Proteomic Analyses of Human Regulatory T Cells Reveal Adaptations in Signaling Pathways that Protect Cellular Identity

Resource

Proteomic Analyses of Human Regulatory T Cells Reveal Adaptations in Signaling Pathways that Protect Cellular Identity Graphical Abstract

Authors Eloy Cuadrado, Maartje van den Biggelaar, Sander de Kivit, ..., Rene A.W. van Lier, Jannie Borst, Derk Amsen

Correspondence [email protected] (E.C.), [email protected] (D.A.)

In Brief Using high-resolution mass spectrometry and transcriptomics, Cuadrado et al. provide a molecular characterization of regulatory and conventional CD4+ T cell subsets, yielding markers to distinguish cells with different properties and insights into mechanisms that prevent regulatory T cells from exhibiting undesirable functional activities of the related but functionally antithetical conventional T cells.

Highlights d

Treg cell protein signatures were defined that are stable even after culture in vitro

d

Treg cells exhibit strategic changes in signaling pathways to protect their identity

d

FOXP3 protein levels exceed critical partner transcription factors in eTreg cells

d

Surface markers define functional heterogeneity within FOXP3+CD4+ T cells

Cuadrado et al., 2018, Immunity 48, 1046–1059 May 15, 2018 ª 2018 Elsevier Inc. https://doi.org/10.1016/j.immuni.2018.04.008

Immunity

Resource Proteomic Analyses of Human Regulatory T Cells Reveal Adaptations in Signaling Pathways that Protect Cellular Identity Eloy Cuadrado,1,* Maartje van den Biggelaar,2,4 Sander de Kivit,3,4 Yi-yen Chen,3 Manon Slot,1 Ihsane Doubal,1 Alexander Meijer,2 Rene A.W. van Lier,1 Jannie Borst,3,5 and Derk Amsen1,5,6,* 1Department of Hematopoiesis, Sanquin Research and Landsteiner Laboratory, Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands 2Department of Plasma Proteins, Sanquin Research and Landsteiner Laboratory, Academic Medical Center, University of Amsterdam, Amsterdam, the Netherlands 3Division of Tumor Biology and Immunology, the Netherlands Cancer Institute-Antoni van Leeuwenhoek, 1066 CX Amsterdam, the Netherlands 4These authors contributed equally 5Senior author 6Lead Contact *Correspondence: [email protected] (E.C.), [email protected] (D.A.) https://doi.org/10.1016/j.immuni.2018.04.008

SUMMARY

To obtain a molecular definition of regulatory T (Treg) cell identity, we performed proteomics and transcriptomics on various populations of human regulatory and conventional CD4+ T (Tconv) cells. A protein expression signature was identified that defines all Treg cells, and another signature that defines effector Treg cells. These signatures could not be extrapolated from transcriptome data. Unique cellbiological and metabolic features in Treg cells were defined, as well as specific adaptations in cytokine, TCR, and costimulatory receptor signaling pathways. One such adaptation—selective STAT4 deficiency—prevented destabilization of Treg cell identity and function by inflammatory cytokines, while these signals could still induce critical transcription factors and homing receptors via other pathways. Furthermore, our study revealed surface markers that identify FOXP3+CD4+ T cells with distinct functional properties. Our findings suggest that adaptation in signaling pathways protect Treg cell identity and present a resource for further research into Treg cell biology.

INTRODUCTION Treg cells constitute a distinct CD4+ T cell lineage, defined by expression of the transcription factor FOXP3 and the ability to suppress immune responses (Josefowicz et al., 2012). Treg cells safeguard immune tolerance, limit immune-cell-inflicted tissue damage, and promote tissue repair (Arpaia et al., 2015; Nosbaum et al., 2016). Treg cells impede immunity against cancer (Nishikawa and Sakaguchi, 2014) and insufficient Treg cell function may underlie human autoimmune diseases and allergy (Pel1046 Immunity 48, 1046–1059, May 15, 2018 ª 2018 Elsevier Inc.

lerin et al., 2014). Targeted strategies are therefore needed to boost or inhibit Treg cell function in human disease. For selective manipulation of the closely related Treg or Tconv cells, we need a detailed understanding of their distinctions at the molecular level. Treg cells generally do not produce cytokines such as IL-2, IL-4, IL-17, and IFN-g (here referred to as ‘‘effector cytokines’’), even when exposed to inflammatory cytokines (Josefowicz et al., 2012; Miyao et al., 2012; Rubtsov et al., 2010). Treg cells do, however, respond to inflammatory cues, which can enhance suppressive function (Arvey et al., 2014) and induce expression of Th-cell-lineage-specific transcription factors and homing receptors (Josefowicz et al., 2012; Tan et al., 2016; Yu et al., 2015). Under conditions that are not understood, inflammation may paradoxically also destabilize Treg cells, provoking production of effector cytokines (Chen et al., 2013; Gao et al., 2015; van Loosdregt et al., 2013). In cancer, the presence of Treg cells that produce effector cytokines marks a relatively favorable prognosis (Overacre-Delgoffe et al., 2017; Saito et al., 2016). In autoimmune diseases, such Treg cells may, however, contribute to development of pathology (Pandiyan and Zhu, 2015). Understanding the mechanisms that generally prevent production of effector cytokines by Treg cells and identification of markers that discriminate between Treg cells with and without the capacity to produce effector cytokines is important for diagnostic and therapeutic purposes. How Treg cells normally resist acquisition of Tconv cell properties is incompletely understood. FOXP3 can directly suppress effector cytokine-encoding genes and prevent their expression through physical interaction with other transcription factors (Kwon et al., 2017; Li and Zheng, 2015; Li et al., 2014; Rudra et al., 2012). Expression of Tconv cell-specific genes is furthermore opposed by an epigenetic layer of regulation (Ohkura et al., 2012). Here, we aimed to define human Treg cell identity at the molecular level through label-free quantitative proteomics and transcriptomics. We specifically examined blood-derived naive (n) and effector (e) Treg cells that, according to studies in mouse

models (Siegmund et al., 2005; Smigiel et al., 2014), respectively inhibit spurious priming of nTconv cells in lymphoid tissue and effector functions of T cells in non-lymphoid tissue. We furthermore analyzed an incompletely defined CD4+ T cell population, similar to ‘‘fraction (Fr.) III’’ (Miyara et al., 2009), that produces effector cytokines, despite expressing FOXP3. Finally, we determined protein expression patterns in Treg cells cultured in vitro, with and without T cell receptor (TCR) stimulation. These analyses identified markers for the discrimination of functionally distinct FOXP3+CD4+ T cell types in human blood and defined a common Treg cell signature, and one specific to eTreg cells. The signatures showed that signaling pathways that control effector cytokine genes in Tconv cells are desensitized at strategic points in Treg cell activation. Low expression of, for example, STAT4, enabled Treg cells to respond to inflammatory cytokines without compromising Treg cell identity. Our findings provide insight into mechanisms protecting Treg cell identity in an inflammatory environment and present a resource for the further study of human Treg cell biology. RESULTS Protein Expression Signatures Define Treg Cell Populations CD4+ T cell subsets were isolated from peripheral blood mononuclear cells (PBMCs) of healthy human donors. These subsets comprised nTconv cells (CD45RA+CD25
), memory (m)Tconv cells (CD45RA
CD25
), nTreg cells (CD45RA+CD25hi), and eTreg cells (CD45RA
CD25hi) (Figures 1A and S1A). Both Treg cell populations expressed FOXP3 and Helios and lacked IL-7 receptor a chain (CD127), while nTconv and mTconv cell populations conversely lacked FOXP3 and Helios, but did express CD127 (Figures S1B and S1C). We also isolated a CD45RA
CD25intCD4+ T cell population, similar to the previously described Fraction (Fr).III (Miyara et al., 2009), which contains cells that produce effector cytokines, despite expressing FOXP3 (Figure S1B). As originally defined, ‘‘Fr.III’’ was heterogeneous in expression of FOXP3, Helios, and CD127 (Figures S1B and S1C) and therefore likely included activated Tconv cells. Unlike nTconv and mTconv cells, both Treg cell populations were demethylated at the Treg cell-specific demethylated region (TSDR) in the FOXP3 gene locus (Figure S1D; Polansky et al., 2008). Methylation of the TSDR in Fr.III cells was intermediate, likely reflecting the mixed composition of this cell population. Using high-resolution mass spectrometry (MS), we identified an average of 35,744 ± 3,757 peptides corresponding to 5,955 ± 344 protein groups in these samples. In total, 4,358 distinct proteins were quantified in all five CD4+ T cell subsets. Of these, 422 proteins exhibited differential expression (FDR < 0.05) among the CD4+ T cell subsets, based on their label-free quantification (LFQ) values. Principal component analysis (PCA, Figure 1B) and Pearson correlation (Figure 1C) of the total datasets and unsupervised hierarchical clustering of the differentially expressed proteins (Figure 1D and Table S1) confirmed the close relatedness of biological replicates. Furthermore, these analyses separated the CD4+ T cell subsets primarily on basis of CD45RA expression, rather than by lineage (Figure 1D). Enrichment analysis revealed that naive Treg and Tconv cells shared high expression of proteins involved in pro-

cesses such as pentose phosphate metabolism, NADP metabolism, and chromatin organization (Figure 1D, clusters 7–8 and Table S1), while eTreg and mTconv cells shared high expression of proteins involved in protein synthesis, cell trafficking, signaling, and apoptosis (Figure 1D, clusters 3–4 and Table S1). The similarity in protein expression based on differentiation stage rather than lineage may in part reflect preferential localization in lymphoid tissue (naive cells) versus non-lymphoid tissue (effector cells) (Campbell, 2015; Smigiel et al., 2014). Despite their markedly different proteomes, nTreg and eTreg cells uniquely shared high (cluster 1) and low (cluster 9–10) expression of certain proteins that thus defined a common Treg cell protein signature. No signature uniquely defined nTreg cells. However, eTreg cells uniquely exhibited high expression of proteins in cluster 2 and low expression of proteins in cluster 6 (Table S1), which thus defined an eTreg cell protein signature. mRNA Expression Signatures Define Treg Cell Populations The proteomic Treg cell signatures defined overlapped little with published transcriptomic signatures of similar cell populations (Miyara et al., 2009; Ferraro et al., 2014; Schmidl et al., 2014). To allow direct comparison, we performed genome-wide mRNA deep sequencing of the five CD4+ T cell subsets. Also based on their transcriptome, nTreg and nTconv cells clustered together and away from the three effector and memory phenotype CD4+ T cell populations (Figure 2A). A total of 649 mRNAs (p < 0.05) was differentially expressed between the five CD4+ T cell subsets (Figure 2B and Table S2), including expected hallmark genes such as IL-7R, IKZF2 (HELIOS), and effector cytokines (Figure S2A). This analysis yielded an eTreg cell-specific (Figure 2B, cluster 7; Table S2) and a common Treg cell mRNA signature (Figures 2B and 2C, cluster 6) that featured expected molecules such as FOXP3, IL2RA, TIGIT, and others. Indeed, this common Treg cell mRNA signature overlapped strongly with published signatures (Miyara et al., 2009; Schmidl et al., 2014), especially with the one reported by Miyara et al. (2009) whose cell populations were closest to the ones studied here (Figure 2E and S2B). In sharp contrast, only three molecules (FOXP3, SHMT2, and SWAP70) overlapped between our own proteomic and transcriptomic core Treg cell signatures (Figure 2F). The discordance between mRNA and protein was not specific for the Treg cell signature. In fact, of the 553 mRNAs and the 409 proteins that were significantly differentially expressed between the five CD4+ T cell subsets and could be quantified at both levels, only 48 overlapped (Figures 2G and S2C). Direct measurement of proteins can therefore yield markedly different results than measurement of mRNAs (at least in cells analyzed at steady state), as is often found in these types of studies (Vogel and Marcotte, 2012). Protein Signatures Differ from mRNA Signatures To quantitatively compare protein and mRNA levels, protein data were analyzed using intensity-based absolute quantification €usser et al., 2011). Protein and mRNA abun(iBAQ; Schwanha dance spanned more than five orders of magnitude (Figures 3A and 3D). Most abundant were ribosomal and metabolic proteins (Figure 3B and Table S3). The most abundant mRNAs encoded ribosome components and molecules involved in (immune) cell Immunity 48, 1046–1059, May 15, 2018 1047

A

n=422 proteins (p<0.05)

D

CD45RA

mTconv D3

mTconv D2

mTconv D1

Fr.IIID1

Fr.III D2

Fr.III D3 eTreg D3

eTreg D1

eTreg D2

nTreg D3

nTreg D1

5

nTreg D2

3

nTconv D3

Cluster:

2

nTconv D2

4

nTconv D1

1

1 2

CD25

B

3 10

2 3 3

0

4 1

2

2

4

5

5

3 -10

5

5 4 -15

-10

-5

0

5

10

15

PC1 (29.7%)

7

0.99

Fr.III

nTconv

nTreg

eTreg

mTconv

6

nTreg (CD25hi, CD45RA+) 5 eTreg (CD25hi, CD45RA-) 4

eTreg nTconv

0.95

mTconv

Fr.III

0.91

nTreg

8 Pearson Correlation

nTconv (CD25-, CD45RA+) 2 mTconv (CD25-, CD45RA-) 3 Fr.III (CD25int, CD45RA-) 1

C

4

1

-20

PC2 (16.4%)

1

9 10 -2.5

0

+2.5

Z-score (expression)

