Hoxa3 and Pax1 Regulate Epithelial Cell Death and Proliferation During Thymus and Parathyroid Organogenesis

Developmental Biology 236, 316–329 (2001) doi:10.1006/dbio.2001.0342, available online at http://www.idealibrary.com on Hoxa3 and Pax1 Regulate Epithelial Cell Death and Proliferation during Thymus and Parathyroid Organogenesis

Dong-ming Su, Steve Ellis,1 Audrey Napier,2 Kristin Lee, and Nancy R. Manley3 Institute of Molecular Medicine and Genetics and Department of Pediatrics, Medical College of Georgia, Augusta, Georgia 30912

The thymus and parathyroid glands in mice develop from a thymus/parathyroid primordium that forms from the endoderm of the third pharyngeal pouch. We investigated the molecular mechanisms that promote this unique process in which two distinct organs form from a single primordium, using mice mutant for Hoxa3 and Pax1. Thymic ectopia in Hoxa3؉/؊Pax1؊/؊ compound mutants is due to delayed separation of the thymus/parathyroid primordium from the pharynx. The primordium is hypoplastic at its formation, and has increased levels of apoptosis. The developing third pouch in Hoxa3؉/؊Pax1؊/؊ compound mutants initiates normal expression of the parathyroid-specific Gcm2 and thymus-specific Foxn1 genes. However, Gcm2 expression is reduced at E11.5 in Pax1؊/؊ single mutants, and further reduced or absent in Hoxa3؉/؊Pax1؊/؊ compound mutants. Subsequent to organ-specific differentiation from the shared primordium, both the parathyroids and thymus developed defects. Parathyroids in compound mutants were smaller at their formation, and absent at later stages. Parathyroids were also reduced in Pax1؊/؊ mutants, revealing a new function for Pax1 in parathyroid organogenesis. Thymic hypoplasia at later fetal stages in compound mutants was associated with increased death and decreased proliferation of thymic epithelial cells. Our results suggest that a Hoxa3–Pax1 genetic pathway is required for both epithelial cell growth and differentiation throughout thymus and parathyroid organogenesis. © 2001 Academic Press Key Words: Hox; Pax; Gcm2; thymus; parathyroid; pharyngeal pouch; endoderm; organogenesis; proliferation; apoptosis.

INTRODUCTION don et al., 2001). Gcm2 and Foxn1 expression in the primordium predates the morphological separation of para- Most of the pharyngeal glandular organs originate from thyroid and thymus domains by two full days of develop- transient outpocketings of the pharyngeal endoderm called ment. However, the mechanisms controlling the initial pharyngeal pouches. The thymus and parathyroid glands in formation of the common primordia and subsequent devel- mice develop from bilateral shared thymus/parathyroid opment of the thymus and parathyroids are just beginning primordia arising from interactions between the third pha- to be understood. ryngeal pouch endoderm and surrounding neural crest cells Mutations in several transcription factors including (reviewed in Manley, 2000). Although the parathyroid- and Hoxa3, Pax1, Pax9, and Foxn1 cause defects in thymus thymus-speciﬁc domains in the early primordia are mor- organogenesis. Hoxa3 is a member of the Hox family of phologically indistinguishable, each primordium is divided transcription factors, which specify positional identity in into organ-speciﬁc regions marked by the expression of the the developing embryo (Krumlauf, 1994). Hoxa3 homozy- transcription factors Gcm2 and Foxn1, respectively (Gor- gous mutants have the most severe thymic defects of the

1 mutants listed above, as they are athymic and aparathyroid, Current address: Gene Evaluation and Mapping Laboratory, and fail to initiate formation of the thymus/parathyroid USDA Agricultural Research Center, Beltsville, MD 20705. 2 Current address: Department of Biological Sciences, Alabama primordia (Chisaka and Capecchi, 1991; Manley and Capec- State University, Montgomery, AL 36101-2071. chi, 1995, 1998; S.E., J. Koushik, J. Gordon, and N.R.M., 3 To whom correspondence should be addressed. Fax: (706) 721- submitted for publication). The closely related paired box 8685; E-mail: [email protected]. transcription factors Pax1 and Pax9 are also required for

