Content uploaded by Damien Portevin
Author content
All content in this area was uploaded by Damien Portevin
Content may be subject to copyright.
Available via license: CC BY 2.5
Content may be subject to copyright.
Natural killer cells are recruited during pulmonary
tuberculosis and their ex vivo responses to
mycobacteria vary between healthy human
donors in association with KIR haplotype
Damien Portevin,1* Laura E. Via,2Seokyong Eum3
and Douglas Young1
1Division of Mycobacterial Research,MRC National
Institute for Medical Research,The Ridgeway, Mill Hill,
London NW7 1AA, UK.
2Tuberculosis Research Section,LCID, NIAID, NIH,33
North Drive, Bethesda, MD 20892, USA.
3International Tuberculosis Research Center,
Gapo-dong, Masan, Changwon, Korea.
Summary
Humans vary widely in their susceptibility to
tuberculosis. While only a minority will progress
to disease, the majority of healthy individuals
exposed to Mycobacterium tuberculosis mount an
immune response that can clear or contain the
infection in a quiescent form. Using immunofluo-
rescence on human clinical samples, we identified
natural killer (NK) cells infiltrating granulomatous
pulmonary lesions during active disease. In order
to compare the NK cell ability to react to free myco-
bacteria in the context of tuberculosis infection
and Mycobacterium bovis BCG vaccination, NK
cells were isolated from the peripheral blood of
anonymous healthy human donors, and stimulated
with M. tuberculosis H37Rv or M. bovis BCG.
Extracellular M. tuberculosis and M. bovis BCG
could equally trigger the release of IFNgand TNFa
from NK cells in the presence of IL-2. However, we
found that this response varied 1000-fold between
individuals (n=52), with differences in KIR haplo-
type providing a significant criterion to distinguish
between low and high responders. Our findings
suggest that variations at the KIR locus and there-
fore of the NK cell repertoire may affect cytokine
production in response to mycobacteria and we
propose that this innate variability could sustain
different levels of susceptibility to M. tuberculosis
infection.
Introduction
Aerosol transmission of Mycobacterium tuberculosis
during active pulmonary disease results in exposure of a
substantial proportion of the global population, although
only a fraction of individuals develop clinical tuberculosis
(de Jong et al., 2008). Based on priming of an antigen-
specific immune response, it is estimated that two billion
people have been infected by M. tuberculosis, of whom
5–10% will go on to develop disease (WHO report,
updated annually). Epidemiology studies suggest that
around 20% of individuals intensively exposed to
M. tuberculosis show no evidence of a memory response,
suggesting a level of innate resistance that can function
prior to engagement of the adaptive immune system
(Cobat et al., 2009). We aimed to identify innate immuno-
logical mechanisms that underlie this diverse spectrum in
clinical outcome (Barry et al., 2009). In light of the fact that
natural killer (NK) cells can produce IFNg(Schoenborn
and Wilson, 2007), express cytolytic activity (Lanier et al.,
1986), and respond to conserved determinants of micro-
bial pathogens through Toll-like and other innate immune
receptors (Chalifour et al., 2004), we wished to investi-
gate whether these cells could constitute one variable in
the immune response to M. tuberculosis.
Human NK cells display extensive phenotypic heteroge-
neity and plasticity within and between individuals. For
instance, the level of CD56 surface expression discrimi-
nates two major subsets of NK cells whose frequencies in
the blood vary significantly between individuals (Cooper
et al., 2001). Such variation could have functional
immune consequences since CD56bright cells are usually
associated with cytokine production, and CD56dim cells with
natural cytotoxicity (Poli et al., 2009) although this
dichotomy has been recently refined (De Maria et al.,
2011). Furthermore, each individual possesses a most
likely unique repertoire of NK cells due to (i) the inherited
Received 5 April, 2012; revised 13 June, 2012; accepted 28 June,
2012. *For correspondence. E-mail dportev@nimr.mrc.ac.uk; Tel.
(+44) 2088162694; Fax (+44) 2088162694.
Re-use of this article ispermitted in accordance with the Terms
and Conditions set out at http://wileyonlinelibrary.com/onlineopen#
OnlineOpen_Terms
Cellular Microbiology (2012) 14(11), 1734–1744 doi:10.1111/j.1462-5822.2012.01834.x
First published online 30 July 2012
© 2012 Blackwell Publishing Ltd
cellular microbiology
set of genes and alleles coding for each different NK
cell receptor, and (ii) the stochastic expression of these
genes among NK cells from the same individual (Uhrberg,
2005). Moreover, there is growing evidence that NK cell
responses are tuned by a process that involves an inter-
action between Killer Immunoglobulin-like Receptors (KIR)
on NK cells and host-specific MHC class I molecules
(Brodin et al., 2009). A consequence is that variations
between individuals in the repertoire of KIR alleles
expressed by NK cells can be associated with differences
in responses to pathogen-associated signals and with
resistance or susceptibility to different diseases (Parham,
2005; Kulkarni et al., 2008; Korbel et al., 2009). In the
context of tuberculosis, a higher prevalence of KIR2DL3
among TB patients has been observed in two independent
studies (Mendez et al., 2006; Mahfouz et al., 2011).
In a murine model of tuberculosis, NK cells were found
to be recruited to the lung and to produce IFNgand per-
forin, although the absence of an infection phenotype
following antibody-mediated depletion led the investiga-
tors to conclude that their functional role was redundant
(Junqueira-Kipnis et al., 2003). In experiments using
g-chain-/-RAG-/-mice in comparison with RAG-/-immuno-
deficient mice, significant NK cell contribution was evident
to the control of M. tuberculosis infection in the absence of
T cell function (Feng et al., 2006), consistent with a model
in which NK cells provide an alternative source of activi-
ties overlapping with those of other immune cells.
However the extent to which NK cell contribution derived
from mice studies can be extended to humans is notably
limited by the inherent expression of independently
evolved and structurally unrelated set of MHC class
I-specific NK cell receptors belonging to the C-type lectin-
like family as opposed to the immunoglobulin-like family in
humans (Parham, 2005).
