T cell differentiation is a highly regulated process that takes place in the thymus. During fetal life, the first wave of lymphoid precursors arrive from the fetal liver to the thymic rudiment (Manley,2000; Anderson and Jenkinson,2001; Blackburn and Manley,2004) around day 11 of gestation. Around 12.5 days of gestation, patterning and differentiation begin, a process that is highly dependent on interactions between epithelial cells and early thymocyte precursors. From day 14 until birth, thymocyte maturation takes place and thymic organogenesis is completed. After birth, bone marrow lymphoid progenitors will continue to seed the thymus throughout life.
The stages of T cell development can be identified by examining the surface expression of coreceptors CD4 and CD8 on thymocytes. The earliest thymocyte subset, double negative (DN, CD4−CD8−), can be further subdivided into four populations based on the expression of CD44 and CD25 (DN1: CD44+CD25−; DN2: CD44+ CD25+; DN3: CD44−CD25+; DN4: CD44−CD25−) (Ceredig and Rolink,2002). At stage DN3, thymocytes express an immature receptor known as pre-TCR (pre-T Cell Receptor), which induces a signal (β selection) that allows initiation of TCR α chain rearrangement and maturation to the second developmental stage, CD4+CD8+ or double positive (DP) stage (von Boehmer et al.,1999). DP thymocytes then express a mature αβ T Cell Receptor (TCR), which allows thymocytes to recognize endogenous peptides expressed on MHC-bearing stromal cells and to undergo positive and negative selection. Positive selection allows maturation to the final developmental stage of mature CD4+ or CD8+ single positive thymocytes (SP), while negative selection eliminates the majority of potentially autoreactive T lymphocytes (Starr et al.,2003). Once mature thymocytes express a single coreceptor, they migrate to secondary lymphoid organs were they can be activated and become CD4+ helper or CD8+ cytotoxic effector T cells.
Thymocyte differentiation depends on interactions between thymocytes and stromal cells as well as on the presence of cytokines and chemokines in the microenvironment of developing thymocytes. Transforming growth factor β (TGF-β) superfamily comprises a group of ubiquitous growth factors widely expressed and capable of inducing a variety of responses, depending not only on the cell type and its differentiation state, but also on the milieu of cytokines present (Cottrez and Groux,2001; reviewed in Massague,1998). This superfamily is divided into three different subfamilies: TGF-βs, BMPs (Bone Morphogenetic Proteins) and Activins. TGF-β superfamily members transduce signals through a couple of transmembrane receptors with serine/threonine kinase activity, known as type I Activin like kinases (ALKs) (ten Dijke et al.,1994a,b) and type II receptors (Heldin et al.,1997). Briefly, the ligand (TGF-β or Activin) binds to a type II receptor, to further recruit a type I receptor. Type II receptor phosphorylates type I receptor, which, in turn, phosphorylates particular members of cytoplasmic proteins named Smads (R-Smads or receptor-regulated Smads) to induce cellular responses (Wrana et al.,1994). Once the R-Smads are phosphorylated, they interact with Smad-4 or Co-Smad (“common-mediator-Smad” or “collaborating-Smad”) (Zhang et al.,1997; reviewed in Shi and Massague,2003) and translocate to the nucleus, where they interact with other transcription factors and regulate gene expression (Abdollah et al.,1997; Souchelnytskyi et al.,1997; reviewed in Derynck et al.,1998). On the other hand, I-Smads (“Inhibitory-Smads”) inhibit the signal induced by R-Smads/Co-Smads complexes, by interfering with R-smad binding to type I receptors or by competing for the Co-smad. Other mechanisms known to inhibit the signal include the presence of soluble proteins such as Noggin and Chordin for BMPs (Piccolo et al.,1996; Zimmerman et al.,1996), Betaglycan for TGF-βs (Lopez-Casillas et al.,1994) and Follistatin (de Winter et al.,1996) and Inhibins (reviewed in Bernard et al.,2001) for Activins, among others (reviewed in Massague,1998).
