Human primed ILCPs support endothelial activation through NF-κB signaling

Innate lymphoid cells (ILCs) represent the most recently identified subset of effector lymphocytes, with key roles in the orchestration of early immune responses. Despite their established involvement in the pathogenesis of many inflammatory disorders, the role of ILCs in cancer remains poorly defined. Here we assessed whether human ILCs can actively interact with the endothelium to promote tumor growth control, favoring immune cell adhesion. We show that, among all ILC subsets, ILCPs elicited the strongest upregulation of adhesion molecules in endothelial cells (ECs) in vitro, mainly in a contact-dependent manner through the tumor necrosis factor receptor- and RANK-dependent engagement of the NF-κB pathway. Moreover, the ILCP-mediated activation of the ECs resulted to be functional by fostering the adhesion of other innate and adaptive immune cells. Interestingly, pre-exposure of ILCPs to human tumor cell lines strongly impaired this capacity. Hence, the ILCP–EC interaction might represent an attractive target to regulate the immune cell trafficking to tumor sites and, therefore, the establishment of an anti-tumor immune response.


Introduction
Innate lymphoid cells (ILCs) constitute the latest described family of innate lymphocytes with key functions in the preservation of epithelial integrity and tissue immunity throughout the body (Mjö sberg and Spits, 2016). Besides conventional natural killer (cNK) cells, three main distinct subsets of non-NK helper-like ILCs have been described so far, mirroring the transcriptional and functional phenotype of CD4 + T helper (Th) cell subsets (Diefenbach et al., 2014): ILC1s, ILC2s, and ILC3s, that mainly produce IFN-g, IL-4/IL-5/IL-13, and IL-17A/IL-22 respectively (Mjö sberg and Spits, 2016).
In human tissues, the majority of ILCs is mainly terminally differentiated, while a population of circulating Lin -CD127 + CD117 + CRTH2 À ILCs, able to differentiate into all ILC subsets, has been recently identified in the periphery and named ILC precursors (ILCPs, Lim et al., 2017). ILCPs are characterized by the expression of CD62L that drives their migration to the lymph nodes (Bar-Ephraim et al., 2019). Enriched at surface barriers, ILCs rely on IL-7 for their development and circulating ILC3s (Shikhagaie et al., 2017) and was not upregulated after in vitro expansion. Compared to ex vivo ILCPs, in vitro-expanded ILCPs downregulated the expression of CD28, although only 20% of circulating ILCPs expressed it (Figure 1-figure supplement 1c and d). Overall, these data suggest that not only in vitro-expanded ILCPs acquire an activated phenotype in vitro, but are also skewed toward an ILC3-like phenotype and share some phenotypical markers with LTi-like cells, while maintaining multipotent features as shown by the expression of T-bet and GATA3. Circulating human ILCs are identified as lineage negative CD127 + cells; within this population, we discriminate ILC1s as c-Kit -CRTH2 -, ILC2s as CRTH2 + c-Kit +/-, and ILCPs as c-Kit + CRTH2cells. HUVEC cells were co-cultured for 3 hr at 1:1 ratio in direct contact with either in vitro-expanded (b) or directly ex vivo-sorted (c) ILC1s, ILC2s, and ILCPs. Untreated ECs were employed as negative control (CTRL). ECs were harvested and analyzed for cell-surface adhesion molecule expression by flow cytometry. Graphs show representative histograms (panels b and c, top) and the summary (panels b and c, bottom) of the induction of the indicated adhesion molecules on the EC surface (n = 6). Ordinary one-way ANOVA-Tukey's multiple comparison test (panel b); Ordinary one-way ANOVA-Friedman test (panel c).
The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Raw data of panels b and c.    To understand if the ability of ILCPs to interact with ECs is an intrinsic property of these cells or if they need to be primed to acquire it, we decided to expose ECs directly to ex vivo-sorted ILC subsets. As shown in Figure 1c, none of the isolated ILC subsets could induce a significant activation of ECs, suggesting that the EC-activating capacity of ILCPs is acquired during the in vitro expansion process. Since ILCPs were expanded in the presence of feeder cells, PHA, and IL-2, it is conceivable that feeder-derived cytokines such as IL-12 and IL-1b are involved in the priming. As ILCs constitute the innate counterpart of CD4 + T cells, we tested if in vitro-expanded individual T-helper (Th) subsets, i.e., Th1, Th2, Th17, and Th* (i.e., Th cells with a Th1/Th17 intermediate phenotype [Sallusto, 2016; Figure 1-figure supplement 2a]) could also interact, at steady state, with ECs. Following the same expansion protocol employed for ex vivo-isolated ILC subsets, Th subsets were employed in 3 hr co-culture experiments with ECs. As reported in the Figure 1-figure supplement 2b, except for a statistically significant Th1-mediated upregulation of VCAM-1, still not to the same extent as the ILCP-mediated induction, all Th subsets failed to upregulate adhesion molecule expression on the EC surface. Overall, these data suggest that in vitro-expanded ILCPs not only acquire a more activated/ILC3-like phenotype in vitro, but also the ability of interacting with ECs by means of mediating the upregulation of adhesion molecule expression on the EC surface.

