Abstract
BMI-1 and EZH2 Polycomb-group (PcG) proteins belong to two distinct protein complexes involved in the regulation of hematopoiesis. Using unique PcG-specific antisera and triple immunofluorescence, we found that mature resting peripheral T cells expressed BMI-1, whereas dividing blasts were EZH2+. By contrast, subcapsular immature double-negative (DN) (CD4−/CD8−) T cells in the thymus coexpressed BMI-1 and EZH2 or were BMI-1 single positive. Their descendants, double-positive (DP; CD4+/CD8+) cortical thymocytes, expressed EZH2 without BMI-1. Most EZH2+ DN and DP thymocytes were dividing, while DN BMI-1+/EZH2− thymocytes were resting and proliferation was occasionally noted in DN BMI-1+/EZH2+ cells. Maturation of DP cortical thymocytes to single-positive (CD4+/CD8− or CD8+/CD4−) medullar thymocytes correlated with decreased detectability of EZH2 and continued relative absence of BMI-1. Our data show that BMI-1 and EZH2 expression in mature peripheral T cells is mutually exclusive and linked to proliferation status, and that this pattern is not yet established in thymocytes of the cortex and medulla. T cell stage-specific PcG expression profiles suggest that PcG genes contribute to regulation of T cell differentiation. They probably reflect stabilization of cell type-specific gene expression and irreversibility of lineage choice. The difference in PcG expression between medullar thymocytes and mature interfollicular T cells indicates that additional maturation processes occur after thymocyte transportation from the thymus.
Lymphocyte differentiation involves developmental decisions that must be propagated to the next generation of cells. This process is regulated both by extracellular and intracellular proteins, including cytokines and transcription factors (1, 2, 3, 4, 5, 6). Recent studies identified Polycomb-group (PcG)2 proteins as a new class of transcription regulators that contribute to regulation of hematopoiesis and the cell cycle (7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18). PcG genes were originally discovered in Drosophila as suppressors of homeobox gene expression. They form a cellular memory system that ensures stable transmission of developmental decisions and cell identity (19, 20, 21, 22).
PcG proteins form large multimeric protein complexes that bind to DNA and probably function by altering the conformation of chromatin (23, 24, 25, 26, 27, 28, 29, 30, 31, 32). Two distinct complexes were identified in humans (reviewed in Ref. 22). One complex consists of the BMI-1, RING1, HPH1, HPH2, HPC1, HPC2, and HPC3 PcG proteins and the C-terminal binding protein CtBP (23, 24, 25, 26, 28, 29, 30, 31, 33). A second complex contains the ENX/EZH2 and EED PcG proteins, the HDAC1/2 histone deacetylases, and the YY1 transcription factor (32, 34, 35, 36). The two PcG protein complexes differ in tissue distribution and probably regulate different target genes (22, 37, 38).
Differentiation of lymphocyte precursors in bone marrow is associated with profound alterations in PcG gene transcription (10). The earliest precursor cells preferentially transcribe the BMI-1 PcG gene, and transcription of other PcG genes occurs at later stages of B cell development while transcription of BMI-1 gradually disappears (10). The most convincing evidence to support a role for PcG genes in lymphopoiesis, however, comes from studies of mutant mice. For instance, overexpression of the Bmi-1 PcG gene in transgenic mice resulted in enhanced lymphoproliferation and ultimately in development of B cell lymphomas (13, 39). By contrast, mice with a targeted deletion of the Bmi-1 and Mel-18 PcG genes develop severe hypoplasia, while loss of the Eed PcG gene results in increased lymphoproliferation (7, 8, 11). These observations suggest that the two PcG complexes have opposing roles in the regulation of hematopoiesis. We recently demonstrated that expression of the two PcG complexes, reflected by detection of BMI-1 and RING1, and EZH2 and EED, is mutually exclusive at different B cell differentiation stages in germinal centers (GCs) (15, 16). These results showed that expression of PcG genes is strictly regulated during follicular B cell development and suggested a role for PcG proteins in the GC cell reaction.
Despite a role for PcG genes in regulation of lymphocyte development in experimental model systems, little is known about PcG expression during human hematopoiesis. In the present study, we addressed the question whether defined stages of T cell development are associated with distinct expression patterns of the two human PcG complexes (identified by the BMI-1 and EZH2 PcG proteins). We performed tricolor immunofluorescence for these proteins in combination with T cell differentiation stage-specific markers in peripheral lymphoid tissue and thymus. Similar to follicular B cells, BMI-1 and EZH2 expression were mutually exclusive in interfollicular and follicular T cells. By contrast, developing thymocytes exhibited three PcG expression profiles, including BMI-1+/EZH2−, BMI-1+/EZH2+, and BMI-1−/EZH2+ stages. These patterns correlated with distinct T cell differentiation stages as characterized by detection of CD3, CD4, CD8, TCRαβ, and TCRγδ. Our results demonstrate that T cell differentiation is associated with discrete expression patterns of the BMI-1- and EZH2-containing PcG complexes and suggest that human PcG genes contribute to regulation of T cell development. Distinct PcG expression profiles at various T cell differentiation stages probably reflect stabilization of gene expression patterns and may be related to irreversibility of lineage choice.
