resource
Transcriptional programs define molecular characteristics
of innate lymphoid cell classes and subsets
npg
© 2015 Nature America, Inc. All rights reserved.
Michelle L Robinette1, Anja Fuchs2, Victor S Cortez1, Jacob S Lee3, Yaming Wang1, Scott K Durum4,
Susan Gilfillan1, Marco Colonna1 & the Immunological Genome Consortium5
The recognized diversity of innate lymphoid cells (ILCs) is rapidly expanding. Three ILC classes have emerged, ILC1, ILC2 and
ILC3, with ILC1 and ILC3 including several subsets. The classification of some subsets is unclear, and it remains controversial
whether natural killer (NK) cells and ILC1 cells are distinct cell types. To address these issues, we analyzed gene expression in
ILCs and NK cells from mouse small intestine, spleen and liver, as part of the Immunological Genome Project. The results showed
unique gene-expression patterns for some ILCs and overlapping patterns for ILC1 cells and NK cells, whereas other ILC subsets
remained indistinguishable. We identified a transcriptional program shared by small intestine ILCs and a core ILC signature.
We revealed and discuss transcripts that suggest previously unknown functions and developmental paths for ILCs.
The Immunological Genome (ImmGen) Project is a collaborative
effort by immunologists and computational biologists who seek,
through rigorously controlled protocols for data generation and
analysis, to delineate gene-expression patterns of cell types across
the immune system so as to better understand the immune response
and comprehensively define its regulatory networks1. In this context, we investigated the global gene-expression profiles of innate
lymphoid cells (ILCs) from mice following the ImmGen Project’s
stringent standards1.
ILCs are non-T, non-B lymphocytes present throughout the body
that show enrichment in frequency at mucosal surfaces2,3. Developing
from an Id2+ common helper-like ILC progenitor4, three classes of
ILCs, now known as ILC1, ILC2 and ILC3, have emerged that mirror
helper T cells in both their cytokine-production profiles and their
transcriptional circuitry2,3. Functionally, T-bet+ ILC1 cells respond
to interleukin 12 (IL-12), IL-15 and IL-18 to produce interferon-γ
(IFN-γ); GATA-3+ ILC2 cells react to IL-33, IL-25 and thymic stromal
lymphopoietin to produce type 2 cytokines, including IL-5 and IL-13;
and RORγt+ ILC3 cells are activated by IL-1 and IL-23 to produce
IL-22 and/or IL-17. As innate sources of distinct cytokines, ILCs have
roles in early defense against infections, modulation of the adaptive
immune response, the development of lymphoid tissue, and repair
and homeostasis of tissues2,3.
Whereas shared transcription factors, cell-surface markers and
functional properties of cytokine production define classes, deviations in one or more categories by subpopulations define subsets of
ILCs within the larger class. In the mouse, ILC2 seems to be the most
homogeneous class, defined by expression of the IL-7 receptor (IL-7R;
also called CD127), Sca-1 and the IL-33 receptor ST2. In comparison,
at least five subsets of mouse ILC3 cells have been reported, four
of which are found in the greatest numbers during steady state in
the adult small intestine, and one in the adult large intestine. These
include CD4+ and CD4− subsets of NKp46−RORγt+ lymphoid tissue–
inducer (LTi)-like cells5; RORγt+T-bet+ receptor Notch–dependent
NKp46+ ILC3 cells5–9; a potential ILC3-ILC1 transitional subset that
has downregulated RORγt, produces IFN-γ and expresses T-bet and
larger amounts of the activating natural killer (NK) cell receptor
NK1.1 compared to NKp46+ ILC3 cells (‘ex-RORγt+ ILC3 cells’)4; and
IL17-producing RORγt+NKp46− ILC3 cells in the large intestine10.
ILC1 cells have the most complicated and controversial distinction,
both of the ILC class itself and between subsets within the class. ILC1
cells and NK cells have functional similarities, mainly IFN-γ production, and share expression of T-bet and many cell-surface markers,
such as NKp46 and NK1.1. However, NK cells are currently thought
to have more cytotoxic potential than ILC1 cells have. Surface expression of CD127 and the integrin subunit CD49a are used to distinguish
ILC1 cells from NK cells in many but not all mouse tissues, as there is
considerable diversity of ILC1 subsets among tissues2. Lineage-tracing
experiments have shown that ILC1 cells and NK cells originate from
distinct progenitors4,11, and mature NK cells are dependent on the
transcription factor eomesodermin (Eomes), whereas ILC1 cells are
not. However, immature NK cells also share many markers with ILC1
cells12 and lack expression of Eomes12,13.
The breadth of polarization and relationships among ILC subsets has remained incompletely understood. To better understand
the functional differences between ILC classes and reported subsets
within a class, we discriminated seven populations of ILCs in the
small intestine lamina propria (siLP) (NK cells, ILC1 cells, ILC2 cells
1Department of Pathology and Immunology, Washington University School of Medicine, St. Louis, Missouri, USA. 2Department of Surgery, Washington University
School of Medicine, St. Louis, Missouri, USA. 3Merck Research Laboratories, Palo Alto, California, USA. 4Cancer and Inflammation Program, Center for Cancer
Research, National Cancer Institute, Frederick, Maryland, USA. 5A list of members and affiliations appears at the end of the paper. Correspondence should be
addressed to M.C. ([email protected]).
Received 17 October 2014; accepted 22 December 2014; published online 26 January 2015; doi:10.1038/ni.3094
306
VOLUME 16 NUMBER 3 MARCH 2015 nature immunology
resource
npg
© 2015 Nature America, Inc. All rights reserved.
and four subsets of ILC3 cells); three additional subsets of ILC1 cells
from liver, spleen and small intestine intraepithelial lymphocytes
(siIELs); and two NK cell subsets from liver and spleen. Our findings provide a molecular definition of ILC classes and subsets, and
also identify a core signature in ILCs distinct from that in NK cells.
We also identify novel targets for future investigation and generate
a comprehensive, high-quality and publically available resource of
ILC transcriptomes.
RESULTS
Analysis of ILC frequency and diversity
We isolated all reported ILC subsets in the siLP of 6-week-old C57BL/6
male mice. We isolated ILC2 cells from wild-type C57BL/6 mice
and isolated all other subsets from RORγteGFP reporter mice, which
express enhanced green fluorescent protein (eGFP) driven by the gene
encoding RORγt. These subsets included CD127− NK cells, CD127+
ILC1 cells, CD4−NKp46− and CD4+NKp46− LTi-like ILC3 cells,
NKp46+RORγthi ILC3 cells and NKp46+RORγtlo ILC3 cells (Fig. 1a).
Notably, NKp46+RORγthi and NKp46+RORγtlo ILC3 cells cannot be
discriminated by intracellular staining of RORγt in wild-type C57BL/6
mice, in which only one NKp46+RORγt+ subset is detectable; thus,
the use of RORγteGFP reporter mice provided the unique opportunity to profile RORγtlo and RORγthi subsets, as described14. As the
RORγtlo ILC3 subset had higher expression of NK1.1 than RORγthi
ILC3 cells had (Supplementary Fig. 1a–c), we reasoned that this
subset would show enrichment for converted or actively converting
‘ex-RORγt+’ ILC3 cells. We also profiled liver CD49b+TRAIL− NK
cells and CD49b−TRAIL+ ILC1 cells and spleen CD127− NK cells
and CD27+CD127+ ILC1 cells (Fig. 1a), the last of which have been
reported but have not previously been called ‘ILC1 cells’15. Small
intestine intraepithelial ILC1 cells were isolated from the intestinal
epithelium as NK1.1+NKp46+ (Fig. 1a). These cells are phenotypically
distinct from NK cells as a result of imprinting with transforming
growth factor-β16, although their developmental origin and transcriptional relationship to siLP ILC subsets remain unclear.
Cytospins showed that ILC subsets were morphologically pure lymphoid populations (Fig. 1b). We assessed the frequency of these 12
ILC subsets within the lymphocyte populations of the small intestine,
liver and spleen in naive mice (Fig. 1c). ILCs were most abundant
in the siLP, followed by liver and spleen; siIEL populations had the
lowest frequency of ILCs. Next we sorted these populations according to the ImmGen Project’s standardized protocol for data generation and analyzed gene expression by means of whole-mouse genome
array. Principal-component analysis (PCA) showed a greater degree
of diversity generated by ILC2 and ILC3 subsets than by ILC1 cells
(Fig. 1d and Supplementary Fig. 1e). In hierarchical clustering, three
pairs of ILC subsets were computationally indistinguishable (Fig. 1e):
splenic and liver NK cells, RORγthiNKp46+ and RORγtloNKp46+ ILC3
cells, and CD4+NKp46− and CD4−NKp46− LTi-like ILC3 cells were
intermixed. All ILC1 subsets clustered separately from NK cells from
the same tissue. However, siIEL ILC1 cells, siLP NK cells and siLP
ILC1 cells clustered in a separate branch of the dendrogram from
liver and spleen NK cells and ILC1 subsets. We concluded that beyond
classical polarizations, environmental factors in the small intestine
differentiated intestinal subsets from those in liver and spleen.
Expression profiles of individual ILC subsets
We assessed the gene-expression profile of each sorted subset (Fig. 2a–h)
by identifying characteristic transcripts that were expressed at least
twofold or fourfold higher in the index subset than in all other profiled
subsets. Heat maps demonstrate the extent to which the identified
nature immunology VOLUME 16 NUMBER 3 MARCH 2015
transcripts were specifically expressed by the index subset (Fig. 2a–h).
We found the greatest number of characteristic transcripts in the siLP
ILC2 cells, with 100 transcripts showing expression more than twofold higher than that in other subsets (Supplementary Table 1) and
34 transcripts showing expression more than fourfold higher than that
in other subsets (Fig. 2a). These included expected transcripts, such as
Il13, Il5, IL4, Il9r and Il17rb (which encodes the IL-25 receptor)17–19,
as well as several transcripts not previously shown to be expressed
by ILC2 (to our knowledge), including Rxrg, Pparg, Mc5r, Dgat2 and
Alox5 (Fig. 2a). The transcriptional repressor RXRγ, which binds the
vitamin A metabolite 9-cis retinoic acid20, has not been previously
described in ILC2 cells. However, vitamin A is known to directly
inhibit ILC2 differentiation by an unknown mechanism21, which suggests that RXRγ may mediate this effect. We also found previously
unrecognized expression of the gene encoding the transcription factor
PPARγ (Fig. 2a), which forms heterodimers with repressor retinoid
X receptors and transcriptionally mediates lipid homeostasis 20. ILC2
cells expressed several other genes encoding molecules linked to lipid
metabolism. These included genes encoding Dgat2, which mediates
the final reaction of triglyceride synthesis; Mc5r, a receptor that causes
lipid mobilization in adipocytes; and Alox5, a lipoxygenase that catalyzes synthesis of leukotriene A4 (Fig. 2a). ILC2 cells are known to
regulate the cellular immune system within visceral adipose tissues22,
but our data suggested that ILC2 cells might also directly sense lipids
and produce lipid mediators.
NKp46− LTi-like ILC3 subsets expressed the second greatest
number of characteristic transcripts. As suggested by clustering
analysis (Fig. 1e), CD4+ and CD4− LTi-like cells had overlapping
gene-expression patterns. CD4+ LTi-like cells expressed only four
transcripts more than twofold higher when compared to all other
subsets, one of which was Cd4, encoding the monomorphic coreceptor CD4; in contrast, no transcripts had characteristic higher expression in CD4− LTi-like ILC3 cells (Fig. 2b). However, 65 transcripts
had expression more than twofold higher in both subsets together
relative to other subsets (Supplementary Table 1), and 9 of these transcripts were expressed more than fourfold higher than in other subsets
(Fig. 2b). Eight additional transcripts were expressed at least fourfold
higher than in other subsets by either CD4+ LTi-like or CD4− LTi-like
ILC3 cells, with differences in expression among LTi-like ILC3 subsets
probably due to replicate variation (Fig. 2b). Transcripts expressed at
higher levels by both LTi-like ILC3 subsets included those encoding
the chemokine receptor CCR6, which has been used as a marker for
LTi-like ILC3 cells7, and the chemokine receptor CXCR5 (Fig. 2b).
