A Bimodal Influence of Thyroid Hormone on Cerebellum Oligodendrocyte Differentiation

Abstract

Thyroid hormone (T₃) can trigger a massive differentiation of cultured oligodendrocytes precursor cells (OPC) by binding the nuclear T₃ receptor α1 (TRα1). Whether this reflects a physiological function of TRα1 remains uncertain. Using a recently generated mouse model, in which CRE/loxP recombination is used to block its function, we show that TRα1 acts at two levels for the in vivo differentiation of OPC in mouse cerebellum. At the early postnatal stage, it promotes the secretion of several neurotrophic factors by acting in Purkinje neurons and astrocytes, defining an environment suitable for OPC differentiation. At later stages, TRα1 acts in a cell-autonomous manner to ensure the complete arrest of OPC proliferation. These data explain contradictory observations made on various models and outline the importance of T₃ signaling both for synchronizing postnatal neurodevelopment and restraining OPC proliferation in adult brain.

Oligodendrocytes (OL) synthesize myelin in the vertebrate central nervous system. They develop from oligodendrocyte precursor cells (OPC), which undergo terminal differentiation at a postnatal stage of neurodevelopment in mice (1). The switch between the self-renewal and proliferation of OPC toward differentiation is sensitive to a number of diffusible factors, produced by other neural cell types in their immediate environment. The best studied are platelet-derived growth factors (PDGF) (2), neurotrophin-3 (NT3) (3), brain-derived neurotrophic factor (BDNF) (4), nerve growth factor (NGF) (5), and Sonic hedgehog (Shh) (6). Because OPC and OL also secrete growth factors (7), they participate to a complex network of cellular interactions that coordinates neurodevelopment. OPC also interact with neighboring cells by direct contact, receiving synaptic input from neurons (8). Newly formed OL can respond to these cell contacts and neurotrophins during a short time frame (9). Timely OPC differentiation is important, because myelination, which usually begins just after axonal growth and can last for more than 1 wk, is necessary to maintain axon integrity (10). OPC are maintained in adult brain, and their function and lineage potential are matters of intense investigations (11).

Culture experiments have shown that OPC can be maintained in chemically defined medium, in which purified PDGF and/or NT3 is sufficient to promote their self-renewal (3, 12). Retrieval of these growth factors combined with addition of the active form of thyroid hormone (T₃) triggers a rapid and massive differentiation into postmitotic OL. Unlike peptidic neurotrophic factors, T₃ acts directly on transcription by binding to TRα1, TRβ1, or TRβ2 nuclear receptors encoded by the THRA and THRB genes (13). THRA is expressed in both OPC and OL, whereas THRB expression starts only at a late stage of OPC differentiation (14). Moreover, circulating T₃ peaks at postnatal d 6 (P6) in mice (15). These observations suggest an elegant and simple mechanism for the temporal control of OL differentiation in the rodent brain white matter: OPC proliferation is first stimulated by NT3 and PDGF. Subsequently, as NT3 production decreases and OPC become less sensitive to PDGF (16), a peak of T₃ would be sufficient to ensure the timely differentiation of OPC into OL.

The importance of T₃ signaling for OPC in vivo differentiation, however, remains unclear. T₃ deficiency delays myelination and the onset of OL differentiation, in both mouse and rats (17–19) whereas excess of T₃ has the opposite effect (20). OPC differentiation is slightly delayed in THRA, but not THRB, knockout (21) whereas a population of slow cycling OPC persists in the adult optic nerve in mice devoid of all receptors (22). The in vivo consequences of altering T₃ signaling are thus much less dramatic than in cultured cells, raising the possibility that other factors can compensate for T₃ deficiency. Moreover, THRA knockout can affect OPC differentiation indirectly, as it is expressed in all the major cerebellar cell types including Purkinje neurons, granular neurons, OL, and astrocytes (23–25). In fact, T₃ deficiency can affect the ability of these cell types to secrete factors acting on OPC proliferation including NT3, BDNF, and nerve growth factor (26–29).

In the present study, we asked whether the function of T₃ signaling in cerebellum OPC differentiation is cell autonomous or indirect, by using the so-called TRα^AMI THRA allele (30). Cerebellum was chosen because it is a region in which consequences of impaired T₃ signaling are most striking (31), affecting γ-aminobutyric acid (GABA)ergic Purkinje neurons, GABAergic interneurons (32), granular glutamatergic neurons (33), Bergmann glia (34), astrocytes, and oligodendrocytes (35). Cre-mediated recombination of the TRα^AMI allele activates the expression of TRα1^L400R, a dominant-negative mutant receptor. Unlike what happens in THRA knock-out mice, in which corepressors recruitment by unliganded TRα1 is lost (36, 37), expression of T₃ target genes is strongly repressed in TRα1^L400R-expressing cells. As a consequence, when TRα1^L400R expression is made ubiquitous by early Cre expression, heterozygous mice display many features of congenital hypothyroidism and do not usually survive beyond 3 wk. However, T₃ level is not changed and TRβ1/2 functions seem to be preserved (30).

