Transcriptional profiling of eosinophil subsets in interleukin‐5 transgenic mice

Abstract Eosinophils are important in fighting parasitic infections and are implicated in the pathogenesis of asthma and allergy. IL‐5 is a critical regulator of eosinophil development, controlling proliferation, differentiation, and maturation of the lineage. Mice that constitutively express IL‐5 have in excess of 10‐fold more eosinophils in the hematopoietic organs than their wild type (WT) counterparts. We have identified that much of this expansion is in a population of Siglec‐F high eosinophils, which are rare in WT mice. In this study, we assessed transcription in myeloid progenitors, eosinophil precursors, and Siglec‐F medium and Siglec‐F high eosinophils from IL‐5 transgenic mice and in doing so have created a useful resource for eosinophil biologists. We have then utilized these populations to construct an eosinophil trajectory based on gene expression and to identify gene sets that are associated with eosinophil lineage progression. Cell cycle genes were significantly associated with the trajectory, and we experimentally demonstrate an increasing trend toward quiescence along the trajectory. Additionally, we found gene expression changes associated with constitutive IL‐5 signaling in eosinophil progenitors, many of which were not observed in eosinophils.


INTRODUCTION
Eosinophils are granulocytes that play a role in the pathogenesis of asthma, atopic dermatitis, and allergy. Although they have been associated with defense against parasites such as helminths, it is their pathogenic inflammatory role that has significant implications for human health in the developed world.
Eosinophils are produced in the bone marrow (BM) and differentiate from myeloid progenitors in response to IL-3, GM-CSF, and more selectively, IL-5. 1,2 Although eosinophils are normally a rare population, upon IL-5 stimulation, which can occur in response to a parasitic infection and inflammatory disease, the BM can produce several orders of magnitude more eosinophils than are seen at steady state. have a permanent and extreme eosinophilia, where eosinophils make up over half of all leukocytes in the BM. 3 IL-5 acts at multiple levels throughout the eosinophil lineage, regulating eosinophil production, activation, migration, and survival. [3][4][5][6] Murine eosinophils can be identified using flow cytometry by their granularity and their expression of IL-5 receptor alpha (IL-5R ) and the surface lectin Siglec-F. 7,8 Siglec-F possesses an ITIM characteristic of the Siglec family that can mediate inhibitory functions including induction of apoptosis. 9 In mouse models of allergy in which Siglec-F ligand has been deleted from bronchial epithelial cells and some immune cells, eosinophil populations are expanded due to reduced apoptosis. 10  www.jleukbio.org 195 spleen (spln), marking them for clearance. 11 Siglec-F has a functional analogue in humans-Siglec-8-rather than a genuine paralogue, which convergent evolution has filling a similar role. 9 Novel therapies targeting IL-5 signaling are approved for the treatment of eosinophilic asthma, 12,13 and anti-IL-5 and anti-Siglec-8 therapies are currently in development for the treatment of diverse eosinophilic diseases. [14][15][16] Here, we have studied the gene expression changes associated with IL-5 up-regulation. We noted the expansion of 2 eosinophil populations in IL5T mice that were distinguished on the basis of medium and high Siglec-F expression. We transcriptionally profiled these populations in BM and blood and have made these transcriptional analyses available at haemosphere.org. Transcriptional profiling and in silico approaches were used to reconstruct the eosinophil lineage (from common myeloid progenitors (CMP) through to mature eosinophils), and show that the Siglec-F hi eosinophil population from the PB, which is rare in wild type (WT) mice, falls at the end of this linear trajectory. We identified gene sets that were associated with eosinophil lineage progression and increased Siglec-F expression, which include eosinophil specific transcription factors and cell cycle regulators.
Furthermore, eosinophils along this trajectory were shown to have demonstrably different cell cycle profiles and to trend toward quiescence. Finally, we explored the transcriptional consequence of constitutive IL-5 signaling on the lineage, noting gene expression changes that were specific to eosinophil precursors (EoPs).

Cell purification
All cells were purified from mice between 7 and 12 weeks of age. BM was collected from femurs and tibiae. PB was collected from the retro-

Data availability
All data can be viewed and are available for download on haemosphere.org. The dataset is also available in the Gene Expression Omnibus under the accession number GSE112010.
Full details of DNA content analyses, ELISA, microarray, cytocentrifuge, RNA isolation, RT qPCR, and bioinformatics analyses can be found in the supplementary methods.

