Directed Neural Differentiation of Mouse Embryonic Stem Cells Is a Sensitive System for the Identification of Novel Hox Gene Effectors

The evolutionarily conserved Hox family of homeodomain transcription factors plays fundamental roles in regulating cell specification along the anterior posterior axis during development of all bilaterian animals by controlling cell fate choices in a highly localized, extracellular signal and cell context dependent manner. Some studies have established downstream target genes in specific systems but their identification is insufficient to explain either the ability of Hox genes to direct homeotic transformations or the breadth of their patterning potential. To begin delineating Hox gene function in neural development we used a mouse ES cell based system that combines efficient neural differentiation with inducible Hoxb1 expression. Gene expression profiling suggested that Hoxb1 acted as both activator and repressor in the short term but predominantly as a repressor in the long run. Activated and repressed genes segregated in distinct processes suggesting that, in the context examined, Hoxb1 blocked differentiation while activating genes related to early developmental processes, wnt and cell surface receptor linked signal transduction and cell-to-cell communication. To further elucidate aspects of Hoxb1 function we used loss and gain of function approaches in the mouse and chick embryos. We show that Hoxb1 acts as an activator to establish the full expression domain of CRABPI and II in rhombomere 4 and as a repressor to restrict expression of Lhx5 and Lhx9. Thus the Hoxb1 patterning activity includes the regulation of the cellular response to retinoic acid and the delay of the expression of genes that commit cells to neural differentiation. The results of this study show that ES neural differentiation and inducible Hox gene expression can be used as a sensitive model system to systematically identify Hox novel target genes, delineate their interactions with signaling pathways in dictating cell fate and define the extent of functional overlap among different Hox genes.


