Acinetobacter species in the skin microbiota protect
against allergic sensitization and inflammation
Nanna Fyhrquist, PhD,a Lasse Ruokolainen, PhD,b Alina Suomalainen, BSc,a Sari Lehtima
Johanna Vendelin, PhD, Piia Karisola, PhD, Maili Lehto, PhD, Terhi Savinko, PhD, Hanna Jarva, MD,e
a a a d
Timo U. Kosunen, MD,e Jukka Corander, PhD,f Petri Auvinen, PhD,d Lars Paulin, PhD,d Leena von Hertzen, PhD,g
Tiina Laatikainen, PhD,h,i Mika Ma €kela€, MD,g Tari Haahtela, MD,g Dario Greco, PhD,a Ilkka Hanski, PhD,b and
a
Harri Alenius, PhD Helsinki, Turku, and Kuopio, Finland
Background: The human commensal microbiota interacts in a
complex manner with the immune system, and the outcome of
these interactions might depend on the immune status of the
subject.
Objective: Previous studies have suggested a strong
allergy-protective effect for Gammaproteobacteria. Here we
analyze the skin microbiota, allergic sensitization (atopy), and
immune function in a cohort of adolescents, as well as the
influence of Acinetobacter species on immune responses in vitro
and in vivo.
Methods: The skin microbiota of the study subjects was
identified by using 16S rRNA sequencing. PBMCs were
analyzed for baseline and allergen-stimulated mRNA
expression. In in vitro assays human monocyte-derived dendritic
cells and primary keratinocytes were incubated with
Acinetobacter lwoffii. Finally, in in vivo experiments mice were
injected intradermally with A lwoffii during the sensitization
phase of the asthma protocol, followed by readout of
inflammatory parameters.
From athe Unit of Systems Toxicology, Finnish Institute of Occupational Health,
Helsinki; bthe Department of Biosciences and dthe Institute of Biotechnology,
University of Helsinki; cthe Molecular Immunology Group, Turku Centre for
Biotechnology; ethe Haartman Institute, Department of Bacteriology and Immunology
and Research Programs Unit, Immunobiology, University of Helsinki, and Helsinki
University Central Hospital Laboratory (HUSLAB); fthe Department of Mathematics
and Statistics, University of Helsinki; gthe Allergy Department, Skin and Allergy
Hospital, Helsinki University Hospital; hthe Department of Chronic Disease
Prevention, National Institute for Health and Welfare, Helsinki; and ithe Institute of
Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio.
The research leading to these results has received funding from the Academy of Finland
(grants nos. 138695, 255350, 121025, and 251170), the European Research Council
(grant no. 239784), the Jane and Aatos Erkko Foundation, Helsinki University Hospi-
tal (HUS) (grant no. 8361), and by the European Union’s Seventh Framework Pro-
gramme FP7/2007-2013 under grant agreements 261366 (MAARS) and 261357
(MeDALL).
Disclosure of potential conflict of interest: This study was funded by the Academy of
Finland (grants nos. 138695, 255350, 121025, and 251170), the European Research
Council (grant no. 239784), the Jane and Aatos Erkko Foundation, the European
Union’s Seventh Framework Programme (grant agreements 261366 and 261357),
and the Helsinki University Hospital (grant no. 8361). T. Haahtela is on the board of
the Global Initiative for Asthma, and has received payment for delivering lectures
from OrionPharma and Merck. The rest of the authors declare that they have no rele-
vant conflicts of interest.
Received for publication April 28, 2014; revised July 1, 2014; accepted for publication
July 7, 2014.
Available online September 26, 2014.
Corresponding author: Harri Alenius, PhD, Unit of Systems Toxicology, Finnish Institute
of Occupational Health, Topeliuksenkatu 41 b, Helsinki-FIN 00250, Finland. E-mail:
harri.alenius@ttl.fi.
0091-6749/$36.00
Ó 2014 American Academy of Allergy, Asthma & Immunology
http://dx.doi.org/10.1016/j.jaci.2014.07.059
€ki, PhD,c Ville Veckman, PhD,a
Results: In healthy subjects, but not in atopic ones, the relative
abundance of Acinetobacter species was associated with the
expression of anti-inflammatory molecules by PBMCs.
Moreover, healthy subjects exhibited a robust balance between
anti-inflammatory and TH1/TH2 gene expression, which was
related to the composition of the skin microbiota. In cell assays
and in a mouse model, Acinetobacter species induced strong TH1
and anti-inflammatory responses by immune cells and skin cells
and protected against allergic sensitization and lung
inflammation through the skin.
Conclusion: These results support the hypothesis that skin
commensals play an important role in tuning the balance of
TH1, TH2, and anti-inflammatory responses to environmental
allergens. (J Allergy Clin Immunol 2014;134:1301-9.)
Key words: Atopy, Gammaproteobacteria, Acinetobacter species,
PBMC, anti-inflammatory gene expression, dendritic cells, keratino-
cytes, mouse asthma model
The incidence of atopic disorders has increased steadily in
developed countries for several decades,1 now affecting approxi-
mately 40% of children in the United Kingdom.2 This epidemic
has been related to changes in lifestyle.3 The ‘‘old driends’’4-6 and
biodiversity3 hypotheses postulate that the increase in chronic inflam-
matory disorders is caused by reduced exposure to environmental mi-
crobes, which in turn influences the composition of the human
commensal microbiota and its interactions with the immune system.
Microbes and vertebrates have coevolved over millennia,4,7 and the
sudden altered microbial contact in urban societies in developed
countries8 might lead to dysregulation of host immunity. Revealingly,
children growing up on traditional farms have an especially low risk
of allergic sensitization (atopy), with the protective phenotype sus-
tained into adult life9; a farm environment exposes children to micro-
bial pressure that has few equivalents in the developed world. In a
study in eastern Finland, land use in the surroundings of children’s
homes was significantly associated with the prevalence of atopy,
with children living in homes surrounded by much forest and agricul-
tural land showing less atopy.10 The causal factor might be microbial
exposure because the generic diversity of Proteobacteria on the skin
was significantly associated with environmental land use.10
Microorganisms inhabiting mammalian body surfaces have a
highly coevolved relationship with the host’s immune system.
Although the immune system is essential in maintaining homeo-
stasis with resident microbes, the latter shape immunity by inducing
protective and regulatory responses.11 The molecular and cellular
pathways that sense and transduce signals leading to protection
are still largely unknown, but they are likely to primarily target reg-
ulatory immune processes. Thus bacteria in the intestine can
1301
1302 FYHRQUIST ET AL
Abbreviations used
ACTB: b-Actin gene
AHR: Aryl hydrocarbon receptor gene
DC: Dendritic cell
FOXP3: Forkhead box P3 gene
GAPDH: Glyceraldeyde-3-phosphate dehydrogenase gene
moDC: Monocyte-derived dendritic cell
OVA: Ovalbumin
PE: Phycoerythrin
qPCR: Quantitative PCR
TLR2: Toll-like receptor 2
Treg: Regulatory T
promote the activity of regulatory T (Treg) cells by inducing IL-
10 production. For instance, Bacteroides fragilis causes CD41 T
cells to secrete IL-10 through the action of polysaccharide A on
Toll-like receptor 2 (TLR2),12 and a mixture of Clostridium
strains promotes intestinal Treg cell activity, possibly through
the induction of TGF-b by the production of short-chain fatty
acids.13 In germ-free mice IL-10 expression is markedly reduced
and Treg cells in the colon are less abundant in comparison with
those seen in normal mice.14 Microbes can also influence the im-
mune system and protect against allergies through induction of
TH1-type immune responses, which inhibit the development of
TH2 cells. Endotoxins (constituents of the outer membrane of
gram-negative bacteria) stimulate macrophages and antigen-
presenting cells to produce IL-12, which triggers the development
of TH1 immune responses. Microbe-induced immune program-
ming might further involve epigenetic modifications at immune-
related genes. A recent study using a mouse model demonstrated
modified histone acetylation of the IFN-g promoter of CD41 T
cells in the offspring of A lwoffii–exposed mothers.15 The effect
of microbes at other sites than the gut has been less extensively
investigated, but there is evidence that skin commensals autono-
mously control local inflammatory responses16 and intranasal de-
livery of microbial material provides significant protection from
experimental allergy.17,18
Here we analyze associations between the relative abundances
of bacterial genera on the skin and expression of selected genes
coding for proinflammatory and anti-inflammatory molecules in a
cohort of adolescents. We report significantly dissimilar patterns of
association between the microbes and gene expression in PBMCs
