilitation (or
suppression) of infection by anotherthese
hypothetical processes might apply to co-
occurring symbionts, generally (Z
el
e et al. 2018).
When, for example, one parasite species shows a
strong sex-biased pattern of infection, a sec-
ondary facilitated parasite likely also shows this
sex-biased pattern, because the original parasite
species creates conditions suitable for the next
species. Parasites also can reduce host exposure
or susceptibility to infective stages of another
parasite species, or reduce coinfection through
direct competitive interactions (Goater et al.
2014, Karvonen et al. 2019). Such occurrences
might explain, for example, the paucity of coin-
fections by trematode species infecting snails
(Sousa 1993). In either case of parasite facilitation
or suppression, infection by the second parasite
is more inuenced by the rst parasite altering
host susceptibility or exposure (or secondary par-
asite establishment) than by host factors per se.
Such potentially spurious hosttrait patterns
are clearly not occurring frequently for Rock
Ptarmigan parasites. We found evidence of a
spurious hosttrait relation in only one of nine
focal parasites where this problem could be
addressed. When not taking coinfections into
account, both L. afnis and G. lagopi mean abun-
dance was higher in juveniles than adults, and a
signicant female bias emerged exclusively
among adults (Nielsen et al. 2020). Such patterns
were previously explained as a consequence of
brood rearing and its costs falling on adult
females alone, thereby increasing directly trans-
mitted parasites among juveniles and adult
females (Nielsen et al. 2020). Remarkably, L. af-
nis (but not G. lagopi) lost any signicant interac-
tive effect between host sex and age when other
vwww.esajournals.org 15 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
predictor parasites in its group were considered.
This result may be due to L. afnisstrong, posi-
tive correlation with G. lagopi: The association
between the two causes a mirroredinteractive
pattern with respect to host traits to emerge in L.
afnis. Alternatively, including strongly correlat-
ing coinfecting parasites in L. afnis models
reduced residual variance that could be
explained by a possibly present, but relatively
weak, age:sex interaction. However, note that in
hosts that are coinfected with G. lagopi, there is a
pattern more reminiscent of a juvenile bias with
a female bias specic to adults in parasitism by
L. afnis (Fig. 2).
Documenting cases of direct or indirect facili-
tation or suppression among parasite species is
notoriously difcult (Hellard et al. 2015). Such
documentation requires detailed parasitehost
natural history observations at different ecologi-
cal scales, and possibly also measurements of
host behavior, physiology, and/or parasitism
levels following experimental infections. As men-
tioned in Introduction, we also can envision a sit-
uation where two parasite species are positively
or inversely related, but it is differential age or
sex biases in parasitism that produce those asso-
ciations without the need to invoke either direct
or indirect interactions between the parasite spe-
cies. Clearly, there is a need for detailed observa-
tions such as direct interactions between parasite
species and/or experiments (e.g., Gopko et al.
2018, Halliday et al. 2018) when such interactions
are being invoked to explain patterns.
Parasiteparasite correlations also might
emerge due to unconsidered infections by
another species. Many studies concern pairwise
associations between coinfecting parasites (Kau-
nisto et al. 2018). It is quite possible, however,
that a real association (either direct or indirect)
with another species drives the observed pattern;
that is, an emergent correlation between the rst
two species does not reect a real association.
For example, we found instances of a strong neg-
ative association between O. chloropus and M.
borealis when measuring pairwise associations
alone (results not presented). These correlations
disappeared once models of O. choloropus or M.
borealis incorporated M. islandicus abundance.
That is, the correlations were spurious due to
shared, strong associations with M. islandicus,
which was positively associated with M. borealis
and negatively associated with O. chloropus.We
caution against making inferences from pairwise
correlations between parasite species where
other coinfecting parasites are present.
Relative variation in focal parasite abundance
explained by predictors, and total variation
explained in relation to prevalence of focal
parasites
Caution in making inferences about species
species interactions in observational studies is
further warranted as associations might be dri-
ven by unconsidered environmental variables or
issues of spatial (or temporal) scale (Blanchet
et al. 2020). For example, correlations between
parasite species across years might reect simi-
lar or different responses of those parasite (or
host) species to annual variation in weather con-
ditions, without the need to invoke parasite
parasite interactions. Remarkably, the random
effect of year explains a similar or greater
amount of variance in focal species mean abun-
dance as does either xed effects (host traits or
coinfecting parasites) for six of nine species. We
do not yet know whether this yearly variation
relates to anything biologically (e.g., abundance
of alternative host species for many of the gener-
alist parasites considered). Similar cautions of
co-occurrence not evidencing speciesspecies
interactions may also inform recently developed
multivariate approaches to joint species distribu-
tion modeling, including the hierarchical model-
ing of species communities (HMSC; Ovaskainen
et al. 2017). Such approaches can disentangle
various host- and environment-related drivers
of multiple parasite speciesoccurrences, while
highlighting resilient parasiteparasite associa-
tions (Dallas et al. 2019, Krasnov et al. 2020,
Veitch et al. 2020). However, the HMSC frame-
work is currently limited to analyzing parasite
presenceabsence data as it is ill-equipped to
handle highly positively skewed response vari-
ables such as parasite abundance (Tikhonov
et al. 2020).