ILCPs activate ECs primarily in a contact-dependent mechanism
Inflammation triggers the upregulation of adhesion molecules in ECs, promoting the accumulation of leukocytes and their adhesion to the blood vessel walls. This phenomenon is mediated by proinflammatory mediators, such as TNF and IL-1b (Collins et al., 1995). As a consequence, to discriminate whether the EC activation by ILCPs was due to contact-dependent or soluble factor(s)-dependent mechanism(s), supernatants from the EC/ILCP co-cultures were analyzed. Significantly higher levels of IL-6, IL-8, GM-CSF, and IFN-g were observed (Figure 2a). To address which cell type was producing the pro-inflammatory cytokines that accumulate in the cell-free supernatants, qPCR analysis of ECs and ILCPs (CD31-based FACS-sorted after 3 hr co-culture) was performed and compared to untreated ECs and steady-state ILCPs. As reported in Figure 2b, high levels of IL-6 and IL-8 transcripts were found in ECs exposed to ILCPs, whereas TNF transcripts were high only in steady-state ILCPs, indicating that IL-6 and IL-8 measured in the supernatant (Figure 2a) derive from ECs, and TNF from ILCPs. GM-CSF and IFN-g transcripts were observed in both ECs and ILCPs before and after co-culture, indicating that both cell types contribute to the accumulation of these two cytokines in the supernatant. To experimentally verify if the upregulation of adhesion molecules in ECs was dependent on these soluble factors, 0.4 mm pore transwell chambers were employed, to allow cytokine exchange between the two compartments yet avoiding the cell contact. In this context, ILCPs failed to induce the expression of adhesion molecules on EC surface ( Figure 2c). Of note, the production of the pro-inflammatory cytokines was dramatically reduced in the presence of the transwell insert ( Figure 2-figure supplement 1a). To further prove the direct contact-dependency of the EC-ILC interaction, ECs were incubated during 3 hr in the presence of cell-free supernatant collected from previous EC-ILCP co-culture. As reported in Figure 2d, cell-free supernatant did not lead to the upregulation of the adhesion molecules E-Selectin and VCAM-1 in ECs, although ICAM-1 levels were found to be significantly increased if compared to unstimulated ECs, yet not to the same extent as for ILCP-exposed ECs. Finally, we analyzed the production of IL-6, IL-8, TNF, GM-CSF, and IFN-g by ex vivo-and in vitro-expanded ILCPs. As shown in Figure 2-figure supplement 1b, no difference in terms of secretion of the indicated cytokines was observed. Indeed, incubation of ECs during 3 hr with cell-free supernatant collected from pure ILCPs at the end of the in vitro expansion did not provoke upregulation of adhesion molecules on EC surface (Figure 2-figure supplement 1c), correlating with the very low amount of the pro-inflammatory cytokines as shown in Figure 2figure supplement 1b. Overall, these data suggest that ILCPs are superior to other ILC subsets in inducing the upregulation of adhesion molecules on ECs and can also favor the release of proinflammatory cytokines, primarily in a contact-dependent manner.