Introduction
The evolutionarily conserved Hox family of homeodomain transcription factors plays fundamental roles in conferring regional identity and regulating cell specification along the anteriorposterior (AP) axis during development of all bilaterian animals [1,2]. Hox genes are expressed in rather broad domains but control cell fate choices in a highly localized, extracellular signal and cell context dependent manner [3,4,5]. Evidence from diverse organisms suggests that Hox proteins act partly as high-level regulators dictating the expression levels of other regulatory proteins including themselves [6,7,8]. They also act partly as ground level regulators, or 'realizators', as initially proposed by Garcia-Bellido [9], fine-tuning very diverse processes such as cell adhesion, cell division rates, cell death and cell movement [10,11,12,13]. Considering their numbers, the scope of their functions, the context dependence of their actions and more than thirty years devoted to their study, few Hox target genes have been identified. Some studies have established direct and downstream target genes in specific systems but their identification is insufficient to explain either the ability of Hox genes to direct homeotic transformations or the diversity of their patterning potential.
Two main general approaches have been used, a candidate target gene approach [14,15,16,17,18] and differential gene expression analysis comparing wild type (wt) tissue with tissue in which specific Hox gene expression has been genetically manipulated [19,20,21,22]. However, the inherent bias in choosing candidate downstream targets, functional redundancy among Hox genes and accumulation of secondary effects in gain or loss of function genetic models present serious limitations. The elucidation of the precise roles that Hox genes play in cell fate specification as well as the identification of target genes and processes are key goals to deciphering the regulatory network underlying morphogenesis of the body plan. Furthermore, this may allow harnessing their patterning potential in the directed differentiation of embryonic stem (ES) cells and induced pluripotent stem (iPS) cells to specific cell types.
During development of vertebrate neural tube the combinatorial use of Hox gene expression and specific dorsoventral (DV) patterning cues define specific subclasses of neuronal progenitors in the developing hindbrain and spinal cord [23]. Genetic evidence suggests that Hox genes act as integrators of AP and DV patterning mechanisms to generate specific classes of neuronal progenitors and neurons for the appropriate AP levels of the hindbrain and the spinal cord. For example, Hoxb1 is specifically expressed in rhombomere 4 of the developing hindbrain. The specification of this territory and subsequent generation of r4 specific neuronal progenitors and neurons depend largely on Hoxb1 function. Disruption of the Hoxb1 gene in mice leads to transformation of the r4 territory into an r2-like state [24,25], whereas retroviralmediated over-expression of Hoxb1 in r2 causes homeotic transformation of r2 to a r4-like identity in chick [26]. In the ventral region of r4, Hoxb1 expression is responsible for the generation of facial branchiomotor neurons and the suppression of serotonergic fate specification [24,27]. Similarly, in more posterior regions of the developing CNS, specific Hox genes direct the generation of distinct motor neuron (MN) subtypes at hindbrain, brachial, thoracic and lumbar regions [28,29,30].
To bypass limitations in delineating Hox gene function in neural development we modeled the role of Hox genes in neural cell fate specification using a mouse ES cell based system that affords the possibility of inducible Hoxb1 expression. Using a differentiation protocol that generates a highly homogeneous population of neural stem (NS) cells and inducible expression of Hoxb1 we showed that timely long term induction (8 days) of the Hoxb1 transgene in ES cell derived NS cells resulted in the specification of NS cells toward a hindbrain specific identity through the activation of a rhombomere 4-specific genetic program and the repression of anterior neural identity [31]. These effects were accompanied by specific changes in the expression of neural progenitor markers some of which suggested that Hoxb1 mediates neural crest cell fate induction. This was subsequently verified in vivo [32]. Furthermore, up regulation of the known Hoxb1 target genes, Hoxb2, Hoxa2, EphA2 and Phox2b [31] suggested that this approach could be used to identify novel Hoxb1 target genes.
Here we use this approach and microarray gene expression profiling to identify potential novel Hoxb1 target genes and processes. To compare the long and short term effects of Hoxb1 function and limit the number of potential target genes we used a short term and a long term induction protocol. To validate the approach and elucidate aspects of Hoxb1 in vivo function we used loss and gain of function approaches using the chick and mouse developing embryos as model systems and investigated the in vivo response of two up (CRABPI, II) and two down (Lhx5, 9) regulated genes in ES derived NS cells. Hoxb1 is itself regulated by retinoic acid [33,34] and we found intriguing the possibility that it may regulate the expression of RA signaling effectors such as CRABPI and II. On the other hand, Lhx5 and 9 mediate neuronal differentiation [35] and their in vivo repression would correlate well with the finding that Hoxb1 blocks ES derived NS cell differentiation after mitogen withdrawal [31]. Notably, these genes have not been identified as Hoxb1 downstream target genes in other approaches [19,20,36] demonstrating that ES neural differentiation and Hox inducible gene expression can be used as a sensitive model system to identify novel Hox target genes and processes, define binding sites and elucidate the interactions of Hox genes and extracellular signals in dictating neural cell fate.

Animals
Animal studies were conducted in accordance with international guidelines and after ethical approval of the competent Veterinary Service of Athens. The Hoxb1 mouse mutants were described and genotyped as reported [24]. Fertilized chick eggs were obtained from Pindos Hellas (Ioannina, Greece) and incubated in a humidified incubator at 38uC.

Microarray gene expression profiling
The generation and neural differentiation of the mouse ES Tet-On/Hoxb1 cells were as described previously [31]. For the short Hoxb1 induction scheme doxycycline (dox) was added during the last day of the selection period and for one additional day during the expansion stage (Fig. 1A). Gene expression profiling was carried out for biological triplicates for both dox induced (Hoxb1 + ) and uninduced (Hoxb1 2 ) cells as described earlier [31] and the Affymetrix Mouse Genome 430A array was used. Microarray data are deposited in the public access Array Express database (Experiment ID E-MIMR-441). The list of regulated genes for the short induction scheme was restricted to genes with 0.75. fold regulation .1.3 and genes that were also present in the long induction scheme.