With respect to variance in focal parasite abun-
dance explained by xed and random effects,
three broad patterns emerge (Fig. 3). First, both
host trait and coinfecting parasites relate signi-
cantly to focal parasite abundance, but which
xed effect type explains the most variance var-
ies among focal species (or the group to which it
vwww.esajournals.org 16 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
belongs). Groups differ in transmission path-
ways and degree of niche overlap (e.g., niche
overlap is higher among the skin mites of the sec-
ond group than across all the plumage ectopara-
sites of group 3). In addition, group 2 mites are
both thought to be vectored by O. chloropus,
which would make their abundances correlated.
With respect to group 3 parasites, host behavior-
or condition-dependent establishment of directly
transmitted ectoparasites (and minimal competi-
tion between them) could explain why host trait
xed effects are consistently more important
therein than coinfecting parasite factors.
The second broad pattern is that the random
effect year explains a similar or greater amount
of variance in focal parasite abundance as at least
one of the two types of xed effects for six of nine
focal species. We currently do not know why
year-to-year variation in mean abundance is rela-
tively inconsequential for A. lagopi,M. borealis,
and S. holoaspis. The third pattern is that across
all focal parasites, the majority of variance in
abundance is left unexplained. These unex-
plained portions may be due to unconsidered
host-related or environmental factors or be statis-
tical in origin. The prevalence of parasites (per-
cent occurrence) is both a biological and a
statistical measure, as it relates to parasite aggre-
gation and mean abundance, and their interplay
with host sampling methodology (Poulin 1993).
Intriguingly, prevalence of focal parasites related
to the overall variance in our models that was
explained. This nding may be due to the nega-
tive and positive relationships between parasite
overdispersion and, respectively, prevalence and
variance: Lower prevalence could indicate higher
aggregation and a higher total variation in abun-
dance requiring explanation. This overall vari-
ance explained is consistent with average
amounts of variance explained in ecological stud-
ies (Møller and Jennions 2002).
One specic pattern emerging from our
results also requires explanation, that is, the
occurrence in two instances of signicant corre-
lations between two parasite species but the
absence of any reciprocal correlation depending
on which species is modeled as the response (fo-
cal) parasite. For example, M. borealis was nega-
tively predicted by A. lagopi though the
reciprocal pattern did not emerge, and T. lagopi
positively predicted L. afnis abundance but not
vice versa. We suspect that these inconsisten-
ciesare purely statistical in origin. One source
of the contrasting patterns might be the limits
on explainable variance in parasite abundances:
Even with an association between two parasites,
no correlation can arise when there is insuf-
cient variance remaining in the response vari-
able to infer the pattern, especially when it is
subject to many competingexplanatory pre-
dictors. Thus, a further caution is that absences
of associations between parasite pairs in models
do not necessarily indicate that no such patterns
exist. Rather, lack of patterns may simply arise
from a lack of sufcient observed variation in
focal parasite abundance to detect the associa-
tion. Direct observations of associations between
parasite species might still uncover biological
interactions that are difcult to detectstatisti-
cally using data on associations against many
predictors.
Summary and conclusions
In summary, inclusion of coinfecting parasite
species did explain variation in abundance of
focal parasite species in addition to signicant
effects of host traits on focal parasite species, pre-
viously reported and conrmed in this study.
Parasiteparasite associations, both negative and
positive, were widespread for parasites of Ice-
landic Rock Ptarmigan. It is possible that impor-
tant factors were missed, such as within- or
between-year variation in condition of birds or
weather-related survival of infective stages of
parasites. For example, birds in poor relative con-
dition can accumulate parasites of different spe-
cies and show further declines in condition as a
result (Beldomenico and Begon 2010). Such birds
would thus show positive associations between
parasite species when lightly parasitized good-
condition birds are also included in samples.
Although realistic, such a scenario cannot
explain negative associations between parasite
species. Such ndings might be attributable to
real interactions between species or again some
third undocumented factor affecting both species
differently (such as weather-related transmission
success). Finally, future studies should not only
report whether host traits and coinfecting para-
sites explain signicant variation in abundance
of focal species, but also use advanced
approaches to report on the relative contribution
vwww.esajournals.org 17 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
of each, as well as the percentage of variation left
unexplained.