Treg signature

Figure 1. Proteomic Profiles of CD4+ T Cell Subsets (A) Representative flow cytometric CD45RA and CD25 profiles of the human CD4+ T cell subsets subjected to MS analysis: nTconv (1, blue), mTconv (2, cyan), Fr. III population (3, green), nTreg (4, red), and eTreg (5, orange) cells. (B) PCA plot of proteomes (each square represents subset of a single donor). (C) Heatmap depicting PC coefficients of subset proteomes. (D) Hierarchical clustering and heatmap showing relative protein expression values (z-score and log2-transformed LFQ protein intensities) of the 422 differentially expressed proteins (FDR < 0.05; S = 0) between the subsets. Columns correspond to samples from individual donors (D1–3). Grey-scale boxes (left) indicate clusters. Dashed lines demarcate the common Treg cell protein signature (see Table S1). Data are from three independent experiments, each with technical triplicate samples.

signaling and function (Figure 3E and Table S3). Differentially expressed proteins, but not mRNAs, were generally more abundant (Figures 3C and 3F). Expression levels of all 4,792 mRNA-protein pairs from our datasets, that could be mapped at both the transcriptome and proteome level, had a correlation coefficient of approximately 0.44 ± 0.01 (Figure S3), similar to the values described for other human cells (Wilhelm et al., 2014). Among the 422 differentially expressed proteins, 409 were detected at both mRNA and protein level and could be quantitatively compared (Figure 3G). Protein and mRNA expression levels generally correlated (e.g., FOXP3, 1048 Immunity 48, 1046–1059, May 15, 2018

IKZF2). However, the difference in expression was often sufficiently large only for one of the two (mRNA or protein) to reach the statistical cut off, explaining the limited overlap between the differentially expressed proteins and mRNAs (Figure 2F). Other studies commonly report such discrepancies between transcriptome and proteome data (Vogel and Marcotte, 2012), especially for cells in steady state. This likely reflects differences in stability between mRNA and protein (Jovanovic et al., 2015). These differences are likely less pronounced in cycling cells that have a greater need to synthesize proteins. Indeed, the most proliferative cell types in our analysis (Fr.III and eTreg cells)

A

2 3

5

4

4

0 -5,000

nTconv D1

nTconv D3

nTconv D2

nTreg D1

nTreg D2

nTreg D3

eTreg D3

eTreg D1

1

eTreg D2

1

mTconv D2

3

Cluster:

1

mTconv D1

2

Fr.III D1

2

mTconv D3

4

Fr.III D2

5

4

n=649 transcripts (p<0.05)

Fr.III D3

PC2 (19.8%)

5,000

5 5

B

nTconv (CD25-,- CD45RA+) mTconv (CD25-, CD45RA-) Fr.III (CD25int, CD45RA-) nTreg (CD25hi, CD45RA+) eTreg (CD25hi, CD45RA-)

1

1

2

2

3 3 -10,000

5,000

0

5,000

10,000

15,000

3

PC1 (59.7%)

C

4

D nT

v

n co

T

m

.III eg eg Fr nT r eT r

nv

co

nT

v

T

n co

m

.III eg eg Fr nT r eT r

nv

co

C21orf91 RP11-798M19 BCAS2P2 HERC6 PLCL1 SLC14A1 DNAH8 SWAP70 SWT1 CDK14 FCRL3 FOXP3 IL2RA HMCN1 RTKN2 CTD-2619J13.3 MIR146A RP11-105N14.1 SHMT2 CENPL TIGIT Y_RNA

Current Study

-1.5

5

C21orf91 RGS1 VAV3 CLNK STAM PTPLA TNFRSF13B TNFRSF1B BCAS2P2 TIGIT FOXP3 IL2RA HMCN1 RTKN2 CDK14 FCRL3 CENPL MIR146A RP11-105N14.1 SHMT2 CTD-2619J13.3 PLCL1 Y_RNA

7

8 9

10

+2.5

0

-2

GSE15659 0

6

Treg signature

+1.5

Z-score

Z-score

E

F GSE15659

Up in Protein

G Dif. exp.RNA (n=649)

Current study

19 4

Up in RNA

16

3

18

RNAs 15,486

6

96

505

shared up

FOXP3 SHMT2 SWAP70

48 361

PROTEINS 4,252

Dif. exp. on both (n=48) Dif. exp. PROTEIN (n=409)

Figure 2. mRNA Expression Profiles of CD4+ T Cell Subsets (A) PCA plot of subset transcriptomes (subsets isolated from three donors as in Figure 1). Each square represents subset of a single donor. (B) Hierarchical clustering and heatmap showing relative mRNA expression levels (z-score) of the 649 differentially expressed genes (p < 0.05) among the subsets. Columns correspond to individual donors (D1–3). Dashed white line demarcates the common Treg cell mRNA signature. (C and D) Heatmap of expression levels (z-scores) of mRNAs in the common signature of nTreg and eTreg cells defined here (C) and by Miyara et al. (2009) (D; GEO: GSE15659). Overlapping transcripts are in blue. (E) Venn diagram showing overlap between (C) and (D). (F) Venn diagram showing the overlap in relatively highly expressed (compared to the other cell subsets) proteins and mRNAs in the common Treg cell signatures. (G) Venn diagram showing overlap in differentially expressed mRNAs and proteins. Note that mRNAs were measured for all proteins detected, but not vice versa.

contained the lowest number of molecules differentially regulated at the protein level only (Figure 3H). Some molecules were truly differentially expressed at the mRNA or protein level only (Figure 3H), suggesting specific

layers of regulation. For example, protein but not mRNA abundance was high in both nTreg and eTreg cells for FTH1 (Figure 3G), a protein known to be under stringent translational control (Hentze et al., 2010). STAM2 protein was likewise highly Immunity 48, 1046–1059, May 15, 2018 1049

A

B

All proteins (n=5,732) Dif. expressed (n=422*)

C 1000

600 400 200 0 3

4

5

6

7

8

9 10 11

400 200 0 0

1

2

Normalized iBAQ (log10)

E

F Norm. FPKM (log10)

3000

n e tio om ta os en ib s s R ho pre g g xP n O ge alin lin i na nt gn A si me sig R o n TC eas phi on ot ro ati Pr rot ad r eu g N de sis A o N yt R oc e ng i d l m En so gna so si Ly lin g su le lin In yc a g l c gn lin el i a C s gn g bB si lin Er R na b. b. ta ta TO ig m K s me ir me P de a te A p M oti re pha . le ch s b uc t ho ta N ma r p m e s is o M it ne n si os di tio he In imi ica nt l sy r es Py rep bio ul ion A n t ec N D yca ling ol rac l m te -g na O sig ion r in s es pto de 3 p5 adh ce ca e s l r a el C ine nt c ok e yt m g C ple alin . . om n b b C sig eta eta + m m ng on d li i a2 C ine ci na act ur ic a ig er Ta le g s int no ho r Li ge pto e ed ec -r M

2000 1000 0 0

1

2

3

4

5

6

7

8

H

EC

Transcripts in bin

4000

-1

0

1

2

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

RNA (z-score FPKM)

5

6

-1.0

-0.5

0.0

0.5

1.0

1.5

RNA (z-score FPKM)

Metabolism

FTH1 -1.5

-1.0

-0.5

0.0

0.5

1.0

1.5 1.0 0.5 0.0 -1.5 -1.0 -0.5

1.5 1.0 0.5

STAM2

1.5

RNA (z-score FPKM)

FTH1 FOXP3

-1.5

-1.0

-0.5

0.0

0.5

g

re g

re

1.5

Cell signaling

Division & Migration

Legend

Other

▲ Protein ▲ Protein = RNA ▲ RNA

▼ Protein = Protein ▼ RNA ▲ RNA RNA abundance

to

Fr

uc

Pe

nt

os e

in

eT

nv

Fr .II I nT

co

nv

0 co

1.0

RNA (z-score FPKM)

eTreg Fr.III mTconv

10

T

9 10 11

I

20

nT

8

STAM2

nTconv nTreg

m

7

mTconv PROTEIN (z-score iBAQ)

IKZF2 -1.5

-1.5 -1.0 -0.5 0.0

FOXP3

PROTEIN (z-score iBAQ)

1.5 1.0 0.5 0.0

PROTEIN (z-score iBAQ)

STAM2

FTH1 STAM2

-1.5 -1.0 -0.5

1.5 1.0

FOXP3

IKZF2

1

RNA only Protein only

30

4

se Pe ter an nto con d se ve St M p rs ar an h io ch o an Pyr nos sp ns d uv e ha su at m te G e G lu cro meta ly th s e b. co at e ta io m b. sa 2ne et m O Bi ino xoc T m ab. os g a C et r l y y Am n c bo A c ab. on xy y t h in e c oa sis de lic le cy o gr ac l-t f a ad id R m at N in io Ph A o n os s a ph T yn ci at M CR the ds id A s s yl P ig is in K n os s al i ig . N tol na Fk si l. H B s gna IF ig l. AM -1 na P si l. C R K gna he a si l. N p1 gn m O D P ok sig al. -li I3 in n ke K e a R IG T re Ak sign l. -I- N ce t s a lik F- pto ig l. e fam r na cG rece ily sign l. M pt sig al. P- or n PK si al G gn . p s al. m 53 igna TO si l. J R gn Le A AK Wn sig al. uk dh -ST t s na oc er AT ig l. R yt en s na eg e t ig l. ul tra ju n at a io D ns ncti l. n N m on ac A ig s tin re rat i o p cy li n to ca s t C k io C T e el n el l a F igh ll ceton dh oc t j yc R es al a unc le ib io d ti os n h on Pr o m e ot Mme ole sion ei i b c n sm io ul pr a ge es N oc tc n K es h es ce s re is ll m Ph ing pai ed ag in r ia o E te d Lys somR ph o e ag so oc me yt os is

% Regulated proteins/transcripts

40

3

nTconv

Fr.III

0.5

PROTEIN (z-score iBAQ)

FTH1

RNA (z-score FPKM)

H

9 10 11

0

PROTEIN abundance

-2

8

Normalized FPKM (log10)

-1.5 -1.0 -0.5 0.0

1.5 1.0 0.5

PROTEIN (z-score iBAQ)

-1.5 -1.0 -0.5 0.0

STAM2

7

1000

0

eTreg

IKZF2 FOXP3

6

2000

9 10 11

nTreg

FTH1

5

3000

Normalized FPKM (log10)

G

4

All transcripts (n=20,748) Dif. expressed (n=649*)

4000

Transcripts in bin

D

3

Normalized iBAQ (log10)

high

2

600

Density

1

800

low

0

Proteins in bin

Norm. IBAQ (log10)

800

e om me os o ib os R e s lic si ay Sp oly w c b. th ly s G ho e eta pa l xP c m te O cy te ha A oa sp TC an ho . op p ab et Pr ose e b. m is . nt so ta e m es ab Pe tea me os th et o te nn yn m Pr va /Ma on ios de ru e ti b ti o a Py tos d A le a N c uc gr tR u Fr de yl- d n g A c n . N a a b lin R ino cid eta na m a m b. ig ir A o e s a a n t n i r m di e o ep A imi d m ept n r r ci c io Py y a e re cis s or tt lik ex ct Fa -I- de fa IG i R ot is on le os ti uc t rip s N cy sc si do an he nt g E n l tr a le sy lin as c io a B l cy b ign n el a s on C lyc ine cti -g k ra N mo nte i ng he C RE ali ase A gn er SN si ym ing R ol al TC p n A ig N s R R s TO si m pto s i po tos ling n A cy na tio do sig ada En lin gr e su d In ine s Ly

Proteins in bin

1000

Figure 3. Pathway Enrichment in Differential mRNA or Protein Expression €usser et al., 2011), and (D) all (A and D) Normalized abundance distribution of (A) all proteins, expressed in log10 values of absolute iBAQ intensities (Schwanha mRNAs expressed as log10 values for fragments per kilobase of transcript per million mapped reads (FPKM) (colors match intensity quartiles). (B and E) KEGG Gene ontology (GO) analysis, showing pathway enrichment in all proteins (B) or mRNAs (E) measured. (C and F) Abundance distribution of differentially expressed proteins (C) and mRNAs (F) among CD4+ T cell subsets (black bars) compared to all proteins or mRNAs identified (gray bars). (G) Density plots showing relative protein and mRNA levels (expressed as in A and D) of the mapped 409 differentially expressed molecules (examples annotated). (H) Percentages of molecules differentially expressed at protein (black) or mRNA (gray) level only in each subset. (I) Clustering of enriched KEGG pathways according to their significant elevated or decreased expression at mRNA and/or protein level (color key on the right) in the CD4+ T cell subsets.

abundant in mTconv cells, despite low mRNA levels (Figure 3G). On the other hand, it was highly expressed at the mRNA but not at the protein level in both nTreg and eTreg cells (Figure 3G), explaining why it was in the common Treg cell transcriptome but not proteome signature. Combined proteome and transcriptome analysis reinforces confidence in pathway analyses. Such combined data strongly argued for example that the nuclear factor kB (NF-kB) and 1050 Immunity 48, 1046–1059, May 15, 2018

JAK-STAT pathways are desensitized in eTreg cells as compared to the other CD4+ T cell subsets (Figure 3I). Together, our findings underscore the importance of proteomic analysis for the functional characterization of cell types. Treg Cells Share a Stable Protein Signature The common Treg cell signature defined above consisted of 22 proteins with higher expression and 29 proteins with lower

