Introduction

Waves of neuronal apoptosis are crucial for shaping the complex architecture of the developing brain, whereas progressive, regional loss of neurons underlies the irreversible pathogenesis of various neurodegenerative diseases in adult brain. During development, the decision between neuronal survival or death is determined by access to neurotrophic support and activation of excitatory neurotransmitter receptors. Both events activate signal transduction pathways that eventually protect neurons from apoptotic cell death. In the case of chronic neurological diseases, various triggering mechanisms, including metabolic impairment, excitotoxicity, free radical production or biochemical abnormalities resulting from genetic mutations, have been identified as causative agents in neuronal death1. However, the molecular pathways involved in the transduction of these signals into cell death are largely unknown. The mechanisms responsible for neuronal apoptosis in neurodegenerative diseases may at least partially overlap with those that are involved in development.

Anti-apoptotic signals exert their effect on cell survival through various mechanisms, including modulation of gene expression. Neurotrophin- and activity-dependent gene expression is mediated by several neuronal signal transduction cascades, such as the PI3K/Akt pathway, the MAPK pathway, the Ca2+/CaMK and cAMP/PKA pathways, that converge on the CREB family of leucine-zipper transcription factors2. This transcription factor family is comprised of CREB, the related proteins cAMP response element modulatory protein (CREM) and the activating transcription factor 1 (ATF1). All three are activated by phosphorylation and bind as homo- or heterodimers to the cAMP response element (CRE) in the promoters of their target genes3. CREB has long been implicated in neuronal function, with much recent interest centered around its role in the maintenance of long-term memory4. More recently, several studies involving overexpression of dominant-negative CREB suggested a role for CREB as a survival factor in various cellular models5,6,7,8,9, possibly acting downstream of the Akt/PKB survival pathway10.

We have used genetic disruption to study the function of the CREB family members. We previously generated two different Creb1 mutations in the mouse. The first, Creb1αΔ, led to the generation of a hypomorphic mutation11,12, whereas the second, Creb1−/−, is a true null mutation that results in perinatal death of homozygous mice13. Determining the significance of complete CREB loss in adult mice thus required the generation of conditional Creb1 mutant mice. One additional complicating factor is that in both Creb1αΔ and Creb1−/− mutants, Crem expression is upregulated in many organs, including brain12,13. The functional importance of CREM upregulation has until now remained unknown. Here we report the generation of mice with a conditional mutation of Creb1 (Creb1loxP). We use these mice to show the result of disrupting both Creb1 and Crem in brain either during development or postnatally. These studies show that CREB and CREM regulate similar neuronal survival processes, as only neurons devoid of both CREB and CREM undergo apoptosis. Moreover, our genetic dissection reveals an unexpected difference between developmental and postnatal roles of CREB and CREM. Specifically, when both Creb1 and Crem are inactivated in neuronal and glial precursors during development, generalized cell death occurs in the nervous system. By contrast, postnatal disruption of both CREB and CREM function leads to progressive neurodegeneration that is restricted to the dorsolateral striatum and to the CA1 and dentate gyrus regions of the hippocampus.

Results

Generation of mice lacking both Creb and Crem in brain

Mice either hypomorphic for CREB11,12 or completely lacking CREB13 or CREM14,15 show normal embryonic brain development. In Creb1αΔ and Creb1−/− mutant mice, reduction or loss of CREB is accompanied by upregulation of CREM12,13. As all three members of the CREB family can potentially compensate for one another, it is notable that only CREB and CREM, but not ATF1, are detectable in wildtype mouse neurons (data not shown).

As an initial step toward obtaining mice devoid of CREB in the nervous system, we generated Creb1loxP/loxP mice by homologous recombination in embryonic stem (ES) cells (Fig. 1a). Creb1loxP/loxP mice showed normal expression of CREB in all tissues, including brain (Fig. 1c). These mice were crossed to mice harboring a nestin-driven Cre recombinase transgene16, to generate Nescre Creb1loxP/loxP (Creb1Nescre) mice. The nestin promotor induces Cre expression early, before separation of neuronal and glial lineages (Fig. 1b). CREB immunoreactivity in Creb1Nescre mice was lost in most of the developing brain by embryonic day (E) 12.5 (data not shown). By E18.5, Creb1Nescre mice showed extensive loss of CREB in brain (Fig. 1d). The high CREM upregulation observed in these mice (data not shown) suggested a compensatory role for CREM in the absence of CREB. To circumvent this possible compensation, Creb1Nescre mice were further crossed to Crem−/− mice15, yielding Creb1NescreCrem−/− double-mutant mice, which lacked both Creb and Crem in brain.