0012-1606/01 $35.00 Copyright © 2001 by Academic Press 316 All rights of reproduction in any form reserved. Genetic Control of Thymus/Parathyroid Development 317 normal thymus development (Dietrich and Gruss, 1995; 2000). Hoxa3ϩ/ϪPax1ϩ/Ϫ double heterozygotes have a thy- Peters et al., 1998; Wallin et al., 1996). Loss of Pax1 mus phenotype indistinguishable from Pax1 single mu- function results in a hypoplastic thymus that is deficient in tants, indicating a strong genetic interaction between these thymocyte development (Dietrich and Gruss, 1995; Wallin two loci. Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutant newborns et al., 1996). Loss of Pax9 function results in an early failure have a hypoplastic, ectopic thymus with much more severe of both thymus and parathyroid organogenesis, although defects in thymocyte development than Pax1 single mu- the exact nature and timing of this phenotype is not yet tants. Chimera analysis showed that thymocyte develop- clear (Neubu¨ser et al., 1995; Peters et al., 1998). Foxn1 ment defects were due to defective TEC function. These (previously whn/Hfh11; Kaestner et al., 2000) is the gene results indicate that Hoxa3 and Pax1 (and presumably mutated in the classical nude mouse strain, and encodes a Pax9) are in a common pathway regulating thymus devel- transcription factor of the winged helix/forkhead class opment and TEC differentiation. (Nehls et al., 1994). Foxn1 is required cell-autonomously for In the current study, we investigated the mechanisms thymic epithelial cell differentiation (Blackburn et al., directing the development of the thymus/parathyroid pri- 1996), but is not required for initiation of thymus organo- mordium and subsequent organogenesis of the thymus and genesis (Nehls et al., 1996). Consistent with this function, parathyroid glands using the Hoxa3 and Pax1 mutants. The Foxn1 is first expressed in the common primordium after it thymus/parathyroid primordia in Hoxa3ϩ/ϪPax1Ϫ/Ϫ com- forms and is restricted to the more caudal and ventral pound mutants form at the correct location and develop- portion of the primordium, where it marks the presumptive mental time. However, the primordia are hypoplastic from thymus-specific region (Gordon et al., 2001). their earliest stage of development and their separation Less is known about the molecular mechanisms support- from the pharynx is delayed by up to a full day. These ing parathyroid organogenesis and differentiation. Both phenotypes are associated with a significant increase in Hoxa3 and Pax9 mutants affect parathyroid development in apoptosis throughout the thymus/parathyroid primordium combination with other pharyngeal region defects (Chisaka at the initiation of organogenesis. Gcm2 and Foxn1 expres- and Capecchi, 1991; Peters et al., 1998). Recently, Gcm2 sion are both initiated normally in Hoxa3ϩ/ϪPax1Ϫ/Ϫ com- has been shown to be specifically required for parathyroid pound mutants, showing that the developing third pouch is development (Gunther et al., 2000). One of two mammalian specified correctly into parathyroid and thymus domains homologues of the Drosophila gene Glial cells missing, (Gordon et al., 2001). However, Gcm2 expression is reduced Gcm2 encodes a transcription factor with a novel DNA in Pax1Ϫ/Ϫ single mutants by E11.5, and further reduced or binding domain (Akiyama et al., 1996; Kim et al., 1998). absent in compound mutants, suggesting that maintenance Gcm2 mutants have no identifiable parathyroid glands, but of Gcm2 expression and subsequent parathyroid develop- have apparently normal thymus development (Gunther et ment may be dependent on both Hoxa3 and Pax1. al., 2000). Gcm2 expression begins at E9.5 in the caudal After the common primordium divided into distinct pharyngeal pouches and is progressively restricted to a thymic and parathyroid rudiments at E13.5, the thymic small domain within the third pharyngeal pouch endoderm lobes themselves continued to be hypoplastic and ectopic. by E10.5 (Gordon et al., 2001). Within the developing Later, thymic organogenesis defects included increased cell common thymus/parathyroid primordium, Gcm2 expres- death and decreased proliferation of TECs at E15.5 as sion marks the presumptive parathyroid domain in a measured by cell cycle analysis. In addition, Hoxa3 contin- complementary pattern to the thymus marker Foxn1 (Gor- ues to be expressed in thymic epithelial cells at E15.5 and don et al., 2001), and is subsequently restricted to the even in adult thymus. This expression not only suggests developing parathyroid glands (Gunther et al., 2000). Gcm2 that TEC defects at later fetal stages are direct effects of the expression is absent in Hoxa3Ϫ/Ϫ embryos at E10.5, suggest- Hoxa3 and Pax1 mutations, but raises the possibility that ing it may be regulated by Hoxa3 (S.E., J. Koushik, J. Hoxa3 is involved in adult thymus function. Parathyroid Gordon, and N.R.M., submitted for publication). organogenesis was also abnormal, resulting in initial hyp- Our previous studies suggested that there is a functional oplasia and subsequent loss of parathyroid glands in com- link between Hoxa3 and Pax1 in thymus development pound mutants. Furthermore, we identified a novel defect (Manley and Capecchi, 1995; Su and Manley, 2000). Hoxa3 in parathyroid development in Pax1 single mutants. Our and Pax1 are both expressed in the third pharyngeal pouch results indicate that Hoxa3 and Pax1 act both at the endoderm, and Pax1 continues to be expressed in thymic initiation of thymus/parathyroid organogenesis and epithelial cells (TECs) throughout thymus development throughout fetal thymus development to regulate thymus (Manley and Capecchi, 1995; Wallin et al., 1996). Pax1 and and parathyroid epithelial cell differentiation. Pax9 are both down regulated in the third pharyngeal pouch in Hoxa3Ϫ/Ϫ embryos at embryonic day 10.5 (E10.5) before morphological differences are present, indicating that they MATERIALS AND METHODS act downstream of Hoxa3 (Manley and Capecchi, 1995; S.E., J. Koushik, J. Gordon, and N.R.M., submitted for publica- Mice and Genotyping tion). Furthermore, mutations in Hoxa3 and Pax1 have The Hoxa3 mutant strain and genotyping by PCR have been synergistic effects on thymus development (Su and Manley, described (Chisaka and Capecchi, 1991; Manley and Capecchi,

FIG. 1. Three-dimensional reconstructions of E12.5 pharyngeal regions. The pharynx including the esophagus and trachea are shown in yellow, thymus in blue, and ultimobranchial bodies (UBs) in green. Individual transverse H&E stained sections from the series used to generate the reconstructions at each of the levels indicated (i–iii) are shown below each reconstruction. (A) Pax1Ϫ/Ϫ embryo showing normal pharyngeal region development. (B) Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutant littermate, with normal pharyngeal and UB development, but delayed separation of the thymic lobes and thymic hypoplasia. The gap in the UB is the result of damage to the sections used in the reconstruction.

FIG. 2. Increased cell death in Hoxa3ϩ/ϪPax1Ϫ/Ϫ embryos at E12.5. H&E (A, B) and TUNEL (C–H)-stained 6-␮m parafﬁn transverse sections of E12.5 control (A, C, E) and littermate Pax1Ϫ/Ϫ (F) and Hoxa3ϩ/ϪPax1Ϫ/Ϫ (B, D, G) embryos. (A–B), (C–D), and (E–G) show sections of thymic lobes from embryos from the same litter. Thymic lobes in each panel are outlined to increase visibility. (A) Wild-type thymic primordium. (B) Thymic lobe from a littermate Hoxa3ϩ/ϪPax1Ϫ/Ϫ embryo, with many apparent apoptotic bodies (arrows). (C, E) Pax1ϩ/Ϫ heterozygous thymic lobes with few TUNELϩ cells. (F) Pax1Ϫ/Ϫ thymic lobes are similar to controls. (D, G) Hoxa3ϩ/ϪPax1Ϫ/Ϫ thymic lobes with very high levels of apoptosis, much higher than was seen in controls. Scale bar, 75 ␮m.

1995). For the Pax1 mutant, we used the Pax1un-ex allele, a deletion were used at 0.5 ␮g/ml. Alkaline phosphatase-conjugated anti- mutant that is viable and fertile as a homozygote, and is pheno- digoxigenin Fab fragments were used at 1:5000. BM-purple (Roche/ typically identical to a Pax1-null allele generated by gene targeting BMB) was used as a chromagen to localize hybridized probe. (Dietrich and Gruss, 1995; Wallin et al., 1996; Wilm et al., 1998). Pax1un-ex mice and embryos were also genotyped by PCR as previously described (Su and Manley, 2000). For simplicity, we will refer Cytokeratin and Propidium Iodide Staining and to Pax1un-ex/un-ex mice as Pax1Ϫ/Ϫ. The day of the vaginal plug was FACS Analysis designated as E0.5. Embryonic stages were confirmed by counting somites, and by evaluating eye development and limb morphology. Thymic lobes were dissected from E15.5 mouse embryos. Single- Both mouse strains are maintained as congenic on a C57BL/6 cell suspensions were made by incubating thymic lobes in 1 ml of genetic background. 0.25% Trypsin/1 mM EDTA at 37°C for 25 min, triturated by pipetting 10–20 times, then incubated again at 37°C for 5 min, and a final pipetting 10–20 times. Cells were collected by gentle Histology and Three-Dimensional Reconstructions centrifugation, then resuspended in a drop (30–50 ␮l) of PBS with 1 Embryos were fixed in 4% formaldehyde and paraffin-embedded mM EDTA. Cells were fixed by slowly adding 1.5 ml of ice-cold by using standard protocols. For three-dimensional reconstruc- 70% EtOH, then incubating on ice for 1 h. Cells were then rinsed tions, serial sections were cut at 6 ␮m, stained with hematoxylin in PBS and resuspended in staining buffer (2% FCS/PBS/0.05% and eosin, and photographed with a Spot digital camera on an NaN3). Cells were stained for 30 min with FITC-conjugated anti- Olympus BX60 microscope. Digital micrographs were aligned and pan cytokeratin antibody (Sigma), then washed with staining buffer ␮ structures to be reconstructed were traced in each section by using and PBS. Cells were then incubated with 0.5 ml 50 g/ml pro- ␮ Adobe Photoshop. The Amira software package (TGS Inc., San pidium iodide in PBS and 50 g/ml RNaseA at 37°C in the dark for Diego, CA) was used to create three-dimensional reconstructions of 30 min. Data generation and analyses were performed by using a the tracings. FACStar Plus flow cytometer with CellQuest software. Significant differences were determined by the unpaired Student’s t test or ANOVA analysis. TUNEL Assay

