Articles https://doi.org/10.1038/s41590-019-0472-4

Follicular regulatory T cells control humoral and allergic immunity by restraining early B cell responses

Rachel L. Clement 1, Joe Daccache1, Mostafa T. Mohammed1, Alos Diallo2, Bruce R. Blazar3, Vijay K. Kuchroo4,5,6, Scott B. Lovitch 7, Arlene H. Sharpe 2,4,7 and Peter T. Sage 1*

Follicular regulatory T (TFR) cells have specialized roles in modulating follicular helper T (TFH) cell activation of B cells. However, the precise role of TFR cells in controlling antibody responses to foreign antigens and autoantigens in vivo is still unclear due to a lack of specific tools. A TFR cell-deleter mouse was developed that selectively deletes TFR cells, facilitating temporal studies. TFR cells were found to regulate early, but not late, germinal center (GC) responses to control antigen-specific antibody and B cell memory. Deletion of TFR cells also resulted in increased self-reactive immunoglobulin (Ig) G and IgE. The increased IgE levels led us to interrogate the role of TFR cells in house dust mite models. TFR cells were found to control TFH13 cell-induced IgE. In vivo, loss of TFR cells increased house-dust-mite-specific IgE and lung inflammation. Thus, TFR cells control IgG and IgE responses to vaccines, allergens and autoantigens, and exert critical immunoregulatory functions before GC formation.

FH cells migrate to B cell follicles to stimulate antibody produc- marrow chimera and/or genetic models, in which the transcription 1 + tion by B cells in the GC reaction . The GC reaction results in factor Bcl6 was deleted in FoxP3 cells, have suggested that TFR Tsomatic hypermutation, affinity maturation and class-switch cells regulate non-antigen-specific B cell responses but do not sub- recombination, although these processes may also occur outside stantially affect GC B cells or antigen-specific IgG levels; however, 2 20–22 GCs . TFH cells provide essential costimulation (through the induc- results have been inconsistent . Moreover, IL-10 produced by TFR ible costimulatory molecule ICOS and CD40L) and cytokines (such cells can promote, rather than inhibit, plasma cell formation23. One as interleukin (IL)-21 and IL-4) to help promote B cell responses3,4. explanation for the variability between studies may be due to the

TFH cells possess a degree of phenotypic plasticity that can be altered models used because Bcl6 can be expressed on Treg cell subsets other by the inflammatory milieu, causing TFH cells to produce cytokines than TFR cells, Bcl6 might not be completely necessary for develop- 5–7 typically made by T helper (TH)1, TH2 and TH17 cells . TFH cells ment of all TFR cells, and compensatory effects may rescue TFR cell are thought to be distinct from TH2 cells because TH2 cells can deletion in non-inducible systems. produce both IL-4 and IL-13 and express the transcription factor To determine the precise role of TFR cells in controlling B cell Gata3, but TFH cells can produce only IL-4 and do not express IL-13 responses a TFR cell-deleter mouse model was developed to induc- 8 or GATA3 (ref. ). Although TH2 cells can mediate IgE responses, ibly delete TFR cells in intact hosts at specific time points during TFH cells might also play a role. Studies have suggested that the TFH immune responses. It was demonstrated that TFR cells potently regu- cell cytokine IL-21 is essential for IgE responses to house-dust mite late antigen-specific and memory IgG levels early during responses

(HDM) antigen, and that TFH cells may convert to TH2-like cells in before GC formation. Using a TH2 cell-like HDM challenge model, 9,10 the lung . IgE responses are not completely dependent on GATA3 it was found that TFR cells can regulate IL-13 production by TFH cells expression, suggesting that cells other than TH2 cells may promote and control IgE responses. Deletion of TFR cells in vivo during HDM IgE. T regulatory (Treg) cells can inhibit allergic inflammation, pos- sensitization resulted in increased HDM-specific IgE and lung 11,12 sibly through suppressing TH2 cells . inflammation. Taken together, these data demonstrate that TFR cells 13,14 TFR cells inhibit TFH cell-mediated B cell responses . In vitro are key regulators of humoral and allergic immunity by controlling assays have shown that TFR cells can inhibit antibody secretion, early GC responses. class-switch recombination and somatic hypermutation through metabolic reprogramming and epigenetic remodeling of B cells15–17. Results

In addition, TFR cells can suppress TFH cell production of effector Development of a specific and inducible TFR cell-deleter mouse cytokines such as IL-4 and IL-21 in vitro, while maintaining the model. To study the role of TFR cells during immune responses 17 TFH cell transcriptional program . The role of TFR cells in control- in vivo, we created a mouse model to perturb TFR cells in an ling TFH cell-mediated B cell responses in vivo is less clear. Adoptive inducible manner. To achieve this, a mouse containing a IRES-LoxP-STOP-LoxP-DTR transfer studies into lymphopenic mice have shown that TFR cells Cxcr5 allele knocked into the Cxcr5 locus was inhibit antigen-specific IgG levels16,18,19. However, studies using bone generated, which was crossed to a FoxP3IRES-CreYFP allele-containing

1Transplantation Research Center, Renal Division, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA. 2Department of Immunology, Harvard Medical School, Boston, MA, USA. 3Department of Pediatrics, Division of Blood and Marrow Transplantation, University of Minnesota, Minneapolis, MN, USA. 4Evergrande Center for Immunologic Diseases, Harvard Medical School and Brigham and Women’s Hospital, Boston, MA, USA. 5Broad Institute, Cambridge, MA, USA. 6Ann Romney Center for Neurologic Diseases, Harvard Medical School and Brigham and Women’s Hospital, Boston, MA, USA. 7Department of Pathology, Brigham and Women’s Hospital, Boston, MA, USA. *e-mail: [email protected]

Nature Immunology | www.nature.com/natureimmunology Articles Nature Immunology

ab DTR – TFR cells CXCR5 Treg cells TFH cells (CD4+ICOS+CXCR5+FoxP3+) (CD4+ICOS+CXCR5–FoxP3+) (CD4+ICOS+CXCR5–FoxP3–)

LoxSTOPLoxDTR WT Cxcr5 Cxcr5 hbegf Foxp3 loxP loxP DTR Cxcr5 IRES STOP hbegf Foxp3 STOP TFR–DTR Cre T –DTR Foxp3 FR CRE Foxp3 IRES Cre Foxp3 Cre

0102 103 104 105 0102 103 104 105 T FR DTR-A647 DTR-A647

cd105 11.9 104 WT 3 Foxp3 WT Foxp3 10 LoxStopLoxDTR 5 5 Cxcr5 10 10 Total T cells 102 reg 4 10 104 0 T deleter FR 103 103 0 2 3 4 5 – Cxcr5 T cells 10 10 10 10 reg 2 2 10 10 105 med 0 0 Cxcr5 TFR cells FoxP3-A488 2 CD19-APC-Cy7 4 10 Cre Cxcr5high T cells 0 2 3 4 5 0 2 3 4 5 Foxp3 FR 10 10 10 10 10 10 10 10 3 2 3 4 5 10 Cxcr5 LoxStopLoxDTR 0 10 10 10 10 CD4-PerCP-Cy5.5 CD4-PerCP-Cy5.5 DTR-A647 102 0 CXCR5-BV421

0 2 3 4 5 10 10 10 10 PD-1-PECy7

efGated on CD4+CD19– 5 10 5 5 10 10 4 61.9 5.35 P = 0.0012 10 4 4 17.7 1.0 15 10 10 P = 0.6199 8 P = 0.8344 3 100 Control 10 3 3 Control 10 10 0.8 2 6 10 80 2 2 0 P = 0.3022 10 10 10 0 0 0.6 0 2 3 4 5 60 4 0 2 3 4 5 0 2 3 4 5 (% CD4) 10 10 10 10 (% CD4 ) 0.4 treg (% CD4)

10 10 10 10 10 10 10 10 – FR 5 5 FH T

10 5 5 T 40 10 10 2 4 65.1 B cells (%) 4.96 0.2

10 T –DTR 4 4 1.96 Cxcr5 FR 10 10 T –DTR 3 20 FR 10 103 103 0.0 0 0 2 10 0 2 2 0 10 10

CD19-APC-Cy7 0 0 –DTR –DTR –DTR

CXCR5-BV421 Control

FoxP3-A488 Control FR Control FR 0 2 3 4 5 T T FR T –DTR 2 3 4 2 3 4 Control 0 5 0 5 10 10 10 10 FR T 10 10 10 10 10 10 10 10 CD4-PerCP-Cy5.5 ICOS-PE CD4-PerCP-Cy5.5

g 25 P = 0.9335 20 5 10 105 13.2 15 4 10 4 Cre P = <0.0001 10 (% gate)

Foxp3 + derived 3 10 10 103

Foxp3WT CXCR 5 102 2 5 10 derived 0 0 FoxP3-APC GFP-FIT C 0 0 2 3 4 5 0 2 3 4 5 10 10 10 10 10 10 10 10 –DTR –DTR CD4-PerCP-Cy5.5 FoxP3-APC Control Control T FR T FR WT Cre derived derived

Fig. 1 | Development of a TFR cell-specific deleter model. a, Schematic diagram of the TFR–DTR strain. Allele details (left) and schematic of events leading – to TFR cell-specific DTR expression (right) are shown. b, DTR expression on TFR cells (left), CXCR5 Treg cells (middle) or TFH cells (right) from control WT DTR (Foxp3 ), Foxp3 or TFR–DTR mice. c, DTR expression on CXCR5-negative, CXCR5-medium or CXCR5-high TFR cells from TFR–DTR mice. d, Quantification Cre LoxStopLoxDTR/WT WT LoxStopLoxDTR/WT of TFR cells from TFR–DTR (Foxp3 Cxcr5 ) or control (Foxp3 Cxcr5 ) mice that were immunized 7 d previously and received DT 2, 4 Cre WT/WT and 6 d after immunization. e, Quantification of B cells from TFR–DTR or control mice (Foxp3 Cxcr5 ) that were immunized 7 d previously and received – DT 2, 4 and 6 d after immunization. f, Quantification of TFR cells, CXCR5 Treg and TFH cells from mice as in e. g, Quantification of TFR cells (by assessing + Cre/WT LoxStopLoxDTR/WT Cre/WT WT/WT CXCR5 Treg cells) from Foxp3 Cxcr5 or control (Foxp3 Cxcr5 ) mice, gated on Cre-derived or WT-derived FoxP3 alleles. Column graphs represent the mean with error bars indicating s.e.m. The P value indicates a two-tailed Student’s t-test. Data are from individual experiments and represent two (b–d,g) or four (e,f) independent experiments with similar results. mouse to generate a Cxcr5IRES-LoxP-STOP-LoxP-DTRFoxP3IRES-CreYFP strain, allele and, hence, DTR expression under the control of Cxcr5. referred to as the TFR–DTR strain (where DTR is diphtheria toxin Therefore, only cells expressing both FoxP3 and CXCR5, such as receptor) (Fig. 1a). In TFR–DTR mice, FoxP3-expressing cells TFR cells, express DTR on the cell surface, making them susceptible produce Cre recombinase which excises the stop cassette in the to deletion with diphtheria toxin (DT). DTR expression was evalu- IRES-LoxP-STOP-LoxP-DTR IRES-DTR – Cxcr5 allele, resulting in an active Cxcr5 ated on TFR cells and CXCR5 Treg cells from wild-type (WT) FoxP3,