Figure 1: Generation of mice deficient for CREB in the nervous system.
figure 1

a, Organization of Creb1 encompassing exons 9–10. Exon 10 was flanked with loxP sites in two steps: First, we generated the modified allele by homologous recombination in ES cells. Second, transient expression of Cre recombinase resulted in removal of the selection cassette (neomycin resistance and thymidine kinase), generating Creb1loxP and Creb null alleles. A scheme depicting the wildtype gene locus, targeting vector and resulting alleles are shown. Triangles represent loxP, black rectangles represent exons; white rectangles represent probes used for Southern-blot analysis and arrows represent primers used for PCR genotyping32. H, HindIII; K, KpnI; X, XbaI. bq, Expression of Cre recombinase in the cre transgenic lines and pattern of recombination in mutants. Nescre transgenics express Cre throughout the developing brain by E12 (data not shown). Cre expression becomes restricted to the proliferating ventricular zones by E16.5, as revealed by Cre immunohistochemistry (b). At E18, CREB is expressed throughout the brain in Creb1loxP/loxP mice (c) but lost in Creb1Nescre brains (d). Camkcre4 transgenics show high Cre expression in all cortical layers (e), in the entire hippocampus (h), in amygdala (k) and striatum (l). Cre-mediated recombination is detected by comparison of Creb protein levels in Creb1loxP/loxP control animals at 12 wk (f,i,m) to those of Creb1Camkcre4 mutant littermates (g,j,n). Recombination affects all cortical layers equally (compare f with g) and is extensive in hippocampus (compare i with j) and striatum (compare m with n). D1cre transgenics show Cre expression restricted to the striatum (o) correlating with CREB loss in Creb1D1cre mutants (q) compared with Creb1loxP/loxP controls (p). Original magnification: ×10 (bd,o), ×30 (el,p,q) and ×400 (m,n,inserts).

At four weeks, all genotypes except Creb1NescreCrem−/− mutants were represented, suggesting embryonal or perinatal death of mice lacking both CREB and CREM in brain. By analyzing litters at E18.5 and newborn (P0) mice, we found that Creb1NescreCrem−/−mice were present at roughly the expected mendelian ratio. Upon closer examination, we saw that newborn Creb1NescreCrem−/− mutants invariably did not suckle, as there was no evidence of milk in their stomachs, and, as a result, they died within one day after birth.

Brain development in Creb1NescreCrem−/− mice

All brain structures were present in Creb1NescreCrem−/− brains at E16.5, E18.5 and P0, indicating that CREB and CREM do not have major roles in brain formation. Histological analysis showed a reduced cell density throughout the brain at E18.5 and P0, accompanied by an increase in cells with pyknotic nuclei and enlarged ventricles (Fig. 2). Notably, these dramatic changes were more extensive in P0 brains, indicating an accumulation or acceleration of cell death over this time period. In the highly affected cortical areas of mutant P0 brains, the cortical plate was completely disorganized and only a few pyramidal cells could be found within loosely arranged clusters of small round cells (Fig. 2d). Instead of a dense layer of large pyramidal neurons, only sparsely distributed, small round cells were present in the pyramidal cell layer of areas CA1 and CA3 of the hippocampus (Fig. 2f). In other areas, such as the thalamus or hypothalamus, the cytoarchitecture was better preserved and specific nuclei could be identified. The number of differentiated neurons in these nuclei was reduced, however, and many neurons were smaller than those of control tissue (Fig. 2g,h). In the olfactory bulb of Creb1NescreCrem−/− brains, the mitral cell layer showed the greatest cell loss (Fig. 2j). Similar changes were seen in rhinal and limbic cortical areas (data not shown).

