embryonic day 12.5
expressed sequence tag
phosphatase 1 regulatory inhibitory subunit 16B
procollagen C-proteinase enhancer 2
procollagen C-proteinase enhancer 1
thymic epithelial cell
Ulex europaeus agglutinin 1
The adaptive immune system relies on the thymic microenvironment for the production of a diverse, self-tolerant T cell receptor repertoire. The central cellular organizer of the thymic microenvironment is the thymic epithelial cell (TEC). The development of TEC from endodermal precursor cells is under the control of the forkhead/winged helix transcription factor Foxn1 but the transcriptional program that leads to this unique epithelial differentiation has not been investigated functionally. Here, we show that expression of procollagen C-proteinase enhancer 2 (PCOLCE2) is absent in Foxn1-deficient TEC. In order to study the functional role of this gene in TEC differentiation, we have genetically inactivated PCOLCE2 and the gene encoding phosphatase 1 regulatory inhibitory subunit 16B (mPPP1R16B), another transcript lacking in Foxn1-deficient TEC. Mice deficient for either one or both of these transcripts presented a normal thymic microenvironment and undisturbed thymopoiesis. While there is no evidence for a functional role of PCOLCE2 and mPPP1R16B in thymus development, our results suggest that the lack of thymopoiesis in Foxn1-deficient mice is caused by multiple functional defects.
The adaptive immune system fights infections and confers immunity by generating effector cells that carry a diverse antigen receptor repertoire. Antigen receptor generation is achieved by the V(D)J recombination system that combines pieces of genomic fragments in a basically random fashion to produce the large number of antigen receptor specificities that can respond to the vast number of potential antigen targets. This process of antigen receptor generation carries the risk of producing antigen receptors that recognize self components and therefore has to be contained by a quality control system 1. For T cells, the thymic microenvironment is the site were immature stages of T cell development mature and potentially self-reactive antigen receptor specificities are removed.
The resident cell type of the thymic microenvironment, the thymic epithelial cell (TEC), emanates from the third pharyngeal pouch endoderm and organizes a functional niche by recruiting and instructing neural crest-derived mesenchymal cells and fetal liver/bone marrow-derived hematopoietic cells 2. We have recently shown that this process of recruiting and instructing non-epithelial cell types can even occur after birth, i.e. long after the physiological initiation of thymus organogenesis, and can be initiated by single TEC precursors 3.
This view of thymus organogenesis places the TEC in the center of the complex interplay between several cell types that form the thymic microenvironment. The formation of the epithelial stroma in the thymus can be functionally separated into two steps: initiation and differentiation 4. In thymus development, organ initiation comprises the mechanisms that lead to the invagination of the third pharyngeal pouch endoderm, the deposition of epithelial cell precursors in the underlying mesenchyme and the separation of the thymic organ anlage from the parathyroid anlage 5. In the next step, epithelial precursors acquire their special differentiation as TEC and go on to adopt the fate of either a medullary or a cortical TEC. These two types of TEC play functionally distinct roles in T cell repertoire selection in that cortical TEC mediate positive selection of developing thymocytes while medullary TEC play a major role in the deletion of self-reactive T lymphocytes, a process termed negative selection, by the expression of tissue-specific antigens 6, 7. Using surface markers, several TEC subsets can be distinguished, one of which appears to be of special interest as it expresses high levels of MHC class II antigen, turns over rapidly and expresses the highest levels of costimulatory molecules and tissue-specific antigens 8.
TEC differentiation is under the control of the forkhead/winged helix transcription factor Foxn1 9, but how TEC acquire specific medullary or cortical characteristics remains elusive. Lack of Foxn1 leads to alopecia and thymic aplasia in mouse and man but the molecular basis for this striking phenotype is poorly understood. In skin, it has been shown that the expression of several hair keratins is under the control of Foxn1, which may explain part of the skin phenotype 10, but the Foxn1-induced transcriptional program that leads to the specific TEC differentiation remains elusive. In the absence of Foxn1, the thymic anlage degenerates into epithelial cysts made up by progenitor-like epithelial cells 3, 11, some of which present characteristics of respiratory epithelium 12. Thus, Foxn1 appears to initiate a transcriptional program that supports the differentiation of TEC while suppressing the phenotypic characteristics of non-TEC fates. Given the central role of TEC differentiation for the generation of a diverse, self-tolerant T cell repertoire, understanding its molecular basis is of prime importance.