To analyze the consequences of restricted expression of the TRα^AMI allele on OPC differentiation, we used a panel of promoter-specific Cre (or the tamoxifen-inducible Cre-ER^T2) transgenic mice that allowed expression of TRα1^L400R in different cell types and time frames. Our data demonstrate that T₃ exerts two completely distinct functions at different stages, explaining apparent contradictions between the conclusions of several previous studies.

Results

Recombinase expression pattern in different Cre and Cre-ER^T2 mice

We generated mice expressing the TRα1^L400R dominant-negative mutation in a restricted manner by crossing TRα^AMI/AMIR26YFP mice with Cre or Cre-ER^T2 transgenics. The Cre-dependent R26YFP transgene was introduced to control by immunostaining the Cre recombination pattern and efficiency in all groups of transgenics. We performed yellow fluorescent protein (YFP) immunostaining in the cerebellum at different postnatal stages and used the result as an indirect evidence for TRα1^L400R expression, because the mutant receptor cannot be distinguished from its wild-type counterpart still present in heterozygous mice. Previous work, in which broad Cre expression allowed the use of PCR analysis, showed that the TRα^AMI and R26YFP display the same pattern and same efficiency of recombination, and that this approach is reliable (Refs. 30 and 38 and Supplemental Table 1 published on The Endocrine Society's Journals Online web site at http://mend.endojournals.org). We first wanted to generate transgenics able to drive recombination specifically in OPC using the two strains that have been reported to do so (39, 40). Oligodendrocyte transcription factor 2 (Olig2), a basic helix-loop-helix transcription factor, was used as a specific marker for immunostaining of OPC (41, 42) and has been combined with YFP immunostaining (Fig. 1). To our surprise neither Cnp-Cre nor PDGFRα-CreER^T2 restricted YFP expression to OPC in the cerebellum. First, YFP was observed in approximately 100% of Purkinje cells in Cnp-CrexR26YFP and Cnp-CrexR26YFPxTRα^AMI mice after P8 but not at P3 (Fig. 1 and Table 1). This is in agreement with published data (43) and the Allen Brain Atlas information. Second, when PDGFRα-CreER^T2xR26YFP mice were treated with tamoxifen either at P2, P5, or P30, the YFP expression pattern, expressed several days later, was much broader than expected, being found not only in OPC and OL but also in astrocytes and most if not all neuronal cell types, including Purkinje and granular cells. The weak PDGFRα expression found outside OPC (44) is thus sufficient to promote recombination at this stage. These unexpected results prompted us to use several other transgenics, to reach a conclusion on a possible cell-autonomous effect of TRα1^L400R expression in OPC. In particular L7-Cre allowed recombination only in Purkinje neurons, at P8 but not at P3, and provides a satisfying control for Cnp-Cre. We also used transgenes recombining in all GABAergic neurons, including Purkinje and GABAergic interneurons (Ptf1a-Cre) or astrocytes lineage (Glast-Cre-ER^T2) to address the possible influence of a cerebellar environment expressing TRα1^L400R on wild-type OPC differentiation. Comparisons within this collection of Cre and Cre-ER^T2 transgenic mice should allow identification of the main glial and neuronal cell types in which TRα1^L400R has cell-autonomous consequences, with the noticeable exception of GABAergic interneurons, the influence of which before P8 cannot be separated from that of Purkinje cells. Expression patterns, recombination efficiencies, and simplified nomenclature are summarized in Table 1 and in Fig 1, panel I.

Fig. 1.

Expression pattern of different Cre-mice in the cerebellum, using ROSA26-lox-STOP-lox-EYFP (R26YFP) as reporter. A–F, Olig2 immunostaining in red; YFP immunostaining in green,. Colocalization between YFP and Olig2 (yellow) indicates Cre-mediated recombination in OPC. White lines delineate the Purkinje cell layer. A, PDGFRα-Cre ER^T2 (tamoxifen injection at P1. Immunostaining at P15) indicates cre-mediated recombination in all cerebellar cell types including oligodendrocyte lineage. B, L7-Cre (P15). Recombination is restricted to Purkinje cells. C, Glast-Cre ER^T2 (tamoxifen injection at P1. Immunostaining at P15). Recombination is restricted to the astrocytes and not found in Olig2+ OPC. D, Ptfa1-Cre (P15): Recombination in GABAergic neurons, but not in oligodendrocytes lineage at P15. E and F, Cnp-Cre (P15). Recombination limited to the oligodendrocytes lineage and Purkinje cells. Stars indicate YFP+Olig2+ OPC. G, Cnp-Cre (P8, green; YFP, red; parvalbumin, blue: DAPI) indicates an expression of Cre recombinase from in Purkinje cells at this stage. H, Cnp-Cre (P3 green: YFP, red: parvalbumin). Absence of recombination at P3 in Purkinje cells at this stage. Scale bar, 100 μm. I, Cell lineages in the cerebellum and recombination patterns observed in various Cre and Cre-ER^T2 transgenic mice. X indicates Cre expression. Counting and nomenclature are summarized in Table 1. GFP, Green fluorescent protein.