Figure 2: Cellular integrity is compromised in brains of mice lacking both Creb and Crem.
figure 2

a,b, Histological analysis of Creb1NescreCrem+/− control (a) and Creb1NescreCrem−/− (b) brain sections at P0. ch, Comparing Nissl stainings reveals extensive loss of large neurons and an increased number of cells with pyknotic nuclei, especially in the cortical plate (c,d) and in the hippocampal area CA1 (e,f) of mutant mice. The thalamus or hypothalamus were also affected (g,h), but to a lesser degree. i,j, In the olfactory bulb, the mitral cell layer was markedly reduced. v, ventricular space. Original magnification: ×10 (a,b), ×200 (ch), ×400 (i,j). Scale bar, 500 μm (a,b); 20 μm (ch); 10 μm (i,j).

Severe neuronal loss during brain development due to apoptosis

To examine whether the reduced number of neurons was a result of increased cellular degeneration, reduced proliferation or both, we labeled cells with Ki-67 and TUNEL and stained for activated caspase 3 in mutant and control brains (Fig. 3). Cell proliferation, as assessed by Ki-67 labeling, revealed no obvious differences between Creb1NescreCrem−/− double-mutant (Fig. 3c) and Creb1loxP/loxPCrem+/− or Creb1NescreCrem+/− control brains (Fig. 3a,b) of mice at E18.5, indicating that proliferation of neurons in the ventricular zones is largely unaffected by the absence of CREB and CREM. In contrast, apoptosis occurred, to varying extents, in all areas of the brain (data not shown). Cells positive for TUNEL and activated caspase 3 were abundant and typically found in the cortical plate, within layers V and VI of Creb1NescreCrem−/− brains (Fig. 3f,i). Apoptotic cells were rare and present only in the subventricular zone of cerebral cortex of control mice (Fig. 3d,e,g,h). Apart from the Creb1NescreCrem−/− brain, other genotypes, including Creb1Nescre or Crem−/− (data not shown) and Creb1NescreCrem+/− (Fig. 3e,h), showed control levels of apoptosis, further supporting the link between neuronal death and the loss of both CREB and CREM.

Figure 3: Normal proliferation, but increased neuronal death, in mice lacking both CREB and CREM in brain.
figure 3

ac, The proliferative activity of the ventricular zones of Creb1loxP/loxPCrem+/− and Creb1NescreCrem+/− control mice (a and b, respectively) was similar to that of Creb1NescreCrem−/− animals (c). Scattered cell death was seen throughout the control brains, but was greatly increased in Creb1NescreCrem−/− brains. di, Compared with control mice (d,e,g,h), cells in layers VI and V of the mutants show intense TUNEL labeling (f) and activated caspase 3 staining (i). Original magnification:×200.

In brain of Creb1NescreCrem−/− mice at E16.5, TUNEL-positive cells were already apparent, but restricted to earlier developing areas, such as the amygdala and small regions of the thalamus (Fig. 4a,b). Areas that develop later, such as the cerebral cortex, the hippocampus and caudate putamen, showed little or no labeling for apoptotic cells at E16.5; however, the number of cells undergoing apoptosis was significantly increased by E18.5 and P0 (Fig. 4a,b). We also observed TUNEL labeling in the trigeminal ganglion and dorsal-root ganglia (data not shown), indicating that the observed increase in apoptotic death was not confined to the central nervous system.

Figure 4: Neuronal degeneration is initiated between E16.
figure 4

5 and E18.5 in the absence of CREB and CREM. a,b, Regional and developmental onset of cell death in NescreCreb1loxP/+Crem+/− controls (a) and Creb1NescreCrem−/− double-mutants (b). Detection of apoptotic cells by TUNEL labeling of 7-μm paraffin sections during different developmental stages is shown. Cells of the cortical plate and the caudate putamen show little staining at E16.5, but the number of positively stained cells is markedly increased in double-mutant animals at E18.5 and at P0. In the reticular thalamic nuclei and the amygdala, which develop early, apoptotic cells are already apparent in double-mutant embryos at E16.5. This results in a high cell loss at later stages. Magnification: ×400.