A

B

D

C

Figure 4. The Common Treg Cell Protein Signature (A) Unsupervised hierarchical clustering and heatmap of expression levels (z-scores) of the 51 proteins in the common signature of nTreg and eTreg cells. (B) Functional and spatial representation of the common Treg cell signature proteins (red denotes higher and blue lower expression in Treg cells than the mean value in all five CD4+ T cell subsets). Proteins in gray refer to pathways in which signature proteins are known to operate. (C and D) Proteomes were measured from in vitro expanded Tconv and Treg cells, unstimulated (US) or stimulated for 24 hr with anti-CD3 and anti-CD28 mAbs. (C) Hierarchical clustering and heatmap showing the z-score and log2-transformed LFQ protein intensities of the 320 differentially expressed proteins (FDR < 0.05) between the indicated populations. GO terms corresponding to numbered clusters demarcated by dashed lines are shown in Figure S4H. Data are from three independent experiments, each with biological triplicates. (D) Unsupervised hierarchical clustering and heatmap of relative expression levels (z-scores) in expanded cells of 46 proteins detected from the common Treg cell signature. Asterisks indicate significantly different expression between in vitro expanded Treg and Tconv cells (FDR < 0.05). Proteins in red deviate from the expression pattern in the ex vivo common Treg cell signature.

expression in Treg cells than in the other CD4+ T cell subsets (Figure 4A). Immunoblotting and flow cytometry confirmed the proteomic data for eight of these proteins (Figures S4A–S4C). This signature cannot be found in an earlier proteomic Treg cell dataset (Figures S4D–S4F; Procaccini et al., 2016), which covered less of the proteome and was based on bulk CD4+CD127
CD25+ and CD4+CD127+CD25
populations. We could, however, trace back most of the core Treg cell signature upon reanalysis of a recently published proteomic dataset covering human CD4+ T cell subsets (Figures S4D–S4F; Rieckmann et al., 2017). The common Treg cell signature included FOXP3 and IKZF2 (Helios), metabolic proteins GK, UGP2, and SHMT2, iron storage proteins Ferritin heavy and light chains (FTH1, FLT), as well as lysosomal proteins ASAH1, GGH, GUSB, SGSH, and PLBD2 that were all expressed at high abundance in Treg compared to Tconv cells (Figures 4A and 4B). The signature also included the glycolytic enzymes HK1 and ME2, the fatty acid oxidation enzyme APOO, and the mitochondrial fatty acid transporter (CPT1A), all expressed at low abundance in Treg compared to Tconv cells. A number of signaling molecules was prominent in the common Treg cell signature. High expression of the deubiquitinase OTULIN, an inhibitor of TNFa-induced NF-kB activation (Keusekotten et al., 2013) and low expression of TNFRSF1A adaptor

TRADD (Figures 4A and 4B), together with high expression of TNFRSF1B at the mRNA level (Table S2), suggest that Treg cells exhibit adaptations in TNFR signaling. The signature also suggests adaptation of the PI3K/AKT/mTOR pathway (Figures 4A and 4B), with high expression of the lipid phosphatase INPP5D (SHIP-1), that inhibits PI3K-AKT signaling (Liu et al., 1999), and low expression of the mTOR activators RPS6KA1 and RPS6KA3. Finally, the low expression of STAT4 and NFATc2 in Treg cells stood out. To determine the stability of the protein expression patterns identified above, we performed a proteomic analysis after T cell expansion in vitro. nTreg and nTconv cells were expanded for 2 weeks in vitro by CD3, CD28, and IL-2 stimulation, rested for 4 days, and validated for their identity by FOXP3, Helios, CTLA-4, and CD25 staining and FOXP3 gene TSDR methylation (Figures S4G–S4J). Remarkably, the Treg cell-specific expression pattern of most of the proteins in the Treg cell core signature was largely conserved, even after re-activation via the CD3 and CD28 pathways (Figures 4C, 4D, and S4K and Table S4). These findings suggest that the common Treg cell signature reflects core Treg cell properties, rather than their recent in vivo exposure to certain stimuli. We conclude that Treg cells intrinsically differ from Tconv cells in certain household and metabolic functions, as well as in the configuration of important signaling pathways. Immunity 48, 1046–1059, May 15, 2018 1051

B

A mTconv D3

FEN1

CD74

MCM3

HLA-DR

MCM2

YWHA3 MCM4 HSPA1A

HAT1 CASP3 EZR

FAS

TUBB DNAJB11

Apoptosis

TUBA1B CORO1C

Phagosome VDAC1 H1F0

eTreglow

PRKCB

IL7R

Chemokine signaling pathway

CARD11

STAT3 DPP4

NFKB1

NR3C1

NF-kappa B signaling pathway NFKB2

SVIL

C 1000

nTreg eTreg mTconv

Ratio over FOXP3

100

10

1

0.1

N FA TC 1

ST AT 4

FO XP 1

YY 1 R U N X1 N FA TC 2 BC L1 1B

0.01

ST AT 3

+2.5

MCM6

ST AT 1

mTconv D2

mTconv D1

Fr.III D1

Fr.III D2

eTreg D1

Fr.III D3 eTreg D3

eTreg D2

nTreg D3

nTreg D2

nTreg D1

nTconv D2

nTconv D3

nTconv D1

0

DNA replication

Ag processing & presentation

SAMHD1 RFTN1 FEN1 TUBB YWHAE MYH9 DHRS1 LRRC59 TFRC HSPA1A CORO1C DPP3 HIST1H1B TUBA1B VDAC1 FAS EZR DNAJB11 APOL2 GBP5 FAM160B1 GBP2 LAP3 STXBP2 CISD2 MCM6 MCM4 MCM2 PGM2L1 HLA-DRA ENTPD1 MTHFD1 NSF CASP3 CD74 MCM3 ATAD2 HAT1 GNB2 SMC6 CCR4 FHL1 STAM2 CKAP4 CBR3 APPL1 SVIL TRAT1 ASAP1 TTC37 PITPNC1 H1F0 ACO1 NDRG2 GIMAP7 GIMAP8 GIMAP5 FKBP5 SERPINB1 ESYT2 CARD11 NFKB2 THEMIS AMPD3 VAV1 CCT8 NFKB1 INPP4A STAT3 PRKCB FAM175A DPP4 INPP4B PDCD4 IARS2 ZNF512B SEPT9 LGALS3BP IL7R NR3C1

-2.5

eTreghigh

Z-score (expression) Figure 5. The eTreg Cell Protein Signature (A) Heatmap showing z-score and log2-transformed LFQ protein intensities across the five CD4+ T cell populations of preferentially high (eTreghi) or low (eTreglo) abundance proteins in eTreg cells (center three columns). Each column corresponds to cells from a different donor (D1–3). (B) GO analysis (STRING) of pathways (confidence view) enriched within eTreghi (top) and eTreglo (bottom) protein clusters. Thicker lines represent stronger associations. (C) Estimated copy number ratio of FOXP3-interacting transcription factors to FOXP3 in nTreg (red), eTreg (orange), and mTconv (cyan) cells, determined by the proteomic ruler approach (Wisniewski et al., 2014). Dashed line represents a ratio of 1. Each symbol in represents an individual donor (same as Figure 1).

A Specific Protein Signature Defines eTreg Cells Apart from the common Treg cell signature, eTreg cells were defined by unique protein clusters with relatively high (eTreghi) and low (eTreglo) expression compared to the other CD4+ T cell subsets (Figure 5A). The eTreghi cluster included proteins 1052 Immunity 48, 1046–1059, May 15, 2018

involved in DNA replication (MCM2, -3, -4, -6, -7, FEN1), mitosis (CORO1C, TUBA1B, TUBB, MYH9) (Figures 5A and 5B), and apoptosis (FAS, CASP3), suggesting that these cells are actively dividing, as reported (Miyara et al., 2009). Of the 41 proteins in the eTreghi cluster, 17 still exhibited an upward trend after Treg

B

C 500

6

pSTAT1 (MFI)

8

4

pSTAT4 (MFI)

AA

C TT T

150

C AG C TG TT G TT C T TT TA TG TA T AA TM TT C TG AG T G TG R AC C TG AC AC AR C TY SN W TT TT C N AN N TG TC A G A TT AA AN TC RT AA A KG AC G A G TG G TC

10

100

50

0

0

0

10

20

30

***

**

30

60

eTreg nTreg Fr.III mTconv nTconv

**

*

300 200

pSTAT4 (MFI)

500

* *

10

exp. Tconv

20000

exp. Treg

15000 10000 5000

20

30

8

***

nTreg eTreg

*

6 4 2

CD3/28: - + + + IFN : - - + IL-12: - - - +

Time (min)

nTconv mTconv Fr.III

***

0

60

10

Time (min)

20

U S

60

30

10

20

0

U S

0

10

Time (min)

F 25000

1000

0

Time (min)

E 1500

0

60

IFN (% positve)

D

pSTAT1 (MFI)

400

100

2

U nk no R wn EP I M N1 LL T LE 7 TA F1 U F6 nk L no wn TC R F3 UN ES X1 R ST RA U A nk T4 no N wn FA T PA C FO X4 XF 2 ZI C 2

Enrichment of TF binding sites (-log p-value)

A

-+ + + - - + - - - +

-+ + + - - + - - - +

-+ + + - - + - - - +

-+ + + - - + - - - +

I H

*

mCherry STAT4

0

60 40 20

+ IL-12 + IFN

STAT4-mCherry

+

IF N

IL -1 2

A /IO PM

IF N +

A / IO

PM

PM

PM

PM

+

A /IO

IL -1 2

0

A /IO

IF N +

A /IO PM

PM

A /IO

+

PM

A / IO

IL -1 2

0

80

+

5

5

mCherry STAT4

A /IO

10

mCherry Control

*

100

PM

IL-2 (%positive)

IFN (% positive)

*

15

FOXP3 (% positive)

10

20

Control-mCherry

mCherry Control

A /IO

G

+ IL-12 + IFN

Day 7 (ex vivo) Unstimulated 0 0.82

CD3/28 0

29.8

CD3/28 + IFN 0 60.9

FOXP3

K 2500

T-bet

nTconv 96.7

0.23

69.9

0.25

38.8

0

0.63

0

1.61

0.037

46.4

nTreg 3.16

96.2

1.70

96.7

2.59

2000

T-bet (MFI)

2.45

50.9

CXCR3

15000

*

1500

CXCR3 (MFI)

J

*

1000

**

10000

nTconv nTreg

**

**

5000

500 0

0

CD3/28: - + + - + + IFN : - - + - - + d3

d7

- + + - + + - - + - - + d3

d7

CD3/28: - + + - + + IFN : - - + - - + d3

d7

- + + - + + - - + - - + d3

d7

Figure 6. Selectively Deficient STAT4 Activation in Treg Cells (A) Enrichment (as –log p value of the enrichment score) of transcription factor (TF) target motifs (above dots) of underrepresented genes in the eTreg cell transcriptome. (legend continued on next page)

Immunity 48, 1046–1059, May 15, 2018 1053

cell expansion in vitro, with 5 reaching statistical significance (Figure S5A). Likewise, 25 out of 39 proteins in the eTreglo cluster retained low expression in in vitro expanded Treg cells. The eTreglo cluster contained multiple GIMAPs that regulate apoptosis sensitivity (Figures 5A and 5B). Importantly, almost 40% of the proteins in the eTreglo cluster have functions in cellular signaling (Table S5). They include multiple components of the NF-kB pathway (PRKCB, DPP4, NF-kB1, NF-kB2), cytokine receptor pathways (IL-7R, INPP4B, STAM2, STAT3), and the TCR pathway (TRAT1, VAV1, THEMIS). The fact that comparatively low expression was maintained for many of these proteins in Treg cells even after in vitro culture (Figure S5A) suggests that this apparent desensitization of signaling pathways is not just a reflection of negative feedback to activating signals via these pathways in vivo. FOXP3 physically interacts with key transcription factors and can quench their ability to transactivate Tconv cell effector genes, including IL17A and IL4 (Li and Zheng, 2015; Li et al., 2014; Rudra et al., 2012). Vice versa, a factor such as YY1 can inhibit FOXP3mediated control of the Treg cell program (Hwang et al., 2016). To determine whether the relative concentration of FOXP3 in Treg cells is sufficient to ‘‘overwhelm’’ such interacting factors, we determined protein copy numbers per cell, using the proteomic ruler method (Wi sniewski et al., 2014). We found that FOXP3 outnumbers many of its partner proteins in eTreg cells, with the greatest excess of FOXP3 over transcription factors such as NFATc1 and STAT4 (Figure 5C). For some transcription factors (YY1, RUNX1, and NFATc2), however, FOXP3 excess is only 2- to 3-fold, which may explain why even modest reduction in FOXP3 allows the production of effector cytokines by Treg cells (Wan and Flavell, 2007). Finally, the results show that the expression level of FOXP3 in mTconv cells may be insufficient to effectively neutralize its partner transcription factors (Figure 5C). eTreg Cells Exhibit Blunted Pathways Leading to Expression of Effector Genes Ingenuity pathway analysis (IPA) highlighted adaptations in TCR, pattern recognition receptors (PRR), cytokine receptor, and TNFRSF family-induced NF-kB pathway signaling in eTreg as compared to Tconv cells (Figure S5B). Furthermore, protein but not mRNA levels of NF-kB1 (p50), NF-kB2 (p52), and RELA (p65) were lower in eTreg cells (Figures S5C and S5D), as confirmed by flow cytometry for NF-kB1 and NF-kB2 (Figures S5E–S5G). Accordingly, anti-CD3- and anti-CD28-induced nu-

clear translocation of NF-kB1 was blunted in eTreg cells, but not in nTreg cells (Figures S5H and S5I). NFATc1 and NFATc2 protein levels were lower in both nTreg cells and eTreg cells (Figure S5J). NFATc2, furthermore, appeared as low abundant in the common Treg cell signature and low expression of NFATc2 was maintained after Treg cell expansion in vitro (Figures 4A and 4D). Finally, genes with NFATc2 binding elements in their promoters were underrepresented in the eTreg cell transcriptome (Figure 6A), suggesting that this factor is less active in these cells in vivo. Inflammatory cytokines control critical aspects of Treg cell behavior but may challenge their stability (Arvey et al., 2014; Josefowicz et al., 2012; Tan et al., 2016; Yu et al., 2015). Among STAT transcription factors, which relay cytokine receptor signals, STAT3, STAT6, and especially STAT4 exhibited low expression in eTreg cells. STAT4 expression was low both at the mRNA and protein level (Figures S6A and S6B) and low protein expression of STAT4 was part of the common Treg cell signature (Figure 4). Genes with STAT4 binding elements in their promoters were also underrepresented in the eTreg cell transcriptome (Figure 6A), supporting the notion that STAT4 activity is reduced in eTreg cells in vivo. STAT4 is activated in human CD4+ T cells by phosphorylation of Y693 in response to IL-12 and type I IFN, with the latter inducing the most robust early STAT4 phosphorylation in human CD4+ Tconv cells (compare Figures 6C and S6C). These cytokines consistently induced less phosphorylation of STAT4 in Treg cells than in Tconv cells, freshly isolated from blood (Figure 6C), and this difference was even more profound in in vitro expanded cells (Figure 6E). STAT4 is a major regulator of IFNg gene expression (O’Shea et al., 2011). We therefore reasoned that low expression of STAT4 might insulate Treg cells against induction of IFN-g by inflammatory cytokines. Consistent with this hypothesis, IFN-a and IL-12 failed to induce IFN-g production in Treg cells (Figures 6F and S6D–S6H). However, when STAT4 was deliberately overexpressed, these cytokines did provoke production of IFN-g in Treg cells (Figures 6G and S6I). Remarkably, overexpression of STAT4 also induced production of IL-2 (Figure 6G) and marked loss of FOXP3 expression (Figures 6H and 6I), especially when Treg cells were stimulated with IFN-a or IL-12. These findings suggest that low expression of STAT4 helps to protect Treg cell identity in an inflammatory environment. Because Treg cells require input from inflammatory cytokines (Arvey et al., 2014; Josefowicz et al., 2012; Tan et al., 2016; Yu