Nature Immunology | www.nature.com/natureimmunology Nature Immunology Articles

24 FoxP3−DTR (ref. ) or TFR–DTR mice using an anti-DTR antibody. to a compensatory mechanism to overcome TFR cell deficiency (see + + + + It was found that TFR cells (gated as CD4 ICOS CXCR5 FoxP3 Supplementary Fig. 2b). cells) had substantial DTR expression in TFR–DTR mice, albeit Next, it was determined whether TFR cells can regulate B cell slightly lower compared with FoxP3–DTR mice (Fig. 1b and see responses after GCs have already formed. TFR–DTR or control – cre wt/wt Supplementary Fig. 1a,b). Importantly, CXCR5 Treg cells expressed (Foxp3 Cxcr5 ) mice were immunized with NP-OVA and given DTR only in FoxP3–DTR mice and not in TFR–DTR mice. TFH cells DT on days 10–14 to delete TFR cells. A reduction in TFR cells was + + + – (gated as CD4 ICOS CXCR5 FoxP3 cells) did not express DTR, found in the TFR–DTR mice, and slight increases in the frequency except for a very small population which are probably ex-TFR cells of TFH cells, polarizing the follicular T cell subset toward TFH cells (Fig. 1b and see Supplementary Fig. 1a,b)25. DTR expression was (Fig. 2f). However, when GC B cells were assessed, no differences highest on TFR cells expressing high levels of the CXC chemokine were found between TFR–DTR and control mice (Fig. 2g). Likewise, receptor type 5 (CXCR5), further demonstrating CXCR5-driven no increases in total or NP-specific IgG were found, although DTR expression (Fig. 1c). plasma cells were slightly elevated, as were levels of IgA and IgE

To determine the efficiency of TFR cell deletion, mice were immu- (Fig. 2g,h and see Supplementary Fig. 2c). The few TFR cells remain- nized with (4‐hydroxy‐3‐nitrophenyl)acetyl-ovalbumin (NP-OVA) ing in TFR-DTR mice had a similar phenotype to control mice except and given DT. It was found that TFR cells were deleted from TFR–DTR but for elevated Ki67 (see Supplementary Fig. 2d). The lack of increased IRES-LoxP-STOP-LoxP-DTR wt Cre wt/wt not from either Cxcr5 FoxP3 or FoxP3 Cxcr5 antigen-specific antibodies when TFR cells were deleted post-GC control mice (Fig. 1d–f). Deletion of TFR cells was selective in TFR– formation was due to the stage of GC, and not the total duration of DTR mice because B cells, TFH cells, total CXCR5-Treg cells or activated TFR cell deletion, because pre-GC deletion strategies have a pheno- + Ki67 Treg cells were not deleted in TFR–DTR mice (Fig. 1e,f and see type as early as day 5 after TFR cell deletion, and deletion of TFR cells Supplementary Fig. 1c). Moreover, deletion of TFR cells in TFR–DTR after GC formation does not result in a phenotype, even at day 26 + mice resulted in the loss of FoxP3 cells within individual GCs (see (see Supplementary Fig. 2e,f). These data demonstrate that TFR cells Supplementary Fig. 1d). To determine whether deletion of TFR cells can regulate GC B cell development and antigen-specific antibody was cell intrinsic, Cxcr5IRES-LoxP-STOP-LoxP-DTRFoxP3Cre/wt mice (in which responses early, before GC formation, and have less regulatory con- only ~50% of the TFR cells will express DTR on the surface) were trol after GCs have been initiated. immunized and given DT. DTR-expressing TFR cells were deleted, but not non-DTR-expressing TFR cells from the same mouse, demonstrat- TFR cells regulate autoreactive IgG and IgE antibodies. Next, we ing cell-intrinsic deletion in the TFR–DTR model (Fig. 1g). assessed whether TFR cells can regulate autoreactive antibodies. TFR cells were deleted in TFR–DTR mice before GC formation (at TFR cells potently regulate early GC formation. Previous data sug- days 5–9), and the sera analyzed at day 21 with autoantigen protein 15 gest that TFR cells can be limited by cytokines produced in GCs . arrays. Autoreactive IgG was increased for a third of the autoanti- Moreover, TFR cells seem to be less frequent in large, developed GCs gens in TFR–DTR mice compared with control mice using a strin- 26 (data not shown) . These findings suggest that TFR cells might reg- gent cutoff (Mann–Whitney U-test, P < 0.01) (Fig. 3a). In most ulate B cell responses most potently before mature GCs form. To cases, control mice had antibodies that recognize autoantigens, but assess the role of TFR cells before GC initiation, TFR–DTR or control levels were higher in the TFR–DTR mice. However, in one example, cre wt/wt (Foxp3 Cxcr5 ) mice were immunized with NP-OVA and TFR there was a substantial signal for anti-histone H1 autoantibodies in cells were deleted on days 5–9 with administration of DT, and B cell TFR–DTR mice, but no detectable signal in control mice (Fig. 3a). responses at day 21 were assessed. TFR cells were largely attenuated, These data demonstrate that TFR cells can regulate levels and forma- even 12 d after the last DT injection (Fig. 2a). There were minor, but tion of autoreactive IgG antibodies. notable, increases in the frequency of TFH cells compared with con- Next, it was determined whether any of the substantial amounts + + + trol mice. The CD19 GL7 FAS GC B cell frequency was approxi- of IgE in the TFR cell-deleted mice were specific for autoantigens. mately twofold higher in TFR–DTR mice compared with control Such autoreactive antibodies could be pathogenic, because patients mice, demonstrating that TFR cells potently regulate initial GC for- with systemic lupus erythematosus (SLE) can generate autoreac- mation (Fig. 2b). In addition, naive B cells, gated as CD38+IgG1– B tive IgE responses, and autoreactive IgE can exacerbate disease in + 27,28 cells, were slightly attenuated in TFR–DTR mice. CD138 plasma mouse models of lupus . Fifteen autoantigens were recognized + + + cells, IgG1 class-switched B cells and IgG1 CD38 memory-like B to a higher degree by IgE in TFR–DTR compared with control mice cells were also increased in TFR–DTR mice, suggesting that TFR cells (Mann–Whitney U-test, P < 0.01) (Fig. 3b). These autoantigens regulate many arms of B cell effector responses (Fig. 2b,c). TFR cells include La/SSB and Ro-52/SSA, which are targets for autoreac- have previously been shown to cause metabolic reprogramming, tive IgE responses in SLE patients; anti-SSB and -SSA IgE may be including inhibition of glycolysis, in B cells15. It was found that indicators of immune complex-mediated disease28. In addition,

Glut1 expression was higher in B cells from TFR cell-deleted com- complement had higher IgE autoantibody scores in TFR–DTR com- pared with control mice (Fig. 2d). These results are consistent with pared with control mice. When a less stringent cutoff of P < 0.05 a mechanism in which TFR cells inhibit metabolism in B cells in vivo. was used, anti-β2-microglobulin and anti-GP2 IgE were present in To determine whether TFR cells regulate antigen-specific anti- TFR–DTR mice, but not in control mice (data not shown). Using body production by regulating early GC responses, total and the stringent P < 0.01 cutoff, there was evidence of increases in IgG NP-specific antibody levels were assessed in TFR–DTR mice. There and IgE targeting the same eight autoantigens, including Ro-52/ was a twofold increase in total IgG and a ~2.5-fold increase in SSA, MPO, CENP-B, PL-7, TTG and M2 (Fig. 3c,d). Some of these

NP-specific IgG, demonstrating that TFR cells regulate antigen-spe- IgG and IgE autoantibodies, such as anti-Ro-52/SSA, are increased cific antibody responses (Fig. 2e and see Supplementary Fig. 2a). in SLE patients.

Moderate increases in total IgA and substantial increases in total Next, it was determined whether TFR cells can regulate initial IgE were also found in TFR cell-deleted mice (Fig. 2e). The latter activation and class-switch recombination of autoreactive B cells. A result was unexpected because TFR–DTR mice are on a C57bl/6 TFH cell-mediated, antigen-specific, autoreactive B cell, class-switch background. These results were not due to preferential deletion of recombination assay was developed. Myelin oligodendrocyte glyco- a TFR cell subset because the small number of TFR cells remaining protein (MOG) was chosen as the model autoantigen because MOG 18,29 in TFR–DTR mice phenotypically resemble TFR cells from control immunization generates functional TFH and TFR cells , MOG- mice (see Supplementary Fig. 2b). Increased Ki67 expression was specific B cells cause a Devic-like disease in experimental autoim- 30,31 found in the TFR cells remaining in TFR–DTR mice, probably due mune encephalomyelitis models and B cell depletion has a large

Nature Immunology | www.nature.com/natureimmunology Articles Nature Immunology

a Gated on b + – CD4 CD19 5 5 5 10 2.57 10 4 10 12.5 30.7 6 P = 0.0002 20 110 4 4 10 8 10 10 P = 0.0414 3 P = 0.0250 3 3 10 10 10 2 P < 0.0001 100 2 15 10 102 10 Control 0 6 Control 0 0 4 0 2 3 4 5 90 2 3 4 5 0 2 3 4 5 0 10 10 10 10 10 4 10 10 10 10 10 10 10 10 5 5 5 10 3.88 80 10 4 Naive B 10 cells of CD 4 8.86 cells of CD 4 4 4 13.6 2 10 FH 10 10 FR 5 3 2

