The Discovery of Chicken Foxp3 Demands Redefinition of Avian Regulatory T Cells

Ever since the publication of the first chicken genome in 2004 (1), the avian research community has been struggling with the apparent absence of avian orthologs of genes that are considered vitally relevant to mammalian physiology (2–4). As evolutionary success requires adaptation to environmental changes, speciation events go hand in hand with gain or loss of genes. Therefore, it is likely that some, or even most, of these genes are indeed not present in birds (2, 4). However, some presumably missing genes have recently been described in chickens (2, 5, 6). As the original publication on the chicken genome had indicated—and others emphasized later—avian genomes feature a high overall GC content, which impairs sequencing (1, 2). This characteristic has been reported to be displayed by certain mammalian genes as well: due to GC-biased gene conversion, genes that have been the subject of vast evolutionary changes are especially GC-rich and usually underrepresented in sequencing datasets (7).

One of the genes that was assumed to be absent in birds until recently is Foxp3. Foxp3 plays a central role in the maintenance of immune homeostasis, representing the master transcription factor of regulatory T cells (T_regs) (8). All attempts to find traces of avian Foxp3 yielded no results (9, 10). In 2013, however, the highest quality avian genome so far, namely that of the Tibetan ground tit (Pseudopodoces humilis), was published (11) and, surprisingly, P. humilis Foxp3 was automatically annotated. Subsequently, Denyer et al. (12) used the ground tit sequence as a template for further database searches and identified single Foxp3 exons in two falcon species (Falco peregrinus and Falco cherrug). However, the function of avian Foxp3 and whether it possesses any relevance for T_regs have not been evaluated in any bird species yet.

T_regs were first described as suppressors of T cell help in mice by Gershon et al. (13) in 1972. Even though a lot of knowledge remains to be gathered, they represent a well-established cell population today. T_regs in mice are defined as a CD4⁺CD25⁺Foxp3⁺ T lymphocyte subset (14), but it has been shown that their phenotype is highly variable depending on species and condition (15). Generation of T_regs can either occur by thymic selection and priming (naive T_regs [nT_regs]) or by peripheral induction of Foxp3 in CD4⁺CD25⁺ naive T cells (iT_regs) (16). Both pathways are promoted by IL-2/IL-2R ligation in the presence of TGF-β in a non-inflammatory environment. As TCR stimulation together with coreceptor ligation is reported to be crucial for effector T cell induction, T_reg development can be triggered earliest during the double-positive (DP) stage (i.e., CD4⁺CD8⁺) of T cell development in the thymus or in mature single-positive (SP) peripheral T cells (i.e., CD4⁺ or CD8⁺). The expression of Foxp3 then indicates lineage determination toward the regulatory branch (8). T_regs express a distinct set of effector cytokines, among which are IL-10 and TGF-β. Along with other suppressor functions, these cytokines diminish proinflammatory immune response mechanisms (17, 18). This is achieved mainly by the IL-10–mediated downregulation of IL-2, which is not critical for the maintenance of T_regs or their induction outside the thymus, but for the activation of naive Th cells (19).

Due to the presumed absence of avian Foxp3, the body of literature on T_regs in birds is comparably small and there is low consensus on the cells’ phenotype. Nevertheless, only a few years after the initial description of mouse suppressor T cells (13), suppression of Ab production in response to infection by a subset of T cells was described in the chicken (20). Various groups have since worked on avian suppressor T cells: the group who discovered the suppressive T cell–mediated effect on Ab production later characterized a subset of these cells as CD8⁺ γδ T cells (21). A number of studies of another group concluded that most CD4⁺CD25⁺ avian T cells display regulatory properties (9, 22–24), and another study recently suggested TGF-β as a marker for T_regs during pathology (25). This heterogeneity of data calls for a universal definition of avian T_regs.

Therefore, the discovery of avian Foxp3 finally paves the way for standardized T_reg research in birds. This study, aiming to characterize chicken Foxp3, is hence the crucially required next step toward a new and fundamentally more precise description of avian T_regs.

To assemble the chicken Foxp3 and chCcdc22 sequences, dataset PRJEB4677, obtained from the National Center for Biotechnology Information (NCBI) (https://www.ncbi.nlm.nih.gov) Sequence Read Archive (SRA) database, was used. Sequences of non-avian vertebrate Foxp3 genes and P. humilis Foxp3 (Table I) were used as probes in basic local alignment search tool (BLAST) searches of the SRA dataset. The sequences obtained were downloaded and assembled manually either with CLC genomics workbench 8.0.1 (www.clcbio.com) or with Lasergene 10.0.0 (DNASTAR, Madison, WI). The resulting short contigs were used as probes in subsequent rounds of BLAST searches against SRA datasets, until the full open reading frame was completed. The assemblies of chicken Foxp3 and chicken Ccdc22 have been published in the GenBank database (https://www.ncbi.nlm.nih.gov/genbank/) with the respective accession numbers MT133687 and MT133688.