The mechanism by which loss of CREB leads to neuronal death is unknown, but may involve pro-survival factors such as Bcl2, which is a putative CREB target gene6,17,18. We therefore carried out an analysis of factors involved in regulation of cellular survival by RNase protection assay. We found no difference in the mRNA levels of both pro- and anti-apoptotic factors Bcl2, Bcl2l, Bcl2l2, Bak1, Bax and Bad (see Web Fig. A online). Loss of expression of anti-apoptotic Bcl2 family members is thus unlikely to be responsible for the neuronal apoptosis seen in our model.

Progressive post-natal neurodegeneration

As Creb1NescreCrem−/− mice lose CREB during early neuronal development and die perinatally, we were only able to study these mice until birth, when neurons are still differentiating. To determine the consequences of CREB loss in postnatal neurons, we crossed the Creb1loxP/loxP mice with transgenic mice (Camkcre4) expressing Cre recombinase postnatally, under the control of the 8.5-kb promoter fragment of the calcium/calmodulin-dependent protein kinase II-α gene (Camk2a). We found high levels of Cre expression in striatum, nucleus accumbens (part of the basal ganglia), thalamus, amygdala, cortex and the hippocampus of the Camkcre4 transgenic mice (Fig. 1e,h,k,l). In Camkcre4 Creb1loxP/loxP mutants (Creb1Camkcre4), Cre-mediated recombination was extensive in all areas of the brain where the recombinase is expressed (Fig. 1, compare panels f,i,m with panels g,j,n). CREB was not detectable in pyramidal cells of hippocampal CA1, CA2, and CA3 (Fig. 1j) or in granule cells of the dentate gyrus, except for a thin strip of cells on the internal face. Extensive recombination was also observed in striatum (Fig. 1n), amygdala (data not shown) and in all cortical layers (Fig. 1g) as indicated by the absence of detectable CREB. The functional compensation by Crem as seen in Creb1Nescre mutants prompted us to generate Creb1Camkcre4Crem−/− double-mutant mice. Offspring were viable, and genotyping revealed the expected mendelian ratios. Notably, we saw a neurological phenotype in older (more than six months) Creb1Camkcre4Crem−/− double-mutants. When suspended by their tail for more than 20 seconds, double-mutant animals retracted their hind limbs toward their trunks in a dystonic fashion, rather than extending them as did control littermates (Fig. 5a,b). This feet-clasping phenotype is observed in several mouse mutants with neurological impairment due to neurodegeneration19,20,21,22,23. At the morphological level, the brain of the Creb1Camkcre4Crem−/− double-mutants showed considerable atrophy of the striatum and the hippocampus (Fig. 5c–f).

Figure 5: Neurological and morphological phenotype of Creb1Camkcre4Crem−/− double-mutants.
figure 5

a,b, Creb1Camkcre4Crem+/− control animals at 12 mo have normal limb reflexes when suspended by the tail (a), whereas Creb1Camkcre4Crem−/− double-mutants show an abnormal feet-clasping behavior (b). cf, Dark-field, low-power magnification (×10) of brain sections from Creb1Camkcre4Crem+/− controls (c,e) and Creb1Camkcre4Crem−/− double-mutants (d,f) at 12 mo. Note the shrinkage of the striatum (compare c with d) and the marked atrophy of the hippocampus in the double-mutants (compare e with f).