T 10 3 3 T TFR–DTR 70 TFR–DTR 10 10 2 2 10 2 0 GC B cells (% ) 10 10 GL7-FITC 0 0 0 0 60

0 FoxP3-A488 CXCR5-BV421 0 2 3 4 5 2 3 0 2 3 4 5 0 4 5 10 10 10 10 10 10 10 10 10 10 10 10 –DTR –DTR FAS-PE- –DTR –DTR CD4-PerCP- Control Control Control Control ICOS-PE T FR T FR Cy7 T FR T FR Cy5.5

c 8 8 d ) 0.5 ) 3,000 P = 0.0482 + P = 0.0186 + P = 0.004 P = 0.0198 0.4 6 6 2,500 0.3 (% CD19 (% CD19 + – 4 4 0.2 2,000 CD38 CD38 + + 2 2

0.1 Glut1 MFI of B cells 0 2 3 4 5 1,500 IgG1 IgG1 Plasma cells (% lymph) 0.0 0 0 10 10 10 10 Glut1 –DTR Control T FR –DTR –DTR Control –DTR FR Control Control T T FR T FR e 4 5 15 8 P = 0.0393 P < 0.0001 P = 0.0012 P < 0.0001 3 4 6 10 3 2 4 2 5 1 1 2 IgG (normalized ) IgA (normalized ) IgE (normalized ) 0 NP IgG (normalized) 0 0 0

–DTR –DTR Control –DTR –DTR FR Control Control T T FR FR Control T T FR f Gated on g CD4+CD19– 5 5 10 5.81 10 P = 0.2936 P = 0.0058 4 21.8 8 0.25 104 10 4 P = 0.0033 15 P = 0.0046 3 Control 103 10 0.20 2 6 102 10 3 0 0 10 0.15 0 2 3 4 5 0 2 3 4 5 2 4 10 10 10 10 10 10 10 10 0.10 5 5 10 6.3 10 5 cells (% CD4 ) 4 4 14 1 cells (% CD4 ) 2 FR 10 10 FH 0.05 GC B cells (% B) T 3 3 T TFR–DTR 10 10 Plasma cells (% live ) 2 2 0 0 0.00 10 10 0

0 FoxP3-A488 0 CXCR5-BV42 1 0 2 3 4 5 0 2 3 4 5 –DTR –DTR Control –DTR –DTR 10 10 10 10 10 10 10 10 FR Control Control Control T T FR T FR T FR ICOS-PE CD4-PerCP- Cy5.5

h P = 0.0785 2.0 8 P = 0.3463

1.5 6

1.0 4

0.5 2 NP IgG (normalized ) Total IgG (normalized) 0.0 0

–DTR –DTR Control Control T FR T FR

+ + + + – Fig. 2 | TFR cells potently regulate early GC formation. a, Quantification of TFR (gated as CD4 ICOS CXCR5 FoxP3 CD19 ) and TFH (gated as + + + - – Cre LoxStopLoxDTR/WT Cre WT/WT CD4 ICOS CXCR5 FoxP3 CD19 ) cells from dLNs of TFR–DTR (Foxp3 Cxcr5 ) or control (Foxp3 Cxcr5 ) mice 21 d after immunization. + + + DT was administered on days 5, 7 and 9 to delete TFR cells before GC initiation. b, Quantification of GC B cells (gated as CD19 GL7 FAS ) and naive B cells (gated as CD38+IgG1–) from dLNs at day 21 after immunization as in a. c, Quantification of plasma cells (gated as CD138+), class-switched B cells (gated as CD19+IgG1+CD38–) and memory-like B cells (gated as CD19+IgG1+CD38+) at day 21 after immunization as in a. d, Glut1 expression on B cells from mice as in a. Representative histogram is shown on the left and quantification on the right. MFI, mean fluorescence intensity. e, Quantification of total IgG (far left), NP- + + specific IgG (middle left), total IgE (middle right) and total IgA (far right) analyzed from serum of mice as in a. f, Quantification of TFR (gated as CD4 ICOS C + + – + + + – – Cre LoxStopLoxDTR/WT Cre WT/WT XCR5 FoxP3 CD19 ) and TFH (gated as CD4 ICOS CXCR5 FoxP3 CD19 ) cells from dLNs of TFR-DTR (Foxp3 Cxcr5 ) or control (Foxp3 Cxcr5 ) mice at day 21 after immunization. DT was administered on days 10, 12 and 14 to delete TFR cells after GC formation. g, Quantification of GC B cells (gated as CD19+GL7+FAS+) and plasma cells (CD138+) from dLNs at day 21 after immunization as in f. h, Quantification of total IgG (left) and NP-specific IgG (right). Column graphs represent the mean with error bars indicating s.e.m. The P value indicates a two-tailed Student’s t-test. Data are either combined results from four (a–c,e) or three (f–h) independent experiments, or from an individual experiment that represents two independent experiments (d).

Nature Immunology | www.nature.com/natureimmunology Nature Immunology Articles

a Control TFR–DTR 20

15

10

IgG (antibody score) 5

nd 0

M2 M2 BPI BPI TTGTTG TPOTPO PL-7PL-7 MPOMPO LC1 LC1Jo-1Jo-1 GP2GP2 dsDNAdsDNA PL-12PL-12 Scl-70Scl-70 PCNAPCNA POLBPOLB NucleolinNucleolin CENP-BCENP-B HistoneHistone H1 H1 HistoneHistone H4 H4 Ro-52/SSARo-52/SSA HistoneHistone H3 H3 Ro-60/SSARo-60/SSA CollagenCollagen IV IV ProteoglycanProteoglycan GliadinGliadin (IgG)U1-snRNP-C (IgG)U1-snRNP-C HistoneHistone H2B H2B IntrinsicIntrinsic Factor Factor U1-snRNP-68U1-snRNP-68EntaktinEntaktin EDTA EDTA KU (P70/P80)KU (P70/P80) U1-snRNP-BB’U1-snRNP-BB’ b2-microglobulinb2-microglobulin ComplementComplement C9 C9 HeparanHeparan HSPG HSPG ComplementComplement C1q C1q MitochondrialMitochondrial antige antige NucleosomeNucleosome antigen antigen RiboRibo phasphoprotein phasphoprotein P0 P0

b c IgG IgE 20

Control TFR–DTR 30 87 15

10 d IgG IgE

ControlTFR–DTR ControlTFR–DTR

IgE (antibody score) Ro-52/SSA 5 Ribophosphoprotein P0 MPO CENP-B PL-7 0 TTG M2 M2 MPO MPO PL-7 PL-7 TTG TTG Complement C9 La/SSBLa/SSB CENP-BCENP-B M2 FibronectinFibronectin Ro-52/SSARo-52/SSA

ComplementComplement C4 C4 ComplementComplement C9 C9 ComplementComplementComplement C8 Complement C8 C5 C5 ComplementComplement C7 C7 ComplementComplement C3 C3 Antibody score

Ribo phasphoproteinRibo phasphoprotein P0 P0 Row min. Row max.

efg Gated on B cells IgHMOG Foxp3 GFP 1.31 P = 0.0054 40,000 30 MOG-B P = 0.0323 30,000 MOG 35–55 20

20,000 24.6

d7 (% CD19) MOG-B + 10 TFH cells

10,000 IgG 1 T cells T cells Number of B cell s MOG-B FH FR rMOG 0 0 4.57 MOG-B +++ MOG-B+ ++ MOG-B _ 6 d _ T cells TFH ++ Analysis TFH cells ++ FH TFR cells _ _ _ _ TFR + TFR cells + IgG1 GL7

Fig. 3 | TFR cells control autoreactive IgG and IgE during foreign antibody responses. a, Quantification of IgG scores for indicated autoantigens Cre LoxStopLoxDTR/WT Cre WT/WT from TFR–DTR (Foxp3 Cxcr5 ) or control (Foxp3 Cxcr5 ) mice that were immunized with NP-OVA and given DT to delete TFR cells before GC initiation, as in Fig. 2a. Serum was collected at day 21 after immunization. Autoantigens that showed a significant difference between control and TFR–DTR mice (38 out of 123, P < 0.01, Mann–Whitney U-test) are shown. b, Quantification of IgE scores for indicated autoantigens as in a. Autoantigens that showed a significant difference between control and TFR–DTR mice (15 out of 123, P < 0.01, Mann–Whitney U-test) are shown. c, Venn diagram showing overlap of differentially expressed autoantibodies for IgG and IgE isotypes in TFR–DTR mice. d, Heatmap showing IgG and

IgE autoantibody scores for the eight overlapping autoantigens in TFR–DTR compared with control mice. e, Schematic of in vitro, autoreactive, B cell MOG suppression assay. TFH and TFR cells from MOG35–55-immunized mice were cultured with IgH B cells in the presence of rMOG for 6 d. f, Relative count of B cells from suppression assays as in e. g, Quantification of class switching to IgG1 in suppression assays as in e. Representative plots (left) and quantification (right) are shown. Graphs are box-and-whisker plots with horizontal lines indicating the mean and bars indicating the range of values (a,b). Column graphs represent the mean with error bars indicating s.e.m. (f,g). Data are from an individual experiment with five mice per group (a,b), or are replicate suppression assays from an individual experiment and represent three independent experiments (e–g). The P value indicates a two-tailed Student’s t-test.