In humans, several clinical studies have explored the
potential association between peripheral blood NK cell
counts and resistance or susceptibility to tuberculosis
(Barcelos et al., 2008; Bozzano et al., 2009; Leung et al.,
2009). A reduction in frequency and functionality of
CD56bright NK cells has been observed in patients with
active tuberculosis and reciprocally, high blood levels
were associated with protection in putative tuberculosis-
resistant individuals. Variations in the number of NK cells
in cord blood have also been suggested to influence the
efficacy of BCG vaccination (Watkins et al., 2008; van den
Biggelaar et al., 2009).
The present study provides the first evidence of
the presence of NK cells in human granulomatous
lesions showing that these cells are taking part in the
immune response during pulmonary tuberculosis. The
cytotoxicity against M. tuberculosis-infected cells has
been addressed and the mechanisms characterized
(Vankayalapati et al., 2002; Garg et al., 2006). However,
direct IFNgresponses of peripheral blood NK cells to
mycobacterial preparations focused only on attenuated
strains, gamma-irradiated or heat-killed mycobacteria
rather than fully virulent M. tuberculosis (Wang et al.,
2004; Batoni et al., 2005; Schierloh et al., 2007). There-
fore, the consequence of a direct interaction between live
M. tuberculosis and human NK cells is still unknown.
Hence, we performed a systematic analysis of the
responses of NK cells from various anonymous human
blood donors, comparing cytokine response intensity to
extracellular virulent M. tuberculosis H37Rv with the
response to the attenuated Mycobacterium bovis BCG
Pasteur strain. We observed that the major determinant of
the NK cell response to mycobacteria is coming from the
host and is independent of mycobacterial virulence. We
describe an important variation of the cytokine response
intensity between NK cells from different individuals and
demonstrate a correlation with KIR gene content.
Results
NK cells are recruited to the lungs during
M. tuberculosis infection
Tuberculosis is generally treated by chemotherapy.
However, tuberculous patients suffering from multi-
drug-resistant tuberculosis may undergo surgery as an
adjunctive approach to reduce disease burden, which
gives access to resected lung tissue. Based on NKp46, a
single universal marker for mammalian NK cells (Walzer
et al., 2007), we used immunofluorescent microscopy to
look for NK cells, screening sections from five formalin-
fixed and paraffin embedded tuberculous lesions cover-
ing most of the different types of human granulomas as
reviewed by Leong et al. (2011). We were able to detect
numerous NK cells especially within inflammatory cell
infiltrates in a sample representative of a necrotizing
lesion (Fig. 1, panel b, c) and also within the well-
vascularized fibrotic wall delimiting this granuloma
(Fig. 1, panel a, d). The proximity of the NK cells to blood
vessels was almost universal, suggesting recent extrava-
sation. We observed a similar distribution within the con-
solidated area of a tuberculous pneumonia within the
fibrotic surroundings of a large necrotizing granuloma
(Fig. S1). In another sample, NK cells could be found
infiltrating the epithelioid macrophage layer of a well-
cuffed granuloma where liquefaction was evident, and
within the chronic inflammatory area juxtaposed to it
(Fig. S2). NKp46 signals were rarely observed in unaf-
fected airway tissues and only few signals could be
detected in the surroundings of a calcified granuloma or
within its sclerotic rim (Fig. S3). These observations
suggest that, during active tuberculosis, NK cells are
recruited at the site of disease especially within highly
Human NK cells and tuberculosis 1735
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744
inflamed granulomatous lesions. At this stage, NK cells
can interact with infected cells and also extracellular
mycobacteria released following lysis of infected cells by
specific CD8 T cells or NK cells themselves.
IFNgproduction by NK cells in response to extracellular
mycobacteria requires cytokine co-stimulation
We aimed to study the consequences of a direct interaction
between NK cells and a virulent strain of M. tuberculosis
and to determine whether mycobacterial virulence could
affect this interaction. We therefore started screening for
optimal time and conditions in which NK cells would
respond to mycobacterial stimulation (Fig. 2). We culti-
vated purified human NK cells with or without single cell
suspensions of M. tuberculosis H37Rv or M. bovis BCG
(MOI 1:1) in the presence or absence of two common
co-stimulatory cytokines for NK cell activity (i.e. IL-2 [100 U
ml-1) or IL-12p70 (1 ng ml-1)]. We collected supernatants
every 24 h for 3 days and measured release of IFNg. In this
Fig. 1. NK cells are recruited to the lungs during M. tuberculosis infection. Bottom left, H&E stain of a section from a necrotizing lesion
resected from the lung of a tuberculous patient that was used for immunofluorescence microscopy assays. (a to d) Insets from a
representative immunostained serial section showing the presence of NK cells (NKp46+in red) within fibrotic and vascularized regions
surrounding the necrotic centre (BV: blood vessel). Each inset shows a focus on positive cells at higher magnitude. Slides were
counter-stained with DAPI (blue).
Fig. 2. NK cell IFNgresponse to mycobacteria requires cytokine stimulation. NK cells purified from human PBMCs were cultivated with or
without single cell suspensions of M. tuberculosis H37Rv (triangles) or M. bovis BCG (circles) at a multiplicity of infection (MOI) of 1:1 in the
presence (filled) or in the absence (opened) of IL-2 (100 U ml-1) (continuous lines) or IL-12p70 (1 ng ml-1) (dashed lines). Supernatants
were collected every 24 h for 3 days and IFNgmeasured by ELISA. Mycobacteria alone were not able to trigger the production of IFNgby
resting NK cells. However, NK cells released IFNgfrom 24 h reaching a plateau 72 h post contact when co-activated with IL-2 or IL-12p70.
M. tuberculosis H37Rv and M. bovis BCG showed comparable potency to induce the production of IFNgby NK cells. Comparable kinetics for
mycobacterial sensing by NK cells were observed across three independent separate experiments.
1736 D. Portevin, L. E. Via, S. Eum and D. Young
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744
experimental setting, cytokines or mycobacteria alone
were not sufficient to independently trigger IFNgproduction
by NK cells. However, we observed progressive accumu-
lation of IFNgin culture supernatants from 24 h to 48 h that
began to plateau after 72 h of contact with the mycobacte-
ria and IL-2 or IL-12p70. In both cytokine environments, the
attenuated BCG vaccine strain elicited a comparable
response to virulent M. tuberculosis H37Rv. Although the
plateau value varies between donors, this kinetic pattern of
IFNgproduction was found consistent across three inde-
pendent experiments.