RESULTS AND DISCUSSION
Cells of the eosinophil lineage are rare in hematopoietic tissues of healthy WT mice but dramatically expanded in IL5T mice as has been reported by Dent et al. 3 We confirmed this expansion in BM, Spln, and PB of IL5T mice, from EoPs to mature eosinophils ( Fig. 1A

Cluster analysis of eosinophil and eosinophil progenitor expression profiles
Given that EoH were expanded upon constitutive IL-5 stimulation (IL5T mice) and have been reported but not characterized in allergy models, 11 it is possible they may play a role in disease. To explore the relationship between this population and other cells in the eosinophil lineage, we collected EoM and EoH from the BM and EoH from the PB of IL5T mice and analyzed gene expression using microarrays. We also collected EoPs, CMPs, and GMPs from WT and IL5T mice (Supplementary Fig. 1). These cells were all on a BALB/c background. We compared them to each other, and to transcriptional profiles of WT C57BL/6 eosinophils (bulk sorted without Siglec-F substratification), EoPs, CMPs, and GMPs that have been previously published by our laboratory as part of a general survey of transcription in blood cells 20 (Supplementary Table 1).
Hierarchical clustering of these samples showed that the progenitors all grouped by cell type, with little difference between "like" cell types from different genetic backgrounds ( Fig. 2A). Eosinophils formed a separate cluster, within which the IL5T eosinophil samples clustered together by Siglec-F expression and tissue of origin.
C57BL/6 eosinophils formed a subcluster, rather than grouping with the IL5T EoM, to which they were closest by surface phenotype, a difference which may be driven by the effect of chronic IL-5 stimulation on these cells. Although the algorithm allows for branching between cell types, in our analyses we found the cell types formed a linear trajectory from CMP through to EoH PB (Fig. 2C). Based on these results, we propose an order for the IL5T eosinophil lineage: a linear trajectory from EoP to EoM BM to EoH BM to EoH PB.
Together, these data support the notion that transcriptionally, EoH are the terminal definable state in our eosinophil trajectory. We interpret this as EoHs representing activated or stimulated eosinophils, which are present in WT mice and driven by constitutive IL-5 stimulation in IL5T mice.

Identification of eosinophil trajectory genes
Having demonstrated the positioning of cells along a linear trajectory, we hypothesized that as cells progress along the series genes they would be gradually up-or down-regulated. We have termed such genes "trajectory genes". To identify trajectory genes, we focused on genes that were differentially expressed (DE) between (1) EoPs versus EoH BM cells and (2) EoM BM versus EoH PB from IL5T mice. These comparisons spanned the key "eosinophil specific" parts of our developmental progression-as opposed to including multi-potential progenitors such as CMPs and GMPs. We did not compare the cells that were immediately adjacent to each other in the series because adjacent cells were likely to have smaller transcriptional differences and thus a higher background to signal ratio leading to a lower ability to detect key genes. We reasoned that this would identify key trajectory genes, without being unduly restrictive. We selected genes that were significantly up-or down-regulated with a log 2 fold change of at least 0.5. Using this approach, we identified 330 up-regulated genes (353 probes; Fig. 3A(i)) and 271 down-regulated genes (290 probes; Fig. 3B(i)) that were significantly altered across the eosinophil trajectory ( Supplementary Tables 1 and 2). Expression profiles of 4 such trajectory genes (Ramp1, Cebpb, Tlr4, and Trem14) were validated using RT qPCR (Fig. 3C). Moreover, there was high concordance with our data and published WT BALB/c RNASeq data generated by the Fulkerson laboratory 21 comparing EoPs to eosinophils ( Fig. 3A(ii) and B(ii)). Of our 601 DE genes, 261 were significantly DE in the same direction with at least a log 2 fold change of 0.5, and only 8 were significantly DE in the opposite direction (Fig. 3A(ii) and B(ii)).
To investigate if trajectory gene expression changes were specific to eosinophils or occurred generally during hematopoietic differentiation, we examined their expression in a broad range of hematopoietic cells 20 (Fig. 3A(iii) and B(iii)). Down-regulated eosinophil trajectory genes generally had higher expression in progenitors than mature cells, implying general functions within progenitors and stem cells that are down-regulated with differentiation into various lineages. Downregulated trajectory genes could be separated into 3 broad classes by expression pattern-those that were expressed in many cell types (ubiquitous), those that were expressed more highly in progenitors, and those that were expressed at higher levels in EoPs (denoted as early eosinophil, Fig. 3A(iii)).
We also examined the expression of up-regulated eosinophil trajectory genes across wider hematopoiesis and found that the majority of them had their expression restricted to particular cell types. We grouped them into 3 categories: genes with ubiquitous expression, those that were more highly expressed in mature cell types, and those that were more restricted to eosinophils (Fig. 3B(iii)).
We further examined up-regulated trajectory genes with restricted expression in mature cells and eosinophils as these were the most likely to have functional importance. In the mature cell set, we found 9 transcription factors. These included Ikzf3 (Aiolos), Nod2, Bcl3, and Foxo1, which are reportedly induced in eosinophil maturation. 21 Ikzf3 binding sites have also been shown to be enriched in active genes regulated during eosinophil development. 21 Six eosinophil specific transcription factors-Arnt, Crtc1, Hic1, Hoxc9, Zfp689, and Zfp282-were also found to be up-regulated To discover more about the function of eosinophil trajectory genes, we tested gene sets derived from gene ontology (GO) terms sourced from MSigDB 22 with Fisher's exact test for overrepresentation. Multiple GO terms associated with the "cell cycle" (genes including Cdk1, Cdca3, and mini-chromosome maintenance genes) were statistically overrepresented in down-regulated trajectory genes, whereas "negative regulation of the cell cycle" (including the cyclin-dependent kinase inhibitor encoding genes Cdkn2a and Cdkn2b) was overrepresented in the up-regulated trajectory genes. Moreover, subsequent testing of trajectory gene sets for overrepresentation of transcription factor binding motifs revealed enrichment for the E2F family (Fig. 4A), known to play a key role in cell cycle regulation. 23 Together, these data are consistent with cell cycle arrest and entry into quiescence across the eosinophil trajectory.
To test directly whether EoP, EoM, and EoH populations differed in terms of cell cycle in vivo, we assayed these populations by flow cytometry. In the WT context, we examined eosinophils from transgenic Fucci mice, 17 in which fluorescent reporters had been linked to proteins Cdt1 and geminin that oscillate inversely during the cell cycle.
Cdt1, linked to Kusabira Orange, accumulates during G 1 and G 0 and is promptly degraded at the onset of S-phase. Geminin, fused to Azami-Green, accumulates in S/G 2 /M phases, and is degraded at the completion of mitosis. 17,24 In Fucci mice, EoP, EoM, and EoH cells significantly differed in their cell cycle distribution (Fig. 4B and C). EoPs were As IL5T mice were not available on the Fucci background, we examined the DNA content in IL5T eosinophil populations with DAPI to assess cell cycling ( Fig. 4D and E). Again, we see that EoP and EoM in the BM have a significant population with >2N DNA content (denoting cells in S/G 2 /M phases), whereas EoH BM and PB eosinophils had almost exclusively 2N content and were likely in G 1 /G 0 arrest.
Together, these data demonstrate that cells become quiescent as they progress along the eosinophil trajectory, and that EoP, EoH, and EoM differ in terms of their cell cycle profiles.