Staged embryos were fixed in 4% formaldehyde and processed RT-PCR for paraffin embedding by using standard techniques. Sections were cut at 6 ␮m and stained with hematoxylin and eosin (H&E), or RNA was prepared from E15.5–E16.5 day embryonic or adult processed for TUNEL. TUNEL reactions (Gavrieli et al., 1992) were thymus, either from whole thymus or from highly enriched thymic performed using direct detection of fluorescent nucleotide incor- epithelial cells (TECs) prepared by culturing in deoxyguanosine to poration on paraffin sections. Sections were dewaxed through a remove hematopoetic cells. To isolate TECs, thymic lobes from xylene-EtOH series and treated with proteinase K at 10 ␮g/ml for E15.5 embryos were isolated and cultured at 37°C in high-oxygen 15 min at 37°C. Terminal transferase reactions containing 20 ␮M submersion culture as described previously (Kishi et al., 1995; Su et ␮ ϫ tetramethylrhodamine-dUTP, 20 M dTTP, 1.5 mM CoCl2,1 al., 1997). Lobes were cultured in 1 ml RPMI medium with 10% TdT buffer (BMB), and 1 unit/100 ␮l TdT (BMB) were performed for fetal bovine serum and 5 ϫ 10Ϫ5 M 2-mercaptoethamol containing 1 h at 37°C. Sections were washed between steps in phosphate- 1.35 ␮M deoxyguanosine (dGuo) for 5 days to deplete thymocytes buffered saline. Sections were counterstained with DAPI and and dendritic cells (Jenkinson et al., 1982; Ready and Jenkinson, mounted in ProLong Anti-Fade reagent (Molecular Probes). 1987). Whole or dGuo-treated thymic lobes were homogenized in Analysis of TUNEL-stained sections was performed on a Power TRIzol (Gibco) by using a freshly autoclaved Micro Pellet Pestle Macintosh 7600 computer by using the NIH Image program (de- (KONTES) followed by a 5-min incubation at room temperature. veloped at the U.S. National Institutes of Health and available on Chloroform was added (1/5 ratio of TRIzol), and samples were the Internet at http://rsb.nih.gov/nih-image/). Cell counts and area vortexed for 15 s. After a 2- to 3-min incubation at room tempera- measurements were performed on serial sections by using digitized ture, the samples were centrifuged at 12,000g for 15 min at 4°C. ϩ/Ϫ Ϫ/Ϫ Ϫ/Ϫ photomicrographs of thymic lobes from Hoxa3 Pax1 , Pax1 , The top phase containing total thymus RNA was harvested into a ϩ/Ϫ and Pax1 E12.5 littermates from two different litters. Compari- fresh tube. The RNA was precipitated using isopropyl alcohol. sons were restricted to littermates to control for age differences Total thymus RNA samples were either left at Ϫ70°C in 75% between litters. Both lobes from each embryo were analyzed and ethanol or dissolved in 20 ␮l of DEPC-treated water. each lobe was considered separately. TUNEL-positive cells were For RT-PCR, genomic DNA was removed from total RNA identified and counted with a density slice. At least 90% of preparations by using DNase I. The DNase I enzyme was inacti- sections from each lobe were measured and counted; sections were vated using 25 mM EDTA and heated for 15 min at 65°C. Reverse excluded due to tissue damage or if TUNEL staining outside the transcription was performed by using the GIBCO SuperScript II thymus in the same section was not detectable. Average numbers method. cDNA was subjected to 30 cycles of 94°C (0.5 min), 62°C of TUNEL-positive cells and average areas for thymus sections (1.0 min), and 70°C (1.0 min) using Qiagen PCR Taq polymerase. were compared by using ANOVA analysis. Each sample was screened for Hoxa3 and CD45 by using the following primers: Hoxa3 forward 5Ј-TGCCAGCACAGCCAAG- Ј Ј In Situ Hybridization AGCCCCC-3 ; Hoxa3 reverse 5 -GGCCATCTCCACCCGGC- GCGG-3Ј; CD45 forward 5Ј-TGCTCCTCAAACTTCGACGGA-3Ј; Whole-mount in situ hybridizations were performed as described CD45 reverse 5Ј-CTGCCAAAAGTCACCGATCGT-3Ј. These prim- (Carpenter et al., 1993; Manley and Capecchi, 1995). Probes for ers generate predicted products of 254 bp for Hoxa3 and 655 bp for Pax9 (Neubu¨ser et al., 1995), Foxn1, and Gcm2 (Gordon et al., CD45. Primers for both Hoxa3 and CD45 cross an intron, and the 2001) were previously described. Digoxigenin-labeled RNA probes specificity of the PCR products were confirmed by sequencing.