Materials and Methods
Lymph nodes and thymuses were obtained from the surgery room and immediately frozen. Consecutive sections were cut (3 μm) and fixed in 2% formaldehyde. Endogenous peroxidase was inhibited with 1% H2O2, diluted in PBS. Following preincubation with 5% BSA, primary Abs against BMI-1 and EZH2 (32) were applied in combination with antisera against one of the following markers (Table I): CD3, CD4, CD8, CD19, CD68, CD86, Mib-1/Ki-67, TCRαβ, or TCRγδ. BMI-1 was detected by incubation with goat anti-mouse (GaM) IgG2bHRP using the streptavidin-biotin-avidin complex/HRP method and rhodamine/tyramine intensification (excitation 550, emission 570; red fluorescence). EZH2 was detected by incubation with goat anti-rabbit antiserum coupled to ALEXA (excitation 495, emission 519; green fluorescence). The other markers were detected by incubating the slides with GaMIgG1BIO or GaMIgG2aBIO (depending on the subclass of the primary antiserum and as indicated in Table I), followed by incubation with StrepAPC (streptavidin coupled to allophycocyanin (excitation 650, emission 660); blue fluorescence). Cross-reactivity of the antisera was excluded by appropriate controls, and PcG expression patterns were confirmed in at least three separate experiments on tissues derived from different individuals. Sections were analyzed with a Leica DMR Confocal LaserScan microscope (Leica, Deerfield, IL). Images were stored digitally at 1024 dpi and processed using Corel Photo-Paint 8.
Antisera and detection system
Antiserum . | Specificity . | Secondary Antiseruma . | Fluorochrome . | Color . |
---|---|---|---|---|
6C9 | PcG BMI-1 | GaMIgG2bHRP | Rhodamine/tyramine | Red |
K358 | PcG EZH2 | NA | GaRALEXA | Green |
Mib-7 | Mib-1/Ki-67 | GaMIgG1BIO | StrepAPC | Blue |
Leu-4 | CD3 | GaMIG1BIO | StrepAPC | Blue |
Leu3a | CD4 | GaMIgG1BIO | StrepAPC | Blue |
C8/1448 | CD8 | GaMIgG1BIO | StrepAPC | Blue |
HD37 | CD19 | GaMIgG1BIO | StrepAPC | Blue |
Kp1 | CD68 | GaMIgG1BIO | StrepAPC | Blue |
B7-2 | CD86 | GaMIgG1BIO | StrepAPC | Blue |
SAG.E9 | TCRαβ | GaMIgG1BIO | StrepAPC | Blue |
5A6.F9 | TCRγδ | GaMIgG1BIO | StrepAPC | Blue |
Antiserum . | Specificity . | Secondary Antiseruma . | Fluorochrome . | Color . |
---|---|---|---|---|
6C9 | PcG BMI-1 | GaMIgG2bHRP | Rhodamine/tyramine | Red |
K358 | PcG EZH2 | NA | GaRALEXA | Green |
Mib-7 | Mib-1/Ki-67 | GaMIgG1BIO | StrepAPC | Blue |
Leu-4 | CD3 | GaMIG1BIO | StrepAPC | Blue |
Leu3a | CD4 | GaMIgG1BIO | StrepAPC | Blue |
C8/1448 | CD8 | GaMIgG1BIO | StrepAPC | Blue |
HD37 | CD19 | GaMIgG1BIO | StrepAPC | Blue |
Kp1 | CD68 | GaMIgG1BIO | StrepAPC | Blue |
B7-2 | CD86 | GaMIgG1BIO | StrepAPC | Blue |
SAG.E9 | TCRαβ | GaMIgG1BIO | StrepAPC | Blue |
5A6.F9 | TCRγδ | GaMIgG1BIO | StrepAPC | Blue |
APC, Allophycocyanin.
Results
Expression of BMI-1 and EZH2 PcG genes in T cell areas of the lymph node
As a first step toward definition of BMI-1 and EZH2 expression in human T cells, we analyzed expression of these PcG proteins in mature T cells of the lymph node. The majority of these cells are situated in the paracortex or interfollicular region, where they can interact with Ag. Ag-mediated activation results in formation of enlarged lymphoblasts and differentiation into Ag-specific effector cells (40).
BMI-1 and EZH2 PcG proteins were chosen as representatives of two different human PcG complexes and were detected in nuclear staining patterns. Interfollicular cells most frequently expressed BMI-1 and expression of EZH2 was less abundant (Fig. 1, A–C). In the majority of cells, detection of BMI-1 and EZH2 was mutually exclusive, and double staining cells (identified by yellow fluorescence) were rarely observed. The BMI-1+ population included CD3+ T cells (Fig. 1, A–C, interfollicular region; D–F, follicular T cells), CD19+ B cells (Fig. 1, G–I), and CD68+ monocytic/dendritic cells (Fig. 1, J–L). EZH2+ cells were a diverse population of CD3+ T cells (dotted arrows in Fig. 1, A–C, and inset in 1B), CD68+ monocytic/dendritic cells (solid arrow in Fig. 1, J–L and inset in 1K), and unidentified cells. No CD19+/EZH2+ cells were observed in the interfollicular region. In rare cases, BMI-1/EZH2 coexpression was seen in interfollicular cells. However, such BMI-1+/EZH2+ cells were CD3− (solid arrow in Fig. 1, A–C and inset in 1C) and CD19− (solid arrows in Fig. 1, G–I and inset in 1, H and I) and did not belong to the T or B cell lineage. In some instances, BMI-1/EZH2 double-positive (DP) cells were associated with CD68+ monocytic/dendritic cells (see for instance the BMI-1+/CD68+ cell in Fig. 1, J–L, indicated by a solid arrow, which also expresses EZH2 (inset in Fig. 1, K and L)). Since we focused our study on T cell development, these CD3−/BMI-1+/EZH2+ cells were not further investigated.