The protein sequences listed in Table I were aligned using the ClustalX 2.1 algorithm (http://www.clustal.org). The similarity between sequence stretches of chicken and human Foxp3 was calculated using the EMBOSS needle pairwise alignment tool (https://www.ebi.ac.uk/Tools/psa/emboss_needle/).

The maximum likelihood method based on the JTT matrix–based model (26) was used by Mega6 (27) to create 500 individual trees using all sites of a previously made MUSCLE alignment of the protein sequences listed in Table I. The bootstrap method was used to assess the likelihood of the displayed taxa association.

To obtain long-read sequences from the chicken genome, we used genomic DNA samples from a female inbred line C White Leghorn chicken (28). Genomic DNA was quantified on the Qubit fluorometer (Invitrogen, Carlsbad, CA) and used for sequencing library preparation (SQK-LSK109; Oxford Nanopore Technologies, Oxford, U.K.). The sequencing was performed on the GridION X5 platform using R9.4.1 chemistry, and base calling was performed in real time using the GridION sequencing software version 19.12.2 (all from Oxford Nanopore Technologies). To select reads from the Foxp3 locus, a BLAST library was generated using CLC Genomics Workbench, and searched with Foxp3 coding sequences. Two reads were selected that covered the entire Foxp3 coding exons and neighboring genomic sequence, including the Ccdc22 gene. Individual exons were mapped to these Nanopore reads by pairwise alignment. The Nanopore reads used to assess Foxp3 synteny have been assigned the NCBI SRA accession number PRJNA610550 (https://www.ncbi.nlm.nih.gov/sra).

White Leghorn line M11 chickens (Federal Research Institute for Animal Health, Neustadt, Germany) were hatched, vaccinated only against Marek’s disease, and conventionally housed. Water and a commercial diet were provided ad libitum. At the age of 3–6 mo, the chickens were euthanized for tissue collection.

Organs were collected in sterile PBS for immediate lymphocyte preparation. Organs were mashed to create a cell suspension as described elsewhere (29). Cell suspensions were subsequently density centrifuged over 1.077 g/ml Biocoll separating solution (Merck, Burlington, MA) and the lymphocyte layer was extracted and washed.

Lymphocytes from respective organs were incubated for 20 min with a mix of anti-CD25, anti-CD4, and anti-Bu1 mAbs, followed by 20-min staining with the respective isotype-matched and fluorochrome-labeled secondary Abs (Table II). Cell sorting for viable single cells was performed with a FACSAria IIIu (Becton Dickinson, Franklin Lakes, NJ) to a purity of >91%. Purified cells were pelleted and stored at −80°C until RNA extraction.

Unless otherwise indicated, total RNA was isolated from sort-purified lymphocyte preparations using the ReliaPrep cell RNA isolation system (Promega, Madison, WI). RNA quality was determined with a Bioanalyzer (Agilent Technologies, Santa Clara, CA); only RNA with a minimum RNA integrity number of 7.5 was used for further analyses. Reverse transcription of 400 ng of RNA was carried out using GoScript reverse transcription master mix (Promega).

Total RNA for RT-PCR was isolated from chicken spleen tissue using TRI reagent (Sigma-Aldrich, St. Louis, MO). Reverse transcription was performed using the SMART RACE (Takara Bio, Kusatsu, Japan) protocol with 1 µg of RNA as a template. For the amplification of the highly GC-rich Foxp3 gene, previously described conditions (5, 30) were used: a 1:200 mixture of Deep Vent and Taq polymerases (both New England Biolabs, Ipswich, MA), and PCR primers with high melting temperature and long amplification times (Table III). The PCR program was set to 30 cycles of 95°C for 30 s and 65°C for 10 min. The PCR product was extracted from a 0.6% agarose gel using a Qiaex II gel extraction kit (Qiagen, Hilden, Germany) and directly sequenced.

SYBR Green–based real-time RT-PCR (subsequently called quantitative PCR [qPCR]) was performed with cDNA in a dilution ensuring linear amplification of each target gene, using GoTaq probe qPCR master mix or GoTaq qPCR master mix (Promega) to amplify Foxp3 or 18S rRNA, IL-10, IL-2, and CD8 (Table III), respectively. For the amplification of CD8, the QuantiTect primer assay for chicken CD8α (Qiagen) was used according to the manufacturer’s instructions. All other primers and the Foxp3 probe were synthesized by Eurofins Genomics (Ebersberg, Germany) and used at a final concentration of 300 nM (primers) or 150 nM (probe) each. Primer efficiency and linear amplification range was determined using serial dilutions of a cDNA mix. In addition to the primers, each 25-µl reaction contained 12.5 µl of the GoTaq qPCR master mix, 0.25 µl of 100× CXR reference dye, 4.25 µl of nuclease-free water, and 5 µl of diluted cDNA. The cycling conditions were set to 95°C for 2 min, followed by 40 cycles of 95°C for 15 s, 59 (Eurofins primers without probe) or 60°C (Foxp3 and CD8) for 30 s, and 72°C for 30 s. Each assay included relevant non-template controls. A melting curve performed following each run with the qPCR master mix confirmed a single, specific PCR product. Raw data created by a 7300 real-time PCR system was analyzed using the SDS 7300 software (both from Applied Biosystems, Foster City, CA). Relative expression fold changes of Foxp3, IL-10, IL-2, and CD8 were assessed using normalized cycle threshold values (ΔCt). Normalization was carried out against 18S rRNA expression for each sample. ΔCt values were subtracted from the total number of cycles (40 − ΔCt) and used as an exponent of 2 to calculate fold changes (31). The mean of all ΔCt values measured for each gene and organ was set to 1.