␤-actin primers generating a 91-bp product were previously described apoptosis than in controls (Figs. 2A and 2B). Condensed and (Mansour et al., 1993). fragmented nuclei were seen throughout both common primordia of Hoxa3ϩ/ϪPax1Ϫ/Ϫ embryos at this stage (Fig. 2B). TUNEL staining of three additional E12.5 Hoxa3ϩ/Ϫ- RESULTS Pax1Ϫ/Ϫ embryos and three Pax1ϩ/Ϫ littermate controls confirmed that these represented apoptotic cells, and that Delayed Separation of the Thymus/Parathyroid these cells were more numerous in compound mutants Primordium in Compound Mutants Ϫ/Ϫ (Figs. 2D and 2G) than in controls (Figs. 2C and 2E). Pax1 We used histological analysis and three-dimensional re- littermates were identical to heterozygous controls (Fig. constructions to determine the time course and morpho- 2F), indicating that this phenotype was specific to com- logical characteristics of pharyngeal organ development. pound mutants. Quantitation of the percentage of TUNEL- The thymus and parathyroids in mice originate from shared positive cells in Hoxa3ϩ/ϪPax1Ϫ/Ϫ thymic lobes at E12.5 bilateral thymus/parathyroid primordia that develop from showed a significant increase of 2- to 6-fold relative to the third pharyngeal pouches. At E11.5, the thymus/ Pax1ϩ/Ϫ and Pax1Ϫ/Ϫ littermates (P Ͻ 0.05). Since increased parathyroid primordia have formed but are still attached to apoptosis was seen throughout the common thymus/ the pharyngeal endoderm (data not shown). By E12.5, the parathyroid primordia, it could provide a mechanism for thymus/parathyroid primordia (in blue) and the ultimo- both thymus and parathyroid hypoplasia in the compound branchial bodies (UBs; in green) that form from the fourth mutants. pouch had separated completely from the pharynx in both controls and Pax1 single mutants (Fig. 1A). The thymus/ parathyroid primordia had begun their migration toward Gcm2 and Foxn1 Expression in Pax1 Single the anterior thoracic cavity (Fig. 1A). At this stage, the Ϫ Ϫ Mutants and Hoxa3-Pax1 Compound Mutants organ primordia in Pax1 / mutants were similar in size and Ϫ/Ϫ location to controls; Pax1 mutants were not significantly To investigate potential molecular mechanisms underly- different from controls until after E13.5 (data not shown). ing the observed morphological changes in the thymus/ ϩ/Ϫ Ϫ/Ϫ In contrast, defects were apparent in Hoxa3 Pax1 parathyroid primordia, we examined the expression of compound mutants as early as E11.5, just after the thymus/ Foxn1 and Gcm2. Foxn1 and Gcm2 are required for thymus parathyroid primordia form. Even at this early stage, the ϩ/Ϫ Ϫ/Ϫ and parathyroid differentiation, respectively (Gunther et Hoxa3 Pax1 mutant primordia were already smaller al., 2000; Nehls et al., 1996), and their expression patterns than in controls (not shown). At E12.5, the organ primordia prefigure division of the shared primordium into distinct Hoxa3ϩ/ϪPax1Ϫ/Ϫ in embryos remained smaller than in thymus and parathyroid organ rudiments (Gordon et al., controls and Pax1 single mutants. In some cases, one or 2001). Changes in the expression of these genes before the both primordia were still attached to the pharynx, although appearance of overt phenotypes could provide a possible the pharynx and trachea (in yellow) and UBs appeared mechanism for the observed defects. Foxn1 was expressed normal (Fig. 1B). These results suggested that the thymic ϩ/Ϫ Ϫ/Ϫ in the presumptive thymus-specific region of the thymus/ ectopia seen at E17.5 in Hoxa3 Pax1 compound mu- Ϫ Ϫ parathyroid primordia at E11.5 in control, Pax1 / , and tants (Su and Manley, 2000) is most likely caused by ϩ Ϫ Ϫ Ϫ Hoxa3 / Pax1 / compound mutant embryos (Fig. 3A). delayed separation of the thymic/parathyroid primordium Both the timing and location of Foxn1 gene expression from the pharyngeal pouch. The phenotype in the com- appear normal. Gcm2 expression also appeared normal in pound mutants was variable, but always more severe than the third pouch of both control and Hoxa3ϩ/ϪPax1Ϫ/Ϫ com- that of Pax1 single mutants. We also noted that cysts which are apparent in the thymic lobes of both Pax1 single pound mutants at E10.5 (Figs. 3B and 3C). However, by mutants (Wallin et al., 1996) and Hoxa3ϩ/ϪPax1Ϫ/Ϫ com- E11.5, Gcm2 expression was variably but consistently reduced in Pax1 single mutant embryos (Fig. 3E) compared to pound mutants (Su and Manley, 2000) are not seen early in ϩ/Ϫ Ϫ/Ϫ organogenesis, but develop after E14.5 (not shown), and control embryos (Fig. 3D). Furthermore, in Hoxa3 Pax1 therefore may be a secondary consequence of abnormal compound mutants, Gcm2 expression was reduced even epithelial cell differentiation. further or absent (compare Fig. 3F with 3G). This loss was striking in contrast to the relatively normal expression of Foxn1 at this stage (Fig. 3A). These results show that the Increased Cell Death at Early Stages of developing third pouch is initially designated correctly into Organogenesis in Compound Mutants parathyroid and thymus-specific domains, and that thymic Ϫ Ϫ The histological analysis showed that the thymus/ epithelial cell fate is appropriately specified, in both Pax1 / ϩ Ϫ Ϫ Ϫ parathyroid primordia in Hoxa3ϩ/ϪPax1Ϫ/Ϫ embryos are hy- single and Hoxa3 / Pax1 / compound mutants. However, poplastic from a very early stage. A possible reason for this parathyroid development is specifically impaired within hypoplasia was seen by closer inspection of the H&E- the common thymus/parathyroid primordium as early as stained sections. In all seven Hoxa3ϩ/ϪPax1Ϫ/Ϫ E12–12.5 E11.5. This impairment is seen in Pax1 single mutants, and embryos examined, there was evidence of higher levels of is exacerbated in Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants.

FIG. 3. Expression of organ-speciﬁc marker genes in Pax1Ϫ/Ϫ and Hoxa3ϩ/ϪPax1Ϫ/Ϫ embryos. Whole mount in situ hybridization analysis of E10.5 (B, C) or hemisected E11.5 (A, D–G) embryos. Identity of probes used is indicated in the upper left corner, genotypes are indicated below the embryos. (A) Foxn1 expression is relatively normal in both Pax1Ϫ/Ϫ and Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutant embryos at E11.5. (B, C) Gcm2 expression at E10.5 is identical in controls (B) and Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants (C). (D, E) By E11.5, Gcm2 expression is noticably reduced in Pax1Ϫ/Ϫ embryos (E) relative to littermate controls (D). (F, G) This reduction is even more dramatic in Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants (G) relative to littermate controls (F).

Multiple Organogenesis Defects in Compound identifiable (Fig. 4B). To further investigate this phenotype, Mutants at E13.5 we looked for parathyroid glands in control, Pax1ϩ/Ϫ, and Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants at the newborn stage. By E13.5 in controls, the thymus (blue) and parathyroids Both parathyroid glands were easily found in wild-type and (red) were easily identifiable as separate structures (Fig. 4A). Hoxa3 heterozygous mice (Figs. 5A and 5B, and not shown). The thymic lobes had increased in size and migrated Ϫ/Ϫ substantially, with the most caudal point near the final In contrast, the parathyroids in Pax1 newborns were severely hypoplastic, although they were readily identified adult position of the thymus (Fig. 4A, ii). At E13.5, the Ϫ/Ϫ primordia in the Hoxa3ϩ/ϪPax1Ϫ/Ϫ mutants had all sepa- in all five E17.5 Pax1 embryos examined (Figs. 5C and Ϫ/Ϫ rated from the pharynx, and most had divided into distinct 5D). Parathyroids in Pax1 newborns had estimated vol- thymus and parathyroid rudiments (Figs. 4B and 4C). How- umes that were approximately 6- to 8-fold smaller than ever, mild to severe defects were seen in the parathyroids, controls. In contrast, no identifiable parathyroids were seen ϩ/Ϫ Ϫ/Ϫ ultimobranchial bodies, and thymus at this stage. in four out of five E17.5 Hoxa3 Pax1 compound mutants. In the one remaining E17.5 compound mutant, a

؊/؊ ؉/؊ ؊/؊ single epithelial structure that may have been a parathyroid Parathyroid Defects in Pax1 and Hoxa3 Pax1 remnant was seen next to the thymus; however, it had an Mutants abnormal, follicle-like epithelial cell structure (Figs. 5E and Parathyroid primordia (red) in compound mutants at 5F). The presence of parathyroid rudiments at E13.5 (Fig. 4) E13.5 were smaller than in controls, and were not always showed that this phenotype was not due to failure of initial

FIG. 4. Three-dimensional reconstructions of E13.5 pharyngeal regions from one control (A) and two littermate compound mutants (B, C). The esophagus and trachea are shown in yellow, thymus in blue, ultimobranchial bodies (UBs) in green, parathyroids in red, and thyroid in purple. Individual transverse H&E stained sections from the series used to generate the reconstructions at each of the levels indicated (i–iii) are shown below each reconstruction. (A) Wild-type control. Both parathyroids are identiﬁable, with one still closely contacting the thymus. The UBs have initiated fusion with the thyroid gland, and are therefore not shown in a separate color. (B) Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutant with a single hypoplastic parathyroid and unfused UBs. Both thymic lobes are smaller than control and have not yet reached the anterior thoracic cavity (B, iii). (C) A second Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutant from the same litter has a less severe phenotype. Both parathyroids are present although small, and one of the thymic lobes is located further posterior. The UBs have not yet fused with the thyroid, although they are closely associated with it.