While TGF-β superfamily members are known to regulate proliferation, migration, differentiation, and many other responses in different tissues (reviewed in Massague,1998), their possible role in lymphoid development has not yet been extensively characterized. Recently, knock-out mice for a transcription factor, Schnurri-2 (Shn-2) (Takagi et al.,2001), which in Drosophila mediates Dpp (“Decapenthaplegic”) signaling pathway and has BMP-4 as its murine homologue, showed absence of thymocyte-positive selection, indicating that BMPs and/or other TGF-β superfamily members, including Activins, may have a role during thymocyte differentiation and selection. In this context, the expression and function of BMPs has been recently described and it was found that these molecules regulate the DN1 to DN2 and DN4 to DP transitions during thymocyte development (Graf et al.,2002; Hager-Theodorides et al.,2002; Varas et al.,2003). On the other hand, TGF-β (reviewed in Letterio and Roberts,1998) was shown to regulate the DN to DP transition (Plum et al.,1995), and the in vitro induction of CD8 expression on pre-T cells during thymocyte development in fetal thymic organ cultures (Suda and Zlotnik,1992a,b).
Activins were first described by their role to regulate follicle-stimulating hormone (FSH) secretion from the anterior pituitary, and as important regulators of gonadal functions. However, Activins are also involved in many other processes during embryogenesis, such as left-right asymmetry (Levin et al.,1997) and digit development (Merino et al.,1999). During thymocyte differentiation, Activins were also shown to inhibit proliferation of rat thymocytes (Hedger et al.,1989,2000) and more recently the expression of Activin type I (Alk-4) and type II receptors in thymic subpopulations (DN and SP) as well as Smad-2 phosphorylation (Rosendahl et al.,2003) has been demonstrated. However, the expression pattern of members of the Activin signaling pathway during fetal thymic development has not been studied. Therefore, in the present study we analyzed the expression of Activins and Activin receptors at different developmental stages of mouse embryo. Our results surprisingly showed that Inhibins may be the major ligands of the Activin family expressed during thymus organogenesis and early thymocyte differentiation.
RESULTS AND DISCUSSION
In order to determine the presence of members of the Activin family and their receptors during thymocyte differentiation, we performed real time RT-PCR analysis, in fetal and adult thymus using primers specific for the three ligand subunits (Activin βA, Activin βB, and Inhibin α), type I receptor Alk-4 and type II receptors ActRIIA and ActRIIB as well as Smads 2, 3, and 4. Whole fetal thymi were analyzed at different developmental stages corresponding to important checkpoints occurring during early T cell development. In order to compare thymocyte before and after pre-TCR signaling (β selection), we used E14 and E15 thymi, respectively: E14 fetal thymus is enriched in DN1 and DN2 stage subpopulations while E15 fetal thymus already contains DN1 to DN4 subpopulations. We also used E16 thymi since this developmental stage correlates with the generation of the first DP thymocytes (Baird et al.,2000). Finally, whole adult thymi was also analyzed, where DN, DP, and SP thymic subpopulations are represented, and positive and negative selection processes are also taking place. As shown in Figure 1, expression of βA, βB, and α subunits (Activins/Inhibins A and B), as well as type I and type II receptors, decreases during fetal development: Activin βA is much more abundant at day E14 and significantly decreases at days E15 and E16 of gestation, as well as in adult thymus (Fig. 1A). Activin βB and Inhibin α subunits also showed significantly higher expression levels at day E14 compared to day E15, E16, and adult thymus (Fig. 1B and C). It is worth mentioning that expression of α subunits had never been previously reported in murine thymus. Interestingly, expression of type II (ActRIIA and ActRIIB) receptors (Fig. 1E,F) showed the same expression pattern as the ligands described above (E14>E15>E16), while type I receptor Alk-4 showed no significant differences between developmental stages (Fig. 1D). On the other hand, Smad expression seemed to be higher at day E15 than at days E14, E16, and adult thymus. However, statistical analysis did not show significant differences between stages (Fig. 1G–I). This indicates that the complete Activin signaling machinery may be active during all stages of thymocyte development analyzed.
