• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of plosbiolPLoS BiologySubmit to PLoSGet E-mail AlertsContact UsPublic Library of Science (PLoS)View this Article
PLoS Biol. Feb 2007; 5(2): e37.
Published online Feb 6, 2007. doi:  10.1371/journal.pbio.0050037
PMCID: PMC1790951

Specification of Neuronal Identities by Feedforward Combinatorial Coding

William A Harris, Academic Editor

Abstract

Neuronal specification is often seen as a multistep process: earlier regulators confer broad neuronal identity and are followed by combinatorial codes specifying neuronal properties unique to specific subtypes. However, it is still unclear whether early regulators are re-deployed in subtype-specific combinatorial codes, and whether early patterning events act to restrict the developmental potential of postmitotic cells. Here, we use the differential peptidergic fate of two lineage-related peptidergic neurons in the Drosophila ventral nerve cord to show how, in a feedforward mechanism, earlier determinants become critical players in later combinatorial codes. Amongst the progeny of neuroblast 5–6 are two peptidergic neurons: one expresses FMRFamide and the other one expresses Nplp1 and the dopamine receptor DopR. We show the HLH gene collier functions at three different levels to progressively restrict neuronal identity in the 5–6 lineage. At the final step, collier is the critical combinatorial factor that differentiates two partially overlapping combinatorial codes that define FMRFamide versus Nplp1/DopR identity. Misexpression experiments reveal that both codes can activate neuropeptide gene expression in vast numbers of neurons. Despite their partially overlapping composition, we find that the codes are remarkably specific, with each code activating only the proper neuropeptide gene. These results indicate that a limited number of regulators may constitute a potent combinatorial code that dictates unique neuronal cell fate, and that such codes show a surprising disregard for many global instructive cues.

Author Summary

The nervous system contains a daunting number of different cell types, perhaps as many as 10,000 in mammals, far outnumbering regulatory genes in many animal species. Studies of the determinants of cell fate in many systems during the last decade have supported the conclusion that cell fate is not determined by any one regulatory gene, but results from the combinatorial action of several regulators. Many questions about the nature of such codes, however, remain. It is not known, for example, how complex such codes are or how they are established. It is also unclear whether they are confined in their action or if they act outside of their normal context. To address these outstanding issues, we have used two unique subsets of Drosophila neurons, identifiable by their specific expression of two different neuropeptide genes. We have identified two partially overlapping and relatively simple codes, consisting of four to seven regulators that act to specify these two cell types. Intriguingly, specification is achieved in a feedforward manner such that A activates B, followed by A/B activating C, and A/B/C activating D. Each code is surprisingly potent, and can ectopically activate neuropeptide gene expression in a variety of neurons, with a surprising disregard for many early patterning events.

Introduction

Animals have a daunting number of different cell types, and this cellular diversity is profound in the nervous system. During the last two decades, it has become increasingly apparent that neuronal cell fate is not dictated by the action of any single regulatory gene, but results from the combinatorial action of a number of genes that comprise a regulatory “code” [13]. Given the enormous diversity of neuronal cell identities, and the great number of regulatory genes in the genome(s), the “decoding” of neuronal cell fate specification is only just beginning.

The appearance of the proper combinatorial code in an early postmitotic neuron is the consequence of a sequence of earlier and increasingly restricted regulatory events [4,5]. How do the earlier regulatory events influence postmitotic cell fate decisions? In one extreme model, early regulators may act simply to ensure proper combinatorial coding in early postmitotic neurons. At another extreme, they may confer a state of competence that to some extent reduces the need for precise and extensive coding of cell fate. If the first model is correct, combinatorial coding should have great informational value and easily traverse many developmental boundaries—once in place, a combinatorial code should be able to act without major restrictions. If the second model is correct, combinatorial codes should have low informational value and be restricted by developmental boundaries—although the proper combinatorial code may be in place, its activity will often be restricted by the history of the cell. Given the extent of neuronal diversity in the mammalian nervous system, and the number of regulatory genes involved, distinguishing between these two models relies on the identification of specific codes in more easily accessible genetic model systems.

We are focusing our studies on a specific subset of Drosophila ventral nerve cord (VNC) neurons. Within the VNC, only about 90 out of approximately 10,000 cells express the LIM-homeodomain transcription factor Apterous (Ap), but this group of neurons consists of several classes [6] (Figure 1), which are located in three positions in each hemisegment; the single dorsal Ap (dAp) neuron, a ventral pair of (vAp) neurons, and a lateral “Ap cluster” of four neurons found only in each of the six thoracic hemisegments. Most Ap neurons are interneurons that project along a common fascicle. However, in each Ap cluster, the Tv neuron projects its axon out of the VNC at the dorsal midline and innervates the dorsal neurohemal organ (DNH) [7]. Tv neurons uniquely express the FMRFamide (FMRFa) neuropeptide gene [8]. A number of regulatory genes have been found to be important for Tv specification and FMRFa regulation. These include ap itself, the zinc finger gene squeeze (sqz), the bHLH gene dimmed (dimm), as well as the transcriptional co-factors encoded by the eyes absent (eya) and dachshund (dac) genes [7,912]. In addition, FMRFa expression is completely dependent upon retrograde BMP signaling, specifically mediated by the Glass bottom boat (Gbb) ligand acting on the Wishful thinking (Wit) receptor [9,13] (Figure 1G). Two other ap neurons, dAp and Tvb, specifically express the dopamine D1 receptor (DopR) and are also peptidergic [10,14], although the neuropeptide(s) expressed by these neurons was hitherto unknown. Similar to Tv neurons, Tvb/dAp express ap, dimm, and eya, and all three regulators are necessary for DopR expression in these neurons [14] (this study). This indicates that additional regulators are needed to distinguish Tv neurons from Tvb/dAp neurons.

