Natural Tr1-like cells do not confer long-term tolerogenic memory

IL-10-producing Tr1 cells promote tolerance but their contributions to tolerogenic memory are unclear. Using 10BiT mice that carry a Foxp3-eGFP reporter and stably express CD90.1 following IL-10 production, we characterized the spatiotemporal dynamics of Tr1 cells in a house dust mite model of allergic airway inflammation. CD90.1+Foxp3-IL-10+ Tr1 cells arise from memory cells and rejoin the tissue-resident memory T-cell pool after cessation of IL-10 production. Persistent antigenic stimulation is necessary to sustain IL-10 production and Irf1 and Batf expression distinguishes CD90.1+Foxp3-IL-10+ Tr1 cells from CD90.1+Foxp3-IL-10- ‘former’ Tr1. Depletion of Tr1-like cells after primary sensitization exacerbates allergic airway inflammation. However, neither transfer nor depletion of former Tr1 cells influences either Tr1 numbers or the inflammatory response during subsequent allergen memory re-challenge weeks later. Together these data suggest that naturally-arising Tr1 cells do not necessarily give rise to more Tr1 upon allergen re-challenge or contribute to tolerogenic memory. This phenotypic instability may limit efforts to re-establish tolerance by expanding Tr1 in vivo.


Introduction
Allergic asthma is a common childhood illness and can be triggered by exposure to aeroallergens such as house dust mite (HDM). Hallmark features of the disease include aberrant T helper 2 (Th2) type responses, airway hyperreactivity, eosinophilic inflammation, increased IgE and mucus hypersecretion (Finkelman et al., 2010).
The balance between pro-inflammatory CD4+ Th2 cells and regulatory T cells, including CD4 +Foxp3+ regulatory T cells (Treg) and CD4+Foxp3-, interleukin-10 (IL-10)-producing Type one regulatory T cells (Tr1) cells, is a significant determinant in the development of allergic disease (Robinson, 2009). Allergen-specific immunotherapy may re-establish tolerance in part by expanding these regulatory T cell populations (Akdis and Akdis, 2014;Meiler et al., 2008). There is therefore great interest in developing stable, antigen specific regulatory T cells for treatment of asthma and allergies (Bassirpour and Zoratti, 2014).
However, much about naturally-arising Tr1 cells and endogenous IL-10 production in allergic airway disease remains unknown (White and Wraith, 2016). In particular, the contribution of antigenspecific Tr1 cells to tolerogenic memory is unclear. Given the persistence of allergen-specific immunity (Hondowicz et al., 2016) this information may prove vital for the development of successful Tr1-directed immunotherapies.
Here, we characterize the spatial and temporal dynamics of endogenous IL-10-producing T cells in a mouse model of HDM-induced allergic airway inflammation. For this purpose, we have used a mouse strain carrying a stable IL-10 reporter, the 10BiT mouse (Maynard et al., 2007). This strain carries multiple copies of a bacterial artificial chromosome (BAC) transgene containing a CD90.1 construct under the control of the IL-10 promoter such that cells that previously made IL-10 express CD90.1 which persists for some time after cessation of IL-10 production. This strain was previously crossed against the Foxp3-eGFP mouse strain to facilitate discrimination between Foxp3-Tr1 from Foxp3+ Treg (Maynard et al., 2007).
Using these animals, we have analyzed the population frequencies and cytokine production profiles of Tr1-like cells in multiple tissues at various stages of allergic airway inflammation including sensitization, challenge, resolution, and memory. To determine the functional contribution of Tr1-like memory T-cell populations we perform loss and gain of function studies by deleting these cells or adoptively transferring them before re-challenge with the allergen. Together, these studies elucidate the cellular source of IL-10 in allergic inflammation, their functional stability, and their contribution to tolerogenic memory in the lung.