ILCPs engage the NF-kB pathway in ECs
It has been shown that adhesion molecule expression can be induced in ECs during inflammatory responses by the activation of different signaling pathways, among which the NF-kB pathway (Rahman and Fazal, 2011). To test whether the induction of adhesion molecules by ILCPs was dependent on NF-kB, ECs were pre-treated during 1 hr with a IkB kinase (IKK) complex inhibitor (BAY 11-7082, Mori et al., 2002) to specifically prevent NF-kB activation. In this context, ILCPs failed to significantly induce the expression of adhesion molecules on pre-treated ECs (Figure 3a), indicating that ILCPs need to engage the NF-kB pathway to activate ECs in vitro. Similar to what we observed in the context of ILCPs cultured with ECs in the presence of a transwell insert, the prevention of NF-kB activation in ECs led to a significant decrease of IL-6, as well as reduction in IL-8, GM-CSF, and IFN-g secretion (Figure 2-figure supplement 1d). Next, to understand which molecular ECs CTRL ILCP-exposed ECs ILCPs CTRL EC-exposed ILCPs Figure 2. Human innate lymphoid cell precursors (ILCPs) activate ECs primarily in a contact-dependent mechanism in vitro. (a) The supernatant of the 3 hr co-culture experiments between ECs and ILCPs was analyzed for its cytokine contents (n = 4). The composition of the supernatant of ECs in EC growth medium was used as negative control (CTRL). (b) The expression of IL-6, IL-8, GM-CSF, TNF, and IFN-g was analyzed by qPCR in ECs and ILCPs after being cultured for 3 hr at 1:1 ratio and FACS-sorted according to CD31 expression. Untreated ECs and ILCPs were employed as controls (CTRL). (c) HUVEC cells were co-cultured for 3 hr at 1:1 ratio in direct contact with in vitro-expanded ILCPs either in the absence (red dots) or presence (red circles) of a transwell (TW) insert (0.4 mm pore polycarbonate filter) or (d) in the presence of pre-conditioned media coming from previous EC-ILCP 3 hr co-cultures. ECs were harvested and analyzed for cell-surface adhesion molecule expression by flow cytometry (n = 6). The dotted lines indicate the level of average expression of adhesion molecules by unstimulated ECs. Statistical tests used: Unpaired t-test (panels a and d); paired t-test (panel c).
The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Raw data of panels a-d.  players were involved in the ILC-EC cross-talk, we screened ECs and, both ex vivo and in vitroexpanded, ILCPs for the presence on their surface of receptors and ligands, respectively, known to be involved in the NF-kB pathway activation. On one side, we observed that untreated ECs constitutively expressed the lymphotoxin-b receptor (LT-bR), as well as the tumor necrosis factor (TNF) receptors 1 and 2 (TNFR-1 and TNFR-2, respectively), whereas B-cell activating factor receptor (BAFF-R), CD40, and RANK were expressed only at low levels ( Figure 3b). Following stimulation with TNF, CD30 expression became detectable and BAFF-R and RANK expression increased, while CD40 and LT-bR expression remained unchanged ( Figure 3b). On the other side, when looking at extracellular NF-kB activating ligands on ex vivo ILCPs, we observed that they expressed high levels  of the transmembrane form of lymphotoxin (LTa 1 b 2 ), a described ligand for LT-bR (Madge et al., 2008), if compared to in vitro-expanded ILCPs (Figure 3c and d). Both BAFF and CD30L were undetectable and low levels of CD40L and RANKL were observed. In contrast, in vitro-expanded ILCPs upregulated the expression of RANKL and downregulated that of LTa 1 b 2 (Figure 3c and d). It has been reported that pro-inflammatory cytokines, such as interleukin 12 (IL-12), can induce RANKL on human periodontal ligament cells in vitro (Issaranggun Na Ayuthaya et al., 2017). Since it is known that feeder cells can produce a wide array of cytokines, among which IL-1b and IL-12, we decided to test whether RANKL expression might be upregulated by one of these factors. Surprisingly, after 24 hr stimulation of freshly ex vivo isolated ILCPs with IL-1b ( Figure 3e), but not with IL-12 (data not shown), we observed increased expression of RANKL compared to untreated ILCPs. The transmembrane form of TNF (tm-TNF) constitutes another described NF-kB activating ligand. However, the detection of the membrane-bound form of TNF could not be tested due to the lack of a specific antibody. Moreover, the discrimination between the soluble and the membrane forms of TNF at mRNA levels is not possible, since TNF is transcribed (and also translated) as a full-length membrane-bound precursor (Black et al., 1997). However, at the end of the in vitro expansion, ILCPs showed higher levels of TNF transcripts compared to ex vivo ILCPs (data not shown). Overall, these data show that in vitro-expanded ILCPs express TNF, possibly present on the ILCP surface, to in vitro interact with ECs via TNFRs and upregulate RANKL expression, possibly via feeder-cell-derived IL-1b, to engage RANK on ECs.