We also found several genes not previously described in LTi-like ILC3
cells (to our knowledge), including Gucy1a3, Cntn1, Slc6a7, Cacna1g
and Nrp1 (Fig. 2b). Gucy1a3 encodes the α-subunit of the soluble
guanylate cyclase receptor, which transduces signals from nitric
oxide; however, we found no expression in LTi-like ILC3 cells of the
other components of the guanylate cyclase receptor (data not shown).
Cntn1, which encodes a glycosylphosphatidylinositol-linked member
of the immunoglobulin family, is best known for its role in regulating
axonal guidance and neural system development23 and has not previously been studied in an immune context. The L-proline transporter
encoded by Slc6a7 and the voltage-gated calcium channel encoded
by Cacna1g are similarly atypical and are not expressed in other cells
of the immune system (data not shown). We concluded that LTi-like
ILC3 cells specifically expressed several transcripts that are unique
in the immune system, including those encoding putative factors
involved in neural crosstalk.
The remaining ILC subsets had fewer candidate characteristic markers, probably because of multiple comparisons with other ­subsets in
307
resource
the same class. As indicated by PCA, siLP NKp46+RORγthi ILC3 cells
and siLP NKp46+RORγtlo ILC3 cells had overlapping gene-­expression
profiles. Sixteen transcripts showed expression that was twofold
higher in NKp46+RORγthi ILC3 cells than in all other profiled ILCs
except NKp46+RORγtlo ILC3 cells, although a heat map revealed that
most of these genes were also expressed at lower levels by other siLP
c
a
Spleen
NK
siLP
NKp46+NK1.1+CD127–
Liver
NKp46+NK1.1+CD49b+TRAIL–
siLP
RORγ t NKp46 NK1.1 CD127
Spleen
–
+
+
+
+
silEL
Liver
d
+
NKp46 NK1.1 CD127
–
5
NKp46+NK1.1+CD49b– TRAIL+
© 2015 Nature America, Inc. All rights reserved.
npg
silEL
RORγ t– NKp46+NK1.1+CD127+
PC3 = 10.25%
ILC1
ILC3
NK
–5
–10
–15
–20
NKp46+NK1.1+
–25
ILC2
siLP
+
+
+
–30
30
+
CD127 Sca-1 ST2 KLRG1
ILC2
20
10
PC1 =
siLP
siLP
NK.CD127 .Sp
NK.CD49b+ .Lv
NK.CD127– .SI
ILC1.IEL.SI
+
ILC1.CD127 .Sp
ILC1.CD49b– .Lv
ILC1.CD127+ .SI
ILC3.NKp46+ .RORγtIo.SI
+
hi
ILC3.NKp46 .RORγt .SI
– –
ILC3.NKp46 .4 .SI
ILC3.NKp46– .4+ .SI
ILC2.SI
ILC1
0
siLP
Spleen
–
10
Liver
T CD3+
B CD19+
–
NK CD127 SI
ILC1 CD127+ SI
+
lo
ILC3 NKp46 RORγt SI
+
hi
ILC3 NKp46 RORγt SI
–
–
ILC3 NKp46 CD4 SI
–
+
ILC3 NKp46 CD4 SI
ILC2 SI
ILC1 IEL
NK CD49b+ Lv
ILC1 CD49b– Lv
–
NK CD127 Sp
ILC1 CD127+ Sp
RORγ t+ NKp46– CD4–
+
–
hi
+
–40
0
0
–10
49.68%
–20
–30
40
20
2=
PC
–20
54%
20.
e
+
RORγ t NKp46 CD4
ILC3
siLP
RORγ t NKp46
siLP
RORγ tIo NKp46+
b siLP
CD127+
ILC1
siLP
ILC2
siLP
CD127–
NK
silEL
ILC1
siLP
hi
RORγ t
ILC3
siLP
RORγ tIo
ILC3
siLP
CD4–
LTi-like
siLP
CD4+
LTi-like
Liver
+
CD49b
NK
Liver
–
CD49b
ILC1
Spleen
–
CD127
NK
Spleen
+
CD127
ILC1
NK.CD127– .Sp (1)
NK.CD49b+ .Lv (1)
NK.CD127– .Sp (2)
NK.CD127– .Sp (3)
NK.CD49b+ .Lv (2)
NK.CD49b+ .Lv (3)
ILC1.CD127+ .Sp (1)
ILC1.CD127+ .Sp (2)
ILC1.CD127+ .Sp (3)
ILC1.CD49b– .Lv (1)
ILC1.CD49b– .Lv (2)
ILC1.CD49b– .Lv (3)
ILC1.IEL.SI (1)
ILC1.IEL.SI (2)
ILC1.IEL.SI (3)
NK.CD127– .SI (1)
NK.CD127– .SI (2)
ILC1.CD127+ .SI (1)
ILC1.CD127+ .SI (2)
ILC3.NKp46+ .RORγ tIo.SI (1)
ILC3.NKp46+ .RORγ tIo.SI (3)
ILC3.NKp46+ .RORγ thi.SI (3)
ILC3.NKp46+ .RORγ tIo.SI (2)
ILC3.NKp46+ .RORγ thi.SI (2)
ILC3.NKp46+ .RORγ thi.SI (1)
ILC3.NKp46– .4– .SI (1)
ILC3.NKp46– .4– .SI (2)
ILC3.NKp46– .4+ .SI (2)
ILC3.NKp46– .4– .SI (3)
ILC3.NKp46– .4+ .SI (1)
ILC2.SI (1)
ILC2.SI (2)
Expression (relative)
Row min
Row max
Figure 1 Analysis of ILC frequency and diversity. (a) Sorting strategy for array analysis after gating on live CD45 + and CD3−CD19− cells. (b) Cytospins of
cells using sorting strategies presented in a for each population (above and below images). Original length (of each panel), 14.6 µm. (c) Flow cytometry
of cells from various tissues, showing the frequency of ILCs among the lymphocyte population. The gating strategy in a was used to distinguish CD3+
T cells, CD19+ B cells and ILCs within the same sample, except the siLP ILC2 cell marker KLRG1 and the spleen ILC1 cell marker CD27 were
excluded and liver ILCs were distinguished with NKp46, CD49b and CD49a. T, T cell; B, B cell; NK, NK cell; SI, small intestine; Lv, liver;
Sp, spleen. (d) PCA of gene expression by subsets of ILCs and NK cells. Numbers along axes indicate relative scaling of the principal variables.
ImmGen nomenclature: spleen NK, NK.CD127−.Sp; liver NK, NK.CD49b+.Lv; siLP NK, NK.CD127−.SI; siIEL ILC1, ILC1.IEL.SI; spleen ILC1, ILC1.
CD127+.Sp; liver ILC1, ILC1.CD49b−.Lv; siLP ILC1, ILC1.CD127+.SI; siLP NKp46+RORγtlo ILC3, ILC3.NKp46+.RORγtlo.SI; siLP NKp46+RORγthi ILC3,
ILC3.NKp46+.RORγthi.SI; siLP NKp46−CD4− LTi-like ILC3, ILC3.NKp46−.4−.SI; siLP NKp46−CD4+ LTi-like ILC3, ILC3.NKp46−.4+.SI; and siLP ILC2,
ILC2.SI. (e) Hierarchical clustering of subsets of ILCs and NK cells based on the 10% of genes with the greatest variability. ImmGen nomenclature
as in d. Data are representative of two independent experiments (b,c) with n = 1–2 mice per tissue (b) or n = 2–4 mice per tissue (c) or are pooled
from one to three experiments per sample with cells pooled from 3–5 mice each (d,e).
308
VOLUME 16 NUMBER 3 MARCH 2015 nature immunology
resource
subsets (Fig. 2c). Liver and splenic ILC1 cells each expressed some
characteristic transcripts with a change in expression of greater than
twofold relative to their expression in all other subsets (Fig. 2d,e),
h
N
K.
C
N D1
K. 2
N CD 7 .–S
K. 49 p
IL CD b .+
C 12 Lv
1
IL .IE 7 .–
C L. S
IL 1.C SI I
C D1
1
IL .CD 27 +
C
1. 49 .S
IL CD b .– p
C 12 Lv
3
IL .N 7 +
C Kp .S
IL 3.N 46 + I
C K
IL 3.N p46 .+RO
C K
R
IL 3.N p4 .RO γ t lo
C K 6 – R .S
2. p .4 – γ I
SI 46 – . t hi
.4 +SI .SI
.S
I
N
K.
C
D
K. 12
C
N D 7 .–
K. 49 Sp
IL CD b .+
C 12 L
IL 1.IE 7 – v
C L .S
IL 1.C .SI I
C D
IL 1.C 127 +
C D4
1
.S
IL .CD 9b – p
C 1 .L
3 2
IL .NK 7 + v
.
C
3 p S
IL .N 46 + I
C Kp .R
3
IL .N 46 + O
R
C K
IL 3.N p4 .RO γ t lo
C K 6 – R .S
2. p4 .4 – γ I
SI 6 – . t hi
.4 +SI .SI
.S
I
c
N
Hist1h2bb
Gpx8
Rac3
2010012O05Rik
Cym
48314216l19Rik
Cma1
Nrarp
Gm10522
Cyp17a1
Mcam
Cdc20b
Lair1
Sord
As3mt
Itgam
Cmklr1
Car5b
St5
Prr5l
S1pr5
A430084P05Rik
Acpl2
Zeb2
Klra1
Nmbr
Fah
B3galt2
Gjb2
Upp1
Dpep2
Gstm5
Eps8l3
II1r2
Gda
Ryk
Capg
Dusp4
Tmtc2
A830080D01Rik
Ccne2
Expression (relative)
Row min
Row max
C
1.
C
D
49
b
–
.L
v
d
i
N
1.C
C
K.
.S
IL
C
D
Row max
p
Row min
+
e
12
7
–
.S
I
Expression (relative)
D
1
3 27
IL .NK .+S
C
3. p4 I
IL NK 6 .+
C
3 p R
IL .N 46 + OR
C Kp .R γ lo
IL 3.N 46 – OR t .S
C K 4– γ l
2. p4
SI 6 – .S t h. i
Sl
.4 + l
.S
l
Row max
C
Tnfsf10
Cd200r2
LOC100503878
Cd3g
IL
IL
Expression (relative)
Row min
D
12
7
N
Hes1
Pdcd1
Ccl1
Rad51l1
Lrrc52
Slc7a8
ll9r
Pparg
ll4
Ccr8
4930412O12Rik
Ptgir
A630001O12Rik
Ppp2r3a
Alox5
Rxrg
Hs3st1
ll17rb
ll13
ll1rl1
Ccr4
Dgat2
Stxbp6
ll5
Neb
Tph1
Mc5r
Calca
ll6
Ptpn13
Lpcat2
Ar
Acer2
Cybb
C
1.C
Expression (relative)
Expression (relative)
Row max
IL
H2-Ob
Sidt1
Cd22
Treml2
Gpr15
Dapl1
Cd3e
Cd3d
Expression (relative)
Row min
Row max
IL
C
1.I
E
L.
SI
f
siLP
Itgae
Expression (relative)
Row min
Row max
C
D
12
7–
.S
I
g
K.
