A stromal cell niche sustains ILC2-mediated type-2 conditioning in adipose tissue

Rana et al. demonstrate that white adipose tissue-resident multipotent stromal cells (WAT-MSC) support IL-33– and LFA-1/ICAM-1–mediated proliferation and type-2 cytokine expression by ILC2. Inversely, ILC2-derived IL-4 and IL-13 promote WAT-MSC eotaxin production and eosinophil recruitment, further maintaining type-2 immune homeostasis.

In mice, administration of IL-33 induced ILC2 activation and promoted beiging of adipocytes, a process associated with increased metabolic consumption (Brestoff et al., 2015;Lee et al., 2015). A number of mechanisms appear to underlie these changes. ILC2s were shown to produce the endogenous opioid peptide Met-enkephalin that may contribute to regulating obesity (Brestoff et al., 2015). Others reported that ILC2, along with eosinophils (Wu et al., 2011), produce IL-4 and/or IL-13, which directly promote beiging of adipocyte progenitors (Lee et al., 2015). IL-5-producing ILC2s were also required to sustain IL-4-secreting eosinophils in visceral adipose tissue (Molofsky et al., 2013), and mice lacking eosinophils gained more weight (Wu et al., 2011). These studies support key roles for IL-33, ILC2, and eosinophils in regulating a lean phenotype.
Here, we aim to clarify the stromally elicited signals sustaining a type-2 immune microenvironment in healthy adipose tissue homeostasis by interrogating the mechanisms by which multipotent stromal cells (MSCs) and ILC2s communicate within this niche.
WAT stromal cells have a gene expression profile characteristic of MSCs, which have the ability to differentiate into multiple cell lineages (Cawthorn et al., 2012;Farahani and Xaymardan, 2015). Indeed, PDGFRα + Il33-citrine +orstromal cells cultured in vitro with defined adipogenic factors readily differentiated into LipidTOX Red-staining lipid-storing mature adipocytes ( Fig. 1 J and data not shown), similar to a positive control constituting the SVF from perigonadal adipose tissue ( Fig. 1 J). Furthermore, PDGFRα + Il33-citrine + WAT stromal cells also possessed myogenic potential when cultured in vitro with PDGF and TGFβ, giving rise to proliferating 5-ethynyl-29-deoxyuridine (EdU) + α-smooth muscle actin-positive cells in culture, resembling a positive control from aorta adventitia ( Fig. 1 K). Consequently, we subsequently refer to these cells as WAT-MSCs. Recently, IL-33 expression has been reported in stromal cells in multiple tissues (Heredia et al., 2013;Kuswanto et al., 2016;Dalmas et al., 2017;Dahlgren et al., 2019;Mahlakõiv et al., 2019;Spallanzani et al., 2019), suggesting that similar cells may represent an important source of IL-33 for maintaining homeostasis and repair throughout the body.
Given the conserved phenotype and gene expression of Il33citrine +and-WAT-MSCs, we subsequently used total WAT-MSCs, independent of their Il33-citrine expression, to assess their potential to regulate ILC2 biology. We found that co-culture of WAT-MSCs with naive WAT-ILC2s (Fig. S3, A and B), induced ILC2 proliferation (Fig. 2, A and B) and IL-5 expression, as compared with cultures lacking WAT-MSCs (Fig. 2, C and D). Unexpectedly, these effects were predominantly IL-33-independent, as ILC2 proliferation was largely unchanged when IL-33-deficient WAT-MSCs were used in this assay (Fig. 2, A-D). We also excluded a potential role for contaminating MSCs by confirming the result using both IL-33-deficient ILC2 and MSCs (Fig. S3 C). Nevertheless, we did observe that MSCderived IL-33 was required to elicit up-regulation of the ILC2 activation marker KLRG1 in two of three experiments, but not GATA3 (Fig. 2, E and F). It has been proposed that cell stress and damage are required to trigger IL-33 release in vivo (Moussion et al., 2008;Scott et al., 2018). Consequently, to replicate in situ cell stress/damage, the SVF from WT or IL-33-deficient mice was freeze-thawed to induce IL-33 release ( Fig. 2 G), and supernatants were added to ILC2 cultures. Supernatants from IL-33-deficient SVF were less efficient at stimulating ILC2 proliferation ( Fig. 2 H), KLRG1 up-regulation ( Fig. 2 I), and GATA3 expression (Fig. 2 J) than WT samples. Taken together, these results suggest that WAT-MSCs express IL-33 in healthy WAT and are capable of releasing IL-33 release upon cellular stress or damage. Indeed, a recent report has demonstrated that absence of IL-33 from PDGFRα + stromal cells reduced the type-2 response in the lung following helminth infection (Dahlgren et al., 2019).