Copyright © 2001 by Academic Press. All rights of reproduction in any form reserved. 324 Su et al. parathyroid rudiment formation from the thymus/ freshly isolated E15.5 cytokeratin-positive TEC stained parathyroid primordium. Of three Hoxa3ϩ/ϪPax1Ϫ/Ϫ com- with propidium iodide by flow cytometry (Figs. 6A–6C). pound mutants examined at E13.5, two had either one (Fig. Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants had significant 4B) or two (Fig. 4C) parathyroids that were clearly hypoplas- changes in the frequencies of both cell death and prolifera- tic, while the remaining one had no identifiable parathy- tion in TECs relative to control and Pax1Ϫ/Ϫ mutants at roids. These data show that Hoxa3 and Pax1 mutations had E15.5. Cell death was increased approximately 4-fold (P Ͻ profound and synergistic effects on parathyroid organogen- 0.01; Fig. 6D), while the percentage of proliferating cells esis, and are consistent with the Gcm2 expression patterns was decreased by 30% (P Ͻ 0.05; FIg. 6E). This result shows in these mutants. that both the survival and growth of TECs in Hoxa3ϩ/Ϫ- Pax1Ϫ/Ϫ compound mutants continued to be affected at later stages of fetal development. Ultimobranchial Body Development Is Relatively Normal in Compound Mutants Hoxa3 Is Expressed in TECs throughout Thymus UBs develop from the fourth pharyngeal pouch and fuse Organogenesis with the dorsal and anterior aspect of the main body of the thyroid gland, and subsequently form the calcitonin- The changes in TEC proliferation and cell death at E15.5 producing C cells in the thyroid (Moseley et al., 1968). suggested that the Hoxa3 and Pax1 mutations were having Hoxa3 mutants have defects in UB development, and some- a direct effect on later TEC development. Pax1 is expressed times lack UB-derived C cells (Manley and Capecchi, 1995), both in the pharyngeal endoderm and in TECs through although early UB organogenesis is normal (S.E., J. Koushik, mid-gestation (Wallin et al., 1996). Hoxa3 is expressed in J. Gordon, and N.R.M., submitted for publication). The UBs the pharyngeal pouch endoderm at the initiation of thymus in our reconstructed E13.5 control embryos had already organogenesis (Manley and Capecchi, 1995). However, fused to the thyroid, and are no longer visible as distinct Hoxa3 has not been shown to be expressed in the develop- structures (Fig. 4A). In contrast to the thymus and parathy- ing thymus at later fetal stages. We used RT-PCR to show roid phenotypes in Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants, that Hoxa3 is expressed in TECs at E15.5 (Fig. 7). TECs ultimobranchial body development appeared relatively nor- were prepared by deoxyguanosine treatment of E15.5 fetal mal, although there was a slight delay in the fusion of the thymic lobes (Jenkinson et al., 1982; Ready and Jenkinson, UBs to the thyroid proper (Figs. 4B and 4C). Ectopic or 1987). This treatment resulted in a dramatic reduction in persistent UBs were never seen at later stages in Hoxa3ϩ/Ϫ- hematopoietic cells, confirmed by reduced expression of the Pax1Ϫ/Ϫ compound mutants. hematopoetic-specific marker CD45 (Fig. 7). These results suggest that Hoxa3 continues to play a role in TEC development beyond the initiation of organogenesis, and that the Defects in Thymus Organogenesis at E13.5 in ϩ/Ϫ Ϫ/Ϫ defects in Hoxa3 Pax1 compound mutants at E15.5 Compound Mutants could be direct effects, rather than secondary effects of an After the thymic lobes in Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound earlier phenotype. To our surprise, Hoxa3 expression was mutants separated from the thymus/parathyroid primordia also seen in adult thymus RNA (Fig. 7). This later expres- at E13.5, they remained smaller than controls throughout sion implicates Hoxa3 in mature thymus function, as well the remainder of fetal development (Figs. 4B and 4C; Su and as in fetal thymus development. Manley, 2000). Although they had begun their caudal migration, most thymic lobes at this stage had not yet reached the anterior thoracic cavity (Figs. 4B, iii and 4C, iii). DISCUSSION Comparison of external morphology (including limb and tail morphology, eye pigmentation, and overall embryonic The development of the thymus and parathyroids from size) and embryonic body weights at E12.5, E15.5, and E17.5 the third pharyngeal pouch represents a unique program of showed no differences between genotypes (not shown). organogenesis. Our results provide evidence that the mo- Therefore, the observed phenotypes in the thymus size and lecular mechanisms directing this process, as well as those migration were not due to a general developmental retarda- required for later thymus and parathyroid growth and tion. differentiation, depend on the functions of the same set of transcription factors, including Hoxa3 and Pax1. Mutations in Hoxa3 and Pax1 have synergistic effects on the develop- Changes in TEC Death and Proliferation at E15.5 ment of these organs, suggesting that they are in a common in Compound Mutants ϩ/Ϫ genetic pathway. We have previously shown that Hoxa3 - The observed thymic hypoplasia in compound mutants at Pax1Ϫ/Ϫ compound mutants at E17.5 had ectopic, cystic, E13.5 either was due to a continuing defect in TEC devel- and hypoplastic thymic lobes (Su and Manley, 2000). In the opment, or was a consequence of the earlier defects at current study, we provide evidence for important functions initial primordium formation. To distinguish between for this pathway from the earliest stages of organogenesis these possibilities, we performed a cell cycle analysis of throughout fetal development.