Serpine2
Fermt2
Sdc2
Cacna1g
Slc6a7
Parvb
Dsg2
Olfr1342
Gm626
Gucy1a3
Igfbp4
Cxcr5
P2ry1
Ccr6
Ptges
Cntn1
Stom
Nrp1
Gria3
Hmgn3
Cd4
Row min
Cd83
Pde7b
Csf2
Ramp3
Cnnm2
Rora
Epas1
Lztfl1
Irak3
Gjb2
Tiparp
Maff
Atf3
Fam110a
Odc1
Rnf19b
Idi1
Ern1
Ifrd1
Tex2
Pcyt1a
Dgat1
Nrip1
Per1
Tgif1
Nr4a1
Bhlhe40
Csrnp1
Nfkbia
Sik1
Irs2
Cd69
Dtx4
Arrdc4
II2ra
N
b
Row max
IL
C
3
IL .NK
C
3. p46
–
N
Kp .4 –
46 .S
–
.4 + I
.S
I
Row min
npg
© 2015 Nature America, Inc. All rights reserved.
N
K.
C
D
K. 12
C 7 –
N D
K. 49 .Sp
IL CD b .+
C 12 Lv
1
IL .IEL 7 .–
C
S
1 .S I
IL .CD I
C
1
1
IL .CD 27 +
C
1. 49 .Sp
IL CD b .–
C 12 Lv
3
IL .NK 7 .+
SI
C
3 p
IL .N 46 +
C Kp .R
3
IL .NK 46 + OR
C
3 p .R γ t lo
IL .N 46 – O .S
C Kp .4 – Rγ I
2. 4
SI 6 – .S t h. i
.4 + I SI
.S
I
a
but we found no transcripts with relative expression greater than
twofold higher in siLP ILC1 cells. This suggested there were few
characteristic factors expressed by individual ILC1 subsets among
Tfam
Vegfa
Myc
Ccbl2
Expression (relative)
Expression (relative)
Row min
Row max
Row min
Row max
Figure 2 Characteristic transcripts of individual ILC subsets. Transcripts upregulated more than fourfold (red font) or twofold (black font) in individual
subsets (a,d–g), two similar subsets (b,c,h) or all subsets from siLP (i), based on pairwise comparisons between the index subset(s) (in bold above heat
map) and all other subsets (full gene lists, Supplementary Table 1). For similar ILC3 cell subsets consisting of NKp46 −CD4+ LTi-like and CD4− LTi-like
cells (b) or NKp46+RORγthi and NKp46+RORγtlo cells (c), some transcripts had expression that was more than fourfold higher (b) or twofold higher (c) in
one subset but not the other; in this case, blue indicates shared transcripts expressed by both subsets. Gray (b, right margin) indicates four transcripts
with twofold higher expression in CD4+ LTi-like ILC3 cells than in CD4− LTi-like ILC3 cells and fourfold higher expression in CD4 + LTi-like ILC3 cells
than in all other subsets. ImmGen nomenclature as in Figure 1. Data are pooled from one to three experiments per sample with cells pooled from three
to five mice each; two to three replicate samples are shown per subset.
nature immunology VOLUME 16 NUMBER 3 MARCH 2015
309
resource
npg
© 2015 Nature America, Inc. All rights reserved.
all ILCs and NK cells. Unexpectedly, the only transcript that showed
expression greater than twofold higher in siIEL ILC1 cells compared
to other ILC subsets was Itgae, which encodes CD103 (integrin αEβ7)
(Fig. 2f). Human IEL ILC1 cells express CD10316, but it was not previously known to be expressed by mouse IEL ILC1 cells, which lack
surface expression of CD103; this would suggest post-transcriptional
regulation of CD103 in mouse IEL ILC1 cells. Focusing on NK cells,
we identified four genes with expression that was twofold higher in
siLP NK cells than in all ILCs (Fig. 2g). Splenic and liver NK cells
expressed no characteristic transcripts. However, when we assessed
both liver and spleen NK cells as a group, we found 25 transcripts with
expression that was at least twofold higher than in all other subsets,
although siLP NK cells, IEL ILC1 cells and splenic ILC1 cells also
expressed these transcripts at lower levels (Fig. 2h). We concluded
that within our dataset, the mRNA profiles of ILC2 cells and LTi-like
ILC3 cells were unique, whereas the profiles of NKp46+ ILC3 cells,
ILC1 cells and NK cells showed considerable overlap.
A transcriptional signature shared by all siLP subsets
We next sought to determine whether there were any transcripts that
were expressed in all subsets from an individual tissue among siLP,
liver and spleen. Given that siLP NK cells and ILC1 cells were found to
cluster further from liver and splenic subsets by hierarchical clustering
and PCA, we focused on the siLP (Fig. 1d,e). Pairwise comparisons of
all subsets from the siLP with remaining subsets from liver and spleen
revealed that all siLP subsets expressed a core 35-transcript signature (Fig. 2i), which included several transcription factor–encoding
transcripts such as Rora, Atf3, Nr4a1, Maff, Epas1, Bhlhe40 and Per1.
Furthermore, all siLP-resident ILCs had high expression of transcript
encoding the activation marker CD69 and varied expression of Csf2,
which encodes the cytokine GM-CSF. Although the production of
GM-CSF by ILC2 cells18, ILC3 cells24,25 and NK cells26 is known,
its production by ILC1 cells has not been reported, to our knowledge. Thus, ILC subsets in the siLP seemed to be more activated than
ILCs in other tissues, probably because of their constant exposure
to varied environmental signals from the microbiome and incoming nutrients, including well-documented transcriptional activators such as vitamin A21,27 and ligands of the transcription factor
AhR (aryl hydrocarbon receptor)6,28.
Transcription factors, cytokines and chemokines of ILC subsets
To address broad patterns of gene expression among ILC subsets
and classes, we investigated the expression of previously reported
and other (unreported) transcription factors, chemokines, cytokines
and other secreted factors. The transcription factors with the highest
relative expression included the well-documented ILC-defining Id2
(encoded by Id2), the ILC3 class–defining RORγt (encoded by Rorc),
the ILC1- and NKp46+ ILC3-defining T-bet (encoded by Tbx21) and
the NK cell–defining Eomes (encoded by Eomes) (Fig. 3a). ILC2 cells
showed higher expression of the ILC2-defining transcription factors
GATA-3 (encoded by Gata3) and RORα (encoded by Rora), but these
were also expressed by all ILCs (Fig. 3a), consistent with an early
role in ILC development, at least for GATA-3 (ref. 19). Nfil3, which
encodes the transcription factor NFIL3 (also known as E4BP4), is
required for the development of NK cells29,30, ILC2 cells and ILC3
cells31,32, and had its highest expression in siLP subsets; NKp46+ ILC3
cells, ILC1 cells and NK cells showed the highest Nfil3 transcript levels, followed by LTi-like ILC3 and ILC2 cells, whereas much lower
Nfil3 transcript levels were present in liver and spleen ILC1 and NK
cells (Fig. 3a). Furthermore, two transcription factors identified in
the siLP signature, ATF3 (encoded by Atf3) and Nur77 (encoded
310
by Nr4a1) (Fig. 2i), were expressed at levels similar to those of
lineage-defining transcription factors (Fig. 3a). Collectively, these
data suggested a substantial role for the intestinal microenvironment in the expression of certain transcription factors, which might
subsequently have diverging roles among ILC classes. Analysis of
chemokines and their receptors (Fig. 3b), as well as of cytokines and
their receptors (Fig. 3c), revealed both shared and distinct expression patterns. Beyond the known signature cytokine and chemokine
circuitries, we identified a candidate feed-forward loop for ILC2 cells,
which expressed both the chemokine receptor CCR8 and its ligand
CCL1 (Fig. 3b). We also identified ILC2 cell expression of Bmp7 and
Bmp2, the latter of which encodes a protein known to modulate intestinal peristalsis by binding the BMP receptor on enteric neurons 33.
In addition, we found expression of Il2 in several ILC populations
(Fig. 3c), which suggested that ILCs might be able to activate T cells
or other ILCs through signaling via its receptor, IL-2R.
Shared and distinct expression profiles among siLP subsets
We next focused our analysis of transcriptional profiles on the four
major CD127+ ILC subsets within the siLP: ILC1 cells, ILC2 cells,
NKp46+ RORγthi ILC3 cells and CD4− LTi-like ILC3 cells (Fig. 3d).
Comparison of siLP ILC subsets revealed overlapping patterns of gene
expression that were not identified in individual subset signatures
(Fig. 3d and Supplementary Table 2). For example, ILC2 cells and
LTi-like ILC3 cells shared 17 transcripts, including Arg1 and Ret.
Arginase-1 (encoded by Arg1) marks fetal and adult ILCs and facilitates the identification of developing ILCs in the siLP34. The receptor
tyrosine kinase encoded by Ret is also expressed by fetal CD11c+
lymphoid tissue initiator cells and is required for the development of
Peyer’s patches35, but it has not, to our knowledge, been previously
reported to be expressed by fetal or adult ILCs, including LTi or
LTi-like ILC3 cells. Together these results suggested that, at least
in fetal mice, ILC2 cells and LTi-like ILC3 cells share a common
progenitor34, although their functional relevance in adult siLP
ILCs remains to be investigated. ILC1 cells and ILC2 cells shared
19 transcripts, including Ets2. Ets-1 and Ets-2 interact with proteins
of the Id family, and whereas Ets-1 has been linked to the early
development of NK cells36, Ets-2 has not. Thus, Ets-2 might be
relevant to the development or maintenance of ILC1 and ILC2 cells.
NKp46+ and LTi-like ILC3 cells have long been known to share
many characteristics, owing to their mutual production of IL-22
and expression of RORγt. However, NKp46+ ILC3 cells and NKp46+
ILC1 cells shared higher relative expression of a greater number of
transcripts (Tbx21, Ifng and Il12rb) than did NKp46+ ILC3 cells and
LTi-like ILC3 cells (Fig. 3d and Supplementary Table 2). Although
T-bet is required for the development of NKp46+ ILC3 cells8–10, these
cells produced little IFN-γ in response to IL-12 and IL-23 (data not
shown). Because IFN-γ is well documented as being post-transcriptionally regulated37, the presence of the Ifng transcript in NKp46+
ILC3 cells suggested that T-bet might be sufficient to induce transcription but that other factors are needed for protein production,
such as bacterial infections in vivo4. In addition to their shared transcripts, NKp46+ ILC3 cells and NKp46+ ILC1 cells had significantly
different expression (by greater than twofold) of 213 genes, with the
genes upregulated in NKp46+ ILC3 cells including many genes that
were also expressed at higher levels by LTi-like ILC3 cells (Fig. 3e).
Thus, NKp46+ ILC3 cells had a transcriptional profile with characteristics intermediate between those of NKp46− LTi-like ILC3 cells and
NKp46+ ILC1 cells, which might enable functional plasticity. Their
functional polarization toward ILC3 or ILC1 cells probably depends
on the tissue microenvironment.
VOLUME 16 NUMBER 3 MARCH 2015 nature immunology
resource
also had substantial expression of Il21r, which encodes a member of
the common γ-chain cytokine family (Fig. 3d). When we included an
additional comparison between ILC1 cells and NK cells in the siLP,
we found that ILC1 cells expressed only four transcripts at least twofold higher than other siLP subsets: Gpr55, Trat1, Mmp9 and Cpne7
(Supplementary Table 2). Thus, although ILC1 cells from the small
intestine were more like NK cells than were other ILC subsets from
the small intestine, they showed no obvious characteristic markers
when we included NK cells in our comparisons.
ILC2 & NKp46– CD4– ILC3
Io
Row max
ILC2
ILC2 & ILC1
NKp46– CD4– ILC3
ILC1
NKp46– CD4– ILC3 & NKp46+ ILC3
+256
17
+128
Ret
+64
Expression (fold)
+
hi
siLP ILC2/siLP NKp46 RORγ t ILC3
hi
N
K.