We next analyzed expression of Activin ligands in adult thymic subpopulations. For these experiments, we performed FACS sorting of DN, DP, CD4 SP, and CD8 SP thymocytes to a purity > 95%. As a source of stromal cells, we used deoxiguanosine-treated E15 thymus as has been previously reported (Jenkinson et al.,1982; Hager-Theodorides et al.,2002). We observed that Activin βA subunit is mainly expressed in stromal cells and, among thymocyte subpopulations, was preferentially expressed in DN cells, although it was also expressed in DP and SP subpopulations at lower levels (Fig. 2A). Statistical analysis showed that DNs expressed significantly higher levels of Activin βA compared to CD4 SP and CD8 SP subsets. Interestingly, among SP thymocytes, CD8 SP appeared to express significant higher levels than CD4 SPs. (Fig. 2A). Our data demonstrate for the first time Activin βA expression in thymocytes, which previously had been reported to be exclusively expressed in stromal cells (Rosendahl et al.,2003). However, Rosendahl et al. did not use purified stromal cells or individual thymocyte subpopulations to assess the source of Activin βA expression and they used a less sensitive method of detection. In contrast, Activin βB showed a different expression pattern, which was restricted to stromal cells (Fig. 2B). Since stromal cells were prepared after deoxiguanosine treatment, while thymocyte subpopulations were obtained from freshly isolated adult thymi, there might be a concern that some of the differences observed might be partially dependent on the treatment. However, the absence of βB subunits in adult sorted subpopulations suggests that the expression of Activin βB observed in whole adult untreated thymus was of stromal origin (compare Fig. 1B with Fig. 2B). These results are in contrast to previously reported data by Rosendahl et al. (2003) who described expression of βB subunits in “sorted” thymocytes. However, these authors had used a different purification method. Finally, as shown in Figure 2C, Inhibin α subunits were mainly expressed in thymic stromal cells. However, early DN, DP, and CD4 SP subsets, but not CD8 SP thymocytes, expressed low levels of Inhibin α subunits.
On the other hand, analysis of Activin receptors showed that type I receptor Alk-4 was mostly expressed in DN thymocytes and stromal cells, although it was also expressed to a lesser extent in DP, CD4 SP, and CD8 SP thymocytes (Fig. 2D), implying that both stromal cells and thymocytes may potentially respond to the ligands. Statistical analysis showed that there were significant differences in expression between DN and the rest of thymocyte subsets or adult thymus. Type II receptors ActRIIA and ActRIIB were expressed in stromal cells as well as in all thymocyte subpopulations. However, ActRIIA was significantly higher expressed in stromal cells compared to thymocyte subpopulations (Fig. 2E), while ActRIIB was not preferentially expressed in stromal cells, since DN cells expressed comparable levels of this receptor. In addition, comparison between thymocyte subsets showed statistical differences between DNs and DP, CD4 SP, or CD8 SP subpopulations (Fig. 2F). The correlation between Alk-4 and ActRIIB receptor expression is in agreement with the requirement of both type I and type II receptors to mediate Activin signals (reviewed in Massague,1998).
Finally, Smad proteins were also expressed by both stromal cells and thymocytes and, as seen for Alk 4 and ActRIIB, the expression was significantly higher in DN cells compared to the rest of thymocyte subpopulations, while it was not significantly different compared to stromal cells, except for Smad 3 expression (Fig. 2G–I).
In conclusion, Activin βB expression seems to be restricted to stromal cells while Activin βA, Inhibin α, Alk-4, ActRIIA, ActRIIB, and Smad expression is shared by stromal and thymocyte subpopulations (mainly DN cells). The fact that both Alk 4, ActRIIB, and Activin-mediated signaling molecules (Smad 2, 3, 4) are preferentially expressed in DN thymocytes (Fig. 2D–I) suggests that this particular early precursor represents the main target for Activin ligands among thymocyte subsets. It is important to note that based on quantitative PCR analysis, stromal cells were always the major population expressing α and β subunits (Fig. 2A–C) so they appear to be the subpopulation responsible for the ligand production. In fact, βB subunit was exclusively detected in stromal cells, while α subunit was also present at low levels in some thymocyte subpopulations, but absent in CD8 SP cells (Fig. 2C). In this context, due to the preferential expression of ligand subunits by stromal cells, we cannot exclude the possibility that the decrease in Activin ligand signal observed in fetal thymi (Fig. 1A–C) might be due to a dilution effect of stromal cells, compared to the increasing numbers of thymocytes, which are strongly proliferating during the fetal developmental stages analyzed.