ILCPs activate ECs via the engagement of TNFR and RANK
To test which of the NF-kB activating molecules was responsible for the upregulation of adhesion molecules on EC surface, a series of blocking experiments using different soluble Fc fusion proteins were performed to prevent the binding of defined ligands to their receptors on ILCPs. Since we observed increased levels of RANKL on in vitro-expanded ILCPs as compared to their ex vivo counterparts (Figure 3e), and higher levels or RANK on ECs following 3 hr co-culture with ILCPs ( . Therefore, we hypothesized a major involvement of tm-TNF in the induction of adhesion molecules. Thus, we pre-incubated ILCPs in the presence of TNFR1:Fc and/or TNFR2:Fc and we observed that the EC expression of adhesion molecules was significantly reduced ( Figure 4b). In all cases, inhibition with TNFR2:Fc was slightly more efficient than with TNFR1:Fc, which could be explained by the greater affinity of TNFR2 for TNF (Grell et al., 1995). Of note, no difference in the cytokine secreted levels was observed (Figure 4-figure supplement 1d), suggesting that interfering with the TNF-TNFR signaling does not impact cytokine production in both cell types. Addition of RANK:Fc to TNFR1:Fc and TNFR2:Fc further slightly reduced E-Selectin, ICAM-1 and VCAM-1 levels, although the contribution of RANK:Fc was not significant ( Figure 4c). However, we could observe a decreased production of the pro-inflammatory cytokine IL-6, IL-8, TNF, and GM-CSF (Figure 4-figure supplement 1e) when blocking ligands of TNFR1, TNFR2, and RANK in ILCPs/ECs co-cultures. Taken together, our data suggest that ILCPs activate EC primarily through the engagement of TNFRs to upregulate adhesion molecules expression on EC surface. The engagement of RANK in ECs does not seem to have an additive effect in inducing adhesion molecules expression, but might act in synergy with tm-TNF to control the cytokine secretion and further support the EC activation.

ILCP-mediated EC activation favors the adhesion of freshly isolated PBMCs in vitro
To address the functionality of the EC-ILCP interaction, i.e., the adhesion of freshly isolated PBMCs to ILCP-exposed EC, a static adhesion assay was performed. Briefly, following the 3 hr co-cultures, CD31 + ECs were isolated by FACS, to remove adherent ILCPs, and re-plated. After the sorting, untreated ECs (negative control) did not upregulate adhesion molecule expression on their cell surface, and ILCP-exposed ECs maintained comparable surface levels of adhesion molecule as before the FACS isolation procedure, showing that the sorting procedure did not affect the activation state of ECs in any of the conditions (Figure 5a). The day after, the assay was performed and ECs, together with adherent PBMCs, were detached and stained for flow cytometry analyses. Interestingly, ECs pre-exposed to ILCPs led to the adhesion of a significantly higher number of freshly isolated PBMCs compared to unstimulated ECs. As shown in Figure 5b and c, the ILCP modification of EC allowed a strong adhesion of T, B as well as NK cells and monocytes. To understand if the adhesion of freshly isolated PBMCs is itself dependent on NF-kB, we repeated the experiment by exposing untreated or NF-kB-inhibited ECs to TNF for 3 hr the day before performing the static adhesion assay. As shown in Figure 5d, the inhibition of NF-kB activation prior stimulation with TNF strongly reduced the numbers of adhered T, B, NK cells, and monocytes. In this setting, we could also observe that ILCs themselves could adhere to TNF-treated ECs ( Figure 5d). Interestingly, a trend for a reduction in the number of adhered PBMCs to ECs was also observed when NF-kB activation was prevented in ECs 30 min before performing the static adhesion assay (Figure 5d) although not significant. Since we showed that NF-kB engagement is crucial for the ILCP-mediated adhesion molecule upregulation in ECs (Figure 3a), it was not surprising to observe the impaired adhesion of PBMCs to ECs in vitro. Overall, these data suggest that the adhesion molecule expression induced by the ILCPs is functional, i.e., it supports the adhesion of other immune cell types to ECs in vitro and relies on NF-kB activation.

Tumor-derived factors impair ILCP ability to activate ECs in vitro
The poorly functional and altered structural organization of the vascular bed has an important impact on tumor progression and affects endothelial-leukocyte interactions (Cedervall et al., 2015). Hence, we were interested in studying the impact that the tumor and/or the tumor microenvironment could exert on ILCPs and, therefore, on their ability to modulate the EC activation. First, we observed that CD3 -RORgt + ILCs are present in low-grade transitional bladder carcinoma in close proximity to CD31 + blood vessels (Figure 6a, panels 1-4) but are barely detected in high-grade bladder carcinoma ( Figure 6a, panels 5-8), suggesting a protective role of RORgt-expressing ILCs, at least at early stage of disease. Interestingly, since we also observed that ILCPs are expanded in the PB of non-muscle-invasive bladder cancer (NMIBC) patients, but reduced in muscle-invasive stage of the The graphs show the level of expression of adhesion molecules by ILCP-exposed ECs after the sorting and before performing the static adhesion assay, compared to untreated ECs (gray). Graphs show representative dot plots (b) and the summary (c) of the number of adhered CD3, CD4, CD8, CD14, CD16, and CD19 expressing cells assessed by flow cytometry with the use of CountBright Absolute Counting Beads (blue gate in the dot plots). (d) The day before the assay, HUVEC cells were cultured for 3 hr in the presence of 20 ng/mL of TNF and treated during 1 hr with 2.5 mM NF-kB inhibitor BAY 11-7082 (Adipogen), either before the TNF treatment (half-full red square dots) or directly on the day of the assay (empty red square dots), before incubation with total PBMCs at 1:4 ratio for 30 min. The graphs show the summary of the number of adhered CD3, CD4, CD8, CD14,, CD56 dim CD16 + , CD56 bright CD16 low , CD19 expressing cells, and ILCs assessed by flow cytometry with the use of CountBright Absolute Counting Beads. Statistical test used: Unpaired t-test (panels c and d).