(B and C) Geometric MFI of phosphorylated STAT1 (B) and STAT4 (C) determined by FACS in CD4+ T cell subsets at indicated time points after stimulation with IFN-a (mean ± SEM of n = 3 experiments). Asterisks indicate significant differences between Fr.III compared to nTreg cells. (D and E) Activation of STAT1 (D) and STAT4 (E) in in vitro expanded Tconv and Treg cells measured as in (B) and (C) (mean values ± SEM of n = 3 experiments). (F) Frequency of IFN-g-producing cells at 6 hr after anti-CD3+anti-CD28 mAb and IFN-a or IL-12 stimulation (mean ± SEM of n = 6 experiments). In (B)–(F), asterisks indicate significant differences according to two-way ANOVA followed by a Tukey’s multiple comparison test (*p < 0.05; **p < 0.01; ***p < 0.001). (G and H) Treg cells were transduced with control- or STAT4-expressing vector. Frequency of IFN-g- or IL-2-positive (G) or FOXP3-positive (H) cells after indicated stimulation. Asterisks indicate significant differences according to RM-two-way ANOVA followed by a Bonferroni’s multiple comparison test (*p < 0.05; mean ± SEM of n = 4 experiments). (I) Representative FOXP3 profiles of cells treated as in (H). (J and K) nTconv and nTreg cells were cultured for 3 or 7 days in absence or presence of anti-CD3 and anti-CD28 mAb with or without IFN-a and analyzed for T-bet and CXCR3 expression by flow cytometry. (J) Representative FACS plots. (K) MFI (mean ± SEM of n = 4 experiments). Asterisks indicate differences according to RM one-way ANOVA followed by a Dunnett’s multiple comparison test (*p < 0.05; **p < 0.01).

1054 Immunity 48, 1046–1059, May 15, 2018

A

B

1

2 2

0

2

4

1

2

3 23 3

3 3’ 3’

5

3’

3’

1 1

4

5

2 Loading...

nTconv (CD127+, CD25-, CD45RA+) mTconv (CD127+, CD25-, CD45RA-) Fr.III 127+ (CD127+, CD25int, CD45RA-) Fr.III 127- (CD127-, CD25int, CD45RA-) nTreg (CD127-, CD25hi, CD45RA+) eTreg (CD127-, CD25hi, CD45RA-)

F

Fr.III CD127+

eTreg

Fr.III CD127-

Total

5

4

5

-10

nTreg nTreg nTreg nTconv nTconv nTconv mTconv mTconv mTconv mTconv mTconv mTconv Fr.III 127+ Fr.III 127+ Fr.III 127+ Fr.III 127Fr.III 127Fr.III 127eTreg eTreg eTreg eTreg eTreg

PC2 (13.1%)

10

2

5

5

CCR4-CD49d-

4

-10

0

10

re g

re g

CCR4-CD49d+

eT

TSDR Methylation (%)

nv

IL-2

CCR4+CD49d-

co

-1.5

CD49d C

CD49d +2.5

30

100

# ***

20 10

**

# ***

0

80

25

**

60 40 20 0

IL-17+ (%cells)

IFNg+ cells (%)

eTreg

40

C C C R4 - T C C R4 C- D ota C C 4 l R D 9d 4+ 4 C 9d + D 49 C C dC R4 - T C C o C R4 - D ta C C 4 l R D 9d 4+ 4 C 9d + D 49 C C dC R4 - T C C R4 C- D ota C C 4 l R D 9d 4+ 4 C 9d + D 49 d-

Fr.III CD127

-

CCR4 -2.5

IL-17

IFN-γ

20

Fr.III CD127+

# #

15

Fr.III CD127eTreg

10 5

**

0 C C C R4 - T C C R4 C- D ota C C 4 l R D 9d 4+ 4 C 9d + D 49 C C dC R4 - T C C o C R4 - D ta C C 4 l R D 9d 4+ 4 C 9d + D 49 C C dC R4 - T C C R4 C- D ota C C 4 l R D 9d 4+ 4 C 9d + D 49 d-

G Fr.III CD127

IL-17

IFN-γ

+1.5

Z-score

E +

IL-17

IFN-γ

CCR4 C

IL-2+ cells (%)

eT

re g

D

Fr .II I1 27 Fr .II I1 27 + m T

80 60 40 20

C C C R4 - T C C R4 C- D ota C C 4 l R D 9d 4+ 4 C 9d + D 49 C C dC R4 - T C C o C R4 - D ta C C 4 l R D 9d 4+ 4 C 9d + D 49 C C dC R4 - T C C R4 C- D ota C C 4 l R D 9d 4+ 4 C 9d + D 49 d-

nv co

m T

nT

co

nv

C

Fr .II I1 27 + Fr .II I1 2 nT 7

PC1 (25.9%)

Z-score

Figure 7. Proteomics Reveals Functional Heterogeneity among Fr.III and eTreg Cells (A) Hierarchical clustering and heatmap showing z-score and log2-transformed LFQ protein intensities of 164 differentially expressed proteins (FDR < 0.05; S = 0.4) among 6 CD4+ T cell subsets defined as indicated. Each column corresponds to cells from a different donor (n = 3–6). Arrow highlights a cluster differentiating eTreg cell and Fr.III CD127
cells. (B) PCA plot of the six CD4+ T cell subset proteomes (squares represent different donors). (C) Heatmap representing percentage of FOXP3 TSDR DNA methylation per subset (see color scale). (D) Heatmap of average CCR4 and CD49d expression levels (z-scores) measured by MS. (E) Representative FACS plots of CD49d and CCR4 in CD127+ Fr.III (dark green), CD127
Fr.III (light green), and CD127
eTreg cells (orange). (F) Representative FACS plots of cytokines produced by the different cell subsets stimulated with PMA and Ionomycin for 4 hr (total) or gated for CD49d and CCR4. (G) Percentage of cells IFN-g-, IL-17-, and IL-2-positive cells after 4 hr of treatment with PMA and ionomycin (mean ± SEM of n = 3 experiments). Statistical analysis by two-way ANOVA followed by a Turkey’s multiple comparison test (**p < 0.01; ***p < 0.001; indicate significant differences compared to the total population; #p < 0.05 indicate significant differences compared to CCR4+CD49d
population).

et al., 2015), we examined whether Treg cells can respond to type I IFN at all, despite low expression of STAT4. Indeed, STAT1 phosphorylation was induced equally in both Treg and Tconv cells (Figures 6B and 6D). Moreover, IFN-a readily induced expression of the transcription factor T-bet and the chemokine receptor CXCR3 in anti-CD3- and anti-CD28activated Treg cells (Figures 6J, 6K, and S6J–S6L). Together, these findings suggest that selective low expression of STAT4 allows Treg cells to respond to inflammatory cytokines (such as for homing to inflamed tissues), without compromising Treg cell identity. Different Populations of Cells Can Be Distinguished in Fr.III As originally defined (Miyara et al., 2009), Fr.III cells are CD4+CD25+ and express FOXP3 but can produce inflammatory cytokines and lack robust suppressive capacity in vitro. The fact that many cells in this population express CD127 (Figure S1B) suggests that it is a mixed population containing cells resembling either Treg or Tconv cells. To better characterize this population, we separated it into two fractions based on CD127 expression. Proteomic analysis of each fraction was then performed and

analyzed together with the proteomes of nTconv, mTconv, nTreg, and eTreg cells (Figures S7A and S7B). Hierarchical clustering and PCA showed that the CD127+ subpopulation was closely related to mTconv cells (Figures 7A and 7B), confirming the usefulness of CD127 to distinguish between Tconv and Treg cells. Remarkably, the CD127
subpopulation was almost indistinguishable from eTreg cells. Only three proteins (LGAL10, PRG2, and FAU) were sufficiently differentially expressed between these two subsets to be identified as such (Table S6). The close proteomic relationship between CD127
Fr.III and eTreg cells suggests that the former population may be a genuine Treg cell population. However, the TSDR in FOXP3 was much more methylated in CD127
Fr.III cells than in eTreg cells (Figure 7C). Furthermore, CD127
Fr.III cells contained a large proportion of cells that produced one or more effector cytokines (Figure 7F, top row), unlike eTreg cells, which mostly lacked this ability. Not all cells in the CD127
Fr.III population produced effector cytokines and not all cells in the eTreg cell population lacked the capacity to produce such cytokines, suggesting that even these better-defined populations might still be heterogeneous. We therefore searched for markers in our proteome dataset that Immunity 48, 1046–1059, May 15, 2018 1055

might help distinguish between cells with different functional properties. Two markers, which have previously been found on Treg cells (Iellem et al., 2001; Kleinewietfeld et al., 2009), correlated positively (CD49d) or negatively (CCR4) with the relative ability of populations to produce effector cytokines (Figures 7D, S7C, and S7D). Flow cytometric analysis showed that these markers exhibit a mutually exclusive expression pattern (Figure 7E) and identify cells with different functional properties within the eTreg cell and Fr.III populations. Those few cells in the eTreg cell population that produced IL-2 or IL-17 were most prominently found among cells expressing CD49d, consistent with an earlier report (Kleinewietfeld et al., 2009). However, only a CD49d
CCR4+ phenotype completely excluded eTreg cells producing effector cytokines altogether (Figure 7G). A similar trend was observed within the CD127
Fr.III population, where IL-17- and IFN-g-producing cells were absent from the CD49d
CCR4+ population. This population was, however, still different from the corresponding CD49d
CCR4+ eTreg cell subset in its ability to produce IL-2 (Figure 7G). The relative protein levels of FOXP3 correlated inversely with the proportions of cells producing cytokines in each population. Thus, FOXP3 expression gradually decreased from eTreg cells (highest) to Fr.III CD127
and Fr.III CD127+ (lowest, but still higher than in mTconv cells). Furthermore, in each of the populations, the CCR4+CD49d
cells always expressed higher FOXP3 than the CCR4
CD49d+ cells (Figures S7E and S7F). Finally, the TSDR was only fully demethylated in eTreg cells, and mostly methylated in CD127
CCR4+CD49d
Fr.III cells, which only produced IL-2, but none of the other inflammatory cytokines (Figures S7E and S7F). Our proteomic data thus led us to the discovery that CCR4 and CD49d distinguish cells with different abilities to produce effector cytokines within the eTreg cell and Fr.III populations. Combination of these markers with the commonly used panel of Treg cell markers may be useful for cell purification and diagnostic purposes. Their inclusion may, for instance, help clarify the association between the presence of FOXP3+CD4+ T cells in tumors and patient prognosis, especially for cancers such as colon carcinoma, where this relationship has been tenuous (Saito et al., 2016). DISCUSSION We identified proteomic signatures reflecting common Treg and eTreg cell-specific properties. These signatures comprised proteins involved in diverse cellular processes such as iron storage, lysosomal biogenesis, and transport, metabolism, cell signaling, and transcription. The composition of these signatures could not have been predicted on the basis of our own and earlier transcriptome analyses (Miyara et al., 2009; Schmidl et al., 2014), but could be traced back in a recently published proteomic dataset covering human CD4+ T cell subsets (Rieckmann et al., 2017). Importantly, the relative expression pattern of the proteins in the common Treg cell signature and, to a lesser degree, the eTreg cell signature, was maintained after culture in vitro and even upon stimulation via CD3 and CD28. Therefore, our data are a robust resource to generate new hypotheses regarding human Treg cell biology, especially because the maintenance of the signatures in vitro provides an assay system to query their function. 1056 Immunity 48, 1046–1059, May 15, 2018