C
N D1
K. 2
C 7–
N D4 .S
K. 9 p
IL CD b .+
C 12 Lv
1. 7 –
IL IEL .S
. I
C
1. SI
C
D
IL
12
C
1
7+
IL .CD .S
C
1. 49 p
IL CD b .–
C
3. 127 +Lv
IL NK .S
C
3. p46 + I
IL NK .R
C
3 p4 O
IL .NK 6 .+ Rγ t
C
p R
3
IL .N 46 – OR
C K .4 γ t
2. p4
–
SI 6 – .S
.4 + I
.S
I
Row min
Heat map max
NKp46+ ILC3
Ccr8 II17rb 137
Ccr4
+
NKp46 ILC3 & ILC1
19
II5
Pdcd1
+32
npg
Expression (relative)
Expression (relative)
d
II9r
II4
II5
II1rI1
II13
II17rb
II6
Bmp2
Csf2
Areg
Bmp7
Ifngr2
II18r1
II1r1
II17re
II22
II23r
II7r
II2ra
II17f
II17a
II1r2
II2
II12rb1
Prf1
Gzma
II17ra
Ifngr1
II4ra
Ifng
II12rb2
II2rb
II21r
II17b
II17c
Cxcr4
Ccl4
Ccl3
Ccl5
Ccr5
Cx3cr1
Cxcr3
CcI1
Ccr8
Ccr4
CcI17
Ccr6
Cxcr5
Ccr9
Cxcr6
Ccr1
II10ra
II4
Calca
Cd27
Cd81
Expression (relative)
+16
Arg1
+8
Vipr2
+4
–16
–32
10–1
CcI5
Cxcr5
Klrd1
II1r1
Igfbp7
–64
–128
–256 61
II22
Rorc
II23r
–256 –128 –64 –32 –16
Tbx21
Upp1
Chad
CtIa4 Ifng
17
–8
–4
–2
+1
+2
+4
+8
CcI3
XcI1
–
–
siLP ILC1/siLP NKp46 CD4 ILC3
10
–3
Row max
94
CcI5 KIra10
Cd27
Gzma
Rorc
II23r
Cxcr5
10–5
Igfbp7
II21r
Itgax
10–4
119
II22
Upp1
Cd2
KIrd1
122
+16 +32 +64 +128 +256
Expression (fold)
10
–2
II12rb2
Ncr1
KIrb1f
100
75
Prf1
Stom
–4
e
Itgax
Ptges
–2
Gzma
II21r
Cntn1
+1 109
Row min
Myc
Nrp1
+2
–8
Ets2
P
© 2015 Nature America, Inc. All rights reserved.
Id2
Rara
Tox2
Rorc
Tox
Nfil3
Maff
Ahr
Nr4a1
Atf3
Rora
Zbtb16
Rxrg
Pparg
Gata3
Gfi1
Hnf1a
Tbx21
Notch1
Irf8
Eomes
Heat map min
c
.S
I
.S
I
b
N
K.
C
N D12
K.
7–
C
N D .S
K. 49 p
C
IL D b .+
C 12 Lv
1. 7 –
IL IEL .S
C
1 .S I
IL .C I
C D1
1. 2
IL C 7 +
C D .
1 4 S
IL .C 9b – p
C D1 .L
3
IL .N 27 + v
C Kp .S
3. 4 I
IL NK 6 +
C
p .R
IL 3.N 46 + OR
C K
3. p4 .RO γ t Io
IL N 6 – R .
S
C Kp .
2. 4 4 – γ t hi I
SI 6 – .S .
.4 + I SI
.S
I
a
N
K.
C
N D1
K. 27 –
C
N D .S
K. 49 p
C b+
IL D1 .L
C
1. 2 7 – v
IL IEL .SI
C
1. .SI
IL CD
12
C
1
IL .CD 7 .+S
C
4
1
9 p
IL .CD b –
.
C
3. 12 Lv
IL NK 7 .+
S
C
3. p46 + I
IL NK .R
p4 O
C
3
R
IL .NK 6 .+R γ t Io
C
O
.S
3 p
IL .NK 46 – Rγ I
h
.
C
2. p4 4 .– t .Si
SI 6 – SI
I
.4 +
.S
I
As discussed above, a pairwise comparison did not identify characteristic transcripts in siLP ILC1 cells compared to all profiled ILC
subsets. However, in comparisons only to siLP ILC2 cells and ILC3
subsets, we found 75 transcripts expressed at least twofold higher
in ILC1 cells (Fig. 3d). For example, ILC1 cells demonstrated a
greater cytotoxic capacity, as indicated by their expression of Gzma
and Prf1 (Fig. 3d), which respectively encode granzyme A and perforin, although this might have been due to an imperfect distinction
between siLP ILC1 cells and NK cells (discussed below). ILC1 cells
II1r1
II1r2
1300014I06Rik
Lrrc49
–64 –32 –16 –8 –4 –2 +1 +2 +4 +8 +16 +32 +64
Fold change
siLP NKp46+RORγ t hi ILC3/siLP ILC1
Figure 3 Spectrum of distinct and shared transcriptional profiles among ILC subsets. (a) Heat map of the expression of selected transcription factors,
normalized by expression within the entire heat map. (b,c) Heat maps of the expression of selected chemokine receptors and ligands (b) and of cytokine
receptors and ligands (c), normalized by expression within each row. (d) Transcripts upregulated by a single subset or two subsets among selected siLP
ILCs (colors in plot match key; full gene lists, Supplementary Table 2). Numbers in plot indicate transcripts upregulated by at least twofold in that
subset (colors match key). (e) Comparison of gene expression in RORγthiNKp46+ ILC3 cells (n = 3 replicates) versus that in ILC1 cells (n = 2 replicates)
by volcano plot; numbers in corners indicate transcripts significantly upregulated by at least twofold. P ≤ 0.05 (t-test). ImmGen nomenclature as in
Figure 1. Data are pooled from one to three (a–c) or one to two (d,e) experiments per sample with cells pooled from three to five mice each; two to three
replicate samples are shown per subset (a–c).
nature immunology VOLUME 16 NUMBER 3 MARCH 2015
311
resource
most highly and most significantly upregulated by CD4− LTi-like ILC3
cells, we found Nrp1. Also identified as part of our LTi-like ILC3 cell
signature (Fig. 2b), Nrp1 is a coreceptor for several ligands, including
the immunoregulatory factor VEGF, transforming growth factor-β1
and semaphorins, and may have a function in negatively regulating
the immune response, in part through enhanced survival of regulatory T cells38. It is also one of a few markers that distinguish natural regulatory T cells from peripherally generated mucosa-derived
induced regulatory T cells39,40. However, to our knowledge, until now
Nrp1 has never been reported to be expressed by LTi-like ILC3 cells.
We confirmed by flow cytometry that Nrp1 was expressed in greater
amounts in LTi-like ILC3 cells than in other siLP subsets (Fig. 4c).
We also found higher expression of CD25 protein in LTi-like ILC3
cells than in other siLP subsets (Fig. 4c). Gating on Nrp1+CD25+ cells
among CD3−CD19− siLP lymphocytes resulted in a cell population
with considerable enrichment for LTi-like ILC3 cells (Fig. 4d) and
production of IL-22 in naive IL-22 reporter mice41 (Fig. 4e). Among
Defining novel transcripts within ILC3 cells
Pairwise comparison of all ILC3 subsets versus other profiled cells
revealed a 42-transcript ILC3 cell signature (Fig. 4a). This signature included well-studied transcripts such as Il23r, Rorc and Il22,
as well as several molecules previously unreported to be expressed
by ILC3 cells, such as Pram1, a target of retinoic acid and activator
of the kinase Jnk. This result supported the role for retinoic acid in
the development of ILC3 cells21,27. Notably, Il17 was not among the
transcripts with expression more than twofold higher in ILC3 cells
than in other ILCs, nor was it expressed uniquely by any individual
ILC3 cell subset (Fig. 2b,c). This result suggested that IL-17 was not
a major product of ILC3 cells in the small intestine, at least in young
adult mice at steady state.
Comparison of the major adult LTi-like population, CD4 − LTilike ILC3 cells, with NKp46+RORγthi ILC3 cells revealed a total of
508 genes with a significant difference in expression of greater than
twofold (Fig. 4b and Supplementary Table 3). Among the transcripts
Row min
Row max
f
213
295
10
0
3
1
–1
10
10–3
Nrp1
P2yr1
Ncr1
–7
–128 –64 –32 –16 –8 –4 –2 +1 +2 +4 +8 +16 +32
Expression (fold)
NKp46–CD4– ILC3/NKp46+RORγ thi ILC3
Cd4
–3
10
Cd81
Fasl
Ncr1
+4
–6
Zfp319
10–2
Ptges
ll12rb2 Tbx21
Parvb
+2
10–5
10–1
–2
+1
Ccr6
Ret
+8
+1
6
+3
2
Vipr2
–4
Sl6a20a
10–3
10–4
c
Klri2
Ccl3
2
–1
6
–8
10–2
P
10
Expression (fold)
NKp46–CD4– ILC3/NKp46–CD4+ ILC3
g
100
11
3
NK
ILC1
+
Io
NKp46–CD4+ ILC3
Nrp1
d
NKp46–CD4+ ILC3
CD25
10
10
104
14.5
NKp46
103
2
10
0
0 102
CD25
103 104 105
0.073
5.54
105
10
104
103
103
102
0 0.097
10
0
0 102
RORγ t
103 104 105
Cxcr4
10
Ccl5
10–4
–4
–2
+1
+2
+4
+8
+
4
94.3
Vegfa
Klrd1
–3
Expression (fold)
NKp46 RORγ thi ILC3/NKp46+RORγ tlo ILC3
e
5
Cxcr5
Ccr5
–8
Gated on CD45+CD3–CD19–
5
Nrp1
NKp46 RORγ t ILC3
NKp46+RORγ thi ILC3
NKp46–CD4– ILC3
Cells
Cells
NKp46+RORγ thi ILC3
NKp46–CD4– ILC3
Io
10–2
3.1
100
2.59
80
60
40
2
85.8
0 102
CD25
8.52
103 104 105
Cells
NK
ILC1
NKp46 RORγ t ILC3
+
P
10–1
Nrp1
Expression (relative)
100
P
K.
C
N D1
K. 2
C 7–
N D4 .Sp
K. 9
C b+
IL D1 .L
C 2 v
1 7–
IL .IEL .S
C
1. .SI l
IL CD
C
1 12
IL .CD 7 .+
C
1. 49 Sp
IL CD b –
C
.
3 12 Lv
IL .NK 7 .+
C
S
p
3
4 I
IL .NK 6 .+
C
R
3. p4 O
IL NK 6 + Rγ
C
p4 .RO t Io
3
IL .N 6 – R .S
C K
2. p4 .4 – γ t hi I
SI 6 – .
S .SI
.4 + I
.S
I
b
N
ll1r1
GIt25d2
Smyd2
Rhobtb1
Ncoa7
ll22
Chad
Tns3
Scpep1
Nr1d1
Ly86
Dlg5
Pram1
Zfp57
Tnfrsf21
Prelid2
Ablim3
Lipo1
Cd93
Rorc
Lingo4
Eps8l3
Tifa
Fnbp1l
Gpr157
St3gal3
Serinc2
Ras/11a
lgfbp7
Mapk10
Asphd2
II17re
II23r
Tacstd2
Ucp3
Kcnk1
Large
Cxcr5
Lrrc49
Ccr1
Hs6st2
Xkrx
npg
© 2015 Nature America, Inc. All rights reserved.
a
20
0
0 102 103 104 105
IL-22 reporter
Figure 4 ILC3-specific transcriptional programs and cell-surface markers. (a) Heat map of mRNA transcripts upregulated more than fourfold (red font)
or twofold (black font) in RORγt+ ILC3 cell subsets relative to their expression in all other subsets. ImmGen nomenclature as in Figure 1. (b) Comparison
of gene expression in NKp46+RORγthi ILC3 cells with that in CD4− LTi-like ILC3 cells (n = 3 replicates each), by volcano plot; numbers in corners
indicate transcripts significantly upregulated (P ≤ 0.05 (t-test)) by at least twofold (colors in plot match those below plot; full gene lists, Supplementary
Table 3). (c) Expression of Nrp1 and CD25 by siLP NK cell, ILC1 cell, NKp46 +RORγtlo ILC3 cell, NKp46+RORγthi ILC3 cell, NKp46−CD4− LTi-like ILC3
cell and NKp46−CD4+ LTi-like ILC3 cell subsets from RORγteGFP reporter mice. (d) Percentage of siLP cells from C57BL/6 wild-type mice electronically
gated on CD45+ and CD3−CD19− cells that expressed both CD25 and Nrp1 (left), and extent of LTi-like ILC3 phenotype among these cells (from left
outlined area) based on surface NKp46 and intracellular RORγt stainings (right). Number adjacent to outlined area (left) indicates percent Nrp1 +CD25+
cells; numbers in quadrants (right) indicate percent cells in each throughout. (e) Production of the IL-22 reporter (right) by cells obtained from naive
IL-22 reporter mice and shown (left) with gating on CD45 +, CD3−CD19− and NKp46− cells (colors of lines (left) match those in quadrants (right)).