Chicken Foxp3 was codon optimized for expression in human cells using the GeneArt Thermo Fisher Scientific web tool. The resulting sequence was synthesized as gBlocks Gene Fragment including a 3xFLAG tag on the 5′ end. The gBlocks Gene Fragment was cloned into pcDNA3.1 using the NEBuilder kit (New England Biolabs) according to the manufacturer’s instructions. Obtained clones were checked by Sanger sequencing. One correct clone was expressed in HEK293 cells, and cDNA of these cells was subsequently used as a template. The cloning template for full-length murine Foxp3 was synthesized without any modifications (GeneArt, Regensburg, Germany). Assembly primers were designed to overlap the junction of the pSBbi(GP)rev vector backbone and the respective Foxp3 sequence; the forward primers contain the Myc-tag sequence (see Table III). The NEBuilder HiFi DNA assembly cloning kit (New England Biolabs) was used to assemble the constructs according to the manufacturer’s instructions. The resulting plasmids were expressed in Escherichia coli and purified with the PureYield plasmid miniprep system (Promega).

293T cells (10⁵) were cultured overnight and transfected with 250 ng of the mouse or chicken Foxp3 construct, respectively, and 12.5 ng of a transposase-containing plasmid using X-tremeGENE 9 DNA transfection reagent (Sigma-Aldrich) according to the manufacturer’s instructions. GFP expression was confirmed microscopically and the cells were harvested for flow cytometry with a 0.02% EDTA solution in PBS 48 h after transfection.

For intracellular staining, 293T cells transfected with either chicken or mouse Myc-Foxp3 plasmid were incubated with intracellular fixation buffer for 30 min in the dark at room temperature and permeabilized using 1× permeabilization buffer in every subsequent washing step. Viability staining was performed with fixable viability dye (all from eBioscience, San Diego, CA). Abs against the Myc epitope, mouse Foxp3, respective isotype-matched negative controls, and, where necessary, fluorochrome-labeled isotype-specific secondary Abs (see Table II) were diluted in 1× permeabilization buffer. The staining protocol used is described elsewhere (32). For cell surface staining, splenic lymphocytes were stained with anti-CD25, anti-CD8, anti-CD4, and anti-CD3 mAbs or respective control Abs (see Table II) according to standard procedures. Viability was assessed using 7-aminoactinomycin D (Invitrogen). Analyses were performed with a FACSCanto II, with sort purification of cell populations performed with a FACSAria IIIu using FACSDiva and FlowJo software (all from Becton Dickinson).

Using previously published avian Foxp3 sequences (12) for homology searches in the NCBI SRA, we assembled the entire coding sequence of putative chicken Foxp3 from the database short reads. Subsequently, we confirmed the correctness of the assembly by RT-PCR amplification and sequencing of full-length Foxp3 from chicken spleen RNA. Previously missing avian genes that have recently been assembled manually are similar in respect to their comparably high overall GC content and long GC stretches (2). To assess whether such features might have impeded the annotation of the newly identified chicken sequence as Foxp3 in the past, we plotted them for each Foxp coding sequence (Foxp1–Foxp4) of six vertebrate species (Supplemental Fig. 1). We found that avian Foxp sequences do not generally differ from mammalian and reptilian sequences concerning GC characteristics. Foxp1 and Foxp2 sequences display no significant species differences (Supplemental Fig. 1A), whereas Foxp4 is slightly more variable and displays a higher GC content (Supplemental Fig. 1C). Foxp3, however, shows a clear increase in both, that is, length of GC stretches and overall GC content, that peaks in the two avian species (Supplemental Fig. 1B). The hypothesis that chicken Foxp3 could not be annotated before the present due to an extraordinary GC content is thereby further supported.