Apart from identifying heterogeneity within Treg cell populations, our proteomic analysis sheds new light on mechanisms that help Treg cells protect their identity. Key in this identity are stable expression of FOXP3 and inability to produce effector cytokines. How and when Treg cells lose their identity is under debate (Sakaguchi et al., 2013). They may be provoked to produce effector cytokines by chronic inflammation (Bailey-Bucktrout et al., 2013; Feng et al., 2014), but this does not generally happen under acute inflammatory conditions, implying that barriers exist to prevent production of effector cytokines by Treg cells (Miyao et al., 2012; Rubtsov et al., 2010). One such barrier involves inhibition by physical interaction of FOXP3 with transcription factors that promote Tconv effector cell gene expression (Rudra et al., 2012). We found that FOXP3 is sufficiently abundant in eTreg cells to allow such a mechanism to operate. For some transcription factors, the difference in abundance was, however, no greater than 3-fold, such that even modest reduction in FOXP3 expression may dismantle this protective mechanism, consistent with experimental observations (Wan and Flavell, 2007). The Treg cell signatures reveal another barrier against production of effector cytokines: adjustments in signaling pathways downstream of CD3, CD28, TNFRSF, and cytokine receptors. These adjustments are most pronounced in eTreg cells, the cell type most likely to be exposed to inflammatory signals. Weak expression of proteins involved in CD3 and CD28 signaling is consistent with earlier findings that reduced TCR signaling capacity is critical for Treg cell stability in mice (Park et al., 2016; Yan et al., 2015). Treg cells furthermore expressed high amounts of INPP5D (SHIP-1) and low amounts of RPS6KA1 and -3, suggesting desensitization of the AKTmTOR pathway, a key regulator of Tconv cell effector capacity (Chi, 2012). Functional analyses have indeed indicated that this pathway is attenuated in Treg cells (Chi, 2012; Crellin et al., 2007), although this finding is not unanimous (De Rosa et al., 2015; Procaccini et al., 2016). TNFa can destabilize Treg cells but can paradoxically also boost Treg cell expansion without compromising stability (Nie et al., 2013; Valencia et al., 2006; Zanin-Zhorov et al., 2010; Zaragoza et al., 2016), suggesting that unknown elements may modify the response to this cytokine. We found that Treg cells exhibited adaptations in TNFR signaling, including weak expression of TNFR1 mRNA (the protein escaped detection) and its adaptor protein TRADD. On the other hand, these cells abundantly expressed TNFR2 mRNA and the OTULIN protein, an inhibitor of NF-kB downstream of TNFR1 and a suppressor of inflammatory disease (Damgaard et al., 2016). It will be interesting to study whether modulation of these molecules underlies the different effects that TNFa can have on Treg cell stability (Nie et al., 2013; Valencia et al., 2006; Zanin-Zhorov et al., 2010; Zaragoza et al., 2016). Our data suggest that Treg cell identity may be insulated against destabilization by specific holes in its repertoire of transcription factors. For instance, several NF-kB pathway components were relatively weakly expressed. The role of NF-kB in Treg cells is complex. In mice, the NF-kB components c-Rel and p65 are required for thymic differentiation and maintenance of Treg cell identity, respectively (Grinberg-Bleyer et al., 2017; Oh et al., 2017). However, unspecified NF-kB activity

destabilized mature human Treg cells in response to TNFa (Nie et al., 2013; Valencia et al., 2006; Zanin-Zhorov et al., 2010). Low abundance was most pronounced for (p105) NF-kB1 and (p100) NF-kB2 in eTreg cells. These molecules can act as activators and as repressors of NF-kB activity. It can therefore not be predicted whether their low expression inhibits or stimulates NF-kB activity or favors the formation of complexes lacking the p50 and p52 subunits derived from these precursors. Nonetheless, at least NF-kB1 recruits histone deacetylases to FoxP3 to inhibit its expression (Jana et al., 2009; Xiao et al., 2015), suggesting that low expression of NF-kB1 helps to avoid loss of FOXP3 expression in Treg cells. Several STAT transcription factors were weakly expressed in Treg cells. Low abundance of STAT4 protein was part of the common Treg cell signature and was reflected also in the Treg cell transcriptome. STAT4 is a major activator of IFNG in response to certain cytokines and antagonizes Treg cell development in mice (O’Shea and Plenge, 2012; Xu et al., 2011). We here showed that forced expression of STAT4 in mature human Treg cells permitted type I IFN and IL-12 to provoke production of IFN-g. It also inhibited expression of FOXP3, perhaps through recruitment of methyl transferases to the CNS2 enhancer of FOXP3 (Wu et al., 2017). As STAT4 itself contains a binding site for FOXP3, both factors may have a mutually antagonistic relationship (STAT4 mRNA abundance was also low in Treg cells). Murine Treg cells weakly express the IL-12 receptor b2 chain to prevent IFN-g production (Koch et al., 2012). In human T cells, both IL-12 and type I IFN can induce IFN-g production via STAT4 (Rogge et al., 1998). Low abundance of this transcription factor therefore prevents induction of IFN-g production by either of these inflammatory cytokines. Having the cytokine receptors, but lacking STAT4, apparently allows Treg cells to respond to inflammatory cues (e.g., to instill homing properties) without compromising their identity. Perhaps, production of effector cytokines by Treg cells is useful under specific conditions and elevation of STAT4 expression may be one way to accomplish this. Indeed, expression of STAT4 was high in IFNg-producing ‘‘fragile’’ Treg cells, which are associated with relatively favorable prognosis in some tumors (Overacre-Delgoffe et al., 2017). Collectively, our proteomic measurements revealed specific core programs of human CD4+ Treg cell populations that will help increase the understanding of their function and perhaps enable their selective manipulation. The identification of surface markers that distinguish between bona fide Treg cells and CD4+FOXP3+ T cells with the ability to produce inflammatory cytokines may furthermore help to select purer and more stable Treg cells for adoptive transfer therapies. STAR+METHODS Detailed methods are provided in the online version of this paper and include the following: d d d

KEY RESOURCES TABLE CONTACT FOR REAGENT AND RESOURCE SHARING EXPERIMENTAL MODEL AND SUBJECT DETAILS B Healthy blood donors B Cell Lines

d

d d

METHOD DETAILS B Cell isolation and cell sorting B Antibodies B Sample preparation for mass spectrometry (MS) B Mass spectrometry data acquisition B Data Processing and Analysis B RNA Isolation B TruSeq stranded mRNA sample preparation B RNA-Seq data processing B Protein and RNA comparison B In vitro cell expansion B Methylation status of the Treg specific demethylated region (TSDR) in the FOXP3 gene B Western blot B Flow cytometry B Nuclear translocation analysis B Cell stimulation assays B Cloning, lentivirus production and gene transduction QUANTIFICATION AND STATISTICAL ANALYSIS DATA AND SOFTWARE AVAILABILITY

SUPPLEMENTAL INFORMATION Supplemental Information includes seven figures and six tables and can be found with this article online at https://doi.org/10.1016/j.immuni.2018.04.008. ACKNOWLEDGMENTS We thank Jelle ten Hoeve and Ran Elkon for assistance with bioinformatics, Evert de Vries for help with biochemistry, Floris van Alphen for help in processing MS samples, Simon Tol and Erik Mul for help with FACS sorting, and Celia Berkers for insightful discussions. This work was supported by a ZonMW-TOP grant 40-00812-98-13071 to R.A.W.v.L. and J.B., grant ICI-00014 from the Institute for Chemical Immunology to J.B., and grant 1430 from the LSBR (Landsteiner Foundation for Blood Transfusion Research) to D.A. AUTHOR CONTRIBUTIONS E.C., J.B., and D.A. designed the study; E.C., S.d.K., and Y.-y.C. designed and performed experiments and analyzed data, with assistance from M.S. and I.D. M.v.d.B. performed MS measurements, analysis, and interpretation; A.M. and R.A.W.v.L. provided conceptual advice and contributed to the manuscript. E.C., J.B., and D.A. analyzed data and wrote the manuscript. DECLARATION OF INTERESTS The authors declare no competing interests. Received: December 1, 2016 Revised: November 2, 2017 Accepted: April 9, 2018 Published: May 8, 2018 REFERENCES Arpaia, N., Green, J.A., Moltedo, B., Arvey, A., Hemmers, S., Yuan, S., Treuting, P.M., and Rudensky, A.Y. (2015). A distinct function of regulatory T cells in tissue protection. Cell 162, 1078–1089. Arvey, A., van der Veeken, J., Samstein, R.M., Feng, Y., Stamatoyannopoulos, J.A., and Rudensky, A.Y. (2014). Inflammation-induced repression of chromatin bound by the transcription factor Foxp3 in regulatory T cells. Nat. Immunol. 15, 580–587. Bailey-Bucktrout, S.L., Martinez-Llordella, M., Zhou, X., Anthony, B., Rosenthal, W., Luche, H., Fehling, H.J., and Bluestone, J.A. (2013). Self-antigen-driven

Immunity 48, 1046–1059, May 15, 2018 1057

activation induces instability of regulatory T cells during an inflammatory autoimmune response. Immunity 39, 949–962.

(2013). OTULIN antagonizes LUBAC signaling by specifically hydrolyzing Met1-linked polyubiquitin. Cell 153, 1312–1326.

Campbell, D.J. (2015). Control of regulatory T cell migration, function, and homeostasis. J. Immunol. 195, 2507–2513.

Kleinewietfeld, M., Starke, M., Di Mitri, D., Borsellino, G., Battistini, L., Ro¨tzschke, O., and Falk, K. (2009). CD49d provides access to ‘‘untouched’’ human Foxp3+ Treg free of contaminating effector cells. Blood 113, 827–836.

Chen, Z., Barbi, J., Bu, S., Yang, H.-Y., Li, Z., Gao, Y., Jinasena, D., Fu, J., Lin, F., Chen, C., et al. (2013). The ubiquitin ligase Stub1 negatively modulates regulatory T cell suppressive activity by promoting degradation of the transcription factor Foxp3. Immunity 39, 272–285. Chi, H. (2012). Regulation and function of mTOR signalling in T cell fate decisions. Nat. Rev. Immunol. 12, 325–338. Cox, J., and Mann, M. (2008). MaxQuant enables high peptide identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat. Biotechnol. 26, 1367–1372. Cox, J., and Mann, M. (2012). 1D and 2D annotation enrichment: a statistical method integrating quantitative proteomics with complementary highthroughput data. BMC Bioinformatics 13 (Suppl 16 ), S12. Crellin, N.K., Garcia, R.V., and Levings, M.K. (2007). Altered activation of AKT is required for the suppressive function of human CD4+CD25+ T regulatory cells. Blood 109, 2014–2022. Damgaard, R.B., Walker, J.A., Marco-Casanova, P., Morgan, N.V., Titheradge, H.L., Elliott, P.R., McHale, D., Maher, E.R., McKenzie, A.N.J., and Komander, D. (2016). The deubiquitinase OTULIN is an essential negative regulator of inflammation and autoimmunity. Cell 166, 1215–1230.e20. De Rosa, V., Galgani, M., Porcellini, A., Colamatteo, A., Santopaolo, M., Zuchegna, C., Romano, A., De Simone, S., Procaccini, C., La Rocca, C., et al. (2015). Glycolysis controls the induction of human regulatory T cells by modulating the expression of FOXP3 exon 2 splicing variants. Nat. Immunol. 16, 1174–1184. Feng, Y., Arvey, A., Chinen, T., van der Veeken, J., Gasteiger, G., and Rudensky, A.Y. (2014). Control of the inheritance of regulatory T cell identity by a cis element in the Foxp3 locus. Cell 158, 749–763. Ferraro, A., D’Alise, A.M., Raj, T., Asinovski, N., Phillips, R., Ergun, A., Replogle, J.M., Bernier, A., Laffel, L., Stranger, B.E., et al. (2014). Interindividual variation in human T regulatory cells. Proc. Natl. Acad. Sci. USA 111, E1111–E1120. Gao, Y., Tang, J., Chen, W., Li, Q., Nie, J., Lin, F., Wu, Q., Chen, Z., Gao, Z., Fan, H., et al. (2015). Inflammation negatively regulates FOXP3 and regulatory T-cell function via DBC1. Proc. Natl. Acad. Sci. USA 112, E3246–E3254. Grinberg-Bleyer, Y., Oh, H., Desrichard, A., Bhatt, D.M., Caron, R., Chan, T.A., Schmid, R.M., Klein, U., Hayden, M.S., and Ghosh, S. (2017). NF-kB c-Rel is crucial for the regulatory T cell immune checkpoint in cancer. Cell 170, 1096–1108.e13. Hentze, M.W., Muckenthaler, M.U., Galy, B., and Camaschella, C. (2010). Two to tango: regulation of Mammalian iron metabolism. Cell 142, 24–38. Hwang, S.S., Jang, S.W., Kim, M.K., Kim, L.K., Kim, B.-S., Kim, H.S., Kim, K., Lee, W., Flavell, R.A., and Lee, G.R. (2016). YY1 inhibits differentiation and function of regulatory T cells by blocking Foxp3 expression and activity. Nat. Commun. 7, 10789. Iellem, A., Mariani, M., Lang, R., Recalde, H., Panina-Bordignon, P., Sinigaglia, F., and D’Ambrosio, D. (2001). Unique chemotactic response profile and specific expression of chemokine receptors CCR4 and CCR8 by CD4(+)CD25(+) regulatory T cells. J. Exp. Med. 194, 847–853. Jana, S., Jailwala, P., Haribhai, D., Waukau, J., Glisic, S., Grossman, W., Mishra, M., Wen, R., Wang, D., Williams, C.B., and Ghosh, S. (2009). The role of NF-kappaB and Smad3 in TGF-beta-mediated Foxp3 expression. Eur. J. Immunol. 39, 2571–2583.