ACKNOWLEDGMENTS
We thank the Icelandic Institute of Natural History,
M
yvatn Research Station, and Northeast Iceland Nat-
ure Centre for logistical support. We thank all the peo-
ple that have contributed in the eld and laboratory,
and to data analysis. The authors have no conicts of
interest to declare. Funding was provided by the Ice-
landic Research Fund (Grant No. 090207021), the
Hunting Card Fund, the Landsvirkjun Energy Fund,
the University of Iceland Research Fund, the Icelandic
Institute of Natural History, and a Natural Sciences
and Engineering Research Council (NSERC) Discovery
Grant (Grant No. 100118). Data were originally col-
lected for a study conceived by
Olafur K. Nielsen
(OKN) and Karl Sk
ırnisson (KS). Ute Stenkewitz (US)
initiated analyses of parasiteparasite interactions in
her PhD thesis, while ideas for the current study were
developed by Andr
e Morrill (AM) and Mark R. Forbes
(MRF). Guðn
yR.P
alsd
ottir, KS, and US performed
laboratory diagnostics. AM performed all statistical
analyses and co-wrote the initial draft with MRF. OKN
and KS reviewed and edited the manuscript with AM
and MRF. All authors contributed to analysis of results
and biological interpretations, and have read and
approved the nal manuscript.
LITERATURE CITED
Barton, K. 2019. MuMIn: multi-model inference. R
package version 1.43.6. https://cran.r-project.org/
package=MuMIn
Bates, D., M. M
achler, B. Bolker, and S. Walker. 2015.
Fitting linear mixed-effects models using lme4.
Journal of Statistical Software 67:148.
Beldomenico, P. M., and M. Begon. 2010. Disease spread,
susceptibility and infection intensity: Vicious circles?
Trends in Ecology & Evolution 25:2127.
Blanchet, F. G., K. Cazelles, and D. Gravel. 2020. Co-
occurrence is not evidence of ecological interac-
tions. Ecology Letters 23:10501063.
Burnham, K. P., and D. R. Anderson. 2002. Model
selection and multimodel inference: a practical
information-theoretic approach. Springer-Verlag,
New York, New York, USA.
Carius, H. J., T. J. Little, and D. Ebert. 2001. Genetic
variation in a host-parasite association: potential
for coevolution and frequency-dependent selection.
Evolution 55:11361145.
Carvalho-Pereira, T. S. A., et al. 2019. Coinfection mod-
ies carriage of enzootic and zoonotic parasites in
Norway rats from an urban slum. Ecosphere 10:
e02887.
Dallas, T. A., A.-L. Laine, and O. Ovaskainen. 2019.
Detecting parasite associations within multi-
species host and parasite communities. Proceed-
ings of the Royal Society B: Biological Sciences
286:20191109.
Do~
na, J., H. Proctor, D. Serrano, K. P. Johnson, A.-O.-V.
Oploo, J. C. Huguet-Tapia, M. S. Ascunce, and R.
Jovani. 2019. Feather mites play a role in cleaning
host feathers: new insights from DNA metabarcod-
ing and microscopy. Molecular Ecology 28:203218.
Ezenwa, V. O., R. S. Etienne, G. Luikart, A. Beja-
Pereira, and A. E. Jolles. 2010. Hidden conse-
quences of living in a wormy world: Nematode-
induced immune suppression facilitates tuberculo-
sis invasion in African buffalo. American Natural-
ist 176:613624.
Fecchio, A., M. Lima, P. Silveira, A. Ribas, R. Caparroz,
and M. Marini. 2015. Age, but not sex and season-
ality, inuence Haemosporida prevalence in White-
banded tanagers (Neothraupis fasciata) from central
Brazil. Canadian Journal of Zoology 93:7177.
Forbes, M. R. 1993. Parasitism and host reproductive
effort. Oikos 67:444450.
Goater, T. M., C. P. Goater, and G. W. Esch. 2014. Para-
site community ecology. Pages 356378 in Para-
sitism: the diversity and ecology of animal
parasites. Second edition. Cambridge University
Press, Cambridge, UK.
Gopko, M., M. M. R. Chowdhury, and J. Taskinen.
2018. Interactions between two parasites of brown
trout (Salmo trutta): consequences of preinfection.
Ecology and Evolution 8:99869997.
Halliday, F. W., J. Umbanhowar, and C. E. Mitchell.
2018. A host immune hormone modies parasite
species interactions and epidemics: insights from a
eld manipulation. Proceedings of the Royal Soci-
ety B: Biological Sciences 285:20182075.
Hellard, E., D. Fouchet, F. Vavre, and D. Pontier. 2015.
Parasiteparasite interactions in the wild: How to
detect them? Trends in Parasitology 31:640652.
Holmstad, P. R., K. H. Jensen, and A. Skorping. 2008.
Ectoparasite intensities are correlated with
endoparasite infection loads in willow ptarmigan.
Oikos 117:515520.
Jackson, J., R. Pleass, J. Cable, J. Bradley, and R. Tins-
ley. 2006. Heterogenous interspecic interactions in
a hostparasite system. International Journal for
Parasitology 36:13411349.