In situ hybridization and immunofluorescence
For in situ hybridization embryos were fixed overnight at 4uC in 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer saline (PBS). In situ hybridization was performed in whole embryos using probes for mouse CRABPI and CRABPII [37], mouse Lhx9 [38] and for chick Lhx9 [39] and Lhx5 [40]. Antisense digoxigenin-labelled riboprobes were synthesized from linearized templates by the incorporation of digoxigenin-labelled UTP (Boehringer) using T3 or T7 polymerase. Processing of the embryos and hybridization with 500 ng/ml of the probe was as described previously [25]. After whole mount in situ hybridization, embryos were fixed again overnight at 4uC and then processed for immunofluorescence. For immunofluorescence embryos were fixed in 4% PFA in PBS for 1-2 h at 4u. Embryos were cryoprotected with 30% sucrose in PBS and cryosectioned. Blocking was carried out in 10% normal goat serum (NGS) with 0.1% triton for 1 h at RT. The cryosections were incubated overnight at 4uC with the primary antibody diluted in 1% NGS, 0.1% triton in PBS. Primary antibodies used were as follows: rabbit anti-Hoxb1, 1:400 (Covance), mouse Lhx5, 1:100. Secondary antibodies were anti-mouse and anti-rabbit Alexa 488 or Alexa 568 (Molecular Probes) used at 1:500. Images were acquired using a Leica TCS SP5 confocal microscope.

Chick in ovo electroporation
Chick embryos were staged according to Hamburger and Hamilton (HH) (Hamburger and Hamilton, 1951) and electroporated at HH stage 10-11. Chick embryos were electroporated with plasmid DNA at a concentration of 1.5 mg/ml. The coding regions of mouse Hoxb1 cDNA was inserted into the pCAGGS-IRES-NLS-GFP expression vector [41] upstream of the IRES. As a control, pCAGGS-IRES-NLS-GFP was included at 0.5 mg/ml. Electroporation was carried out using a BTX ECM830 electroporator delivering five 20 V pulses of 50 millisecond duration each. Electroporated embryos were dissected at the desired stage and fixed for in situ hybridization or immunofluorescence.

Identification of Hoxb1 target genes
To identify potential Hoxb1 target genes and processes we used the stable line ES Tet-On/Hoxb1 that allows for tight dox mediated inducible expression of the Hoxb1 transgene at both the ES cell and NS cell stages. However, inducible expression of the transgene could mobilize the endogenous Hoxb1 autoregulatory loop only at the NS cell stage demonstrating the importance of cellular context for Hoxb1 function and its analysis. Hoxb1 induction using an 8-day long dox exposure resulted in the generation of r4 specific neuronal progenitors [31]. Microarray gene expression analysis was used to identify the genes that were regulated at the end of that period (Table S1). To reduce the number of likely Hoxb1 downstream effectors and compare the short term and long term effects of Hoxb1 expression we performed microarray gene expression analysis after a two day long exposure to dox (Fig. 1A). Analysis of the microarray data using fold regulation cut offs (0.75, fold regulation .1.3) and stringent statistical criteria (FDR ,0.005) showed that the number of regulated probe sets increased with time from 209 regulated genes at 2 days of exposure to 1017 regulated genes at 8 days of exposure (Fig. 1B, Table S1 and Table S2). Interestingly, the percentage of repressed genes increased from 55% to 73% with time suggesting that the long-term effects of Hoxb1 expression were primarily to repress genes and thus exclude alternative fates, consistent with Hoxb1 acting as a cell fate selector gene (Fig. 1C). To identify Hoxb1 regulated processes we performed Gene Ontology (GO) analyses for the genes identified in the long-term induction scheme. Strikingly, repressed and activated genes segregated in distinct GO processes. Up regulated genes were associated with early patterning and developmental activities including signaling whereas down regulated genes were associated with late, differentiation processes (Table 1).
We then turned to choosing genes for in vivo validation. To increase specificity, we focused on genes regulated in both short and long term exposure experiments (Table 2). Regulation was towards the same direction with the notable expression of only three genes and generally stronger in the long term (for a full list see Table S2). We then used qPCR and in vivo loss and gain of function approaches to validate the results for two up regulated and two down regulated genes. Real Time PCR analyses for the regulation of CRABPI, CRABPII, Lhx5 and Lhx9 using the long induction scheme yielded results that were in good agreement with the microarray results ( Fig. 2) suggesting that they were appropriate candidates for further, in vivo analyses.