Expression of BMI-1 and EZH2 PcG genes and Mib-1/Ki-67 in interfollicular cells of the lymph node. BMI-1 and EZH2 were detected by red and green fluorescence, respectively. BMI-1 and EZH2 expression is visualized in combination with in CD3+ (A–C for interfollicular cells and D–F for follicles), CD19 (G–I), and CD68 (J–L), identifying T cells, B cells, and monocytic/dendritic cells, respectively, by blue fluorescence. A–C, BMI-1 and EZH2 expression in interfollicular CD3+ T cells. Note that most CD3+ T cells are BMI-1+ (B and C), while few CD3+ cells express EZH2 (dotted arrows). There are many CD3−/BMI-1+ cells visible. Inset in B, Detail of two CD3+/EZH2+ T cells, identified by ∗. Note also that a large BMI-1+/EZH2+ cell, identified by solid arrow, is CD3− (inset in C shows detail of this cell). D–E, Detail of GC showing BMI-1/EZH2 expression in follicular CD3+ T cells. G–I, BMI-1/EZH2 expression in CD19+ B cells. CD19+/BMI-1+ B cells are abundant in the interfollicular T cell area (IFT). Note that EZH2+/BMI-1+ yellow fluorescent cells are CD19− (solid arrow), as exemplified by insets in E (EZH2 staining pattern for selected BMI-1+/CD19− cells identified by solid arrow and ∗ and F (detail of this CD19−/BMI-1+/EZH2+ cell). J–L, BMI-1/EZH2 expression in CD68+ monocytoid/dendritic cells. Although many CD68+ monocytoid dendritic cells are BMI-1+/EZH2−, BMI-1/EZH2 double expression appears to be occasionally associated with CD68+ cells (solid arrow). Inset in K, EZH2 staining of CD68+/BMI-1+ cell indicated by solid arrow showing that selected BMI-1+/EZH2+ cells express CD68. Inset in L, Detail of the cell identified by solid arrow showing CD68/BMI-1/EZH2 triple positivity. M–O, Expression of BMI-1 and EZH2 in relation to Mib-1/Ki-67 (blue fluorescence). Solid arrow, BMI-1-/EZH2+/Mib-1+ cell; dotted arrow, BMI-1+/EZH2+/Mib-1+ cell. Original magnification, ×400.
Expression of BMI-1 and EZH2 PcG genes and Mib-1/Ki-67 in interfollicular cells of the lymph node. BMI-1 and EZH2 were detected by red and green fluorescence, respectively. BMI-1 and EZH2 expression is visualized in combination with in CD3+ (A–C for interfollicular cells and D–F for follicles), CD19 (G–I), and CD68 (J–L), identifying T cells, B cells, and monocytic/dendritic cells, respectively, by blue fluorescence. A–C, BMI-1 and EZH2 expression in interfollicular CD3+ T cells. Note that most CD3+ T cells are BMI-1+ (B and C), while few CD3+ cells express EZH2 (dotted arrows). There are many CD3−/BMI-1+ cells visible. Inset in B, Detail of two CD3+/EZH2+ T cells, identified by ∗. Note also that a large BMI-1+/EZH2+ cell, identified by solid arrow, is CD3− (inset in C shows detail of this cell). D–E, Detail of GC showing BMI-1/EZH2 expression in follicular CD3+ T cells. G–I, BMI-1/EZH2 expression in CD19+ B cells. CD19+/BMI-1+ B cells are abundant in the interfollicular T cell area (IFT). Note that EZH2+/BMI-1+ yellow fluorescent cells are CD19− (solid arrow), as exemplified by insets in E (EZH2 staining pattern for selected BMI-1+/CD19− cells identified by solid arrow and ∗ and F (detail of this CD19−/BMI-1+/EZH2+ cell). J–L, BMI-1/EZH2 expression in CD68+ monocytoid/dendritic cells. Although many CD68+ monocytoid dendritic cells are BMI-1+/EZH2−, BMI-1/EZH2 double expression appears to be occasionally associated with CD68+ cells (solid arrow). Inset in K, EZH2 staining of CD68+/BMI-1+ cell indicated by solid arrow showing that selected BMI-1+/EZH2+ cells express CD68. Inset in L, Detail of the cell identified by solid arrow showing CD68/BMI-1/EZH2 triple positivity. M–O, Expression of BMI-1 and EZH2 in relation to Mib-1/Ki-67 (blue fluorescence). Solid arrow, BMI-1-/EZH2+/Mib-1+ cell; dotted arrow, BMI-1+/EZH2+/Mib-1+ cell. Original magnification, ×400.
Expression patterns of BMI-1 and EZH2 in the interfollicular region correlated with the stage of the cells in the cell cycle, as determined by coexpression of the cell proliferation marker (41) Mib-1/Ki-67 (blue fluorescence in Fig. 1, M–O). The majority of interfollicular BMI-1+ cells (red fluorescence in Fig. 1, M–O) were resting, because BMI-1 expression rarely overlapped with the blue Mib-1/Ki-67 signal (Fig. 1, N and O). By contrast, green fluorescent EZH2+ interfollicular cells (Fig. 1, M and O) coexpressed Mib-1/Ki-67 (for instance solid arrow in Fig. 1, M–O) and were dividing. Note that this population included the cells that coexpressed BMI-1 and EZH2 (for instance dotted arrow in Fig. 1, G–L) and were CD3−/CD19− (see earlier).
Detection of BMI-1 and EZH2 in GCs confirmed previously determined expression profiles in follicular B cells (15, 16) and showed that expression of these proteins is mutually exclusive in the majority of follicular cells (GC in Fig. 1, A–C, and detail in Fig. 1, D–F). Most CD3+ T cells within GCs expressed BMI-1+ at varying levels (Fig. 1, E and F) and rarely expressed EZH2 (dotted arrows in Fig. 1, A–C, and inset in 1B). BMI-1/EZH2 staining patterns within CD3+ cells were identical for TCRαβ+ cells and no differences were noted between CD4+ and CD8+ cells (data not shown). Mib-1/Ki-67 expression in GCs fully overlapped with EZH2 expression (Fig. 1, N and O) and was not observed in BMI-1+ follicular lymphocytes (Fig. 1 N and Refs. 15, 16). Therefore, the majority of BMI-1+ follicular lymphocytes, which includes CD3+ T cells, is resting.