The online version of this article includes the following source data for figure 5: Source data 1. Raw data of panels c and d.  Figure 6b), and, following in vitro expansion, ILCPs acquire RORgt expression, we hypothesized that the presence of ILCPs in NMIBC patients might underline the attempt of this cell population to support the infiltration of immune cells into the tumor site. To this aim, ILCPs were pre-exposed to human bladder cancer cell lines, originating either from non-muscle-invasive (NMIBC, early stage) or muscle-invasive (MIBC, late-stage) epithelial bladder carcinoma, thus allowing us to mimic in vitro early and late tumor stage conditions. As shown in Figure 6c, the capacity to upregulate adhesion molecule expression on ECs by ILCPs was significantly reduced after the overnight incubation with bladder carcinoma cell lines, if compared to resting ILCPs. Interestingly, the co-culture with MIBC lines showed the highest capacity to modify ILCP ability to activate the ECs ( Figure 6c). Moreover, the analysis of the cytokine composition of the supernatants from 3 hr EC-ILCP co-culture revealed statistically significant reduced levels of the pro-inflammatory cytokines IL-8, GM-CSF, and IFN-g when ECs where co-cultured with MIBC pre-exposed ILCPs (Figure 6d). Similar observations were obtained using tissue sections of colon adenocarcinoma patients and the SW1116 colon cancer cell line ( Figure 6-figure supplement 1a and b). To further understand which could be the mechanisms underlying the tumor cell-mediated impairment of ILCPs, we wondered whether the tumor cells were affecting ILCPs via adenosine and/or kynurenines, two metabolites with potent immune-inhibitory effects in the TME (Vigano et al., 2019;Labadie et al., 2019). As shown in Figure 6e, TCC-Sup did not express IDO-1, suggesting that the tumor-mediated effects on ILCPs might not depend on kynurenines. However, following the overnight incubation with ILCPs, TCC-Sup strongly upregulated CD39 and further increased the CD73 expression ( Figure 6e). Interestingly, steady-state ILCPs also expressed CD39, but very low levels of CD73 ( Figure 6-figure supplement 1c), suggesting that, in the presence of CD73 + cells, ILCPs might process ATP and support adenosine production. Interestingly, as shown in Figure 6-figure supplement 1d, in vitroexpanded ILCPs upregulated the expression of A2A, A2B, and A3 receptors. Of note, pre-exposure of ILCPs to 2-Chloroadenosine (a stabilized form of adenosine) reduced their EC-activating capacity ( Figure 6f). Taken together, these results suggest that tumor cells might impair or deviate, at least in part via adenosine production, the capacity of ILCs to modulate vascular activation through the upregulation of cell surface adhesion molecules, and affect the production of pro-inflammatory cytokines upon EC-ILCP encounter. Therefore, this could represent a mechanism through which tumors can prevent and block immune cell infiltration into the tumor site.