IFNgproduction by NK cells in response to extracellular
mycobacteria is independent of mycobacterial virulence
We subsequently compared the NK cell response from
three anonymous donors that were isolated, cultivated for
72 h in the presence or in the absence of mycobacteria
(MOI 1:1) and/or co-stimulatory cytokines, and analysed
simultaneously. Figure 3 illustrates the donor variability in
the final amount of IFNgreleased by NK cells following
contact with mycobacteria, independently of mycobacte-
rial strain. Indeed when looking at each donor individually,
we confirmed that M. tuberculosis was able to trigger very
similar cytokine response intensities as M. bovis BCG in
both cytokine environments. Using intracellular antibody
staining and polychromatic flow cytometry on another set
of donors, we confirmed that IFNgoriginated from NK cells
(Fig. 4A). We also observed that the amount of IFNgfound
Fig. 3. NK cell IFNgresponse to mycobacteria is independent
of mycobacterial virulence. NK cells were purified from three
independent donors and cultured in parallel in the presence of
IL-2 (100 U ml-1) or IL-12 (1 ng ml-1) and live M. bovis BCG or
M. tuberculosis H37Rv (MOI 1:1). Supernatants were harvested
after 72 h and assessed for IFNgcontent. When looking at the
cytokine response for each donor taken individually, M. bovis BCG
shows comparable antigenicity to M. tuberculosis. However, the
IFNgresponse intensity was found very variable between the
different NK cell preparations. Donors are numbered arbitrarily to
be presented in ascending order of response. (One of three
representative experiments)
Fig. 4. Intracellular cytokine staining of NK cells exposed to extracellular mycobacteria correlates with the amount of IFNgdetected in
supernatants. NK cells were purified from three independent donors (‘a’, ‘b’ and ‘c’) and cultured in parallel in the presence of IL-2
(100 U ml-1) and live M. bovis BCG (MOI of 1:1) for 18 h before brefeldin A treatment, antibody staining and flow cytometry analysis.
A. Pseudo-colour dot-plots showing variable de novo induction of IFNgproduction by mycobacteria (lower quadrants) across the three NK cell
preparations but not in the presence of IL-2 only (upper quadrants).
B. Histogram comparing the frequency of IFNg-positive NK cells after 24 h of mycobacterial exposure in the presence of IL-2 (100 U ml-1)
(white bars) and the amount of IFNgmeasured after 72 h of contact with mycobacteria and IL-2 (black bars). (One of three representative
experiments)
Human NK cells and tuberculosis 1737
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744
in the supernatants after 72 h reflected the frequency of
IFNgpositive NK cells 24 h post stimulation (Fig. 4B).
NK cell secretion profile after mycobacterial stimulation
highlights substantial donor variability
Since our previous observations suggested quantitative
differences in the predisposition of NK cells from indi-
vidual anonymous donors to respond to mycobacteria, we
evaluated this variability in a larger donor sample size.
Using standardized culture conditions, we recorded the
cytokine response of purified NK cells from 52 independ-
ent donors after 72 h of contact with mycobacteria (MOI
1:1). Since neither the mycobacterial virulence nor the
nature of the co-stimulatory cytokine influenced the NK
cell response in the previous experiments, we arbitrarily
chose to limit this screen to M. bovis BCG and IL-2
(100 U ml-1) as co-stimulatory cytokine. As a result of the
lower threshold of detection of the bead fluorescent tech-
nology used in this experiment to measure cytokine pro-
duction, we could now observe a slight but significant
induction of IFNgproduction by IL-2 alone but not with
mycobacteria (Fig. 5A). IFNgproduction in response to
co-stimulation with IL-2 and live mycobacteria extended
over three orders of magnitude when comparing the
different donor responses. In addition, since a previous
study suggested that M. bovis BCG could trigger the pro-
duction of TNFaby NK cells (Marcenaro et al., 2008), we
also measured the production of this cytokine in the same
set of samples. As with IFNg, we observed a slight but
significant induction of TNFasecretion by IL-2 but also by
mycobacteria alone and a synergistic effect of NK expo-
sure to both mycobacteria and IL-2 (Fig. 5B). TNFapro-
duction was also variable across the set of donors and
there was a significant correlation between the ability
of each independent donor to produce IFNgand TNFa
simultaneously (Spearman r0.7426, P<0.0001).
Mycobacteria preferentially trigger NK cell donor
associated with KIR B haplotype
There is increasing evidence that different KIR/HLA geno-
types influence NK cell potency and the threshold of their
responsiveness (Kim et al., 2008; Brodin et al., 2009). We
therefore characterized the gene content of the KIR
cluster for each of the 52 donors screened in the stand-
ardized assay. This characterization was performed by
PCR using two independent sets of primers per gene as
previously described (Martin and Carrington, 2008) and
summarized in Table 1. KIR haplotypes can be segre-
gated in two groups (A and B) according to their gene
content (Garcia et al., 2003). In contrast to B haplotypes,
which show higher genetic diversity, the haplotype A
group is characterized by the absence of KIR2DL5,
KIR2DS1,KIR2DS2,KIR2DS3,KIR2DS5 and KIR3DS1
and would express one single activating KIR, KIR2DS4.
Interestingly, we found a significantly higher proportion of
donors that were homozygous for the AA haplotypes
within the group of low responders, i.e. below the median
response (Fig. 6). Therefore, donors harbouring one or
several activating receptors other than KIR2DS4 are sig-
nificantly more represented in the group of high respond-
ers. When we looked for the activating KIR that could
drive this association, we found that KIR2DS3 and
KIR2DS5, two similar KIR with unidentified ligands, were
significantly over-represented within the group of high
responders [c2, d.f. (3.82, 1), P=0.0253].
Discussion
Consideration of the potential contribution of NK cells is
generally restricted to the early phase of mycobacterial
infection, prior to engagement of adaptive immunity. Our
demonstration of the presence of NK cells within mature
granulomatous lesions is consistent with an involvement
that extends into later stages of mycobacterial pathogen-
Fig. 5. NK cell secretion profile highlights
substantial donor variability. Purified NK cells
from 52 independent donors were cultured
in the presence or in the absence of IL-2
(100 U ml-1) and live M. bovis BCG (MOI 1:1)
for 72 h. Cell-free supernatants were
subjected to simultaneous multiple analytes
measurement technology (A: IFNg;B:TNFa).
