Distinct and shared gene expression for human innate versus adaptive helper lymphoid cells

Abstract Innate lymphoid cells (ILCs) are the latest identified innate immune cell family. Given their similarity in transcription factor expression and cytokine secretion profiles, ILCs have been considered as the innate phenocopy of CD4 Th cells. Here, we explored the transcriptome of circulating human ILC subsets as opposed to CD4 Th cell subsets. We describe transcriptomic differences between total ILCs and total CD4 Th cells, as well as between paired innate and adaptive cell subsets (ILC1 vs. Th1; ILC2 vs. Th2; and ILC3 vs. Th17 cells). In particular, we observed differences in expression of genes involved in cell trafficking such as CCR1, CCR6 and CXCR3, innate activation and inhibitory functions, including CD119, 2B4, TIGIT, and CTLA‐4, and neuropeptide receptors, such as VIPR2. Moreover, we report for the first time on distinct expression of long noncoding RNAs (lncRNAs) in innate vs. adaptive cells, arguing for a potential role of lncRNA in shaping human ILC biology. Altogether, our results point for unique, rather than redundant gene organization in ILCs compared to CD4 Th cells, in regard to kinetics, fine‐tuning and spatial organization of the immune response.


INTRODUCTION
Innate lymphoid cells (ILCs) have recently emerged as a key contributors to host defense and tissue homeostasis, given their capacity to rapidly respond to microenvironmental cues. 1 ILCs are a family of lymphoid cells that, in contrast to adaptive T and B lymphocytes, do not express receptors for antigens. Similar to T cells, ILCs exhibit functional specialization with the ability to produce cytokine in patterns resembling helper CD4 T cells. Helper ILCs are subdivided into 3 main subsets, ILC1s, ILC2s, and ILC3s, according to their master transcription factor expression and cytokine production profiles. ILC1s express Tbet and mainly secrete IFN during infections by intracel-Several bulk and single-cell transcriptomic analyses have been performed in both mouse and human ILCs. In mice, these studies allowed to clarify the developmental positioning of each ILC subset, but also to appreciate the impact of the microenvironment, including the microbiota, in shaping their transcriptional identity. By comparing the gene expression of ILCs and NK cells from different organs, it was shown that unique profiles are acquired by ILC subsets, whereas overlaps in gene expression are present among ILC1s and NK cells. 3,4 In another study, the authors showed a functional subcompartmentalization within the main ILC subsets, mainly within ILC3s, in tissues such as the intestinal mucosa. 5 These properties might be key for ILCs to rapidly integrate and adapt to environmental stimuli.
In humans, a single-cell mRNAseq analysis of tonsil derived ILCs also revealed distinct transcriptional programs in each ILC population and functional sub-specialization within ILC subsets, particularly in tonsilderived ILC3s. 6 Yudanin and colleagues have provided a transcriptomic mapping of ILCs in nondiseased human tissues, highlighting spatial and temporal characteristics of individual ILC subsets. 7 Last, a recent gene expression profiling of human circulating ILCs enabled to identify unique ILC signatures of each individual ILC subset. 8 These advances in our understanding of ILC biology argues for complementary, but also divergent functions of ILCs and CD4 Th cells in host immune responses. In fact, despite the presence of the exact same master transcription factors in mirrored ILC and CD4 Th subsets, functional distinction cannot be excluded. Evidence in that direction comes from studies on ILC3 activity using RAG knockout and wild-type mice where antagonist interactions have been described for gut resident ILC3s and Th17 cells. 9,10 In line with these findings, a distinct temporal activity for ILC3s and their adaptive counterparts was also reported in the context of Citrobacter rodentium infections. In that setting, sequential ILC and Th functions were shown to be necessary for pathogen clearance, the action of T cells contributing to extinguish the early functions of ILC3. 12,13 Further, parallel gene expression profiling and epigenetic analysis of murine ILC and CD4 Th subsets at steady state revealed shared, but also different networks of functional regulators between innate and adaptive cytokine secreting cells, as well as among subsets within the same lineage. Interestingly, regulatory circuits in both lineages are dramatically altered in the context of Type 2 infection models, but at the same time converge to a similar epigenetic signature. Strikingly, ILC regulomes appear to be already poised before cell activation, possibly explaining the ability of ILCs to rapidly respond to infections. In contrast, CD4 Th regulatory elements undergo considerable remodeling during antigen stimulation. Whereas these comparisons have allowed us to revise our view on ILC-CD4 Th cell analogies in model organisms, knowledge about the transcriptomic similarities between human ILCs and CD4 Th cells is still limited. In a study, regulomes of human tonsil-derived ILC1s and ILC3s were compared to the ones of Th1 and Th17 cells, respectively, showing the presence of both unique and overlapping pathways in innate and adaptive mirror cells. 3 However, due to the paucity of ILC2s and Th2s in tonsils, the investigation of these cells was not included in that analysis. Furthermore, no data are available on the comparison of ILCs and CD4 Th cells in the human peripheral blood.
In the current study, we compared gene expression profiles of human circulating helper ILCs and CD4 Th cells. We show transcriptomic differences in expression of genes involved in cell trafficking, innate activation, and inhibitory functions, supporting distinct temporal and spatial activation of ILCs and Th cells in vivo.
Moreover, we report on distinct expression of long noncoding RNAs (lncRNAs) in innate vs. adaptive cells, arguing for a subtle and different cellular fine-tuning of human ILCs as compared to their adaptive counterparts.