In an attempt to identify potential Foxn1 target genes that are part of the transcriptional program that results in the differentiation and proliferation of functional TEC, we have previously reported the comparison of the transcriptome of nude and wild-type thymic epithelial anlagen by highly parallel microarray analysis 13. While all of the differentially expressed transcripts derived from the few hematopoietic cells in the embryonic day 12.5 (E12.5) thymic anlage were found to be encoded by annotated genes of known function, all but one of the transcripts derived from TEC were defined only by expressed sequence tags (EST) that could not be attributed to a known gene. Here, we report the functional analysis of two of these transcripts in the formation of a functional thymic microenvironment in vivo.
Results and discussion
PCOLCE2 expression is missing in Foxn1-deficient thymic epithelial cells
We previously reported the identification of genes that are regulated by the nude locus gene product, Foxn1 13. By profiling the expression of 22 000 transcripts in the E12.5 thymic anlage of wild-type and nude mice and confirming differential expression by both quantitative RT-PCR and in situ hybridization, we identified eight transcripts. Three transcripts (germ-line TCRγ, granzyme A and LEI) were found to be of hematopoietic and five (PD-L1 and the novel genes WDT1–WDT4) of epithelial origin. Additionally, after careful re-analysis and several attempts of in situ hybridization, a transcript encoding the procollagen C-proteinase enhancer 2 (PCOLCE2) protein 14, 15 could be confirmed as a differentially expressed epithelial transcript (Fig. 1). PCOLCE2 expression is reduced 14-fold in the nude E12.5 thymic anlage as shown by both microarray analysis and quantitative RT-PCR (Fig. 1A). By in situ hybridization, PCOLCE2 is undetectable in the nude thymic anlage and shows an epithelial expression pattern (Fig. 1B), which is easily distinguishable from the hematopoietic pattern that is caused by the earliest immigrating hematopoietic precursors 13. In contrast to the homogeneous staining restricted to the thymic anlage found for epithelial transcripts, hematopoietic transcripts show a punctate staining within and outside the thymic anlage 13. Consistent with the epithelial pattern, sorted adult TEC but not sorted thymocytes express this transcript (Fig. 1C). In conclusion, the data identify PCOLCE2 as a transcript that is associated with normal TEC differentiation but fails to be expressed in nude epithelium. Thus, PCOLCE2 might be part of the transcriptional program that is initiated by the Foxn1 transcription factor in the wild-type epithelial anlage at E12.5 and maintained in adult TEC.
T cell development in the thymus of PCOLCE2-deficient mice
The finding that PCOLCE2 might be part of the Foxn1-induced transcriptional program suggested that it plays a role in TEC differentiation, a process for which an in vitro model assay is not available. It has been reported that the culture of TEC in vitro leads to the rapid loss of Foxn1 expression which can only be prevented by maintaining TEC in 3D cultures such as fetal thymic organ cultures 16. As a first approach, we therefore attempted to knock down the PCOLCE2 transcript in vitro by delivering morpholino oligonucleotides to TEC in fetal thymic organ cultures. This appeared to be a valid approach since the closely related procollagen C-proteinase enhancer 1 (PCPE1) gene that could have replaced PCOLCE2 function is not expressed in the thymic anlage (Fig. 1D), ruling out the possibility that PCPE1 could compensate for decreased PCOLCE2 levels in TEC. Unfortunately, several attempts to knock down epithelial transcripts with well-characterized roles in thymopoiesis, like IL-7, by the delivery of morpholino oligonucleotides into TEC were unsuccessful.