To verify the identity of the newly assembled Foxp3 sequence, we aligned it with all four previously described members of the Foxp family from a total of six mammalian and sauropsid species. Phylogenetic analysis (Fig. 1) generated from this alignment reveals that the protein sequences of Foxp1 and Foxp2 are highly conserved among species. A slightly elevated substitution rate separates mammalian and sauropsid Foxp4 sequences. Foxp3, however, displays considerable variation between species, whereas mammalian sequences, on the one hand, and sauropsid sequences, on the other hand, cluster together. The non-archosaur sauropsid lizard Anolis carolinensis represents a separate branch within the latter clade. Consistent with previous analyses (33), it is also shown that Foxp3 sequences are generally less conserved than those of the other Foxp family members.

We also aimed to assess whether the chicken Foxp3 locus shows synteny. To this end, we assembled the sequence of the closest neighboring gene, chicken Ccdc22, using a similar approach as for chicken Foxp3. Next, we mapped the position of both genes in long-read Nanopore sequences from the chicken genome. Indeed, both chicken Foxp3 and Ccdc22 lie on a single contiguous Nanopore read (Supplemental Fig. 2), suggesting strongly that the two-gene synteny is conserved in the chicken genome, as it is in the Tibetan ground tit genome (12). Additionally, this analysis indicates that chicken Foxp3 is divided into 11 coding exons, similar to the human ortholog. Thus, both, phylogenetic analysis and synteny, indicate that the newly assembled sequence is a true chicken Foxp3 ortholog.

As a next step, we analyzed the Foxp3 protein sequences of the aforementioned six species for conservation of functionally relevant domains (Fig. 2). The N terminus contains the repressor domain, which is represented by a proline-rich region and conveys the repressive function of Foxp3 (34). This domain is poorly conserved in chickens. It is only ∼100 aa long, instead of 170 aa in human Foxp3, and displays low similarity values. Nevertheless, with roughly 30% proline residues compared with 17% in the human sequence, it is highly proline rich. The zinc finger domain, located downstream of the proline-rich region at aa 198–221 of human Foxp3, is absent in non-mammalian species. However, the leucine zipper domain (aa 240–261) (35), which in non-mammals connects the N- and C-terminal domains, is highly conserved, displaying 67% similarity. It contains a conserved glutamic acid residue essential for dimerization of the protein (E251) (34, 36). The C-terminal forkhead (FKH) domain is 91% similar to the mammalian domain. It includes the sites of the nuclear localization signals essential for the translocation of Foxp3 to the nucleus (R347 and aa 414–417) (34), which are 100% similar to the respective positions in the mammalian protein.

These findings prove structural homology of the newly assembled chicken Foxp3 gene and its previously described orthologs. In addition, they hint on possibly conserved functions despite considerable differences in the sequence and in domain organization.

Our next approach aimed to determine the expression pattern of chicken Foxp3 in relevant lymphocyte populations. We sort purified cells from spleen, cecal tonsil, and thymus and analyzed them using real-time RT-PCR (Fig. 3). We first distinguished chicken T cells according to the human T_reg phenotype (CD4⁺CD25⁺Foxp3⁺) (14) by analyzing CD4⁺CD25⁺ and CD4⁺CD25⁻ lymphocyte subsets. Chicken Foxp3 mRNA was highly abundant in CD25⁺ but not in CD25⁻ subsets of CD4⁺ lymphocytes from secondary lymphatic organs (Fig. 3A, left and left-middle panels, dark gray bars).

As CD25 expression in CD4⁺ splenocytes was quite variable, we performed a separate analysis of Foxp3 expression for CD25⁻, CD25^low, and CD25^high subsets of chicken CD4⁺ splenocytes (Supplemental Fig. 4). We found that CD25^low cells express 66-fold more Foxp3 than does the CD25⁻ population. Furthermore, in CD25^high cells, Foxp3 expression is 130-fold increased as compared with CD25⁻ cells (Supplemental Fig. 4).

Thymic CD4⁺CD25⁺ cells expressed very low levels of Foxp3 (Fig. 3A, right-middle panel, dark gray bars). Taken together, chicken Foxp3 is expressed considerably by CD4⁺CD25⁺ cells in secondary lymphatic organs, but not in the thymus.

To investigate possibly deviating pathways of Foxp3 induction, we next investigated the CD4⁻ thymocyte population. Much to our surprise, we found that, in association with CD25, CD4⁻ thymocytes expressed very high levels of Foxp3 (Fig. 3A, right panel, dark gray bars). To exclude cytotoxic T cells as the source of the detected Foxp3 in this CD4⁻ subset, analysis of splenic CD8⁺ lymphocytes was carried out and revealed very low Foxp3 mRNA levels (data not shown). Due to very low cell numbers, a sort of cytotoxic T cells for the activation marker, that is, of CD8⁺CD25⁺ lymphocytes, was not possible from any of the examined lymphocyte sources. However, real-time RT-PCR analysis of the CD4⁻CD25⁺ thymocyte subset showed CD8 expression levels equal to those of any CD4⁺ peripheral lymphocyte subsets and way below those of CD25⁻ thymocytes (Fig. 3B). Furthermore, B cells from secondary lymphatic organs were also excluded as Foxp3-expressing cells in chickens by analysis of AV20⁺ lymphocyte populations (Fig. 3C, dark gray bars). Thus, Foxp3 expression seems to be restricted mainly to CD25⁺ subsets of CD4⁺ peripheral and CD4⁻CD8⁻ thymic T cells (Fig. 3D). Further examination of the Foxp3-expressing thymocytes revealed that only 6.64% of these double-negative (DN) CD25⁺ cells were CD3⁺ (Supplemental Fig. 3).