32 therapeutic benefit in multiple sclerosis . For this assay, TFH and TFR of recombinant MOG (rMOG) (Fig. 3e). TFH cells could stimulate GFP MOG cells were sorted from MOG35–55-immunized Foxp3 mice and cul- IgH B cells to expand and undergo class-switch recombination MOG MOG tured with B cells isolated from naive IgH mice in the presence (Fig. 3f,g). Importantly, TFR cells could potently suppress IgH B

Nature Immunology | www.nature.com/natureimmunology Articles Nature Immunology

a b c 500 P = 0.0005 1.4 P = 0.0046 1.2 400

) 1.0 Deletion pre-GC –1 300 0.8 P = 0.0059 Day 5Day 7Day 9 Day 30 Day 38 200 0.6 NP2/NP16 ratio

DT DT DT Boost Harvest NP IgG ( µ g ml 0.4 NP-OVA 100 NP-OVA PBS 0.2 Addavax 0 0.0

–DTR –DTR –DTR Control Control Control T FR T FR T FR Day 30 Day 8 Day 8 post-boost post-boost def 4 P = 0.0142 15 40 P = 0.7587 P = 0.6070 ) + 3 ) + 30 10 cells) FH 2 20 (% T + cells (% CD4 5 FH T

1 Ki67

GC B cells (% CD19 10

0 0 0

–DTR –DTR Control Control FR –DTR T FR T Control T FR

Fig. 4 | TFR cells regulate antibody memory responses. a, Schematic of TFR cell deletion to assess memory responses. TFR–DTR or control mice were immunized with NP-OVA in Addavax and DT was given on days 5, 7 and 9 to delete TFR cells before GC formation. Mice received a boost of NP-OVA without adjuvant at day 30. Mice were harvested on day 38. b, Analysis of NP-specific IgG levels before and after NP-OVA boost as in a. c, Quantification of the NP2/NP16 ratio in experiments as in a. d, Quantification of GC B cells (gated as CD19+GL7+FAS+) from dLNs of mice at + + + – – day 38 as in a. e, Quantification of TFH (gated as CD4 ICOS CXCR5 FoxP3 CD19 ) cells at day 38 as in a. f, Quantification of Ki67 expression in TFH cells gated as in e. Column graphs represent the mean with error bars indicating s.e.m. The P value indicates a two-tailed Student’s t-test. Data represent combined data from three independent experiments. cell expansion and class-switch recombination. These data demon- Memory B cell responses probably require secondary GCs to 33 strate that TFR cells can regulate initial autoreactive B cell activation produce high-affinity antibody . Therefore, the study next analyzed and class-switch recombination. GC responses in TFR–DTR mice in which TFR cells were deleted and boosted with NP-OVA. Increases in GC B cells were found in

TFR cells regulate antigen-specific antibody during memory TFR–DTR compared with control mice, suggesting augmented sec- responses. Next, we determined whether TFR cells regulate memory ondary GCs after rechallenge (Fig. 4d). However, it is important to Cre wt/wt B cell responses. To do this, TFR–DTR or control (Foxp3 Cxcr5 ) note that this assay cannot distinguish GC B cells that form from mice were immunized with NP-OVA containing a mild adjuvant, memory B cells or naive B cells. In contrast to GC B cells, there were

MF59-like Addavax, and given DT from day 5 to day 9. At day 30, no significant increases in TFH cell frequency or expression of Ki67 mice were boosted intraperitoneally with NP-OVA that did not in TFR–DTR mice after rechallenge (Fig. 4e,f). Taken together, these contain adjuvant (Fig. 4a). It was found that NP-specific antibody data indicate that TFR cells restrain the quantity, but promote the responses were ~2.8 times higher before, and ~3.3-fold higher after, affinity, of antigen-specific antibody during memory responses by the boost in TFR–DTR compared with control mice (Fig. 4b). These regulating early GCs. data suggest that antigen-specific memory responses are substan- tially increased when TFR cells are deleted before GC initiation, HDM antigen generates distinct populations of TFH and TFR cells. suggesting, in turn, that TFR cells control B cell memory to limit The finding that TFR cell deletion results in increased IgE levels in antibody responses. Next, it was determined whether the increased mice immunized with NP-OVA suggested that TFR cells may be able antigen-specific antibody had altered affinity because it was hypoth- to regulate TH2-like responses. Therefore, it was next determined esized that TFR cells can set thresholds on B cell responses. ELISAs whether HDM exposure, a TH2-like response, generated TFH and were performed with low and high ratios of NP and the NP2/NP16 TFR cells. C57bl/6 mice were challenged with HDM intranasally ratio was calculated to approximate the affinity of NP-specific every 2 d and mediastinal lymph nodes were assessed on day 7. antibody. After a boost, antibody had a lower NP2/NP16 ratio in Unimmunized mice had a very small ICOS+CXCR5+CD4+ popula- 13 TFR–DTR compared with control mice, suggesting lower affinity tion made up of ~60% TFR cells, a ratio that is typical in basal states (Fig. 4c and see Supplementary Fig. 3a). Notable changes in the (Fig. 5a). In comparison, HDM-exposed mice developed a sub- + + + NP2/NP16 ratio were not found in the non-boosted experiments stantial population of ICOS CXCR5 CD4 cells in which TFR cells (see Supplementary Fig. 3b,c). were only ~30%. Moreover, a subpopulation of TFH and TFR cells was

Nature Immunology | www.nature.com/natureimmunology Nature Immunology Articles

a Gated on Gated on CD4+CD19– CD4+CD19– 5 5 5 5 10 4.12 10 10 0.82 10 4 104 4 4 10 63.5 T 10 10 49.9 FR 15,000 ‘GC’ TFR 3 3 3 103 10 10 10 cells P = 0.0006 Control 2 2 34.3 T 2 2 47.8 Control 10 10 FH 10 10 ‘GC’ TFH 0 0 10,000 0 0 cells

0 2 3 4 5 0 2 3 4 5 0 2 3 4 5 0 2 3 4 5 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 P = 0.0005 5 105 105 5,000 5 10 12.6 Cells (count) 10 3.53 4 104 104 4 10 32 10 18 ‘GC’ T TFR FR 3 3 3 HDM 10 10 0 HDM 103 10 cells 2 2 66.6 2 80.7 ‘GC’ T 10 10 T 102 10 FH FH HDM HDM 0 cells 0 0 Control Control 0 FoxP3-A488 CXCR5-BV421 FoxP3-A488 CXCR5-BV421 2 3 4 2 3 4 5 2 3 4 5 0 5 2 3 0 0 TFH TFR 0 4 5 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 ICOS-PE CD4-PerCP-Cy5.5 PD-1-PE-Cy7 CD4-PerCP-Cy5.5 b c 5,000 OVA T Gene NES con 1.0 Cell set score OVA T FH OVA TFH TFH 1.69

OVA T OVA TFH TH21.34 0 FR HDM TFH TFH 1.68 HDM TFH 0.5 HDM TFH TH21.17 PC2 HDM TFR –5,000 Enrichment score 0.0

–10,000 010,00020,00030,00040,000 30,000 40,000 50,000 60,000 PC1 Enrichment TFH Tcon

d Gene NES e f T T Cell FH FR set score Tcon 1.0 OvaOva HDM Ova HDM OVA T T 2.01 FR FR 1700066N19Rik Fan1 HDM TFR TFR 1.97 OVA versus HDM OVA versus HDM T T Itga4 0.5 FH FR E030030I06Rik_1 E030030I06Rik_1 Gfi1 335 39 626 Kif13b Fbf1 Enrichment score 0.0 Simc1 Tatdn3 0 10,000 20,000 30,000 40,000 Anxa1 Gcnt1 Dsn1 Enrichment TFR Tcon Fbxl22 Rbm44 + – Gm13009 g T T h Gated on CD4 CD19 T FH FR + + Prr13 con ICOS CXCR5 Irx5 5 10 Zfp239 OVA OVA HDM OVA HDM Rnf128 104 P = 0.0397 Ascl2 0 20 Traf3ip1 Bcl6 103 Acn9 Ctla4 FMO F13a1 2 ) 10 + 15 Klf11 Cxcr5 + Gata3 0 Ctdspl Mettl21d Gfi1 0 2 3 4 5 10 10 10 10 Batf3 10 CXCR5 Icos (% CD4 + + Il17re Id2 5 10 Tyrobp Id3 4 IL-13 ICO S Serpinb1a 10 5 Il10 6.48 Cd163l1 Il13 Full 103 Wnt10a Il2 stain Lrsam1 2 0 Il21 10 Stx3 Il2ra 0 Tctn3 IL-13 PE-Cy7 FMO IL-13 Zik1 Il4 0 2 3 4 5 Maf 10 10 10 10 Ly6c1 Pdcd1 CD4-PerCP- Mafk Cy5.5 Trim17

Fig. 5 | HDM antigen generates distinct populations of TFH and TFR cells. a, Quantification of TFH and TFR cells in response to HDM challenge. WT mice were either challenged or not challenged (control) with HDM intranasally on days 0, 2, 4 and 6. The dLNs were harvested on day 7. The gating strategy to identify TFH and TFR cells (left), total numbers of TFH and TFR cells (middle), and gating strategy for ‘GC’ TFH and TFR cells (right) is shown. b, PCA showing + + + – – + + + + – the relationship between transcriptional profiles of TFH (CD4 ICOS CXCR5 FoxP3 CD19 ) and TFR (CD4 ICOS CXCR5 FoxP3 CD19 ) cells generated in GFP response to NP-OVA (subcutaneous) or HDM (intranasal) challenge in Foxp3 mice. PC, principal component. c, GSEA comparing TFH cells generated in response to NP-OVA or HDM for TFH or TH2 signatures (GSE14308). NES, normalized enrichment score. d, GSEA comparing TFR cells generated in response to NP-OVA or HDM for TFR signatures. e, Venn diagram demonstrating the overlap of differentially expressed genes (P < 0.05) between NP-OVA and HDM components for TFH and TFR cells. f, Heatmap showing the 39 common differentially expressed genes in TFH and TFR cells in NP-OVA versus

HDM challenge as in e. g, Heatmap of common follicular T cell and TH2 genes in TFH and TFR cells generated in response to NP-OVA or HDM challenge. h, IL-13 production by HDM TFH cells. Intracellular staining was performed on HDM-treated mice as in a. FMO, stain without anti-IL-13 antibody. Column graphs represent the mean with error bars indicating s.e.m. The P value indicates a two-tailed Student’s t-test. Data either represent three independent experiments (a,h) or are combined data from two independent experiments (b–g).