Table 1.

Genotypes, mice, and nomenclature

Promoter	Nomenclature	Granular neurons	GABAergic neurons		Glia		Ref.
			Purkinje cells	GABAergic inter-neurons	OL	Astrocytes
Sycp1 (synaptonemal complex protein 1)	TRαAMI/S	100	100	100	100	100	69
Nes (Nestin)	TRαAMI/N	100	100	100	100	100	70
Cnp (2′,3′-cyclic nucleotide 3′- phosphodiesterase)	TRαAMI/C	0	74.1 ± 11.4	0%	91.6 ± 7.2	0	39
Pcp2 (L7, Purkinje cell protein 2)	TRαAMI/liter	0	98.3 ± 4.1	0	0	0	71
Ptf1a (pancreas transcription factor 1a)	TRαAMI/P	0	72.5 ± 17.3	87.3 ± 7.1	0	0	72
Tamoxifen inducible CreERt2
Slc1a3 (Glast, solute carrier 1a3)	TRαAMI/S/Gt2/Pna	0	0	0	0	92.8 ± 4.9	73
Pdgfra (platelet-derived growth factor receptor, α polypeptide)	TRαAMI/S/Pt2/Pna	>99b	>99b	>99b	98.7 ± 2.3	>99b	40

% of YFP+ observed at P15 in TRα^AMI/ROSAYFP/CRE cerebellum.

^a Pn, Postnatal day of tamoxifen treatment (P1 for initial characterization); OL, oligodendrocyte lineage, characterized using Olig2 staining. n = 3 for each phenotype.

^b No DAPI+/YFP− cells detected.

TRα1^L400R expression increases OPC postnatal proliferation in an indirect manner

First, study of TRα^AMI/S mice was performed to address the influence of ubiquitous expression of TRα1^L400R on OPC differentiation and proliferation. Because the intact TRα^AMI allele is a nul allele, expressing none of the known TRα isoforms, we used heterozygous TRα^AMI/+ mice as controls rather than wild-type animals. Quantitative RT-PCR analysis of whole cerebellum RNA indicated increased expression of PDGFRα, a marker of OPC encoding the platelet-derived growth factor receptor α, and decreased expression of Mbp, a specific marker of mature oligodendrocyte encoding myelin basic protein (Fig. 2A). Quantitative Western blot analysis showed an equivalent reduction in the MBP protein level at the P15 (Fig. 2B). Because these methods provide only indirect indications that OPC differentiation is affected, we used immunostaining to count Olig2+ cells in cerebellum lobule VIII of TRα^AMI/S. This revealed a slight excess of OPC at P15 (Fig. 3A and Table 1) and confirmed that increased proliferation and/or delayed OPC differentiation change the cellular composition of TRα^AMI/S cerebellum. Pan-neural expression of TRα1^L400R in TRα^AMI/N and TRα^AMI/P^T2/P5 (tamoxifen injection at P5) mice also resulted in an increased Olig2+ cell numbers (Fig. 3A and Table 2). Delay in the onset of OL appearance was confirmed by using the CC-1 monoclonal antibody, which marks mature OL (45) and reveals a reduced density at P15 (Supplemental Fig. 1). Altogether, these data indicate that ubiquitous, pan-neural and pan-cerebellar (from P5) expression of TRα^AMI allele delays OPC differentiation. This phenotype is mainly visible at P15.

Fig. 2.

Delayed postnatal differentiation of OPC after ubiquitous TRα1^L400R expression. A, Quantification of mRNA relative abundance in whole cerebellum at different ages (P8, P15, and P21). Comparison between ubiquitous mutants (TRα^AMI/S) mice (n = 6, in black) and control littermates (n = 6, in gray). PDGFRα is specific of OPC, and Mbp is a marker of mature OL. P15 wild-type whole cerebellum RNA was used as a reference. B, Myelin basic protein quantification at P15 by Western blot analysis (n = 5), normalized by β-tubulin (β-Tub) indicates a delay in myelination in mutant expressing the TRα1^L400R ubiquitously (TRα^AMI/S). Data are expressed ± sd. *, Significant difference between mutants and wild types, P < 0.05. MBP, Myelin basic protein.

Fig. 3.