IL-2 is expressed in effector T cells and conveys their activation via the IL-2 receptor complex, of which CD25 is a part. Besides other mechanisms, T_regs suppress IL-2 expression. IL-10 and TGF-β are the major suppressive cytokines used by mammalian T_regs to achieve regulatory function. To investigate the cytokine pattern of putative chicken T_regs, we measured IL-2, IL-10, and TGF-β mRNA levels in the sort-purified lymphocyte subsets mentioned above. IL-10 expression was mostly limited to CD4⁺CD25⁺ subsets in secondary lymphatic organs and to CD4⁻CD25⁺ thymocytes (light gray bars in (Fig. 3A). As described above, this pattern was in line with the expression of Foxp3. IL-2, however, showed no signs of association with Foxp3, but was expressed only by CD4⁺CD25⁺ cells from thymus, and, to a much lesser extent, from cecal tonsil (white bars in (Fig. 3A). B cells expressed neither IL-2 nor IL-10 (Fig. 3C). Surprisingly, TGF-β was not differentially expressed between any of the analyzed subsets (data not shown). These results show that IL-10 and IL-2 expression levels in chicken T cell subsets generally resemble their mammalian counterparts, whereas TGF-β expression is not associated with Foxp3 or with CD25. These data suggest the existence of several distinct T cell phenotypes associated with regulatory pathways in chickens.

Future analyses of avian T_regs must extend to the protein level. Therefore, specific mAbs are indispensable. To assess cross-reactivity with chicken Foxp3, we hence transiently transfected HEK293T cells with full-length mouse or chicken Foxp3, respectively, and tested binding of selected anti-mouse Foxp3 mAbs with proven mammalian interspecies cross-reactivity by cytoplasmic staining. Using the mouse Foxp3-transfected cells as a positive control, we confirmed that both Abs bound mouse Foxp3 in the experimental setup used in the present study. However, no cross-reactivity with the chicken protein could be observed (Fig. 4).

The ongoing search for “hidden” genes in birds is rewarded with increasing success due to progression in fields of sequencing techniques and bioinformatics. Using a large chicken RNA sequencing dataset, we were able to manually assemble a gene sequence homologous to mammalian and reptilian Foxp3, as well as to the sequence automatically annotated as Foxp3 in the ground tit (P. humilis) genome published in 2013 (11). Phylogenetic analysis and sequence alignment revealed unique characteristics of Foxp3 among the Foxp family members. The various analyzed Foxp3 sequences cluster to one clade. Furthermore, the chicken sequence clusters closest with the automatically annotated sequence of the ground tit, followed by alligator Foxp3. This mirrors the proven common archosaur origin of the crocodilian reptiles and avian species (37) and thereby emphasizes the accuracy of the chicken Foxp3 assembly (Fig. 1). PCR amplification of the assembled Foxp3 sequence from chicken tissue mRNA subsequently proved the gene’s transcription in vivo. In addition, we succeeded in mapping Ccdc22, a gene within the syntenic locus of mammalian Foxp3, to the same long sequencing read as the assembled chicken Foxp3 sequence. We thereby prove conserved synteny (Supplemental Fig. 2). The Foxp3 clade, however, displayed the highest heterogeneity compared with the clades of Foxp1, Foxp2, or Foxp4 proteins, respectively: only the two included mammalian species, mice and humans, showed vast similarity for Foxp3. Nevertheless, these species’ Foxp3 sequences showed higher numbers of substitutions per site than did the other three members of the Foxp family between these species. In the course of evolution, Foxp3 therefore appears to have been the subject of fundamental structural changes. This assumption is underlined by the amount of guanine and cytosine residues present in Foxp3 transcripts (Supplemental Fig. 1). It is known that structural damage to DNA is more likely to be repaired using G or C rather than A or T bases (7). As the class of birds is phylogenetically older than mammals, their genomes have been subject to a higher accumulated mutation probability throughout evolution. Consequently, high GC content is observable especially in immune-related genes, which are liable to a high selection pressure (1). This led to delayed annotation of Foxp3 in birds, because the strong triple covalent bonds between G and C impair amplification. Thereby, high GC content, especially in combination with long stretches of only G and C, significantly reduces sequencing accuracy using classical methods (2). Such sequences are also predicted to form various secondary structures that might impede analysis (2, 38). However, in this study we present the first successful full-length assembly of a chicken homolog of mammalian Foxp3.