To determine whether the atrophy of striatum and hippocampus was caused by a wave of cell death at a particular stage of development or by progressive cell loss, we analyzed the time-course of neuronal loss in brain of Creb1Camkcre4Crem−/− double-mutants. At 1.5 months, neuronal loss, as visualized by neuN or Cre immunostaining, is modest in hippocampus and almost undetectable in striatum (Fig. 6, compare panels a,f,k with panels b,g,l). The lesion becomes progressively larger with time (2, 3 and 6 months) in both structures, eventually leading to extensive atrophy of the dorsolateral striatum, a marked reduction in thickness of dentate gyrus and a complete elimination of CA1 (Fig. 6e,j,o). Thus, the hippocampal and striatal lesions are the result of the accumulation of scattered neuronal loss during a progressive neurodegenerative process. Striatal degeneration was restricted to the dorsolateral region and did not affect the nucleus accumbens. The CA1 region of the hippocampus showed the most severe cell loss, whereas degeneration in the dentate gyrus was milder and the CA3 region was unaffected. The specificity of the lesion was revealed by GFAP immunostaining, which indicates the areas where astrogliosis has been induced by neuronal injury. We detected massive astrogliosis in the dorsal striatum, CA1 and dentate gyrus of Creb1Camkcre4Crem−/− double-mutants (Fig. 7b,d–f). Fainter gliosis was observed in amygdala, cortex and thalamus (data not shown). By contrast, control animals carrying a single functional copy of Crem showed no signs of gliosis, demonstrating the ability of CREM to compensate for the loss of Creb (Fig. 7a,c).

Figure 6: The striatum and the hippocampus from Creb1Camkcre4Crem−/− double-mutants undergo progressive neurodegeneration.
figure 6

ao, Cre (ae) and neuN (fo) immunostaining of brain sections from Creb1Camkcre4Crem+/− controls at 1.5 mo (a,f,k) and Creb1Camkcre4Crem−/− double-mutants at 1.5 mo (b,g,l), 2 mo (c,h,m), 3 mo (d,i,n) and 6 mo (e,j,o). Striatal degeneration is progressive (be) and is confined to the dorsolateral area. The nucleus accumbens is still intact at the age of 24 wk (e). In hippocampus, neuronal loss accumulates in CA1 and dentate gyrus while CA3 remains intact (gj). Panels ko show high-power magnification of the CA1 region. Magnification: ×25 (aj), ×400 (ko).

Figure 7: Astrogliosis in striatum and hippocampus.
figure 7

af, GFAP immunostaining reveals no signs of gliosis in Creb1Camkcre4Crem+/− control mice, which retain one functional copy of Crem (a,c). By contrast, massive astrogliosis is visible in striatum (b) and hippocampus (d) of Creb1Camkcre4Crem−/− double-mutants at 12 wk. In the hippocampus of double mutants, gliosis is restricted to CA1 and dentate gyrus (dg, e). High-power magnification of GFAP staining in dorsolateral striatum (f). Magnification: ×10 (ad), ×30 (e), ×400 (f).

As predicted by the massive wave of apoptosis observed in Creb1NescreCrem−/− mutants, staining of Creb1Camkcre4Crem−/− brain sections with the nuclear Hoechst 33342 dye revealed many apoptotic cells with highly condensed and fragmented nuclei in the regions of neurodegeneration (see Web Fig. B online). Similarly, staining for the activated form of caspase 3 revealed scattered positive cells in CA1, dentate gyrus and dorsal striatum, whereas no positive cells could be detected in control animals (see Web Fig. B online). Similar to what we observed in the Creb1NescreCrem−/− mutants, we could not detect significant differences in Bcl2 immunostaining between control and Creb1Camkcre4Crem−/− double-mutant mice (see Web Fig. A online).

Genetic dissection of striatal degeneration

We next sought to verify the pattern of striatal degeneration in an independent mouse line in which disruption of Creb1 was more restricted to the striatum. We therefore used the D1cre transgenic line, in which Cre expression is directed by the dopamine receptor D1A gene (Drd1a) promoter. To achieve faithful expression in this transgenic model, we used a YAC of 140 kb containing the entire Drd1a as an expression vector for Cre recombinase. Expression of the transgene recapitulated the expression pattern of the endogenous Drd1a. Specifically, we found high expression levels of Cre in striatum, nucleus accumbens and olfactory tubercles (Fig. 1o). We also found weaker expression in layer VI of the cortex, CA2 in hippocampus and some nuclei of the thalamus (data not shown). Recombination, as shown by CREB immunocytochemistry, overlapped with regions of Cre expression in Creb1D1cre mutants (Fig. 1p,q).