Our data indicate that the thymic ectopia in Hoxa3ϩ/Ϫ- evidence that the reduction in TEC number in compound Pax1Ϫ/Ϫ compound mutants is caused primarily by delayed mutants is the result of both increased apoptosis and separation from the pharynx. These results suggest that the decreased proliferation of epithelial cells. Effects are seen at problem is not a defect in organ primordia migration per se, both initial organogenesis and at later fetal stages. Since but a result of delayed separation of the primordium from both Hoxa3 (this report) and Pax1 (Wallin et al., 1996) are the pharynx. By the time this separation has occurred, the expressed throughout fetal thymus development, both the environment through which the developing primordium early and later phenotypes could be direct effects of these must migrate has developed further, causing variable envi- mutations. ronmental delays or blocks in descent as a secondary effect. We were unable to detect a defect in TEC proliferation or Another possibility is that the ectopia is a secondary effect apoptosis in the Pax1 single mutants in this study. The lack of hypoplasia. However, other mutants that have small of a phenotype in the Pax1 single mutants in this assay is thymuses do not have ectopic thymic lobes, the nude also consistent with our previous results showing normal mouse being the best example (Cordier and Haumont, 1980; numbers of TECs in Pax1Ϫ/Ϫ mutants (Su and Manley, Schorpp et al., 2000). Also, the phenotype in the compound 2000). Taken together, these results show that the thymic mutants is variable, and the ectopia does not strictly hypoplasia in Pax1Ϫ/Ϫ mice is largely or entirely due to correlate with the degree of hypoplasia. Therefore, the reduced numbers of thymocytes. This result is also consis- weight of evidence suggests that delayed separation from tent with the relatively normal early development of the the pharynx is the primary cause of ectopia in the com- thymus in Pax1 mutants, since thymocytes are not numer- pound mutants. This conclusion is consistent with a model ous in the thymus until about E13.5. As we have also in which separation of the thymus/parathyroid primordium shown that hematopoietic cells in these mice are normal from the pharynx is independent of proliferation and (Su and Manley, 2000), the reduced numbers of thymocytes growth of the rudiment, and that Hox and Pax genes control reflect reduced TEC function. Therefore, the Pax1 mutation both of these processes. alone primarily affects TEC differentiation, rather than The thymic hypoplasia seen in the compound mutants is organogenesis. The role of Pax1 in thymus organogenesis is caused by increased epithelial cell death throughout fetal thus only revealed in combination with mutation of Hoxa3. thymus development, combined with decreased epithelial The relative mildness of the Pax1Ϫ/Ϫ thymus phenotype is proliferation. Mutations in Hox genes have been associated a strong contrast to the severity of the parathyroid pheno- with elevated apoptosis in the hindbrain in Hoxa1Ϫ/Ϫ and type in Pax1 single mutants. We also found evidence for a Hoxa1Ϫ/ϪHoxa2Ϫ/Ϫ mutants, although proliferation was un- link between Hoxa3 and Pax1 in parathyroid development. affected (Barrow et al., 2000; Rossel and Capecchi, 1999). Both Pax1 (this report) and Hoxa3 (Chisaka and Capecchi, Differences in proliferation have also been proposed as a 1991; Kovacs et al., 2001) single mutants have parathyroid mechanism for Hox gene patterning in the limbs and in defects, and in Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants para- cervical vertebrae (Condie and Capecchi, 1993; Davis et al., thyroid glands are apparently absent. The importance of 1995; Duboule, 1995; Goff and Tabin, 1997). Expression of Pax1 in parathyroid development is consistent with the Hox genes has been linked to regulating the balance of expression of Pax1 in the common thymus/parathyroid proliferation and differentiation, either by inhibiting prolif- primordium (Wallin et al., 1996), and with the defective eration to induce differentiation (Bromleigh and Freedman, parathyroid development in Pax9 homozygous mutants 2000), or promoting proliferation at the expense of differen- (Peters et al., 1998). Hoxa3 has an even more important role tiation (Salser and Kenyon, 1996; Sauvageau et al., 1995, in parathyroid development, as it is required for initiation 1997). Pax1 has also been shown to control proliferation; of parathyroid organogenesis (Chisaka and Capecchi, 1991; double mutants for Pax1 and Pax9 have also been shown to S.E., J. Koushik, J. Gordon, and N.R.M., submitted for have decreased proliferation and increased apoptosis in the publication). In addition, Hoxa3 heterozygotes have a de- developing sclerotome (Peters et al., 1999). In the case of crease in serum PTH levels (Kovacs et al., 2001), although thymus and parathyroid development, Hoxa3 and Pax1 the parathyroids are normal sized. This result indicates that would seem to promote proliferation, survival, and differ- Hoxa3 is haploinsufficient for parathyroid differentiation entiation in the developing primordium. Since Foxn1 ex- and/or function. pression appears relatively normal in the compound mu- The parathyroid defects in both Pax1Ϫ/Ϫ single mutants tants, the affect on apoptosis and proliferation may be and Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants may be mediated independent of effects on differentiation, at least at the through Gcm2. Gcm2 mutants have a profound defect in level of specifying the thymic primordium. parathyroid gland development (Gunther et al., 2000), and We have previously shown that Hoxa3ϩ/ϪPax1Ϫ/Ϫ com- Gcm2 expression is lost prior to primordium formation in pound mutants have defects in TEC differentiation that Hoxa3 single mutants (S.E., J. Koushik, J. Gordon, and result in an inability to promote normal thymocyte matu- N.R.M., submitted for publication). In both Pax1Ϫ/Ϫ single ration (Su and Manley, 2000). Effects on TEC development mutants and Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants, Gcm2 included a significant reduction in total TEC number, expression is initiated normally but then downregulated at reduced numbers of MHC class IIϩ TECs, and reduced level formation of the common primordium. This effect is more of MHC class II expression. The current study provides severe in compound mutants than in Pax1Ϫ/Ϫ single mu-

FIG. 5. Parathyroid defects in Pax1Ϫ/Ϫ and Hoxa3ϩ/ϪPax1Ϫ/Ϫ mutants at E17.5. H&E-stained transverse paraffin sections. (B), (D), and (E) are higher magnifications of sections shown in (A), (C), and (E). (A, B) Hoxa3ϩ/Ϫ control with normal parathyroid development. (C, D) Pax1Ϫ/Ϫ single mutant with severe parathyroid hypolasia. (E, F) Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutant with a possible parathyroid remnant. No other compound mutants analyzed had any identifiable parathyroid tissue. pt, parathyroid; ty, thyroid; th, thymus. Scale bars in A, C, E, 200 ␮m; in B, D, F, 50 ␮m.

tants. These results suggest that Hoxa3 is required for mutants, although it could be involved in the compound initial Gcm2 expression, but both Hoxa3 and Pax1 are mutant phenotype. Further analysis of the Gcm2 mutants required for maintenance of Gcm2. Reconstructions show a will be needed to provide deﬁnitive evidence for a causal variable failure of parathyroid formation from the common link between these two phenotypes. primordium, and no identiﬁable parathyroids were present Alternatively, apoptosis of the parathyroid precursors in at later embryonic stages. These results suggest that the Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants could be independent failure to maintain Gcm2 expression in Hoxa3ϩ/ϪPax1Ϫ/Ϫ of loss of Gcm2 expression. This possibility is supported by compound mutants may ultimately result in failure of the delay of up to 2 days between loss of Gcm2 expression parathyroid development. It is not known whether Gcm2 and failure of parathyroid development, by the fact that expression is required for survival of parathyroid cells. The apoptosis is seen throughout the thymus/parathyroid pri- lack of increased cell death in the E12.5 shared rudiment in mordium, and by the relatively normal Foxn1 expression at Pax1Ϫ/Ϫ mutants suggests that the loss of parathyroids is the same stage that Gcm2 is lost. Elegant and detailed not due to apoptosis of parathyroid precursors in these studies of homeobox gene function in both Drosophila