Direct interaction with mycobacteria in
synergy with IL-2 induced the production
of IFNgand TNFaby human NK cells
(Wilcoxon matched-pairs signed rank test,
****P<0.0001). This cytokine production
shows important variation across different
donors. There is a statistically significant
correlation between the intensity of individual
IFNgversus TNFaresponses (Spearman r
0.7426, P<0.0001).
1738 D. Portevin, L. E. Via, S. Eum and D. Young
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744
esis. Given their rapid turnover (<15 days) (Zhang et al.,
2007), the presence of NK cells in lesions suggests con-
stant recruitment, which is reflected by the frequency of
positive signals in the vicinity of blood vessels proximal to
the lesions. Our observations suggest that NK cells are
frequently recruited into inflamed tissues, patrolling the
lesions where they could intercept infected cells migrating
out of the granuloma as well as interact with extracellular
Table 1. Analysis of the Leucocyte Receptor Complex (LRC) locus shows higher representation of KIR haplotypes B within high responders.
IFNg3DL3 2DL4 3DL2 2DL2 2DL3 2DL5 2DL1 3DL1 2DP1 2DS4 2DS1 2DS2 2DS3 2DS5 3DS1 Haplo
Low
E3 4.3 +++-++++-++--++B
D3 4.4 ++++-+++++-++--B
H4 4.7 +++-+-++++-----A
G4 5.8 +++-+-++++-----A
S4 6.1 +++-+-++++-----A
F4 8.4 +++-+-++++-----A
M4 10.0 ++++++++++++-++B
T4 11.1 +++-+-++++-----A
D\4 13.5 +++++-++++-+---B
L4 15.7 +++-+-++++-----A
K3 15.8 +++++-++++-+---B
E4 17.6 +++-+-++++-----A
I4 20.0 +++-+-++++-----A
C4 20.6 +++-+-++++-----A
U3 24.8 +++++++++++++++B
R4 27.0 +++-+++-+-++-++B
N4 29.8 +++++++++++++-+B
G3 31.6 +++-+-++++-----A
K4 33.2 +++++++-+-++---B
J4 36.1 +++-+-++++-----A
Q4 40.0 +++++-++++-+---B
W3 48.9 +++-+-++++-----A
J3 81.3 ++++++++++++-++B
A4 95.4 +++-+-++++-----A
O3 115.4 +++-+-++++-----A
I3 142.5 +++-+++++++--++B
Frequency 100 100 100 34.62 96.15 34.62 100 92.31 96.15 92.31 30.77 38.46 11.54 23.08 30.77 53.85
High
P3 148.9 +++++-++++-+---B
F3 159.2 +++++++++++++-+B
D4 160.2 ++++++++++-++--B
Z3 172.8 +++++-++++-+---B
V3 207.0 +++-++++++---++B
H3 225.9 +++-+-++++-----A
B4 237.8 +++++++-+-+++++B
Z2 256.3 ++++-+-+-+++-++B
E\4 285.9 +++++-++++-+---B
B\4 345.3 ++++-+++++-++--B
B3 399.5 +++-++++++---+-B
A3 427.0 +++-+-++++-----A
X3 446.6 +++-+++++++--++B
Y3 576.2 +++-++++++---++B
T3 608.8 +++-+++-+-+--++B
S3 640.6 +++-+++++++--++B
U4 1207.2 +++++-++++-+---B
Q3 1581.0 +++-+-++++-----A
M3 1770.7 +++-+-++++-----A
C3 1793.6 +++++++++++++++B
P4 2662.4 +++-+-++++-----A
O4 3320.4 +++-+-++++-----A
L3 4416.6 +++-+++++++--++B
C\4 4528.6 +++--+++++--+-+B
R3 4930.4 +++-+-++++-----A
N3 8259.7 +++++++++++++++B
Frequency 100 100 100 42.31 88.46 57.69 96.15 92.31 96.15 92.31 34.62 42.31 26.92 42.31 46.15 26.92
**
DNA was extracted from PBMCs and KIR genotyping established as previously described (Martin and Carrington, 2008). The table summarizes the presence
(+) or the absence (-) of each KIR gene within the LRC. NK cell donors were ranked (top to bottom) according to the amount of IFNgmeasured after 72 h
of contact with mycobacteria (MOI 1:1) in the presence of IL-2 (100 U ml-1). Using the median, two groups referred as low versus high responders were
created for subsequent statistical analysis. Frequency for each gene or haplotype within each group has been calculated and significant changes are
highlighted in bold (*P<0.05).
Human NK cells and tuberculosis 1739
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744
mycobacteria released after lysis by cytotoxic cells.
Therefore, variation in the host genetics that could affect
NK cell responses would give rise to different levels of
innate resistance to M. tuberculosis infection.
We found that human peripheral blood NK cells have
the potential to produce IFNgwhen they encounter
M. bovis BCG as well as M. tuberculosis in an appropriate
cytokine milieu. This indicates that priming of NK cells by
cytokines is sufficient to render NK cells responsive to
mycobacteria. Release of IL-12 by macrophages and
dendritic cells could therefore promote NK cell-mediated
production of IFNgduring the early phase of infection,
while IL-2 from T cells provides a potential stimulus for
activation of NK cells at later stages. The fact that myco-
bacteria synergize with co-stimulatory cytokines to induce
IFNgproduction by NK cells is fully in line with previous
reports showing the requirement of co-stimulatory
cytokines following Toll-Like Receptor agonist stimulation
(Chalifour et al., 2004; Souza-Fonseca-Guimaraes et al.,
2012). We did not perform an exhaustive screen of poten-
tial co-stimulatory cytokines but it is very likely that IL-15
among other cytokines may also synergize with the
stimulation triggered by mycobacteria as observed for
Pathogen-Associated Molecular Pattern recognition.
Moreover, it has been shown that a cellular contact with
APCs can modulate the cytokine response of NK cells
that are recruited at the site of tuberculous pleurisy (Schi-
erloh et al., 2007). Initial studies have described partial
involvement of TLR2 and the natural cytotoxicity receptor
NKp44 in recognition of M. bovis BCG (Esin et al., 2008;
Marcenaro et al., 2008). The fact that we observed similar
induction of IFNgproduction suggests that recognition of
virulent M. tuberculosis and attenuated BCG vaccine by
human NK cells are most likely controlled by the same set
of pathogen receptors.