However, even in the absence of IL-33, we observed that MSCs provided additional signals to support a type-2 immune microenvironment, results supported by the fact that depletion of IL-33 + MSCs, even with supplementation of exogenous IL-33, causes a reduction in the proportion of ILC2 in the lung (Dahlgren et al., 2019). To identify the IL-33-independent pathway through which WAT-MSCs regulate ILC2 proliferation, we determined if the factor was secreted or cell-bound. Time-course analysis confirmed the enhanced proliferation of ILC2 co-cultured in the presence of MSCs over 5 d, accompanied by the up-regulation of the activation marker KLRG1 (Fig. 3 A).
Co-culture of purified WAT-MSCs with ILC2 resulted in increased total cell counts ( Fig. 3 B), which were ablated when the ILC2 and MSCs were separated using transwells (Fig. 3 B), though in part, this was due to the reduced survival of ILC2 on the transwell filter ( Fig. S3 D). However, we also observed enhanced ILC2 proliferation, as determined by cell trace violet (CTV) staining ( Fig. 3 C), which was reversed in the transwell culture, and was independent of the substrate on which they were cultured (Fig. 3 C). Similarly, MSC-induced, ILC2-derived IL-5 production was also reduced when the cells were separated in transwell assays (Fig. 3 D).
To isolate the cell surface-associated proliferative/maintenance signal, we interrogated WAT-MSC and WAT-ILC2 gene expression data for potential ligand/counter-ligand pairs. CRISPR-Cas9-mediated candidate gene deletion was performed in MSCs derived from Cas9-transgenic mice using lentiviral vectors carrying target gene-specific guide RNAs (gRNAs; Fig. 4 A). With this approach, we determined that intercellular adhesion molecule-1 (ICAM-1) was expressed on WAT-MSCs while its counter-ligand LFA antigen-1 (LFA-1) was present on WAT-ILC2s (Fig. 4 B). ICAM-1 and LFA-1 facilitate lymphocyte accumulation at sites of immune activation where ICAM-1 is up-regulated on stromal cells (Verma and Kelleher, 2017;Walling and Kim, 2018) but can also initiate LFA-1 signaling leading to T cell activation and proliferation (Verma and Kelleher, 2017;Walling and Kim, 2018) and promote Th1 cell differentiation of human T cells (Smits et al., 2002). In addition, LFA-1 ligation has been described to mediate TCR-independent signals, such as respiratory burst induction in neutrophils, suggesting that LFA-1 can serve as an activating receptor in cells by itself (Berton et al., 1992). ICAM-1 was efficiently deleted from MSCs ( Fig. 4 C). Notably, co-culture of ICAM-1 gRNAtargeted MSCs with WT ILC2 resulted in impaired ILC2 proliferation, reduced KLRG1 expression, and reduced IL-5 production as compared with controls ( Fig. 4 D). To validate this ICAM-1-mediated interaction, we assessed the involvement of the reciprocal LFA-1 ligand by CRISPR-Cas9-targeting in ILC2s.
We next examined the frequency of ILC2 in WAT from naive LFA-1-deficient (Itgal −/− ) mice and found that ILC2 numbers and frequencies were significantly lower in the absence of LFA-1 (Fig. 4 . Thus, the absence of LFA-1 in vivo leads to a diminution of ILC2s in adipose tissue, and this correlated with fewer eosinophils. Our data suggest that inefficient ILC2 proliferation may contribute to the observed ILC2 deficit, but selective cellular adhesion may also be important for localization and retention of ILC2 in WAT. Icam1 −/− mice have been reported to spontaneously gain weight when maintained on normal diet (Dong et al., 1997), and it is possible, given our data, that the interaction of ICAM-1 with LFA-1 on ILC2s contributes to this phenotype. Interestingly, a recent report has (n = 3). (J) Adipose differentiation determined by lipid droplet analysis. Representative of two experiments. Scale bars, 100 µm. (K) Myocyte differentiation determined by α-smooth muscle actin (αSMA) staining. Representative of three experiments. Scale bars, 100 µm. Data are represented as mean ± SEM. Max, maximum. indicated that signaling via ICAM-1 expressed on common lymphoid progenitors promotes their differentiation to ILC2s, and may contribute to the proliferation of ICAM-1 + ILC2 in response to IL-33 (Lei et al., 2018). Furthermore, stromal cells in the mesentery supported ILC2 differentiation from ILC precursors (Koga et al., 2018). However, although we also detected ICAM-1 on ILC2, the effect on proliferation that we observed was dependent on ICAM-1 present on the MSCs and LFA-1 expression by ILC2s. In combination, these findings indicate the importance of future studies to understand how the LFA-1 and ICAM-1 signaling pathways, such as calcium flux and PI3K activation, modulate ILC2 function (Berton et al., 1992;Verma and Kelleher, 2017).