FIG. 6. Cell cycle analysis of thymic epithelial cells (TECs) by FACS. Single-cell suspensions from whole E15.5 thymic lobes were stained with a pan-cytokeratin antibody and propidium iodide for cell cycle analysis. (A, B) Control genotype showing the parameters used to set the cytokeratin-positive gate (R). (C) Cell cycle analysis of the cytokeratin-positive gated population (R), showing the gates used to measure the percentage of apoptotic cells (Ͻ2N, M1) and proliferating cells (S ϩ G2M, M2). (D) Percentage of cells in the M1 (Ͻ2N) population. Hoxa3ϩ/ϪPax1Ϫ/Ϫ compound mutants have an average level of apoptotic epithelial cells 5-fold higher than controls. (E) Percentage of cells ϩ/Ϫ Ϫ/Ϫ in the M2 (S ϩ G2M) population. Hoxa3 Pax1 compound mutants have a 30% decrease in proliferating epithelial cells. Asterisks (*) in (D) and (E) indicate signiﬁcant differences (P Ͻ 0.05) by Anova analysis. Each bar represents data from at least 10 separate embryos of each genotype; data is shown as the average value Ϯ SEM.

(Weatherbee et al., 1998) and Caenorhabditis elegans such a later role, at least during fetal stages. Similarly, our (Salser and Kenyon, 1996) have provided evidence that Hox data shows that the thymus defect in Pax1 single mutants genes may act at multiple points within the development of is primarily in TEC differentiation, with a role in thymus a cell type or structure, regulating cell death, proliferation, organogenesis revealed only in combination with the and differentiation at different points along a developmen- Hoxa3 mutation. However, the normal expression of Foxn1 tal pathway. Hoxa3 could be playing a similar role in in the compound mutants shows that speciﬁcation of the parathyroid and thymus development, promoting initial TEC lineage is unaffected. This result is consistent with the formation of the common primordium, expression of differ- lack of a genetic interaction between Hoxa3 and Foxn1 entiation markers, and regulating apoptosis and prolifera- (A.N., C. Raines, J. Chen, and N.R.M., unpublished data), tion at different times within the developing primordium. and is consistent with there being at least two separate Our data provide new information about the structure of pathways controlling fetal TEC differentiation. Further- the genetic pathway for parathyroid and thymus organogen- more, the expression of Hoxa3 in the adult thymus raises esis, and the role of Hoxa3 and Pax1 during embryonic the intriguing possibility that Hoxa3 may continue to development. The Hoxa3 homozygous mutant fails to ini- function in the thymus beyond fetal stages. tiate formation of the thymus/parathyroid primordium In contrast, there is so far evidence for only a single (S.E., J. Koushik, J. Gordon, and N.R.M., submitted for pathway for parathyroid development, which includes publication), preventing analysis of later roles for Hoxa3 in Hoxa3, Pax1/9, and Gcm2. Both initiation and mainte- thymus and parathyroid development. The compound mu- nance of Gcm2 expression appear to be under the control of tant analysis described here, combined with the expression this pathway. There is also evidence for a separate role for of Hoxa3 in the developing thymus, provides evidence for Hoxa3 in the differentiation or function of PTH-producing

Blackburn, C. C., Augustine, C. L., Li, R., Harvey, R. P., Malin, M. A., Boyd, R. L., Miller, J. F., and Morahan, G. (1996). The nu gene acts cell-autonomously and is required for differentiation of thymic epithelial progenitors. Proc. Natl. Acad. Sci. USA 93, 5742–5746. Bromleigh, V. C., and Freedman, L. P. (2000). p21 is a transcrip- tional target of HOXA10 in differentiating myelomonocytic cells. Genes Dev. 14, 2581–2586. Carpenter, E. M., Goddard, J. M., Chisaka, O., Manley, N. R., and Capecchi, M. R. (1993). Loss of Hox-A1 (Hox-1.6) function results in the reorganization of the murine hindbrain. Development 118, 1063–1075. Chisaka, O., and Capecchi, M. R. (1991). Regionally restricted developmental defects resulting from targeted disruption of the mouse homeobox gene hox-1.5. Nature 350, 473–479. Condie, B. G., and Capecchi, M. R. (1993). Mice homozygous for a targeted disruption of Hoxd-3 (Hox-4.1) exhibit anterior transfor- mations of the first and second cervical vertebrae, the atlas and the axis. Development 119, 579–595. Cordier, A. C., and Haumont, S. M. (1980). Development of thymus, parathyroids, and ultimobranchial bodies in NMRI and Nude mice. Am. J. Anat. 157, 227–263. Davis, A. P., Witte, D. P., Hsieh-Li, H. M., Potter, S. S., and FIG. 7. Expression of Hoxa3 in fetal and adult thymus and in TEC Capecchi, M. R. (1995). Absence of radius and ulna in mice by RT-PCR. Identities of RNA samples are indicated below the lacking hoxa-11 and hoxd-11. Nature 375, 791–795. figure. All samples were analyzed with (ϩ) or without (Ϫ)RT Dietrich, S., and Gruss, P. (1995). Undulated phenotypes suggest a treatment. Adult and E16.5 thymus samples are from whole role of Pax-1 for the development of vertebral and extravertebral thymus, E15.5 stroma is from dGuo-treated thymic lobes (see structures. Dev. Biol. 167, 529–548. Materials and Methods). The 254-bp Hoxa3-specific band was Duboule, D. (1995). Vertebrate Hox genes and proliferation: An amplified in all three RNA samples. The hematopoetic-specific alternative pathway to homeosis? Curr. Opin. Genet. Dev. 5, marker CD-45 was dramatically reduced by dGuo treatment, while 525–528. ␤ the Hoxa3-specific band remained relatively constant. -actin Gavrieli, Y., Sherman, Y., and Ben-Sasson, S. A. (1992). Identifica- primers generating a 91-bp product were used as a control. tion of programmed cell death in situ via specific labeling of nuclear DNA fragmentation. J. Cell Biol. 119, 493–501. Goff, D. J., and Tabin, C. J. (1997). Analysis of Hoxd-13 and Hoxd-11 misexpression in chick limb buds reveals that Hox cells (Kovacs et al., 2001), independent of Gcm2. Taken genes affect both bone condensation and growth. Development together, these results provide evidence for multiple roles 124, 627–636. for Hoxa3 during fetal thymus and parathyroid organogen- Gordon, J., Bennett, A., Blackburn, C., and Manley, N. (2001). esis and differentiation, and in mature organ function. Gcm2 and Foxn1 mark early parathyroid- and thymus-specific domains in the developing third pharyngeal pouch. Mech. Dev. 103, 141–143. Gunther, T., Chen, Z. F., Kim, J., Priemel, M., Rueger, J. M., ACKNOWLEDGMENTS Amling, M., Moseley, J. M., Martin, T. J., Anderson, D. J., and Karsenty, G. (2000). Genetic ablation of parathyroid glands We thank D. Platt for help with reconstructions and B. Headrick reveals another source of parathyroid hormone. Nature 406, in the IMMAG Electron Microscopy/Histology Core Facility for 199–203. histology support. We also thank S. Stadler, A. Godwin, C. Kovacs, Jenkinson, E. J., Franchi, L. L., Kingston, R., and Owen, J. J. (1982). and B. Condie for critical reading of the manuscript. This work was Effect of deoxyguanosine on lymphopoiesis in the developing supported by NIH Grant R01-HD35920 (to N.R.M.) and a Research thymus rudiment in vitro: Application in the production of Supplement for Underrepresented Minority Postdoctoral Training chimeric thymus rudiments. Eur. J. Immunol. 12, 583–587. (to A.N. and N.R.M.). Kaestner, K. H., Knochel, W., and Martinez, D. E. (2000). Unified nomenclature for the winged helix/forkhead transcription factors. Genes Dev. 14, 142–146. REFERENCES Kim, J., Jones, B. W., Zock, C., Chen, Z., Wang, H., Goodman, C. S., and Anderson, D. J. (1998). Isolation and characterization of Akiyama, Y., Hosoya, T., Poole, A. M., and Hotta, Y. (1996). The mammalian homologs of the Drosophila gene glial cells missing. gcm-motif: A novel DNA-binding motif conserved in Drosophila Proc. Natl. Acad. Sci. USA 95, 12364–12369. and mammals. Proc. Natl. Acad. Sci. USA 93, 14912–14916. Kishi, H., Su, D. M., Muraguchi, A., and Watanabe, T. (1995). A Barrow, J. R., Stadler, H. S., and Capecchi, M. R. (2000). Roles of novel cell surface antigen, immature thymocyte antigen-1, is Hoxa1 and Hoxa2 in patterning the early hindbrain of the mouse. involved in the differentiation of murine thymocytes. J. Immu- Development 127, 933–944. nol. 155, 568–577.