Koch, M.A., Thomas, K.R., Perdue, N.R., Smigiel, K.S., Srivastava, S., and Campbell, D.J. (2012). T-bet(+) Treg cells undergo abortive Th1 cell differentiation due to impaired expression of IL-12 receptor b2. Immunity 37, 501–510. Kwon, H.-K., Chen, H.-M., Mathis, D., and Benoist, C. (2017). Different molecular complexes that mediate transcriptional induction and repression by FoxP3. Nat. Immunol. 18, 1238–1248. Li, X., and Zheng, Y. (2015). Regulatory T cell identity: formation and maintenance. Trends Immunol. 36, 344–353. Li, P., Spolski, R., Liao, W., and Leonard, W.J. (2014). Complex interactions of transcription factors in mediating cytokine biology in T cells. Immunol. Rev. 261, 141–156. Liu, Q., Sasaki, T., Kozieradzki, I., Wakeham, A., Itie, A., Dumont, D.J., and Penninger, J.M. (1999). SHIP is a negative regulator of growth factor receptor-mediated PKB/Akt activation and myeloid cell survival. Genes Dev. 13, 786–791. Matys, V., Kel-Margoulis, O.V., Fricke, E., Liebich, I., Land, S., Barre-Dirrie, A., Reuter, I., Chekmenev, D., Krull, M., Hornischer, K., et al. (2006). TRANSFAC and its module TRANSCompel: transcriptional gene regulation in eukaryotes. Nucleic Acids Res. 34, D108–D110. Miyao, T., Floess, S., Setoguchi, R., Luche, H., Fehling, H.J., Waldmann, H., Huehn, J., and Hori, S. (2012). Plasticity of Foxp3(+) T cells reflects promiscuous Foxp3 expression in conventional T cells but not reprogramming of regulatory T cells. Immunity 36, 262–275. Miyara, M., Yoshioka, Y., Kitoh, A., Shima, T., Wing, K., Niwa, A., Parizot, C., Taflin, C., Heike, T., Valeyre, D., et al. (2009). Functional delineation and differentiation dynamics of human CD4+ T cells expressing the FoxP3 transcription factor. Immunity 30, 899–911. Nie, H., Zheng, Y., Li, R., Guo, T.B., He, D., Fang, L., Liu, X., Xiao, L., Chen, X., Wan, B., et al. (2013). Phosphorylation of FOXP3 controls regulatory T cell function and is inhibited by TNF-a in rheumatoid arthritis. Nat. Med. 19, 322–328. Nishikawa, H., and Sakaguchi, S. (2014). Regulatory T cells in cancer immunotherapy. Curr. Opin. Immunol. 27, 1–7. Nosbaum, A., Prevel, N., Truong, H.-A., Mehta, P., Ettinger, M., Scharschmidt, T.C., Ali, N.H., Pauli, M.L., Abbas, A.K., and Rosenblum, M.D. (2016). Cutting edge: regulatory T cells facilitate cutaneous wound healing. J. Immunol. 196, 2010–2014. O’Shea, J.J., and Plenge, R. (2012). JAK and STAT signaling molecules in immunoregulation and immune-mediated disease. Immunity 36, 542–550. O’Shea, J.J., Lahesmaa, R., Vahedi, G., Laurence, A., and Kanno, Y. (2011). Genomic views of STAT function in CD4+ T helper cell differentiation. Nat. Rev. Immunol. 11, 239–250. Oh, H., Grinberg-Bleyer, Y., Liao, W., Maloney, D., Wang, P., Wu, Z., Wang, J., Bhatt, D.M., Heise, N., Schmid, R.M., et al. (2017). An NF-kB transcription-factor-dependent lineage-specific transcriptional program promotes regulatory T cell identity and function. Immunity 47, 450–465.e5.

Josefowicz, S.Z., Lu, L.-F., and Rudensky, A.Y. (2012). Regulatory T cells: mechanisms of differentiation and function. Annu. Rev. Immunol. 30, 531–564.

Ohkura, N., Hamaguchi, M., Morikawa, H., Sugimura, K., Tanaka, A., Ito, Y., Osaki, M., Tanaka, Y., Yamashita, R., Nakano, N., et al. (2012). T cell receptor stimulation-induced epigenetic changes and Foxp3 expression are independent and complementary events required for Treg cell development. Immunity 37, 785–799.

Jovanovic, M., Rooney, M.S., Mertins, P., Przybylski, D., Chevrier, N., Satija, R., Rodriguez, E.H., Fields, A.P., Schwartz, S., Raychowdhury, R., et al. (2015). Immunogenetics. Dynamic profiling of the protein life cycle in response to pathogens. Science 347, 1259038.

Overacre-Delgoffe, A.E., Chikina, M., Dadey, R.E., Yano, H., Brunazzi, E.A., Shayan, G., Horne, W., Moskovitz, J.M., Kolls, J.K., Sander, C., et al. (2017). Interferon-g drives Treg fragility to promote anti-tumor immunity. Cell 169, 1130–1141.e11.

Keusekotten, K., Elliott, P.R., Glockner, L., Fiil, B.K., Damgaard, R.B., Kulathu, Y., Wauer, T., Hospenthal, M.K., Gyrd-Hansen, M., Krappmann, D., et al.

Pandiyan, P., and Zhu, J. (2015). Origin and functions of pro-inflammatory cytokine producing Foxp3+ regulatory T cells. Cytokine 76, 13–24.

1058 Immunity 48, 1046–1059, May 15, 2018

Park, Y., Jin, H.-S., Lopez, J., Lee, J., Liao, L., Elly, C., and Liu, Y.-C. (2016). SHARPIN controls regulatory T cells by negatively modulating the T cell antigen receptor complex. Nat. Immunol. 17, 286–296.

Tan, T.G., Mathis, D., and Benoist, C. (2016). Singular role for T-BET+CXCR3+ regulatory T cells in protection from autoimmune diabetes. Proc. Natl. Acad. Sci. USA 113, 14103–14108.

Pellerin, L., Jenks, J.A., Be´gin, P., Bacchetta, R., and Nadeau, K.C. (2014). Regulatory T cells and their roles in immune dysregulation and allergy. Immunol. Res. 58, 358–368.

Valencia, X., Stephens, G., Goldbach-Mansky, R., Wilson, M., Shevach, E.M., and Lipsky, P.E. (2006). TNF downmodulates the function of human CD4+CD25hi T-regulatory cells. Blood 108, 253–261.

Perez-Llamas, C., and Lopez-Bigas, N. (2011). Gitools: analysis and visualisation of genomic data using interactive heat-maps. PLoS ONE 6, e19541.

van Loosdregt, J., Fleskens, V., Fu, J., Brenkman, A.B., Bekker, C.P.J., Pals, C.E.G.M., Meerding, J., Berkers, C.R., Barbi, J., Gro¨ne, A., et al. (2013). Stabilization of the transcription factor Foxp3 by the deubiquitinase USP7 increases Treg-cell-suppressive capacity. Immunity 39, 259–271.

Polansky, J.K., Kretschmer, K., Freyer, J., Floess, S., Garbe, A., Baron, U., Olek, S., Hamann, A., von Boehmer, H., and Huehn, J. (2008). DNA methylation controls Foxp3 gene expression. Eur. J. Immunol. 38, 1654–1663. Procaccini, C., Carbone, F., Di Silvestre, D., Brambilla, F., De Rosa, V., Galgani, M., Faicchia, D., Marone, G., Tramontano, D., Corona, M., et al. (2016). The proteomic landscape of human ex vivo regulatory and conventional T cells reveals specific metabolic requirements. Immunity 44, 406–421. Rappsilber, J., Ishihama, Y., and Mann, M. (2003). Stop and go extraction tips for matrix-assisted laser desorption/ionization, nanoelectrospray, and LC/MS sample pretreatment in proteomics. Anal. Chem. 75, 663–670. Rieckmann, J.C., Geiger, R., Hornburg, D., Wolf, T., Kveler, K., Jarrossay, D., Sallusto, F., Shen-Orr, S.S., Lanzavecchia, A., Mann, M., and Meissner, F. (2017). Social network architecture of human immune cells unveiled by quantitative proteomics. Nat. Immunol. 18, 583–593. Rogge, L., D’Ambrosio, D., Biffi, M., Penna, G., Minetti, L.J., Presky, D.H., Adorini, L., and Sinigaglia, F. (1998). The role of Stat4 in species-specific regulation of Th cell development by type I IFNs. J. Immunol. 161, 6567–6574. Rubtsov, Y.P., Niec, R.E., Josefowicz, S., Li, L., Darce, J., Mathis, D., Benoist, C., and Rudensky, A.Y. (2010). Stability of the regulatory T cell lineage in vivo. Science 329, 1667–1671. Rudra, D., deRoos, P., Chaudhry, A., Niec, R.E., Arvey, A., Samstein, R.M., Leslie, C., Shaffer, S.A., Goodlett, D.R., and Rudensky, A.Y. (2012). Transcription factor Foxp3 and its protein partners form a complex regulatory network. Nat. Immunol. 13, 1010–1019. Saito, T., Nishikawa, H., Wada, H., Nagano, Y., Sugiyama, D., Atarashi, K., Maeda, Y., Hamaguchi, M., Ohkura, N., Sato, E., et al. (2016). Two FOXP3(+) CD4(+) T cell subpopulations distinctly control the prognosis of colorectal cancers. Nat. Med. 22, 679–684. Sakaguchi, S., Vignali, D.A.A., Rudensky, A.Y., Niec, R.E., and Waldmann, H. (2013). The plasticity and stability of regulatory T cells. Nat. Rev. Immunol. 13, 461–467. Schmidl, C., Hansmann, L., Lassmann, T., Balwierz, P.J., Kawaji, H., Itoh, M., Kawai, J., Nagao-Sato, S., Suzuki, H., Andreesen, R., et al.; FANTOM consortium (2014). The enhancer and promoter landscape of human regulatory and conventional T-cell subpopulations. Blood 123, e68–e78. €usser, B., Busse, D., Li, N., Dittmar, G., Schuchhardt, J., Wolf, J., Schwanha Chen, W., and Selbach, M. (2011). Global quantification of mammalian gene expression control. Nature 473, 337–342. Siegmund, K., Feuerer, M., Siewert, C., Ghani, S., Haubold, U., Dankof, A., Krenn, V., Scho¨n, M.P., Scheffold, A., Lowe, J.B., et al. (2005). Migration matters: regulatory T-cell compartmentalization determines suppressive activity in vivo. Blood 106, 3097–3104. Smigiel, K.S., Richards, E., Srivastava, S., Thomas, K.R., Dudda, J.C., Klonowski, K.D., and Campbell, D.J. (2014). CCR7 provides localized access to IL-2 and defines homeostatically distinct regulatory T cell subsets. J. Exp. Med. 211, 121–136.   Supek, F., Bo snjak, M., Skunca, N., and Smuc, T. (2011). REVIGO summarizes and visualizes long lists of gene ontology terms. PLoS ONE 6, e21800.

Vizcaı´no, J.A., Csordas, A., del-Toro, N., Dianes, J.A., Griss, J., Lavidas, I., Mayer, G., Perez-Riverol, Y., Reisinger, F., Ternent, T., et al. (2016). 2016 update of the PRIDE database and its related tools. Nucleic Acids Res. 44 (D1), D447–D456. Vogel, C., and Marcotte, E.M. (2012). Insights into the regulation of protein abundance from proteomic and transcriptomic analyses. Nat. Rev. Genet. 13, 227–232. Wan, Y.Y., and Flavell, R.A. (2007). Regulatory T-cell functions are subverted and converted owing to attenuated Foxp3 expression. Nature 445, 766–770. Wilhelm, M., Schlegl, J., Hahne, H., Gholami, A.M., Lieberenz, M., Savitski, M.M., Ziegler, E., Butzmann, L., Gessulat, S., Marx, H., et al. (2014). Massspectrometry-based draft of the human proteome. Nature 509, 582–587. Wisniewski, J.R., Zougman, A., Nagaraj, N., and Mann, M. (2009). Universal sample preparation method for proteome analysis. Nat. Methods 6, 359–362. Wisniewski, J.R., Hein, M.Y., Cox, J., and Mann, M. (2014). A ‘‘proteomic ruler’’ for protein copy number and concentration estimation without spike-in standards. Mol. Cell. Proteomics 13, 3497–3506. Wu, D., Luo, Y., Guo, W., Niu, Q., Xue, T., Yang, F., Sun, X., Chen, S., Liu, Y., Liu, J., et al. (2017). Lkb1 maintains Treg cell lineage identity. Nat. Commun. 8, 15876. Xiao, X., Shi, X., Fan, Y., Zhang, X., Wu, M., Lan, P., Minze, L., Fu, Y.-X., Ghobrial, R.M., Liu, W., and Li, X.C. (2015). GITR subverts Foxp3(+) Tregs to boost Th9 immunity through regulation of histone acetylation. Nat. Commun. 6, 8266. Xu, J., Yang, Y., Qiu, G., Lal, G., Yin, N., Wu, Z., Bromberg, J.S., and Ding, Y. (2011). Stat4 is critical for the balance between Th17 cells and regulatory T cells in colitis. J. Immunol. 186, 6597–6606. Yan, D., Farache, J., Mingueneau, M., Mathis, D., and Benoist, C. (2015). Imbalanced signal transduction in regulatory T cells expressing the transcription factor FoxP3. Proc. Natl. Acad. Sci. USA 112, 14942–14947. Yu, F., Sharma, S., Edwards, J., Feigenbaum, L., and Zhu, J. (2015). Dynamic expression of transcription factors T-bet and GATA-3 by regulatory T cells maintains immunotolerance. Nat. Immunol. 16, 197–206. Zanin-Zhorov, A., Ding, Y., Kumari, S., Attur, M., Hippen, K.L., Brown, M., Blazar, B.R., Abramson, S.B., Lafaille, J.J., and Dustin, M.L. (2010). Protein kinase C-theta mediates negative feedback on regulatory T cell function. Science 328, 372–376. Zaragoza, B., Chen, X., Oppenheim, J.J., Baeyens, A., Gregoire, S., Chader, D., Gorochov, G., Miyara, M., and Salomon, B.L. (2016). Suppressive activity of human regulatory T cells is maintained in the presence of TNF. Nat. Med. 22, 16–17. Zhuo, C., Li, Z., Xu, Y., Wang, Y., Li, Q., Peng, J., Zheng, H., Wu, P., Li, B., and Cai, S. (2014). Higher FOXP3-TSDR demethylation rates in adjacent normal tissues in patients with colon cancer were associated with worse survival. Mol. Cancer 13, 153.