Hoxb1 modulates RA signaling by regulating expression of CRABPI and CRABPII in r4
The results presented above suggested that Hoxb1 patterns the hindbrain at least partly by modulating the cellular response to RA through the regulation of CRABPI and CRABPII. To examine this hypothesis we compared the CRABPI and CRABPII expression in wt and Hoxb1 2/2 mouse embryos at 10.5 dpc using in situ hybridization.
CRABPI and CRABPII are both expressed in the developing hindbrain in a rhombomere specific manner. CRABPI expression first appears at the five-somite stage caudal to the preotic sulcus. During subsequent stages, expression spreads to the rest of the hindbrain but remains stronger in the caudal hindbrain, particularly in r4, 5 and 6 [37] (Fig. 3A). CRABPII expression appears at the same early stage as CRABPI in the post-otic region of the hindbrain and its expression subsequently spreads to the rest of the hindbrain [37]. CRABPI and CRABPII expression is generally stronger in r4 and the caudal hindbrain. At 10.5 dpc neural progenitors acquire specific identity and both CRABPI and II are expressed in rhombomere specific longitudinal stripes prefiguring sites of generation and differentiation of defined neuronal subtypes (Fig. 3A, C). In wt r4, strong CRABPI expression extends to a ventral domain corresponding to the resident site of facial motor neuron progenitors (arrows, Fig. 3A). Compared to more anterior rhombomeres, there is also stronger expression of CRABPI in dorsomedial positions of r4 (arrowheads, Fig. 3A). In wt r4, CRABPI expression is excluded from the resident site of facial motor neuron progenitors but there is strong expression in an adjacent domain (arrows, Fig. 3C) as well as in medial and dorsomedial positions of r4 (brackets, Fig. 3C).
The r4 expression pattern of CRABPI and II in Hoxb1 2/2 embryos changed dramatically. The ventral most expression domain of CRABPI was lost and expression of both CRABPI and CRABPII in medial and dorsal stripes was either lost or weakened (asterisks, Fig. 3B, D). Overall, consistent with an r4 to r2 homeotic transformation [24,42], the r4 expression patterns of CRABPI and CRABPII in the Hoxb1 2/2 embryos became identical to those of r2.
Thus the identification of CRABPI and II as Hoxb1 downstream genes in our screen suggested that part of Hoxb1 patterning activity may be mediated by regulation of the RA signaling activity through the up regulation of CRABPI and CRABPII gene expression.