In contrast, the level of IFNgproduction by NK cells in
response to mycobacterial stimulation varies over a 1000-
fold between donors. Preferential expansion of particular
NK cell subsets during viral infection has been shown to
establish a long-lived protective memory response in mice
(Sun et al., 2009), and it is attractive to speculate that
mycobacterial infection or BCG vaccination could similarly
establish an NK memory response. Having used anony-
mous blood donors, and therefore in the absence of
medical records, we could not directly correlate NK cell
responsiveness with BCG vaccination or previous M.
tuberculosis infection for instance. However, using M. tu-
berculosis Protein Purified Derivative (PPD) to stimulate
PBMCs from our panel of anonymous donors and IFNg
release assay as a read-out, we did not find any significant
correlation between the NK cell responsiveness and
memory immune response indicative of previous expo-
sure to mycobacteria (Fig. S4). However, we cannot rule
out the possibility of bystander amplification or modulation
of NK cell activity by other infections (Marras et al., 2011).
A major component of NK cell inter-individual variability
is associated with the expression of specific HLA/KIR
haplotypes revealed as a crucial determinant of NK cell
responsiveness to tumour cell lines (Kim et al., 2008) and
pathogen-associated signals (Korbel et al., 2009). We
have shown here that KIR B haplotypes, i.e. those har-
bouring multiple activating KIRs, and especially KIR2DS3
or KIR2DS5, correlates with a higher responsiveness
to extracellular mycobacteria. These two receptors are
similar to each other and together form the activating
lineage III KIR (Moesta et al., 2010). Unlike other activat-
ing KIR, KIR2DS3 and KIR2DS5 were not derived from
a paired inhibitory receptor by recombination and their
ligands still remain unknown (Cheent and Khakoo, 2009).
Our observations suggest a link between KIR genotype
and the ability of the respective NK cell repertoire to react
to mycobacteria. However, despite being statistically sig-
nificant, the contingency test performed to study this
association has limited statistical power. For instance, the
use of the median as a cut-off to distinguish low from high
responders has been performed arbitrarily in order to use
all the collected data but other grouping would have led to
different interpretations. Sample size calculation indicates
that further study would require a much bigger sample
size to fully attest the link between KIR haplotype and the
cytokine response to M. tuberculosis. For instance, 2236
samples for each arm should be screened in order to
detect a difference of 160 units of IFNgon average
between each haplotype group (5% significance level,
80% power). It also important to highlight the fact that KIR
Fig. 6. KIR B haplotype is associated with a higher IFN-gamma
response to mycobacterial stimulation.
A. KIR genotype frequency, extracted from Table 1 revealed a
significant association between B haplotypes and high responders,
defined as those above the median response of IFNg.
B. The activating receptors KIR2DS3 and KIR2DS5 were found to
drive this association with a significant higher prevalence within the
group of high responders (chi-squared test, P<0.05).
1740 D. Portevin, L. E. Via, S. Eum and D. Young
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744
haplotype does not fully segregate low from high respond-
ers and some NK cell preparation from KIR A individuals
showed high IFNgresponse indicating that KIR haplotype
is certainly not the only determinant of the cytokine
response to M. tuberculosis. Still, this donor variability
highlights a substantial potential impact on the pathogen-
esis of M. tuberculosis that has to be considered. One
could argue that higher IFNgproduction should confer a
better protection against an intracellular pathogen through
macrophage activation, although higher inflammation
could also contribute to exacerbated tissue destruction
that is required for transmission in tuberculosis. Three
published studies have addressed the influence of KIR
genotype in tuberculosis in Mexican, Lebanese and
Iranian populations respectively (Mendez et al., 2006;
Mahfouz et al., 2011; Tajik et al., 2012). In the Lebanese
study, KIR A haplotype was found more prevalent in the
group of tuberculosis patients when compared with
healthy controls (2.6:1 versus 1.5:1). However, no asso-
ciation could be found in the Iranian study. Therefore, the
relations between NK cell repertoire, mycobacterial
responsiveness and the protection or the sensitivity
to tuberculosis need to be tested in various population
settings.
To conclude, our results contribute to the growing evi-
dence that NK cell activities are regulated by a complex
interplay between multiple stimulatory and inhibitory
signals which generates extensive functional diversity in
NK cell populations between individuals (Boyton and
Altmann, 2007; Kim et al., 2008). Thus, with their ability to
deliver a range of functions that complement various
aspects of innate and adaptive immunity, NK cells may
make an important contribution to diversity of the human
immune response to tuberculosis. Efforts to identify corre-
lates of immune susceptibility to tuberculosis and indica-
tors of successful vaccination are therefore likely to be
enhanced by evaluation of NK cell responses alongside
measurement of T cell markers in immunological analyses.
Experimental procedures
Ethics statement
The tissues used in this study were collected between 2003 and
2006 as previously described (Leong et al., 2011) with written
consent of the subjects, approval of the National Masan Tuber-
culosis Hospital IRB and an exemption by the US NIH, Office of
Human Subjects Research.
Blood samples, cells and cell cultures
Fresh blood packs (Buffy coats and Cones) from healthy adult
donors were purchased anonymously from National Blood Serv-
ices, London, UK. Peripheral blood mononuclear cells (PBMCs)
were prepared on a Ficoll-Paque density gradient (Amersham
Biosciences AB, Uppsala, Sweden) by centrifugation (800 g,
30 min at room temperature), washed twice and frozen in RPMI
1640-FCS (5%)-DMSO (8.7%)-methyl-cellulose (0.1%). NK
cells were selected from PBMCs by magnetic cell sorting using
indirect NK isolation kit (Miltenyi Biotec, Auburn, CA, USA)
according to manufacturer’s recommendations. Average NK cell
purity checked by flow cytometry (CD3-/CD16+/-/CD56+/-) was
97.51% ⫾2.47 (standard deviation) across all donors. NK cells
were cultured in complete RPMI 1640 medium, including 1 mM
sodium pyruvate, and 1% heat-inactivated fetal calf serum.
Recombinant human IL-2 and IL-12 were obtained from
PeproTech EC.