IL-10-producing T cells accumulate at site of allergen sensitization
To interrogate the patterns of endogenous IL-10 production in allergic airway inflammation we used a house dust mite (HDM) induced murine model ( Figure 1A). In this model, animals are sensitized to crude house dust mite (HDM) protein intranasally (i.n.) over 2 weeks while control animals are given phosphate buffered saline (PBS). Analysis of the inflammatory response is carried out 4 days after completion of the challenge series. This model recapitulates key features of allergic responses associated with type two inflammation, with HDM-sensitized animals showing an increase in total number of cells ( Figure 1B) and total number of eosinophils ( Figure 1C) in the broncho-alveolar lavage (BAL) as compared to controls. Also, the lungs of HDM-sensitized animals exhibited an increase in peribronchiolar and perivascular inflammation ( Figure 1D). The levels of total IgE were also elevated in the BAL supernatants obtained from HDM-sensitized animals ( Figure 1E).
Next, using the 10BiT/Foxp3eGFP strain (Tg(Il10-Thy1)1Weav crossed to Foxp3 tm2Tch ) (Maynard et al., 2007), we characterized the endogenous IL-10 response in this model. This strain carries dual reporters for IL-10 and Foxp3. A bacterial artificial chromosome (BAC) containing CD90.1 under the control of an IL-10 promoter enables the detection of cells that have produced IL-10 via the expression of cell surface CD90.1 while a GFP knocked into the Foxp3 locus tracks endogenous Foxp3 expression (Maynard et al., 2007). Using this model, we found that the frequency of CD90.1+ Foxp3-CD4+ Tr1-like cells was increased in the HDM-sensitized animals in the BAL and the lungs ( Figure 1F,G), the sites of allergen challenge, but not in the pulmonary lymph node ( Figure 1H) or the spleen ( Figure 1I). Thus, CD90.1+Foxp3-Tr1-like cells specifically accumulated at the site of allergen challenge.

Lung Tr1-like cells are located in the lung parenchyma
We used intravenous labeling to further discriminate between lung-resident versus circulating cells (Anderson et al., 2014). For staining non-tissue-resident circulating cells in the lung, we injected CD45 antibody retro-orbitally 2 min before mice were euthanized. We found that the majority of CD90.1+Foxp3-CD4+ T cells were parenchymal (Figure 2A-D) as were the CD90.1+Foxp3+CD4+ T cells ( Figure 2E). In contrast more than half of CD90.1-Foxp3+CD4+ T cells and about half of the Figure 1. IL-10-producing T cells accumulate at site of allergen sensitization. C57Bl/6 mice were administered either PBS as control or crude HDM containing 20mg total protein in 50ml intranasally (i.n.) Six times over 2 weeks as shown. Terminal analysis was performed 4 days after last challenge. (B) The total number of cells and (C) total number of eosinophils in the BAL were determined. Data are pooled from two experiments. Each symbol represents a single animal. Error bars represent standard error of mean. n=10 for control group, n=10 for HDM group. (D) Representative haematoxylin and eosin stained lung sections showing perivascular and peribronchiolar inflammation in PBS control and HDM-sensitized animals. (E) The level of IgE in BAL supernatant was determined by ELISA. n=5 for control group, n=5 for HDM group. Error bars represent standard deviation of mean. Statistical significance was determined using an unpaired two tailed students t-test. Representative flow cytometry plots showing different T cell subsets as identified by surface expression of CD90.1 and Foxp3eGFP in the (F) BAL, (G) lungs, (H) pulmonary lymph nodes, and (I) spleens of control or HDMtreated mice 4 days after final challenge. The frequency of IL-10-producing Foxp3-(CD90.1+Foxp3-), IL-10-producing Foxp3+ (CD90.1+Foxp3+) and Foxp3+ cells which do not produce IL-10 (CD90.1-Foxp3+) within all CD4+ T cells in the four sites are plotted. Data are pooled from two experiments. Error bars represent standard error of mean. n=10 for control group, n=10 for HDM group. Data are representative of 4 independent experiments. Statistical significance was determined using 2-way ANOVA (post hoc test: Sidaks). *P0.05, **P0.01, ***P0.001 ****P0.0001. PBS=Phosphate buffered saline, HDM=house dust mite, i.n=intranasally, BAL=bronchoalveolar lavage. DOI: https://doi.org/10.7554/eLife.44821.002 The following source data is available for figure 1: Source data 1. IL-10-producing T cells accumulate at site of allergen sensitization. DOI: https://doi.org/10.7554/eLife.44821.003 CD90.1-Foxp3-CD4+ T cells were in the lung vasculature ( Figure 2F-G). These data indicated that both CD90.1+Foxp3-and CD90.1+Foxp3+ T cells are primarily within the lung parenchyma. Figure 2. IL-10-producing cells are located in the lung parenchyma at peak of inflammation. Intravascular (IV) labeling of cells was performed by retro orbital injection of CD45 antibody and animals were euthanized 2 minutes after injection. Analysis was carried at peak of inflammation day 16-post first house dust mite challenge. (A) Efficacy of labeling of cells was assessed in blood and in the lungs. (B) The proportion of cells in blood and lung, which were labeled with antibody IV, was quantified. Error bars are standard deviation. Statistical significance was determined using a paired two tailed students t-test. n. CD4+Foxp3-T cells are the primary source of IL-10 at the peak of the inflammatory response Since different cell types can produce IL-10, we further distinguished T cell-derived and non-T cellderived sources of IL-10 using CD90.1 expression in control and HDM-sensitized animals ( Figure 3A). We found marked differences in the composition of CD90.1+ cells between control and HDM-sensitized animals. In control animals, non-T cells (CD3-) comprised the majority of CD90.1+ whereas in HDM-sensitized animals, the majority of CD90.1+ were Foxp3-CD4+ T cells ( Figure 3B). The total number of CD90.1+ Foxp3-cells was predominantly increased in the lungs in comparison to other CD90.1 + subsets ( Figure 3C). The MFI of CD90.1 as a correlate of IL-10 production was also increased in the CD90.1+Foxp3-subsets as compared to CD3+CD4-and CD3-CD4-subsets ( Figure 3D). Thus, the increase in the frequency of Tr1-like cells at the site of allergen sensitization could be attributed to the accumulation of CD90.1+ Foxp3-CD4+ T cells in the lungs. These changes were specific to the site of allergen challenge as neither the frequency of total CD90.1+ Previously described Tr1 markers do not distinguish IL-10-producing Foxp3-T cells from IL-10-producing Foxp3+ Treg in this model We next investigated whether CD90.1+ Foxp3-cells in the lungs of HDM-sensitized animals expressed phenotypic markers previously associated with Tr1 cells in comparison to the other CD4+ T cells subsets (Gagliani et al., 2013;Roncarolo et al., 2006;Yao et al., 2015).
We measured the expression of CD25, the alpha subunit of the IL-2 receptor, which is highly expressed on Foxp3+Tregs and correlates with their suppressive function (Sakaguchi et al., 2009). CD25 was explicitly increased on Foxp3+ cells irrespective of IL-10 production, though CD90.1 +Foxp3+ exhibited a lower level of expression in comparison to CD90.1-Foxp3+ cells ( Figure 4A).
We also quantified the expression of KLRG1 and PD1, markers of senescence or exhaustion respectively that correlate with the suppressive function of regulatory T cells (Burton et al., 2014;Shevach, 2006;Stephens et al., 2007). Tr1 cells also express PD-1 during immunotherapy (Burton et al., 2014). We found that both KLRG1 and PD-1 are highly expressed on CD90.1+ cells irrespective of Foxp3 expression ( Figure 4B,C). These differences were reflected in the frequency of cells expressing these markers ( Figure 4D-F) All IL-10-producing CD4+ T cells irrespective of Foxp3 expression express high levels of CD44 ( Figure 4G). CD44 is a receptor for hyaluronan and studies from our group and others have shown that it can potentiate IL-10 responses in CD4+ T cells (Bollyky et al., 2011;Yao et al., 2015). Despite this association, CD44-/-mice did not show a decrease in the frequency of CD90.1+Foxp3or CD90.1+Foxp3+ cells in the HDM model, suggesting that CD44 is dispensable for their induction (Figure 4-figure supplement 1).
Together these data indicate that CD90.1+Foxp3-cells in this model are typically CD44 hi , CD25 lo , PD-1 hi , LAG3 hi , and KLRG int . Thus, these cells share most markers previously associated with Tr1 cells (i.e. they are 'Tr1-like'). However, these markers do not distinguish between Tr1-like cells and other IL-10-producing cells in this model.