Since the data obtained from Real time RT-PCR only represents differences in expression of a particular gene between thymocyte differentiation stages or adult cell subsets, but does not allow us to compare expression of different genes in a particular thymic subpopulation, we decided to evaluate the differences in gene expression within the same tissue (mRNA abundance) (see Experimental Procedures section). As shown in Figure 3A, the analysis of mRNA abundance at different developmental stages showed that Inhibins A or B (formed by α/βA or α/βB subunits) rather than Activins (formed by β subunits), seem to be the major ligands expressed in fetal and adult thymus, since expression of the α subunit was 2–4 times higher than the βA subunit in early E14 and adult thymus, while this difference increased up to 15–20 times in E15 and E16 thymus. On the other hand, expression of the α subunit was 15–20 times higher than the βB subunit, in all developmental stages (Fig. 3A). This finding may have important implications on the role of Activins in T cell development, since Inhibins have been shown to bind to ActRII receptors, inhibit binding of Alk4 to ActRIIs, and therefore may antagonize Activin-mediated functions (Vale et al.,2004). Furthermore, the expression levels of Alk 4 and ActRIIB observed in all developmental stages suggest that they are the most likely receptor pair to mediate Activin signaling during thymocyte development (Fig. 3B). These results are in contrast to data recently reported by Rosendahl et al. (2003) who were unable to detect significant amounts of ActRIIB in thymocytes. This difference might be explained by the fact that they used a qualitative RT-PCR analysis, instead of Real Time PCR. Finally, analysis of Smad expression showed that Smad 2 was consistently higher than Smad 3 and 4 in all differentiation stages (Fig. 3C). These results are consistent with the recent report by Rosendahl et al. (2003) showing Smad 2 phosphorylation in thymocytes stimulated with Activin, and may have important implications in Activin-mediated signaling, since Smad 2 and Smad 3 may compete for binding to the co-smad (Smad-4). However, we cannot rule out a role for Smad 3 in Activin signaling, since phosphorylation of Smad 3 has not been yet investigated in the thymus.
Similarly, we performed the same analysis in thymic subpopulations from adult mice and E15 stromal cells (Fig. 4A–C). Our results showed that Inhibin α is the most abundant subunit produced by E15 stromal cells, while similar levels of α and βA subunits were expressed in thymocytes (Fig. 4A), although as shown in Figure 2A and C, expression was much lower than that observed in stromal cells. Among Activin type II receptors, ActRIIB was always expressed at higher levels than ActRIIA in all thymic subpopulations tested (Fig. 4B), further supporting the idea that Alk4/ActRIIB may be mediating most of Activin signaling in the thymus, acting as an autocrine pathway in stromal cells. Furthermore, receptors Alk 4/Act RIIB are present in thymocytes (mostly DN), but levels of ligand are low in these cells, suggesting that stromal cells may regulate thymocyte differentiation in a paracrine pathway. Whether Inhibin and or Activin are promoting or inhibiting transition of DN thymocyte differentiation towards DP thymocytes is currently unknown and will be addressed in future studies. In addition, although the receptors Alk 4 and Act RIIB are expressed in DP and SP thymocytes at lower levels (Fig. 2D and F), we cannot rule out a functional role of Activin or Inhibin in these cells. On the other hand, we do not exclude the possibility that other members of the TGF-β superfamily expressed in the thymus, such as BMPs, which have been shown to share type II receptors with Activins (Yamashita et al.,1995), could also signal through these receptors. Finally, as previously shown for fetal thymus, Smad2 expression was higher than Smad3 and Smad4 in all the thymic subpopulations tested (Fig. 4C), further supporting the role of this protein, during T cell development (Rosendahl et al.,2003).