The phenotype of Creb1D1CreCrem−/− double-mutants was most striking in 7-month-old mice. As seen with the Creb1Camkcre4Crem−/− double-mutants, degeneration specifically affected the dorsolateral striatum of Creb1D1creCrem−/− double mutants (Fig. 8a,b). Notably, not only was the spatial specificity of the striatal degeneration retained in the Creb1D1creCrem−/− strain, but the kinetics were also virtually identical to that observed in Creb1Camkcre4Crem−/− double-mutants (Fig. 8c–f). These observations suggest that the pattern of degeneration in striatum is dependent on the intrinsic sensitivity of striatal neurons.

Figure 8: Striatal neurodegeneration in Creb1D1creCrem−/−mice.
figure 8

a,b, At 7 mo, immunostaining for Cre in Creb1D1creCrem+/− control animals (a) and Creb1D1creCrem−/− double-mutants (b) reveals a clear lesion in the dorsolateral striatum. cf, Cre immunostaining showing the evolution of the lesion in Creb1D1creCrem−/− mice at 1.5 mo (c), 2 mo (d), 3 mo (e) and 7 mo (f). Magnification: ×10 (a,b), ×25 (cf).

Discussion

This study demonstrates that CREB family members are crucial in neuronal survival in vivo. Absence of CREB and CREM in developing brain results in generalized cell death, whereas postnatal disruption of transcription mediated by CREB or CREM triggers selective and progressive neurodegeneration. Loss of only CREB in brain, either during development or postnatally, had no impact on neuronal survival or survival of the mice. The same holds true for brains from Crem−/− mice, which show no defect in neuronal survival. As mice devoid of both CREB and CREM in the brain during development show a severe loss of neurons (Fig. 2), we conclude that CREM upregulation in Creb1 mutant mice is sufficient to maintain cellular survival. Decreased cell density in Creb1NescreCrem−/− brain resulted from the increased incidence of neuronal apoptosis and not from decreased proliferative activity (Figs 3 and 4). The first evidence of apoptosis occurred at E16.5, in the earlier-developing areas, such as the amygdala and reticular thalamic nuclei (Fig. 4). Later-developing areas, such as the cerebral cortex, hippocampus and caudate putamen, showed little evidence of apoptosis at E16.5; however, the number of cells undergoing apoptosis was significantly increased by E18.5 and P0 (Fig. 4). Increased apoptotic nuclei were also seen in the trigeminal ganglion and dorsal-root ganglia (data not shown), indicating that the observed increase in apoptotic death was not confined to the central nervous system. As significant CREB loss occurs as early as E12.5, it is unlikely that CREB influences neuronal survival at periods between E12.5 and E16.5. However, we can not rule out the possibility that small amounts of residual CREB may be sufficient to support survival during this time period.

The selective loss of neurons in the Creb1Camkcre4Crem−/− and Creb1D1creCrem−/− mice also supports a role for CREB and CREM in adult neuronal survival. However, the specificity of neuronal loss observed in the postnatal forebrain contrasts with the earlier generalized effects observed in Creb1NescreCrem−/− mutants. Thus, developmental and postnatal functions of CREB and CREM are only partially overlapping. We hypothesize that loss of CREB in differentiated neurons reduces the probability of neuronal survival rather than being an absolute requirement for survival, as observed in the developing brain. In both postnatal models presented here, the sensitive areas degenerate progressively because of scattered apoptotic cell death, mimicking the typical 'one-hit' dynamics of cell loss described for several neurodegenerative disorders24. It is possible that differential survival in the postnatal brain is modulated by both region-specific expression of protective factors and the network properties of the vulnerable structures. The dynamics of degeneration is then likely to be determined by intrinsic properties of the affected areas, such as CA1, dentate gyrus and striatum, which may explain why the time-courses of striatal degeneration are so similar in the two independent postnatal models presented here. If the effect of CREB loss is to lower the threshold of survival, it is possible that death may be triggered by a particular pattern of neuronal electrical activity. In conjunction with the role of CREB in synaptic plasticity4, these observations open up the possibility that CREB may control processes commonly involved in both apoptosis and synaptic plasticity.