of healthy versus atopic subjects, suggesting that the interactions
between the immune system and the microbiota are significantly
altered by atopy (allergic sensitization). The analysis highlights a
robust positive association between the genus Acinetobacter (Gam-
maproteobacteria) and anti-inflammatory molecules in healthy but
not atopic subjects. Having identified the special role for Acineto-
bacter species among some hundreds of bacterial genera, we inves-
tigated the effect of heat-inactivated A lwoffii on immune responses
and report strong induction of anti-inflammatory gene expression
in cell assays and protection against allergic sensitization and
lung inflammation after exposure to microbial material through
the skin in a mouse model.
METHODS
Study subjects and atopic sensitization
The study subjects (n 5 118) had been previously selected for a long-term
allergy study.19 All subjects provided written informed consent, and
J ALLERGY CLIN IMMUNOL
DECEMBER 2014
institutional ethics committees approved the protocol. Atopy was defined as
a specific IgE level to inhalant allergens of greater than 2.5 kUA/L. On the basis
of the previous study,10 the cutoff value was located between 2 clear humps in
the log-transformed distribution of IgE values for this cohort.
Skin microbiota
The skin microbiota was analyzed by using 16s rRNA sequencing, as
described previously.10
PBMCs
PBMCs were separated from whole blood and stimulated with allergens for
6 or 24 hours in complete RPMI.
Microbes for in vitro and in vivo experiments
A lwoffii strains from blood culture isolates were identified by means if 16S
sequencing or API 20NE strips (BioMerieux, Lyon, France), grown on
chocolate agar plates, collected, and heat inactivated.
In vitro cell assays
Monocytes were isolated from healthy donor buffy coats and differentiated
into monocyte-derived dendritic cells (moDCs). The moDCs and human
epidermal keratinocytes were stimulated for 6 hours with heat-killed A lwoffii
(at 1:5 cell/bacteria ratio).
Real-time quantitative PCR and Luminex
RNAwas extracted from PBMCs, moDCs, keratinocytes, and mouse tissue;
reverse transcribed into cDNA; and analyzed by using real-time quantitative
PCR (qPCR). As endogenous controls, we used 18S rRNA, b-actin gene
(ACTB), and glyceraldeyde-3-phosphate dehydrogenase gene (GAPDH) for
human targets and 18S and TATA-binding protein for mouse targets. The
results are expressed as relative quantity, which was calculated by using
the comparative cycle threshold method, according to the manufacturer’s
instructions. Supernatants from the PBMC cultures were analyzed by using
Luminex with Bio-Rad multiplex assays (Bio-Rad Laboratories, Hercules,
Calif).
Animal model
All animal experiments were approved by the Social and Health Care
Department of the State Provincial Office of Southern Finland. Mice were
shaved on the back, tape stripped 3 times, and injected intradermally with
ovalbumin (OVA) and heat-killed A lwoffii, followed by intranasal OVA chal-
lenge and collection of samples. Lung, skin, and blood samples were prepared
for histologic analysis, RNA extraction, and measurement of IgE serum levels.
Statistical analysis
The association between and effect of gene expression, the relative
abundance of Acinetobacter species on the skin, and atopy were examined
by using Pearson correlation and analysis of covariance, respectively. Global
associations of gene expression with the microbiota were inferred by using
network analysis.
For a detailed description of the methods, see the Methods section in this
article’s Online repository at www.jacionline.org.
RESULTS
Associations between the skin microbiota and
PBMC gene expression
The skin microbiota was identified to the genus level by
sequencing the 16S rRNA gene from DNA samples obtained
from the volar surface of the forearm. Altogether, 1017 bacterial
genera were identified in the 118 study subjects. The PBMCs
J ALLERGY CLIN IMMUNOL
VOLUME 134, NUMBER 6
FIG 1. Network analysis of the relative abundance of skin microbial genera and PBMC gene expression in
healthy (A) and atopic (B) subjects. Th1, T helper type 1 (blue); Th2, T helper type 2 (red); Th17, T helper type
17 (yellow); Treg, T regulatory/immunoregulatory (green) genes. Red and blue edges indicate positive and
negative correlations, respectively, and the color hue indicates the strength of the correlation. Sample size is
46 healthy and 28 atopic subjects.
were analyzed for baseline and allergen-stimulated mRNA
expression of inflammatory and regulatory genes. We used
network-inferring algorithms to highlight patterns of association
between the relative abundances of the microbial genera and
gene expression. Microbial genera that were present in at least
75% of the study subjects were included in the analysis (see
Table E1 in this article’s Online repository at www.jacionline.
org), and those without direct edges to genes were removed
from the final networks.
The distribution of edges between microbes and gene
expression in unstimulated 24-hour cultures of PBMCs deviated
significantly between healthy and atopic subjects (defined by
increased specific IgE levels to inhalant allergens). In healthy
subjects the most significant direct links were between
Acinetobacter species and IL-10 and between Diaphorobacter
species (Proteobacteria) and TLR2, which is a strong inducer
of IL-10 (Fig 1, A). IL-10 further bridged into a network of
anti-inflammatory transcripts, including forkhead box P3
(FOXP3), TGFB, TLR2, and aryl hydrocarbon receptor gene
(AHR; Fig 1, A). In the network for atopic subjects, Microbacte-
rium and Alcaligenes species correlated negatively with the
expression of IL13 and IL17, respectively (Fig 1, B). The
positive correlation between Acinetobacter species and IL-10
in healthy subjects and the negative, although not significant,
association in atopic subjects were similar at the protein level,
as measured from the supernatants of the PBMCs (Fig 2, A).
The protein concentrations corresponded closely to the level of
RNA expression (see Fig E1 in this article’s Online repository
at www.jacionline.org).
Association of anti-inflammatory responses with
Acinetobacter species
The network inference was carried out stringently. To
facilitate interpretation, we highlight here the simple correla-
tions behind the network results for the biologically most
interesting relationships, considering data for all stimuli (Bet v
1, Phl p 1, and anti-human CD3/CD28) and time points (6 and
FYHRQUIST ET AL 1303
24 hours) used in the PBMC cultures. In healthy subjects Aci-
netobacter species influenced most strongly the expression of
IL10 (r 5 0.51, P 5 .0004), FOXP3 (r 5 0.42, P 5 .004),
and AHR (r 5 0.35, P 5 .018; Fig 2, B, and see Fig E2 in
this article’s Online Repository at www.jacionline.org) in 24-
hour unstimulated PBMCs, as well as TGFB (r 5 0.40, P 5
.0053; Fig 2, B) and FOXP3 (r 5 0.30, P 5 .041) in Bet v
1–stimulated PBMCs. No such correlations were observed in
atopic subjects (see Table E2 in this article’s Online Reposi-
tory at www.jacionline.org).
We used analysis of covariance to explain gene expression
by diagnosis (healthy vs atopic), the relative abundance of
Acinetobacter species, and their interaction. The interaction
term between Acinetobacter species and atopy was significant
for IL10 (24-hour unstimulated cultures, P 5 .005), FOXP3
(Phl p 1– and Bet v 1–stimulated and unstimulated cultures:
P 5 .029, P 5 .031, and P 5 .042, respectively), TGFB (Bet v
1–stimulated 6-hour cultures, P 5 .031), and IL1B (6-hour
unstimulated cultures, P 5.043), indicating opposite associations
with Acinetobacter species in healthy and atopic subjects (see
Table E3 in this article’s Online Repository at www.jacionline.
org). As expected, the expression of IL4 (Bet v 1, P < .0001;
Phl p 1, P 5 .006) and IL13 (Bet v 1, P 5 .0002) in allergen-
stimulated PBMCs is significantly different between atopic and
healthy subjects, and atopy had a weak effect on the expression
of IL10 and TGFB in unstimulated 24-hour PBMC cultures
(P 5 .024 and P 5 .025, respectively; see Table E3).
To find out the cellular source of the above genes, we isolated
CD31 T cells and CD141 monocytes from PBMCs with magnetic
beads, followed by RNA extraction from the different cell subsets.
The magnetic bead–enriched CD141 monocytes expressed high
levels of IL10 and, to some extent, TLR2 and AHR, whereas
CD31 T cells contributed a small part of TGFB expression and
most of the FOXP3 expression (see Fig E3 in this article’s Online
Repository at www.jacionline.org). Given the high expression of
TLR2, AHR, and TGFB in PBMCs, the analysis likely misses
additional cellular sources, such as B cells, natural killer cells,
and dendritic cells (DCs).
1304 FYHRQUIST ET AL J ALLERGY CLIN IMMUNOL
DECEMBER 2014