Granule protein production in IL5T mice
The production and release of granule components, including the cytotoxic proteins eosinophil peroxidase (EPX), major basic protein, and eosinophil associated RNAses, are central to eosinophil function. Analysis of transcript expression for genes associated with granules in IL5T mice (Fig. 5A)  To compare the pattern of transcript production to protein storage, we performed an EPX ELISA on IL5T eosinophil populations (Fig. 5B).
EPX protein levels were not significantly different between these populations. This supports the notion that EPX production occurs early in eosinophil development in the BM and is then stored until it is required in the periphery. This is in contrast to the expression of the CCR3, the eotaxin receptor, which has the levels of protein surface expression more correlated with its transcript levels (Fig. 5C).

Effects of constitutive IL-5 on EoPs
EoPs express IL-5R and are responsive to IL-5 stimulation. Gene expression changes have been examined in BM cells cultured in vitro with IL-5, 26 but little is known about the transcriptional differences caused in eosinophil progenitors by chronic IL-5 stimulation in vivo.
We therefore compared the transcription profiles of EoPs, CMPs, and GMPs isolated from IL5T to their counterparts from WT mice. In EoPs, the influence of IL-5 stimulation was prominent with 325 DE probes ( Fig. 6A), as opposed to the few significantly DE genes in the CMPs and GMPs (51 and 11 probes, respectively, at a false discovery rate of <0.05), suggesting IL-5 stimulation has a major influence on cells committed to the eosinophil lineage.
We examined the expression of EoP IL-5-responsive genes in the eosinophil-committed section of the eosinophil trajectory (from EoP to EoH PB; Fig. 6B and C). There were 3 major groups of genes ( Fig. 6C): (1) those that were DE from EoP to eosinophils, which we considered associated with maturation, (2) those that were generally IL-5 responsive, that is, not DE during maturation but also DE between WT and IL5T eosinophils-termed IL-5 shared, and (3) those which did not fall into either previous category, so were only responsive to IL-5 in EoPs (Supplementary Tables 1 and 3).
More than half the differential expression (58%) is due to specific effects of IL-5 on EoPs, with few genes DE due to a general IL-5 effect on both EoPs and eosinophils. A total of 33% of the EoP IL-5 responsive genes were associated with eosinophil maturation including down-regulation of cytokine signaling genes Il12a, Il17rb, Il15, Il31ra, Cd72, and Stat4.
Notably, genes involved in granule production were not significantly affected by IL-5 stimulation in EoPs, despite being an early sign of EoP differentiation from less committed progenitors.
In designed and performed research, analyzed data, performed transcriptional analyses, and wrote the manuscript.