Culture and preparation of mycobacterial strains
Mycobacterium tuberculosis H37Rv and M. bovis BCG Pasteur
were grown at 37°C in Middlebrook 7H9 broth supplemented with
ADC (Becton Dickinson, Sparks, USA). The strains were grown
to mid-exponential growth phase and pelleted at room tempera-
ture. Single cell bacterial suspensions were then prepared as
previously described (N’Diaye et al., 1998). Briefly, the medium
was discarded, bacteria were dispersed by shaking for 1 min with
glass beads (3 mm diameter), and resuspended in PBS, pH 7.4.
The remaining clumps were removed by centrifuging the super-
natant for 10 min at 200 g. Bacteria were then plated on Middle-
brook 7H11 agar plates supplemented with OADC (Becton
Dickinson, Sparks, USA) to establish precise bacterial counts
before and after freezing aliquots with glycerol (5% final v/v) and
storage at -80°C.
Cytokine production analysis
Cell culture supernatants were filtered using 0.2 mm 96-well filter
plates (Corning) before detection of cytokines and chemokines
using either ELISA kits (Peprotech) or combined cytokine
singleplex assays (Invitrogen) on a Luminex100,cMAPTM Technol-
ogy. For intracellular cytokine detection, NK cells were stimulated
for 24 h. Brefeldin A (BioLegend) was added during the last
6 h of culture before harvesting cells for antibody staining
with anti-CD16-FITC (Miltenyi Biotec) and anti-CD56-PC7 (BD
Bioscience), then fixed and permeabilized using BD Cytofix/
Cytoperm™ buffer and stained with anti-IFNg-PE (BD Bio-
sciences). Cells were run on a BD Biosciences FACSCalibur flow
cytometer and data analysed using FlowJo 7.6.4.
Immunohistochemistry
Formalin fixed, paraffin embedded lung specimens obtained from
lung resection surgery of TB patients with chronic MDR-TB were
provided by the Tuberculosis Research Section of the Laboratory
of Clinical Disease, NIAID, NIH, Bethesda, MD, directed by Dr
Clifton E. Barry, III. Tissue sections (5 mm) were deparaffinized
and rehydrated before high temperature antigen retrieval (citric
acid 10 mM, pH 6) followed by a 20 min blocking step in PBS/
0.15% BSA/0.1% Tween20 and further blocking with Fc Receptor
Block solution (Innovex Biosciences) according to manufacturer’s
recommendations. Sections were then incubated overnight at
37°C with anti-human NKp46/NCR1 MAb (MAB 1850 & AF1850,
R&D systems) (5 mgml
-1) followed by 2 h with Alexa Fluor®
594 chicken anti-mouse and donkey anti-goat IgG (Invitrogen)
Human NK cells and tuberculosis 1741
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744
(20 mgml
-1) at room before mounting with DAPI (Vectashield).
Sections were washed with PBS between each step. Bright-field
and fluorescence image acquisition was performed using a Mirax
MIDI Scan with an HXP120 lamp (Zeiss) and a HV-F22 camera
(Hitachi). Annotated scaled images were converted to TIFF files
using Mirax viewer software.
Kirotyping
DNA was extracted from 5·106PBMCs using QIAmp DNA mini kit
(Qiagen). Oligonucleotide sequences and PCR amplifications
were performed as previously described (Martin and Carrington,
2008).
Statistical analysis
Data analysis, correlation, paired t-tests, and contingency
tests were performed using GraphPad Prism software. Without
assuming a pre-defined distribution of the response tested,
non-parametric statistical analysis has been used all across the
study. Unless the direction of the association was expected
prior to performing the assays, KIR2DS3/5 association study
for instance, two-tailed statistical test were always performed.
Statistical analysis of the KIR association study has been sub-
jected to external review to the UCL statistical support services
who performed sample size calculation.
Acknowledgements
We would like to thank NIMR Histology Services and especially Dr
Radma Mahmood for her technical support, Dr Helene Royo for
fruitful briefing on fluorescence microscopy and Dr Andreas Wack
for critical reading of the manuscript. We would also like to thank
the patients and staff of NMTH for their kind participation
in the research protocols at ITRC and NMTH. This work was
entirely supported by the MRC core funds (U117581288). D.P.
was holding a Career Development Fellowship from the MRC.
The tissue collection, L.E.V. and S.E. were supported in part by
the Intramural Research Program of the National Institute of
Allergy and Infectious Disease, National Institutes of Health, USA,
and in part by the Korean Ministry of Health, Welfare and Family.
References
Barcelos, W., Sathler-Avelar, R., Martins-Filho, O.A., Car-
valho, B.N., Guimaraes, T.M., Miranda, S.S., et al. (2008)
Natural killer cell subpopulations in putative resistant indi-
viduals and patients with active Mycobacterium tuberculo-
sis infection. Scand J Immunol 68: 92–102.
Barry, C.E., 3rd, Boshoff, H.I., Dartois, V., Dick, T., Ehrt, S.,
Flynn, J., et al. (2009) The spectrum of latent tuberculosis:
rethinking the biology and intervention strategies. Nat Rev
Microbiol 7: 845–855.
Batoni, G., Esin, S., Favilli, F., Pardini, M., Bottai, D.,
Maisetta, G., et al. (2005) Human CD56bright and
CD56dim natural killer cell subsets respond differentially to
direct stimulation with Mycobacterium bovis bacillus
Calmette-Guerin. Scand J Immunol 62: 498–506.
van den Biggelaar, A.H., Prescott, S.L., Roponen, M., Nadal-
Sims, M.A., Devitt, C.J., Phuanukoonnon, S., et al. (2009)
Neonatal innate cytokine responses to BCG controlling
T-cell development vary between populations. J Allergy
Clin Immunol 124: 544–550, 550 e541–542.
Boyton, R.J., and Altmann, D.M. (2007) Natural killer cells,
killer immunoglobulin-like receptors and human leucocyte
antigen class I in disease. Clin Exp Immunol 149: 1–8.
Bozzano, F., Costa, P., Passalacqua, G., Dodi, F., Ravera, S.,
Pagano, G., et al. (2009) Functionally relevant decreases
in activatory receptor expression on NK cells are associ-
ated with pulmonary tuberculosis in vivo and persist after
successful treatment. Int Immunol 21: 779–791.
Brodin, P., Karre, K., and Hoglund, P. (2009) NK cell educa-
tion: not an on-off switch but a tunable rheostat. Trends
Immunol 30: 143–149.
