The LIM-Homeobox Gene Islet-1 Is Required for the Development of Restricted Forebrain Cholinergic Neurons

Forebrain cholinergic neurons modulate complex mammalian behaviors such as reward-related learning and cognitive functions. Although their dysfunction is implicated in various psychiatric and neurodegenerative diseases, the factors governing cholinergic neuron differentiation and diversity are mostly unknown. We tested the role of the LIM-homeobox gene Isl1 in the development of forebrain cholinergic neurons by conditionally deleting Isl1 using a Six3-cre transgene. A depletion of cholinergic interneurons in the dorsal and ventral striatum, and cholinergic projection neurons in the nucleus basalis is observed and is ascribed to an early and persistent defect in cholinergic neuron differentiation. Notably, cholinergic innervation to the neocortex is abolished, whereas that to the hippocampus is unaltered. The unique pattern of cholinergic hypoinnervation encountered is supported by the presence of cholinergic projection neurons in the medial septum, the magnocellular preoptic area, and the substantia innominata. Together, these results demonstrate the requirement for Isl1 in the development of restricted telencephalic cholinergic neurons and link the development of cholinergic neurons in anatomically disparate sites to Isl1 function.


Introduction
Neurotransmitter identity is a fundamental characteristic of the nervous system, yet the genetic determinants of neurotransmitter selection are incompletely understood. Cholinergic neurons, for example, are found in discrete locations along the neuraxis and possess a striking level of anatomic heterogeneity (Woolf, 1991). Some are local circuit neurons, modulating reward-related learning in the striatum (Kitabatake et al., 2003), whereas others are projection neurons, targeting cortical and subcortical structures (Johnston et al., 1979;Lehmann et al., 1980;Woolf et al., 1983). Yet little is known about the genetic mechanisms underlying cholinergic neuron diversity.
Forebrain cholinergic neurons have demanded considerable attention, because disruption in subsets of these neurons has been implicated in the cognitive dysfunction associated with Alzheimer's disease (Davies and Maloney, 1976;Whitehouse et al., 1981). Forebrain cholinergic projection neurons are found in the medial septum (MS) (Ch1), the vertical (Ch2) and horizontal limbs (Ch3) of the nucleus of the diagonal band, and the basal magnocellular complex (Ch4) (Mesulam et al., 1983). Ch4 includes cholinergic neurons in the ventral pallidum, the nucleus basalis, ansa lenticularis, substantia innominata (SI), and the magnocellular preoptic nucleus. These designations correspond approximately to the axonal trajectories of the groups; Ch1 and Ch2 provide major innervation to the hippocampus; Ch3, the olfactory bulb; and Ch4, the neocortex (Mesulam et al., 1983).
Fate-mapping studies have suggested that cholinergic neurons emerge primarily from the ventral-most regions of the developing telencephalon in the anterior entopenduncular/preoptic area (AEP/POa) (Marin et al., 2000). Analysis of loss-offunction mutants for Nkx2-1, Ascl1, Dlx1/2, and Arx reveal varying degrees of cholinergic marker loss, but a corresponding disruption of GABAergic interneurons suggests a broader role for these factors in basal telencephalic development (Sussel et al., 1999;Marin et al., 2000;Kitamura et al., 2002;Cobos et al., 2005;Colombo et al., 2007). Two genes downstream of Nkx2-1, Lhx6 and Lhx8 are required for forebrain GABAergic and cholinergic neuron development, respectively (Zhao et al., 2003;Alifragis et al., 2004;Mori et al., 2004;Liodis et al., 2007). But Lhx8independent forebrain cholinergic neurons exist, and Xenopus Lhx8 overexpression in mouse subpallial neurons is not sufficient to expand cholinergic fates (Bachy and Retaux, 2006). Thus, other factors in addition to Lhx8 may be necessary for the differentiation of cholinergic neurons. LIM-homeodomain factors also function coordinately to instruct cell fate (Thor et al., 1999). The requirement for the LIM-homeobox gene Isl1 in spinal motor neuron and starburst amacrine differentiation (Pfaff et al., 1996;Elshatory et al., 2007b), both cholinergic neurons, makes Isl1 a candidate for regulating forebrain cholinergic differentiation. The expression of Isl1 in developing rat forebrain (Wang and Liu, 2001), the later restriction to subsets of cholinergic neurons (Wang and Liu, 2001), and the downregulation of Isl1 in Lhx8-null forebrains (Zhao et al., 2003) further support a role for Isl1 in forebrain cholinergic development.