Immunity 48, 1046–1059, May 15, 2018 1059

STAR+METHODS KEY RESOURCES TABLE

REAGENT or RESOURCE

SOURCE

IDENTIFIER

PE Anti-human CD25 (Clone 2A3)

BD Biosciences

Cat#341011

PE-Cy7 Anti-human CD45RA (Clone HI100)

BD Biosciences

Cat#560675 (RRID: AB_1727498)

APC Anti-CD49d (Clone 9F10)

BD Biosciences

Cat#559881

PE-Cy7 Anti-human CD194/CCR4 (Clone 1G1)

BD Biosciences

Cat#557864 (RRID: AB_396907)

Brilliant Violet 421 Anti-human CD127 (Clone A019D5)

Biolegend

Cat#351309 (RRID: AB_10898326)

Alexa Fluor 488 anti-human NF-kB p50 (Clone 4D1)

Biolegend

Cat#616704 (RRID: AB_493279)

Antibodies

PE Anti-human IFN-g (Clone 4S.B3)

Biolegend

Cat#502510 (RRID: AB_315235)

Brilliant Violet 605 anti-Human IL-17A (Clone BL168)

Biolegend

Cat#512325 (RRID: AB_11218595)

Brilliant Violet 421 anti-Human CXCR3 (Clone G025H7)

Biolegend

Cat#353716 (RRID: AB_2561448)

Alexa Fluor 488 anti-Human IL-2 (Clone MQ1-17H12)

Biolegend

Cat#500314 (RRID: AB_493368)

APC anti-Human T-bet (Clone 4B10)

Biolegend

Cat#644813 (RRID: AB_10896913)

PE anti-human CTLA-4(Clone L3D10)

Biolegend

Cat#349906 (RRID: AB_10641842)

Alexa Fluor 488 anti-Human NFATc1 (Clone 7A6)

Biolegend

Cat#649603 (RRID: AB_2561822)

PE-Cy7 anti-Human FOXP3 (Clone 236A/E7)

eBioscience

Cat#25-4777-42 (RRID: AB_2573450)

eFluor 450 anti-Human HELIOS (Clone 22F6)

eBioscience

Cat#48-9883 (RRID: AB_2574136)

PE anti-Human Phospho-STAT4 (Tyr693) (Clone 4LURPIE)

eBioscience

Cat#12-9044-42 (RRID: AB_2572689)

eFluor 450 anti-Human Phospho-STAT1 (Tyr701) (Clone KIKSI0803)

eBioscience

Cat#48-9008-41 (RRID: AB_2574118)

PE-Cy7 anti-Human Granzyme A (Clone CB9)

eBioscience

Cat#25-9177-41 (RRID: AB_2573537)

PE anti-Human Gata-3 (Clone TWAJ)

eBioscience

Cat#12-9966-42 (RRID: AB_1963600)

APC anti-Human ROR gamma (t) (Clone AFKJS-9)

eBioscience

Cat#17-6988-80 (RRID: AB_1633425)

Rabbit Polyclonal Anti-Human SHMT2

Sigma-Aldrich

Cat#HPA020549 (RRID:AB_1856834)

Mouse Monoclonal Anti-human Actin (Clone AC-40)

Sigma-Aldrich

Cat#A3853 (RRID:AB_262137)

FITC AntiHuman -NFkB p52

Sigma-Aldrich

Cat#FCMAB346F (RRID:AB_11205225)

PE anti-Human Granzyme K (Clone 24C3)

Immunotools

Cat#21144054

Rabbit Polyclonal Anti-Human Ferritin Light Chain

Abcam

Cat#ab69090 (RRID:AB_1523609)

Mouse Monoclonal Anti-Human Galectin-10

R&D Systems

Cat#MAB5447 (RRID:AB_10889649)

Mouse Monoclonal anti-Human Annexin I (Clone EH17a)

Santa Cruz Biotechnology

Cat#sc-12740 (RRID:AB_2057007)

Alexa Fluor 488 anti-Human NFAT1(Clone D43B1)

Cell Signaling Technology

Cat#14324 (RRID:AB_10834808)

CD4 MicroBeads, human

Miltenyi Biotec

Cat#130-045-101

anti-CD3 mAb (Clone 1XE)

Pelicluster

Cat# M1654

anti-CD28 mAb (Clone CD28.2)

eBioscience

Cat#16-0289-85 (RRID:AB_468927)

Goat anti-Rabbit-HRP antibody

DAKO

Cat#P0448 (RRID: RRID:AB_2617138)

Goat anti-Mouse-HRO antibody

DAKO

Cat#P0447 (RRID:AB_2617137)

Invitrogen

Cat#18265017

Bacterial and Virus Strains Subcloning Efficiency DH5a Competent Cells Chemicals, Peptides, and Recombinant Proteins Sequencing Grade Modified Trypsin

Promega

Cat# V5111

TRIzol Reagent

ThermoFisher Scientific

Cat#15596018

SuperScript II Reverse Transcriptase

Invitrogen

Cat#18064-014

iQ SYBR Green Supermix

Bio-Rad

Cat#1708880

IL-2 (Proleukin)

Novartis

N/A

TO-PRO-3 Iodide

ThermoFisher Scientific

Cat#T3605

DRAQ5

ThermoFisher Scientific

Cat#65-0880-92

Phorbol 12-myristate 13-acetate

Sigma Aldrich

Cat#P8139 (Continued on next page)

e1 Immunity 48, 1046–1059.e1–e6, May 15, 2018

Continued REAGENT or RESOURCE

SOURCE

IDENTIFIER

Ionomycin

Sigma Aldrich

Cat#I3909

IFN alpha

Peprotech

Cat#300-02A

IL-12

Peprotech

Cat#200-12

Brefeldin A

eBioscience

Cat#4506-51

Polyethylenimine

Polysciences

Cat#23966-1

RetroNectin

Clontech

Cat#T100B

Critical Commercial Assays RNeasy MinElute Cleanup Kit

QIAGEN

Cat# 74204

EZ Methylation-Direct kit

Zymo Research

Cat#D5020

Foxp3 Transcription Factor Staining Buffer Kit

eBioscience

Cat#A25866A

LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit

Invitrogen

Cat#L34976

Cytofix/Cytoperm Plus reagents

BD Biosciences

Cat#554714

Raw data files Mass Spectrometry analysis

ProteomeXchange

PDX007745

Raw data files Mass Spectrometry analysis

ProteomeXchange

PDX007744

Raw data files Mass Spectrometry analysis

ProteomeXchange

PXD005477

Raw data files RNASeq analysis

NCBI GEO

GSE90600

Deposited Data

Experimental Models: Cell Lines Human: primary T lymphocytes

This paper

N/A

Cell line: HEK293/T17

ATCC

Cat#CRL-11268

FOXP3-TSDR demethylation-specific F

(Zhuo et al., 2014)

TAGGGTAGTTAGTTTTTGGAATGA

FOXP3-TSDR demethylation-specific R

(Zhuo et al., 2014)

CCATTAACATCATAACAACCAAA

FOXP3-TSDR methylation-specific F

(Zhuo et al., 2014)

CGATAGGGTAGTTAGTTTTCGGAAC

FOXP3-TSDR methylation-specific R

(Zhuo et al., 2014)

CATTAACGTCATAACGACCGAA

STAT4 cDNA forward

This paper

TGGTGGTAGGGAATTCATGTCTCAGTGGAATCAAGT

STAT4 cDNA reverse + HA-tag

This paper

CGTAGCGGCCGCTCAAGCGTAATCTGGAACAT CGTATGGGTATTCAGCAGAATAAGGA

psPAX

Addgene

Cat#12260

pMD2.G

Addgene

Cat#12259

pCDH-EF1

Addgene

Cat#72266

FlowJo

Tree Star

RRID:SCR_008520

GraphPad Prism 6.0

GraphPad software

RRID:SCR_002798

MaxQuant version 1.5.2.8

(Cox and Mann, 2008)

www.coxdocs.org/doku.php?id=maxquant:start

Perseus version 1.5.0.31

(Cox and Mann, 2012)

http://www.coxdocs.org/doku.php?id=perseus:start

R environment for statistical computing v 3.2.0

N/A

https://www.r-project.org/

Xcalibur Software

ThermoFisher Scientific

Cat#OPTON-30487

Gitools version 2.2.3

(Perez-Llamas and Lopez-Bigas, 2011)

www.gitools.org/

Ingenuity Pathway Analysis

QIAGEN

https://www.qiagenbioinformatics.com

REVIGO

(Supek et al., 2011)

N/A

Molecular Signatures Database v6.1

(Matys et al., 2006)

http://software.broadinstitute.org/gsea/msigdb

Qlucore Omics Explorer (3.1)

Qlucore

https://www.qlucore.com/

IDEAS Software

Amnis

N/A

Eppendorf

Cat#0030108094

Oligonucleotides

Recombinant DNA

Software and Algorithms

Other Protein LoBind Tubes

Immunity 48, 1046–1059.e1–e6, May 15, 2018 e2

CONTACT FOR REAGENT AND RESOURCE SHARING Requests for resources and reagents and further information on protocols and experimental designs should be directed to and will be fulfilled by the Lead Contact, Derk Amsen ([email protected]). EXPERIMENTAL MODEL AND SUBJECT DETAILS Healthy blood donors Blood samples were obtained from anonymized healthy male donors with written informed consent in accordance to guidelines established by the Sanquin Medical Ethical Committee. Cell Lines HEK293T cells were cultured in DMEM with HEPES (Life Technologies) supplemented with 10% FCS and 1% penicillin/streptomycin and maintained at 37 C degrees Celsius with 5% CO2. METHOD DETAILS Cell isolation and cell sorting Human PBMCs were isolated from fresh buffy coats from using Ficoll-Paque Plus (GE Healthcare) gradient centrifugation. Total CD4+ T cells were isolated using magnetic sorting with CD4 microbeads (Miltenyi Biotec) and viable cells were separated using flow cytometric sorting based on the expression of CD25, CD45RA, and CD127 on a FACS Aria III (BD Biosciences). Blood samples were obtained from anonymized healthy male donors with written informed consent in accordance to guidelines established by the Sanquin Medical Ethical Committee. Antibodies The following antibodies were used: CD25 (341011), CD45RA (560675), CD49d (559881), and CCR4 (557864) from BD Biosciences. CD127 (351309), NFkB p50 (616704), IFNg (502510), IL-17A (512325), CXCR3 (353716), IL-2 (500314), T-bet (644813), CTLA-4 (349906), and NFATc1 (649603) from Biolegend. FOXP3 (25-4777-42), Helios (48-9883), phospho-STAT4 Y693 (12-9044-42), phospho-STAT1 Y701 (48-9008-41), GZMA (25-9177-41), GATA3 (12-9966-42), and RORgt (17-6988-80) from eBioscience. SHMT2 (HPA020549), Actin (A3853), and NFkB p52 (FCMAB346F) from Sigma. GZMK (21144054, Immunotools), FTL (ab69090, Abcam), Galectin-10 (MAB5447, R&D Systems), ANXA1 (sc-12740, Santa Cruz), and NFATc2 (14324, Cell Signaling). Sample preparation for mass spectrometry (MS) FACS-purified CD4+ subsets (at least 1x106 cells per subset) from 3-6 donors were placed in Protein LoBind tubes (Eppendorf), washed in PBS and lysed in 100 mM Tris HCl pH 8.0 with 4% SDS, 100 mM DTT. Samples were heated at 95 C, sonicated (Bioruptor) and after centrifugation (16,000 g), the supernatant cell lysates were isolated and stored at
80 C. Cell lysates were processed using filter aided sample preparation (FASP) as described (Wi sniewski et al., 2009). Briefly, proteins were alkylated and digested into peptides using sequencing-grade Trypsin (Promega). Samples were desalted using StageTips (Rappsilber et al., 2003), the aqueous phase was evaporated in a speedvac, and proteins were reconstituted in 2% acetonitrile in 0.1% TFA in water before analysis by MS. Mass spectrometry data acquisition Tryptic peptides were separated by nanoscale C18 reverse chromatography coupled on line to an Orbitrap Fusion Tribrid mass spectrometer (Thermo Scientific) via a nanoelectrospray ion source (Nanospray Flex Ion Source, Thermo Scientific). Peptides were loaded on a 20 cm 75–360 mm inner-outer diameter fused silica emitter (New Objective) packed in-house with ReproSil-Pur C18-AQ, 1.9 mm resin (Dr Maisch GmbH). The column was installed on a Dionex Ultimate3000 RSLC nanoSystem (Thermo Scientific) using a MicroTee union formatted for 360 mm outer diameter columns (IDEX) and a liquid junction. The spray voltage was set to 2.15 kV. Buffer A was composed of 0.5% acetic acid and buffer B of 0.5% acetic acid, 80% acetonitrile. Peptides were loaded for 17 min at 300 nl/min at 5% buffer B, equilibrated for 5 minutes at 5% buffer B (17-22 min) and eluted by increasing buffer B from 5%–15% (22-87 min) and 15%–38% (87-147 min), followed by a 10 minute wash to 90% and a 5 min regeneration to 5%. Survey scans of peptide precursors from 400 to 1500 m/z were performed at 120K resolution (at 200 m/z) with a 1.5 3 105 ion count target. Tandem mass spectrometry was performed by isolation with the quadrupole with isolation window 1.6, HCD fragmentation with normalized collision energy of 30, and rapid scan mass spectrometry analysis in the ion trap. The MS2 ion count target was set to 104 and the max injection time was 35 ms. Only those precursors with charge state 2–7 were sampled for MS2. The dynamic exclusion duration was set to 60 s with a 10 ppm tolerance around the selected precursor and its isotopes. Monoisotopic precursor selection was turned on. The instrument was run in top speed mode with 3 s cycles. All data were acquired with Xcalibur software. Data Processing and Analysis RAW mass spectrometry files were processed with the MaxQuant computational platform (Cox and Mann, 2008) version 1.5.2.8 using label-free quantitation (LFQ). Peptides were identified using the Andromeda search engine by querying the human Uniprot e3 Immunity 48, 1046–1059.e1–e6, May 15, 2018