IL-10 production by Tr1-like cells is transient and wanes after the peak of inflammation
We next sought to determine the kinetics of IL-10 production and airway inflammation in this model. To this end, we analyzed responses at day 2, day 6, day 16 and day 30 after allergic sensitization. For these experiments, the same protocol was used as in Figure 1 but with additional analysis timepoints; a schematic of these is shown in Figure 5A.
Together, these data indicate that Tr1-like cells and not Foxp3+ Treg comprise the majority of IL-10+ T cells in this model. Further, these data suggest that some CD90.1+cells may contribute to the allergen specific memory T-cell pool.
Active IL-10 production in Tr1 like cells is associated with Irf1 and Batf expression Given the low levels of IL-10 production in CD90.1+ cells 30 days after antigenic challenge ( Figure 5G), we questioned whether CD90.1+ cells require persistent antigenic signals for active IL-10 production. To address this, we isolated CD90.1+ cells from spleens of 10BiT mice and cultured them with or without anti-CD3 and anti-CD28 as described previously (Chihara et al., 2016) Only cells which were activated continued to produce IL-10 after 5 days in cell culture ( Figure 6A,B). Moreover, the viability of cultured cells was severely affected in the absence of TCR stimulation over time ( Figure 6-figure supplement 1).
We found that Irf1 and Batf are upregulated only in cells actively producing IL-10 ( Figure 6C,D). These transcription factors therefore characterize active IL-10 production by Tr1-like cells in this model.

Tr1-like cells are a part of allergen-specific memory in the lungs of previously sensitized animals
We next investigated the role of previous allergen exposure in IL-10 production. For this we established a long-term model of HDM challenge ( Figure 7A). In brief, we sensitized animals to HDM over 2 weeks as before. Following this, the animals were rested for close to 2 months. Then, on day 67 after the first challenge, both PBS and HDM-sensitized animals were challenged with one dose of HDM, and the airway inflammatory response was analyzed 24 hr later. Looking this early after allergen challenge allowed us to better assess the rapid CD4 T cell allergen specific memory responses.
Using this protocol, we found that animals that received primary sensitization with HDM exhibited heightened airway inflammation ( Figure 7B), increased eosinophils ( Figure 7C), and heightened IgE upon rechallenge ( Figure 7D). CD90.1+Foxp3-cells were increased in HDM-sensitized animals ( Figure 7E) while the frequency of CD90.1+Foxp3+ as well as CD90.1-Foxp3+ cells were unchanged. Consequently, the frequency of Foxp3-CD4+ T cells within all IL-10-producing cells was also elevated in HDM-sensitized animals in comparison to control, although non-T cells remained the main producers of IL-10 in both groups ( Figure 7F). These changes were specific to the lung as we did not observe any changes in these subsets in the draining lymph node nor spleens (not shown).
To further define the functional phenotype of these CD90.1+ Foxp3-T cells, we measured cytokine production after ex vivo stimulation. We found that the majority of CD90.1+Foxp3-cells produced IL-10 and fewer cells produced other cytokines such as IL-4, IL-13, IFNg and IL-17 ( Figure 7I, Figure 7-figure supplement 2). Additionally, we found that these Tr1-like cells also expressed low levels of GATA3 and higher levels of Tbet ( Figure 7J, Figure 7-figure supplement 2).
Together, these data indicate that Tr1-like cells arise from the allergen-specific memory response in this model. Further, the tetramer-positive cells that produced IL-10 in this model were Tr1-like cells and not Foxp3+ Treg.

IL-10-producing T cells in the lung can arise from tissue resident memory cells
We next asked whether the Tr1-like cells found upon memory challenge are resident to the lung or circulating. Using the same long-term model of HDM challenge used in Figure 6A, we observed that most CD90.1+ cells exhibited an effector memory phenotype (CD44hi, CD62Llow) as compared to the CD90.1-Foxp3-cells ( Figure 8A). In addition, using intravascular labeling, we found that CD90.1+Foxp3-cells had a higher ratio of resident to circulating T cells ( Figure 8B).