Due to the comprehensive remodeling of Foxp3 since the split of the avian and mammalian lineages from the common ancestor, Foxp3 proteins in modern species are likely to vary in domain organization. To investigate this hypothesis further, we aligned the translation of the newly assembled chicken Foxp3 transcript sequence with Foxp3 of five other vertebrates (Fig. 2). The highest consensus is present at the C-terminal FKH domain. It is, in mammals, responsible for binding of the transcription factor Foxp3 to DNA, particularly conveyed by the nuclear localization signals on both ends of the domain (34). Both sites are conserved in the chicken sequence, indicating Foxp3’s ancient function as a transcription factor. However, single residues with predicted crucial roles in NFAT interaction and DNA binding (33) are only partially conserved, suggesting distinct class-specific differences for functional pathways that Foxp3 is involved in. The FKH domain is connected to the N-terminal effector domain by a leucine zipper, which is also highly conserved among all species. It conveys dimerization. To influence the transcription of >700 genes in total (39), Foxp3 forms large multiprotein complexes. The initial dimerization, however, results in homodimers or in heterodimers with Foxp1 (40, 41). The glutamic acid residue at position 251, within the leucine zipper domain, is crucial for these dimerization events. We found E251 to be conserved in the chicken sequence, rendering a conservation of Foxp3’s overall quaternary structure in this species likely. The zinc finger domain, in contrast, does not seem to be present in any non-mammalian species we investigated, which supports findings by an earlier study on Foxp3 ontogeny (33). However, a nonredundant role of the Foxp3 zinc finger domain has been doubted by a study examining truncated versions of the protein that reported unrestricted Foxp3 functionality regardless of the presence or absence of the zinc finger (34). The repressive function itself is conveyed by the N-terminal proline-rich domain (34). This domain distinguishes Foxp3 from the other members of the Foxp family, which are characterized by a glutamic acid–rich N terminus. Several studies describe functional motifs within the N-terminal domain (42, 43), but the exact mechanisms by which the proline content affects repressor function remain to be elucidated (44). However, even though consensus of the six examined vertebrate species is poor in the N-terminal region, the chicken protein features an even higher proline content than for the mammalian sequences. This finding confirms results of a previous study on avian Foxp3, which assembled partial sequences for two falcon species (12). Taken together, the comparison of Foxp3 sequences among species implies that Foxp3 adapted its functional repertoire throughout evolution to meet newly arisen requirements of avian and mammalian physiology, respectively. Nevertheless, structural similarities indicate that Foxp3 fulfilled certain basic tasks, which are likely to be of repressive nature, ever since the common vertebrate ancestor (33). Therefore, chicken Foxp3 promises to be an excellent candidate for studying the protein’s nonredundant functions in the future.

The common development of all T cell subsets takes place in the thymus. It proceeds from a CD4/CD8 DN via a DP to finally an SP state. These mature, primed SP cells are (revocably) committed to a T cell lineage before they leave the thymus and enter peripheral lymphatic organs, for example, spleen or cecal tonsil. For murine T_regs, lineage determination is characterized by the expression of CD4, CD25, and Foxp3 (14). Hence, we examined Foxp3 expression in lymphocyte subsets that had been extracted from lymphatic organs and sort purified for the presence or absence of surface CD4 and CD25. We found that Foxp3 expression is associated with CD4⁺CD25⁺ subpopulations in peripheral lymphatic organs, that is, spleen and cecal tonsil, but not in the thymus. As the thymus represents the major organ for T cell development, this finding raised the question on the origin of Foxp3⁺ avian T cells. To further investigate this issue, we subsequently analyzed CD4⁻ thymic lymphocytes. Within this subset, surprisingly, we detected high Foxp3 expression in the rare CD25⁺ cells (Fig. 3A, dark gray bars). We could not detect any CD8 expression in this population, whereas the CD25⁻ subsets of CD4⁻ as well as CD4⁺ thymocytes were highly CD8⁺ (Fig. 3B). This distribution of CD8 to the Foxp3⁻ subsets only suggests a minor role of Foxp3-expressing CD8⁺ T cells in chickens. In line with what is known about mammalian thymic T cell development, the CD25⁻ (and CD8⁺) subpopulations of CD4⁺ or CD4⁻ lymphocytes are therefore likely to represent the CD4/CD8 DP and CD8 SP developmental stages, respectively. We also ruled out B cells as a source of Foxp3 in the CD4⁻ population by analysis of AV20⁺ subsets. These cells did not express Foxp3 (Fig. 3C, dark gray bars). From these results, we conclude that Foxp3 is mainly expressed by DN CD25⁺ T cells in the chicken thymus and in a subset of CD4⁺CD25⁺ cells in secondary lymphatic organs.