An indirect effect of T₃ for OPC differentiation at P15. A, Olig2 staining at P15 (in green. DAPI Counterstain blue) indicates an increased density of OPC in TRα^AMI/S (ubiquitous expression), TRα^AMI/N (brain specific), and TRα^AMI/P^T2/P5 (pan-cerebellar expression from P5) compared with littermate control. See Table 2 for counts. B, Olig2 staining in the white matter of control (without Cre), oligodendrocyte specific (TRα^AMI/C), GABAergic lineage specific (Purkinje cell and GABAergic interneurons, TRα^AMI/P), and astrocyte specific from P1 (TRα^AMI/G^T2/P1) indicate a not cell-autonomous effect of thyroid hormone in oligodendrocyte differentiation at P15. C, Same staining at P21 indicates a transient phenotype, mainly observable at P15. D, 24 h BrdU labeling of proliferating cells. Double BrdU/Olig2+ (red/green, colocalization in yellow) E, Counting of double BrdU/Olig2+ cells indicates an increased density in proliferating OPC at P15 when TRα1^L400R is expressed in GABAergic neurons (TRα^AMI/P) mutants or in astrocytes (TRα^AMI/G^T2/P1) but not in OPC (TRα^AMI/C). Data are expressed ± sd. *, Significant difference between mutants and wild types, P < 0.05. Scale bar, 50 μm.

Table 2.

Counts of olig2+ OPC on cerebellum lobule VIII slices

Genotype	Relative Olig2+ density at P15 mean ± sd (±sd of control littermates)			Relative Olig2+ density at P21 mean ± sd (±sd of control littermates)
	White matter	Total	n	White matter	Total	N
TRαAMI/S	0.94 ± 0.30 (±0.01)	1.39 ± 0.21 (±0.06)a	3	0.85 ± 0.1 (±0.01)a	0.93 ± 0.14 (±0.15)	2
TRαAMI/N	1.24 ± 0.11 (±0.13)a	1.23 ± 0.14 (±0.18)a	4	0.91 ± 0.11 (±0.06)	1.14 ± 0.27 (±0.22)	4
TRαAMI/C	1.12 ± 0.11 (±0.16)	1.19 ± 0.19 (±0.16)	5	0.97 ± 0.25 (±0.24)	1.09 ± 0.33 (±0.20)	6
TRαAMI/liter	0.98 ± 0.14 (±0.12)	0.93 ± 0.13 (±0.29)	4	1.06 ± 0.19 (±0.19)	1.07 ± 0.21 (±0.36)	3
TRαAMI/P	1.61 ± 0.22 (±0.18)a	1.57 ± 0.15 (±0.11)a	6	1.09 ± 0.10 (±0.09)	1.11 ± 0.06 (±0.06)	5
TRαAMI/GT2/P1	1.53 ± 0.36 (±0.12)	2.39 ± 0.46 (±0.05)a	3	1.21 ± 0.17 (±0.01)	1.19 ± 0.17 (±0.01)	4
TRαAMI/PT2/P5	1.49 ± 0.28 (±0.22)a	1.51 ± 0.30 (±0.13)a	5	Not determined	Not determined

Relative density was measured on cerebellum slices as described in Materials and Methods.

^a Significant changes, compared with control littermates, are in bold (P < 0.05).

We next asked whether the changes observed at P15 in mice expressing TRα1^L400R ubiquitously were due to the cell-autonomous inhibition of T₃ signaling. We therefore counted Olig2+ cells of mice cerebellum expressing TRα1^L400R in a restricted manner. Surprisingly Olig2+ cell density remained unchanged when TRα1^L400R expression was in OPC (TRα^AMI/C; Fig. 3B and Table 2), suggesting that the observed phenotype is not a cell-autonomous consequence of TRα1^L400R expression. Conversely, Olig2+ cell density was increased at P15), at the expense of OL (Supplemental Fig. 1), when expression was either in Purkinje neurons and GABAergic interneurons (TRα^AMI/P) or in astrocytes and Bergmann glia (TRα^AMI/G^T2/P1, Fig. 3B and Table 2). Although we cannot formally rule out a possible intervention of GABAergic interneurons in the determination of the TRα^AMI/P phenotype, Purkinje cells are more likely to be involved (see Discussion). If this is true, expression of TRα1^L400R in Purkinje neurons impacts OPC differentiation when it occurs early in development (TRα^AMI/P) but not after P8 (TRα^AMI/L, TRα^AMI/C). Therefore, we conclude that TRα1^L400R expression impairs OPC differentiation at this stage in an indirect manner, by acting on both astrocytes and Purkinje cells, and perhaps GABAergic interneurons, before P8. This phenotype is mainly visible at P15, a recovery being apparent at P21 (Fig. 3C and Table 2).