(f,g) Comparison of gene expression in CD4+ LTi-like cells (n = 2 replicates) with that in CD4− LTi-like ILC3 cells (n = 3 replicates) (f) and in
NKp46+RORγthi ILC3 cells with that in NKp46+RORγtlo ILC3 cells (n = 3 replicates each) (g) by volcano plot (additional information, Supplementary
Table 3). P ≤ 0.05 (t-test). Data in c–e are representative of three experiments.
312
VOLUME 16 NUMBER 3 MARCH 2015 nature immunology
resource
CD127
Cd200r2
Klra1
68
Sell
–1
10
10–2
–3
10
106
CXCR6
ll7r
Gzma
Transcriptional differences between ILC1 cells and NK cells
One subject that has been particularly controversial is the difference
between ILC1 cells and NK cells, in part because of a lack of markers
characteristically and consistently expressed in NK cells and ILC1 cells
in various organs. We used different sorting strategies in each tissue
to discriminate between ILC1 cells and NK cells, consistent with what
has been reported before2,42. We achieved the best separation of subsets in the liver and the worst such separation in the siLP (Fig. 5a).
Comparisons of NK cells and ILC1 cells from liver, spleen and siLP
reflected the degree of separation between populations during sorting,
with the greatest number of significantly differently expressed genes
in the liver and the least in the siLP. Nonetheless, ILC1 replicates clustered together (Fig. 1e) and were transcriptionally distinct from NK
cells in liver, spleen and siLP (Fig. 5b–d and Supplementary Table 4).
As expected, Eomes expression was significantly different, with a
difference in expression of greater than twofold in NK cells relative
to its expression in ILC1 cells in all tissues analyzed (Fig. 5b–d and
Supplementary Table 4). The amount of transcripts encoding proteins of the cytotoxic machinery was generally greater in NK cells
than in ILC1 cells (Fig. 5b–d). However, in the siLP, the number of
perforin-encoding transcripts was only 1.7-fold greater in NK cells
than in ILC1 cells, and in the liver, only granzyme K was expressed at
nature immunology VOLUME 16 NUMBER 3 MARCH 2015
Expression (fold)
Spleen ILC1/Spleen NK
e
0
10 191
–1
10
10–2
Lmo4
Npas2
10–4
–3
10
10–4
ltga1
Gzmc
Gzmb
Cmah
S1pr1
Mmp9
Entpd1
Sell
Expression (fold)
Liver ILC1/Spleen ILC1
Expression (fold)
siLP ILC1/siLP NK
f
73
Itga2
Cxcr4
Kit
Eomes
10–5
Tmem64
Ahr
Prf1
ll2ra Tcrg-v3
Eomes
Klrg1
Klra10
Klra9
ll7r
–6
–34
2
–1
6
–8
–4
–2
+1
+2
+4
+8
+1
+36
+62
4
0
10
Ltb
Eomes
Klrg1
Expression (fold)
Liver ILC1/Liver NK
d
126
Gzma
Gzmb
ltga1
Spleen
Liver
100
siLP
Isotype
NK
ILC1
80
60
40
Cells
LTi-like ILC3 cells, distinctions in gene expression between CD4+ and
CD4− subsets were not robust and probably do not have functional
significance (Fig. 4f), although they might reflect different developmental lineages5,9. Similarly, comparison of siLP NKp46+RORγtlo
ILC3 cells with siLP NKp46+RORγthi ILC3 cells revealed substantial
overlap in gene expression, with few significant differences (Fig. 4g
and Supplementary Table 3). The greater expression of NK cell–like
transcripts such as Ccl5 and Klrd1 in siLP RORγtlo ILC3 cells than in
siLP RORγthi ILC3 cells suggested that the population of siLP RORγtlo
ILC3 cells might have included a minor ILC3 cell population ‘converting’ into ‘ex-RORγt+’ ILC3 cells. Robust identification of ‘converting’
ILC3 cells can be established only through fate mapping experiments,
as has been described4.
P
10–4
10
104 105
Figure 5 Transcripts expressed differently by NK cells versus ILC1 cells.
(a) Gating strategy for liver, spleen and siLP ILC1 and NK cells after the
first round of sorting from pooled samples. (b–e) Comparison of gene
expression in liver ILC1 cells with that in liver NK cells (n = 3 replicates
each) (b), in spleen ILC1 cells versus spleen NK cells (n = 3 replicates
each) (c), in siLP ILC1 cells versus siLP NK cells (n = 2 replicates each) (d),
and in liver ILC1 cells versus spleen ILC1 cells (e), by volcano plot;
numbers in corners indicate transcripts significantly upregulated
(P ≤ 0.05 (t-test)) by at least twofold (colors in plot match those below
plot; full gene lists, Supplementary Table 4). (f) Expression of CD127 and
Eomes in ILC1 cells and NK cells from liver, spleen and siLP, as well as
staining with isotype-matched control immunoglobulin (Isotype).
Data are representative of four to six experiments (a), are pooled from
one to three (b–e) experiments per sample with cells pooled from three
to five mice each, or are representative of two experiments (f).
Cd200r1
Sell
P
103
10–3
57
–6
4
–3
2
–1
6
–8
–4
–2
+1
+2
+4
+8
+1
6
+3
2
+6
4
P
0 102
Gzmc
6
104 105
Klrg1
+1
CD27
103
–5
10–2
Eomes
–1
2
–68
–34
2
–1
6
–8
–4
–2
+1
+2
+4
+
+18
+3 6
+2
+164
28
0 102
10–4
102
0
–1
CXCR6
6
104 105
86.8
10–3
100
10
Tnfsf10
20
0
0 10 10
CD127
2
3
10
4
10
5
2
3
10
4
105
100
80
60
40
Cells
© 2015 Nature America, Inc. All rights reserved.
CD49b
103
102
0
Prf1
10–2
–1
72.7
64.9
103
NK1.1
102
0
0 102
npg
8.95
103
CD127
TRAIL
103
27.6
+4
+8
104
c
ll7r
+2
10
104
10–1
5
175
–2
10
104
5
100 190
+1
105
21
b
siLP
–4
Spleen
–8
Liver
P
a
20
0
0 10 10
Eomes
higher levels by NK cells than by ILC1 cells (Supplementary Table 4).
In fact, granzyme A and granzyme C had higher expression in liver
ILC1 cells than in liver NK cells (Fig. 5b and Supplementary Table 4),
consistent with the reported cytolytic activity of liver ILC1 cells43.
Liver and spleen ILC1 cells selectively expressed transcripts encoding known ILC1 cell-surface markers. In liver ILC1 cells, these transcripts included Itga1 and Tnfrsf10 (Fig. 5b), which encode CD49a
and TRAIL, markers that have been used for separating liver ILC1
cells from NK cells40. Splenic ILC1 cells had high expression of IL2ra
and IL7r (Fig. 5c), which encode cytokine receptors that collectively
enable this population to escape regulation by regulatory T cells15. The
transcripts with the most significant higher expression in siLP ILC1
cells than in siLP NK cells included only previously uncharacterized
transcripts such as Tmem64, Npas2 and Lmo4; however, these transcripts were expressed in other ILCs of the small intestine (Fig. 5d)
and therefore probably would not be useful markers. Comparison
of splenic and liver ILC1 cell subsets revealed that splenic ILC1 cells
expressed markers of immature and mature NK cells, including CXCR4,
c-Kit and Eomes13,42 (Fig. 5e). These differences in transcription could
be explained by the finding that our sorting strategy based on CD127
and CD27 included an Eomes+ subset among the splenic ILC1 cells
(Fig. 5f). We also noted that the siLP NK cell subset identified by
313
Expression (fold) spleen ILC1/spleen NK
+32
Tnf
+4
Tnfsf10
+2
–
12
7
D
D
C
N
C
K.
Eomes
–4
Klra1
–8
Serpinb9b
–16
Klra3
Klrg1
Khdc1a
–32
15
36
siLP ILC2
5
10 99.9
39.8
0.32
2
3
4
ST2
2
2
10
0
0
2
5
0 10 10 10 10
3
4
Sp NK
Lv ILC1
0.10
98.5
4
10
4
10
3
10
3
10
2
10
0
0.03
2
5
3
4
5
10
100
Chr 13
0
qA1 qA2 qA3.2 qA4
4
2
10
0
0
5
2
3
4
0 10 10 10 10
0
2
3
4
0 10 10 10 10
5
10
2
10
0
10
0
0.07
3
2
4
5
0 10 10 10 10
Intracellular anti-TCRδ
5
4
10
3
10
3
10
10
0
4
2
10
0
5
*
*
*
40
20
4
2
5
80
75
70
65
0
2
3
4
0 10 10 10 10
Cxcr6
eGFP/eGFP
Cxcr6
eGFP/+
5
0
2
3
4
0 10 10 10 10
Liver NK CD49b (1)
Liver NK CD49b+(2)
[0 – 160]
Liver ILC1 CD49a+(1)
+
Liver ILC1 CD49a (2)
RefSeq annotation
5
tcrgv2
g
P
I
si NK
LP p4 LC
N 6– 2
Kp IL
46 + C3
si ILC
lE
3
L
si ILC
LP 1
IL
si C1
LP
Li
ve NK
rI
L
Li C1
v
e
Sp r
le NK
en
Sp IL
le C1
en
N
K
tcrgv1 tcrgj4
siLP NK
sILP NK & ILC1
siLP ILC1
+4
10
si
L
+
2
20
P
+ 3
si ILC
lE
3
L
I
si LC
LP 1
IL
si C1
LP
Li
N
ve K
rI
L
Li C1
Sp ver
le NK
en
Sp IL
le C1
en
N
K
2
C
C
IL
46
IL
–
N
Kp
P
46
si
L
Kp
N
P
si
L
qD2.2
+
[0 – 75]
[0 – 125]
3
10
0
0
0
P
CD3–CD19– cells (%)
f
60
si
L
0.03
3
2
4
10
0
0 10 10 10 10
0 10 10 10 10
si
L
CXCR6eGFP+ cells (%)
Spleen NK CD49b (2)
0
5
10 100
0
10
3
3
Lv NK
Lv ILC1
100
4
10
2
Sp NK
0.04
5
4
2
Sp ILC1
3.81
10
10
100
80
Spleen NK CD49b+(1)
[0 – 111]
Expression
(fold) IEL ILC1/siLP NK
3
10
0
NKp46
10
100
NKp46
4
10
e
5
IL7R
10
ST2
Extracellular
anti-TCRδ
siLP ILC2
29.8
[0 – 109]
[0 – 102]
IEL γδ
10.2
5
qD1
26 kb 19,428 kb 19,430 kb 19,432 kb 19,434 kb 19,436 kb 19,438 kb 19,440 kb 19,442 kb 19,444 kb 19,44
2
10
0
5
qC3
20 kb
Extracellular anti-TCRδ
10
qB1 qB3 qC1
3
0 10 10 10 10
0 10 10 10 10
Row max
d
Lv NK
1.54
5
10
10
3.7
Expression (relative)
Row min
10
3
3
10
0
5
10
10
10
10
4
4
10
Sp ILC1
0.06
+64 +128
NKp46
IEL γδ
54.8
5
10
+16 +32
–8
–4
–2
1
+2
+4
+8
Expression (fold) liver ILC1/liver NK
NKp46
c
–32 –16
IL7R
–128 –64
CD3 & CD19
Itga1
1
–2
–64
npg
Atp8a2
Socs2
Tcrg-V3
Tcrg-V2
II2ra
II7r
Ckb
Gpr97
Podnl1
Tmem176a
Tmem176b
Cdon
Gpr114
Tmem154
Slc27a6
St6gaInac3
Cxcr6
Klrg1
Khdc1a
Serpinb9b
Cym
Zeb2
A430084P05Rik
CmkIr1
Car5b
Itgam
Klra1
Eomes
Klra3
Klra9
Klra10
Irf8
Ahr
+8
–128
© 2015 Nature America, Inc. All rights reserved.