Here, we present data using Isl1 conditional knock-outs showing the regulation by Isl1 of restricted populations of telencephalic cholinergic neurons that overlap with the pool of cholinergic neurons disrupted in Lhx8-null mice. The data provide evidence that forebrain cholinergic differentiation depends on the function of multiple LIM-homeobox genes, which may act hierarchically or coordinately to promote the differentiation of different subsets of cholinergic neurons.

Materials and Methods
Animals. The generation of and genotyping for the Isl1 lacZ knock-in and floxed Isl1 mice, Isl1 fl/ϩ , has been described previously (Elshatory et al., 2007a,b). The Six3-cre transgenic line used in the present study has also been reported previously (Furuta et al., 2000). All animal procedures described herein were approved by the University Committee of Animal Resources at the University of Rochester.
For the purposes of staging embryos, noon of the day a vaginal plug was detected was taken to be embryonic day 0.5 (E0.5). Embryos were anesthetized on ice and immersion fixed for 3 h in 4% paraformaldehyde (PFA) at 4°C, washed several times in PBS, and then cryoprotected in 30% sucrose in PBS before rapid freezing in OCT compound (Tis-sueTek). To fix brains from postnatal day 0 (P0) pups and 3-week-old mice, animals were anesthetized, perfused transcardially with PBS, followed by 10 or 60 ml of 4% PFA in PBS, pH 7.3. Brains were removed subsequently and postfixed overnight at 4°C in the same fixative, and then cryoprotected and frozen as described for embryos.

Expression of Isl1-lacZ in the mouse basal forebrain
Isl1 expression has been detected in diverse cholinergic pools in the rat nervous system (Thor et al., 1991;Galli-Resta et al., 1997;Wang and Liu, 2001). In the adult mouse brain, colocalization of Isl1-lacZ expression with markers of cholinergic neurons revealed expression of Isl1-lacZ in cholinergic interneurons of the dorsal striatum ( Fig. 1 A) and ventral striatum ( Fig. 1 B), but also in cholinergic projection neurons in the nucleus basalis (Fig. 1C), and fewer cholinergic projection neurons in the magnocellular preoptic area (MCPO) (Fig. 1 D) and MS ( Fig. 1 E). The percentage of colocalization of Isl1-lacZ and ChAT varied across the groups, with the highest percentages being encountered in the caudate-putamen and nucleus basalis (average Ϯ SD, n ϭ 3: 95.4 Ϯ 0.6 and 86.9 Ϯ 7.3%, respectively). The lowest percentages of colocalization were encountered in the MCPO/SI and Ch1-2 (average Ϯ SD, n ϭ 3: 35.6 Ϯ 14.4 and 24.6 Ϯ 6.1%, respectively). However, expression was not detected in striatal projection neurons immunoreactive for DARPP-32 (Ppp1r1b) ( Fig. 1 F) or GABAergic interneurons in the striatum immunoreactive for NPY ( Fig. 1G), parvalbumin ( Fig. 1 H), or calretinin (calbindin 2) ( Fig. 1 I). The colocalization of the recombination pattern of Six3-cre with cholinergic and striatal interneuron markers was gauged to determine the appropriateness of its use to study the function of Isl1 in the forebrain. Most cholinergic interneurons were positive for the lineage of Six3-cre ( Fig. 1 J, black arrowhead), although a minor population of reporter-negative cholinergic interneurons was encountered ( Fig. 1 J, white arrowhead). Broader Six3-Cre-mediated recombination was encountered in the nucleus basalis and medial septum, encompassing most cholinergic projection neurons in these regions ( Fig. 1 K, L, black arrowheads). The percentages of colocalization of Six3-cre-R26R and ChAT indicate that Six3-cre recombination occurs in the majority of forebrain cholinergic neurons. For