To better characterize the indirect consequences of expressing TRα1^L400R in astrocytes or Purkinje neurons, mitotic activity of OPC was assayed by treating mice with bromodeoxyuridine (BrdU) 24 h before histochemical analysis (Fig. 3, D and E). The number of Olig2+/BrdU+ cells was increased when TRα1^L400R expression was in Purkinje neurons (TRα^AMI/P) or astrocytes (TRα^AMI/G^T2/P1), indicating that the increase of Olig2+ cells number correlated with an increase in OPC mitotic activity (Fig. 3E). In an attempt to identify signaling pathways involved in these indirect effects, we performed gene expression profiling by quantitative RT-PCR at P8. Pan-neural expression of TRα1^L400R in TRα^AMI/N mice correlated with low Ntf3, Bdnf, Pdgf, P75ntr mRNA levels, and down-regulation of three genes encoding component of the hedgehog pathways: Shh, Ptch1, and Gli2 (Table 3). Decrease in NT3, BDNF, and Shh signaling suggests a global deficiency in the cerebellum developmental process. Among other cell types, it can thus affect granular neurons, which express Ntf3 and Bdnf, and Purkinje cells, which express Shh. OPC differentiation can then be indirectly influenced because it is sensitive to the corresponding signaling pathways. However, TRα1^L400R expression in early Purkinje neurons (TRα^AMI/P mice) affects only Ntf3 mRNA level, indicating that Bdnf, Shh, and Ntf3 deregulation, as observed in TRα^AMI/N mice, is not a consequence of TRα1^L400R expression in Purkinje cells. Because Ntf3 expression gradually decreases during normal cerebellar development, we addressed the possibility that changes in Ntf3 mRNA level at P15 were only a consequence of delayed granular neurons maturation, by measuring Gabra6 mRNA level. This gene encodes the α6-subunit of the GABA receptor, and its expression, restricted to inner granular layer neurons, increases gradually during postnatal maturation. Gabra6 mRNA level was changed by broad TRα1^L400R expression (TRα^AMI/N) but not when it was in Purkinje neurons (TRα^AMI/P). Therefore, delayed granular neurons maturation cannot be the only explanation for the observed changes in Ntf3 expression (Table 3). The mechanism underlying the indirect effect of TRα1^L400R on early OPC therefore involves several cell types, including astrocytes, Purkinje neurons, and granular neurons and permanent interactions mediated by cell contacts and exchange of diffusible factors.

Table 3.

Quantitative RT-PCR analysis of gene expression in whole cerebellum at P8 and P60

A. Total cerebellar mRNA at P8
Gene	Control group mean (±sd) (n = 6)	TRαAMI/N mean (±sd) (n = 6)	TRαAMI/P mean (±sd) (n = 6)
Ntf-3	100 ± 6.2	67 ± 6.6a	144.0 ± 14.9a
Ngf	100 ± 28.2	99.3 ± 12.3	96.6 ± 11.1
Bdnf	100 ± 49.6	20 ± 2.4a	152.0 ± 54
Pdgf	100 ± 5.4	79 ± 8.4a	101.9 ± 22.1
p75	100 ± 6.5	37 ± 4.4a	89.3 ± 5.1
Shh	100 ± 13.2	46 ± 4.7a	105.4 ± 35
Gli2	100 ± 8.9	26 ± 3.4a	126.8 ± 39.7
Ptch1	100 ± 18.1	33 ± 5.0a	114.8 ± 16.5
Gabra6	100 ± 16.5	42 ± 18.3	109 ± 8.7

B. Total cerebellar mRNA at P60
Gene	Control group cre-mean (±sd) (n = 5)	TRαAMI/Pt2/P30 mean (±sd) (n = 5)	TRαAMI /Pt2 no tamoxifen mean (±sd) (n = 5)
Olig2	100 ± 25.3	198.0 ± 34.2a	93.0 ± 16.2
Pdgf-aa	100 ± 14.2	102.3 ± 34.5	107.8 ± 20.1
Plp	100 ± 16.5	92.5 ± 43.9	128.0 ± 17.1
Mbp	100 ± 16.3	76.1 ± 15.3	97.8 ± 19.0
Golli Mbp	100 ± 24.0	78.3 ± 14.7	108.2 ± 25.3
TRa1	100 ± 20.8	135.0 ± 25.6	106.3 ± 12.5

Data normalized using Hprt expression as a reference.

^a Significant difference (bold characters), P < 0.05.