Nature Immunology | www.nature.com/natureimmunology Articles Nature Immunology

found, which assumed a GC-like phenotype, suggesting proper TFH undergo class switching to IgG1 and, to a lesser extent, IgE (Fig. 6f). and TFR cell effector differentiation. Importantly, addition of TFR cells resulted in near-complete reduc- + + To determine whether after HDM exposure TFH and TFR cells tion in IgE B cells and a substantial reduction in IgG1 B cells transcriptionally resemble TFH and TFR cells, transcriptional anal- (Fig. 6f,g). No evidence was found of class switching to IgE when ysis was performed. Foxp3IRES−GFP mice were immunized with HDM was omitted from the wells, or when similar cultures NP-OVA (emulsified in Freund’s complete adjuvant) subcutane- were performed using cells from NP-OVA immunization (see ously on day 0 or HDM intranasally on days 0, 2, 4 and 6, and Supplementary Fig. 4; data not shown). Although HDM TFR cells + + + − − TFH (gated as CD4 ICOS CXCR5 FoxP3 CD19 ), TFR (gated as suppressed class switching of NP-OVA B cells to IgG1, NP-OVA TFR + + + + – CD4 ICOS CXCR5 FoxP3 CD19 ) or T conventional (‘Tcon’, gated cells may suppress TFH13 cell-mediated IgE class switching of B cells + - – – – as CD4 ICOS CXCR5 FoxP3 CD19 ) cells were sorted on day 7 less potently than HDM TFR cells (see Supplementary Fig. 4). and RNA sequencing (RNA-seq) transcriptional analysis was To determine whether TFR cells prevent class switching to IgE, performed. Using principal component analysis (PCA), most fol- suppress already class-switched IgE+ B cells, or both, levels of GL7, licular T cell populations clustered separately from Tcon cells; how- a GC B cell-expressed molecule that is attenuated on B cells during + ever, HDM TFH cells clustered closer to Tcon cells than other cells TFR suppression, were assessed. It was found that IgE B cells had (Fig. 5b). Next, it was determined whether TFH cells from HDM- lower expression of GL7 in the presence of TFR cells, suggesting that + challenged mice transcriptionally resemble TFH cells or whether IgE class-switched B cells are less activated (Fig. 6h). Protein levels they take on a TH2-like phenotype. OVA- and HDM-specific TFH of IgE and IgG1 were also analyzed within switched B cells. Both cells had a similar enrichment for TFH genes and HDM TFH cells IgE- and IgG1-expressing B cells had lower expression of IgE and did not have enrichment for TH2 genes (Fig. 5c). Similarly, TFR cells IgG1, respectively, if TFR cells were present, although this did not from both OVA and HDM challenges had strong enrichment for reach statistical significance for IgE (Fig. 6i). To determine whether

TFR genes. (Fig. 5d). TFH13 cell cytokines were essential for full IgE responses, IL-13- or Although TFH and TFR cells from HDM-challenged mice had IL-4-blocking antibodies were added to cultures and the levels of intact transcriptional programs, differentially expressed genes were IgE assessed. TFH cells stimulated large amounts of IgE secretion, found between these cells and their OVA challenge counterparts. which was strongly attenuated when IL-13- or IL-4-blocking anti- There were 374 genes differentially expressed (P < 0.05) between bodies were added, demonstrating that cytokines produced by TFH13 OVA and HDM TFH cells, and 665 genes differentially expressed cells stimulate IgE (Fig. 6j). Importantly, IgE and IgG production (P < 0.05) between OVA and HDM TFR cells, with 39 genes being dif- was substantially suppressed by the presence of TFR cells (Fig. 6j). ferentially expressed in both TFH and TFR cells (Fig. 5e). When these Taken together, these data demonstrate that TFR cells can suppress 39 genes were assessed in more detail, a subset of genes was found TFH cell-mediated IL-13 and IgE responses in vitro. expressed in HDM TFH and TFR cells, but not OVA populations, such as Gfi1 (TFH cells: P = 0.0130; TFR cells: P = 0.0424), which has a role TFR cells regulate antigen-specific IgE responses in vivo. Next the 34 in stabilizing TH2 cells (Fig. 5f). Genes commonly expressed in TFH role of TFR cells was assessed in allergic immunity in vivo. For this, and TFR cells were also evaluated. Some genes, such as Icos and Id2, a HDM sensitization and challenge model was used that results in seemed to be expressed less in HDM TFH cells compared with OVA antigen-specific IgE responses and IL-13-dependent eosinophilic 10,35 Cre wt/wt TFH cells, although only Id2 was statistically significant (P = 0.0400) lung inflammation . TFR–DTR or control (Foxp3 Cxcr5 ) (Fig. 5g). A low, but positive, transcript for Il13 was found in HDM mice were sensitized with HDM and given DT to delete TFR cells. TFH cells that was not present in OVA TFH cells. In addition, HDM On day 7 mice with HDM were challenged (Fig. 7a). Robust dele- TFH cells expressed more Gata3. To assess whether a subset of TFH tion of TFR cells was found in TFR–DTR mice as a percentage of both cells could produce IL-13, intracellular cytokine staining was per- total CXCR5+CD4+ cells and total CD4+ cells (Fig. 7b). No evidence formed and a small proportion of TFH cells was found that produced was found of deletion of activated Treg cells in TFR–DTR mice (see IL-13 (Fig. 5h). Taken together, these data demonstrate that TFH and Supplementary Fig. 5a). Deletion of TFR cells did not result in altered + + TFR cells from HDM challenge have TFH and TFR transcriptional pro- frequencies of TFH cells, GC B cells, IgG1 B cells, IgE B cells or grams, but also some distinct transcriptional characteristics, such plasma cells (Fig. 7c). In addition, deletion of TFR cells did not alter as an IL-13 transcript in HDM TFH cells. As HDM TFH cells have an the relative class switching to IgG or IgE in GC B cells (Fig. 7d). intact TFH, but not a TH2, program yet express IL-13, these cells are However, when plasma cells were assessed, it was found that dele- referred to as ‘TFH13-like’ cells. tion of TFR cells resulted in small increases in IgE plasma cells (Fig. 7e). Moreover, TFR–DTR mice had substantially higher levels of TFR cells regulate TFH13 cell-mediated IgE responses to HDM total and HDM-specific IgE compared with control mice (Fig. 7f). in vitro. As it was found that TFH13-like cells from HDM-challenged These data demonstrate that TFR cells can control HDM-specific IgE mice expressed IL-13, and TFR cell-deleted mice had elevated levels responses in vivo. of autoreactive IgE, it was hypothesized that TFR cells may regu- To determine whether deletion of TFR cells results in altered late IL-13 and IgE responses in the context of TH2-like allergic lung inflammation, bronchoalveolar lavage fluid was analyzed responses. To test this hypothesis, an in vitro suppression assay was and increases in eosinophils found (Fig. 7g). Histological analysis developed in which TFH13-like cells mediate class switching of B of lungs showed increased inflammation consisting of cell infiltra- cells to IgE in response to HDM antigen. HDM was given intra- tion to the airway/vessel walls and alveolar parenchyma in TFR cell- IRES-GFP nasally to Foxp3 mice every 2 d and, on day 7, total B, TFH deleted mice compared with control mice (Fig. 7h). It was found + + + – – (gated as CD4 ICOS CXCR5 FoxP3 CD19 ) and TFR (gated as CD4 that the immune cell infiltrate in the lungs of TFR cell-deleted mice +ICOS+CXCR5+FoxP3+CD19–) cells were sorted from mediastinal was positive for Gr1 or SiglecF, suggesting the presence of granulo- lymph nodes. Cells were cultured together along with HDM for 6 d cytes and eosinophils, respectively (Fig. 7i and see Supplementary

(Fig. 6a). TFR cells inhibited TFH cell proliferation and were still Fig. 5b). Taken together, these data demonstrate that TFR cells regu- present at the end of the culture (Fig. 6b). It was found that cultures late HDM-specific IgE responses in vivo and control immune cell containing TFH13 and B cells contained large amounts of TH2-like infiltrate during HDM sensitization and challenge. cytokines, including IL-5, IL-13 and IL-4, all of which were sup- pressed by the addition of TFR cells (Fig. 6c,d). In addition, TFR cells Discussion + − suppressed the frequency of TFH cells (gated as CD4 FoxP3 cells) The precise role of TFR cells in modulating B cell responses has been expressing IL-13 protein (Fig. 6e). TFH13 cells stimulated B cells to elusive due to the lack of specific mouse models to study TFR cells.

Nature Immunology | www.nature.com/natureimmunology Nature Immunology Articles

a b c B B + B + B + 2,500 60 TFH TFH + TFH P < 0.0001 ) P = 0.0019

+ –1 TFR T (pg ml ) TFH B HDM 2,000 FR 40 IL-18 6,236 1,500 IL-5 5,983 IL-13 4,168 6 d 1,000 IL-10 3,094 cells (count) 20

Analysis cells (% CD 4

FH IL-4 2,103 FR

T 500

T IL-9 1,776 IL-6 805 0 0 IL-23 561 _ _ TFH ++ TFH ++ IL-2 557 __ __ IL-22 465 TFR + TFR + TNF-α 423 deGM-CSF 305 5 IL-17A 275 10 8.46 IL-27 251 P = 0.0005 4 10 P = 0.0018 8,000 5,000 P = 0.0027 10 IL-12p70 144 P = 0.0002 2,500 3 10 IFN- 117 8 γ 102 IL-1 79 ) 4,000 β ) 2,000

6,000 )

0 cells) –1 –1 –1

FH 6 1,500 3,000 0 2 3 4 5 Row min. Row max. 10 10 10 10 4,000 5 Below detection

10 1.18 (% T 4

1,000 2,000 + 104 IL-5 (pg ml IL-4 (pg ml 3 2,000 IL-13 (pg ml 10 2 500 1,000 IL-13 102 0 0 0 0 0 IL-13-PE-Cy7 _ _ _ 0 2 3 4 5 T ++ TFH ++ TFH ++ TFH ++ FH 10 10 10 10 ______T _ + TFR + TFR + TFR + CD4-PerCP- FR Cy5.5

f 2.5 P = 0.0019 30 P = 0.0013 B cells B + TFH cells B + TFH + TFR cells 5 5 5 2.0 10 1.07 0010 23.2 .018 10 5.56 0.015 20 104 104 104 1.5

3 3 3

10 10 10 (% B cells) +

(% B cells) 1.0

+ 10 102 102 102 IgE 0 0.02 0 2.01 0 0.185 0.5 IgG1 IgG1-PE 0 2 3 4 5 0 2 3 4 5 0 2 3 4 5 0.0 0 10 10 10 10 10 10 10 10 10 10 10 10 _ ++ _ ++ TFH IgE-BV421 TFH __ __ TFR + TFR +

g h Gated on Gated on 80 P = 0.0046 5,000 CD19+ CD19+IgE+ 10,000 P = 0.0066 P = 0.0002 )