It has to be pointed out that this approach could only have identified a role for PCOLCE2 in maintaining the integrity of an embryonic thymus for the few days of the culture period, but it would have been unable to identify a potential role in the formation of a functional thymus microenvironment. Therefore, the only way to conclusively address such a role was to investigate the in vivo role of PCOLCE2 in the thymus by the genetic inactivation of the PCOLCE2 gene in mice. The PCOLCE2 expression pattern in adult tissues has been shown to be restricted to heart, bladder, mammary gland and trachea 15. PCPE1 is much more widely expressed and was found to be present in these organs in addition to PCOLCE2. We therefore argued that the risk of generating an embryonic lethal phenotype should be low and opted for a straight knockout approach.
Functional inactivation of the PCOLCE2 gene was achieved by replacing large parts of exon 3 that encodes the first of two CUB domains that are thought to be important in protein-protein interactions by a reporter/neomycin resistance cassette (Fig. 2A–E). Correctly targeted ES cell lines were identified and injected into blastocysts for germ-line transmission. The targeted allele was transmitted according to Mendelian distribution and was found to be a null allele, as PCOLCE2 transcript was detectable in thymus stroma RNA of wild-type and heterozygous but not PCOLCE2-deficient mice (Fig. 2E). Furthermore, since exons 2 and 4 are not in phase and the genomic PCR shown in Fig. 2E remained negative in PCOLCE2-deficient mice, we found no evidence for the possibility that the targeted exon 3 is spliced out to yield an aberrant transcript with some residual function. The findings indicate that exon 3 is replaced by the reporter/neomycin cassette in the aberrant PCOLCE2 transcript that therefore fails to encode a functional CUB domain.
PCOLCE2-deficient mice were viable, fertile and did not show gross developmental abnormalities. Analyzing thymopoiesis in PCOLCE2-deficient mice, we found no evidence for abnormal thymus or T cell development (Fig. 2F–H). The thymus of PCOLCE2-deficient mice is located orthotopically in the mediastinum, is of normal size and shows the typical compartmentalization into cortical and medullary areas with normal distribution of the markers cytokeratin 5 (K5) and cytokeratin 8 (K8) (Fig. 2H) and MTS-10, Ulex europaeus agglutinin 1 (UEA-1) and ERTR-7 (data not shown). T cells develop normally in PCOLCE2-deficient thymi. The different stages of mainstream αβ T cell development were found in the expected ratios (Fig. 2F) and these cells are exported in normal numbers to the periphery (Fig. 2F, G). All minor populations in the thymus, including NKT cells, γδ T cells, regulatory T cells, B cells and recent thymic emigrants were found to be present at normal numbers (data not shown). Thus, the absence of the Foxn1-dependent transcript PCOLCE2 in TEC has no detectable effect on thymus development.
The mPPP1R16B transcript is also absent in Foxn1-deficient TEC
In parallel to PCOLCE2, we analyzed the in vivo role of another transcript identified in our screen, namely WDT4 13, in thymus development. Like PCOLCE2, WDT4 shows an epithelial expression pattern in the thymic organ anlage of Foxn1-sufficient but not -deficient mice (Fig. 3A, B) and it encodes a protein with predicted protein-protein interaction domains. The foremost reason, however, to study this transcript was the recent demonstration that its expression is regulated by TGF-β signaling 17. Indeed, there is good evidence that signaling through TGF-β family members is involved in thymus development 18, 19.
The full-length sequence of WDT4, which was represented by oligonucleotide sequence of the EST cluster TC40806 on the microarray 13, was determined by 5′ prime RACE and screening of a genomic phage library. The resulting 6.4-kb transcript was found to encode a 569-amino acid protein with five predicted ankyrin repeats which are thought to mediate protein-protein interactions. Homology of the predicted protein was found to be highest to regulatory subunits of protein phosphatase 1 that target this serine/threonine phosphatase to distinct subcellular locations or specific substrates 20. Based on sequence homology, the WDT4 gene product has now been annotated as murine protein phosphatase 1 regulatory inhibitory subunit 16B (mPPP1R16B) 17, 21. A functional role has so far not been attributed. Collectively, the mPPP1R16B transcript might be part of the Foxn1-dependent transcriptional program in TEC.