In summary, expression of BMI-1− and EZH2-containing PcG complexes in interfollicular and follicular T cells is mutually exclusive in the majority of mature CD3+ T cells. BMI-1+ T cells are Mib-1/Ki-67− and resting, while EZH2+ T cells are Mib-1/Ki-67+ and in cycle.
BMI-1 and EZH2 expression in human thymus
We next questioned whether the mutually exclusive expression of BMI-1 and EZH2 in peripheral interfollicular T cells is already established during T cell differentiation in the thymus. We determined BMI-1 and EZH2 expression in thymocytes of the subcapsular layer of the thymus, cortex, and medulla (Figs. 2-4). These regions correlate with distinct steps in T cell differentiation (40). The majority of thymic progenitors are located in the subcapsular layer, and differentiation of immature thymocytes primarily occurs in the cortex. The medulla mainly contains mature thymocytes.
BMI-1, EZH2, and Mib-1/Ki-67 expression in thymocytes. BMI-1 and EZH2 were detected by red and green fluorescence, respectively. A–F, BMI-1/EZH2 expression in cortex (Co) and subcapsular layer (SCL, A–C) and cortex and medulla (Me, D–F. The vast majority of cortical thymocytes expresses EZH2 (B, C, E, and F) whereas BMI-1 is relatively underrepresented (A and D). BMI-1+ cells in the subcapsular layer and cortex frequently coexpress EZH2, producing yellow/orange nuclear fluorescence (C). Dotted arrows in C identify BMI-1+/EZH2− cells in the subcapsular layer. Note that EZH2 appears down-regulated in medullar thymocytes (E and F), whereas BMI-1 does not appear to change (D). Solid arrows in D–F, Medullar thymocytes with low-level BMI-1/EZH2 coexpression. G–L, Expression of BMI-1 and EZH2 in relation to Mib-1/Ki-67 (blue fluorescence) in the subcapsular layer and cortex (G–I) and at the cortical-medullary region (J–L). G and J, Superimposition of the BMI-1 and EZH2 signal; H and K, superimposition of the BMI-1 and Mib-1/Ki-67 signal; I and L, superimposition of the BMI-1, EZH2, and Mib-1/Ki-67 signal. BMI-1+/EZH2+ cells are identified by yellow fluorescence; BMI-1+/Mib-1+ cells stain purple. Note that Mib-1 expression overlaps mainly with EZH2, particularly in the cortex. BMI-1/Mib-1+ (purple cells) belong to the BMI-1+/EZH2+ (yellow) subset (solid arrows), but within the BMI-1+/EZH2+ population not all cells are dividing (dotted arrows). Original magnification, ×400.
BMI-1, EZH2, and Mib-1/Ki-67 expression in thymocytes. BMI-1 and EZH2 were detected by red and green fluorescence, respectively. A–F, BMI-1/EZH2 expression in cortex (Co) and subcapsular layer (SCL, A–C) and cortex and medulla (Me, D–F. The vast majority of cortical thymocytes expresses EZH2 (B, C, E, and F) whereas BMI-1 is relatively underrepresented (A and D). BMI-1+ cells in the subcapsular layer and cortex frequently coexpress EZH2, producing yellow/orange nuclear fluorescence (C). Dotted arrows in C identify BMI-1+/EZH2− cells in the subcapsular layer. Note that EZH2 appears down-regulated in medullar thymocytes (E and F), whereas BMI-1 does not appear to change (D). Solid arrows in D–F, Medullar thymocytes with low-level BMI-1/EZH2 coexpression. G–L, Expression of BMI-1 and EZH2 in relation to Mib-1/Ki-67 (blue fluorescence) in the subcapsular layer and cortex (G–I) and at the cortical-medullary region (J–L). G and J, Superimposition of the BMI-1 and EZH2 signal; H and K, superimposition of the BMI-1 and Mib-1/Ki-67 signal; I and L, superimposition of the BMI-1, EZH2, and Mib-1/Ki-67 signal. BMI-1+/EZH2+ cells are identified by yellow fluorescence; BMI-1+/Mib-1+ cells stain purple. Note that Mib-1 expression overlaps mainly with EZH2, particularly in the cortex. BMI-1/Mib-1+ (purple cells) belong to the BMI-1+/EZH2+ (yellow) subset (solid arrows), but within the BMI-1+/EZH2+ population not all cells are dividing (dotted arrows). Original magnification, ×400.
By contrast to the abundance of BMI-1+ T cells in interfollicular and follicular regions of the lymph node, the vast majority of subcapsular and cortical thymocytes were EZH2+ (green fluorescence in Fig. 2, B and C). BMI-1 was infrequently detected (red fluorescence in Fig. 2,A), and superimposition of the BMI-1 expression pattern demonstrated three patterns of BMI-1/EZH2 expression: detection of EZH2 in the absence of BMI-1 (BMI-1−/EZH2+, green fluorescent cells in Fig. 2, B and C), detection of BMI-1 in the absence of EZH2 (BMI-1+/EZH2−, red fluorescent cells in Fig. 2, A and C), and detection of both proteins in the same nucleus (BMI-1+/EZH2+, orange and yellow fluorescent cells in Fig. 2, B and C, suggesting different expression levels of BMI-1 and EZH2 in DP cells). The most abundant group of thymocytes were BMI-1−/EZH2+ and were found throughout the cortex (Fig. 2, B and C). BMI-1+/EZH2+ thymocytes were present at a lower frequency (yellow and orange fluorescence in Fig. 2, B and C). These cells were most abundant in the subcapsular layer, but were observed throughout the cortex (Fig. 2,C). Subcapsular BMI-1+/EZH2+ cells were larger than cortical BMI-1−/EZH2+ cells. BMI-1+/EZH2− cells were the least abundant of cortical thymocytes. Most of the BMI-1+/EZH2− cells with large nuclei were CD68+ macrophages (data not shown), but rare BMI-1+/EZH2− cells with small nuclei could be detected in the subcapsular layer (dotted arrows in Fig. 2 C).