To functionally address tissue residency of these cells, we treated mice sensitized to HDM with the S1P receptor agonist FTY720 to block T cell egress from lymph nodes prior to memory rechallenge ( Figure 8C, Figure 8-figure supplement 1). We found that disruption of lymphocyte egress   did not affect cellular infiltration or eosinophilia in the BAL ( Figure 8D,E). However, the ratio of CD62L+ to CD62L-CD4+ T cells in the lungs was significantly reduced ( Figure 8F). This is in line with a previous report that lung resident CD62L negative memory cells are sufficient to induce airway inflammation (Hondowicz et al., 2016). Notably, the frequency of Tr1-like cells within the CD4 +T cells in the lungs was unaffected ( Figure 8G). Unlike CD90.1-Foxp3-T cells, most of the CD90.1 +Foxp3-Tr1-like cells were consistently effector memory (CD44hi CD62L low) both with and without FTY270 administration ( Figure 8H,I).
To address the memory phenotype of our CD90.1+Foxp3-CD4 T cells in the lung parenchyma, we analyzed lungs from memory mice on day 68 without allergen rechallenge looking for common markers of tissue-residency Schenkel and Masopust, 2014). We find that when compared to unsensitized mice, CD90.1+Foxp3-cells in HDM-sensitized mice show higher frequencies of CD69 expression and lower frequencies of Ly6C expression. There was no difference in CCR7, CD103, IL7Ra or KI67 expression in CD90.1+Foxp3-cells between sensitization conditions. CD90.1-Foxp3-cells showed no difference in expression of any of the markers analyzed between sensitization conditions. (Figure 8-figure supplement 2).
Together, these data indicate that the majority of CD90.1+Foxp3-CD4+ Tr1-like cells following allergen challenge have a memory phenotype and can arise from lung-resident memory cells.

Neither depletion nor adoptive transfer of Tr1-like cells influences tolerogenic responses to allergen re-challenge
To elucidate the functional contribution of Tr1-like cells to allergic inflammation, we next examined their function using depletion experiments. To this end we administered an antibody directed against CD90.1 to deplete IL-10 producing cells, as has been done previously to deplete CD90.1+ cells in this model (Maynard et al., 2007). The strategy was likewise successful here in depleting all CD90.1+ cells including CD90.1+Foxp3-and CD90.1+Foxp3+ T cells from the lungs ( Figure 9A,B). Importantly the antibody clone used for depleting these cells did not mask CD90.1 staining with other clones (Figure 9-figure supplement 1). In addition, it depleted CD90.1+CD3-cells however, IL-10 production by them was unaffected (Figure 9-figure supplement 2).
We first depleted the CD90.1+ cell population during the sensitization phase ( Figure 9A,B). This exacerbated airway inflammation as measured by total cell and eosinophil infiltration in the BAL ( Figure 9C,D) as well as IL-13 production by CD90.1-Foxp3-CD4+T cells ( Figure 9E).
We next depleted the CD90.1+ cell population during the memory phase ( Figure 9F,G). This did not impact the inflammatory response ( Figure 9H,I), cytokine production by CD90.1-Foxp3-cells ( Figure 9J), nor the frequency of tetramer + cells within CD4+T cells in the lungs ( Figure 9K).
We also performed intra-tracheal adoptive transfer of memory CD90.1+ Tr1-like cells or CD90.1-T effector cells from lungs of memory rechallenged mice sensitized to HDM or an unrelated allergen (ovalbumin) into naïve recipients who were subsequently challenged with HDM. Consistent with our observations in the depletion studies, we did not see any beneficial impact of the transfer of memory Tr1-like cells on cellular infiltration or eosinophilia in the BAL ( Figure 10A-C). Moreover, the transferred CD90.1+ T cells were not more likely than transferred CD90.1-T cells to produce IL-10 upon rechallenge with allergen ( Figure 10D) and did not alter the levels of IL-10 in the lung ( Figure 10E). In contrast, CD90.1+Foxp3-cells could suppress cell infiltration in the BAL when transferred in an acute model and were also suppressive in vitro (Figure 10-figure supplement 1). This suggests that CD90.1+Foxp3-cells are unable to suppress the allergen specific inflammatory memory responses. An inherent problem of this model is that neither CD90.1+ nor CD90.1 negative cells engraft efficiently in the lungs (Figure 10-figure supplement 2).   Together these data indicate that during primary allergen challenge CD90.1+ cells, the majority of which are Tr1-like cells but also include some CD90.1+Foxp3+, promote immune tolerance. However, endogenous Tr1-like cells do not contribute to tolerogenic memory in this model.