database (2015-02, 89796 entries) with a 1% false discovery rate (FDR) cut-off both at peptide and protein level. Potential contaminants and reverse hits were eliminated using Perseus version 1.5.0.31 (Cox and Mann, 2012). To compare proteomes of different cell populations, LFQ values were log2-transformed and when appropriate the 3 technical replicates per experimental condition grouped. For further analysis, only proteins with 3 valid values in at least one of the groups were included. Missing values were imputed by normal distribution (width = 0.3, shift = 1.8). Intergroup differences in protein levels were calculated using either Kruskal-Wallis or ANOVA test followed by a Benjamini-Hochberg multiple testing correction with a 5% FDR. General relatedness of samples was assessed through Principal Component Analysis (PCA) using Perseus’ built-in tool. Calculation of Pearson’s correlations was assessed using Gitools (version 2.2.3)(Perez-Llamas and Lopez-Bigas, 2011). Clusters of proteins with specific expression patterns were identified using unsupervised hierarchical clustering based on K-means and Euclidian distance after z-score normalization. Cell type specific clusters were further analyzed for enrichment of functional and biological categories, using STRING and DAVID bioinformatics tools. Pathway diagrams were generated through the use of Ingenuity Pathway Analysis (IPA, QIAGEN). GO biological processes redundancy was removed using REVIGO (Supek et al., 2011). Absolute protein abundance was estimated using the proteomic ruler approach using a plug-in built in Perseus as described by the authors (Wisniewski et al., 2014). The abundances were expressed as protein copy numbers and the values were log10 transformed and presented in graphs using Prism 6.0 (GraphPad Software, San Diego, CA, USA). Gene set enrichment analysis was performed using the geneset collection C3 from the Molecular Signatures Database (MsigDB v4.0) that contains information about curated, motif, transcription factor target gene sets. The genes that share a transcription factor binding site are defined and annotated by TRANSFAC (Matys et al., 2006) (version 7.4). RNA Isolation Total RNA was extracted from 1 3 106 cells using TRIzol reagent (15596-018, Ambion Life Technologies) according to the manufacturer’s protocol. The total RNA pellet was air-dried for 8 minutes and dissolved in an appropriate volume of nuclease free water (AM9937, Ambion Life Technologies) followed by a total RNA quantification using the Nanodrop UV-VIS Spectrophotometer (Thermo Scientific). The total RNA was further purified using the MinElute Cleanup Kit (74204, QIAGEN) according to the manufacturer’s instructions. Quality and quantity of the total RNA was assessed by the 2100 Bioanalyzer using a Nano chip (Agilent, Santa Clara, CA). Total RNA samples having RNA Integrity Number (RIN)>8 were subjected to library generation. TruSeq stranded mRNA sample preparation Strand-specific libraries were generated using the TruSeq Stranded mRNA sample preparation kit (ref. RS-122-2101/2, Illumina) according to the manufacturer’s instructions. Briefly, polyadenylated RNA from intact total RNA was purified using oligo-dT beads. Following purification the RNA was fragmented, random primed and reverse transcribed using SuperScript II Reverse Transcriptase (18064-014, Invitrogen) with the addition of Actinomycin D. Second strand synthesis was performed using Polymerase I and RNaseH with replacement of dTTP for dUTP. The generated cDNA fragments were 30 end adenylated and ligated to Illumina Paired-end sequencing adapters and subsequently amplified by 12 cycles of PCR. The libraries were analyzed on a 2100 Bioanalyzer using a 7500 chip (Agilent), diluted and pooled equimolar into a 15-plex, 10 nM sequencing pool and stored at
20 C. RNA-Seq data processing RNA-Seq raw reads Fastq were aligned to the Ensembl reference genome (Homo_sapiens.GRCh38.dna.primary_assembly) with TopHat (version 2.0.12, Bowtie version 1.0.0, Samtools version: 0.1.19). Read counts were generated by HTseq-count with uniquely mapped reads. Unmapped reads were discarded. Sequence reads were normalized to 10 million reads per sample and log2 transformed with the formula, log2(((expression gene 3 O library size)106)+1), where the library size was the sum of all expression values per sample. Read-counts were further analyzed by Qlucore Omics Explorer (3.1) for differential expression. Protein and RNA comparison The comparison of protein and transcript abundances across cell types was performed as published before (Wilhelm et al., 2014). Briefly, absolute transcript abundances (expressed as fragment per kilobase per million, FPKM) were extracted from the RNASeq data. Absolute protein abundance was estimated using the intensity based absolute protein quantification (iBAQ) €usser et al., 2011) approach with few modifications described elsewhere (Wilhelm et al., 2014). Briefly, peptide intensities (Schwanha were summed up and divided by the number of observable peptides per protein. The iBAQ values were then normalized based on the total sum of all protein intensities in order to be able to compare abundances across samples. Proteins were mapped to transcripts using the Uniprot.ws package (Bioconductor) resulting in 4,792 transcripts/proteins with expression data. For the comparison of protein and mRNA abundances, protein and transcript expression of 409 differentially expressed proteins/RNAs was z-score-transformed. All the analyses were performed using R (v 3.2.0). In vitro cell expansion Naive Tconvs (CD4+, CD127+, CD25-, CD45RA+) and naive Tregs (CD4+, CD127-, CD25+, CD45RA+) were isolated by FACS sorting (FACS Aria III, BD Biosciences). The cells were then cultured in presence of 0.1 mg/ml of anti-CD3 mAb (M1654, clone 1XE, PeliCluster) and anti-CD28 mAb (16-0289-85, clone CD28.2, eBioscience) for 14 days in IMDM containing 10% FCS and 300 U/ml IL-2. The expanded cells were directly used for cytokine stimulation assays. For proteomics, the expanded cells were Immunity 48, 1046–1059.e1–e6, May 15, 2018 e4

rested for 4 days in medium with 300 U/ml IL-2 prior to restimulation with 0.1 mg/ml anti-CD3- and anti-CD28 mAb. After 24 h, unstimulated and stimulated cells were washed, lysed, and processed for MS. Methylation status of the Treg specific demethylated region (TSDR) in the FOXP3 gene Sorted and expanded cells were collected in PBS. Bisulfite treatment of genomic DNA was performed using the EZ MethylationDirect kit according to manufacturer’s protocol (Zymo Research). Methylation-specific quantitative PCR was performed in a 10 ml reaction containing 25 ng bisulfite-converted DNA, iQ SYBR Green Supermix (Bio-Rad) and 0.5 mM forward and reverse primers identifying FOXP3 gene methylation or demethylation status, as previously described (Zhuo et al., 2014). The methylation status was calculated using the formula: 100/(1+2[Ct(CG)-Ct(TG)]), where Ct(CG) is the Ct-value using the methylation-specific primers and Ct(TG) is the Ct value using the demethylation-specific primers. Western blot Cells were lysed in RIPA buffer and cell lysates were boiled in sample buffer prior to gel electrophoresis. SDS-PAGE gel electrophoresis was performed using the NuPAGE electrophoresis system (Novex, Life Technologies). Proteins were transferred using the iBlot system (Thermo Scientific) and analyzed using the corresponding antibodies. ECL signals on western blots were developed using the Pierce ECL substrate kit (Pierce) followed by autoradiographic detection on film (Fuji Medical). Flow cytometry Cells were labeled with fluorochrome-conjugated antibodies in PBS, 0.5% BSA for 30 min at 4 C. For intracellular and nuclear staining, cells were fixed and permeabilized using the FOXP3/Transcription Factor Fixation/Permeabilization buffers (eBioscience) according to the manufacturer’s instructions. Near-IR (L10119, ThermoFisher) or TOPRO-3 (T3605, Invitrogen) dyes (ThermoFisher) were used to stain dead cells. Cells were analyzed on LSR Fortessa or LSR II cytometers (BD Biosciences). Nuclear translocation analysis Cells were stimulated with anti-CD3- and anti-CD28 mAbs, fixed in 1.6% paraformaldehyde, permeabilized in ice-cold methanol (Sigma), and stained for NF-kB1, NFATc2, and with the nuclear dye DRAQ5 (65-0880-92, ThermoFisher Scientific). Images were recorded (10,000 events per condition) on Imagestream (Amnis) and analyzed using the nuclear localization analysis template included in the IDEAS software (Amnis). Cell stimulation assays For analysis of cytokine production, cells were stimulated for 6 h with 0.1 mg/ml anti-CD3- and anti-CD28 mAbs, or 3h with phorbol 12-myristate 13-acetate (PMA; 10 ng/mL; P8139, Sigma) and ionomycin (IO; 1 mg/mL; I3909, Sigma), without or with IFNa (20 ng/ml; 300-02A, Peprotech) or IL-12 (40 ng/ml; 200-12, Peprotech), in presence of Brefeldin A (3 mg/ml; 00-4506-51, eBioscience). Next, cells were fixed and permeabilized using Cytofix/Cytoperm Plus reagents (BD) according to manufacturer’s instructions. To assay STAT phosphorylation, cells were stimulated with 20 ng/ml IFNa or 40 ng/mL IL-12 (both from Peprotech), fixed in 1.6% paraformaldehyde, permeabilized in ice-cold methanol (Sigma), and stained for phospho-STAT4 (Y693) and phospho-STAT1 (Y701). Cells then were analyzed on a LSR Fortessa or LSR II cytometer (BD Biosciences). Cloning, lentivirus production and gene transduction The IRES-mCherry fragment of the pMSCV-IRES-mCherry vector was cloned into the lentiviral vector pCDH-EF1 linearized with EcoRI and SalI (pCDH-EF1 was a gift from Kazuhiro Oka - Addgene plasmid # 72266). The cDNA encoding STAT4 was PCR amplified from pLX304-STAT4-V5-mCherry using primers containing a 50 EcoRI and a 30 NotI restriction site. A C-terminal HA-tag was added to the reverse primer (fwd: 50 - TGGTGGTAGGGAATTCATGTCTCAGTGGAATCAAGT-30 and rev: 50 - CGTAGCGGCCGCTCAAGCG TAATCTGGAACATCGTATGGGTATTCAGCAGAATAAGGAG-30 ). The STAT4-HA cDNA was cloned into the EcoRI and SnaBI site of the pCDH-EF1-mCherry construct. All generated inserts were sequence verified. Lentivirus were produced as described previously by transfecting confluent HEK293T cells with packaging (psPAX2) and envelope plasmids (pMD2.G) with pCDH-EF1-mCherry or pCDH-EF1-STAT4-mCherry. HEK293T cells were cultured in DMEM with HEPES (Life Technologies) supplemented with 10% FCS and 1% penicillin/streptomycin. Polyethylenimine (Polysciences, Hirschberg an der Bergstrasse, Germany) was used as a transfection reagent. After 24 h, the cultures were refreshed with medium with 2% FCS and 24 h later, lentiviral particles were concentrated and purified by ultracentrifugation at 50,000 g for 2.5 h at 8 C. Freshly sorted nTconvs and nTregs were expanded for 5 days as described above and restimulated one day prior to transduction. Cells were then transduced in RetroNectin (Clontech) coated-plates for 24 h. After that, the cells were replated into tissue culture-treated plates with medium containing 100 U/ml IL-2 with or without IL-12 or IFNa. After 5 days the cells were stimulated with PMA/IO for 3 h. QUANTIFICATION AND STATISTICAL ANALYSIS Statistical tests, excluding those evaluating proteomic and RNaseq data, were performed using Prism 6.0 (GraphPad Software). All experiments were performed at least three independent times. The variable distribution was assessed by the Kolmogorov–Smirnov test. Nonparametric Kruskal–Wallis test was used followed by a Dunn’s multiple comparison test to assess intergroup differences. e5 Immunity 48, 1046–1059.e1–e6, May 15, 2018

Two factor group differences were assessed using Row matched or normal Two-way ANOVA followed by Tukey’s or Bonferroni’s multiple comparison test. A p < 0.05 was considered statistically significant at a 95% confidence level. Data are presented as mean ± SEM unless otherwise indicated in figure legends. Significance levels are: * p < 0.05; ** p < 0.01; *** p < 0.001. DATA AND SOFTWARE AVAILABILITY The mass spectrometry data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange. org) via the PRIDE (Vizcaı´no et al., 2016) repository with the dataset identifier PDX007745, PDX007744, and PXD005477. RNASeq data can be accessed with accession number GSE90600.

Immunity 48, 1046–1059.e1–e6, May 15, 2018 e6