Kovacs, C., Manley, N., Moseley, J., Martin, T., and Kronenberg, H. Rossel, M., and Capecchi, M. R. (1999). Mice mutant for both Hoxa1 (2001). Fetal parathyroids are not required to maintain placental and Hoxb1 show extensive remodeling of the hindbrain and defects calcium transfer. J. Clin. Invest. 107, 1007–1015. in craniofacial development. Development 126, 5027–5040. Krumlauf, R. (1994). Hox genes in vertebrate development. Cell 78, Salser, S. J., and Kenyon, C. (1996). A C. elegans Hox gene switches 191–201. on, off, on and off again to regulate proliferation, differentiation Manley, N. (2000). Thymus organogenesis and molecular mecha- and morphogenesis. Development 122, 1651–1661. nisms of thymic epithelial cell differentiation. Semin. Immunol. Sauvageau, G., Thorsteinsdottir, U., Eaves, C. J., Lawrence, H. J., 12, 421–428. Largman, C., Lansdorp, P. M., and Humphries, R. K. (1995). Manley, N. R., and Capecchi, M. R. (1995). The role of hoxa-3 in Overexpression of HOXB4 in hematopoietic cells causes the mouse thymus and thyroid development. Development 121, selective expansion of more primitive populations in vitro and in 1989–2003. vivo. Genes Dev. 9, 1753–1765. Manley, N. R., and Capecchi, M. R. (1998). Hox group 3 paralogs Sauvageau, G., Thorsteinsdottir, U., Hough, M. R., Hugo, P., regulate the development and migration of the thymus, thyroid Lawrence, H. J., Largman, C., and Humphries, R. K. (1997). and parathyroid glands. Dev. Biol. 195, 1–15. Overexpression of HOXB3 in hematopoietic cells causes defec- Mansour, S. L., Goddard, J. M., and Capecchi, M. R. (1993). Mice tive lymphoid development and progressive myeloproliferation. homozygous for a targeted disruption of the proto-oncogene int-2 Immunity 6, 13–22. have developmental defects in the tail and inner ear. Develop- Schorpp, M., Wiest, W., Egger, C., Hammerschmidt, M., Schlake, ment 117, 13–28. T., and Boehm, T. (2000). Genetic dissection of thymus develop- Moseley, J. M., Mathews, E. W., Breed, R. H., Galante, L., Tse, A., ment. Curr. Top. Microbiol. Immunol. 251, 119–124. and MacIntyre, I. (1968). The ultimobranchial origin of calcito- Su, D.-M., Wang, J., Lin, Q., Cooper, M. D., and Watanabe, T. ␣ ␤ nin. Lancet i, 108–110. (1997). Interferons / inhibit IL-7-induced proliferation of CD4ϪCD8ϪCD3ϪCD44ϩCD25ϩ thymocytes, but do not inhibit Nehls, M., Kyewski, B., Messerle, M., Waldschu¨tz, R., Schu¨d- that of CD4ϪCD8ϪCD3ϪCD44ϪCD25Ϫ thymocytes. Immunol- dekopf, K., Smith, A. J., and Boehm, T. (1996). Two genetically ogy 90, 543–549. separable steps in the differentiation of thymic epithelium. Su, D. M., and Manley, N. R. (2000). Hoxa3 and Pax1 transcription Science 272, 886–889. factors regulate the ability of fetal thymic epithelial cells to Nehls, M., Pfeifer, D., Schorpp, M., Hedrich, H., and Boehm, T. promote thymocyte development. J. Immunol. 164, 5753–5760. (1994). New member of the winged-helix protein family dis- Wallin, J., Eibel, H., Neubu¨ser, A., Wilting, J., Koseki, H., and rupted in mouse and rat nude mutations. Nature 372, 103–107. Balling, R. (1996). Pax1 is expressed during development of the Neubu¨ser, A., Koseki, H., and Balling, R. (1995). Characterization thymus epithelium and is required for normal T-cell maturation. and developmental expression of Pax9, a paired-box-containing Development 122, 23–30. gene related to Pax1. Dev. Biol. 170, 701–716. Weatherbee, S. D., Halder, G., Kim, J., Hudson, A., and Carroll, S. Peters, H., Neubu¨ser, A., Kratochwil, K., and Balling, R. (1998). (1998). Ultrabithorax regulates genes at several levels of the Pax9-deficient mice lack pharyngeal pouch derivatives and teeth wing-patterning hierarchy to shape the development of the and exhibit craniofacial and limb abnormalities. Genes Dev. 12, Drosophila haltere. Genes Dev. 12, 1474–1482. 2735–2747. Wilm, B., Dahl, E., Peters, H., Balling, R., and Imai, K. (1998). Peters, H., Wilm, B., Sakai, N., Imai, K., Maas, R., and Balling, R. Targeted disruption of Pax1 defines its null phenotype and proves (1999). Pax1 and Pax9 synergistically regulate vertebral column haploinsufficiency. Proc. Natl. Acad. Sci. USA 95, 8692–8697. development. Development 126, 5399–5408. Ready, A. R., and Jenkinson, E. J. (1987). Deoxyguanosine-induced Received for publication April 9, 2001 survival of class II-expressing fetal mouse thymic allografts is Revised June 5, 2001 associated with the depletion of donor lymphoid and dendritic Accepted June 5, 2001 cells. Transplant. Proc. 19, 4225–4226. Published online July 10, 2001