Chalifour, A., Jeannin, P., Gauchat, J.F., Blaecke, A.,
Malissard, M., N’Guyen, T., et al. (2004) Direct bacterial
protein PAMP recognition by human NK cells involves
TLRs and triggers alpha-defensin production. Blood 104:
1778–1783.
Cheent, K., and Khakoo, S.I. (2009) Natural killer cells: inte-
grating diversity with function. Immunology 126: 449–457.
Cobat, A., Gallant, C.J., Simkin, L., Black, G.F., Stanley, K.,
Hughes, J., et al. (2009) Two loci control tuberculin skin
test reactivity in an area hyperendemic for tuberculosis.
J Exp Med 206: 2583–2591.
Cooper, M.A., Fehniger, T.A., and Caligiuri, M.A. (2001)
The biology of human natural killer-cell subsets. Trends
Immunol 22: 633–640.
De Maria, A., Bozzano, F., Cantoni, C., and Moretta, L. (2011)
Revisiting human natural killer cell subset function
revealed cytolytic CD56(dim)CD16+NK cells as rapid pro-
ducers of abundant IFN-gamma on activation. Proc Natl
Acad Sci USA 108: 728–732.
Esin, S., Batoni, G., Counoupas, C., Stringaro, A., Brancati-
sano, F.L., Colone, M., et al. (2008) Direct binding of
human NK cell natural cytotoxicity receptor NKp44 to the
surfaces of mycobacteria and other bacteria. Infect Immun
76: 1719–1727.
Feng, C.G., Kaviratne, M., Rothfuchs, A.G., Cheever, A.,
Hieny, S., Young, H.A., et al. (2006) NK cell-derived IFN-
gamma differentially regulates innate resistance and neu-
trophil response in T cell-deficient hosts infected with
Mycobacterium tuberculosis.J Immunol 177: 7086–7093.
Garcia, C.A., Robinson, J., Guethlein, L.A., Parham, P., Mad-
rigal, J.A., and Marsh, S.G. (2003) Human KIR sequences
2003. Immunogenetics 55: 227–239.
Garg, A., Barnes, P.F., Porgador, A., Roy, S., Wu, S., Nanda,
J.S., et al. (2006) Vimentin expressed on Mycobacterium
tuberculosis-infected human monocytes is involved in
binding to the NKp46 receptor. J Immunol 177: 6192–
6198.
de Jong, B.C., Hill, P.C., Aiken, A., Awine, T., Antonio, M.,
Adetifa, I.M., et al. (2008) Progression to active tuberculo-
sis, but not transmission, varies by Mycobacterium tuber-
culosis lineage in The Gambia. J Infect Dis 198: 1037–
1043.
Junqueira-Kipnis, A.P., Kipnis, A., Jamieson, A., Juarrero,
M.G., Diefenbach, A., Raulet, D.H., et al. (2003) NK cells
respond to pulmonary infection with Mycobacterium tuber-
culosis, but play a minimal role in protection. J Immunol
171: 6039–6045.
1742 D. Portevin, L. E. Via, S. Eum and D. Young
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744
Kim, S., Sunwoo, J.B., Yang, L., Choi, T., Song, Y.J., French,
A.R., et al. (2008) HLA alleles determine differences in
human natural killer cell responsiveness and potency. Proc
Natl Acad Sci USA 105: 3053–3058.
Korbel, D.S., Norman, P.J., Newman, K.C., Horowitz, A.,
Gendzekhadze, K., Parham, P., and Riley, E.M. (2009)
Killer Ig-like receptor (KIR) genotype predicts the capacity
of human KIR-positive CD56dim NK cells to respond to
pathogen-associated signals. J Immunol 182: 6426–6434.
Kulkarni, S., Martin, M.P., and Carrington, M. (2008) The Yin
and Yang of HLA and KIR in human disease. Semin
Immunol 20: 343–352.
Lanier, L.L., Phillips, J.H., Hackett, J., Jr, Tutt, M., and Kumar,
V. (1986) Natural killer cells: definition of a cell type rather
than a function. J Immunol 137: 2735–2739.
Leong, F.J.W.-M., Eum, S., Via, L.E., and Barry, C.E., 3rd
(2011) Pathology of tuberculosis in the human lung. In A
Color Atlas of Comparative Pathology of Pulmonary Tuber-
culosis. T.F. group (ed.). Boca Raton, FL, USA: CRC press,
pp. 53–82.
Leung, W.L., Law, K.L., Leung, V.S., Yip, C.W., Leung, C.C.,
Tam, C.M., and Kam, K.M. (2009) Comparison of intracel-
lular cytokine flow cytometry and an enzyme immunoassay
for evaluation of cellular immune response to active tuber-
culosis. Clin Vaccine Immunol 16: 344–351.
Mahfouz, R., Halas, H., Hoteit, R., Saadeh, M., Shamsed-
deen, W., Charafeddine, K., et al. (2011) Study of KIR
genes in Lebanese patients with tuberculosis. Int J Tuberc
Lung Dis 15: 1688–1691.
Marcenaro, E., Ferranti, B., Falco, M., Moretta, L., and
Moretta, A. (2008) Human NK cells directly recognize
Mycobacterium bovis via TLR2 and acquire the ability to kill
monocyte-derived DC. Int Immunol 20: 1155–1167.
Marras, F., Bozzano, F., and De Maria, A. (2011) Involvement
of activating NK cell receptors and their modulation in
pathogen immunity. J Biomed Biotechnol 2011: 152430.
Martin, M.P., and Carrington, M. (2008) KIR locus polymor-
phisms: genotyping and disease association analysis.
Methods Mol Biol 415: 49–64.
Mendez, A., Granda, H., Meenagh, A., Contreras, S., Zava-
leta, R., Mendoza, M.F., et al. (2006) Study of KIR genes in
tuberculosis patients. Tissue Antigens 68: 386–389.
Moesta, A.K., Graef, T., Abi-Rached, L., Older Aguilar, A.M.,
Guethlein, L.A., and Parham, P. (2010) Humans differ from
other hominids in lacking an activating NK cell receptor that
recognizes the C1 epitope of MHC class I. J Immunol 185:
4233–4237.
N’Diaye, E.N., Darzacq, X., Astarie-Dequeker, C., Daffe, M.,
Calafat, J., and Maridonneau-Parini, I. (1998) Fusion of
azurophil granules with phagosomes and activation of the
tyrosine kinase Hck are specifically inhibited during phago-
cytosis of mycobacteria by human neutrophils. J Immunol
161: 4983–4991.