Cell preparation
Buffy coats were obtained from healthy donors at the local Blood Transfusion Center, Lausanne, Switzerland. PBMCs were isolated by density-gradient centrifugation and immediately used.

ILC and CD4 Th cell evaluation by flow cytometry
Human total ILCs and ILC subsets were identified using lineage markers, all FITC conjugated, that include: anti-human CD3 (UCHT1, Beck- and data were analyzed using FlowJo software (TreeStar V.10).

mRNA extraction and sequencing library preparation
Pure FACS-sorted cells were stored at −80 • C in RNAlater (Thermo Fisher, Carlsbad, CA, USA), as previously reported. 11  Analyzer and the Nextera XT DNA Library preparation kit (Illumina).

mRNA sequencing: data processing and statistical analysis
The raw sequencing reads were trimmed to remove the adapters and filtered for low quality and low complexity (Cutadapt v. We subsequently calculated the average expression level of all genes within a signature in each cell subset independently, and drew the spider chart using the fmsb package (v.0.6.3).

RNA purification and qPCR
Total RNA was isolated from highly pure, sorted human ILC and with specific primers (hIL18R1 5 ′ -TGGTGTGGCAGTTAAGAGATG-3 ′ ,

Statistical analysis
Statistical analysis was performed with GraphPad Prism software version 6 using parametric t-test. The data is shown by plotting individual data points and the mean ± SEM. A P-value <0.05 (2-tailed) was considered statistically significant and labelled with *. P values <0.01 were labelled with **.