Jolles, A. E., V. O. Ezenwa, R. S. Etienne, W. C. Turner,
and H. Olff. 2008. Interactions between macropara-
sites and microparasites drive infection patterns in
free-ranging African buffalo. Ecology 89:22392250.
vwww.esajournals.org 18 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
Karvonen, A., J. Jokela, and A.-L. Laine. 2019. Impor-
tance of sequence and timing in parasite coinfec-
tions. Trends in parasitology 35:109118.
Kaunisto, K. M., A. Morrill, and M. R. Forbes. 2018.
Negative covariance between water mite and gre-
garine parasitism for adult dragonies, Leucorrhinia
intacta (hagen): An age-related pattern? Parasitol-
ogy Research 117:39093915.
Krasnov, B. R., M. Stanko, M. Lareschi, and I. S.
Khokhlova. 2020. Species co-occurrences in
ectoparasite infracommunities: Accounting for con-
founding factors associated with space, time, and
host community composition. Ecological Entomol-
ogy 45:11581171.
Krasnov, B. R., M. Stanko, and S. Morand. 2006. Age-
dependent ea (Siphonaptera) parasitism in
rodents: A hosts life history matters. Journal of
Parasitology 92:242248.
Little, T. J., and D. Ebert. 1999. Associations between
parasitism and host genotype in natural popula-
tions of Daphnia (Crustacea: Cladocera). Journal of
Animal Ecology 68:134149.
Mideo, N. 2009. Parasite adaptations to within-host
competition. Trends in Parasitology 25:261268.
Mironov, S. V., A. V. Bochkov, and A. Fain. 2005. Phy-
logeny and evolution of parasitism in feather mites
of the families Epidermoptidae and Dermationidae
(Acari: Analgoidea). Zoologischer Anzeiger-A Jour-
nal of Comparative Zoology 243:155179.
Mironov, S. V., K. Skirnisson, S. T. Thorarinsdottir, and
O. K. Nielsen. 2010. Feather mites (Astigmata:
Psoroptidia) parasitising the rock ptarmigan Lago-
pus muta (Montin)(Aves: Galliformes) in Iceland.
Systematic Parasitology 75:187206.
Møller, A., and M. D. Jennions. 2002. How much vari-
ance can be explained by ecologists and evolution-
ary biologists? Oecologia 132:492500.
Morrill, A., F. Dargent, and M. R. Forbes. 2017.
Explaining parasite aggregation: more than one
parasite species at a time. International Journal for
Parasitology 47:185188.
Nakagawa, S., P. C. Johnson, and H. Schielzeth. 2017.
The coefcient of determination R
2
and intra-class
correlation coefcient from generalized linear
mixed-effects models revisited and expanded. Jour-
nal of the Royal Society Interface 14:20170213.
Nielsen,
O. K., A. Morrill, K. Sk
ırnisson, U. Stenkewitz,
G. R. P
alsd
ottir, and M. R. Forbes. 2020. Host sex
and age typically explain variation in parasitism of
Rock Ptarmigan: implications for identifying deter-
minants of exposure and susceptibility. Journal of
Avian Biology 51:e02472.
Ovaskainen, O., G. Tikhonov, A. Norberg, F. Guil-
laume Blanchet, L. Duan, D. Dunson, T. Roslin,
and N. Abrego. 2017. How to make more out of
community data? A conceptual framework and its
implementation as models and software. Ecology
Letters 20:561576.
Paquette, C., D. Garant, J. Savage, D. R
eale, and P.
Bergeron. 2020. Individual and environmental
determinants of Cuterebra bot y parasitism in the
eastern chipmunk (Tamias striatus). Oecologia
193:359370.
Poulin, R. 1993. The disparity between observed and
uniform distributions: a new look at parasite
aggregation. International Journal for Parasitology
23:937944.
Poulin, R. 2001. Interactions between species and the
structure of helminth communities. Parasitology
122:S3S11.
R Core Team. 2019. R: a language and environment for
statistical computing. R Foundation for Statistical
Computing, Vienna, Austria.
Ray-Mukherjee, J., K. Nimon, S. Mukherjee, D. W.
Morris, R. Slotow, and M. Hamer. 2014. Using com-
monality analysis in multiple regressions: a tool to
decompose regression effects in the face of multi-
collinearity. Methods in Ecology and Evolution
5:320328.
Read, A. F. 1988. Sexual selection and the role of para-
sites. Trends in Ecology & Evolution 3:97102.
Reiczigel, J., M. Marozzi, I. F
abi
an, and L. R
ozsa. 2019.
Biostatistics for parasitologists a primer to quan-
titative parasitology. Trends in Parasitology
35:277281.
Rigaud, T., M.-J. Perrot-Minnot, and M. J. Brown. 2010.
Parasite and host assemblages: embracing the reality
will improve our knowledge of parasite transmission
and virulence. Proceedings of the Royal Society B:
Biological Sciences 277:36933702.