Finally, to confirm the RT-PCR results at the protein level, we performed immunohistochemistry assays with antibodies directed against type I receptor Alk-4, Activin βA, and α subunits, in fetal and adult tissue sections. As shown in Figure 5A, Activin βA expression appeared to be higher at E14 and decreased at later developmental stages (E15, E16, and adult thymus). Expression of Inhibin α subunit was also confirmed in fetal and adult thymic sections, showing a strong staining in all developmental stages (Fig. 5B). Furthermore, as expected, we observed a pronounced staining for Alk-4 in E14 thymus and also at days E15, E16, and adult thymus, although the intensity of staining seemed to be lower (Fig. 5C). As we showed in the RT-PCR results, Activin ligands were preferentially expressed by stromal cells. To further confirm this result, we performed staining on serial sections including other markers known to be expressed exclusively in stromal cells. For fetal E14 thymus analysis, we used an anti-cytokeratin antibody, specific for K5/K8 keratins, which are highly expressed in thymic epithelial cells (Fig. 6A), and for adult thymus we used anti MHC-Class II, highly expressed by medullary dendritic cells (Fig. 6C). We compared the stromal staining with the staining obtained with anti-Inhibin α antibody, demonstrating a similar expression pattern (Fig. 6B, fetal E14 thymus; 6D adult thymus).
Altogether, our results demonstrate for the first time the expression of the α subunit of the Activin family in the thymus, indicating that Inhibins may play an important role during thymocyte differentiation. Moreover, we could speculate that Activins/Inhibins might have a role in thymus organogenesis since we found expression of those ligands in stromal cells as early as day 14 of gestation, when interactions between incoming early lymphoid progenitors and stromal cells are crucial for further differentiation of stromal subtypes and thymocytes, leading to the anatomical compartmentalization of the thymus (Manley,2000). Furthermore, due to the antagonizing role of Inhibins on Activin-mediated functions, the balance Activin/Inhibin expression may regulate many of the cellular responses that take place during thymocyte development (such as proliferation and apoptosis) and, by this means, may regulate important “checkpoints” during T cell maturation.
Four- to six-week-old adult thymus and E14–16 fetuses thymus from B6/D2 F1 hybrid mice obtained by crossing C57/BL6 females with DBA-2 males (obtained from Jackson Laboratories, Bar Harbor, ME) were used in our analysis.
RNA Sources and Primers
RNA was obtained from total adult and fetal E14–16 thymi, sorted adult thymocytes, E15 thymic stromal cells, and testicle as a positive control (not shown) (Meunier et al.,1988). For thymic subpopulations, thymi from 4–6-week-old B6/D2 mice were stained with CD4-phycoerythrin and CD8-CyChrome (Pharmingen, San Diego, CA) and CD4, CD8, DP, and DN subpopulations were sorted using the flow-cytometer FACS Aria (Becton & Dickinson) to a purity of >95%. For stromal cell purification, E15 fetal thymi were treated for 5 days with deoxiguanosine (1.35 mM [Sigma Chemicals D-0901, St. Louis, MO]) in 5% CO2 at 37°C.
RNA was isolated using RNA-60 reagent (Tel-Test Inc, Friendslaw, TX) according to the manufacturer's protocol. DNA was removed using the “DNA-free” reagent from Ambion Inc. (Austin, TX). cDNA was synthesized from total DNA-free RNA, using Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA) and oligo dT (Roche, Mannheim, Germany).
Real-time PCR was performed by amplifying 1 μl of cDNA (1/40 of total cDNA yield obtained from 1–3 μg of mRNA) with SYBR Green PCR Core Kit (Applied Biosystems, Foster City, CA) on a ABI PRISM 7000 Sequence Detection System (Applied Biosystems). Amplification conditions for Activin βB, Inhibin α, ActRIIA, ActRIIB, Smad-2, Smad-3, and Smad-4, were 95°C for 10 min, followed by 40 cycles of 94°C for 15 s, 55°C for 1 min, and 72°C for 1 min. Amplification conditions for Activin βA and Alk-4 were 95°C for 10 min, followed by 45 cycles of 94°C for 15 s, 58°C for 45 s, and 60°C for 1 min. Following amplification, a melting point program was run to create melting profiles for each reaction described. The number of PCR cycles required for SYBR Green fluorescence to cross a threshold where there was a significant increase in change of fluorescence (CT = threshold cycle) was obtained using ABI PRISM 7000 Sequence Detection System Software. Total cDNA input was normalized to mouse β-Actin expression measured in parallel PCR reactions. For statistical analysis of RT-PCR data, the program Qgene was used (Muller et al.,2002). The mean of the normalized expression was calculated by averaging the CT values (from triplicate samples of the target gene and of the reference) and subsequent calculation of the mean normalized expression and its standard error (equations 3 and 4, respectively, according to Muller et al.,2002). The calculation of the normalized expression is supported by the amplification efficiency values calculated from the amplification efficiency plots constructed for each gene These results were used to calculate the relative amount of each gene mRNA present in fetal thymi and between sorted thymic supopulations or E15 stromal cells, compared to adult thymi. Unpaired two-tailed Student's t-test was performed. P < 0.05 was considered significant.