Hippocampus and striatum are particularly vulnerable to insult in several contexts, including neurometabolic disorders, seizure, ischemia episodes and neurodegenerative diseases such as Huntington disease. Several recent studies have shown that mutated forms of huntingtin, ataxin-1, atrophin-1 and the androgen receptor proteins, all of which carry an expanded polyglutamine stretch and cause neurodegenerative diseases, interact and sequester CREB-binding protein (CBP) and TAF2C1 transcriptional co-factors25,26,27,28. These interactions are thought to deplete these co-factors from their normal locations and to perturb transcription that is mediated by CBP or TAF2C1. Consistent with this, transcriptional activities of CREB and Elk1, two transcription factors that interact with CBP or TAF2C1, are reduced by mutant huntingtin25,26,29. CBP and TAF2C1, however, interact with many transcription factors; thus, it was unclear which of the transcriptional pathways acting downstream of CBP or TAF2C1 was involved in triggering neurodegeneration. Our study clearly shows that disruption of the CREB signaling pathway alone is sufficient to produce a neurodegenerative phenotype in the mouse. In light of the reported inhibitory effect of the expanded polyglutamine stretch of huntingtin on CBP function, it is notable that the conserved pattern of degeneration involves the dorsolateral striatum in both Creb1Camkcre4Crem−/− and Creb1D1creCrem−/− models. Striatum is the main region of brain to undergo degeneration in Huntington disease, both in humans and in several rodent models30. It is therefore probable that interference with CREB-mediated transcription contributes to neuronal loss in a subset of polyglutamine diseases such as Huntington disease.

Mice lacking CREB and ATF1 show early developmental arrest and increased cell death at a time when CREM is not yet present31. Increased apoptosis of spermatids, which normally express only CREM, is also responsible for the infertility of Crem−/− male mice14. Together, these observations provide genetic evidence that all three CREB family members are crucial for the maintenance of cell viability in a variety of tissues and at various stages of development. Moreover, this study shows that CREB and CREM are crucial for the maintenance of neuronal survival in vivo, raising the possibility that pharmacological manipulation of the CREB signaling pathway may exert positive therapeutic action on both the functional and anatomical deterioration that follows the progression of neurodegenerative disorders.

Methods

Generation of Creb1loxP mice.

To generate the nervous system-specific CREB mutant mice, we used homologous recombination in ES cells to modify the Creb1 allele such that Creb1 exon 10, encoding the first part of the bZIP domain, was flanked by loxP sites. Cre-mediated recombination of the Creb1loxP allele leads to a Creb1 null allele that encodes a truncated CREB protein devoid of DNA-binding and dimerization domains. This truncated protein is unstable; thus, successful recombination results in loss of CREB. Mice harboring the Creb1loxP allele were crossed with transgenic mice carrying a transgene for Cre recombinase under the control of the nestin promoter and enhancer16. To generate Creb1NescreCrem−/− mice, we crossed the Creb1loxP mice with Crem−/− mice15. The only two possible kinds of breeding pairs that could yield useful numbers of double-mutant Creb1NescreCrem−/− (NescreCreb1loxP/loxPCrem−/−) mice were NescreCreb1+/loxPCrem+/− males crossed with Creb1loxP/loxPCrem−/− females, and Creb1loxP/loxPCrem+/− males crossed with NescreCreb1+/loxPCrem−/− females. Male Crem−/− mice are sterile, and both male and female Creb1Nescre mutant mice are subfertile, to the point that it was impractical to use this genotype for breeding. In each case, the expected number of Creb1NescreCrem−/− mice was 1 in 8. Similarly, to generate mice with postnatal Creb loss (see Camkcre4 and D1cre transgenic lines below), we used the following mating schemes to obtain double-mutants: (i) Camkcre4 Creb1loxP/loxPCrem+/− males × Creb1loxP/loxPCrem−/− females (1 in 4 probability of obtaining a double mutant) and (ii) D1creCreb1loxP/+Crem+/− males × Creb1loxP/loxPCrem−/− females (1 in 8 probability of obtaining a double-mutant). Mice were bred in a genetic background comprising a mix of 129SvOla, C57/BL6 and FVB/N. We genotyped mice with the Creb1loxP allele by PCR (Fig. 1a) as previously described32 and detected cre transgenes by dot-blot analysis of tail DNA. The Crem null allele was detected by PCR using primers designed within the Crem allele and lacZ cassette.

cre transgenic lines.