FIG 2. The expression of IL10 RNA (***P 5 .0004) and IL-10 protein (P 5 .079; A) and FOXP3 RNA (**P 5 .004;
B, left panel) in 24-hour unstimulated PBMC cultures and TGFB RNA (**P 5 .0053; B, right panel) in 6-hour Bet
v 1–stimulated PBMC cultures correlated with the relative abundance of Acinetobacter species on the skin in
healthy (open symbols) but not atopic (solid symbols) subjects. Sample size is 46 healthy and 28 atopic sub-
jects. RQ, Relative quantity.

Balanced proinflammatory and anti-inflammatory
responses in healthy subjects
In atopy there is a lack of balance between TH1, TH2, and Treg
cell responses. We characterized the expression of TH1, TH2, and
anti-inflammatory genes in stimulated and unstimulated PBMCs
with a principal component analysis and found that the first
principal component of TH2 cytokines (accounting for 51% of
the variance) correlated strongly with the first principal
component of TH1 and anti-inflammatory molecules (47% of the
variance) in healthy subjects (P 5 .021; Fig 3, A). For instance,
IL13 and IFNG expression correlated in healthy subjects in both
unstimulated (P 5 .0009; see Fig E4, A, in this article’s Online
Repository at www.jacionline.org) and allergen-stimulated
(Bet v 1, P 5 .0005; Fig 3, B) PBMCs. In striking contrast the
atopic subjects lacked such a balance and had significantly lower
expression of IFNG relative to IL13 (Fig 3, B). Similar patterns
were observed for IL13 and IL4 against IL27, with a strong
correlation in healthy subjects but none in atopic subjects
(see Fig E4, B-D). Furthermore, IL4 expression correlated strongly
with that of the Treg cell marker FOXP3 (P 5.0023; see Fig E4, E)
and the anti-inflammatory cytokine TGFB (P 5 .0002; Fig 3, C)
but again only in the healthy subjects. Finally, there was a highly
significant balance between IL10 and IL4 expression in unstimu-
lated long-term PBMC cultures in healthy subjects (P < .0001)
but not in atopic subjects (see Fig E4, F).
Acinetobacter species–induced immune responses
in vitro
To examine the influence of Acinetobacter species on the
immune system, we incubated human moDCs and human
keratinocytes with A lwoffii and 2 other skin commensals,
Staphylococcus aureus and Staphylococcus epidermidis. Consis-
tent with previous studies,18,20,21 we observed strong induction of
IL12 (Fig 4, A), IL10, and IL-10–inducing genes (Delta-like-4 and
IL27; Fig 4, B) in A lwoffii–stimulated moDCs. In keratinocytes A
lwoffii–induced expression of TNF and IL10, but not thymic stro-
mal lymphopoietin (TSLP), which was induced by S epidermidis
(Fig 4, C).
Acinetobacter species–induced protective immune
responses in vivo
The effect of microbial materials on the immune system has
been studied in several mouse models by using the airways or the
J ALLERGY CLIN IMMUNOL FYHRQUIST ET AL 1305
VOLUME 134, NUMBER 6

FIG 3. A, TH2-type responses correlated (*P 5 .021) with TH1/Treg cell–type responses in the healthy
(open symbols) but not atopic (solid symbols) subjects. B and C, In the healthy subjects IFNG expression
correlated with IL13 expression (***P 5 .0005) and was at a relatively higher level (**P < .01; Fig 3, B),
and IL4 expression correlated with TGFB expression (**P 5 .0002; Fig 3, C). Sample size is 46 healthy
and 28 atopic subjects. RQ, Relative quantity.

gut as exposure routes. To examine how Acinetobacter species
influences the immune system through the skin, we injected
mice intradermally with heat-inactivated A lwoffii or S epidermi-
dis during the sensitization phase of the asthma protocol (Fig 5,
A). Four injections of A lwoffii together with the allergen (OVA)
resulted in reduced lung inflammation after allergen challenge
through the airways compared with that seen in OVA-treated
control mice or mice treated with OVA plus S epidermidis
(Fig 5, B). The number of eosinophils in the bronchoalveolar
lavage fluid, the expression of IL5 and IL13 mRNA in lung tissue
(Fig 5, C, and see Fig E5 in this article’s Online Repository at
www.jacionline.org), and the serum levels of OVA-specific IgE
and IgG2a were significantly lower in the A lwoffii–treated mice
(Fig 5, D). One week after the last exposure to microbes, the A
lwoffii–treated mice had increased levels of IL-10 and IFN-g in
the skin at the site of the injections (Fig 5, E). These results clearly
demonstrate strong modulation of the local environment in the
skin by A lwoffii toward TH1-polarized anti-inflammatory
immune responses, resulting in reduced allergen-induced
production of TH2 cell–associated cytokines and IgE production.
DISCUSSION
Here we present evidence of a robust positive association
between the relative abundance of Acinetobacter species on the
skin and the expression of anti-inflammatory genes in PBMCs
in healthy subjects, an association that is strikingly lacking in
atopic subjects. Furthermore, healthy subjects display a
controlled balance between anti-inflammatory TH1 and TH2
gene transcription, which is related to the composition of the
skin microbiota. Finally, we show that Acinetobacter species
triggers anti-inflammatory and TH1-polarized immune responses
in human keratinocytes and moDCs and inhibits the development
of allergic airway inflammation through the skin in a mouse
model. These observations indicate a critical role for the skin
microbiota in tuning immune responses and inducing tolerance
against allergic sensitization, with effects beyond the skin.
The allergic phenotype is the product of both genetic
predisposition and gene-environment interactions, with the latter
heavily influencing the development of atopy and TH2-
dominated immunity.22 Adaptive immune responses are largely
immature and suppressed in the newborn, being dependent on
programming by environmental factors. Exposures promoting
the development of TH1 and anti-inflammatory immune re-
sponses protect against atopy, and their timing is crucial, with
the strongest effect early in life. Nonetheless, even in adoles-
cents, the immune system is susceptible to lifestyle factors that
provoke immune imbalance.23
In healthy subjects the relative amounts or balance of the TH1,
TH2, and anti-inflammatory cytokines appear to be more
1306 FYHRQUIST ET AL J ALLERGY CLIN IMMUNOL
DECEMBER 2014

FIG 4. Heat-inactivated Acinetobacter lwoffii (Al) induced the expression of IL12 (A) and IL10, Delta-like-4,
and IL27 (B) in human moDCs and TNF and IL10, but not TSLP (C), in human primary keratinocytes.
Sa, Staphylococcus aureus; Se, Staphylococcus epidermidis. *P < .05 and **P < .01. Bars represent
means 6 SEMs (n 5 3-5 per group).

important than the absolute amounts, and the consistency of this
balance suggests functional significance.24 The mechanisms that
maintain the balance of immune responses are unknown but might
include induction of IL-10. Considering the hypothesis that IL-10
plays a pivotal role, our results on the networks of microbial
genera on the skin and PBMC gene expression point to Acineto-
bacter species as a potential causative factor. In the network for
healthy subjects, the very strongest positive link among all the
identified microbial genera and the set of 17 immune-related
genes was between the relative abundance of Acinetobacter spe-
cies and IL10, which was further related to a network of other anti-
inflammatory genes, including FOXP3 and TGFB. In the atopic
subjects these correlations were lacking or even reversed. The
anti-inflammatory genes further correlated strongly with TH2-
type cytokine expression in the healthy subjects, suggesting that
the immune balance might be directly linked to the composition
of the skin microbiota. TH1-type responses were tightly correlated
with TH2-type responses but only in healthy subjects.
IL-10 is produced by many different immune cells.25 Of
PBMCs, monocytes are the most efficient producers of IL-10.
Our results point to the same conclusion. Thus the expression
of IL-10 in unstimulated PBMCs, which correlated with the
relative abundance of Acinetobacter species on the skin in the
healthy subjects, is likely derived from monocytes. However,
the potential role of T cells cannot be excluded because
the production of IL-10 also correlated with Acinetobacter
species in 24-hour PBMC cultures that were stimulated with
anti-human CD3 and anti-human CD28 (see Fig E6 in this
article’s Online Repository at www.jacionline.org). The
involvement of T cells is further supported by the association of
J ALLERGY CLIN IMMUNOL FYHRQUIST ET AL 1307
VOLUME 134, NUMBER 6