Discussion
In this study, we characterized for the first time the in vitro interaction between circulating human ILCs and vascular ECs. In particular we identify ILCPs as the only competent circulating ILC subset in inducing EC activation through the upregulation of adhesion molecules on the EC surface. Our results are consistent with previously reported data showing that group 3 ILCs (defined as Lin -CD127 + NKp44 + cells) induce the expression of ICAM-1 and VCAM-1 on mesenchymal stem cells (MSCs) after 4-day co-culture and in the presence of IL-7 (Cupedo et al., 2009;Crellin et al., 2010). According to recent findings, circulating ILCPs constitute a distinct subset from ILC3s, although they share the expression of c-Kit on their cell surface and are CRTH2 À (Lim et al., 2017). Following in vitro priming, the upregulation of RORgt and the expression of activation markers argue for a conversion of ILCPs into committed ILC3-like cells, possibly supported by IL-1b and/or other factors secreted by feeder cells during the expansion phase. This environment mimics the in vivo dynamics observed during inflammatory processes driven by PAMP/DAMP/tumor-dependent DC activation. However, differently from what was described by Lim and colleagues (Lim et al., 2017), the in vitro culture of ILCPs isolated from the PB of HDs did not lead to the expansion of neither ILC1s nor ILC2s, whereas only ILC3-like cells arose. Indeed, the in vitro stimulation applied in that context differs from our in vitro expansion protocol, with the lack of ILC1-, ILC2-, or ILC3-specific cytokines. Overall, our findings support the idea that the in vitro expansion of circulating ILCPs in the presence of feeder cells, PHA, and IL-2 favors their commitment toward an ILC3-like phenotype.
Interestingly, as far as adaptive immune cells are concerned, a previous publication showed that freshly isolated CD4 + CD45RO + lymphocytes are able to induce, to different extents, the expression of VCAM-1 on ECs in a contact-dependent manner (Yarwood et al., 2000). We were unable to recapitulate these findings, most probably due to different culture conditions (isolation and in vitro expansion of T cells, timing, and EC:T-cell ratios). For the innate counterpart, it was shown that the human NK cell line NK92 induces the expression of E-selectin and IL-8 in ECs, which results in EC activation, through the LT-dependent activation of the NF-kB pathway (von Albertini et al., 1998). Yet, in our system, we did not observe the upregulation of adhesion molecules when employing in vitro-expanded purified primary NK cells (data not shown).
In this work, we define the ILCP-mediated activation of ECs primarily as a contact-dependent mechanism. However, we cannot exclude that pro-inflammatory cytokines (IL-6, IL-8, GM-CSF, IFN-g, and TNF), produced during the EC-ILCP interaction, might contribute to the observed EC activation. It is known that pro-inflammatory cytokines, and especially TNF, constitute potent inducers of adhesion molecule expression in ECs (Collins et al., 1995). However, in our hands, ECs upregulate adhesion molecules expression only when short-term exposed to TNF, but not to the other cytokines (data not shown). Moreover, the exposure of ECs to cell-free supernatant recovered from foregoing EC-ILCP co-culture only provoked a significant upregulation of ICAM-1 on ECs, and exposure of ECs to cell-free supernatant collected at the end of the in vitro expansion of ILCPs did not upregulate adhesion molecules on ECs, supporting the idea of a primarily contact-mediated interaction between these two cell types. However, it cannot be excluded that the pro-inflammatory cytokines that are produced during the co-culture can support, at the cell-cell contact region, the in vitro cross-talk.
Upon interaction, ILCPs engage the NF-kB pathway in ECs, most probably via TNFR/tm-TNF and RANK/RANKL interactions that possibly act in synergy. RANKL has been recently described as a negative regulator of CCR6 + ILC3s activation and cytokine production, via the paracrine interaction with its receptor RANK (Bando et al., 2018). On one side, we observed that the expression of RANK on ex vivo ILCPs was not detectable, whereas in vitro-expanded ILCPs acquire transient, intermediate levels of RANK after expansion (data not shown). Nevertheless, the contribution of RANKL to EC activation needs further investigation, as well as the formal evaluation of tm-TNF on ILCP surface. Of note, we observed higher levels of transcripts in in vitro-expanded ILCPs compared to their ex vivo counterparts, but very low levels of soluble TNF at the end of the expansion, suggesting that TNF might be present on the surface of expanded ILCPs. We might speculate that a sequential engagement of these ligand-receptor interactions occurs in the EC-ILCP interface, with initial tm-TNF/TNFR interactions that are needed to induce adhesion molecules expression, together with increased RANK expression in ECs. This could facilitate the sequential RANKL/RANK interactions, possibly required to support the production of pro-inflammatory cytokines, since the prevention of both TNFR/RANK engagement resulted in impaired cytokine production.
By performing a static adhesion assay, we show that the ILCP-mediated EC activation is functional. Therefore, ILCPs might favor the initial tethering of circulating immune cells to vascular ECs via E-Selectin induction, and the subsequent ICAM-1/LFA-1 and VCAM-1/VLA-4-mediated firm adhesion step, and support the EC-dependent recruitment of other immune cell types, thus facilitating their exit from the blood stream through the vessel wall.