Genetic inactivation of mPPP1R16B
To study the functional role of mPPP1R16B in the Foxn1 transcriptional program, we genetically inactivated the mPPP1R16B gene in mice by homologous recombination in ES cells. This appeared to be reasonable since mPPP1R16B shows a thymus-restricted expression pattern in the E12.5 embryo (Fig. 3B). Gene targeting was done by replacing exon 4, which encodes the first ankyrin repeat in mPPP1R16B, by a neomycin resistance cassette (Fig. 3C–G). Correctly targeted ES cell lines were identified and injected into blastocysts for germ-line transmission. The targeted allele was transmitted according to Mendelian distribution.
A wild-type mPPP1R16B transcript was undetectable in the thymic stroma of mPPP1R16B-deficient mice (Fig. 3G), indicating that the mPPP1R16B gene had been disrupted. Instead, a mutant mRNA, apparently reduced just by the size of exon 4, was found. From the mPPP1R16B sequence, translation from this mutant mRNA is predicted to end after two aberrant amino acids at a stop codon in exon 5. Thus, the aberrant transcript encodes only two of the five ankyrin repeats that are known to be required to regulate the function of the protein phosphatase 1 catalytic subunit in the closely related regulatory protein M11020. Thus, targeting the mPPP1R16B locus resulted in a null allele.
The mPPP1R16B-deficient mice were viable, fertile and did not show gross developmental abnormalities. The analysis of thymopoiesis in mPPP1R16B-deficient mice revealed no evidence for abnormal thymus or T cell development (Fig. 3H–K). The thymus of mPPP1R16B-deficient mice is of normal size with all the signs of a normal microenvironment as evidenced by the normal distribution of K5 and K8 (Fig. 3K), MTS-10, UEA-1 and ERTR-7 (data not shown). T cell development was found to be normal in mPPP1R16B-deficient mice. All stages of αβ T cell development were found displaying normal phenotype in the expected ratios (Fig. 3H) and the same is true for NKT cells, γδ T cells, regulatory T cells, B cells and recent thymic emigrants (data not shown). Mature T lymphocytes were exported to the periphery in mPPP1R16B-deficient mice like in wild-type controls (Fig. 3I, H). Collectively, the absence of mPPP1R16B in TEC has no detectable effect on the formation of a functional thymus.
T cell development in PCOLCE2/mPPP1R16B double-deficient mice
The transcriptional program that is initiated by Foxn1 in TEC during embryogenesis and that is maintained in adult TEC results in the formation of a functional thymic microenvironment that supports proper selection and differentiation of T lymphocytes. Since both PCOLCE2 and mPPP1R16B might be part of this program, we hypothesized that only the simultaneous absence of both transcripts might be sufficient to affect TEC differentiation to a measurable degree. We therefore generated double-deficient mice and analyzed their thymus for structural and functional abnormalities.
PCOLCE2/mPPP1R16B double-deficient thymi were found to be of normal size with stromal cells and lymphoid cells located in the correct thymic niches (Fig. 4A, B). No difference was found for the different lymphoid subsets in the thymus between PCOLCE2/mPPP1R16B double-deficient mice and wild-type mice (Table 1). Consistently, we find that the repertoire of Vβ chains used in PCOLCE2/mPPP1R16B double-deficient mice is indistinguishable from that of wild-type mice (Fig. 4C), indicating that the double-deficient microenvironment selects a normal T cell receptor repertoire. To address the question whether mature T cells in double-deficient mice were functional, we immunized mice with the T cell-dependent antigen chicken gammaglobulin-4-hydroxy-3-nitrophenylacetyl (CG-NP) and found that the resulting IgG response was the same as in double-heterozygous controls (Fig. 5). Based on these studies, we did not find any evidence for abnormal thymus development even in PCOLCE2/mPPP1R16B double-deficient mice.