Hoxb1 represses the expression of Lhx5 and Lhx9
We then examined whether Hoxb1 can repress Lhx5 and Lhx9 expression in vivo. To study the expression of Lhx5 in the mouse hindbrain and specifically in r4 we performed whole mount in situ hybridization using a specific Lhx5 probe [38]. At 10.5 dpc in the hindbrain, Lhx5 is expressed in two dorsoventral stripes along r1-r6 in a rhombomere specific pattern. In wt r4 there is a paucity of After gene expression profiling of cells in the long induction scheme upregulated and downregulated genes were separately subjected to GO analysis. The ratio is represented by the number of genes regulated in a particular GO category over the total number of genes in that GO category. doi:10.1371/journal.pone.0020197.t001 Lhx5 expression in the ventral domain corresponding to the site of motor neuron progenitors whereas expression in the dorsal stripe is weaker compared to that of r2 and r3 and similar to that of r5 and r6 (brackets, Fig. 4A). In Hoxb1 2/2 r4 Lhx5 expression increases in both the dorsal and ventral domains and becomes similar with the expression pattern of r2 and r3 (brackets, Fig. 4B). Thus r4 expression of Hoxb1 and Lhx5 appeared to be mutually exclusive. This was confirmed, by Lhx5 and Hoxb1 immunofluorescence on wt r4 transverse sections (Fig. 4C). In Hoxb1 2/2 r4 expression of Lhx5 expanded in both ventral and dorsal expression domains. This was consistent with the in situ hybridization results and suggested that Hoxb1 may repress expression of Lhx5. To address this, we ectopically expressed Hoxb1 in the hindbrain of HH stage 10-11 chick embryos using in ovo electroporation. The embryos were analyzed 48 h post electroporation (PE) (HH stage 20) by whole mount in situ hybridization with the chick Lhx5 in situ hybridization probe [40] and Hoxb1 immunofluorescence. The cLxh5 at HH is expressed in two dorsomedial stripes in r2 and r3 (arrowheads Fig. 4E). Expression of cLhx5 was specifically down regulated in the areas where Hoxb1 was ectopically expressed (asterisks, Fig. 4E, F) and this was confirmed by r2 transverse sections showing that dorsal expression of Lhx5 was lost in the electroporated side of the embryo (Fig. 4G, H).
Lhx9 is broadly expressed in the mouse developing CNS in the forebrain, midbrain, hindbrain and spinal cord. In the mouse, its levels of expression, as detected by RNA in situ hybridization, in the hindbrain were relatively low with no specific r4 pattern [43]. Using a chick Lhx9 in situ probe [39] we found that cLhx9 is expressed in dorsal r1 and in a thin dorsal stripe in the developing chick hindbrain (arrowheads, Fig. 1A, C, D). Thus we choose to do our analysis in chick embryos by ectopically expressing Hoxb1 in the developing hindbrain. Chick embryos were electroporated with Hoxb1 expression vector at HH 10-11 and RNA in situ hybridization was performed 48h PE to detect cLhx9 expression. The expression of cLhx9 in the nonelectroporated side was strong along the whole length of the hindbrain but, in the electroporated side, cLhx9 was down regulated in response to ectopic Hoxb1 expression. This was evident in whole mount embryos and flat mounted hindbrains (asterisks in Fig. 5B, C, D) and these findings were confirmed by cryosections (Fig. 5E, F).
Taken together these results showed that Hoxb1 represses expression of both Lhx5 and Lhx9 thus confirming the results of the microarray gene expression analysis in ES cell derived Hoxb1 2 and Hoxb1 + NS cells.