example, in the caudate-putamen, 78.1 Ϯ 5.9% of ChAT ϩ interneurons were Six3-cre-R26R ϩ , 89.2 Ϯ 3.5% of cholinergic neurons in the nucleus basalis were Six3-cre-R26R ϩ , and 91.9 Ϯ 3.4% were positive in Ch1-2 neurons (averages Ϯ SD; n ϭ 3). NPY ϩ and parvalbumin-positive striatal interneurons similarly expressed the reporter (Fig. 1 M, N, black arrowheads), whereas calretininpositive striatal interneurons were routinely negative for reporter expression (Fig. 1O, white arrowheads). Examination of the developing forebrain at E13.5 revealed Isl1 expression in cells negative for phosphorylated histone H3 expression, an M-phase cell cycle marker, in the subventricular zone of the lateral ganglionic eminence (LGE) and the AEP/POa, as well as scattered cells in between ( Fig. 1 P). The ventral-most expression of Isl1 in postmitotic neurons of the AEP/POa is suggestive of a role in the differentiation of neurons emanating from this region, such as cholinergic neurons. Analysis of Isl1 fl/fl ; Six3-cre mice revealed a depletion of Isl1 expression in the AEP/POa (Fig. 1Q, arrowheads), but its maintenance more dorsolaterally in the LGE (Fig.  1Q, bracket). This pattern of Isl1 deletion is consistent with the greater efficiency of Six3-cre recombination in the AEP/POa compared with the LGE (supplemental Fig. 1, available at www. jneurosci.org as supplemental material). More anteriorly, Isl1 expression in the septal anlagen is also abolished (data not shown).

Marked reduction in cholinergic but not GABAergic interneurons in striata of adult Isl1 conditional knock-outs
The expression of interneuron striatal markers was assessed in Isl1 fl/fl ; Six3-cre animals compared with their littermate controls at 3 weeks of age. Labeling for the acetylcholine synthetic enzyme ChAT demonstrated an 86% reduction in cholinergic interneuron marker expression in the caudate-putamen of Isl1 conditional knock-outs (Fig. 2 A, B) (average number of ChAT ϩ interneurons Ϯ SD: controls, 266 Ϯ 25, n ϭ 5; Isl1 fl/fl ; Six3-cre, 36 Ϯ 20, n ϭ 5; p Ͻ 0.0002, Student's t test). This defect is likely cell-autonomous, because residual Isl1 expression in Isl1 conditional knock-outs is sufficient to confer ChAT expression in the caudate-putamen (supplemental Fig. 2, available at www. jneurosci.org as supplemental material). Cholinergic interneurons in the olfactory tubercle and nucleus accumbens were also markedly reduced (data not shown). The other predominant neurotransmitter phenotype of striatal interneurons is GABAergic, and this group of neurons are further subdivided into subtypes based on the expression of calcium-binding proteins (calretinin, parvalbumin) and neuropeptides (NPY, somatostatin) (Kawaguchi, 1993;Figueredo-Cardenas et al., 1996a,b). No significant difference in the numbers of calretinin, parvalbumin, or NPY-positive interneurons was detected in the caudate-putamen of Isl1 conditional knock-outs compared with controls ( Fig. 2C-H ). The observation that calretinin-positive interneurons are negative for the lineage of Six3-cre in the caudate-putamen (Fig.  1O) indicates a possibility still that Isl1 could regulate the development of this cell type. Using anti-DARPP-32 immunolabeling to reveal the area occupied by striatal projection neurons revealed a slight but significant 26% reduction in Isl1 conditional knockouts (supplemental Fig. 3, available at www.jneurosci.org as supplemental material). Together, Isl1 deletion leads to a profound disruption of cholinergic interneurons in the dorsal and ventral striatum, and does so without disrupting the expression of other interneuron phenotypes significantly.