59
Tcrg-C
CXCR6
II2ra
+16
17
Tcrg-V2 Tcrg-V3
Tmem176b
IL7r
Tmem176a
K.
+128
+64
b
Spleen & liver ILC1
Spleen, liver & siLP ILC1
N
Spleen & liver NK
Spleen, liver & siLP NK
N
a
.S
P
4
K. 9b +
C
.
L
D
IL
C 12 v
1. 7 –
I
IL EL. .SI
C
1. SI
C
D
IL
1
C
1. 27 +
C
IL
D .Sp
4
C
1. 9b –
C
D .Lv
IL
1
C
3. 27 +
N
Kp .SI
IL
C
3. 46 +
.R
IL NK
O
p
C
3. 46 + Rγ I
IL NK .R t .oS
p4 O
C
I
3.
6– R
IL NK .4 – γt h. i
C
2. p46 – .SI SI
SI
.4 +
.S
I
resource
+2
Eomes Klra3
Sell
Serpinb9b
tcrgc4
IEL ILC1/siLP ILC1
IEL ILC1/siLP ILC1 & NK
IEL ILC1/NK
Itgae Irgm1
CD160
+1
–2
–4
–8
–16
II2ra
Icos
B3gaIt5
–8
Tcrg-V3
Tcrg-V2
Plekhg1
Serpinab1a
–4
–2
+2
Expression (fold)
IEL ILC1/siLP ILC1
+4
Figure 6 Generation of a core ILC signature distinct from that of NK cells. (a) Transcripts expressed differently by ILC1 cells relative to their expression
by NK cells from various tissues (colors in plot indicate transcripts upregulated by two or three subsets and match those in key); numbers in plot
indicate transcripts with a significant (P ≤ 0.05 (t-test)) difference in expression of at least twofold in liver and spleen ILC1 and NK cells (n = 3
replicates each) or with an additional filter for difference in expression of at least twofold in siLP ILC1 cells versus NK cells (n = 2 replicates each).
(b) Gene signatures generated according to data in a (colors on left match those in a); red font indicates transcripts upregulated by at least fourfold in
all three comparisons of ILC1 cells versus NK cells (full gene lists, Supplementary Tables 5 and 6). (c) Extracellular and intracellular staining of TCRδ
on (or in) ILCs and NK cells (electronically gated as in Fig. 1c, except splenocytes were positively selected with beads coated with anti-CD49b) and IEL
γδ T cells (positive control), assessed by flow cytometry; for intracellular staining, phycoerythrin-conjugated monoclonal antibody to TCRδ was used to
stain IELs extracellularly before intracellular staining with fluorescein isothiocyanate–conjugated anti-TCRδ. Numbers in outlined areas (gates) indicate
percent TCRδ+ cells (outline colors (red and blue) match those in keys above plots), except IEL γδ cells, which are shown as a percentage of CD45+ cells
(additional information, Supplementary Fig. 2). Sp, spleen; Lv, liver. (d) 3′ end targeted RNA-sequencing data from a publicly available data set (GEO
accession code GSE52043), showing the locus encoding TCRγ (chr13:19,426,000–19,446,000). (e) Frequency of CXCR6+ cells in each ILC subset
from Cxcr6eGFP/+ reporter mice. Each symbol represents an individual mouse (three or four per genotype per tissue); small horizontal lines indicate the
mean (±sem). *P ≤ 0.05 (one-way analysis of variance (ANOVA) with multiple comparisons). (f) Frequency of ILCs in Cxcr6eGFP/+ and Cxcr6eGFP/eGFP
mice. All comparisons are not significant (two-way ANOVA with multiple comparisons). (g) Gene expression in siLP ILC1 cells, siLP NK cells and siIEL
ILC1 cells; colors indicate transcripts upregulated by one or two subsets (match key above plot). Data are representative of one to three experiments
per sample with cells pooled from three to five mice each (a,b,g; n = 2–3 replicates each), two independent experiments (c,d) or three independent
experiments (e,f; error bars, mean ± sem of three or four mice per genotype per tissue).
314
VOLUME 16 NUMBER 3 MARCH 2015 nature immunology
resource
npg
© 2015 Nature America, Inc. All rights reserved.
cell-surface markers contained a large population of Eomes− cells
(Fig. 5f). Thus, NK cells and ILC1 cells could not be discriminated
on the basis of CD127 and/or CD27 in the spleen and siLP. The use of
EomeseGFP reporter mice might be useful in future attempts to better
discriminate NK cells from ILC1 cells, as has been described4,42.
A core, NK cell–distinct ILC signature
If ILC1 cells are different from NK cells as a class, we reasoned that
there should be transcripts common among ILC1 cells from all tissues that differ from those of NK cell subsets, and vice versa (Fig. 6a).
Spleen and liver ILC1 cells shared expression of genes previously
reported to be expressed by ILC1 cells,8,13,15,42,43 such as Tnfsf10, Tnf
and Il2 (Fig. 6a and Supplementary Table 5), although Il2 was filtered
out by our methods because of variability between replicates (Fig. 3c).
Transcripts with higher expression in NK cell subsets included Eomes,
Itgam (which encodes CD11b (integrin αM)) and members of the
Klra family of genes (which encode receptors of the Ly49 family)
(Supplementary Table 5). This NK cell–specific signature was consistent with that identified in a more limited data set comparing
NK cells with ILC1 cells and ex-RORγt+ ILC3 cells from the siLP4.
Visualization of genes expressed differently in the liver, spleen and
siLP for the entire data set revealed that with few exceptions, genes
upregulated in ILC1 cells relative to their expression in NK cells had
even higher expression in many other ILC subsets (Fig. 6b). Genes
upregulated in all NK cells were consistently not expressed in other
ILCs (Fig. 6b), except for low transcript levels in ILC1 cells. The transcripts with the highest expression in ILCs relative to their expression
in NK cells (Fig. 6a) were Tcrg-V3, Tmem176a, Tmem176b, Il7r and
Cxcr6 (Fig. 6b and Supplementary Table 6). The high expression
of several TCRg transcripts by all ILC subsets compared to NK cells
(Fig. 6a,b) was unexpected, as ILCs by definition do not express
recombined antigen receptors. We first used flow cytometry to confirm that no T cell antigen receptor (TCR) δ-chains were expressed in
any of our cell types, either extracellularly or intra­cellularly (Fig. 6c
and Supplementary Fig. 2a); this suggested that they lacked functional TCRγδ expression. Furthermore, from a published RNAsequencing data set that included liver ILC1 cells, liver NK cells and
spleen NK cells from mice deficient in recombination-activating
gene 1 (Rag1−/− mice)43, we found high levels of the 3′ end of the
locus encoding TCRγ, annotated as the ‘Tcrg C4’ transcript, present
in liver ILC1 cells but not in NK cells from liver or spleen (Fig. 6d).
By PCR, we confirmed that the transcript encoding TCRγ-V3 was
a germline transcript (data not shown). As cytokines IL-7 and IL15, which activate the transcription factor STAT5, are well known
to mediate germline expression of the locus encoding TCRγ44,45, we
concluded that ILCs had open chromatin at the locus encoding TCRγ
and germline transcription, probably due to signaling through IL-7.
We also assessed the ability to use CXCR6 as an ILC marker through
the use of a Cxcr6eGFP reporter mouse. Although we found significant
differences between ILC1 and NK cell populations in their frequency
of CXCR6+ cells (Fig. 6e and Supplementary Fig. 2b), not all ILCs
were labeled (Fig. 6e and Supplementary Fig. 2b). Additionally, at
steady state, CXCR6 was not required for the development or tissue
homing of ILCs, as we found no difference between Cxcr6eGFP/+ mice
and Cxcr6eGFP/eGFP mice in their frequency of ILCs (Fig. 6f).
Finally, we sought to determine whether intestinal intraepithelial
ILC1 cells should be classified as an ILC1 or NK cell population, as this
population has unique transcription factors and cell-surface markers
that prevent it from fitting clearly into an ILC1 or NK cell designation16. Comparisons of IEL ILC1 cells, siLP ILC1 cells and siLP NK
cells revealed that IEL ILC1 cells expressed transcripts characteristic
nature immunology VOLUME 16 NUMBER 3 MARCH 2015
of both NK cell and ILC1 populations (Fig. 6g). Whereas the NK cell
signature transcripts Eomes and Klra3 (Fig. 6b) had higher expression
by IEL ILC1 cells than siLP ILC1 cells, the ILC1 signature transcripts
Tcrg-V2 and Tcrg-V3 (Fig. 6b) were more abundant in IEL ILC1 cells
than in siLP NK cells (Fig. 6g). It remains unclear whether IEL ILC1
cells are a single, unique subset with distinct developmental and functional characteristics, or whether siLP NK cells and ILC1 cells both
traffic to the epithelium, where they become phenotypically indistinguishable, possibly in response to tissue factors in the epithelium.
DISCUSSION
Here we have provided the first comprehensive transcriptional analysis
of the spectrum of ILC subsets reported in the siLP, liver and spleen, to
our knowledge. The transcriptional programs we found should allow
better definition of the individual ILC classes, as well as of ILC subsets
within a class. ILC2 was the most homogeneous and distinguishable
ILC class and expressed the greatest number of characteristic genes,
with many transcripts expressed more than fourfold higher than in any
other subset. These genes encoded well-­documented factors as well as
previously unrecognized genes, including the nuclear receptors RXRγ
and PPARγ and several other molecules involved in lipid metabolism.
Within the ILC3 class, LTi-like ILC3 cells expressed several characteristic genes at higher levels than in other subsets, including genes
previously unknown to be expressed in cells of the immune system,
including Cntn1, Slc6a7 and Cacna1g. These cells were distinct from
NKp46+ ILC3 cells and were effectively marked by the previously unreported LTi-like factor Nrp1, as well as by CD25. The definition of the
ILC1 class was the most problematic, because in comparisons of ILC1
cells with all other ILCs and NK cells, we found no markers specific for
siLP ILC1 cells and few for liver and spleen ILC1 cells. The transcript
encoding CD103 was an unexpected characteristic transcript for siIEL
ILC1 cells, given that it was not present as CD103 protein in this subset
in mouse studies and has been noted only in human cells16. Among
siLP subsets, we also found a previously unrecognized ‘intestinal’ signature composed of activation markers such as CD69 and many transcription factors, including those encoded by Rora, Atf3, Nr4a1 and
Maff, probably reflective of continuous environmental exposure.
Beyond the unique factors, ILC classes also shared many transcripts.
For example, siLP NKp46+RORγt+ ILC3 cells shared many transcripts
with siLP ILC1 cells, including Ifng and Il12rb2. These data, which
were consistent with published reports4,14, provide a basis for the proposal of functional plasticity of siLP NKp46+RORγt+ ILC3 cells, which
become similar to ILC1 cells in certain conditions yet to be defined.