FIG 5. Intradermal injections of A lwoffii (Al; A) reduced lung inflammation (B), eosinophil counts in
bronchoalveolar lavage fluid and TH2 cytokine expression in lung tissue (C), and OVA-specific IgE and
IgG2a levels in serum (D). E, A lwoffii induced IL-10 and IFN-g in the skin. S epidermidis (Se) was used
as a control. *P < .05, **P < .01, and ***P < .001. Bars represent means 6 SEMs (n 5 8 per group). i.d.,
Intradermal; i.n., intranasal; RQ, relative quantity.
1308 FYHRQUIST ET AL
Acinetobacter species with the transcription of genes (eg,
FOXP3) that are expressed by T cells. These results suggest
that the tuning of immune responses, which appears to be
influenced by Acinetobacter species, occurs through both the
innate and adaptive arms of immunity.
Our present and previous results10 suggest a protective role for
Gammaproteobacteria and especially for Acinetobacter species
on the skin against the risk of allergic sensitization. Previous
studies have identified Acinetobacter species from mattress dust
and found its relative abundance to be inversely related to the
risk of atopic disease in children.26 A lwoffii isolated from cow-
sheds induces tolerogenic21 and TH1-polarizing18 programming
of DCs and reduce allergic reactions in mice, acting through
TLR-mediated signaling17 and epigenetic modifications.15 In
our experiments A lwoffii induced TH1 programming and
expression of IL10 by moDCs and TNF (which is a known
promoter of Treg cells27) and IL10 by keratinocytes. These
results suggest that both immune cells and skin-resident
cells actively contribute to the induction of tolerance. It is
noteworthy that the skin commensal S epidermidis, but not A
lwoffii, triggered TSLP expression in the keratinocytes.
Skin-derived TSLP is a key initiator of TH2 immune responses,
activating DCs, which in turn induce the expression of TH2-
type cytokines by CD41 T cells.28
The atopic march in human subjects often starts with
sensitization of the skin, progressing through food allergies to
the development of rhinitis or asthma later on in life.29 To imitate
this process in a mouse model, we sensitized mice through the
skin, followed by allergen challenge through the airways. The
sensitizing process is orchestrated by DCs, which, after internal-
izing and processing the allergen, mature and migrate to local
lymph nodes, where they present the processed antigen to naive
T cells. The phenotype of the DC is flavored by danger signals,
cytokines, and chemokines of the local environment, which in
turn influence the programming of T cells into distinct subsets
of effector or regulatory cells. Naik et al16 recently demonstrated
that skin commensals are critical for the tuning of skin immune
responses by balancing effector and Treg cells, controlling skin
cytokine expression, and driving skin immunity against
pathogens in a manner dependent on IL-1 and TLR signaling.
A large part of the skin microbiota resides in subepidermal
compartments, with Gammaproteobacteria accumulating in the
deeper layers of the skin,30 where their influence on the immune
system is likely to be more profound. To mimic the localization of
bacteria in the skin, we injected mice intradermally with
heat-inactivated A lwoffii together with the allergen. Treatment
with A lwoffii resulted in significantly reduced lung inflammation
in terms of decreased infiltration of eosinophils and expression of
TH2 cytokines in the lung tissue, as well as lower levels of OVA
IgE in the serum. Remarkably, the injection of A lwoffii produced
long-lasting effects in the skin, inducing long-term expression of
IL-10 and IFN-g, with likely consequences for the sensitization
process. We conclude that A lwoffii has a strong modulatory
effect of the immune system through the skin, producing a local
TH1- and IL-10–polarized environment, which is translated into
systemic protection from allergic sensitization and inflammation
on allergen challenge.
In conclusion, our results indicate a significant immunomod-
ulatory and protective role for Acinetobacter species in the skin
microbiota, as supported by results from an epidemiologic study
of a human study cohort, cell assays, and an animal model. In the
J ALLERGY CLIN IMMUNOL
DECEMBER 2014
study cohort the healthy subjects displayed a robust balance
between anti-inflammatory and TH1/TH2 gene expression by
PBMCs, which was lacking in the atopic subjects; we suggest
that induction of IL-10 by Acinetobacter species contributes to
this balance in healthy subjects. We found significant induction
of TH1 and protective immune responses in vitro, as well as
protection against allergic sensitization and lung inflammation
in vivo by Acinetobacter species. Our experimental results add
credence to the observed negative correlation between the
diversity and abundance of Gammaproteobacteria on the skin
and the incidence of atopy.10 Given that high relative abundances
of Gammaproteobacteria and other Proteobacteria are associated
with much forest and agricultural land in the environment,10 these
results support the general notion that interactions with a
biologically rich and diverse natural environment might enrich
the commensal microbiota, with far-reaching consequences for
public health.
We thank Ms Katriina Rossi, Ms Sari Tillander, and Mr Sauli Savukoski for
technical assistance.
Key messages
d Abundance of allergy-protective Gammaproteobacteria
on the skin strongly associates with anti-inflammatory
gene expression in human PBMCs.
d Acinetobacter species induce anti-inflammatory and TH1-
type gene expression in DCs and keratinocytes in vitro and
protect against allergic sensitization (atopy) and lung
inflammation in a mouse model when applied
intradermally.
d Skin commensals play a key role in the tuning of immune
responses to environmental allergens.
REFERENCES
1. Bach JF. The effect of infections on susceptibility to autoimmune and allergic
diseases. N Engl J Med 2002;347:911-20.
2. Gupta R, Sheikh A, Strachan DP, Anderson HR. Burden of allergic disease in the
UK: secondary analyses of national databases. Clin Exp Allergy 2004;34:520-6.
3. von Hertzen L, Hanski I, Haahtela T. Natural immunity. Biodiversity loss and
inflammatory diseases are two global megatrends that might be related. EMBO
Rep 2011;12:1089-93.
4. Rook GA. 99th Dahlem conference on infection, inflammation and chronic
inflammatory disorders: darwinian medicine and the ‘‘hygiene’’ or ‘‘old friends’’
hypothesis. Clin Exp Immunol 2010;160:70-9.
5. Rook GA. Regulation of the immune system by biodiversity from the natural
environment: an ecosystem service essential to health. Proc Natl Acad Sci
U S A 2013;110:18360-7.
6. Rook GA, Raison CL, Lowry CA. Microbial ‘‘old friends,’’ immunoregulation and
socio-economic status. Clin Exp Immunol 2014;177:1-12.
7. Rook GA. Review series on helminths, immune modulation and the hygiene
hypothesis: the broader implications of the hygiene hypothesis. Immunology
2009;126:3-11.
8. Schram ME, Tedja AM, Spijker R, Bos JD, Williams HC, Spuls PI. Is there a rural/
urban gradient in the prevalence of eczema? A systematic review. Br J Dermatol
2010;162:964-73.
9. Ege MJ, Mayer M, Normand AC, Genuneit J, Cookson WO, Braun-Fahrlander C,
et al. Exposure to environmental microorganisms and childhood asthma. N Engl
J Med 2011;364:701-9.
10. Hanski I, von Hertzen L, Fyhrquist N, Koskinen K, Torppa K, Laatikainen T, et al.
Environmental biodiversity, human microbiota, and allergy are interrelated. Proc
Natl Acad Sci U S A 2012;109:8334-9.
11. Hooper LV, Littman DR, Macpherson AJ. Interactions between the microbiota and
the immune system. Science 2012;336:1268-73.
J ALLERGY CLIN IMMUNOL
VOLUME 134, NUMBER 6
12. Round JL, Mazmanian SK. Inducible Foxp31 regulatory T-cell development by a
commensal bacterium of the intestinal microbiota. Proc Natl Acad Sci U S A 2010;
107:12204-9.
13. Atarashi K, Tanoue T, Oshima K, Suda W, Nagano Y, Nishikawa H, et al.
Treg induction by a rationally selected mixture of Clostridia strains from the
human microbiota. Nature 2013;500:232-6.
14. Atarashi K, Tanoue T, Shima T, Imaoka A, Kuwahara T, Momose Y, et al.
Induction of colonic regulatory T cells by indigenous Clostridium species. Science
2011;331:337-41.
15. Brand S, Teich R, Dicke T, Harb H, Yildirim AO, Tost J, et al. Epigenetic
regulation in murine offspring as a novel mechanism for transmaternal asthma
protection induced by microbes. J Allergy Clin Immunol 2011;128:618-25, e1-7.
16. Naik S, Bouladoux N, Wilhelm C, Molloy MJ, Salcedo R, Kastenmuller W, et al.
Compartmentalized control of skin immunity by resident commensals. Science
2012;337:1115-9.
17. Conrad ML, Ferstl R, Teich R, Brand S, Blumer N, Yildirim AO, et al. Maternal
TLR signaling is required for prenatal asthma protection by the nonpathogenic
microbe Acinetobacter lwoffii F78. J Exp Med 2009;206:2869-77.
18. Debarry J, Garn H, Hanuszkiewicz A, Dickgreber N, Blumer N, von Mutius E, et al.
Acinetobacter lwoffii and Lactococcus lactis strains isolated from farm cowsheds
possess strong allergy-protective properties. JAllergy Clin Immunol 2007;119:1514-21.
19. von Hertzen L, Makela MJ, Petays T, Jousilahti P, Kosunen TU, Laatikainen T,
et al. Growing disparities in atopy between the Finns and the Russians: a
comparison of 2 generations. J Allergy Clin Immunol 2006;117:151-7.
20. Hessle C, Andersson B, Wold AE. Gram-positive bacteria are potent inducers of
monocytic interleukin-12 (IL-12) while gram-negative bacteria preferentially
stimulate IL-10 production. Infect Immun 2000;68:3581-6.
FYHRQUIST ET AL 1309
21. Zhang M, Liu M, Luther J, Kao JY. Helicobacter pylori directs tolerogenic
programming of dendritic cells. Gut Microbes 2010;1:325-9.
22. Romagnani S. Coming back to a missing immune deviation as the main
explanatory mechanism for the hygiene hypothesis. J Allergy Clin Immunol
2007;119:1511-3.
23. von Mutius E, Weiland SK, Fritzsch C, Duhme H, Keil U. Increasing prevalence of
hay fever and atopy among children in Leipzig, East Germany. Lancet 1998;351:
862-6.
24. Halonen M, Lohman IC, Stern DA, Spangenberg A, Anderson D, Mobley S, et al.
Th1/Th2 patterns and balance in cytokine production in the parents and infants of a
large birth cohort. J Immunol 2009;182:3285-93.
25. Ouyang W, Rutz S, Crellin NK, Valdez PA, Hymowitz SG. Regulation and
functions of the IL-10 family of cytokines in inflammation and disease.
Ann Rev Immunol 2011;29:71-109.
26. Ege MJ, Mayer M, Schwaiger K, Mattes J, Pershagen G, van Hage M, et al.
Environmental bacteria and childhood asthma. Allergy 2012;67:1565-71.
27. Chen X, Baumel M, Mannel DN, Howard OM, Oppenheim JJ. Interaction of TNF
with TNF receptor type 2 promotes expansion and function of mouse
CD41CD251 T regulatory cells. J Immunol 2007;179:154-61.
28. Leyva-Castillo JM, Hener P, Michea P, Karasuyama H, Chan S, Soumelis V, et al.
Skin thymic stromal lymphopoietin initiates Th2 responses through an orchestrated
immune cascade. Nat Commun 2013;4:2847.
29. Spergel JM. From atopic dermatitis to asthma: the atopic march. Ann Allergy
Asthma Immunol 2010;105:99-109, 117.
30. Nakatsuji T, Chiang HI, Jiang SB, Nagarajan H, Zengler K, Gallo RL.
The microbiome extends to subepidermal compartments of normal skin.
Nat Commun 2013;4:1431.
1309.e1 FYHRQUIST ET AL J ALLERGY CLIN IMMUNOL
DECEMBER 2014