Early impairment of forebrain cholinergic differentiation in Isl1 conditional knock-outs
The expression of multiple markers of cholinergic neurons was assessed at P0 in Isl1 fl/fl ; Six3-cre and control littermates to assess earlier differentiation of cholinergic neurons after Isl1 disruption. ChAT expression revealed an early disruption of cholinergic interneurons in the caudate-putamen and cholinergic projection neurons in the nucleus basalis compared with controls ( Fig.  4 A, B). Ngfr (also p75) mRNA expression confirmed the early disruption in cholinergic neuron differentiation in the nucleus basalis (Fig. 4C,D), whereas cholinergic neurons in the magnocellular preoptic area appeared to express Ngfr mRNA normally (Fig. 4 E, F ). TrkA immunohistochemistry, which identifies basal telencephalic cholinergic neurons (Sussel et al., 1999), reaffirmed the defect observed in cholinergic interneuron differentiation of the caudate-putamen (Fig. 4G,H ) and cholinergic projection neuron differentiation in the nucleus basalis ( Fig. 4 I, J ), because TrkA expression was markedly reduced. Although expression levels of Lhx8 mRNA in the caudate-putamen appeared to be reduced in Isl1 conditional knock-outs ( Fig. 4 K, L), it was still present, and more numerous than could be accounted for by residual cholinergic interneurons (Fig. 4 B, H ). This suggests that early markers of cholinergic development persist, whereas acquisition of mature cholinergic markers is disrupted (Fig. 4 M, N ). This is also supported by persistent Isl1-lacZ ϩ cells in the caudate-putamen of Isl1 conditional knock-outs, although a broad overexpression of Isl1-lacZ expression in conditional knock-outs obscures clear identification of putative cholinergic interneurons (supplemental Fig. 4, available at www.jneurosci.org as supplemental material). More caudally, Lhx8 mRNA expression is maintained in the ansa lenticularis ( Fig. 4O,P), which represents a structure also containing cholinergic projection neurons that closely apposes the nucleus basalis. Cholinergic projection neurons here are also decreased in Isl1 conditional knock-outs (data not shown), and may represent persisting neurons expressing early markers of cholinergic neurons but failing to differentiate. The close juxtaposition of the ansa lenticularis to the globus pallidus, a structure in which Lhx8 is also expressed (Zhao et al., 2003), makes it difficult to conclude definitively that early markers of cholinergic projection neurons persist in Isl1 conditional knock-outs. Lhx6 mRNA expression was not apparently affected in the neocortex (Fig. 4Q,R, asterisks) or caudate-putamen (Fig.  4Q,R, black arrowheads), suggesting the specification of GABAergic interneurons is unaffected in Isl1 fl/fl ; Six3-cre animals.

Discussion
Isl1 deletion leads to a profound disruption of cholinergic interneurons in the dorsal and ventral striatum, and appears to do so without disrupting the expression of other interneuron phenotypes significantly. Although the reduction in the area occupied by DARPP-32 ϩ neurons might arise from an intrinsic role for Isl1 in DARPP-32 ϩ neurons as postulated (supplemental Fig. 3, available at www.jneurosci.org as supplemental material) (Wang and Liu, 2001;Stenman et al., 2003), it may also be a consequence of the reported reduction in body size in Isl1 conditional knockouts (Elshatory et al., 2007a). If merely a reflection of reduced body size, the reduction in striatal area should generalize to reductions in other brain structures, such as the thickness of the cerebral cortex. However, cortical thickness was not significantly reduced in Isl1 conditional knock-outs, suggesting that Isl1 likely functions in the development of striatal projection neurons (average cortical thickness Ϯ SD: controls, 1.21 Ϯ 0.09 mm, n ϭ 4; Isl1 fl/fl ; Six3-cre, 1.24 Ϯ 0.06 mm, n ϭ 4; NS, Student's t test). This finding is more compelling given the limited recombination of Six3-cre in LGE-derived cells (supplemental Fig. 1, available at www.jneurosci.org as supplemental material) and suggests more efficient deletion of Isl1 in this region