R
ozsa, L., J. Reiczigel, and G. Majoros. 2000. Quantify-
ing parasites in samples of hosts. Journal of Para-
sitology 86:228232.
Schalk, G., and M. R. Forbes. 1997. Male biases in para-
sitism of mammals: effects of study type, host age,
and parasite taxon. Oikos 78:6774.
Sk
ırnisson, K., and
O. K. Nielsen. 2019. Quill mite
infestation of rock ptarmigan Lagopus muta (Aves:
Phasianidae) in relation to year and host age, sex,
body condition, and density. Parasitology Research
118:26432650.
Skirnisson, K., S. T. Thorarinsdottir, and O. K. Nielsen.
2012. The parasite fauna of rock ptarmigan (tex-
titLagopus muta) in Iceland: prevalence, intensity,
and distribution within the host population. Com-
parative Parasitology 79:4455.
Sousa, W. P. 1993. Interspecic antagonism and
species coexistence in a diverse guild of larval
trematode parasites. Ecological Monographs
63:103128.
vwww.esajournals.org 19 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
Stenkewitz, U. 2017. Parasites and population change
of rock ptarmigan in Iceland. PhD Thesis. Reykja-
vik, Iceland: University of Iceland, Faculty of Life,
Environmental Sciences.
Stenkewitz, U.,
O. K. Nielsen, K. Skirnisson, and G.
Stef
ansson. 2016. Host-parasite interactions and
population dynamics of rock ptarmigan. PLOS
ONE 11:e0165293.
Telfer, S., X. Lambin, R. Birtles, P. Beldomenico, S. Bur-
the, S. Paterson, and M. Begon. 2010. Species inter-
actions in a parasite community drive infection
risk in a wildlife population. Science 330:243246.
Tikhonov, G., O. Ovaskainen, J. Oksanen, M. de Jonge,
O. Opedal, and T. Dallas. 2020. Hmsc: hierarchical
model of species communities. R package version
3.0-6. https://cran.r-project.org/package=Hmsc
Veitch, J. S., J. Bowman, and A. I. Schulte-Hostedde.
2020. Parasite species co-occurrence patterns on
Peromyscus: joint species distribution modelling.
International Journal for Parasitology: Parasites
and Wildlife 12:199.
Wu,Q.,M.Richard,A.Rutschmann,D.B.Miles,andJ.
Clobert. 2019. Environmental variation mediates the
prevalence and co-occurrence of parasites in the com-
mon lizard, Zootoca vivipara. BMC Ecology 19:44.
Z
el
e, F., S. Magalh~
aes, S. K
e, and A. B. Duncan. 2018.
Ecology and evolution of facilitation among sym-
bionts. Nature Communications 9:112.
SUPPORTING INFORMATION
Additional Supporting Information may be found online at: http://onlinelibrary.wiley.com/doi/10.1002/ecs2.
3709/full
vwww.esajournals.org 20 August 2021 vVolume 12(8) vArticle e03709
DISEASE ECOLOGY MORRILL ET AL.
... Icelandic Rock Ptarmigan (Lagopus muta, hereafter ptarmigan) are an ideal host species to study parasite aggregation because they have been subject to intense research as a game bird and have provided large numbers of replicates for ecological studies (Morrill et al., 2021;Nielsen et al., 2020;Stenkewitz et al., 2016); additionally, their parasite fauna are exceptionally well known and can be sampled with standardized protocols (Skírnisson, Thorarinsdottir & Nielsen, 2012;Stenkewitz et al., 2016). Parasites infecting ptarmigan from a focal population in northeast Iceland comprise a diverse set of ecto-and endoparasites, for which prevalence of several species correlates with host health indices and/or host population densities (Stenkewitz et al., 2016). ...
... Assuming the ecto-/endoparasite dichotomy with respect to relationships between aggregation and mean abundance is real, we should consider broad differences in traits between these groups of parasites other than the leading explanations of differences in niche capacity and density-dependent transmission. Transmission of ectoparasites between hosts is expected particularly during brood-rearing and crêche formation (Nielsen et al., 2020) and is expected to be mediated by the louse fly for certain parasites (Morrill et al., 2021). One broad difference between parasite types concerns variation in factors affecting the viability or availability of infective stages. ...
... In the present study, for example, we could not look at the relationship between host age or sex and aggregation for each parasite species, because such stratification would result in different numbers of samples across models (one sample per species per year, vs. two or four if one or both host-level variables were also considered). We could not address whether any potential associations between co-infecting parasites explained patterns of aggregation, as such modeling would necessarily apply exclusively to individual host-level data; however, we do not expect co-infection to impact aggregation in this system as previous research demonstrated that any associations between these parasites explained on average only 2.01% of the observed variation in abundance, and never more than 5.13% (Morrill et al., 2021). To answer those questions regarding host-level predictors of aggregation or co-infection, an alternative modeling approach with a multivariate response could be used that considers parasite abundances for each species as following, for example, negative binomial (Poisson-Gamma mixture) distributions, that then additionally estimates the distributions' dispersion parameters. ...