4,000 B cells T cells + 8,000 60 FH alone T + T cells 3,000 FH FR 6,000 100 40 2,000 80 4,000 (relative count ) + (relative count)

+ 60 20 2,000 1,000 GL7 (MFI of Ig E Ig E

IgG1 40 0 0 20 0 _ ++ T _ ++ T ++ TFH FH 0 FH __ __ 0 102 103 104 105 T _ + T + TFR + FR FR GL7-FITC

i j P = 0.0020 P = 0.0152 P = 0.0293 6,000 1,000 P < 0.0001 P = 0.0006 )

150,000 + 25,000 ) 800 ) + ) P = 0.0599 –1 4,000 –1

IgG1 20,000 Ig E +

+ 600 100,000 15,000 400 2,000 IgG (ng ml

10,000 IgE (ng ml 50,000 200 5,000 0 0

IgE MFI (of CD1 9 _

IgG1 MFI (of CD1 9 +++ + 0 TFH _ ++ 0 TFH T ++ __+ _ _ T ++ FH TFR __+ FH ___ _ TFR _ + aIL-13 + T _ + TFR FR aIL-4 ____+

Fig. 6 | TFR cells regulate TFH13 cell-mediated IgE responses in vitro. a, Schematic of experimental design for an in vitro HDM suppression assay. Total B,

TFH and TFR cells were purified from the dLNs of mice that received HDM on day 0, 2, 4 and 6, and were added to culture wells along with HDM for 6 d. + + – – b, Quantification of total TFH cells (left) and the percentage of TFR cells (FoxP3 of CD4 IA CD19 cells) (right) from cultures as in a. c, Quantification of cytokines in culture supernatants from cultures as in a. Cytokines listed in red have levels notably lower in cultures containing TFR cells compared with cultures without TFR cells. GM-CSF, ganulocyte–macrophage colony-stimulating factor; IFN, interferon; TNF, tumor necrosis factor. d, Column graphs of

IL-5, IL-4 and IL-13 from the data in c. e, Intracellular cytokine staining of IL-13 in TFH cells from cultures as in a. f, Analysis of class switching to IgG1 and IgE in cultures as in a. Representative gating (left, pregated on CD19+IA+CD4– cells), IgE+ B cell quantification (middle) and IgG1+ B cell quantification (right) are shown. g, Counts of total IgE+ (left) and IgG1+ (right) B cells expressed as relative counts from experiments as in a. h, Expression of GL7 on IgE+ B cells from indicated cultures as in a. i, Expression of IgE on IgE+ B cells (left) and IgG1 on IgG1+ B cells are shown from cultures as in a. MFI, mean fluorescence intensity. j, Levels of IgE (left) and IgG (right) in culture supernatants from cultures as in a. Anti-IL-13 (aIL-13) or anti-IL-4 (aIL-4) was added to the indicated wells. Column graphs represent the mean with error bars indicating s.e.m. The P value indicates a two-tailed Student’s t-test (b–i,j; right) or one- way analysis of variance with Tukey’s correction (j; left). Data are from individual experiments and represent three independent experiments.

Nature Immunology | www.nature.com/natureimmunology Articles Nature Immunology

5 a b 105 10 19.4 4 4 ) P < 0.0001 P < 0.0001 Day 0 10 10 23 + 25 8 HDM sens. (i.n.) 3 3 TFR 10 10 ) + + 2 2 20 DT 10 10 TFH 6 Control 0 74.7 CXCR5 0 + Day 2 Day 4 Day 6 Day 7 Day 8 Day 9 Day 10Day 11 Day 15 15 0 2 3 4 5 0 2 3 4 5 10 10 10 10 10 10 10 10 4 DT DT DT Harvest 5 5 HDM 10 10 10 16.1 4 cells (% CD4 challenged 104 10 8.02 2 (i.n.) 3 TFR 5 FR T –DTR 3 T

FR 10 10 cells (% CD4

2 102 T FR 10 91.1 FH T 0 0 0 0 FoxP3-A488 CXCR5-BV421

0 2 3 4 5 0 2 3 4 5 10 10 10 10 10 10 10 10 –DTR –DTR Control Control ICOS-PE CD4-PerCP- T FR T FR Cy5.5

P = 0.8268 c P = 0.5611 15 P = 0.9323 1.5 P = 0.6594 2.0 P = 0.2570 ) 30 15 + ) ) + + ) + 1.5 20 10 10 1.0 1.0 10 5 5 0.5 B cells (% CD19 cells (% CD4 +

B cells (% CD19 0.5 + FH T Plasma cells (% lymph) GC B cells (% CD19 IgE IgG1 0 0 0.0 0 0.0

–DTR –DTR –DTR –DTR –DTR Control Control Control Control FR Control FR FR T T FR T FR T T

deP = 0.8408 P = 0.1754 3 80 100 P = 0.3844 14 P = 0.0035 Control T –DTR ControlT–DTR FR FR 80 5 5 60 5 5 12 10 60.5 10 66.1 2 10 10 4 45.6 4 57.2 10 104 104 10 10 60 40 8 3 3 3 103 10 10 10 6 40 1 (% GC B cells) 2 2 (% GC B cells) +

2 2 (% plasma cells)

+ 10 10 10 10 20 4 + 0 0.498 0 2.71 (% plasma cells)

0 1.09 0 1.22 + 20 IgG1-PE IgG1-PE IgE

IgG1 2 0 2 3 4 5 0 2 3 4 5 2 3 4 5 2 3 4 5 IgG1

0 0 IgE 10 10 10 10 10 10 10 10 10 10 0 0 10 10 10 10 10 10 0 0 IgE-BV421 IgE-BV421 –DTR –DTR –DTR –DTR Control Control Control Control T FR T FR T FR T FR

P = 0.1225 P = 0.3939 f P = 0.0016 g 4 8 i 30 P = 0.0038 8 8 P = 0.0111 7 7 25 6 6 3 6 3 20 4

15 3 4 2 2

10 2 Eos (AU) Total IgE (AU) HDM IgE (AU) 1 CD8 T cells (AU) 2 CD4 T cells (AU) 1 Control 5 1

0 0 0 0 0

–DTR –DTR –DTR Control –DTR –DTR Control Control FR Control Control FR T FR T T FR T FR T

h Naive Control TFR–DTR

P = 0.0117 3 H&E

TFR–DTR 2

1 Inflammation score 0 PAS Actin SiglecF Gr1 I-A

–DTR Control T FR

Fig. 7 | TFR cells regulate HDM-specific IgE responses in vivo. a, Schematic of HDM sensitization-and-challenge model to induce lung inflammation. Cre LoxStopLoxDTR/WT Cre WT/WT TFR–DTR (Foxp3 Cxcr5 ) or control (Foxp3 Cxcr5 ) mice received HDM sensitization (sens.) intranasally (i.n.) at day 0, followed by DT administration at days 0, 2, 4 and 6. Mice were challenged with HDM on days 7–11 and harvested on day 15. b, Analysis of TFR cells from dLNs of HDM- + + + + challenged mice as in a. Representative gating (left) and quantification (right) are shown. c, Quantification of TFH, GC B cells (CD19 GL7 FAS ), total IgE B cells (CD19+IgE+), total IgG1+ B cells (CD19+IgG1+) and total plasma cells (CD138+) from dLNs of HDM-challenged mice as in a. d, Quantification of IgG1 and IgE expression in GC B cells (CD19+GL7+FAS+). e, Quantification of IgG1 and IgE expression in plasma cells (CD138+). f, Quantification of total IgE or

HDM-specific IgE from HDM-challenged mice as in a (n = 30, control; n = 22, TFR–DTR). AU, arbritrary units. g, Quantification of eosinophils (Eos, left), CD4 T cells (middle) or CD8 T cells (right) from the bronchoalveolar lavage fluid of mice as in a (n = 24, control; n = 17, TFR–DTR). h, Lung inflammation scores (left) and representative images of hematoxylin and eosin (H&E) or periodic acid–Schiff (PAS) staining (right) of lung samples. Scale bars, 500 μm. i, Immunofluorescence micrographs of lungs stained for actin, SiglecF, Gr1 and I-A. Scale bars, 100 μm. Column graphs represent the mean with error bars indicating s.e.m. The P value indicates a two-tailed Student’s t-test (b–e) or a Mann–Whitney U-test (f–h). Data are from individual experiments and represent three independent experiments (b,c left), are combined data from four independent experiments (c right, d–g) or are from one experiment (h,i).

Nature Immunology | www.nature.com/natureimmunology Nature Immunology Articles

In the present study, a new TFR–DTR mouse strain was developed by TFR cells. Therefore, TFR cells are likely to have roles in regulat- to study TFR cells at distinct stages of immune responses. It was ing allergic inflammation and immunity to helminth infections, and found that TFR cells potently regulate foreign and self-reactive IgG modulation of TFR cells may help to control these responses. responses, especially before initial GC development. Surprisingly, a population of IL-13-expressing, TFH13-like cells was found dur- Online content ing HDM challenge, and TFR cells were found to potently regulate Any methods, additional references, Nature Research reporting these cells to control IL-13- and HDM-specific IgE responses. Taken summaries, source data, statements of code and data availability and together, these data indicate that TFR cells have a dynamic role in associated accession codes are available at https://doi.org/10.1038/ controlling many types of B cell responses to foreign and self-reac- s41590-019-0472-4. tive antigens, particularly before initial GC formation.