In order to draw a more comprehensive picture of the phenotype of the T cell subsets examined, we next investigated the expression of the marker cytokines IL-10 and IL-2 (Fig. 3A, 3C, light gray and white bars, respectively). In mammals, IL-10 is the major repressive cytokine secreted by T_regs. It suppresses the transcription of IL-2 in Th cells, which is required for the generation of all T cell subsets and for maintenance of Th cells themselves (45, 46). IL-2 binds to naive CD4⁺ T cells expressing the IL-2 receptor complex, part of which is CD25, and induces determination toward either the helper or regulatory lineage, depending on the presence or absence of proinflammatory cytokines (47). We could show on the mRNA level that, also in chickens, IL-10 expression was strongly associated with Foxp3, whereas IL-2 was expressed mostly by the Foxp3⁻ thymic CD4⁺CD25⁺ population and to a lesser extent by peripheral CD25⁺ lymphocytes. As it has been shown that CD25 can act as a naive T cell activation marker in mouse (8) and also in chicken γδ T cells (48), these findings appear to emphasize that mammalian and avian T cell populations share vastly similar phenotypes. The chicken lymphocyte populations from secondary lymphatic organs very likely display the overall cytokine pattern of activated Th cells or mature T_regs, in line with low or very high expression of Foxp3, respectively. These subsets could not be distinguished from each other in this study, which resulted in a mixed CD4⁺CD8⁻CD25⁺Foxp3⁺IL-10⁺IL-2^low population in peripheral lymphatic organs. We hypothesize that this mixed population consists, on the one hand, of one subset of CD4⁺CD25⁺ lymphocytes, which is low in IL-10 and intermediate in IL-2 expression (Fig. 3D). This phenotype resembles murine peripherally activated Th cells (8). The second subset, which presumably merged into the peripheral CD4⁺CD25⁺ sort population, expresses high levels of Foxp3 and IL-10 and therefore resembles mature T_regs. As we were able to show that Foxp3 is expressed most strongly in CD25^high splenocytes (Supplemental Fig. 4), the T_reg subset is likely to be found among this subpopulation.

Another subset of thymic CD25⁺ lymphocytes, on the other hand, indicates a developmental course of chicken T_regs that differs in one decisive point from that in mammals; that is, the vast majority of Foxp3⁺ thymocytes are CD4⁻ (Fig. 3A). We hypothesize that these cells represent immature nT_regs, which in other species do express CD4 (49). Apart from this substantial difference, we found evidence that the classical T_reg developmental cascade in the thymus resembles that between the mouse and chicken: in immature T_regs, IL-2 interaction with CD25 seems to have triggered Foxp3 induction and, in line with that, facilitated IL-10 expression. Due to their immaturity, also chicken thymic T_regs (CD4⁻CD25⁺Foxp3⁺) were yet comparably low in CD25. As opposed to that, Th cells in the thymus, that is, CD4⁺CD25⁺Foxp3⁻ thymocytes, were CD25^high, which indicates that they had been subject to strong activating signals. This confirms our expectations, as the thymus is the location of positive T cell selection via TCR affinity to self-antigens, as shown in mice (50). The chicken CD4⁺CD25^high thymic lymphocyte population is therefore likely to represent naive Th cells in the act of priming. This subset expresses high amounts of IL-2, which is crucial for the thymic generation of nT_regs (51). TGF-β is another important T_reg marker cytokine in mammals (52). In addition, it has recently been discussed as a substitute marker for chicken T_regs in Marek’s disease, defining a T_reg subset mostly distinct from CD4⁺CD25⁺ T cells (25). The study found CD4⁺TGF-β⁺ cells in in vitro–activated lymphocyte preparations from spleen, caecal tonsil, and blood of chickens susceptible to Marek’s disease, but not in the thymus. The authors concluded that TGF-β is a marker for peripherally induced chicken T_regs (25). Surprisingly, we could not detect any differential expression of TGF-β between the analyzed lymphocyte subsets on the mRNA level (data not shown). Nonetheless, we cannot exclude a specific role of TGF-β⁺ T cells with regulatory function in certain diseases. We therefore conclude that the importance of TGF-β for chicken nT_regs is neglectable under physiological conditions.