Article
Full-text available
Aggregation of macroparasites among hosts is a near-universal pattern, and has important consequences for the stability of host-parasite associations and the impacts of disease. Identifying which potential drivers are contributing to levels of aggregation observed in parasite-host associations is challenging, particularly for observational studies. We apply beta regressions in a Bayesian framework to determine predictors of aggregation, quantified using Poulin’s index of discrepancy ( D ), for 13 species of parasites infecting Icelandic Rock Ptarmigan ( Lagopus muta ) collected over 12 years. 1,140 ptarmigan were collected using sampling protocols maximizing consistency of sample sizes and of composition of host ages and sexes represented across years from 2006–2017. Parasite species, taxonomic group (insect, mite, coccidian, or nematode), and whether the parasite was an ecto- or endoparasite were tested as predictors of aggregation, either alone or by modulating an effect of parasite mean abundance on D . Parasite species was an important predictor of aggregation in models. Despite variation in D across samples and years, relatively consistent aggregation was demonstrated for each specific host-parasite association, but not for broader taxonomic groups, after taking sample mean abundance into account. Furthermore, sample mean abundance was consistently and inversely related to aggregation among the nine ectoparasites, however no relationship between mean abundance and aggregation was observed among the four endoparasites. We discuss sources of variation in observed aggregation, sources both statistical and biological in nature, and show that aggregation is predictable, and distinguishable, among infecting species. We propose explanations for observed patterns and call for the review and re-analysis of parasite and other symbiont distributions using beta regression to identify important drivers of aggregation—both broad and association-specific.
Book
Full-text available
Tilraunastöð Háskóla Íslands í meinafræði að Keldum er leiðandi stofnun í rannsóknum á sjúkdómum og sjúkdómavörnum dýra. Rannsakaðir eru sjúkdómar í flestum spendýrategundum Íslands og allmörgum fugla-, fisk- og lindýrategundum. Tilgangur rannsóknanna er að efla skilning á eðli sjúkdóma og skapa nýja þekkingu. Tilraunastöðin sinnir þjónustu og ráðgjöf á sviði dýrasjúkdóma fyrir hagsmunaaðila í samstarfi við Matvælastofnun. Heilbrigð dýr eru forsenda arðvænlegs landbúnaðar og fiskeldis. Mikilvægt er að dýr beri ekki sjúkdóma í menn með matvælum eða öðrum hætti. Stofnunin þarf að geta brugðist sem skjótast við nýjum og aðkallandi vandamálum á sviði sjúkdómagreininga. Viðbúnaður þarf að vera til staðar vegna vandamála sem upp kunna að koma með tilheyrandi mannauði og innviðum. Tilraunastöðin tengist læknadeild Háskóla Íslands og hefur sjálfstæða stjórn og fjárhag. Starfseminni er skipt í þrjár fagdeildir auk stoðdeildar; 1) veiru- og sameindalíffræðideild, 2) bakteríu-, sníkjudýra- og meinafræðideild og 3) rannsóknadeild fisksjúkdóma.
Article
Full-text available
Understanding the interactions between parasites, hosts, and their shared environment is central to ecology. Variation in infestation prevalence may be the result of varying environmental and population characteristics; however, variations in parasitism may also depend on individual characteristics that influence both the exposure and susceptibility to parasites. Using 12 years of data from a population of wild eastern chipmunks relying on pulsed food resources, we investigated the determinants of bot fly parasitism at both the population and individual level. We assessed the relationship between infestation prevalence and weather conditions, population size and food abundance. Then, we assessed the relationship between infestation intensity and chipmunk behavior, sex, age, body mass and food abundance. Precipitation, temperature and population size were positively related to infestation prevalence, while beech masts were negatively related to infestation prevalence, highlighting the importance of local environmental conditions on hosts and parasites. We also found that the influence of activity and exploration on infestation intensity varied according to sex in adults. More active and faster exploring males had more parasites compared to females, suggesting that reproductive behaviors may influence parasite exposure. For juveniles, infestation intensity was greater when juveniles emerged in the spring as opposed to fall, possibly because spring emergence is synchronized with the peak of bot fly eggs in the environment, low food availability and longer activity period. Our results suggest that the environmental, population and host characteristics that are advantageous for reproduction and resource acquisition may come at the cost of increasing parasitism.