Furthermore, ILC1 cells and ILC2 cells shared expression of the transcription factor–encoding transcript Ets-2, which would suggest previously unknown transcriptional pathways shared by ILC1 cells and
ILC2 cells. Notably, ILC2 cells and LTi-like ILC3 cells expressed genes
that would suggest a function in neural and glial crosstalk. ILC2 cells
expressed Bmp2, which encodes a molecule that has been found to
modulate enteric motility in response to the microbiota33; given the
importance of motility in clearing helminth infections, this observation might indicate a previously unknown mechanism of innate
defense. We discovered that both ILC2 cells and LTi-like ILC3 cells also
expressed Ret, a proto-oncogene that encodes a receptor for the gliaderived neurotrophic family of molecules, which are known to drive
the development of Peyer’s patches37 but currently remain unstudied
in ILCs. Thus, in the siLP, ILCs may engage in crosstalk with neurons
and glia in the steady state and during an immune response.
We generated a core ILC signature that included 17 genes, with the
highest expression of germline transcripts encoding TCRγ, as well as
Cxcr6, Tmem176a, Tmem176b and Il7r. We were surprised to find that
315
npg
© 2015 Nature America, Inc. All rights reserved.
resource
the gene with the highest expression by all ILCs relative to its expression in NK cells was the TCRγ-V3 germline transcript, which has not
been previously reported in ILCs, to our knowledge, but has been
reported in a putative ILC3 cell line46. We postulate that this might
reflect signaling by IL-7R, which is expressed by all ILCs but not by
mature NK cells44. In our study, we confirmed that all ILCs had higher
expression of Cxcr6 than did NK cells. Two published studies have
also investigated CXCR6 expression in ILCs. The first study demonstrated that a CXCR6+ early progenitor gives rise to both NK cells and
ILCs but not T cells47. The second study found that CXCR6 deficiency
‘preferentially’ affects the frequency and function of NKp46+ ILC3
cells by preventing appropriate interactions with CD11b+ intestinal
dendritic cells48. We found that CXCR6 did not mark the entire population, nor were any ILC frequencies affected by its loss; however, it
is possible that loss of CXCR6 affects ILC function, and this should
be tested in all ILC classes. Notably, Tmem176b has been reported to
be a marker of innate lymphocytes as one of only three transcripts
shared by NK cells, NKT cells and γδ T cells49, although in our data
set we found that ILC1, ILC2 and ILC3 cells had significantly higher
expression of Tmem176b than did NK cells. These data suggest that
the core ILC signature identified here may also extend to other innate
and/or tissue-resident lymphoid cells.
IFN-γ-producing ILC1 cells and NK cells can develop from different progenitors and are, respectively, independent of and dependent
on Eomes4,11. However, in tissues they show overlapping phenotypes
and functional programs. NK cells are well known to have cytolytic
ability, but liver ILC1 cells express granzyme A and granzyme C and
have also been shown to be cytolytic42,43,50. In our study, liver ILC1
cells were clearly separated by TRAIL and CD49a, but cell-surface
markers of CD127 with and without CD27 in the spleen and siLP,
respectively, were insufficient to discriminate Eomes− ILC1 cells from
Eomes+ NK cells. Thus, transcriptional data generated using Eomes
reporter mice might be useful for comparison to our data set4,42.
Moreover, the induction of cytokines such as IL-15 and/or IL-2 in
certain pathologic conditions might further increase the phenotypic
and functional similarity of ILC1 cells and NK cells. Thus, it remains
unclear whether ILC1 cells and NK cells are truly distinct lineages or
a spectrum of cells within a single lineage that includes ILC1 cells,
immature NK cells and mature NK cells.
Our data offer the most complete transcriptional profile of ILCs
and NK cells so far, to our knowledge, and provide a comprehensive
view of the relationships among ILC subsets at steady state. Our findings should help define new avenues of research and should aid in
the production of new tools for studying ILCs, especially with the
identification of several molecular targets with high expression by all
ILCs. Finally, they should also be a valuable resource for the scientific
community, with access to our data set and comparisons to other
published data sets generated under the same rigorous conditions,
provided by the ImmGen Project.
Methods
Methods and any associated references are available in the online
version of the paper.
Accession codes. GEO: microarray data, GSE37448.
Note: Any Supplementary Information and Source Data files are available in the
online version of the paper.
Acknowledgments
We thank our colleagues in the ImmGen consortium, especially C. Benoist and
L. Lanier, for input and discussion; the core ImmGen team, K. Rothamel and
316
A. Rhodes, for contributions and technical assistance; M. Artyomov, G. Krishnan
and J. Siegel for computational assistance; D. Sojka for discussion; E. Lantelme and
D. Brinja for sorting assistance; P. Wang for microscopy assistance; and eBioscience
and Affymetrix for support of the ImmGen Project. Supported by the US National
Institutes of Health (R24AI072073 to the ImmGen Consortium; 1U01AI095542,
R01DE021255 and R21CA16719 to the Colonna laboratory; MSTP T32 GM07200
to M.L.R.; and Infectious Disease Training Grant T32 AI 7172-34 to V.S.C.).
AUTHOR CONTRIBUTIONS
M.L.R. analyzed data; A.F., M.L.R., J.S.L. and Y.W. sorted cell subsets;
M.L.R., A.F. and V.S.C. performed follow-up experiments and analyzed data;
S.G. maintained mice; S.K.D. provided critical reagents; M.L.R., A.F., S.G. and
M.C. designed studies; M.L.R. and M.C. wrote the paper; and the ImmGen
Consortium contributed to the experimental design and data collection.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
Reprints and permissions information is available online at http://www.nature.com/
reprints/index.html.
1. Heng, T.S. & Painter, M.W. The Immunological Genome Project: networks of gene
expression in immune cells. Nat. Immunol. 9, 1091–1094 (2008).
2. Diefenbach, A., Colonna, M. & Koyasu, S. Development, differentiation, and diversity
of innate lymphoid cells. Immunity 41, 354–365 (2014).
3. McKenzie, A.N., Spits, H. & Eberl, G. Innate lymphoid cells in inflammation and
immunity. Immunity 41, 366–374 (2014).
4. Klose, C.S. et al. Differentiation of type 1 ILCs from a common progenitor to all
helper-like innate lymphoid cell lineages. Cell 157, 340–356 (2014).
5. Sawa, S. et al. Lineage relationship analysis of RORgammat+ innate lymphoid cells.
Science 330, 665–669 (2010).
6. Lee, J.S. et al. AHR drives the development of gut ILC22 cells and postnatal
lymphoid tissues via pathways dependent on and independent of Notch.
Nat. Immunol. 13, 144–151 (2011).
7. Klose, C.S. et al. A T-bet gradient controls the fate and function of CCR6-Rorγt+
innate lymphoid cells. Nature 494, 262–265 (2013).
8. Sciumé, G. et al. Distinct requirements for T-bet in gut innate lymphoid cells.
J. Exp. Med. 209, 2331–2338 (2012).
9. Rankin, L.C. et al. The transcription factor T-bet is essential for the development
of NKp46+ innate lymphocytes via the Notch pathway. Nat. Immunol. 14, 389–395
(2013).
10.Buonocore, S. et al. Innate lymphoid cells drive interleukin-23-dependent innate
intestinal pathology. Nature 464, 1371–1375 (2010).
11.Constantinides, M.G., McDonald, B.D., Verhoef, P.A. & Bendelac, A. A committed
precursor to innate lymphoid cells. Nature 508, 397–401 (2014).
12.Yu, J., Freud, A.G. & Caligiuri, M.A. Location and cellular stages of natural killer
cell development. Trends Immunol. 34, 573–582 (2013).
13.Gordon, S.M. et al. The transcription factors T-bet and Eomes control key checkpoints
of natural killer cell maturation. Immunity 36, 55–67 (2012).
14.Reynders, A. et al. Identity, regulation and in vivo function of gut NKp46+RORγt+
and NKp46+RORγt– lymphoid cells. EMBO J. 30, 2934–2947 (2011).
15.Gasteiger, G., Hemmers, S., Bos, B.D., Sun, J.C. & Rudensky, A.Y. IL-2-dependent
adaptive control of NK cell homeostasis. J. Exp. Med. 210, 1179–1187 (2013).
16.Fuchs, A. et al. Intraepithelial type 1 innate lymphoid cells are a unique subset of
IL-12- and IL-15-responsive IFN-γ-producing cells. Immunity 38, 769–781
(2013).
17.Turner, J.E. et al. IL-9-mediated survival of type 2 innate lymphoid cells
promotes damage control in helminth-induced lung inflammation. J. Exp. Med.
210, 2951–2965 (2013).
18.Mjösberg, J. et al. The transcription factor GATA3 is essential for the function of
human type 2 innate lymphoid cells. Immunity 37, 649–659 (2012).
19.Yagi, R. et al. The transcription factor GATA3 is critical for the development of all
IL-7Rα-expressing innate lymphoid cells. Immunity 40, 378–388 (2014).
20.Evans, R.M. & Mangelsdorf, D.J. Nuclear receptors, RXR, and the Big Bang. Cell
157, 255–266 (2014).
21.Spencer, S.P. et al. Adaptation of innate lymphoid cells to a micronutrient deficiency
promotes type 2 barrier immunity. Science 343, 432–437 (2014).
22.Molofsky, A.B. et al. Innate lymphoid type 2 cells sustain visceral adipose tissue
eosinophils and alternatively activated macrophages. J. Exp. Med. 210, 535–549
(2013).
23.Mohebiany, A.N., Harroch, S. & Bouyain, S. in Cell Adhesion Molecules: Implications
in Neurological Diseases, Advances in Neurobiology Vol. 8 (eds. Berezin, V. &
Walmod, P.S.) Chapter 8 (Springer Science+Business Media, 2014).
24.Cella, M. et al. A human natural killer cell subset provides an innate source of
IL-22 for mucosal immunity. Nature 457, 722–725 (2009).
25.Mortha, A. et al. Microbiota-dependent crosstalk between macrophages and ILC3
promotes intestinal tolerance. Science 343, 1249288 (2014).
26.Levitt, L.J. et al. Production of granulocyte/macrophage-colony stimulating factor
by human natural killer cells. Modulation by the p75 subunit of the interleukin 2
and by the CD2 receptor. J. Clin. Invest. 88, 67–75 (1991).
VOLUME 16 NUMBER 3 MARCH 2015 nature immunology
npg
© 2015 Nature America, Inc. All rights reserved.
resource
27.van de Pavert, S.A. et al. Maternal retinoids control type 3 innate lymphoid cells
and set the offspring immunity. Nature 508, 123–127 (2014).
28.Kiss, E.A. et al. Natural aryl hydrocarbon receptor ligands control organogenesis of
intestinal lymphoid follicles. Science 334, 1561–1565 (2011).
29.Gascoyne, D.M. et al. The basic leucine zipper transcription factor E4BP4 is
essential for natural killer cell development. Nat. Immunol. 10, 1118–1124
(2009).
30.Kamizono, S. et al. Nfil3/E4bp4 is required for the development and maturation
of NK cells in vivo. J. Exp. Med. 206, 2977–2986 (2009).
31.Seillet, C. et al. Nfil3 is required for the development of all innate lymphoid cell
subsets. J. Exp. Med. 211, 1733–1740 (2014).
32.Geiger, T.L. et al. Nfil3 is crucial for development of innate lymphoid cells and host
protection against intestinal pathogens. J. Exp. Med. 211, 1723–1731 (2014).
33.Muller, P.A. et al. Crosstalk between muscularis macrophages and enteric neurons
regulates gastrointestinal motility. Cell 158, 300–313 (2014).
34.Bando, J.K., Liang, H.E. & Locksley, R.M. Identification and distribution of
developing innate lymphoid cells in the fetal mouse intestine. Nat. Immunol. http://
www.nature.com/ni/journal/vaop/ncurrent/full/ni.3057.html (2014).
35.Veiga-Fernandes, H. et al. Tyrosine receptor RET is a key regulator of Peyer’s patch
organogenesis. Nature 446, 547–551 (2007).
36.Ramirez, K. et al. Gene deregulation and chronic activation in natural killer cells
deficient in the transcription factor ETS1. Immunity 36, 921–932 (2012).