Discussion
The Hox patterning genes play diverse roles during embryo development in all three germ layer derivatives. An approach to understand their function was to compare the transcripteomes of wt tissue with tissues where Hox gene expression has been genetically manipulated [19,20,21,22]. However, tissue heterogeneity, accumulation of long term effects that are not directly related to Hox gene function and functional redundancy among Hox genes limit the utility of this approach. Additionally, it is becoming increasingly evident that Hox activity is dependent upon extracellular signals and cellular context [5,31,44,45,46,47,48,49,50,51]. Thus, to identify Hox target genes in a given cell specification process a model system recapitulating key aspects of this process could provide novel insights. We have shown that directed neural differentiation of mouse ES cells and inducible Hoxb1 expression recapitulates key aspects of r4 neural specification [31]. Here we investigated whether this approach could be used to identify novel downstream effectors of Hoxb1. Microarray gene expression analysis identified both induced and repressed genes in response to Hoxb1 expression. Comparison of the effects of short term and long term Hoxb1 induction showed that whereas Hoxb1 acted as both activator and repressor of gene transcription in the short term, its long-term effects were mostly repressive suggesting that its fate selector function included active exclusion of alternative genetic programs. Strikingly, gene ontology (GO) analysis showed that up regulated and down regulated genes related to strictly distinct processes. The Hoxb1 repressing activity was directed primarily towards differentiation related processes whereas its activating functions were directed primarily towards early development, wnt and cell surface receptor linked signal transduction and cell-to-cell communication ( Table 1). These results were consistent with the finding that Hoxb1 expression delayed differentiation of ES derived NS cells in the absence of a mitogen and pinpointed likely effectors of these effects [31]. Thus Hoxb1 plays a role in maintaining neural progenitor state and delaying differentiation. This does not rule out the possibility that Hoxb1 may have distinct functions in post mitotic, maturing neural cells. A role in post mitotic maturation of motor neurons has been assigned to some members of the Hox family [30,52,53] and it is not understood whether distinct Hox genes are involved in either proliferating progenitors or post mitotic neural cells or both and to what extent. The approach described here offers a venue to address these issues.
Three other screens have been conducted to identify Hoxb1 downstream effectors in r4 using tissue from mouse wt and Hoxb1 2/2 hindbrains [19] or zebrafish wt and Hoxb1a knock down hindbrains [20] and by identifying the expression profiles of distinct mouse rhombomeres [36]. It is important to bear in mind that Hox gene activation in the mouse occurs around 7.5 dpc and the screens were performed at 9.5 or 10.5 dpc and, similarly, in zebrafish, Hox gene expression starts at around 10 hpf and the screen was conducted at 20 hpf. Thus there was ample time for multiple intermediate regulatory steps to take place and the observed readout was a combination of direct and indirect Hox targets, other patterning influences and co-regulated genes. In the screen based on ES derived NS these effects are minimized, due mainly to the absence of neighboring tissues, albeit not completely eliminated. In two of the studies selected genes were validated by corroborating changes in their expression profiles in wt and mutants [20,36]. We have identified some, but not all, of these genes as well in our long induction scheme. Surprisingly, some of these genes were repressed in our screen rather than activated. A comparison of regulated genes in our long induction scheme revealed that about 10% (120 out of 1117) of them were also found regulated in the r4 of the Hoxb1 2/2 mouse mutants [19]. Again, many of them were regulated in opposite directions (Table S3 and  Table S4). An important difference between the methods followed previously and the approach described here is that the former combined cells of the ventricular and mantle layers at a time point when post mitotic neuronal cells abound whereas our approach relied on actively dividing neural progenitor cells representative of an earlier time point of development. This raises the intriguing possibility that some Hoxb1 regulated genes switch from repressed to activated (and conversely) upon cell cycle exit. To in vivo validate some of our findings we corroborated the effects of Hoxb1 on the expression patterns of CRABPI, CRABPII, Lhx5 and Lhx9 using in vivo loss and gain of function models. CRABPI, CRABPII and Lhx5 had a Hoxb1 dependent r4 specific expression pattern. It is worth noting that none of them was identified as such in the aforementioned screens underlining the sensitivity of the approach presented here.