using Nestin-cre may reveal a more widespread role for Isl1 in striatal development. The reduced area of the caudate-putamen, however, cannot account for the degree of cholinergic interneuron loss that was observed. The cholinergic neuron loss in Isl1 conditional knock-outs confirms that multiple LIM-homeodomain proteins regulate forebrain cholinergic fate, and establish a scenario in which Isl1, Lhx8, and Lhx6 function in the context of dense transcriptional codes (Flames et al., 2007) to instruct the differentiation of diverse GABAergic and cholinergic neuronal populations. There is a possibility that the remaining cholinergic projection neurons are maintained in Isl1 conditional knock-outs because Isl1 is not adequately deleted, but the widespread recombination of Six3-cre in basal forebrain nuclei ( Fig. 1 K, L), and the robust deletion of Isl1 in the AEP/POa and septal anlagen suggest otherwise (Fig. 1Q) (data not shown). Isl1 conditional knock-outs were found to have varying degrees of cholinergic neuron reduction across forebrain cholinergic groups, but the degree of reduction encountered corresponded approximately to the degree of colocalization of Isl1-lacZ and ChAT in these groups. That is, there was an 86% reduction in cholinergic interneurons in the caudate-putamen, an 83% reduction in cholinergic neurons in the nucleus basalis, a 27% reduction in the MCPO/SI, and no significant difference in Ch1-2 cholinergic neurons (Fig. 3). The high percentage of Six3cre recombination in Ch1-2, and the lack of significant reduction in this pool of cholinergic neurons in Isl1 conditional knock-outs (Fig. 3) suggest that the differentiation of these cells is mostly Isl1-independent.
Fate-mapping experiments have suggested that cholinergic neurons arise from the AEP/POa (Marin et al., 2000). Because Isl1 expression is more completely ablated in the AEP/POa than the LGE in Isl1 conditional knock-outs, and because this correlates with defects in cholinergic neuron differentiation, our findings support earlier assertions regarding the origins of cholinergic   P0. A, B, Light micrographs of anti-ChAT staining at P0 demonstrating its reduction in the nucleus basalis (NB) and caudate-putamen (CP) in knock-outs (B, arrowheads) versus controls (A, arrowheads). Ngfr mRNA expression is considerably reduced in the NB of P0 knock-outs (D, arrowheads) versus controls (C, arrowheads), whereas expression in the MCPO of knock-outs (F , arrowheads) is not clearly disrupted compared with controls (E, arrowheads). Anti-TrkA immunofluorescence shows a reduction in TrkA ϩ neurons in the CP of knock-outs (H, arrowhead) compared with controls (G, arrowheads) at P0, and a similar reduction in TrkA ϩ neurons in the NB of knock-outs (J, arrowheads) compared with controls (I, arrowheads). Lhx8 mRNA expression reveals no clear reduction of Lhx8 expression in the MS, but a decrease in CP expression levels in knock-outs (L, arrowheads) versus controls (K , arrowheads). Higher magnification of Lhx8 mRNA expression in the CP demonstrates the presence of sustained Lhx8 expression despite the lower expression levels in knock-outs (N , arrowheads) compared with controls (M , arrowheads). Lhx8 mRNA is also maintained in the globus pallidus (GP) and, putatively, the ansa lenticularis (AL) in conditional knock-outs (P, arrowheads) as in controls (O, arrowheads). Lhx6 mRNA expression is not clearly disrupted, both in the neocortex (NC) (asterisks) and CP (arrowheads) in conditional knock-outs (R, arrowheads) versus controls (Q, arrowheads). HP, Hippocampus. Scale bars: (in B) A,  neurons. For subsets of forebrain cholinergic neurons in Isl1 conditional knock-outs, early markers of cholinergic development persist, whereas acquisition of mature cholinergic markers is disrupted (Fig. 4). This is also supported by persistent Isl1-lacZ ϩ cells in the caudate-putamen of Isl1 conditional knock-outs (supplemental Fig. 4, available at www.jneurosci.org as supplemental material). Together, Isl1 disruption seems to affect preferentially the differentiation of a subset of forebrain cholinergic neurons.