Although TFR cells can be found inside and outside of GCs, there Received: 27 December 2018; Accepted: 21 July 2019; Published: xx xx xxxx has been a debate in the field as to where and when TFR cells regulate B cells. TFR cells are less frequently found in GCs compared with the B cell follicle, and cytokines produced in high levels in GCs, such References 15,26 1. Crotty, S. Follicular helper CD4 T cells (TFH). Annu. Rev. Immunol. 29, as IL-21, can inhibit TFR cells . Based on these observations, it 621–663 (2011). has been suggested that TFR cells do not substantially regulate B cell 2. Victora, G. D. & Nussenzweig, M. C. Germinal centers. Annu. Rev. Immunol. 15 responses within GCs . However, TFR cells can regulate GC B cells 30, 429–457 (2012). 17 in vitro . In the present study, it has been demonstrated that TFR 3. Crotty, S. T follicular helper cell diferentiation, function, and roles in disease. cells potently regulate antibody responses before, but not after, GC Immunity 41, 529–542 (2014). 4. Vinuesa, C. G., Linterman, M. A., Yu, D. & MacLennan, I. C. Follicular formation. Although the experiments in this study cannot eliminate helper T cells. Annu. Rev. Immunol. 34, 335–368 (2016). the possibility of TFR cells having roles within very early GCs, the 5. Cannons, J. L., Lu, K. T. & Schwartzberg, P. L. T follicular helper cell diversity lack of a phenotype in mature GCs suggests that TFR cells regulate and plasticity. Trends Immunol. 34, 200–207 (2013). 6. Weinstein, J. S. et al. TFH cells progressively diferentiate to regulate the GC development. However, it is important to note that TFR cells may have more subtle roles in GCs such as facilitating GC resolution and germinal center response. Nat. Immunol. 17, 1197–1205 (2016). 7. Luthje, K. et al. Te development and fate of follicular helper T cells defned promoting immune homeostasis. by an IL-21 reporter mouse. Nat. Immunol. 13, 491–498 (2012). It was also found that TFR cells regulate memory B cell responses. 8. Liang, H. E. et al. Divergent expression patterns of IL-4 and IL-13 defne It was proposed that they do this by preventing GC formation unique functions in allergic immunity. Nat. Immunol. 13, 58–66 (2011). where some memory B cells originate. However, it is possible that 9. Ballesteros-Tato, A. et al. T follicular helper cell plasticity shapes pathogenic non-GC memory B cells may also be regulated by T cells in the B T helper 2 cell-mediated immunity to inhaled house dust mite. Immunity 44, FR 259–273 (2016). cell follicle. As reactivation of memory B cells may require second- 10. Coquet, J. M. et al. Interleukin-21-producing CD4+ T cells promote type 2 33 ary GCs , TFR cells probably have two roles in modulating B cell immunity to house dust mites. Immunity 43, 318–330 (2015). memory: regulation of memory B cell formation and regulation of 11. Noval Rivas, M. & Chatila, T. A. Regulatory T cells in allergic diseases. secondary GCs during memory B cell reactivation. In addition it J. Allergy Clin. Immunol. 138, 639–652 (2016). was found that T cells promote antibody affinity during memory, 12. Curotto de Lafaille, M. A. et al. Adaptive Foxp3+ regulatory T cell- FR dependent and -independent control of allergic infammation. Immunity 29, suggesting that TFR cells can modulate not only the quantity of anti- 114–126 (2008). gen-specific antibody, but also the quality of the antibody. Previous 13. Sage, P. T. & Sharpe, A. H. T follicular regulatory cells. Immunol. Rev. 271, studies found that deletion of Bcl6 in all Treg cell subsets from birth 246–259 (2016). resulted in increased autoreactive antibodies, but these antibodies 14. Maceiras, A. R., Fonseca, V. R., Agua-Doce, A. & Graca, L. T follicular 22,36 regulatory cells in mice and men. Immunology 152, 25–35 (2017). develop only after months . It was found that deletion of TFR cells 15. Sage, P. T. et al. Suppression by TFR cells leads to durable and resulted in increases in a variety of autoreactive IgG antibodies and, selective inhibition of B cell efector function. Nat. Immunol. 17, surprisingly, autoreactive IgE antibodies. Autoreactive IgE has been 1436–1446 (2016). found in autoimmune diseases such as SLE and has been suggested 16. Sage, P. T., Paterson, A. M., Lovitch, S. B. & Sharpe, A. H. Te coinhibitory to enhance autoimmune pathology27,28; it has also been found in receptor ctla-4 controls B cell responses by modulating T follicular 37 helper, T follicular regulatory, and T regulatory cells. Immunity 41, models of epithelial damage . 1026–1039 (2014). The role of TFH and TFR cells in TH2-like immunity has been 17. Sage, P. T., Alvarez, D., Godec, J., von Andrian, U. H. & Sharpe, A. H. unclear. Although TFH cells are not thought to make IL-13, TFH Circulating T follicular regulatory and helper cells have memory-like cells have been implicated in controlling IgE responses to HDM9,10. properties. J. Clin. Invest. 124, 5191–5204 (2014). 18. Sage, P. T., Francisco, L. M., Carman, C. V. & Sharpe, A. H. Te receptor Likewise, attenuated percentages of TFR cells correlate with worse 38 PD-1 controls follicular regulatory T cells in the lymph nodes and blood. allergy in patients . Evidence was found of a small frequency Nat. Immunol. 14, 152–161 (2013). of IL-13-producing TFH cells in vivo after HDM administration. 19. Wollenberg, I. et al. Regulation of the germinal center reaction by Foxp3+ Using an in vitro HDM IgE assay, it was shown that these TFH cells follicular regulatory T cells. J. Immunol. 187, 4553–4560 (2011). could produce large amounts of IL-13, IL-5 and IL-4, and potently 20. Wu, H. et al. Follicular regulatory T cells repress cytokine production by stimulate B cell IgE responses. These T cells have been referred follicular helper T cells and optimize IgG responses in mice. Eur. J. Immunol. FH 46, 1152–1161 (2016). as TFH13-like cells to distinguish them from previously described 21. Linterman, M. A. et al. Foxp3+ follicular regulatory T cells control the TFH2 cells. Both IL-13 and IgE responses were potently suppressed germinal center response. Nat. Med. 17, 975–982 (2011). by TFR cells. Interestingly, deletion of TFR cells in vivo during HDM 22. Fu, W. et al. Defciency in T follicular regulatory cells promotes sensitization/challenge resulted in higher levels of antigen-specific autoimmunity. J. Exp. Med. 215, 815–825 (2018). + IgE and increased lung inflammation. High-affinity IgE responses 23. Laidlaw, B. J. et al. Interleukin-10 from CD4 follicular regulatory T cells 39,40 promotes the germinal center response. Sci. Immunol. 2, eaan4767 (2017). occur preferentially through sequential switching of IgG1 to IgE . 24. Kim, J. M., Rasmussen, J. P. & Rudensky, A. Y. Regulatory T cells prevent As there was such a profound increase in IgE serum levels and catastrophic autoimmunity throughout the lifespan of mice. Nat. Immunol. 8, alterations in IgE+, but not IgG1+, plasma cells, these data suggest 191–197 (2007). 25. Hou, S. et al. FoxP3 and Ezh2 regulate TFR cell suppressive function and that TFR cells probably limit IgE plasma cell responses rather than sequential switching. However, more in-depth studies are necessary transcriptional program. J. Exp. Med. 216, 605–620 (2019). 26. Sayin, I. et al. Spatial distribution and function of T follicular regulatory cells to fully determine the role of TFR cells in sequential switching. Taken in human lymph nodes. J. Exp. Med. 215, 1531–1542 (2018). together, these results demonstrate that TFH cells can have roles in 27. Dema, B. et al. Immunoglobulin E plays an immunoregulatory role in lupus. TH2-like responses to HDM and that these responses are controlled J. Exp. Med. 211, 2159–2168 (2014).

Nature Immunology | www.nature.com/natureimmunology Articles Nature Immunology

28. Dema, B. et al. Autoreactive IgE is prevalent in systemic lupus erythematosus 40. Xiong, H., Dolpady, J., Wabl, M., Curotto de Lafaille, M. A. & Lafaille, J. J. and is associated with increased disease activity and nephritis. PLoS ONE 9, Sequential class switching is required for the generation of high afnity IgE e90424 (2014). antibodies. J. Exp. Med. 209, 353–364 (2012). 29. Sage, P. T. et al. Dendritic cell PD-L1 limits autoimmunity and follicular T cell diferentiation and function. J. Immunol. 200, 2592–2602 (2018). Acknowledgements 30. Bettelli, E., Baeten, D., Jager, A., Sobel, R. A. & Kuchroo, V. K. Myelin We would like to thank T. Chatila, R. Anthony, D. Wesemann and M. Carroll for helpful oligodendrocyte glycoprotein-specifc T and B cells cooperate to induce a discussions, the MICRON imaging core for help with microscopy and H. Wekerle and Devic-like disease in mice. J. Clin. Invest. 116, 2393–2402 (2006). S. Zamvil for reagents. This work was supported by the US National Institutes of Health 31. Krishnamoorthy, G., Lassmann, H., Wekerle, H. & Holz, A. Spontaneous through grant nos. K22AI132937 (P.T.S.), P01AI056299 (A.H.S.), R37AI34495 (B.R.B.) opticospinal encephalomyelitis in a double-transgenic mouse model of and R01HL11879 (B.R.B.), and the Evergrande Center for Immunologic Diseases. autoimmune T cell/B cell cooperation. J. Clin. Invest. 116, 2385–2392 (2006). 32. Mulero, P., Midaglia, L. & Montalban, X. Ocrelizumab: a new Author contributions milestone in multiple sclerosis therapy. Ter. Adv. Neurol. Disord. 11, R.L.C, J.D., M.T.M. and P.T.S. performed the experiments. R.L.C, A.D., S.B.L. and 1756286418773025 (2018). P.T.S. analyzed the data. B.R.B., V.K.K. and A.H.S provided key technical help and 33. McHeyzer-Williams, L. J., Milpied, P. J., Okitsu, S. L. & McHeyzer-Williams, reagents. P.T.S. conceived of the project and wrote the manuscript. All the authors M. G. Class-switched memory B cells remodel BCRs within secondary edited the manuscript. germinal centers. Nat. Immunol. 16, 296–305 (2015). 34. Zhu, J., Jankovic, D., Grinberg, A., Guo, L. & Paul, W. E. Gf-1 plays an important role in IL-2-mediated T2 cell expansion. Proc. Natl Acad. Sci. Competing interests USA 103, 18214–18219 (2006). The authors declare no competing interests. 35. Tomlinson, K. L., Davies, G. C., Sutton, D. J. & Palframan, R. T. Neutralisation of interleukin-13 in mice prevents airway pathology caused by Additional information chronic exposure to house dust mite. PLoS ONE 5, e13136 (2010). Supplementary information is available for this paper at https://doi.org/10.1038/ 36. Botta, D. et al. Dynamic regulation of T follicular regulatory cell s41590-019-0472-4. responses by interleukin 2 during infuenza infection. Nat. Immunol. 18, 1249–1260 (2017). Reprints and permissions information is available at www.nature.com/reprints. 37. Crawford, G. et al. Epithelial damage and tissue gammadelta T cells Correspondence and requests for materials should be addressed to P.T.S. promote a unique tumor-protective IgE response. Nat. Immunol. 19, Peer review information: Z. Fehervari was the primary editor on this article, and 859–870 (2018). managed its editorial process and peer review in collaboration with the rest of the 38. Yao, Y. et al. Allergen immunotherapy improves defective follicular regulatory editorial team. T cells in patients with allergic rhinitis. J. Allergy Clin. Immunol 144, 118–128 (2019). Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in 39. He, J. S. et al. IgG1 memory B cells keep the memory of IgE responses. published maps and institutional affiliations. Nat. Commun. 8, 641 (2017). © The Author(s), under exclusive licence to Springer Nature America, Inc. 2019