Within the developing neural tube the diverse cellular distribution patterns of retinoid receptors and retinoid binding proteins indicates that it is necessary to fine-tune levels of RA signaling for the specification of diverse of neural subpopulations. CRABPI and II are located in the cytoplasm and bind RA, a key player in CNS pattern formation, neural specification and differentiation. CRABP expression was initially associated with structures that were more sensitive to excess of RA [54] and subsequent studies shed light in the function of these proteins. CRABPI participates in reducing the cellular RA response and associated differentiation by accelerating RA degradation [55,56]. On the other hand, CRABPII acts as a ligand dependent coactivator of RAR translocating in the nucleus in the presence of RA thus facilitating its channeling to RAR and potentiating RA dependent transcriptional activation. [57,58,59]. Expression of both CRABPI and II was activated by Hoxb1 in ES derived NS and these findings were validated in the mouse embryo since expression of both was down regulated in the r4 of Hoxb1 2/2 reverting to expression patterns identical to those of r2. Intriguingly, CRABPI is up regulated whereas CRABPII is down regulated in the resident territory of r4 motor neurons suggesting that maturation and/or specification of this subpopulation needs particular shielding from RA exposure. Ectopic Hoxb1 expression in r2 through timely supply of extraneous RA converts the r2 trigeminal motor neurons into r4 facial motor neurons [60,61]. Conversely, loss-of function of Hoxb1 converts r4 facial motor neurons into trigeminal motor neurons [24,25]. Thus RA is necessary for facial motor neuron specification acting as an upstream regulator of Hoxb1 [33,34] and in turn, Hoxb1 finetunes RA availability through the regulation of CRABPI and II expression. However, further studies are needed to prove this hypothesis and establish whether CRABPI/II are direct Hoxb1 target genes. The localized expression of RARa in r4 and the localized expression of Cyp1B1, an atypical RA generating cytochrome, in the ventral r4 [36] lends further support for an important role of RA during the patterning of this territory. Both our screen and previous screens [19,20,36] have identified RARa as a Hoxb1 downstream target in r4. The ES derived NS cells are a mixture of different DV characters and this limits the detection capacity for markers that are exclusively expressed in distinct and narrow DV levels. This can be bypassed by dorsalising or ventralising these cells with appropriate DV morphogenetic signals [32]. It will be interesting to determine whether Cyp1b1 is induced in shh treated ES derived Hoxb1 + NS cells as well.  The expression of several members of the LIM domaincontaining subgroup of homeobox transcription factors (Lhx genes) was regulated by Hoxb1 in ES derived NS cells. (Table S2). This subgroup is of considerable interest given that the LIM domain is a modified zinc finger domain that mediates interactions among transcription factors and their major, but not exclusive, role is patterning the CNS. Lhx genes define neuronal identity in a combinatorial manner and they control key aspects of neural cell fate decisions and neuronal differentiation including subtype identity and axonal guidance [35]. Thus they lay temporally downstream of the regionalization of the CNS controlled by Hox genes. In ES derived NS cells, Hoxb1 postpones neural differentiation after mitogen withdrawal through the activation of the Notch signaling pathway [31]. The findings reported here suggest that Hoxb1 may do so partly by temporarily repressing expression of transcription factors such as Lhx. On the other hand, Lhx8 was up regulated in ES derived NS cells by Hoxb1 (Table S2) suggesting that Hox gene patterning activity may be exerted through both repression and activation of Lhx genes. Since Lhx8 is a key player in cholinergic neuron specification [62], Hoxb1 may participate in the specification of this subpopulation in the hindbrain. Lhx5 and Lhx9 are expressed broadly in the developing neural tube in specific subdomains [43]. Our findings suggest that Hoxb1 can repress their expression but it is not yet known whether this is a direct effect. Nevertheless it does imply that Hox genes may act as upstream Lhx regulators in shaping their expression domains and thus participate in neuronal subtype specification.
The results of this study suggest that ES neural differentiation and inducible Hox gene expression can be used as a sensitive model system to address several important open issues pertaining to Hox gene function such as possible differential roles in ventricular and mantle zone neural cells, identify genome wide binding sites by chromatin immunoprecipitation studies, delineate the interactions of Hox genes and DV patterning signals in assigning neural identity and address the issue of specificity and functional overlap among different Hox genes.

Supporting Information
Table S1 List of genes regulated by Hoxb1 induction in the long induction scheme as found by microarray gene expression profiling. (XLS)

Table S2
List of genes regulated in both short (s) and long (l) induction schemes. Classification is according to primary GO process assignment. If not an assignment has been made genes are labelled as non-classified. (XLS) Table S3 List of of common genes induced in Hoxb1-/-r4 and also regulated by Hoxb1 in ES derived neural progenitors after long induction. At the top part of the list the observed regulation is in the same direction and after the space it is in the opposite direction. (XLS)

Table S4
List of common genes repressed in Hoxb1-/-r4 and also regulated in ES derived neural progenitors after long induction. At the top part of the list the observed regulation is at the same direction and after the space it is in the opposite direction. (XLS)