Article
Full-text available
Hosts are often infested by multiple parasite species, but it is often unclear whether patterns of parasite co-occurrence are driven by parasite habitat requirements or parasite species interactions. Using data on infestation patterns of ectoparasitic arthropods (fleas, trombiculid mites, cuterebrid botflies) from deer mice (Peromyscus maniculatus), we analyzed species associations using joint species distribution modelling. We also experimentally removed a flea (Orchopeas leucopus) ectoparasite from a subset of deer mice to examine the effect on other common ectoparasite species. We found that mite (Neotrombicula microti) and botfly (Cuterebra sp.) had a negative relationship that is likely a true biotic species interaction. The flea had a negative association with the mite and a positive association with the botfly species, both of which appeared to be influenced by host traits or parasite life-history traits. Furthermore, experimental removal of the flea did not have a significant effect on ectoparasite prevalence of another species. Overall, these findings suggest that complex parasite species associations can be present among multiple parasite taxa, and that aggregation is not always the rule for ectoparasite communities of small mammals.
Article
Full-text available
There is a rich amount of information in co‐occurrence (presence–absence) data that could be used to understand community assembly. This proposition first envisioned by Forbes (1907) and then Diamond (1975) prompted the development of numerous modelling approaches (e.g. null model analysis, co‐occurrence networks and, more recently, joint species distribution models). Both theory and experimental evidence support the idea that ecological interactions may affect co‐occurrence, but it remains unclear to what extent the signal of interaction can be captured in observational data. It is now time to step back from the statistical developments and critically assess whether co‐occurrence data are really a proxy for ecological interactions. In this paper, we present a series of arguments based on probability, sampling, food web and coexistence theories supporting that significant spatial associations between species (or lack thereof) is a poor proxy for ecological interactions. We discuss appropriate interpretations of co‐occurrence, along with potential avenues to extract as much information as possible from such data. We present seven arguments highlighting why co‐occurrence data is not, and should never be considered, evidence ecological interaction. These arguments are based on probability, sampling, food web and coexistence theories. We discuss appropriate interpretations of co‐occurrence, along with potential avenues to extract as much information as possible from co‐occurrence data. In addition, we present ideas to better describe, understand and predict ecological interactions.
Article
Full-text available
Background: Hosts and their parasites are under reciprocal selection, leading to coevolution. However, parasites depend not only on a host, but also on the host's environment. In addition, a single host species is rarely infested by a single species of parasite and often supports multiple species (i.e., multi-infestation). Although the arms race between a parasite and its host has been well studied, few data are available on how environmental conditions may influence the process leading to multiple infestations. In this study, we examine whether: (1) environmental factors including altitude, temperature, vegetation cover, human disturbance, and grazing by livestock affect the prevalence of two types of ectoparasites, mites and ticks, on their host (the common lizard, Zootoca vivipara) and (2) competition is evident between mites and ticks. Results: We found the probability of mite infestation increased with altitude and vegetation cover, but decreased with human disturbance and presence of livestock. In contrast, the probability of tick infestation was inversely associated with the same factors. Individuals with low body condition and males had higher mite loads. However, this pattern was not evident for tick loads. The results from a structural equation model revealed that mites and ticks indirectly and negatively affected each other's infestation probability through an interaction involving the environmental context. We detected a direct negative association between mites and ticks only when considering estimates of parasite load. This suggests that both mites and ticks could attach to the same host, but once they start to accumulate, only one of them takes advantage. Conclusion: The environment of hosts has a strong effect on infestation probabilities and parasite loads of mites and ticks. Autecological differences between mites and ticks, as indicated by their opposing patterns along environmental gradients, may explain the pattern of weak contemporary interspecific competition. Our findings emphasize the importance of including environmental factors and the natural history of each parasite species in studies of host-parasite coevolution.
Article
Full-text available
We examine coinfection between helminth species and the bacterium Leptospira interrogans in a natural population of the Norway rat, Rattus norvegicus. We ask whether coinfection can influence the probability and intensity of infection of these enzootic and zoonotic parasites in urban rats, which may affect the loads of parasites excreted into the environment. Rodent sampling was carried out during two seasons in 2014 in a Brazilian urban slum. We sampled rats’ feces, kidney imprints, and urine to identify and quantify helminth eggs/larvae and infection by L. interrogans. Eleven species/groups of helminths and L. interrogans were identified among 299 captured rats. Simple correlation tests and generalized linear models were performed to identify general patterns of coinfection and potential direction of effects, respectively, after controlling for consolidated environmental and host biotic variables. Significant associations were illustrated in an interaction network. Focusing on parasites with the potential to cause zoonoses among humans, we observed that coinfection between L. interrogans and the nematode Angiostrongylus cantonensis was significantly more frequent than expected. Reduced prevalence of A. cantonensis was found in the presence of Nippostrongylus brasiliensis, and N. brasiliensis intensities (eggs per gram of feces) were increased with greater intensities of both L. interrogans and Strongyloides sp., the latter of which was, in turn, found to increase the intensities of A. cantonensis. A higher probability of finding L. interrogans was associated with infection by Strongyloides sp. Our study provides a novel perspective on evaluating helminth coinfection profiles in populations of naturally infected urban rats, moving beyond previous studies which have been limited to descriptions of co‐occurrence. Noticeably, infection risk was dependent on coinfection and this should be accounted for when targeting the control of zoonotic pathogens in natural populations.