METHODS Transcription kit (Life Technologies), according to the manufacturer’s

instructions. A reaction was performed in 25 mL at 258C for 20 minutes,
Study subjects and atopic sensitization
followed by 378C for 120 minutes. For mouse lung and skin specimens,
The study subjects were a random sample of 13- to 20-year-old adolescents
0.5 mg of RNA was used for cDNA synthesis.
(n 5 118) selected for a long-term allergy study among school children in
2003E1 and living within an area of 100 3 150 km in eastern Finland. Samples
were collected in September 2010 and 2011. The homes of the study subjects
varied from flats in apartment buildings to row and individual houses in the
Real-time quantitative PCR analysis
mRNA levels of IFNG, IL1B, IL4, IL5, IL10, IL12A, IL12B, IL13, IL17,
town of Joensuu (a small town of 73,000 inhabitants) to villages of different
IL27, AHR, ALDH1A1, CYP1A1, CYP1B1, Delta-like-4, ENTPD1, FOXP3,
sizes and isolated houses in the sparsely populated rural area. For further
GZMB, MAF, TGFB, TNF, TSLP, and TLR2 were analyzed by means of
description of the study cohort, see Hanski et al.E2 Atopy was defined by a
quantitative RT-PCR with TaqMan chemistry and the 7500 Fast Real-Time
specific IgE level to inhalant allergens of greater than 2.5 kUA/L (see Hanski
PCR System (Applied Biosystems, Life Technologies, Foster City, Calif).
et alE2).
Reactions were performed in 1 cycle of 2 minutes at 508C and 30 seconds
at 958C followed by 40 cycles of 3 seconds at 958C and 30 seconds at 608C.
PCR amplification of the endogenous 18S rRNA, ACTB, and GAPDH for
Skin microbiota
human targets and 18S and TATA-binding protein for mouse targets was
The skin microbiota was analyzed, as described previously.E2 In brief,
performed for each sample to control sample loading and to allow normaliza-
study subjects were sampled with sterile nylon swabs, DNA was extracted
tion between samples. Probe and primer sets were purchased from Life
from the swabs with the FastDNA spin kit for soil (MP Biomedicals, Solon,
Technologies. The results are expressed as relative units, which were
Ohio), and the V1-V3 region of the 16s rRNA gene was amplified in a
calculated by using the comparative cycle threshold method, according to
PTC-225 thermal cycler (MJ Research, Bruno, Canada) and sequenced by
the manufacturer’s instructions.E3
using the 454-GS FLX Titanium protocol with an average read length of
approximately 400 bp (Roche Diagnostics, Mannheim, Germany). After
filtering out low-quality data, sequences were aligned and clustered, and oper-
ational taxonomical units were defined. Altogether, 1017 bacterial genera
PBMC cytokine protein quantification
were identified in the 118 study subjects. Supernatants from 24-hour PBMC cultures (unstimulated or anti-human
CD3 [R&D Systems] and anti-human CD28 [BD] stimulated) were analyzed
for cytokine production by using Luminex with Bio-Rad multiplex assays
PBMCs (Bio-Rad Laboratories).
PBMCs were separated from whole blood by using BD Vacutainer CPT
tubes (#362780; BD Biosciences PharMingen, San Jose, Calif), frozen, and
shipped to the site of analysis. The thawed PBMCs were cultured in 24-well
Microbes for in vitro and in vivo experiments
plates at 1 3 106/mL in complete RPMI-1640 medium (Gibco, Life A lwoffii strains were blood culture isolates from the Helsinki University
Central Hospital Laboratory that were identified by means of 16S rRNA
Technologies, Carlsbad, Calif) with 50 U/mL penicillin, 50 mg/mL
sequencing or API 20NE strips. A lwoffii, S epidermidis (ATCC 12228), and
streptomycin (PEST, Life Technologies), and 10% heat-inactivated FBS
S aureus (ATCC 25923) were grown over night at 1378C on chocolate agar
(Gibco, Life Technologies) at 378C in a 5% CO2 atmosphere and stimulated
with Bet v 1 or Phl p 1 at 10 mg/mL (Indoor Biotechnologies, Charlottesville, plates. Bacteria were collected and washed with sterile 0.9% NaCl. The micro-
bial concentration was adjusted to 1.8 3 108 cells/mL by using OD600. The
Va) or soluble mouse anti-human CD3 (1 mg/mL; R&D Systems,
microbes were killed by means of boiling for 15 minutes, placed in aliquots,
Minneapolis, Minn) and anti-human CD28 (1 mg/mL, BD) for 6 or 24 hours.
CD31 T cells and CD141 monocytes were isolated from PBMCs of 1 and stored at 2208C until use.
healthy donor by means of positive selection with human CD31 and
CD141 selection kits (STEMCELL Technologies, Grenoble, France),
respectively. The original PBMC population consisted of 82% lymphocytes
In vitro cell assays
and 14% monocytes according to the cell counter (Beckman Coulter, Monocytes were isolated from buffy coats of healthy donors (Red Cross
Fullerton, Calif). The lymphocyte population consisted of approximately Blood Transfusion Service, Helsinki, Finland) by using standard Ficoll density
gradient centrifugation, further enriched with the Monocyte enrichment kit
69% CD31 cells, and the monocyte population consisted of approximately
(STEMCELL Technologies), and allowed to adhere on 6-well plates at 1378C
90% CD141 cells. Thus the PBMCs contained approximately 12.6%
CD141 monocytes and approximately 56% CD31 T lymphocytes based on for 1 hour. Adherent cells were cultured in DC medium (complete
RPMI-1640, PEST, 10% FBS, GM-CSF [10 ng/mL, Sigma-Aldrich, St Louis,
cell counter values and flow cytometric analysis (Fig E3, A). The purity of
Mo], and IL-4 [20 ng/mL, Immunotools, Friesoythe, Germany]) for 8 days and
the cell populations was analyzed by means of flow cytometry with
analyzed by using flow cytometry (with mouse anti-human fluorescein
mouse anti-human allophycocyanin-CD14, phycoerythrin (PE)–CD11c, and
A488-CD3 antibodies (BD; Fig E3, B and C). The cell fraction values were isothiocyanate–CD209, PE-CD11c, PE-Cy5–CD1a, and allophycocyanin-
CD14 antibodies) to ensure proper differentiation. DCs were stimulated
used for extrapolation of gene expression levels of the 2 cell subsets relative
with heat-killed microbes at a 1:5 ratio (DC/microbe) for 6 hours. Pooled
to those of the mixed PBMC population.
neonatal human epidermal keratinocytes were purchased from Invitrogen Life
Technologies and cultured in Epilife medium supplemented with Human
RNA isolation and cDNA synthesis Keratinocyte Growth Supplement and PEST (Invitrogen, Life Technologies).
After 6 or 24 hours of culture, total RNA was extracted from the PBMCs by Cells were grown until 30% confluence, seeded into 25-cm2 cultures vials, and
using TRIsure reagent (Bioline, London, United Kingdom) and from left to adhere for 24 hours. The adhered cells were stimulated with microbes
human primary keratinocytes and human moDCs by using the Exiqon kit (2.5 3 105/mL) for 6 hours. RNA was isolated from DCs and keratinocytes by
(Vedbaek, Denmark), according to the manufacturer’s instructions. Mouse using an RNA isolation kit from Exiqon, followed by real-time qPCR, as
lung and skin tissue was homogenized in TRIsure by using the FastPrep described above.
instrument (Thermo Fisher Scientific, Waltham, Mass), and RNA was
extracted according to the manufacturer’s instructions.
The purity of RNA was analyzed by using the NanoDrop ND-1000 Murine asthma model
(Thermo Fisher Scientific, Wilmington, Del), wherein a A260/A280 ratio of BALB/c mice (Scanbur, Karlslunde, Denmark) aged 6 to 8 weeks were
greater than 1.8 was considered pure. Three hundred nanograms of RNA maintained on OVA-free diets and water ad libitum. All animal experiments