Intriguingly, our RNA sequencing (RNA-seq) data, concordant with a recently published report (Dahlgren et al., 2019), identified that WAT-MSCs highly expressed the eosinophil chemo-attractant CCL11 (eotaxin; Fig. 1 I and Fig. S3 F), but not CCL24 (eotaxin-2) or CCL26 (eotaxin-3; data not shown), responsible for eosinophil migration to peripheral tissues (Wen and Rothenberg, 2016). This raised the possibility that MSCs might contribute to eosinophil recruitment to adipose tissue and that this may be regulated by ILC2s. For example, eotaxin is upregulated through IL-4 receptor (IL-4Rα) signaling (Moore et al., 2002;Zhou et al., 2012;Lee et al., 2015;Gieseck et al., 2016), which is a component of the IL-4 and IL-13 receptor complexes, and is expressed by WAT-MSCs ( Fig. 1 I and Fig. 5 A). Culture of WAT-MSCs in vitro with IL-4 or IL-13 induced eotaxin expression (Fig. 5, B and C). Furthermore, ILC2 cell-culture supernatant also drove eotaxin expression from WAT-MSCs, as compared with control supernatant (Fig. 5 C). This effect was dependent on WAT-ILC2-derived IL-4/IL-13 as ILC2s from Il4 −/− Il13 −/− BALB/c mice failed to induce eotaxin secretion, as compared with BALB/c WT controls (Fig. 5 D). CRISPR-mediated IL-4Rα-gRNA targeted deletion of IL-4Rα on MSCs confirmed the requirement for IL-4Rα-mediated signaling (Fig. 5, E and F). We also investigated if IL-4 and IL-13 could regulate Il33-citrine expression by WAT-MSCs. Indeed, IL-4 and IL-13 treatment increased IL-33 expression (data not shown). Thus, ILC2secreted IL-4 and IL-13 can prime the expression of the eosinophil-attracting chemokine, eotaxin, from WAT-MSCs and may provide a potential positive feedback loop for IL-33 expression from WAT-MSCs.
To evaluate whether the interaction of WAT-MSCs with IL-4/ IL-13-producing ILC2 mediates eosinophil recruitment in vivo, WAT-MSCs, WT ILC2, or IL-4/IL-13-deficient ILC2s were purified, labeled with CTV, and co-cultured overnight, combined with Matrigel, and transferred subcutaneously into naive WT recipients (Fig. 5 G). After 48 h, the Matrigel plugs were isolated, and flow cytometry analysis revealed that eosinophils were increased when MSCs and WT ILC2 were cotransferred as compared with Matrigel alone (Fig. 5 H and Fig. S3 G). Notably, eosinophil recruitment was not observed when IL-4/IL-13- deficient ILC2s were combined with WAT-MSCs (Fig. 5 H), correlating with the absence of eotaxin (Fig. S3 H). IL-5 is known to promote eosinophil retention and survival in peripheral tissues, as well as acting as a growth and differentiation factor for eosinophil development in the bone marrow (Wen and Rothenberg, 2016), and maintaining eosinophils in adipose tissue (Molofsky et al., 2013). However, IL-5 is unlikely to influence these Matrigel assays as both WT and IL-4/IL-13-null ILC2s  produce similar levels of IL-5 (data not shown). Together, these data highlight the synergy between ILC2s and WAT-MSCs and indicate the potential for ILC2-derived IL-4/IL-13-dependent induction of eotaxin production from MSCs to enhance eosinophil recruitment. IL-4/IL-13-deficient mice have been reported to have normal numbers of WAT eosinophils (Molofsky et al., 2013), though these mice do show eosinophil deficits in other tissues, especially upon immune challenge (Fallon et al., 2002). Furthermore, others have indicated that such a network may play a role in recruiting eosinophils to the liver during fibrosis (Gieseck et al., 2016). Further studies will be required to determine the potential role of this pathway in vivo during homeostasis and during dysregulation, and to understand the interplay between ICAM-1-LFA-1 signals and ILC2-derived IL-4/IL-13.
In summary, we show that adipose MSCs can help sustain the type-2 microenvironment by supporting the proliferation, activation, and type-2 cytokine production of ILC2, and by responding to ILC2-derived IL-4 and IL-13, up-regulate eotaxin to promote eosinophil recruitment. Thus, MSCs can provide a niche for the maintenance of type-2 immune homeostasis in adipose tissue.