|DP thymocytes (% of total thymocytes)||90.3 ± 3.31||90.48 ± 1.7|
|CD4 SP thymocytes( % of total thymocytes)||4.41 ± 1.6||4.27 ± 0.36|
|CD8 SP thymocytes (% of total thymocytes)||1.52 ± 0.82||1.55 ± 0.6|
|DN thymocytes (% of total thymocytes)||0.91 ± 0.11||0.87 ± 0.19|
|TN1 thymocytes (% of TN cells)||6.2 ± 0.6||6.18 ± 1.54|
|TN2 thymocytes (% of TN cells)||0.96 ± 0.75||0.59 ± 0.2|
|TN3 thymocytes (% of TN cells)||49.43 ± 4.24||51.05 ± 3.67|
|TN4 thymocytes (% of TN cells)||33.03 ± 5.8||33.25 ± 5.27|
|Recent thymic emigrants (% of CD4 SP cells)||26.18 ± 8.34||28.65 ± 9.9|
|Thymic Treg (% of total thymocytes)||0.29 ± 0.15||0.26 ± 0.06|
|Thymic NKT cells (% of HSA– thymocytes)||3.87 ± 1.17||4.81 ± 1.97|
|Thymic B cells (% of total thymocytes)||0.04 ± 0.02||0.04 ± 0.02|
|Thymic γδ T cells(% of total thymocytes)||0.08 ± 0.03||0.08 ± 0.02|
|CD4+ splenocytes (% of total splenocytes)||16.96 ± 7.32||14.29 ± 5.45|
|CD8+ splenocytes (% of total splenocytes)||11.7 ± 4.07||11.82 ± 1.52|
Our previous work has demonstrated that subtle abnormalities in thymic stroma, as in the case of LTβR-deficient mice 22, had little effect on T cell development as such but caused the production of auto-antibodies in affected mice. We therefore searched for auto-antibodies in the sera of 6–10-wk-old double-deficient mice and double-heterozygous controls but could not find any signs of auto-antibody production in these mice (Table 2). Thus, we found no structural or functional abnormalities in the thymus of mice that lack both transcripts, PCOLCE2 and mPPP1R16B.
|Stomach||Pancreas||Salivary gland||Ovary||Liver||Adrenal gland||Kidney|
|PCOLCE2/PPP1R16B double-heterozygous control #25||+||+||0||0||0||0||0|
|PCOLCE2/PPP1R16B double-heterozygous control #28||0||0||0||0||0||0||0|
|PCOLCE2/PPP1R16B double-heterozygous control #40||0||0||0||0||0||0||0|
|PCOLCE2/PPP1R16B double-heterozygous control #41||0||0||0||0||0||0||0|
|PCOLCE2/PPP1R16B double-KO #23||0||0||0||0||0||0||0|
|PCOLCE2/PPP1R16B double-KO #50||0||0||0||0||0||0||0|
|PCOLCE2/PPP1R16B double-KO #52||0||0||0||0||0||0||0|
Our studies suggest that the normal differentiation of TEC does not depend on PCOCLE2 and on PPP1R16B and hint at the robustness of this differentiation program. From a practical point of view our data lend support to the notion that mutiple genetic deficiencies need to be combined in order to phenocopy the effect of the loss of the transcription factor Foxn1 in TEC.