A late cell-autonomous effect of T₃ on OPC

In all mice, OPC differentiation appeared to be only delayed, as, by P21, Olig2+ OPC cell numbers fell within control values in all genotypes (Fig. 2C and Table 2). Broad expression of TRα1^L400R, which occurs in TRα^AMI/S, TRα^AMI/N, or TRα^AMI/P^T2/P5 mice is usually lethal before P25, precluding any analysis at a later stage. However, the long-term consequences of TRα1^L400R expression on cerebellum OPC fate can be addressed in the mice expressing TRα1^L400R in a restricted manner, or only after the critical weaning period, by activating recombination with tamoxifen at a late stage (TRα^AMI/P^T2/P30) (Fig. 4). Including the R26YFP reporter transgene allowed us to verify that Cre recombination patterns at P60 were identical to those observed at early stages for TRα^AMI/C, TRα^AMI/P (data not shown) or TRα^AMI/P^T2/P30 (Fig. 4I) and to trace the fate of cells undergoing recombination. In adults, few Olig2+ OPC were present in wild-type cerebellum white matter (Fig. 4, A and E). By contrast to what was observed at P15, the density of Olig2+ OPC was not increased in TRα^AMI/P mice (Fig. 4, D and H), confirming that expression of TRα1^L400R in Purkinje cells has only a transient influence on OPC differentiation. Conversely, when OPC express TRα1^L400R (TRα^AMI/C), density dramatically increased at P60, indicating that this late effect of TRα1^L400R on OPC proliferation is cell autonomous (Fig. 4, B and F). Interestingly, the accumulation of OPC-derived cells, as judged by YFP staining, was not observed in the white matter of other brain areas, including corpus callosum, in which suggestive observations have been made in hypothyroid rats (46).

Fig. 4.

TRα1^L400R stimulates adult OPC proliferation in a cell-autonomous manner. A–H, Immunostaining at P60 of Olig2+ OPC (red; DAPI counterstain in blue) compared with control (A and E). Density is increased when TRα1^L400R is expressed in the oligodendrocyte lineage (B and F, TRα^AMI/C, 12-fold increase compared with littermate controls) or in all cell types starting from P30 (C and G, TRα^AMI/P^T2/P30, 14-fold increase compared with littermate controls), but not if expression is restricted to GABAergic neurons (D and H, TRα^AMI/P). I, YFP/Olig2 immunostaining confirms that Cre recombination occurs in the oligodendrocyte lineage of TRα^AMI/P^T2/P30/R26YFP mice (green, YFP; red, Olig2; yellow, merge). J and K, Expression of TRα1^L400R in oligodendrocyte lineage increases the number of YFP+/EdU+ cells at P60, 7 d after EdU injection (YFP, red; and EdU, green). J, Control CrexR26YFP K, TRα^AMI/CxR26YFP. L and M, Colocalization of Olig2 and EdU in TRα^AMI/C white matter (blue, DAPI; gree:, EdU; red, Olig2). n = 4 for each genotype. Scale bar, 200 μm (panels A–L), 10 μm (panel M).

We then asked whether the late accumulation of OPC in cerebellum white matter requires the permanent expression of TRα1^L400R, by inducing a broad expression only after cerebellum maturation (TRα^AMI/P^T2/P30). The important increase of Olig2+ cell density found in these mice 30 d later showed that proliferation is stimulated by TRα1^L400R, even when expression starts after the completion of the myelination process (Fig. 4, C and G). R26YFP reporter transgene indicated that, in the white matter, most of the cells that underwent recombination were OPC, which did not differentiate within 30 d (Fig. 4I). TRα^AMI/P^T2/P30 cerebellum also showed an increase in Olig2 mRNA level at P60, correlating with the change in cellular content, but mRNA levels were not changed for Mbp, GolliMBP, and Plp, which mark differentiated OL (Table 3). These observations further support the idea that adult OPC expressing TRα1^L400R are unable to differentiate.

5-Ethynyl-2′ deoxyuridine (EdU) injection at P53, i.e. 7 d before analysis, revealed the presence of proliferating cells among the OPC that expressed TRα1^L400R (TRα^AMI/C/R26YFP) indicating that a permanent but slow proliferation of Olig2+ OPC was responsible for the observed accumulation of Olig2+ cells in the white matter (Fig. 4, J, K, L, and M). As recombination takes place at an early stage for these TRα^AMI/C/R26YFP mice, OL also express YFP. Taken together, these data lead us to conclude that T₃ exerts a permanent control of OPC proliferation in adults, in nonpathological situations. This control is mediated by TRα1 and is cell autonomous.