was reverse transcribed into cDNA by using the High Capacity cDNA Reverse were approved by the Social and Health Care Department of the State
J ALLERGY CLIN IMMUNOL
VOLUME 134, NUMBER 6
Provincial Office of Southern Finland. The mice were anesthetized and their
backs were shaved and tape stripped 3 times to introduce skin injury, followed
by intradermal injection of 50 mg of OVA in 100 mL of PBS in the
tape-stripped area. The mice were sensitized to OVA by a total of 4 injections
during a period of 4 weeks. Two groups of mice received 5 3 106 heat-
inactivated A lwoffii or S epidermidis in addition to OVA in the injections. Con-
trol mice received OVA or PBS only. On days 30, 31, and 32, the mice were
challenged with intranasally administered OVA (50 mg of OVA in 50 mL of
PBS), followed by collection of samples on day 33. Inflammatory cells in
the bronchoalveolar lavage fluid were handled and counted, as previously de-
scribed.E4 Skin or lung tissue was fixed in 10% formalin and embedded in
paraffin. Multiple 4-mm-thick sections were stained with the hematoxylin
and eosin protocol. Skin from the injection area or lung tissue was homoge-
nized in TRIsure, and RNA was extracted according to the manufacturer’s in-
structions. Real-time qPCR was performed, as described above. For
measurement of OVA-specific IgE and IgG2a levels, the plate was coated
with rat anti-mouse IgE/IgG2a mAb (BD Biosciences). Diluted sera were al-
lowed to bind overnight, and bound IgE/IgG2a was detected with biotinylated
OVA, Streptavidin-HRP (BD Biosciences), and peroxidase substrate reagents
(Kirkegaard & Perry Laboratories, Gaithersburg, Md). Absorbance at 405 nm
was read with an automated ELISA reader (Titertek Multiscan; Eflab, Turku,
Finland).
Statistical analyses
The association between the expression of each gene and the relative
abundance of Acinetobacter species on the skin of healthy and atopic subjects
was examined by using the Pearson correlation test. Analysis of covariance
was used to examine the effects of the relative abundance of Acinetobacter
species on the skin (ac), atopy status (at), and their interaction on the
expression of each gene separately, as follows:
yi ;ac1at1ac  at1error;
where y is the normalized and standardized PBMC expression. This analysis
was performed with R software, version 3.0.0 (http://cran.r-project.org).
Network analysis
Global associations of gene expression with the microbiota were inferred
by means of network analysis. The data were filtered as follows before
analysis. We selected only those genera that were present in at least 75% of all
samples (n 5 65). To simplify the analysis, we further removed all genera with
a Pearson correlation coefficient of less than 0.2 for healthy subjects and 0.4
for atopic subjects with all cytokines. This step was repeated for the 7 different
stimuli for cytokine expression Healthy and atopic subjects were treated
separately because of a smaller sample size for atopic (n 5 22) than healthy
(n 5 43) subjects; a smaller sample size increases the likelihood of spurious
correlations.
Patterns of association between cytokine expression and the relative
abundances of microbial genera were sought by using 3 different network-
building algorithms: ARACNE, convex optimization, and stepwise selection.
ARACNE (Algorithm for the Reconstruction of Accurate Cellular
Networks) is an information-theoretic algorithm that seeks to eliminate
indirect interactions from a network.E5 In the present case a Spearman
FYHRQUIST ET AL 1309.e2
correlation matrix between cytokines and microbes was used as the starting
point for the algorithm, as implemented in the package minet in R.E6
The convex optimization algorithm searches for a model that maximizes a
penalized log-likelihood for the concentration matrix (inverse of covariance
matrix) of the data (K).E7 This method was applied as implemented in the
package glasso in R,E8 with a penalization parameter r of 0.4 for healthy sub-
jects (0 < r < 1; higher values lead to sparser networks) and r of 0.6 for atopic
subjects (higher penalty for atopic individuals because of smaller sample size).
Finally, we used forward selection of network edges (starting from the
minimal BIC forest for the data in the package gRapHDE9 by using BIC as the
selection criterion [function stepw in package gRapHDE7]).
These algorithms were used to construct cytokine-microbe networks for
both healthy and atopic subjects. The credibility of each network was assessed
with random permutations. For this, the ordering of both rows and columns in
the data matrix were randomized while retaining the original labels (Patients
3 [Cytokines 1 Microbes]), and the randomized data were in turn used to
construct a network graph. This procedure was repeated 9999 times for each
network-building algorithm. Calculating the mean over the adjacency
matrices (a square matrix in which the presence of an edge between 2 nodes
is indicated with 0/1) for these 9999 networks gave the probability of detecting
an edge by chance, which could be compared with the networks inferred from
the original data. On the basis of these permutation analyses, we decided to
trust those edges that were recovered by at least 2 different algorithms with
90% probability. This high probability was chosen because high correlations
arise relatively easily at random in small data sets. However, because a node is
kept only if recovered with at least 2 independent methods, the type I error
probability is a P value of .01 or less. Because we were mainly interested in the
links between cytokines and microbes, we removed all microbial genera
without direct edges to cytokines from the final networks.
REFERENCES
E1. von Hertzen L, M€akal€a MJ, Pet€ays H, Jousilahti P, Kosunen TU, Laatikainen T,
et al. Growing disparities in atopy between the Finns and the Russians:
a comparison of 2 generations. J Allergy Clin Immunol 2006;117:151-7.
E2. Hanski I, von Hertzen L, Fyhrquist N, Koskinen K, Torppa K, Laatikainen T,
et al. Environmental biodiversity, human microbiota, and allergy are interrelated.
Proc Natl Acad Sci U S A 2012;109:8334-9.
E3. Schmittgen TD, Livak KJ. Analyzing real-time PCR data by the comparative
C(T) method. Nat Protoc 2008;3:1101-8.
E4. Haapakoski R, Karisola P, Fyhrquist N, Savinko T, Wolff H, Turjanmaa K, et al.
Intradermal cytosine-phosphate-guanosine treatment reduces lung inflammation
but induces IFN-g-mediated airway hyperreactivity in a murine model of natural
rubber latex allergy. Am J Respir Cell Mol Biol 2011;44:639-47.
E5. Margolin AA, Nemenman I, Basso K, Wiggins C, Stolovitzky G, Dalla Favera R,
et al. ARACNE: an algorithm for the reconstruction of gene regulatory networks
in a mammalian cellular context. BMC Bioinformatics 2006;7(suppl 1):S7.
E6. Meyer PE, Lafitte F, Bontempi G. MINET: An open source R/Bioconductor
package for mutual information based network inference. 2008
E7. H€ojsgaard S, Edwards D, Lauritzen S. Graphical models with R, use R!. Springer
New York Dordrecht Heidelberg London; 2012.
E8. Friedman J, Hastie T, Tibshirani R. glasso: Graphical lasso-estimation of
Gaussian graphical models. R package version 1.7. 2011.
E9. Abreu GCG, Edwards D, Labouriau R. High-dimensional graphical model search
with the gRapHD R package. J Stat Softw 2010;37:1-18.
1309.e3 FYHRQUIST ET AL J ALLERGY CLIN IMMUNOL
DECEMBER 2014