However, the role of the thymic CD4⁻CD25⁺Foxp3^highIL-10^highIL-2^low subset remained unclear at this point. In mice, Foxp3 is induced through TCR signaling and with the help of the coreceptors CD4 or CD8, respectively. Therefore, Foxp3 induction cannot occur before the DP state (49). In a human study, however, a small portion of DN thymocytes was shown to express Foxp3, along with low amounts of CD25, questioning the overall TCR dependency of Foxp3 induction (53). To test whether association of Foxp3 with the expression of a functional TCR in chickens is obligatory, we analyzed the subset of potential pre-T_regs (DN CD25⁺Foxp3^highIL-10^highIL-2^low) for surface expression of CD3 (Supplemental Fig. 3). CD3 serves as co-TCR and as such is part of any functional TCR complex. It is hence essential for activation of T cells; its absence on the cell surface excludes TCR signaling in the respective cell (54). We observed that the vast majority (93%) of assumed pre-T_regs did indeed not express CD3. However, a small proportion of the examined population is CD3⁺ and therefore possesses the tools for TCR signaling. Without appropriate tools on our side, namely a specific anti-chicken Foxp3 Ab, we have no means to determine whether Foxp3 is expressed in the small CD3⁺ or the large CD3⁻ subpopulation. Thus, it yet remains unclear whether the presumed pre-T_reg population has not yet undergone T cell maturation. In this case, the population would be negative for surface CD3 and thereby TCR, and the expression of Foxp3 is hence TCR-independent. A comparably low expression level of CD25 in this subset, which is in line with data for human cells expressing Foxp3 in a TCR-independent manner (53), argues for this hypothesis. In addition, Foxp3 would be expected to be stronger diluted and thus not show very high expression levels (compare with (Fig. 3) in the overall population if only a small fraction of the examined population (7%) would express it. Nevertheless, another possibility is that somewhere in the course of their differentiation, the DN CD25⁺ thymocytes lost CD4 and/or CD8 expression after going through the classical maturation pathway and TCR-dependent Foxp3 induction. In that case, Foxp3⁺ cells would be found in the small CD3⁺ subpopulation, whereas the remaining 97% of this population would represent DN thymocytes activated throughout the course of their maturation toward CD4⁺ or CD8⁺ cells. However, a recent comprehensive study showed that there are at least two distinct pathways for the development of mature T_regs in humans (55). We can therefore also not exclude the option that both of the above-mentioned hypotheses or additional ways that are not discussed in the present study represent the actual events. Hence, the developmental pathways of chicken T_regs remain to be elucidated.

To further investigate the development and phenotype of Foxp3-expressing cells, appropriate tools are required, first and foremost specific anti-chicken Foxp3 Abs. As we showed within this study, chicken Foxp3 contains highly conserved domains as well as such that are highly deviant from other species’ Foxp3. Likely due to the very high conservation of the FKH domain between members of the Fox protein family, and consequently low specificity of this domain for Foxp3, we did not find any commercially available Abs with proven epitopes within the C-terminal region. We therefore tested a pair of anti-mouse Foxp3 mAbs against epitopes within the N terminus of murine Foxp3 for cross-reactivity with the recombinant chicken protein (Fig. 4). The Abs we chose to investigate in this concern are the clones FJK-16s and MF-14. FJK-16s is a standard tool used to study Foxp3 in a number of mammalian species, for example, swine (56), dogs (57), mice (58), and cats (59). Besides FJK-16s, MF-14 is frequently used for mouse Foxp3 studies. Their epitopes (aa 75–125 or aa 72–107 of the mouse protein, for FJK-16s and MF-14, respectively) are located within an area encoded by exon 2 and partly by exon 3, a part of the highly variate repressor domain. As expected, none of the two Ab clones possesses affinity to the chicken protein. A previous study on cross-reactive mAbs showed that only 2 out of 182 mAbs raised against various human leukocyte proteins cross-reacted with chicken leukocytes (60). The mentioned study, however, concerned itself only with Abs against surface molecules. mAbs raised against partially highly conserved intracellular proteins might be more likely to cross-react. Therefore, we cannot exclude the existence of Abs with sufficient affinity against chicken Foxp3. However, the two most widely used anti-murine Foxp3 Ab clones, FJK-16s and MF-14, are not suitable for Foxp3 detection in chickens. The next step toward a comprehensive assessment of the chicken T_reg phenotype will therefore be the generation of a species-specific anti-chicken Foxp3 mAb.

In this study we identify an ortholog of Foxp3 in the chicken genome. We show structural homology as well as a number of features deviating from mammalian sequences and conclude that chicken Foxp3 is a model protein to study nonredundant Foxp3 function. We furthermore examine the phenotype of Foxp3-expressing chicken lymphocytes in secondary lymphatic organs and the thymus. From the results of these experiments we deduct the hypothesis that, except for CD4 expression in immature T_regs, chicken and mammalian T_regs are vastly similar in phenotype and are likely to resemble each other also in function and development. By revealing this similarity, this study allows for, and demands, a new definition of avian T_regs on the background of what is known about Foxp3 in other species. It thus paves the way toward standardized avian T_reg research.

We thank Prof. Thomas Brocker and Prof. Gerhild Wildner for providing FJK-16s anti-mouse Foxp3 Abs, and Prof. Stefan Endres for providing MF-14 anti-mouse Foxp3 Abs. In addition, we thank Marina Kohn for expert technical assistance and Prof. Dr. Steffen Weigend (Institute of Farm Animal Genetics, Friedrich-Loeffler-Institut Mariensee) for providing M11 birds.

The authors have no financial conflicts of interest.