As previously reported (Eisenring et al., 2010), NKp46 + ILCs were described to be crucial, in a subcutaneous melanoma mouse model, for the establishment of an IL-12-dependent anti-tumor immune response. A similar role was proposed for NKp44 + ILC3s in NSCLC patients (Carrega et al., 2015). Beside their putative role in supporting intratumoral TLS formation, an aspect that has been further recently supported in colorectal cancer patients (Ikeda et al., 2020), these cells were suggested to activate tumor-associated ECs and, in turn, favor leukocyte recruitment. Hence, ILCP-EC interactions might represent an early event during a large spectrum of biological reactions, ranging from inflammation, autoimmunity, and cancer. In tumors, leukocytes have to travel across the vessel wall to infiltrate tumor tissue where they contribute to the killing of cancer cells. Further, the vessel wall serves as a barrier for metastatic tumor cells, and the integrity and the activation status of the endothelium serves as an important defense mechanism against metastasis formation (Cedervall et al., 2015).
The infiltration of immune cells in solid tumors often correlates with a better overall survival in cancer patients (Zhang et al., 2003, Tjin and Luiten, 2014, Mina et al., 2015. However, in the tumor microenvironment, ECs are dysfunctional and play a major role in several processes that contribute to cancer-associated mortality. One mechanism by which ECs can actively discourage the tumor homing of immune cells was described by Buckanovich and colleagues (Buckanovich et al., 2008). By transcriptionally profiling the tumor ECs (TECs) isolated from ovarian cancer specimens poorly infiltrated by T cells, the authors describe a mechanism that relies on the interaction between endothelin B receptor (ET B R), found to be highly expressed by TECs, and its ligand endothelin-1 (ET-1), overexpressed in ovarian cancer cells. ET B R signaling was shown to be responsible for the impaired ICAM-1-dependent T cell homing to tumors, and in turn, it correlated with shorter patient survival. Another mechanism that prevents T cell infiltration into the tumors relies on the overexpression of Fas ligand (FasL) on TEC surface (Motz et al., 2014) that causes the selective killing of tumor-specific CD8 + T cells and to the accumulation of FoxP3 + T regulatory (Tregs) cells within the tumors. Finally, it has been reported that Th1 cells can actively influence vessel normalization processes via the production of IFN-g, which positively correlated with a more favorable outcome for cancer patients (Tian et al., 2017).
Therefore, committed ILCPs might represent an additional key regulator of efficient immune cell penetration into the tumor.
Tumors can engage multiple mechanisms to discourage the establishment of anti-tumor immune responses (Vinay et al., 2015). The shaping of an immunosuppressive milieu together with the diversion of the vascular system supports tumor progression and favors metastatic dissemination (Hida and Maishi, 2018). Here we show that RORgt-expressing ILCs, which share the transcription factor with in vitro-expanded ILCPs, infiltrate both human low-grade bladder and colon cancers and are associated with CD31 + vessels, arguing for a potential ILC-EC interaction also in vivo. In vitro, we observed that the ability of ILCPs to induce adhesion molecules on ECs was dampened after the co-culture with bladder-and colon-derived tumor cells. ILCs are very plastic cells (Bal et al., 2020), and it has been reported that, in the cancer setting, tumor-derived TGF-b drives the transition of NK cells to dysfunctional and pro-tumoral ILC1s in vivo, a novel mechanism exploited by tumors to prevent the establishment of an innate anti-tumor response (Gao et al., 2017). One can speculate that a similar conversion also occurs for ILCPs toward a non-EC activating ILC subset. We showed that the mechanism of impairment of ILCPs might rely on adenosine. In vitro-expanded ILCPs express high mRNA levels of the adenosine receptors and CD39 at the protein level, whereas bladder cancer cells express CD73 and potentially also CD39. The presence of these two ectoenzymes, key for adenosine production, suggest that adenosine might be produced during the co-culture between cancer cells and ILCPs and impact ILCP functions. Indeed, by pre-exposing ILCPs to 2-Chloroadenosine, we could observe reduced EC-activating ability.
In conclusion, our data show that ILCPs, upon proper stimulation, might represent novel players in regulating the trafficking of immune cells to tissues, not only during the early phase of inflammation, but also at early phases of anti-tumor immune responses. Such contact-mediated events may be crucial in supporting further EC activation, to favor tumor-specific T-cell adhesion and, in turn, recruitment to the tumor site.