Materials and methods
PCOLCE2 and PPP1R16B genomic clones from a 129/SvJ BAC library (Genome Systems Inc., St. Louis, MO) were used to determine the genomic structure of the two genes under investigation. For PCOLCE2, a targeting construct was prepared that replaces coding sequence of exon 3 corresponding to nucleotides 258 to 431 of the PCOLCE2 mRNA (BC051174) by an IRES-lacZ/neomycin resistance cassette but preserves the splice acceptor and donor of the exon. The 5′ prime and 3′ prime flanking arms comprised 1.6 kb and 7.3 kb of genomic sequence, respectively. NotI-linearized construct was electroporated into R1 ES cells. For PPP1R16B, a targeting construct was prepared that replaces 643 nucleotides of genomic sequence corresponding to nucleotides 59952 to 60594 of AL663077 that encode exon 4 by a neomycin resistance cassette. The 5′ prime and 3′ prime flanking arms comprised 1.7 kb and 7 kb of genomic sequence, respectively. XbaI-linearized construct was electroporated into R1 ES cells.
For both constructs, correctly targeted ES cell clones were identified by Southern blotting, investigated for off-target integration and injected into C57BL/6 blastocysts. Mice were kept under specific pathogen-free conditions in the mouse facility of the Max-Planck-Institute of Immunobiology. Experimentation and animal care was in accordance with the guidelines of the Max-Planck-Institute of Immunobiology. Genotyping was done on genomic DNA from tail biopsies by PCR using the following primers: for PCOLCE2, 5′-CCATCTAATGAGATTAGTCC-3′, 5′-TGTACCGTCATCTTGTTGCC-3′ and 5′-CCCGGGATCAACTACC-3′; for PPP1R16B, 5′-CTGTGTCAGACACAGGCATC-3′, 5′-CTCATTACCCTTTTGTTGGG-3′ and 5′-CGACTGCATCTGCGTGTTCG-3′. Analyzed mice were 4–6 wk old unless indicated otherwise.
Quantitative PCR and RT-PCR
Quantitative PCR was carried out on a LightCycler System (Roche) as described 13. RT-PCR was carried out on RNA isolated using TRI reagent (Sigma) from thymus stroma-enriched material using the primers CB312 5′-TAGCCCCAGGCAGAGCAAGG-3′, CB415A 5′-CTGAAATCATCTTCTGCTCAGG-3′, CB648 5′-ATTCTTACCGGAGAGTCTGG-3′, CB649 5′-AAGGCGTCCTCCACAATACC-3′, CB650 5′-CGTATCTGCAGTAGTTATCGC-3′ according to a previously reported protocol 23. Control samples without reverse transcriptase remained negative throughout.
In situ hybridization and immunohistochemistry
In situ hybridization was carried out as described 23 and immunohistochemistry was done according to a published protocol 22 using Troma-1 (anti-K8; a gift by Rolf Kemler, Max-Planck-Institute of Immunobiology), polyclonal anti-mouse K5 antibody (Covance), UEA-1 (Vector Laboratories) and the following reagents from BD Biosciences: anti-CD4 (GK1.5), anti-CD8α (53-6.7) and anti-B220 (RA3-6B2).
FACS analysis was carried out on single-cell suspensions from thymus, spleen and lymph nodes and Vβ usage was determined as previously described 22, 24. Percentages for a given population were determined by electronic gating using the Flowjo Software Package (Tree Star).
Mice were immunized with 100 µg of CG-NP (a kind gift by Gleb Turchinovich, Max-Planck-Institute of Immunobiology) in aluminum hydroxide (Alu-Gel-S; Serva, Heidelberg, Germany). Serum samples were obtained from mice by tail bleeding. Fourteen days after immunization, IgG1-specific antibody responses were measured by ELISA using BSA-NP (a kind gift by Gleb Turchinovich, Max-Planck-Institute of Immunobiology) and goat anti-mouse IgG1 antiserum (Southern Biotechnology Associates) according to standard protocols.
We are indebted to Sandra Groß, Mahela Konrath and Christiane Happe for excellent technical support, Andreas Würch for FACS sorting, Benoit Kanzler and Elsa Lopez for blastocyst injections and Thomas Boehm for his support and critical comments on the manuscript. K.H. was supported by a grant of the Deutsche Forschungsgemeinschaft (SFB620, A7).