Thymocytes in the medulla had a different expression pattern of BMI-1 and EZH2 as compared with thymocytes in the cortex. Whereas expression of BMI-1 was low in both cortex and medulla (red fluorescence in Fig. 2,D), the intensity of EZH2 staining was less intense in medullar thymocytes than in cortical thymocytes (green fluorescence in Fig. 2, E and F). Occasional cells were observed in the medulla that coexpressed BMI-1 and EZH2 at low levels (solid arrows in Fig. 2, D–F).
Combined analysis of BMI-1, EZH2, and Mib-1/Ki-67 demonstrated that the three patterns of BMI-1 and EZH2 expression correlated with different stages in the cell cycle. The majority of BMI-1−/EZH2+ cells in the cortex were in cycle because they coexpressed Mib-1/Ki-67 (identified by a combination of green (EZH2) and blue fluorescence (Mib-1/Ki-67) in Fig. 2, G–L). However, the overlap between EZH2 and Mib-1/Ki-67 was not complete, suggesting that not all cortical EZH2+ thymocytes were dividing (green fluorescent cells in Fig. 2,I that lack Mib-1/Ki-67 expression). The staining pattern for BMI-1 and Mib-1/Ki-67 (red and blue signal in Fig. 2, H and K, respectively) showed that most cortical BMI-1+ cells were resting and did not express Mib-1/Ki-67. This is best illustrated by detection of red fluorescent (BMI-1+) cells in Fig. 2, H and I and K and L, which lack blue (Mib-1/Ki-67) fluorescence. However, occasional dividing BMI-1+ cells could be observed, as indicated by the presence of purple staining cells in Fig. 2, H and K (a combination of the red and blue signal for BMI-1 and Mib-1/Ki-67, respectively). Such cells always belonged to the BMI-1+/EZH2+ fraction (solid arrows in Fig. 2, G–L), but not all BMI-1+/EZH2+ cells were in cycle because some of them were clearly Mib-1/Ki-67− (dotted arrow in Fig. 2, G–I). We did not observe Mib-1/Ki-67 expression in cortical BMI-1+/EZH2− cells, suggesting that these cells are resting (for instance, ∗ in Fig. 2, G–I). Finally, detection of Mib-1/Ki-67 was infrequent in the medulla and mainly associated with BMI-1−/EZH2+ cells (Fig. 1, J–L). Notably, BMI-1+/EZH2+ cells in the medulla did not express Mib-1/Ki-67.
BMI-1+/EZH2− and BMI-1+/EZH2+ thymocytes have an immature phenotype
Because lymphocytes with strong coexpression of BMI-1 and EZH2 were less abundant than BMI-1−/EZH2+ cells, larger, and most prevalent in the subcapsular layer and the cortex of the thymus, they possibly represented thymocyte precursors. BMI-1+/EZH2− cells, mostly detected in the subcapsular layer, could also belong to this precursor population. Thymocyte subpopulations can be roughly divided using four cell surface molecules (40): CD3 and TCRαβ (both belonging to the TCR complex) and CD4 and CD8. The most immature T cell-committed thymocytes express CD3 at low levels and are CD4−/CD8− (double negative, DN) and TCRαβ−. The onset of rearrangement at the TCRB locus and the decision to become an αβ or a γδ T cell occur in this population, which also encompasses cells that still have the capacity to form NK or B cells (6, 40, 42, 43, 44). These cells differentiate to a CD4+/CD8+ (DP) TCRαβ+ stage in the cortex (40), which generally have rearranged TCRB genes and are in the process of TCRA recombination. Upon expression of the TCRαβ chain, they become subject to positive selection before entering the medulla (45). Here, they give rise to CD4+/CD8− and CD4−/CD8+ (single positive, SP) TCRαβ+ cells, which are the target of negative selection (46) before they leave the thymus and seed to the periphery. To better correlate BMI-1 and EZH2 expression profiles with different stages of T cell development, we performed triple staining for BMI-1 and EZH2 in combination with CD3, CD4, CD8, and TCRαβ (Fig. 3).
Distinct BMI-1/EZH2 expression patterns in DP thymocytes and DN precursors. BMI-1 and EZH2 were detected by red and green fluorescence, respectively. BMI-1+/EZH2+ cells are identified by yellow/orange fluorescence. BMI-1/EZH2 expression is shown in relation to CD3 (blue fluorescence in A–C), CD4 (blue fluorescence in D–F), CD8 (blue fluorescence in G–I), TCRαβ (blue fluorescence in J–L), and TCRγδ (blue fluorescence in M–O). A–C, Detail of the subcapsular layer. Note that the majority of BMI-1+ cells express CD3 (B), including BMI-1+/EZH2− (∗) and BMI-1+/EZH2+ thymocytes. BMI-1+/EZH2− and BMI-1+/EZH2+ cells are generally CD4−, CD8−, and TCRαβ− (E, H, K, and insets in E and F, H and I, and K and L showing details of the subcapsular layer). The majority of BMI-1−/EZH2+ cells are CD4+ and CD8+ (E, F, H, I, and insets), but not all of these cells express TCRαβ (K, L, and inset). BMI-1−/EZH2+ cells in the cortex and medulla were occasionally TCRγδ+ (shown for a single cortical cell in M–O; medulla, data not shown). Original magnification, ×400.