Discussion

We previously reported that, like T₃ deficiency, ubiquitous expression of TRα1^L400R mutation retards myelin formation (47). Because the phenotype was observed in heterozygous mice, and T₃ level was unchanged in this transgenic model, other TR isoforms (TRβ1, TRβ2, TRα2, TRΔα1, etc.) were not involved. However, the ubiquitous expression of TRα1^L400R being usually lethal, we could not decide whether systemic disorders could be compromising neurodevelopment in an indirect and not specific manner. By using somatic recombination we show here that a combination of direct and indirect effects explains the influence of TRα1 on the oligodendrocyte lineage. When T₃ is present at physiological levels, TRα1 promotes OPC cell cycle exit and commitment to differentiation both in juveniles and adults. However, TRα1 exerts its function at different stages in two very different manners. Soon after birth, liganded TRα1 acts primarily on Purkinje cells and astrocytes, triggering a combination of neurotrophic factors secretion and an intricate network of cellular interactions involving granular neurons, Purkinje neurons, and astrocytes. These cellular interactions indirectly ensure the timely differentiation of OPC. Because we are unable to restrict TRα1^L400R to Purkinje cells before P8, we cannot rule out an involvement of immature GABAergic interneurons at an early stage. Based on the parsimony principle, we believe that such a possibility should not be taken into account at this point because, unlike Purkinje cells, GABAergic interneurons are not known to secrete neurotrophins. Second, they differentiate mainly at a later stage. This prompts us to believe that the TRα1 effect on early OPC differentiation is only relayed by Purkinje cells and Bergmann glia paracrine activity. Interestingly, Purkinje cells effect is only transient and ceases between P5 and P8. This conclusion is based on the observation that OPC density at P15 is increased in TRα^AMI/P^T2/P5 mice, which express TRα1^L400R from P5, but not in TRα^AMI/L and TRα^AMI/C, which express the mutation from P8 in Purkinje cells. In vitro experiments also indicate that activation of TRα1 by T₃ is required for their early morphological maturation (48, 49). Recent evidence suggests that TRα1 expression is high in Purkinje cells and then drops (23), whereas TRβ1 expression increases and might become necessary for the maintenance of the function of these cells (50). Impairment of Purkinje cells maturation by early TRα1^L400R expression consequently impacts their environment, increasing Ntf3 expression in granular neurons and delaying OPC differentiation. When TRα1^L400R expression is ubiquitous (TRα^AMI/N), a delay in granular neurons differentiation exerts an additional indirect influence on Ntf3 mRNA level, and probably on NT3 secretion. Although the heterogeneous astrocyte lineage is poorly characterized, it is also sensitive to the direct action of liganded TRα1 (34, 51) and can synthesize a number of factors required for proper cerebellum development and myelin formation (9). The response of astrocytes to T₃ appears to be also required for timely OPC differentiation. T₃ at this early stage appears therefore to be able to synchronize a complex network of cellular interactions involving several glial and neuronal cell types (52, 53) but does not act directly on OPC differentiation.

Our second finding is that T₃/TRα1 signaling is also required to prevent the expansion of a population of slowly proliferative OPC in adults. This second function becomes obvious only at a late stage of development, when most OPC differentiation is normally terminated, and requires the cell-autonomous function of TRα1. This persistence of cycling OPC in adult mice has already been reported in the optic nerve for THRA/THRB double knockout, which are devoid of all T₃ receptors but not for THRA single knockout (22). Similarity between the phenotypes of TRα^AMI/C and THRA/THRB knockout, but not single THRA knockout, can be explained in several ways. First, most manifestations of T₃ deficiency during neurodevelopment are due to permanent negative regulation exerted by unliganded TRα1, rather than loss of transactivation. Knock-in mutations are thus more detrimental than knock-out in which negative regulation is lost (54). Second THRA individual knockout might artificially broaden the repertoire of TRβ1 target genes, as demonstrated for other nuclear receptors (55). Finally TRα1^L400R could act as a TRβ antagonist, although this is not usually the case, in cell types in which both receptors are present (30). We thus propose a hypothetical model in which liganded TRα1 exerts two distinct functions: an early indirect effect involving several neurotrophins and a cell-autonomous regulation in adult OPC (Fig. 5). At this point the early influence is very unclear, but certainly involves several glial and neuronal cell populations, and neurotrophin factors, including NT3, the individual contribution of which was not precisely defined.

Fig. 5.

Hypothetical model proposed for the influence of T₃ signaling on OPC differentiation. During the first postnatal week, T₃ exerts its cell-autonomous function only in Purkinje cells and astrocytes, modifying neurotrophic factor secretion and cell contacts, and indirectly favoring the commitment of proliferating OPC toward terminal differentiation. Although ubiquitous expression of TRα1^L400R affects BDNF, Shh, and PDGF secretion, only secretion of NT3 by granular neurons is presented. By contrast, adult OPC are sensitive to the cell-autonomous signaling by T₃/TRα1, which is required to prevent uncontrolled proliferation (second mode of action).

It is likely that the early and late OPC phenotypes involve distinct cell populations. The origin of murine cerebellar OPC is unknown (56) but recent quail/chick chimera experiments indicate that in birds they migrate from the midbrain alar plate (57). Whether adult OPC derive from the maturation of perinatal OPC (58) or from the expansion of a distinct small population of cells already present at birth (59) remains unclear. The early OPC population cycles rapidly and ensures most of initial myelination, after few symmetric divisions (60). We found that it is mostly sensitive to the indirect effect of T₃. If early OPC and late OPC are truly distinct cell types, they would not necessarily respond in the same manner to T₃ stimulation.