FIG E1. Expression of IL10 at the protein level corresponds closely to the level of RNA transcription in
PBMCs in healthy (left panel) and atopic (right panel) subjects. Sample size is 46 healthy and 28 atopic
subjects. RQ, Relative quantity.
J ALLERGY CLIN IMMUNOL FYHRQUIST ET AL 1309.e4
VOLUME 134, NUMBER 6

FIG E2. Expression of AHR in PBMCs correlates with the abundance of Acinetobacter species on the skin of
healthy (green circles) but not atopic (violet squares) subjects. Sample size is 46 healthy and 28 atopic
subjects. RQ, Relative quantity.
1309.e5 FYHRQUIST ET AL J ALLERGY CLIN IMMUNOL
DECEMBER 2014

FIG E3. A, The original PBMC population consisted of 56% CD31 T lymphocytes (middle panel) and 12.6%
CD141 monocytes (right panel). FSC, Forward scatter; SSC, side scatter. B and C, After cell separation, CD31
cells were 98% pure (Fig E3, B), and the CD141 cell subset was 97% pure (Fig E3, C; isotype control is shown
at right). D, The relative expression of cytokines was extrapolated based on the fraction of each cell type in
donors’ PBMCs. Bars represent mean 6 SEM (n 5 3 per group). RQ, Relative quantity.
J ALLERGY CLIN IMMUNOL FYHRQUIST ET AL 1309.e6
VOLUME 134, NUMBER 6

FIG E4. Linear regression plots of gene expression levels of IL-13 versus IFN-g (6-hour unstimulated PBMCs;
A), IL-13 versus IL-27 (6-hour Bet v 1–stimulated PBMCs; B), IL-13 versus IL-27 (6-hour Phl p 1–stimulated
PBMCs; C), IL-4 versus IL-27 (6-hour Bet v 1–stimulated PBMCs; D), IL-4 versus Foxp3 (6-hour unstimulated
PBMCs; E), and IL-4 versus IL-10 (24-hour unstimulated PBMCs; F). Green open symbols, Healthy; violet
solid symbols, atopic. Sample size is 46 healthy and 28 atopic subjects. RQ, Relative quantity.
1309.e7 FYHRQUIST ET AL J ALLERGY CLIN IMMUNOL
DECEMBER 2014

FIG E5. Representative images of immune cells in bronchoalveolar lavage

fluid cytospin preparations in control (A) and OVA-sensitized (B) mice and
in OVA-sensitized mice that were treated intradermally with A lwoffii (C) or
S epidermidis (D).
J ALLERGY CLIN IMMUNOL FYHRQUIST ET AL 1309.e8
VOLUME 134, NUMBER 6

FIG E6. The production of IL-10 protein by anti-human CD3– and anti-human CD28–stimulated PBMCs
correlates (P 5 .031) with the relative abundance of Acinetobacter species on the skin in healthy (green open
symbols) but not atopic (violet solid symbols) subjects. Sample size is 46 healthy and 28 atopic subjects.
1309.e9 FYHRQUIST ET AL J ALLERGY CLIN IMMUNOL
DECEMBER 2014

TABLE E1. Bacterial genera on the skin of the study subjects

All Healthy Atopic
Class Genus Mean SD Mean SD Mean SD

Acidobacteria Gp2 0.34 0.48 0.29 0.39 0.42 0.59

Actinobacteria Actinomyces 1.61 1.98 1.38 1.50 2.00 2.56
Actinobacteria Brachybacterium 0.39 0.66 0.37 0.74 0.43 0.51
Actinobacteria Corynebacterium 13.63 12.09 12.99 11.76 14.66 12.75
Actinobacteria Dermacoccus 0.95 2.72 0.93 1.95 0.97 3.68
Actinobacteria Kocuria 1.94 3.99 1.96 4.60 1.91 2.82
Actinobacteria Microbacterium 0.45 0.59 0.45 0.62 0.44 0.55
Actinobacteria Micrococcus 13.37 12.24 11.52 11.87 16.37 12.44
Actinobacteria Nocardioides 0.18 0.35 0.12 0.14 0.27 0.53
Actinobacteria Propionibacterium 17.66 18.39 18.91 20.94 15.64 13.35
Alphaproteobacteria Bradyrhizobium 0.12 0.15 0.13 0.16 0.09 0.11
Alphaproteobacteria Brevundimonas 0.49 0.95 0.54 0.86 0.41 1.09
Alphaproteobacteria Methylobacterium 0.30 1.00 0.34 1.24 0.24 0.40
Alphaproteobacteria Paracoccus 0.77 1.08 0.80 1.06 0.71 1.12
Alphaproteobacteria Sphingomonas 0.50 0.90 0.52 0.99 0.46 0.75
Bacilli Bacillus 0.24 0.43 0.16 0.24 0.37 0.62
Bacilli Gemella 0.48 0.70 0.49 0.69 0.46 0.73
Bacilli Granulicatella 0.43 0.68 0.43 0.72 0.43 0.61
Bacilli Lactobacillus 1.65 3.30 1.62 3.60 1.69 2.80
Bacilli Staphylococcus 5.17 6.16 4.03 3.60 7.01 8.64
Bacilli Streptococcus 7.82 9.66 8.17 9.77 7.25 9.63
Bacteroidia Porphyromonas 0.48 0.83 0.54 0.99 0.40 0.47
Bacteroidia Prevotella 1.31 2.74 1.62 3.38 0.80 0.90
Betaproteobacteria Alcaligenes 0.94 1.61 1.07 1.84 0.72 1.13
Betaproteobacteria Burkholderia 0.55 0.76 0.63 0.84 0.42 0.61
Betaproteobacteria Diaphorobacter 0.14 0.20 0.14 0.21 0.13 0.18
Betaproteobacteria Neisseria 0.17 0.29 0.18 0.33 0.16 0.20
Betaproteobacteria Ralstonia 6.94 12.73 8.58 14.39 4.29 9.06
Clostridia Anaerococcus 0.63 1.04 0.75 1.12 0.44 0.89
Clostridia Finegoldia 0.96 1.28 1.14 1.39 0.66 1.03
Clostridia Peptoniphilus 0.71 1.36 0.89 1.63 0.42 0.66
Deinococci Deinococcus 0.54 1.56 0.27 0.68 0.97 2.33
Deltaproteobacteria Desulfocurvus 0.29 0.39 0.31 0.41 0.26 0.35
Flavobacteria Chryseobacterium 0.33 1.12 0.22 0.45 0.50 1.73
Fusobacteria Fusobacterium 0.18 0.23 0.18 0.27 0.16 0.14
Gammaproteobacteria Acinetobacter 0.56 1.71 0.37 0.72 0.87 2.60
Gammaproteobacteria Enhydrobacter 1.31 3.47 1.21 2.76 1.47 4.44
Gammaproteobacteria Haemophilus 0.28 0.47 0.29 0.44 0.26 0.53
Gammaproteobacteria Pseudomonas 0.12 0.27 0.14 0.34 0.08 0.10
Gammaproteobacteria Stenotrophomonas 0.15 0.24 0.15 0.20 0.15 0.29
Negativicutes Veillonella 0.74 1.96 0.55 0.67 1.06 3.06
Total 85.81 85.39 86.47
J ALLERGY CLIN IMMUNOL FYHRQUIST ET AL 1309.e10
VOLUME 134, NUMBER 6