Distinct BMI-1/EZH2 expression patterns in DP thymocytes and DN precursors. BMI-1 and EZH2 were detected by red and green fluorescence, respectively. BMI-1+/EZH2+ cells are identified by yellow/orange fluorescence. BMI-1/EZH2 expression is shown in relation to CD3 (blue fluorescence in A–C), CD4 (blue fluorescence in D–F), CD8 (blue fluorescence in G–I), TCRαβ (blue fluorescence in J–L), and TCRγδ (blue fluorescence in M–O). A–C, Detail of the subcapsular layer. Note that the majority of BMI-1+ cells express CD3 (B), including BMI-1+/EZH2− (∗) and BMI-1+/EZH2+ thymocytes. BMI-1+/EZH2− and BMI-1+/EZH2+ cells are generally CD4−, CD8−, and TCRαβ− (E, H, K, and insets in E and F, H and I, and K and L showing details of the subcapsular layer). The majority of BMI-1−/EZH2+ cells are CD4+ and CD8+ (E, F, H, I, and insets), but not all of these cells express TCRαβ (K, L, and inset). BMI-1−/EZH2+ cells in the cortex and medulla were occasionally TCRγδ+ (shown for a single cortical cell in M–O; medulla, data not shown). Original magnification, ×400.
Membrane expression of CD3 (blue fluorescence in Fig. 3, B and C) was detected in most cortical BMI-1+ thymocytes (red fluorescence in Fig. 3, A and C) and all cortical EZH2+ thymocytes (green fluorescence in Fig. 3, A and C). Superimposition of the BMI-1, EZH2 and CD3 signals demonstrated that CD3 expression occurred in BMI-1−/EZH2+ and BMI-1+/EZH2+ thymocytes (green and yellow cells in Fig. 3, A and C), and BMI-1+/EZH2− thymocytes with a small nucleus (red fluorescent cell indicated by ∗ in Fig. 3 C). This indicated that cells committed to T cell differentiation express BMI-1 and EZH2 in all three patterns.
To further define the CD3+ thymocyte population in DN, DP, and SP subsets, we analyzed BMI-1 and EZH2 expression in combination with CD4 and CD8. The CD4 expression pattern is shown in Fig. 3, D–F. It shows that CD4 expression (the blue signal) is associated with EZH2-expressing cells (green signal in Fig. 3, D and F) but rarely with BMI-1+ cells (red signal in Fig. 3, D–F). This is best illustrated by subcapsular BMI-1+ cells in Fig. 3,E, which are clearly CD4− (a detail of BMI-1+/CD4− cells is shown in the inset of 3E, and the combination with the EZH2 signal in the inset of 3F). We obtained identical results with immunostaining for CD8 (Fig. 3, G–I): detection of CD8 (blue signal in Fig. 3, H and I) was primarily associated with EZH2+ cells (green signal in Fig. 3, G and I), whereas red fluorescent BMI-1+ cells in the subcapsular layer did not stain for CD8 (Fig. 3 H and insets in H and I). We concluded that BMI-1−/EZH2+ cells, the major population in the cortex, expressed both CD4 and CD8 and therefore corresponds with DP thymocytes. By contrast, BMI-1+/EZH2+ cells and BMI-1+/EZH2− cells were generally CD4−/CD8− and belong to the DN stage of T cell differentiation.
This conclusion is further supported by analysis of TCRαβ expression in combination with BMI-1 and EZH2 (Fig. 3, J–L). Membrane staining of TCRαβ (the blue signal) was primarily associated with EZH2+ (green fluorescent) thymocytes (Fig. 3, J–L). By contrast, the majority of BMI-1+ (red fluorescent) thymocytes were TCRαβ− (Fig. 3,K and inset). The presence of TCRαβ within many EZH2+/BMI-1− thymocytes is consistent with successful completion of TCRB and TCRA gene rearrangement in this thymocyte population. However, the fact that not all EZH2+/BMI-1− thymocytes express TCRαβ suggests that some of these cells are still in the process of recombination. Similarly, the absence of TCRαβ on the majority of BMI-1+/EZH2− and BMI-1+/EZH2+ cells (Fig. 3, K, L, and insets) suggests that they represent a less advanced stage of differentiation than TCRαβ+/EZH2+/BMI-1− cells. Using triple staining for BMI-1, EZH2, and TCRγδ (Fig. 3, M–O), we determined that a minority of TCRαβ− cells was TCRγδ+. TCRγδ expression was occasionally detected in cortex and medulla, and only observed in EZH2+ cells that did not express BMI-1. This demonstrated that these cells have concluded TCRGV and TCRDV rearrangement successfully, and represent more mature stages than BMI-1+/EZH2− or BMI-1+/EZH2+ thymocytes.
BMI-1 and EZH2 expression in SP thymocytes of the medulla
We found earlier that the transition between cortex and medulla is associated with a lower intensity of EZH2 staining (Fig. 4, A and D) and decreased detection of Mib-1/Ki-67. In the medulla, expression of CD4 and CD8 is separated and fully mature SP thymocytes leave the thymus at the corticomedullary junction (40). Analysis of CD4 expression (blue fluorescence in Fig. 4, A–C) showed that the majority of CD4+ medullar thymocytes expressed EZH2 at a low level (Fig. 4, A–C), whereas BMI-1 was infrequently detected in this population (Fig. 4,B). Likewise, CD8+ medullar thymocytes (blue fluorescence in Fig. 4, D–F) were primarily associated with low-level EZH2 expression (Fig. 4, D–F) and rarely expressed BMI-1 (Fig. 4, E and F). The infrequent detection of BMI-1 in SP CD4+ and CD8+ medullar thymocytes stands in sharp contrast to the abundance of BMI-1 expression in peripheral T cells.
BMI-1 and EZH2 expression in DP thymocytes of the cortex (Co) and SP thymocytes of the medulla (Me). BMI-1 and EZH2 were detected by red and green fluorescence, respectively. BMI-1/EZH2 expression are shown in relation to CD4 (blue fluorescence in A–C) and CD8 (blue fluorescence in D–F). Note that the majority of CD4+ and CD8+ cells are BMI-1− (B and E) and weakly express EZH2 (A, C, D, and F). Original magnification, ×400.