Interestingly, an early in vitro study concluded that purified newborn OPC are poorly sensitive to T₃-mediated differentiation (61). This prompts us to believe that most of these previous studies, performed in vitro, which very convincingly demonstrated that T₃ is a key factor triggering proliferation arrest and terminal differentiation, are relevant to adult OPC. This second population of OPC has a very different behavior. It is composed of slow cycling cells, which mostly perform asymmetric divisions (58, 60) and have a poor capacity to make myelin (59). This slow cycling can be maintained for months in culture medium without T₃ (12, 62). Adult OPC recently received considerable attention not only because ischemia and demyelination reactivate their proliferation (63) but also because they make synapses with the climbing fibers that excite Purkinje cells (64). They are also believed to constitute a distinct population of dispersed neural stem cells (11, 40, 42, 65), although their ability to generate neurons has been recently challenged (66). The fact that liganded TRα1 exerts a direct and permanent control on adult OPC proliferation suggests a number of interesting possibilities for later investigations.

Materials and Methods

Animals

All animal experimentations were conducted under animal care procedures in accordance with the guidelines set by the European Community Council Directives (86/609/EEC). All mice were of C57/Bl6 genetic background, with a 129Sv contribution. ROSA26-lox-STOP-lox-EYFP (R26YFP) (67) was introduced in TRα^AMI/Cre mice to define recombination patterns (see Table 1c). YFP immunostaining was systematically used after tamoxifen treatment to rule out any variation between treated animals. Mice were genotyped by PCR for Cre, R26YFP, and TRα^AMI alleles using specific primers. Tamoxifen (1 mg/40 g of body weight in 100 μl of sunflower oil), 100 mg BrdU/kg of mice in 100 μl 0.9% NaCl injectable solution, or 25 mg EdU/kg of mice were injected ip. For immunohistochemistry, animals were first anesthetized by ip injection of a lethal dose of ketamine/xylazine and perfused with paraformaldehyde (4% in PBS). Cerebellums were postfixed for 2 h and washed, and sections (50 μm thick) were performed with a vibratome (Campden Instruments, Lafayette, IN).

Immunohistochemistry, cell labeling, and Western blotting

Sections were washed in PBS (pH 7.4) and then blocked in 10% normal donkey serum, 2% gelatin from cold water Fish Skin (Sigma, St. Louis, MO), and 0.02% Triton X-100 in PBS. Sections were incubated overnight at 4 C with primary antibodies (Supplemental Table 2) diluted in 1% dimethylsulfoxide, 1% donkey serum, 1% goat fetal serum in PBS, washed at least three times in PBS, and incubated for 2 h at room temperature with 4′,6-diamidino-2-phenylindole (DAPI) (1:1000; Invitrogen, Carlsbad, CA) and secondary antibodies coupled to a fluorescent dye diluted in 1% donkey serum, 1% dimethylsulfoxide in PBS. For BrdU labeling, HCl treatment (2 m, 30 min, 37 C) was performed before the blocking step. EdU staining was done as previously described (68). Microscopic observations were done using Axioplan2 (Carl Zeiss, Thornwood, NY) or SP5 confocal microscope (Leica Microsystems, Deerfield, IL). Cell counting was performed using Metamorph and ImageJ softwares. Two series of two optical sections (1 μm along the z-axis) were acquired in each slice. Section series were projected using the ImageJ z-projection function set for maximum intensity. Counting was performed manually, and the region of interest was determined using DAPI staining.

For Western blotting, protein extracts (10 μg) were resolved on 10% SDS-polyacrylamide gels and transferred to nitrocellulose membranes before incubation with antibody and then IRDye secondary antibody. Immunoreactive bands were detected and quantified using Odyssey infrared imaging system (LI-COR Biosciences, Lincoln, NE). All antibodies are listed in Supplemental Table 1.

RNA analysis

RNA was extracted from whole cerebellum using a mini-extraction kit (QIAGEN, Chatsworth, CA). cDNA were prepared from 1 μg of RNA using M-MLV reverse transcriptase (Promega Corp., Madison, WI) and random 6-mer primers in 20 μl. cDNA (1 ng) was used for quantitative PCR, using sybrgreen (QIAGEN). Quantification was performed in triplicate using HPRT as internal standard and the 2^–ΔΔCt method for analysis.

Statistical analysis

All statistical analyses were done using R software. ANOVA analysis was used to determine significant difference between Olig2+ cell number or mRNA abundance in different genotypes. Data are presented ± sd. The number of mice in each experiment was at least four per genotype, originating from two different litters, providing both mutant and control mice. All statistical analyses were performed between animals from the same litter.