Nature Immunology | www.nature.com/natureimmunology Nature Immunology Articles

Methods substrate (Sigma), and plates were read on a plate reader (Spectramax). For NP- Mice. Studies of Foxp3IRES-GFP mice on the C57Bl/6 background have been published specific antibody ELISAs, Maxisorp plates were coated with NP-BSA (Biosearch previously41. Foxp3IRES-CreYFP and Foxp3DTR mice on the C57Bl/6 background were Technologies), followed by the secondary reagents above. Anti-HDM IgE kits from Jackson Laboratories. IgHMOG mice were a kind gif from Hartmut Wekerle42. (Chondrex) were used for HDM-specific antibody quantification. Cxcr5IRES-LoxP-STOP-LoxP-DTR knockin mice were generated by constructing a targeting vector in which an IRES-Frt-PGKNeo-Frt-LoxP-STOP-LoxP-hbegf sequence was Autoantigen arrays. Autoantibody reactivities against a panel of autoantigens were placed directly downstream of the stop sequence of Cxcr5. Te targeting vector measured using an autoantigen microarray platform developed by the University was introduced into C57bl/6 embryonic stem cells by electroporation, and the of Texas Southwestern Medical Center. Genepix Pro v.6.0 software was used to resulting neomycin-resistant embryonic stem cells were screened for homologous analyze IgG or IgE fluorescence. The antibody score was generated through the recombination. Positive clones were microinjected into albino-B6 blastocysts following equation: log2(net fluorescence intensity × signal-to-noise ratio + 1). The and implanted into pseudopregnant female mice to generate chimeras. Germline Mann–Whitney U-test was used to determine statistical significance. transmission was achieved and mice were bred to FlpE-carrying mice to remove 15 the NEO cassette29. Te resulting mice were then bred to Foxp3IRES-CreYFP mice to RNA-seq. RNA-seq was performed as described previously . Briefly, RNA was generate the TFR–DTR colony. Mouse progeny were routinely screened for leakiness isolated using MyOne Silane Dynabeads (Thermo Fisher Scientific). RNA was of the Foxp3IRES-CreYFP allele by fow cytometry. All mice were used according to fragmented, and barcoded using 8-basepair barcodes together with standard Brigham and Women’s Hospital and Harvard Medical School Institutional Animal Illumina adaptors. Primers were removed using Agencourt AMPure XP bead Care and Use Committee and National Institutes of Health guidelines. cleanup (Beckman Coulter/Agencourt) and samples were amplified with 14 PCR cycles. Libraries were gel purified and quantified using a Qubit high-sensitivity Immunization. Mice were immunized with 100 μg NP-OVA (Biosearch DNA kit (Invitrogen) and library quality was confirmed using Tapestation high- sensitivity DNA tapes (Agilent Technologies). RNA-seq reactions were sequenced Technologies) or 100 μg MOG35–55 (UCLA Biopolymer Facility) emulsified in H37RA Freund’s complete adjuvant subcutaneously in the mouse flanks as on an Illumina NextSeq sequencer (Illumina) according to the manufacturer’s previously described15,43, unless otherwise noted. For memory studies, mice instructions, sequencing 50-basepair reads. Analysis was performed using the CLC were immunized with 100 μg NP-OVA mixed with Addavax MF59-like adjuvant Genomics Workbench v.8.0.1 RNA-seq analysis software package (Qiagen). Briefly, (Invivogen) subcutaneously in one flank. Then, 30 d later mice received a boost of reads were aligned (mismatch cost = 2, insertion cost = 3, deletion cost = 3, length 100 μg NP-OVA in PBS intraperitoneally. For allergy studies, mice were sensitized fraction = 0.8, similarity fraction = 0.8) to the mouse genome and differential with 10 μg HDM (Greer Labs) in PBS intranasally and then challenged with 10 μg expression analysis was performed (total count filter cutoff = 5.0). Results were HDM in PBS intranasally on days 7, 8, 9, 10 and 11, followed by harvesting at day normalized to reads per million. Gene-e (Broad Institute) was used to generate heatmaps. The datasets generated during the current study are available on the 15. In some cases mice received 1 μg DT in PBS intraperitoneally to delete TFR cells at indicated timepoints. Gene Expression Omnibus (GSE134153) and are available from the corresponding author on reasonable request. Antibodies. The following antibodies were used for surface staining at 4 °C for 25 30 min: anti-CD4 (Biolegend, 1:200, RM4-5), anti-ICOS (Biolegend,1:200, 15F9), Gene set enrichment analysis. RNA-seq data were compared with TFH signatures , 25 anti-CD19 (Biolegend,1:200, 6D5), anti-PD-1 (1:200, RMP1-30), anti-CXCR5 TFR signatures or TH2 signatures (GSE14308) using default settings in gene set biotin (BD Biosciences,1:100, 2G8), anti-GL7 (BD Biosciences,1:200, GL-7), anti- enrichment analysis (GSEA) software (Broad Institute). HB-EGF/DTR (R&D Systems, 1:200, AF-259-NA), anti-CD38 (Biolegend, 1:200, 90), anti-CD138 (Biolegend, 1:200, 281-2), anti-IA (Biolegend,1:200, M5/114.15.2), Histology. Paraffin-embedded lung sections were stained with hematoxylin and anti-SiglecF (BD Biosciences, 1:200, E50-2440), anti-CD8a (Biolegend, 1:200, eosin or periodic acid–Schiff. Images were taken on an Olympus BX41 microscope 53-6.7), anti-CD11c (Biolegend, 1:200, N418) and anti-CD11b (Biolegend, 1:200, and Olympus DP26 camera and images obtained using Olympus CellSens v.1.1 M1/70). For CXCR5 detection, streptavidin-BV421 (Biolegend, 1:400, 405225) software. Lung inflammation was scored as: no inflammation: 0; perivascular/ was used at 4 °C. In some cases, anti-IgE (BD Biosciences, 1:200, R35-72) was peribronchial inflammation without infiltration of airway/vessel walls: 1; included to block IgE bound to cell surfaces. For intracellular staining, samples infiltration of airway/vessel walls without extension into alveolar parenchyma: 2; were fixed with the Foxp3 Fix/Perm buffer set according to the manufacturer’s destruction of alveolar parenchyma: 3. instructions (eBioscience). Samples were then intracellularly stained with anti- IgG1 (BD Biosciences, 1:200, A85-1), anti-IgE (BD Biosciences, 1:200, R35-72), Statistics. Most statistical tests were performed using Prism v.6.0 (GraphPad) anti-FoxP3 (eBiosciences, 1:200, FJK-16S), anti-Ki67 (BD Biosciences, 1:100, B56) using the two-tailed, unpaired, Student’s t-test for normalized data, or the Mann– or anti-Glut1 (Abcam, 1:200, EPR3915). In some cases, a donkey anti-goat A647 Whitney U-test for non-normal data, as indicated. Statistics for RNA-seq were secondary antibody was used (Invitrogen, 1:400, A-21447). See the Reporting performed using CLC Genomics Workbench (Qiagen). Statistics for gene set Summary for more details. enrichment was performed at the GSEA (Broad Institute). All measurements were taken from distinct samples, except for memory experiments in which the same Sorting. The draining lymph nodes (dLNs) were passed through 70-μm filters mice were bled before and after boost. and resuspended in PBS supplemented with 1% fetal bovine serum and 1 mM ethylenediaminetetraacetic acid. CD4+ cells were enriched by magnetic positive Reporting Summary. Further information on research design is available in the selection (Miltenyi Biotec). CD4+-enriched cells were then stained and sorted on Nature Research Reporting Summary linked to this article. an Aria II cell sorter (70-μm nozzle) using optimal purity settings. TFH cells were + + + – – + + gated as CD4 ICOS CXCR5 FoxP3 CD19 and TFR cells were gated as CD4 ICOS Data availability + + – + CXCR5 FoxP3 CD19 . B cells were isolated by flow-through from CD4 selection, The data that support the findings of this study are available from the which was then positively selected using CD19 beads (Miltenyi Biotec). corresponding author upon request. Transcriptomic data have been deposited in the Gene Expression Omnibus with the accession code GSE134153. Suppression assays. In vitro stimulation/suppression assays were performed by 4 4 4 culturing 5 × 10 B cells, 3 × 10 TFH cells and 1.5 × 10 TFR cells (isolated from dLNs –1 of HDM-, NP-OVA- or MOG35–55-challenged mice) along with 20 μg ml of HDM, References NP-OVA or rMOG in 96-well culture plates for 6 d. In some cases 20 μg ml–1 of 41. Bettelli, E. et al. Reciprocal developmental pathways for the generation of anti-IL-4 (Biolegend, 11B11) or anti-IL-13 (R&D Systems, MAB413) was added. pathogenic efector TH17 and regulatory T cells. Nature 441, 235–238 (2006). Cells and supernatants were collected for flow cytometric and ELISA. 42. Litzenburger, T. et al. B lymphocytes producing demyelinating autoantibodies: development and function in gene-targeted transgenic mice. J. Exp. Med. 188, ELISA. For total IgG and IgE levels in serum, Maxisorp (Nunc) plates were 169–180 (1998). coated with anti-mouse Ig (SouthernBiotech) or anti-IgE (BD Biosciences, 43. Sage, P. T. & Sharpe, A. H. In vitro assay to sensitively measure TFR R35-72), respectively, and the serum was incubated. Alkaline phosphatase suppressive capacity and TFH stimulation of B cell responses. Methods Mol. secondary reagents (SouthernBiotech) were added, followed by phosphatase Biol. 1291, 151–160 (2015).

Nature Immunology | www.nature.com/natureimmunology