Co-culture experiments
Following expansion, individual pure (!90%) ILC and Th subsets were rested overnight in RPMI-8% HS medium supplemented with 10 U/mL of rh-IL-2. Then, confluent EC monolayers were either cocultured for 3 hr with individual ILC and Th subsets at 1:1 ratio, treated with 20 ng/mL of rh-TNF (Peprotech) or left untreated as positive and negative controls, respectively. Co-cultures of ECs with ILCPs were performed both in the presence or absence of 0.4 mm pore polycarbonate filter in 24well transwell chambers (Corning). ILCPs were also incubated overnight with epithelial bladder carcinoma cell lines in RPMI 8% HS with 10 U/mL of IL-2 at 1:1 ratio, before exposure to EC monolayers. The day of the experiment, ILCs were collected, washed with PBS, and re-suspended in the respective EC growth medium (Lonza). At least three independent experiments were performed, using individual ILC and Th subsets isolated from a different donor. At the end of the experiment, supernatants were collected and stored at À20˚C, and ECs were washed twice with PBS and detached with Accutase (Gibco) for 5 min at 37˚C. Cell suspensions were then washed with PBS and stained for flow cytometry analyses.

Phenotypic characterization
The phenotypic characterization of both ex vivo-and in vitro-expanded ILCPs from HDs, as well as the quantification of ex vivo ILCPs in the PB of bladder cancer patients, was performed by using the same antibodies as the ones used for isolation by FACS together with the following antibodies:

Static adhesion assay
ECs were plated at 80% confluency in a 24-well plate in complete EGM. Once adherent, the media was removed, ECs were washed with PBS, and 500 mL of complete EGM, containing or not ILCPs at 1:1 ratio, were added to the wells during 3 hr. As positive control, ECs were incubated during 3 hr with 20 ng/mL of TNF. After the co-culture, ECs were detached and stained with FITC anti-CD31 antibody and FACS-sorted to remove adherent ILCPs. Recovered ECs were seeded in a 48-well plate and let to adhere overnight in complete EGM. The morning after, the static adhesion assay was performed (adapted from Safuan et al., 2012). The adhesion of freshly isolated PBMCs was assessed by adding 4:1 cells (PBMC:EC) /well for 30 min at 37˚C. Non-adherent cells were washed away from the EC monolayer by performing 2Â washing steps with PBS. ECs, together with adherent PBMCs, were detached with Accutase (Gibco), and stained for flow cytometry analyses. The number of CD3, CD4, CD8, CD14, CD56 dim CD16 + , CD56 bright CD16 low , and CD19 expressing cells, as well as Lin -CD127 + total ILCs were quantified by flow cytometry by adding 10 mL of CountBright Absolute Counting Beads (Thermo Fisher) to the cell suspensions. 2000 beads/sample were acquired and cell counts normalized.

RNA purification and qPCR
Total RNA was isolated from highly pure ex vivo-and in vitro-expanded ILCPs, from primary ECs (HUVECs) and from sorted human ILC and CD4 Th cell subsets using the TRIZOL reagent according to the manufacturer's instructions (Invitrogen, Carlsbad, CA, USA). Final preparation of RNA was considered protein-free if the ratio of spectrophotometer (NanoDrop, ThermoFischer, Carls-bad, CA, USA) readings at 260/280 nm was !1.7. Isolated mRNA was reverse-transcribed using the iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Watford, UK) according to the manufacturer's protocol. The qPCR was carried out in the ECO Real-time PCR System (Illumina) with specific primers (see Key Resources Table) using KAPA SYBR FAST qPCR Kits (KAPA Biosystems, Inc, MA). Samples were amplified simultaneously in triplicate in one-assay run with a nontemplate control blank for each primer pair to control for contamination or for primer dimerization, and the Ct value for each experimental group was determined. The housekeeping gene (ribosomal protein S16) was used as an internal control to normalize the Ct values, using the 2 ÀDCt formula.

Immunohistochemical staining
Immunohistochemical staining was performed on 2 mm paraffin sections with an automated IHC staining system (Ventana BenchMark ULTRA, Ventana Medical Systems, Italy). Sequential double IHC was performed on Ventana BenchMark ULTRA, using a ultraView Universal DAB detection Kit as the first stain and ultraView Universal Alkaline phosphatase Red detection kit as the second stain. Heatinduced epitope retrieval pre-treatment was performed using CC1 buffer (

Statistical analyses
GraphPad Prism 7 software was used to perform the statistical analyses. Paired or unpaired t-tests were used when comparing two groups. ANOVAs or the non-parametric Kruskal-Wallis test were used for comparison of multiple groups. Data in graphs represent the mean ± SEM, with a p-value <0.05 (two-tailed) being significant and labeled with *. p-values <0.01, <0.001, or <0.0001 are indicated as **, ***, and ****, respectively. Without mention, differences are not statistically significant.