Circulating human ILCs and CD4 Th cells have distinct gene signatures
To gain insight into the transcriptomic profiles of human ILC and In contrast, TRDC and TRDJ2 were more expressed in ILCs than in CD4 Th cells, as previously reported by Li et al. 8 (Supporting Information   Fig. S1D).
Next, we performed GSEA to determine whether defined GO genesets were shared as a common ILC signature as compared to the Th one. A total of 112 GO genesets were significantly enriched across all three individual ILC subsets vs. the Th subsets (Fig. 1B).
These GO genesets related to metabolic processes, lymphocyte activation in immune responses, leukocyte migration, defense responses to pathogens, myeloid cell activation and migration, phago-and endocytosis, and B cell regulation (Fig. 1C). Th cells ( Fig. 2A). High protein expression of CCR1 on circulating ILCs was also confirmed by flow cytometry (Fig. 2A). Regarding cytokines, LIF and IL18 were the most up-regulated in ILCs, mainly due to high expression of these factors in ILCPs. The response of ILCs and CD4 Th cells to cytokines also seems to be distinctly regulated. Indeed, total ILCs expressed significantly higher levels of IL13RA1, LTBR, IL18R1, and IL1R1, as confirmed by qPCR or flow cytometry (Fig. 2B). The presence of these receptors in ILCs might account for the well-known plasticity of these cells that are able to promptly convert one into the other in response to environmental cues. Moreover, given the absence of somatically rearranged antigen receptors on ILCs, these cells are characterized by their ability to rapidly respond to tissue mediators derived from lipid metabolism, bacterial, and dietary products. This is reflected in our mRNA sequencing data by the higher expression in ILCs of TLR4, involved in innate sensing of pathogens, and CIITA, the master regulator of MHC class II (Fig. 2C). Differences in the expression of master transcription factors might also account for the divergent differentiation of ILCs and Th cells. In that regard, we observed higher expression of PAX5 and TCF4 in ILCs than in Th cells (Fig. 2D).
Finally, we compared the expression of costimulatory and coinhibitory receptors, and neuropeptide/adrenergic receptors known to influence the kinetics and magnitudes of immune cell responses. We identified a general absence or very low expression of inhibitory receptors (e.g., TIGIT, CTLA4) on total ILCs compared to CD4 Th cells, whereas only at mRNA level, higher expression of some costimulatory receptors, in particular of 4-1BB (TNFRSF9) (Fig. 2E). Moreover, in line with published work on ILC involvement in neural circuits, 22-26 these cells showed overexpression of ADRB1, VIPR2, NMUR1, and HTR1F and significant differences in their metabolism-related genes (Fig. 2F).
Overall, these results argue for distinct transcriptomic signatures in human ILCs and CD4 Th cells that might be related to their distinct sensing of the environment and kinetics of immune reactivity.

Pairwise comparison of mirror ILCs and CD4 Th cell subsets reveals shared, but also distinct gene profiles
Next, we aimed at directly comparing the gene expression profiles of mirror ILCs and CD4 Th cell subsets. To do that, we performed a PCA of ILC1s, ILC2s, ILCPs, Th1s, Th2s, and Th17s using the top 500 most variable genes among ILCs and CD4 Th subsets (Fig. 3A (Fig. 3C).
F I G U R E 1 (Continued) (C) Heatmap (mRNAseq) of row z-scores of the top 100 up-regulated genes and the top 100 down-regulated genes between total ILCs and total Th cells. Each gene is labeled with colors according to its belonging to the genesets listed. Genes were clustered hierarchically using the complete method on Euclidean distances of scaled log 2 (normalized cpm)

GSEA of differentially expressed genes between mirror subsets
Whereas Th1s are defined as CCR6 − cells, we found a significant higher expression of CCR6 in ILC1s compared to Th1s at both transcriptomic and protein level (Fig. 4A). ILC2s were characterized by high mRNA levels of CCR1 and CXCL2, IFNGR1 (also known as CD119), LTBR, IL1RL1 (also known as ST2, the IL-33 receptor), CD160, 4-1BB, and significantly lower expression of the IL21R, as compared to their Th counterparts (Fig. 4B). Elevated protein levels of CCR1 and IFNGR1 were confirmed by flow cytometry and increased transcript for ST2 were validated by qPCR (Fig. 4B). Expression of VIPR2 was also higher in ILC2s (Fig. 4B). Finally, in the comparison between ILCPs and Th17 cells the most striking differences appeared in CCR1 expression, that was higher in ILCPs, and CCR10 that was increased in Th17 cells. Several chemokines were also overexpressed in ILCPs, including XCL1, PPBP, and CXCL2. Furthermore, when considering costimulatory and coinhibitory receptors, Th17 cells presented elevated levels of TIGIT, LAG3, and CTLA4, whereas ILCPs were overexpressing 2B4, as confirmed at protein level. Neuroregulatory circuits might also be distinct among ILCPs and Th17s, as illustrated by the different expression of VIP1R and RAMP1 (Fig. 4C).
Overall, to visualize the contribution of these genes in defining the ILC subset signature as compared to their CD4 Th counterparts a spider diagram representation was used. As shown in Figure