Discussion
We have investigated the phenotypic stability and contributions to tolerogenic memory of endogenous Tr1-like cells in a mouse model of allergic airway inflammation-induced asthma. We report that natural Tr1-like cells only transiently express IL-10 after activation. Moreover, while cells that actively produce IL-10 are important for immune tolerance to airway allergens, neither depletion nor transfer of Tr1-like cells altered airway inflammation upon subsequent allergen challenge. Together these data suggest that naturally-arising Tr1-like cells may promote tolerance but do not contribute to a functionally stable tolerogenic memory population in this model. Persistent antigenic signals were required for maintenance of IL-10 production and expression of Irf1 and Batf, transcription factors previously linked to Tr1 status. It may be that repeated stimulation, a feature of many immunotherapy regimens, maintains active IL-10 production in airway Tr1like cells by supporting expression of these canonical Tr1 transcription factors. In the gut, a prominent site of IL-10 production, antigen stimulation by commensal bacteria may provide a similar function (Gagliani et al., 2013;Gagliani et al., 2015;Yu et al., 2017).
In light of these data, we propose that IL-10 production is a temporary phenotype that a fraction of memory T cells adopt upon activation in the setting of certain previously defined differentiation signals Huang et al., 2017;Gregori et al., 2010;Li and Flavell, 2008;Coomes et al., 2017;Brockmann et al., 2017). Furthermore, IL-10 production in response to allergen exposure may be stochastic and not necessarily predicated upon past production of IL-10. This model is perhaps consistent with past reports of phenotypic plasticity in induced Foxp3+ regulatory T cells (Joetham et al., 2017;Hwang et al., 2018), and between T-helper subsets in general (Zhu and Paul, 2010). Notably, the data presented here involving naturally arising Tr1-like cells stands in contrast to more phenotypically stable Tr1-like cells that have been engineered or differentiated in vitro (Gregori and Roncarolo, 2018).
Natural Tr1-like cells arise from the memory T-cell subset of mice previously sensitized to HDM. During the acute response, there is a significant increase in the frequency of CD90.1+Foxp3-, Tr1like cells in the lungs and airways, the site of allergen challenge. After cessation of IL-10 production, these cells persisted over time in the lungs of challenged animals as tissue-resident memory cells. This is consistent with evidence that an immune reaction to an antigen can contribute to subsequent regulatory memory responses (Brincks et al., 2013;Rosenblum et al., 2011;Sanchez et al., 2012;Sanchez Rodriguez et al., 2014;Gratz et al., 2013).
Tr1-like cells in this model arose specifically in the lung, the site of allergen challenge and not in distal compartments such as the spleen. This is analogous to the report from Hondowicz et al. that tissue-resident Th2 cells drive allergic responses (Hondowicz et al., 2016). These data are also consistent with previous reports of tissue-specific roles for Tr1-like cells (though there are indications of geographic plasticity as well; Yu et al., 2017). However, currently the most commonly used routes for allergen-specific immunotherapy (SIT) are subcutaneous and oral. The data presented here  suggest that intranasal routes may be more effective at inducing Tr1-like cells in the lung (Takabayashi et al., 2003).
Tr1-like cells and not Foxp3+ Treg were the major source of IL-10 in the lung in this model. Consistent with this, there was a greater increase in the frequency of Tr1-like cells versus Foxp3+ Tregs in the lung during the peak of inflammation. Unlike Tr1 cells, the majority of CD90.1-Foxp3+ Tregs in the inflamed lung were circulating, non-resident cells and their frequency did not increase upon memory allergen rechallenge.