To generate the Camkcre4 transgenic line, we cloned the open reading frame (ORF) encoding Cre recombinase, fused to a nuclear localization signal (nls-cre), into the Camk2a-pMM403 vector33 as previously described34. We injected purified, linearized DNA into pronuclei of FVB/N oocytes. We chose to use the Camkcre4 line because of its robust and widespread pattern of Cre expression in the forebrain. To generate the D1cre line, we isolated a YAC of 140 kb from a C57Bl/6 mouse genomic library35. We introduced the nls-cre ORF into the dopamine receptor-D1A gene (Drd1a) by homologous recombination in yeast. We assembled a pop-in/pop-out modification vector in the pRS306 (ref. 36) vector using a region of 800 bp showing 5′ homology to Drd1a and a region of 580 bp showing 3′ homology to Drd1a. The second region contains a unique XbaI site for linearization of the construct. The modified D1cre YAC DNA was isolated as described37 and injected into pronuclei of FVB/N oocytes. The pattern of Cre expression was identical in four of the five lines obtained. The D1cre line used in this study harbors four copies of the transgene and was selected because of its robust expression levels. cre transgenic mice were genotyped by dot-blot analysis of tail DNA.

Immunohistochemistry.

We perfused mice with cold 4% paraformaldehyde (PFA). Depending on the age of the mice, we either dissected brains (adult) or post-fixed whole heads (E16.5 and E18.5) overnight in 4% PFA at 4 °C; in either case, we embedded the samples in paraffin wax. We sectioned the paraffin blocks on a microtome at a thickness of 7 μm, cleared the sections of paraffin and rehydrated them through an ethanol dilution series. We cut postnatal brains at a thickness of 50 μm on a vibratome (Leica) and further processed floating sections for immunohistochemical detection using the VECTASTAIN ABC system (Vector Laboratories) and diaminobenzidine (DAB; Sigma) incubation. We used the following primary antibodies: polyclonal anti-Creb (N-terminal epitope: H2N-SGADNQQSGDAAVTEC-CONH2,1:9,000), polyclonal anti-Cre (Covance Research Products, 1:3,000), polyclonal anti–Ki-67 (Dianova, 1:100), monoclonal anti-neuN (Chemicon, 1:500), monoclonal anti-GFAP (Sigma, 1:500), polyclonal anti-cleaved caspase 3 (Cell Signaling, 1:100) and polyclonal anti–mouse Bcl2 (PharMingen, 1:400).

Nissl staining.

We rehydrated paraffin sections, rinsed them in water, incubated them for 10 min in 0.1% cresyl violet and rinsed them again in water. We destained the slides in 250 ml 96% ethanol containing 5 or 6 drops of acetic acid, then dehydrated and mounted them.

TUNEL labeling.

We cleared paraffin sections of embryos and rehydrated them through an ethanol series. After a water rinse, we incubated the sections for 5 min in 20 μg ml−1 proteinase K and for 5 min in 2% H2O2 at room temperature. We carried out staining according to the manufacturer's instructions using the In Situ Cell Detection Kit (Roche). We detected positive cells with the VECTASTAIN Elite Kit (Vector Laboratories) and with DAB substrate (Roche), as described for the immunohistochemical staining.

RNase protection assay.

We used 5 μg of total RNA for analysis. For the probe set, we used the mouse Apo-2 RiboQuant set (Pharmingen), which was used as the template to generate radiolabeled ([32P]UTP) riboprobes. We hybridized the RNA and probes for 16 h at 56 °C, then digested them with RNase A and RNase T1 at 30 °C for 45 min, precipitated them and resolved the products on a 6% acrylamide sequencing gel. We carried out image analysis using a Molecular Dynamics Phosphorimager.

Note: Supplementary information is available on the Nature Genetics website.