Article
Full-text available
A total of 1209 ptarmigan were examined for Mironovia lagopus, including 721 juvenile birds (ca. 3 months old) and 488 adult birds (15 months or older). A total of 88 birds or 7.3% (n = 1209, 95% cl 5.9–8.9%) were infested with M. lagopus. There was an age difference in prevalence of infection, and more adults (10.7%, n = 488, 95% cl 8.2–13.7%) than juveniles (5.0%, n = 721, 95% cl 3.6–6.8%) were infested. There was a significant age effect in the mean intensity index, and adult birds had more advanced infestations compared with juvenile birds. There were no significant changes in either the interannual prevalence of infection or the mean intensity index of infection. Of the feather types inspected, there was no age-related difference in selection of feathers, nor was there any preference of mites for any one of the inspected feather types. Body dispersants were all adult females. There was an age-related difference in mean intensities of infection of body dispersants; it was higher in adult birds. The methods used to determine presence or absence of M. lagopus were not 100% accurate especially for juvenile hosts, and this at least partly explains the difference in prevalence among age groups. There was no relationship between host body condition or host density and infection by M. lagopus.
Article
Full-text available
The aggregated distributions of host–parasite systems require several different infection parameters to characterize them. We advise readers how to choose infection indices with clear and distinct biological interpretations, and recommend statistical tests to compare them across samples. A user-friendly and free software is available online to overcome technical difficulties.
Article
1. We explored patterns of co‐occurrence of ectoparasite species on individual hosts from Central Europe (Slovakia) and South America (Argentina) within and between higher taxa while controlling for confounding factors associated with variation between host individuals and host species, as well as spatial and temporal variation. We used a recently developed statistical approach, the hierarchical modelling of species communities. 2. Although the probability of pairwise associations of the majority of species in both regions did not differ from random, all significant species associations in Slovakia were positive, whereas the only three significant species associations in Argentina were negative. On average, associations between ectoparasite species belonging to different higher taxa were positive in Slovakia and negative in Argentina. 3. At the host species level, both positive and negative associations between species were detected in all higher taxa. This was also true for ectoparasite co‐occurrences in the same site, habitat, or year; although the number of co‐occurring species pairs with high posterior probability was much greater in Slovakia than in Argentina. 4. We conclude that consideration of species co‐occurrences across the entire multi‐host assemblage and control for confounding spatial and temporal factors provided important insights into parasite community structure.
Article
Measures of parasitism often differ between hosts. This variation is thought due in part to age or sex differences in exposure to parasites and/or susceptibility to parasitism. We assessed how often age or sex biases in parasitism were found using a large, multi‐year (2006 – 2017) dataset of 12 parasite species of Icelandic Rock Ptarmigan (Lagopus muta) . We found host traits (i.e. age and/or sex) accounted for significant variation in abundance of 11 of the 12 parasite species. We often found increased abundance among juvenile hosts, although significant adult biases were observed for three parasite species. Additionally, higher levels of parasitism by many species were observed for female hosts, contrary to frequent male biases in parasitism reported for other vertebrates. Abundance of six parasite species was best explained by interactions between host age and sex; some degree of decrease in abundance with host age was present for both male and female hosts for four of those parasite species. We consider various host and parasite traits that could account for observed singular and repeated patterns of age and/or sex biases in parasitism (e.g. age‐ and sex‐related grouping behaviours, age‐specific mortality in relation to parasitism, acquisition of greater immunity with age). This work provides a foundation for future studies investigating age‐related differences in acquired immunity and age‐specific parasite‐mediated mortality for males and females, as well as studies on interactions between co‐infecting parasite species. This article is protected by copyright. All rights reserved.
Article
Understanding the role of biotic interactions in shaping natural communities is a long-standing challenge in ecology. It is particularly pertinent to parasite communities sharing the same host communities and individuals, as the interactions among parasites-both competition and facilitation-may have far-reaching implications for parasite transmission and evolution. Aggregated parasite burdens may suggest that infected host individuals are either more prone to infection, or that infection by a parasite species facilitates another, leading to a positive parasite-parasite interaction. However, parasite species may also compete for host resources, leading to the prediction that parasite-parasite associations would be generally negative, especially when parasite species infect the same host tissue, competing for both resources and space. We examine the presence and strength of parasite associations using hierarchical joint species distribution models fitted to data on resident parasite communities sampled on over 1300 small mammal individuals across 22 species and their resident parasite communities. On average, we detected more positive associations between infecting parasite species than negative, with the most negative associations occurring when two parasite species infected the same host tissue, suggesting that parasite species associations may be quantifiable from observational data. Overall, our findings suggest that parasite community prediction at the level of the individual host is possible, and that parasite species associations may be detectable in complex multi-species communities, generating many hypotheses concerning the effect of host community changes on parasite community composition, parasite competition within infected hosts, and the drivers of parasite community assembly and structure.