Finally, Tr1-like cells and not Treg (irrespective of CD90.1+ expression) were also the predominant Derp1:I-A b tetramer-positive IL-10+ cell population in the lungs after secondary rechallenge. One could speculate that Foxp3+ regulatory memory Tregs may regulate inflammatory responses directed against self-rather than non-self-antigens, such as the HDM allergen (Rosenblum et al., 2016;Rosenblum et al., 2011). However, they may be specific for other HDM epitopes not assessed in our study. CD90.1+Foxp3+ cells are fewer in number and their frequency doesn't change significantly upon allergen sensitization. However, like the Tr1-like cells, they are also parenchymal and highly activated expressing CD44, Lag3, PD-1 and KLRG1.Furthermore they are also depleted when we use the CD90.1 antibody. Thus, while Tr1-like cells are the primary source of IL-10 in our model, we cannot rule out their secondary contribution to supressing inflammation in our acute model. Tr1-like cells in our model expressed most of the cell surface markers previously associated with Tr1 cells including CD44, Lag3, PD-1, and KLRG1. However, these markers did not distinguish between Tr1-like cells and CD90.1+Foxp3+ Treg. Other studies have likewise reported that while these Tr1 markers identify a highly suppressive subset, they are not consistently expressed by all IL-10-producing Tr1-like cells (Gagliani et al., 2013;Burton et al., 2014;White and Wraith, 2016). This variation in their phenotypic markers is also consistent with the model that in vivo, natural Tr1like cells may not be functionally stable in the absence of a persistent antigenic stimulus.
These findings may inform efforts to develop Tr1-based tolerogenic therapies. The phenotypic instability reported here may limit efforts to re-establish tolerance by promoting Tr1-based tolerogenic memory in the absence of repeated antigen stimulation. It may prove advantageous to engineer Tr1-like cells in vitro for adoptive transfer (Gregori and Roncarolo, 2018). Insight into the tissue-specific factors required for maintenance of Tr1 function in vivo will help better target both endogenous and engineered Tr1-like cells to induce long-term tolerance in patients.

Isolation of cells from tissues
Tissues were processed as described previously (Gebe et al., 2017). Briefly, bronchoalveolar lavages (BAL) were performed with 1 mL flushes of the lung with sterile PBS containing 0.2% BSA. BAL fluid was treated with ACK RBC lysis buffer (5 min, 37 degrees C) before cell counting and flow cytometry. Following removal of BAL fluid, lungs were perfused with 5 mL sterile PBS injected into the right ventricle. Lungs were removed, placed in RPMI media and cut with scissors to approximately 1 mm pieces before the addition of collagenase IV (Sigma-Aldrich, St. Louis, MO) to 150 U/mL and DNase I (Sigma-Aldrich, St. Louis, MO) to 25 U/mL for digestion. Lung tissue was digested for 37 degrees C for 45 min on a shaker. A single cell suspension was obtained by pressing digested tissue through a 70 mm cell strainer using the plunger of a 3 mL syringe followed by a wash with PBS containing serum and EDTA. Where applicable, spleen and lymph nodes were isolated and separated into a single cell suspension using a 70 mm cell strainer similarly. All tissues were treated with ACK RBC lysis following single cell suspension (5 min, 37 degrees C) for cell counting and flow cytometry.

Blocking lymphocyte migration using FTY720
To block lymphocyte egress from lymph nodes, HDM-sensitized animals were given daily IP injections of FTY720 (Sigma-Aldrich Cat: SML0700-5MG) at 5 mg/kg, estimating 25 g per mouse, or DMSO (sham control) in saline from day 64 to day 68 (Brinkmann et al., 2002).

Depletion of IL-10-producing cells
Depletion of IL-10-producing cells was performed by intraperitoneal injection of 200 ug of anti-Thy1.1 antibody or matched isotype control (BioXCell, West Lebanon, NH). This was done on day eight for the acute model (as depicted in the schematic in Figure 8A) and beginning on day 64 in the memory model (as depicted in the schematic in Figure 8F).  Data availability All data generated or analysed during this study are included in the manuscript and supporting files.