To analyze the relative abundance of different mRNAs within a particular tissue, we used standard curves obtained with known amounts of the specific amplified products, obtained from the same cDNA to interpolate our experimental data normalized to the endogenous invariable β actin control.
The PCR products were analyzed on a 2% (w/v) agarose gel to confirm purity and size product. Primers used in the PCR amplification were the same as described above.
Paraffin-embedded thymus (previously fixed and dehydrated) from fetal and adult mice were sliced (5 μm thick) and rehydrated by serial washes with decreasing concentration of etanol (from 100 to 50%). After quenching endogenous peroxidase activity by 10 min incubation with 3% H2O2, epitope exposure with citrate buffer in high-pressure conditions for 1 min and blocking with super block reagent for 6 min (Sensitek HRP, Scy Tek Laboratories, Logan, UT), sections were incubated with primary antibodies in blocking solution overnight at 4°C. Anti-Alk-4 (ACTR-IB, clone yT-17), anti-Activin-βA (Inhibin βA, clone C-18), and anti-Cytokeratin 5/8 (RCK102) were purchased from Santa Cruz Biotechnology, Santa Cruz, CA. Anti Alk-4 and anti-Activin βA antibodies were biotinylated using a protocol adapted from (Guesdon et al.,1979); Anti-Inhibin α rabbit polyclonal antibody (Vaughan et al.,1989) was kindly provided by Dr. Wylie Vale, The Salk Institute for Biological Studies, La Jolla, CA. Anti-MHC Class II (biotin anti-mouse I-A/I-E clone 2G9), was purchased from Pharmingen, San Diego, CA. Sections were washed 4 times with PBS×1 and streptavidin-HRP secondary reagent (Sensitek HRP, Scy Tek Laboratories) sheep anti-mouse-HRP or goat anti-rabbit HRP (Amersham Biosciences, Piscataway, NJ) were added for 20 min at 37°C. As negative controls, normal rabbit sera or secondary reagents were used after the initial blocking step. Slides were developed for 5–20 min in diaminobenzidine-substrate (Sensitek-HRP) and counter-stained with Harris hematoxylin. Images were analyzed under a Nikon Optiphot-2 microscope equipped with a Nikon Coolpix 4300 camera (Nikon Inc, Tokio, Japan).
We thank David Garciadiego, Pedro Medina, and Ramsés Chávez for technical assistance. We also thank Dr. Wylie Vale for kindly providing us with the anti-Inhibin α polyclonal antibody. We additionally extend our thanks to Dr. Leopoldo Santos-Argumedo and Dr. Eduardo García-Zepeda for a critical reading of the manuscript and Dr. Rafael Camacho and Ileana Licona for assistance in the analysis of the quantitative RT-PCR data. We gratefully acknowledge QFB Carlos Castellanos Barba from the flow-cytometry unit at the Instituto de Investigaciones Biomédicas for technical assistance during the cell sorting experiments. Finally, we extend our appreciation to Dr. Gerardo Arrellín from the mouse facility at the Instituto de Investigaciones Biomédicas, UNAM, México, for technical support and veterinary advice. This work was supported in part by Consejo Nacional de Ciencia y Tecnología (CONACyT; 42797-Q to G.S. and 42568-Q to J.C.M.) and Universidad Nacional Autónoma de México (DGAPA; IN200205 to J.C.M.). P.L. is the recipient of a fellowship from CONACyT supplemented by DGEP from UNAM.