RNA-seq
Cells were flow sorted into in RPMI-1640, 10% FCS plus penicillin/streptomycin, 2-mercaptoethanol and transferred to Trizol (Life Technologies). RNA was extracted using RNeasy kits and reagents (Qiagen). DNA was digested using Turbo DNase (Ambion). RNA was concentrated using an RNeasy Micro Kit (Qiagen) and assessed using a Bioanalyser (Agilent). RNA was processed for RNA-sequencing using an Ovation RNA-seq System V2 (Nugen), fragmented using the Covaris M220, and bar-coded using Ovation Ultralow Library Systems (Nugen). Samples were sequenced using an Illumina Hiseq4000 (Cancer Research UK Cambridge Institute), and sequence data were aligned using Tophat2 with Partek Flow. RNA-seq analysis was performed using Partek Genomics Suite software, version 6.16.
For adipocyte differentiation assays, total SVF from perigonadal fat was dispersed by digestion with collagenase type II before WAT-IL-33-citrine + MSCs were isolated by flow sorting. Cells were grown to confluence in growth media, high-glucose DMEM (glucose 25 mM; Sigma-Aldrich) and 1% penicillin/ streptomycin (Sigma-Aldrich) with 10% neonatal calf serum. For adipose cell differentiation, cells were cultured for a further 10 d in growth media with 10% FBS supplemented with 3-isobutyl-1-methylxanthine (0.5 mM), dexamethasone (Sigma-Aldrich), and insulin (Sigma-Aldrich). LipidTOX Red staining (Thermo Fisher Scientific) was used to identify lipid droplets, and cultures were imaged using an EVOS Cell image system (Thermo Fisher Scientific).
For smooth muscle differentiation assays, cells were plated at 2,600 cells/cm 2 and cultured with platelet-derived growth factor (PDGF)-BB (10 ng/ml; Peprotech) and TGFβ1 (2 ng/ml; Peprotech) in DMEM supplemented with 10% FCS, glutamine, penicillin, and streptomycin for 7 d. Growth factors were replenished every 2 d. Cells from aortic adventitia were treated in parallel as a positive control. To measure proliferation, EdU was added to the media 4 h before fixation in 4% formaldehyde. EdU detection was performed using the Click-it Plus 647 kit (c10640; Life Technologies) following the manufacturer's instructions and differentiation assessed by staining with Alexa Fluor 594conjugated anti-α-smooth muscle actin antibody (ab202368; Abcam) or control IgG (ab202368; Abcam).
Ligand screen using CRISPR-Cas9 targeting Lentiviral vectors were generated by transfecting HEK293(t) cells with TransIT-LT1 transfection reagent (Cambridge Bioscience) along with the plasmids psPAX2, pMD2.G, and lentiGuide-Puro (Addgene) that had been modified to express BFP and individual targeting gRNA (or random controls), designed according to the Broad GPP genome-wide Brie Library (Addgene). Lentivirus-containing supernatants were harvested and lentiviral particles concentrated using Lenti-X (Clontech). MSCs derived from the Cas9-expressing Rosa26-Cas9 knockin mice were plated on a flat-bottom 96-well plate (10 4 /well), and after 24 h transduced with respective lentiviral constructs. At day 6 of culture, purified ILC2s (5 × 10 3 ) were added to the wells and cultured for a further 4 d (with IL-7 as described in In vitro cell culture and differentiation assays). At day 10 of culture, MSCs and ILC2 were analyzed by flow cytometry, and cell supernatants harvested for IL-5 ELISA.
Purified ILC2 from Cas9-expressing Rosa26-Cas9 knockin mice were plated on a round-bottom 96-well plate at 5 × 10 3 ILC2 per well, and cultured for 2 d in complete RPMI (RPMI-1640, 10% FCS plus penicillin/streptomycin, 2-mercaptoethanol), and rmIL-7 and rmIL-33 (both interleukins at 10 ng/ml; BioLegend) before being transduced with respective lentiviral constructs. 4 d after transduction, Lin -CD45 + BFP-positive ILC2s were purified using flow cytometry and plated on roundbottom 96-well plates, 5 × 10 3 ILC2 per well. Target gene knockout was confirmed at this point with an aliquot of retrovirally transduced BFP-positive ILC2. Transduced ILC2s were then cultured for 3 d in complete RPMI with rmIL-7, but without rmIL-33 before co-culturing with MSCs for a further 4 d. MSCs (separately purified biological replicates) were flowcytometrically sorted and plated in a flat-bottom 96-well plate 4 d before co-culturing with either target gene knockout or control transduced ILC2s. ILC2s were then harvested and analyzed by flow cytometry.

Statistical analysis
Graphpad Prism software was used for all statistical analysis. RNA-seq analysis was performed using Partek Genomics Suite software, version 6.16.

Data availability
The accession nos. for RNA sequencing datasets reported in this paper have been deposited with the Gene Expression Omnibus under accession no. GSE132738. Fig. S1 shows the ILC2 and MSC populations in the WATs. Fig. S2 shows gating strategies. Fig. S3 shows gating strategies and additional control experiments.