BMI-1 and EZH2 expression in DP thymocytes of the cortex (Co) and SP thymocytes of the medulla (Me). BMI-1 and EZH2 were detected by red and green fluorescence, respectively. BMI-1/EZH2 expression are shown in relation to CD4 (blue fluorescence in A–C) and CD8 (blue fluorescence in D–F). Note that the majority of CD4+ and CD8+ cells are BMI-1− (B and E) and weakly express EZH2 (A, C, D, and F). Original magnification, ×400.
Discussion
Development of pluripotent hematopoietic stem cells into mature lymphocytes follows a complex pathway of differentiation that is regulated by internal and external factors (1, 2, 3, 4, 5, 6). Each differentiation step is accompanied by expression of specific transcription factors which regulate formation of various lymphoid lineages and determine their identity. This identity must be maintained during cell division to ensure a functional immune system, but how cell type-specific gene expression patterns are propagated to the daughter cells is unclear. Developmental biology identified PcG genes as essential regulators of embryogenesis (for reviews, see Refs. 19, 20, 21, 22). They determine commitment of cells to patterns of differentiation and contribute to maintenance of cellular identity during cell division. Recent studies suggested that PcG genes have a similar role during lymphoid differentiation and maturation.
PcG proteins form multimeric protein complexes that bind and modify chromatin (23, 24, 25, 26, 27, 28, 29, 30, 31, 32). The two human PcG complexes, identified by the presence of the BMI-1 or EZH2 PcG protein (23, 24, 25, 26, 28, 29, 30, 31, 32, 33, 34, 36), are expressed in a mutually exclusive pattern in follicular B cells (15, 16). In the current study, we found that the majority of interfollicular T cells express BMI-1 and EZH2 in a pattern similar to that of follicular B cells. Most of these T cells are BMI-1+/EZH2− and resting (Mib-1/Ki-67−), whereas dividing Mib-1/Ki-67+ T cell blasts are EZH2+ in the absence of BMI-1. Similarly, BMI-1 was detected in Mib-1/Ki-67− follicular T cells while EZH2 was not. These PcG expression patterns collectively suggest that BMI-1 and EZH2 expression is strictly regulated in both T and B cells. This may reflect a difference in target genes of the two PcG complexes. In addition, the absence of BMI-1 in cycling cells is probably related to the observation that chromatin association of the BMI-1-containing PcG complex is cell cycle dependent, because BMI-1 dissociates from chromosomes during the late S-G2-M phase of cell division (47). Suppression of the BMI-1 gene in knockout mice resulted in inhibition of cell proliferation, whereas the absence of EED (belonging to the EZH2-containing complex) correlated with increased proliferation (11). These results suggest that normal regulation of cell division depends on a balance between the BMI-1- and EZH2-containing PcG complex.
The mutually exclusive expression pattern of BMI-1 and EZH2 in peripheral mature T cells is not immediately established in T cell precursors and different differentiation stages of thymocytes correlated with distinct BMI-1 and EZH2 expression profiles (summarized in Fig. 5). We showed that DN cells in the subcapsular layer and cortex coexpress BMI-1 and EZH2. In addition, a minor population of DN cells expressed the BMI-1 gene in the absence of EZH2. The two BMI/EZH2 expression patterns suggest different silencing patterns of PcG target genes and may reflect the heterogeneity of the DN thymocyte population. It is unclear whether BMI-1+/EZH2− and BMI-1+/EZH2+ DN cells represent two separate lineages or whether one originates from the other. However, the frequent detection of CD3 in both populations suggests that each contains T cell-committed precursors. We theorize that subcapsular resting cells, which express BMI-1 in the absence of EZH2, precede BMI-1+/EZH2+ thymocytes. This sequence of events is supported by RT-PCR experiments on purified CD34+ human bone marrow cells, which showed that the most primitive long-term culture-initiating cells contain the highest level of BMI-1 transcripts and the lowest level of EZH2 transcripts (10). In more mature populations, expression of EZH2 progressively increased while BMI-1 transcripts decreased to being minimal in CD34− cells. If PcG expression follows a similar pattern during differentiation of DN thymocyte precursors, CD3+/BMI-1+/EZH2− DN cells may be the precursors of the CD3+/BMI-1+/EZH2+ DN stage. The known steps of T cell differentiation and reactivity with Mib-1/Ki-6 allow us to speculate further about a possible relationship between BMI-1/EZH2 expression and TCR gene rearrangement status. Proliferation in developing thymocytes is induced after TCR β-chain expression in the pre-TCR complex (48), and detection of Mib-1/Ki-67 expression in BMI-1+/EZH2+ DN/TCRαβ− thymocytes suggests that these cells have completed rearrangement of the TCRB locus. Since these cells are TCRαβ−, rearrangement of the TCRA genes has not yet occurred or has yet to produce a functional TCRα protein.
Model of BMI-1 and EZH2 PcG gene expression during human T cell differentiation. Most cortical DN (CD4−/CD8−) thymocyte precursors coexpress BMI-1 and EZH2, although a small subset of DN cells is BMI-1+/EZH2−. These cells mainly reside in the subcortical zone (SCZ) and possibly precede the BMI-1+/EZH2+ stage. The majority of these DN cells are resting (Mib-1−). During transition to the DP stage, BMI-1 expression is lost while TCRαβ or TCRγδ membrane expression occurs. This coincides with induction of proliferation (Mib-1+ stage). Hypothetically, rare proliferating (Mib-1+) cells within the DN subset could reflect expression of the TCR β-chain in the absence of successful TCRA recombination. SP cells in the medulla are resting and express EZH2 at a lower level than cortical DP cells. It is unknown whether the limited number of BMI-1+/EZH2− medullary T cells reflect precursors to BMI-1+/EZH2− interfollicular T cells. For further details, see text.