Expression of lncRNAs is distinct between ILCs and CD4 Th cells
The lncRNAs represent an abundant part of the cellular mRNA content, display high cell subtype specificity, and their expression has been associated with several diseases. 27 Therefore, we explored our dataset for the expression levels of lncRNAs in ILCs as opposed to CD4 Th cells. A sharp differential expression of lncRNAs between the innate and adaptive counterparts could be observed, with 3, 26, and 169 lncR-NAs differentially expressed between ILC1s/Th1s, ILC2s/Th2s, and ILCPs/Th17s, respectively (Fig. 6A). Interestingly, a shared expression of lncRNAs was observed among ILC subsets, as represented in the Venn diagrams of Figure 6B. One lncRNA in particular, CASC15, was overexpressed by ILCs regardless of specific subset, as opposed to their Th counterparts and it might represent an ILC-specific lncRNA signature gene.

DISCUSSION
Here  ILC1s and Th1s, ILC2s and Th2s, and ILCPs and Th17s specification/maintenance. It is tempting to speculate that the presence of these receptors accounts for a preprogrammed subset plasticity enabling the rapid adaptation in effector functions in the microenvironment (e.g., IFNGR on ILC2s). In this regard, evidence of ILC plasticity has been extensively reported in ILC1s, 29 ILC2s, [30][31][32][33] and ILC3s. 34 To some extent, plasticity among Th subsets has also been described, mainly between Th1/Th2 35,36 and Th17/regulatory T cells (Tregs). 37 Further, the observed overall low levels of coinhibitory receptors on ILCs agree with a cell profile characteristic of more immediate activation, to provide protection in a window when CD4 Th cells have not yet developed a specific immune response. However, expression of PD-1 has been reported in ILC2s, 38  Fine-tuning of immune cell functions and acquisition of cell identity might also result from the effect of lncRNAs on gene expression.
In CD4 T cells, LincR-Ccr2-5 ′ AS has been involved in murine CD4 Th2 cell migration 43 and linc-MAF-4 was described as specifically linked to a CD4 Th1 phenotype. 44 With respect to ILCs, a previous work identified the lncRNA Rroid as a key regulator of ILC1 development in mice. 45 In our analysis we observed significantly different expression of lncRNAs in ILCs as compared to CD4 Th cells. This is particularly true for the ILCPs vs. Th17s comparison, where more than 150 lncRNAs were found to be differentially expressed. ILCPs are known to maintain open developmental options toward both NKs and helper ILCs. 46 It might be speculated that this innate cell precursor abilities are at least in part controlled by the action of lncRNAs and other epigenetic cues, as recently shown in ILC2 development. 47 Future studies will be needed to evaluate the impact of distinct lncRNAs on ILCs. Of interest, we identified CASC15 as a signature lncRNA in ILCs as opposed to total CD4 Th cells. Several studies have already evaluated the role of this lncRNA in different types of tumor. In that regard, CASC15 has been reported to directly bind to the enhancer of zeste homolog 2 (EZH2), the key catalytic component of the polycomb repressive com-plex 2 (PRC2) involved in H3-K27 methylation. 48 In melanoma, the CASC15-dependent EZH2-mediated inhibition of PDCD4 resulted in tumor progression. 49 Whether this lncRNA has also an impact on ILC development and/or proliferation remains to be investigated.
Overall, our transcriptomic results support the notion that ILCs might be involved in physiologic process or contribute to pathologies previously attributed to T cell functions/dysfunction. 50 We also provide indication for potential hitherto unknown roles of lncRNA in shaping human ILC biology. Final evidence in support of this assumption will need investigations in mouse models specifically probing gene expression in ILC subsets.

AUTHOR CONTRIBUTIONS
GE performed the experiments, analyzed the data and wrote the paper.

DISCLOSURES
The authors declare no conflicts of interest.