Parham, P. (2005) MHC class I molecules and KIRs in human
history, health and survival. Nat Rev 5: 201–214.
Poli, A., Michel, T., Theresine, M., Andres, E., Hentges, F.,
and Zimmer, J. (2009) CD56bright natural killer (NK) cells:
an important NK cell subset. Immunology 126: 458–465.
Schierloh, P., Yokobori, N., Aleman, M., Landoni, V.,
Geffner, L., Musella, R.M., et al. (2007) Mycobacterium
tuberculosis-induced gamma interferon production by
natural killer cells requires cross talk with antigen-
presenting cells involving Toll-like receptors 2 and 4
and the mannose receptor in tuberculous pleurisy. Infect
Immun 75: 5325–5337.
Schoenborn, J.R., and Wilson, C.B. (2007) Regulation of
interferon-gamma during innate and adaptive immune
responses. Adv Immunol 96: 41–101.
Souza-Fonseca-Guimaraes, F., Parlato, M., Fitting, C.,
Cavaillon, J.M., and Adib-Conquy, M. (2012) NK cell toler-
ance to TLR agonists mediated by regulatory T cells after
polymicrobial sepsis. J Immunol 188: 5850–5858.
Sun, J.C., Beilke, J.N., and Lanier, L.L. (2009) Adaptive
immune features of natural killer cells. Nature 457: 557–
561.
Tajik, N., Shah-Hosseini, A., Mohammadi, A., Jafari, M.,
Nasiri, M., Radjabzadeh, M.F., et al. (2012) Susceptibility
to pulmonary tuberculosis in Iranian individuals is not
affected by compound KIR/HLA genotype. Tissue Antigens
79: 90–96.
Uhrberg, M. (2005) Shaping the human NK cell repertoire: an
epigenetic glance at KIR gene regulation. Mol Immunol 42:
471–475.
Vankayalapati, R., Wizel, B., Weis, S.E., Safi, H., Lakey, D.L.,
Mandelboim, O., et al. (2002) The NKp46 receptor contrib-
utes to NK cell lysis of mononuclear phagocytes infected
with an intracellular bacterium. J Immunol 168: 3451–
3457.
Walzer, T., Blery, M., Chaix, J., Fuseri, N., Chasson, L.,
Robbins, S.H., et al. (2007) Identification, activation, and
selective in vivo ablation of mouse NK cells via NKp46.
Proc Natl Acad Sci USA 104: 3384–3389.
Wang, C., Li, J., Zheng, H., Wei, H., Zhang, R., Li, B., and
Tian, Z. (2004) The role of innate immune cells in the
response of heat-treated Mycobacterium tuberculosis
(M.tb) antigens stimulating PBMCs. Cell Mol Immunol 1:
467–470.
Watkins, M.L., Semple, P.L., Abel, B., Hanekom, W.A.,
Kaplan, G., and Ress, S.R. (2008) Exposure of cord blood
to Mycobacterium bovis BCG induces an innate response
but not a T-cell cytokine response. Clin Vaccine Immunol
15: 1666–1673.
WHO report (updated annually) Global tuberculosis
control [WWW document]. URL: http://www.who.int/tb/
publications/global_report/archive/en/index.html. Geneva.
Zhang, Y., Wallace, D.L., de Lara, C.M., Ghattas, H., Asquith,
B., Worth, A., et al. (2007) In vivo kinetics of human natural
killer cells: the effects of ageing and acute and chronic viral
infection. Immunology 121: 258–265.
Supporting information
Additional Supporting Information may be found in the online
version of this article:
Fig. S1. NK cells presence in a tuberculous pneumonia sample.
At the centre, H&E stain of a section from a tuberculous pneu-
monia sample resected from the lung of a tuberculous patient
that was used for immunofluorescence microscopy assays.
Insets from a representative immunostained serial section
showing the presence of NK cells (NKp46+in red) in various
part of the lesion, within the consolidated area (top left), within
Human NK cells and tuberculosis 1743
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744
lymphoid aggregates filling alveolar spaces (lower right) and at
the border of a well-delimitated necrotic lesion (upper right).
Fig. S2. NK cells presence in a consolidated and necrotizing
tuberculous lesion. At the centre, H&E stain of a section from a
tuberculous necrotizing granuloma that was used for the follow-
ing immunofluorescence microscopy assays. Insets from a rep-
resentative immunostained serial section showing the presence
of NK cells (NKp46+in red) starting notably to infiltrate the epi-
thelioid macrophage layer delimiting a large necrotic lesion.
Fig. S3. NK cells localization in a calcified tuberculous lesion. Top
left, H&E stain of a section of a calcified granuloma that was used
for the following immunofluorescence microscopy assays. Insets
from a representative immunostained serial section reveal the
presence of NK cells (NKp46+in red) recently extravasated from
a blood vessel (BV) (inset b), or within surrounding alveolar
spaces (inset a, d) of a calcified granuloma. Few signals could be
detected at the periphery and infiltrating the sclerotic rim (inset c).
Fig. S4. PBMCs immune response intensity following exposure
to M. tuberculosis antigens does not correlate with the ability of
respective NK cells to respond to mycobacteria. Histogram com-
paring IFNgproduction from (i) PBMC stimulated with Purified
Protein Derivative (PPD) from M. tuberculosis (2 mgml
-1) for 24 h
and (ii) NK cell preparation from the matching donor following
exposure to M. bovis BCG for 72 h (MOI 1:1) in the presence of
IL-2 (100 U ml-1). We observed substantial differences in the
intensity of the immune response to PPD among the 52 donors
(28 donors below 20 pg ml-1, 16 donors comprised between 20
and 100 pg ml-1and 8 donors over 100 pg ml-1). However, there
was no evident correlation between the memory response to
PPD among PBMCs and the variable responsiveness of NK cell
exposed to mycobacteria in the presence of co-stimulatory
cytokine.
Please note: Wiley-Blackwell are not responsible for the content
or functionality of any supporting materials supplied by the
authors. Any queries (other than missing material) should be
directed to the corresponding author for the article.
1744 D. Portevin, L. E. Via, S. Eum and D. Young
© 2012 Blackwell Publishing Ltd, Cellular Microbiology,14, 1734–1744