Model of BMI-1 and EZH2 PcG gene expression during human T cell differentiation. Most cortical DN (CD4−/CD8−) thymocyte precursors coexpress BMI-1 and EZH2, although a small subset of DN cells is BMI-1+/EZH2−. These cells mainly reside in the subcortical zone (SCZ) and possibly precede the BMI-1+/EZH2+ stage. The majority of these DN cells are resting (Mib-1−). During transition to the DP stage, BMI-1 expression is lost while TCRαβ or TCRγδ membrane expression occurs. This coincides with induction of proliferation (Mib-1+ stage). Hypothetically, rare proliferating (Mib-1+) cells within the DN subset could reflect expression of the TCR β-chain in the absence of successful TCRA recombination. SP cells in the medulla are resting and express EZH2 at a lower level than cortical DP cells. It is unknown whether the limited number of BMI-1+/EZH2− medullary T cells reflect precursors to BMI-1+/EZH2− interfollicular T cells. For further details, see text.
In contrast to DN cells, we found that the majority of DP cells express high levels of EZH2 with little or no BMI-1. The transition from the DN to DP stage therefore coincides with loss of BMI-1 expression and the continued presence of EZH2. This change in PcG expression probably means that different gene silencing patterns are established in thymocytes that differentiate from the DN to the DP stage. Most of these cells are Mib-1/Ki-67+, and the absence of BMI-1 in these cycling cells resembles the situation in mature peripheral T cells where Mib-1/Ki-67 and BMI-1 expression are separated. However, the presence of Mib-1/Ki-67 in a subpopulation of BMI-1+/EZH2+ DN cells shows that proliferation starts before the appearance of CD4 and CD8 and loss of BMI-1 expression (possibly related to expression of the TCR β-chain). Most DP EZH2+/BMI-1− cells expressed TCRαβ and have completed recombination of the TCRA and TCRB Ag receptor genes. A minority of BMI-1−/EZH2+ DP cells was TCRαβ− and probably represent cells that failed to generate a functional TCRαβ gene or are still in the process of recombination. This subset also included a low number of TCRγδ+ T cells.
The final phase of thymocyte development is the SP stage, where expression of CD4 and CD8 is separated. SP cells are primarily located in the medulla and undergo negative selection by clonal deletion (40, 46). The transition between DP and SP cells correlated with a lowered expression of EZH2 and a continued relative absence of BMI-1. This expression pattern is markedly different from that in mature T cells of the lymph node, which preferentially express BMI-1. In addition, EZH2 expression in peripheral T cells is associated with cell division (expression of Mib-1/Ki-67), whereas most medullar EZH2+ thymocytes are resting and Mib-1/Ki-67−. This demonstrates that the BMI-1/EZH2 expression pattern of mature peripheral T cells is not yet established in medullar SP thymocytes, and suggests that these cells have not fully matured with respect to PcG expression profile. Recent thymic emigrants are known to further mature in the periphery, as evidenced by acquisition of functional competence and evolution of surface markers (49, 50). This possibly coincides with the appearance of the PcG expression pattern that is observed in T cells of the lymph nodes. Alternatively, the few BMI-1+/EZH2− medullar thymocytes may be the cells that actually leave the thymus, while BMI-1+/EZH2low thymocytes could represent cells that are destined to die.
Experiments in mutant mice convincingly demonstrated that PcG genes encode proteins that are essential for hematopoiesis (for reviews, see Refs. 18, 51). A range of effects on lymphoid tissue of these animals has been observed, including enhanced lymphoid proliferation in BMI-1-transgenic mice and EED knockout mice (11, 39), inhibition of precursor proliferation and severe thymic hypoplasia in BMI-1 and Mel-18 knockout mice (7, 8, 52), B cell maturation defects (53), and development of lymphomas (39, 54, 55). Although their role in human lymphopoiesis is unclear, the distinct patterns of BMI-1/EZH2 expression in DN, DP, and SP thymocytes suggest that PcG genes contribute to the regulation of human T cell development as well.
The most obvious mechanism of regulation is suppression of homeobox gene expression. Homeobox genes are known to affect lymphopoiesis and select groups of homeobox genes are expressed in various T lymphocyte subsets, while others are suppressed (4, 56, 57, 58, 59, 60, 61). Since PcG proteins contribute to silencing of homeobox genes (19, 20, 21), it is reasonable to expect that they are involved in this process. Different PcG expression patterns in distinct lymphocyte subpopulations possibly reflect stage-specific expression of homeobox genes. Two other candidate proteins for PcG-mediated regulation are Vav and E2F6. Vav is a rho family GTP/GDP exchange factor that is involved in regulation of positive and negative selection and Ag receptor-mediated proliferation (62, 63, 64). Vav is preferentially expressed in hematopoietic cells (65). The possibility that EZH2 and Vav affect each other is suggested by the observed interaction between Vav and EZH2 (66). E2F6 is a transcription factor that is preferentially expressed by CD34+ lymphoid precursors and is involved in regulation of apoptosis (67). Similarly, the BMI-1-containing PcG complex is specifically expressed in thymocyte precursors (this study) and may be involved in E2F6 function because E2F6 can be part of this PcG complex (68).
In conclusion, we demonstrated that expression of the BMI-1- and EZH2-containing PcG complexes in mature T cells is mutually exclusive, and that this pattern is not yet established in differentiating thymocytes. The various stages of T cell differentiation are associated with profound changes in PcG expression patterns. This suggests a regulatory role for PcG genes in lymphopoiesis, as previously suggested by PcG knockout mice and PCR assays on lymphoid precursor populations in bone marrow. The change in PcG expression during T cell development probably reflects stabilization of cell type-specific gene expression patterns and irreversibility of lineage choice.
Acknowledgements
We thank Jeroen Belien for assistance with confocal laserscan microscopy and generation of the figures and Jannie Borst for providing the TCRγδ antiserum.
Footnotes
Abbreviations used in this paper: PcG, Polycomb-group; GC, germinal center; DP, double positive; DN, double negative; SP, single positive; GaM, goat anti-mouse.