TABLE E2. Correlation analysis of gene expression in PBMCs TABLE E2. (Continued)
in relation to the relative abundance of Acinetobacter species Healthy subjects Atopic subjects
on the skin of healthy and atopic subjects Gene and stimulus r P value r P value
Healthy subjects Atopic subjects MAF, 6 h unstimulated 0.051 .739 20.156 .420
Gene and stimulus r P value r P value IL13, 6 h unstimulated 20.049 .746 20.089 .648
CYP1A1, 24 h unstimulated 20.046 .765 20.549 .005
IL10, 24 h unstimulated 0.509 4.24E-04 20.193 .367
IFNG, Phl p 1 0.026 .883 0.027 .907
FOXP3, 24 h unstimulated 0.425 4.04E-03 20.125 .562
IL27, 6 h unstimulated 0.020 .893 20.318 .093
TGFB, Bet v 1 0.400 5.31E-03 20.126 .516
IL17, Phl p 1 0.017 .924 0.177 .443
AHR, 24 h unstimulated 0.351 .018 0.038 .861
IL4, 6 h unstimulated 20.009 .954 0.170 .377
ENTPD1, 24 h unstimulated 0.351 .018 0.119 .579
CYP1B1, 6 h unstimulated 20.003 .984 0.009 .965
GZMB, 24 h unstimulated 0.351 .020 20.067 .756
TLR2, 6 h unstimulated 0.000 .998 20.117 .546
CYP1A1, Bet v 1 0.337 .021 0.002 .993
IL13, 24 h unstimulated 0.336 .026 0.245 .248
IL4, 24 h unstimulated 0.325 .029 0.026 .903
TGFB, 24 h unstimulated 0.315 .037 0.169 .430
FOXP3, Bet v 1 0.299 .041 20.220 .251
IL1B, Bet v 1 0.295 .044 0.000 .998
TLR2, 24 h unstimulated 0.303 .046 20.078 .718
TLR2, Phl p 1 0.315 .066 0.173 .454
IFNG, Bet v 1 0.270 .067 0.009 .964
FOXP3, Phl p 1 0.297 .083 20.299 .188
ALDH1A1, 24 h unstimulated 0.258 .091 0.234 .272
AHR, Bet v 1 0.238 .108 20.105 .587
CYP1B1, 24 h unstimulated 0.242 .109 0.010 .962
ENTPD1, Bet v 1 0.230 .120 20.039 .842
ALDH1A1, Bet v 1 0.229 .121 20.008 .968
TGFB, Phl p 1 0.260 .132 20.145 .531
IFNG, 24 h unstimulated 0.226 .141 20.185 .386
TLR2, Bet v 1 0.216 .145 20.062 .751
ALDH1A1, Phl p 1 0.245 .156 0.178 .440
IL27, Phl p 1 0.230 .184 20.302 .184
GZMB, 6 h unstimulated 20.196 .193 20.168 .383
IL13, Bet v 1 0.191 .199 20.083 .668
AHR, Phl p 1 0.219 .206 0.079 .732
MAF, Bet v 1 0.179 .228 20.261 .172
IL4, Bet v 1 0.173 .244 20.222 .247
IL10, Phl p 1 0.201 .246 20.243 .288
IL1B, Phl p 1 0.200 .250 20.157 .497
IL1B, 24 h unstimulated 0.168 .275 20.221 .300
MAF, 24 h unstimulated 0.167 .277 0.170 .426
IL1B, 6 h unstimulated 0.156 .300 20.319 .092
TGFB, 6 h unstimulated 0.148 .327 20.048 .805
GZMB, Bet v 1 0.146 .328 20.029 .883
IL27, Bet v 1 0.145 .329 0.038 .846
IL17, 6 h unstimulated 0.128 .397 20.280 .141
CYP1A1, Phl p 1 0.147 .399 0.197 .393
CYP1B1, Phl p 1 0.144 .410 0.029 .902
CYP1B1, Bet v 1 0.122 .413 20.010 .958
IL13, Phl p 1 0.136 .435 0.005 .982
IL4, Phl p 1 0.130 .455 20.032 .890
GZMB, Phl p 1 0.130 .455 0.124 .602
ALDH1A1, 6 h unstimulated 0.106 .484 20.016 .933
IL10, 6 h unstimulated 0.101 .504 20.308 .104
IL17, 24 h unstimulated 0.098 .527 20.247 .245
IFNG, 6 h unstimulated 0.090 .552 20.205 .286
IL17, Bet v 1 0.088 .554 20.221 .250
MAF, Phl p 1 0.101 .562 0.051 .825
IL27, 24 h unstimulated 0.089 .571 0.117 .576
IL10, Bet v 1 0.084 .575 20.074 .703
AHR, 6 h unstimulated 0.081 .593 20.029 .880
ENTPD1, 6 h unstimulated 0.069 .649 0.058 .766
CYP1A1, 6 h unstimulated 0.068 .654 0.094 .634
ENTPD1, Phl p 1 0.072 .681 20.094 .684
FOXP3, 6 h unstimulated 0.054 .721 20.136 .483

(Continued)
1309.e11 FYHRQUIST ET AL J ALLERGY CLIN IMMUNOL
DECEMBER 2014

TABLE E3. Analysis of covariance of gene expression in TABLE E3. (Continued)

PBMCs in relation to the relative abundance of Acinetobacter Acinetobacter
species on the skin and diagnosis Acinetobacter species
species, Diagnosis, 3 diagnosis,
Acinetobacter
Gene and stimulus P value P value P value
Acinetobacter species
species, Diagnosis, 3 diagnosis, TLR2, 6 h unstimulated .737 .929 .640
Gene and stimulus P value P value P value MAF, Bet v 1 .535 .929 .124
IL4, Bet v 1 .832 5.81E-06 .077 AHR, Phl p 1 .274 .935 .842
IL13, Bet v 1 .329 2.10E-04 .302 GZMB, Bet v 1 .499 .958 .497
IL4, Phl p 1 .491 6.21E-03 .618 IL27, Phl p 1 .587 .964 .077
IL10, 24 h unstimulated .022 .024 .005 IL1B, 24 h unstimulated .560 .967 .186
TGFB, 24 h unstimulated .061 .025 .643 ALDH1A, Phl p 1 .105 .975 .909
IL17, Bet v 1 .898 .031 .154 IFNG, 24 h unstimulated .389 .980 .133
IL4, 24 h unstimulated .100 .043 .245 GZMB, 6 h unstimulated .120 .996 .916
GZMB, 24 h unstimulated .074 .048 .089
ENTPD1, 24 h unstimulated .043 .053 .388
FOXP3, 6 h unstimulated .926 .060 .363
IL1B, Phl p 1 .680 .063 .225
IL10, Phl p 1 .834 .070 .121
IL1B, 6 h unstimulated .655 .093 .043
ALDH1A, Bet v 1 .336 .119 .536
FOXP3, Bet v 1 .321 .121 .031
MAF, 24 h unstimulated .255 .162 .748
IL27, Bet v 1 .288 .173 .704
TLR2, 24 h unstimulated .126 .259 .133
AHR, 24 h unstimulated .067 .290 .264
CYP1B1, 24 h unstimulated .239 .290 .400
FOXP3, 24 h unstimulated .111 .420 .042
ALDH1A1, 24 h unstimulated .057 .421 .831
CYP1B1, Bet v 1 .537 .433 .650
TLR2, Phl p 1 .058 .442 .903
AHR, Bet v 1 .500 .459 .194
GZMB, Phl p 1 .388 .483 .947
FOXP3, Phl p 1 .775 .489 .029
MAF, Phl p 1 .557 .509 .739
IFNG, Bet v 1 .113 .577 .297
IL17, Phl p 1 .732 .586 .698
IL13, 24 h unstimulated .009 .589 .609
AHR, 6 h unstimulated .690 .601 .663
IL13, Phl p 1 .452 .602 .659
CYP1A1, 24 h unstimulated .195 .603 .179
IL1B, Bet v 1 .108 .614 .225
IL4, 6 h unstimulated .556 .633 .419
CYP1A1, unstimulated .492 .646 .945
TGFB, Bet v 1 .068 .647 .031
IL17, 24 h unstimulated .978 .675 .243
CYP1B1, 6 h unstimulated .953 .676 .959
CYP1A1, Phl p 1 .225 .735 .684
IL10, Bet v 1 .904 .738 .514
MAF, 6 h unstimulated .959 .744 .453
IL27, 6 h unstimulated .425 .753 .171
IL17, 6 h unstimulated .950 .762 .124
IL27, 24 h unstimulated .448 .775 .859
ENTPD1, 6 h unstimulated .559 .794 .979
IFNG, 6 h unstimulated .873 .802 .230
ALDH1A1, 6 h unstimulated .641 .811 .638
ENTPD1, Phl p 1 .923 .815 .568
ENTPD1, Bet v 1 .336 .818 .314
IL10, 6 h unstimulated .536 .824 .066
CYP1A1, Bet v 1 .052 .850 .156
TLR2, Bet v 1 .326 .866 .267
CYP1B1, Phl p 1 .526 .873 .814
TGFB, 6 h unstimulated .515 .880 .438
IL13, 6 h unstimulated .576 .897 .839
IFNG, Phl p 1 .844 .923 .984
(Continued)