37.Carpenter, S., Ricci, E.P., Mercier, B.C., Moore, M.J. & Fitzgerald, K.A. Posttranscriptional regulation of gene expression in innate immunity. Nat. Rev. Immunol.
14, 361–376 (2014).
38.Kumanogoh, A. & Kikutani, H. Immunological functions of the neuropillins and
plexins as receptors for semaphorins. Nat. Rev. Immunol. 13, 802–814 (2013).
39.Yadav, M. et al. Neuropilin-1 distinguishes natural and inducible regulatory T cells
among regulatory T cell subsets in vivo. J. Exp. Med. 209, 1713–1722 (2012).
40.Weiss, J.M. et al. Neuropilin 1 is expressed on thymus-derived natural regulatory
T cells, but not mucosa-generated induced Foxp3+ T reg cells. J. Exp. Med. 209,
1723–1742 (2012).
41.Shen, W. et al. Adaptive immunity to murine skin commensals. Proc. Natl. Acad.
Sci. USA 111, E2977–E2986 (2014).
42.Daussy, C. et al. T-bet and Eomes instruct the development of two distinct natural
killer cell lineages in the liver and in the bone marrow. J. Exp. Med. 211, 563–577
(2014).
43.Sojka, D.K. et al. Tissue-resident natural killer (NK) cells are cell lineages distinct
from thymic and conventional splenic NK cells. Elife 3, e01659 (2014).
44.Ye, S.K. et al. Induction of germline transcription in the Tcrγ locus by Stat5:
implications for accessibility control by the IL-7 receptor. Immunity 11, 213–223
(1999).
45.Zhao, H., Nguyen, H. & Kang, J. Interleukin 15 controls the generation of restricted
T cell receptor repertoire of intraepithelial lymphocytes. Nat. Immunol. 6,
1263–1271 (2005).
46.Allan, D.S. et al. An in vitro model of innate lymphoid cell function and
differentiation. Mucosal Immunol. http://www.nature.com/mi/journal/vaop/ncurrent/
full/mi201471a.html (2014).
47.Yu, X. et al. The basic leucine zipper transcription factor NFIL3 directs the
development of a common innate lymphoid cell progenitor. Elife 10, e04406
(2014).
48.Satoh-Takayama, N. et al. The chemokine receptor CXCR6 controls the functional
topograophy of interluekin-22 producing intestinal innate lymphoid cells. Immunity
41, 776–788 (2014).
49.Bezman, N.A. et al. ImmGen report: molecular definition of natural killer cell identity
and activation. Nat. Immunol. 13, 1000–1009 (2012).
50.Peng, H. et al. Liver-resident NK cells confer adaptive immunity in skin-contact
inflammation. J. Clin. Invest. 123, 1444–1456 (2013).
The complete list of authors is as follows:
Laura Shaw6, Bingfei Yu6, Ananda Goldrath6, Sara Mostafavi7, Aviv Regev8, Edy Y Kim9, Dan F Dwyer9,
Michael B Brenner9, K Frank Austen9, Andrew Rhoads10, Devapregasan Moodley10, Hideyuki Yoshida10,
Diane Mathis10, Christophe Benoist10, Tsukasa Nabekura11, Viola Lam11, Lewis L Lanier11, Brian Brown12,
Miriam Merad12, Viviana Cremasco13, Shannon Turley13, Paul Monach14, Michael L Dustin15, Yuesheng Li16,
Susan A Shinton16, Richard R Hardy16, Tal Shay17, Yilin Qi18, Katelyn Sylvia18, Joonsoo Kang18, Keke Fairfax1,
Gwendalyn J Randolph1, Michelle L Robinette1, Anja Fuchs2 & Marco Colonna1
6Division
of Biological Sciences, University of California San Diego, La Jolla, California, USA. 7Computer Science Department, Stanford University, Stanford,
California, USA. 8Broad Institute of MIT and Harvard, Cambridge, Massachusetts, USA. 9Division of Rheumatology, Immunology and Allergy, Brigham and Women’s
Hospital, Boston, Massachusetts, USA. 10Division of Immunology, Department of Microbiology & Immunobiology, Harvard Medical School, Boston, Massachusetts,
USA. 11Department of Microbiology & Immunology, University of California San Francisco, San Francisco, California, USA. 12Icahn Medical Institute, Mount
Sinai Hospital, New York, New York, USA. 13Dana-Farber Cancer Institute and Harvard Medical School, Boston, Massachusetts, USA and Department of Cancer
Immunology, Genentech, San Francisco, California, USA. 14Department of Medicine, Boston University, Boston, Massachusetts, USA. 15Skirball Institute of
Biomolecular Medicine, New York University School of Medicine, New York, New York, USA. 16Fox Chase Cancer Center, Philadelphia, Pennsylvania, USA.
17Department of Life Sciences, Ben-Gurion University of the Negev, Be’er Sheva, Israel. 18Department of Pathology, University of Massachusetts Medical School,
Worcester, Massachusetts, USA.
nature immunology VOLUME 16 NUMBER 3 MARCH 2015
317
ONLINE METHODS
npg
© 2015 Nature America, Inc. All rights reserved.
Mice. Consistent with ImmGen Project standards, 6- to 8-week-old male
C57BL/6J mice and male B6.129P2(Cg)-Rorctm2Litt/J mice (with the sequence
encoding enhanced green fluorescent protein (EGFP) inserted into the locus
encoding RORγt) obtained from Jackson Laboratories and maintained at
the Washington University School of Medicine (WUSM) under specific
pathogen–free conditions were used for cell-sorting and validation
experiments. B6.129P2-Cxcr6tm1Litt/J mice with the sequence encoding EGFP
inserted into the locus encoding CXCR6, obtained from Jackson Laboratories,
were used for validation experiments in which mice were littermates or
age-matched, gender-matched and co-housed when possible. Il-22 tdTomato
reporter BAC transgenic mice were used for validation experiments41. The
WUSM Animal Studies Committee approved all experiments.
Antibodies and flow cytometry. Anti-CD3e (145-2C11), anti-CD19
(eBio1D3), anti-CD27 (LG.7F9), anti-CD49b (HMa2, DX5), anti-γδ (UC713D5; eBioGL3), anti-CD127 (A7R34; eBioSB/199), anti-CD49a (HMa1), antiNkp46 (29A1.4), anti-CD304 (3DS304M), anti-NK1.1 (PK136), anti-Sca-1
(D7), anti-ST2 (RMST2-2), anti-TRAIL (eBioN2B2), anti-CD4 (GK1.5), antiKLRG1 (2F1), anti-RORγt (AFKJS-9), anti-Eomes (Dan11mag), SAv-PE/Cy7,
and isotype-matched control monoclonal antibodies were obtained from
eBioscience. Anti-CD25 (PCB1) and 7-AAD were from BD Biosciences.
Anti-CD45 (30-F11) was from Miltenyi Biotec. LIVE/DEAD Fixable Aqua
and SAv-APC were from Life Technologies. Monoclonal anti-NKp46 (Cs96;
rat immunoglobulin G) was produced and biotinylated in the Colonna lab at
WUSM. Fc receptors were blocked before surface staining with supernatant
from hybridoma cells producing monoclonal antibody to CD32 (HB-197;
ATCC). For intracellular staining, the FOXP3 staining kit (eBioscience) was
used. Data were acquired on a BD FACSCanto II and analyzed with FlowJo
software (Treestar).
Cell identification, isolation and microscopy. All cells were stained and
sorted according to the published standard operations protocol on the ImmGen
website (http://www.immgen.org/Protocols/ImmGen%20Cell%20prep%20an
d%20sorting%20SOP.pdf). Cells were isolated from 3 to 15 mice per sample.
For isolation of siLP and siIEL, small intestine beginning half a centimeter after
the pylorus and ending half a centimeter before the cecum with Peyer’s patches
removed was dissociated using the Miltenyi Lamina Propria Dissociation kit,
which includes two dithiothreitol/EDTA washes and an enzymatic digestion step of 30 min. For liver and spleen isolation, mice were perfused with
20 mL of cold PBS, and tissue was dissociated through a 70 µm cell strainer.
Lymphocytes were enriched at the interface between a gradient of 40% and
70% Percoll in HBSS. siIELs were further preenriched by negative selection
with anti-CD8 MicroBeads (Miltenyi). Liver and spleen were pre-enriched by
negative selection with anti-CD19 and anti-CD4 MicroBeads (Miltenyi). Cells
were double-sorted directly into TRIzol with a Becton-Dickinson FACSAria II.
The data browser of the ImmGen Project website shows flow cytometry plots
from gating strategies and purity from the first round of sorting. Cytospins
were prepared according to the same ImmGen standards, except from only one
to three mice; splenic samples were positively pre-enriched with DX5-coated
beads (Miltenyi). Sorted cells were spun onto slides, left to dry overnight, and
nature immunology
then fixed and stained with Diff-Quik. Pictures were taken at ×60 with oil
immersion using a Nikon Eclipse E800 and Leica Application Suite software.
Microarray and data analysis. Two to three replicates of RNA were obtained
from each sample that passed quality control. RNA amplification and hybridization to the Affymetrix Mouse Gene 1.0 ST array were carried out by ImmGen
with a standardized TRIzol extraction protocol (https://www.immgen.org/
Protocols/Total%20RNA%20Extraction%20with%20Trizol.pdf). Data generation and quality-control documentation also followed the ImmGen protocol;
these methods can be found online (https://www.immgen.org/Protocols/Imm
Gen%20QC%20Documentation_ALL-DataGeneration_0612.pdf), along with
quality-control data, replicate information, and batch information from each
sample. Data were analyzed with GenePattern software51 (Broad Institute).
Raw data were normalized with RMA. Differences in gene expression were
identified with the Multiplot Studio function of GenePattern, from a filtered
subset of genes with coefficients of variation less than 0.1 in all samples and
expression of at least 120 relative units in one subset by the class mean functions, a value that corresponds to 95% confidence of true expression across the
ImmGen data set. For gene-expression signatures of individual subsets, pairwise comparisons were made between subsets, filtering for a change in expression of twofold or fourfold. For comparisons of two to four samples, probe sets
were considered to have differences of expression for transcripts that expressed
>120 relative units, with a change in expression of greater than twofold and
P values of <0.05 (Student’s t-test), except between siLP ILC1 and siLP NK
cells, where the P value was not considered. Volcano plots and plots comparing
change in expression (fold) versus change in expression (fold) were produced
in Multiplot, and the degree of overlapping genes between subsets was calculated in MATLAB. Heat maps were generated with Gene-E (http://www.
broadinstitute.org/cancer/software/GENE-E/). Data were log2-­transformed
and visualized by ‘relative’ expression per row or ‘global’ expression in the heat
map, as indicated, and rows were clustered with the Hierarchical Clustering
function with the Pearson correlation as a metric. Where more than one probe
set with the same gene annotation was found, the probe set with highest average expression was used. For PCA, the top 10% of the most variable probe sets
was calculated with the PopulationDistances PCA program (S. Davis, Harvard
Medical School). This program identifies differently expressed genes through
ANOVA using the geometric standard deviation of populations to weight genes
that vary in multiple populations. The data set for the top 10% of genes with
the most variability was log2-transformed in MATLAB and used to generate
a PCA with the functions pca and scatter3. Hierarchical clustering of sample
replicates was carried out in Gene-E from the same data set, with the Pearson
correlation used as a metric. For RNA sequencing analysis, data from GEO
accession code GSE52043 were visualized in Integrative Genomics Viewer and
aligned to the mm9 National Center for Biotechnology Information assembly
of the mouse genome, as described43.
Statistics. Prism (GraphPad Software) was used for statistical analysis of flow
cytometry data. Data were tested using either one-way ANOVA or two-way
ANOVA, as indicated.
51.Reich, M. et al. GenePattern 2.0. Nat. Genet. 38, 500–501 (2006).
doi:10.1038/ni.3094