Figure 1
A Specific Subset of Ap Neurons, the Tvb and dAp Cells, Expresses the Nplp1 Neuropeptide, the DopR Dopamine Receptor and the Col Transcription Factor

Here we identify the neuropeptide gene expressed by Tvb/dAp as the Neuropeptide like precursor protein 1 (Nplp1) gene. We find that these neurons also express the COE family member collier (col; Flybase knot). In col mutants, as well as in ap, eya, and dimm mutants, expression of Nplp1 and DopR is severely affected or completely lost. Genetic analysis further reveals that col is expressed in the lineage generating the Ap cluster, and acts at an early postmitotic stage to activate ap and eya, but not sqz and dac. col subsequently acts with ap and eya to activate dimm, and together with all three regulators to activate the Nplp1 and DopR terminal differentiation genes in a “feedforward cascade.” Combinatorial misexpression of col, ap, eya, and dimm can potently activate Nplp1 and DopR in many neurons throughout the VNC. Although this code is similar to the FMRFa code, simply replacing col with dac is sufficient to shift the specificity of the code from Nplp1/DopR expression to widespread activation of FMRFa. In summary, these combinatorial codes are established in a multistep feedforward manner. Once established, they have high informational value for the specification of neuronal identity, and potently activate specific terminal differentiation genes, with limited cross-activation of the incorrect ones. Surprisingly, ectopic activation of terminal differentiation genes can be observed throughout the VNC, disregarding many known developmental boundaries such as anteroposterior, dorsoventral, and mediolateral boundaries.

Results

A Subset of Ap Neurons Specifically Express the Neuropeptide Gene Nplp1 and the Dopamine Receptor DopR

In the developing Drosophila VNC, approximately 90 neurons express the LIM-homeodomain regulator Apterous (Ap), and these represent at least six different cell types. Herein, we will focus on three of the Ap neurons: two cells of the Ap cluster—the Tv cells, which express FMRFa, and the Tvb cells, which together with the dAp cells express DopR (Figure 1G). A number of regulators involved in Tv neuron specification have been identified, but to better understand specification of the related Tvb/dAp neurons, we wanted to identify the putative neuropeptide gene expressed by Tvb/dAp neurons. The completion of the Drosophila genome led to the prediction of several additional neuropeptide genes, including the Neuropeptide like precursor protein 1–4 genes (Nplp1–4; Flybase, http://flybase.bio.indiana.edu/). The validity of these predictions has been confirmed by the identification of expressed sequence tags (ESTs) matching these genes (Flybase, http://flybase.bio.indiana.edu/), and by the detection of amidated and secreted peptides in circulation, and/or in brain extracts [15]. Expression of gene products from one of these genes, Nplp1, was found in a set of cells in the VNC reminiscent of the Tvb/dAp neurons [15]. In situ hybridization for Nplp1 verified that these cells indeed correspond to the dAp neurons and to one Ap cluster neuron (Figure S1C–S1E). To further identify this Ap cluster neuron, we raised antibodies against pro-Nplp1 and against one of the processed and amidated peptides, IPNamide, and detected a similar pattern (Figure S1A and S1B). We used markers for specific subsets of Ap neurons, and could identify the Nplp1-expressing cell in the Ap cluster as the Tvb neuron (Figure 1A–1E). In Tvb/dAp neurons, Nplp1 and DopR expression commences in the late embryo (18 h after egg laying [AEL]) and persists at least into the third larval stage (unpublished data). Nplp1 and DopR are thus specifically expressed by the 28 embryonic and larval Tvb/dAp neurons (Figure 1G).

The COE Family Member col Is Expressed in Tvb/dAp Neurons and Regulates Nplp1 and DopR

Recent studies have revealed that ap and dimm are important for DopR expression in Tvb/dAp [14]. We analyzed whether these Ap neuron determinants, as well as eya, also affected Nplp1 expression (Figure 2). We found that Nplp1 expression depended on eya, ap, and dimm (Figure 2C–2E), and that eya also regulates DopR (Figure 2H–2I). How is Tv versus Tvb/dAp cell fate then determined? Although both cell types express ap, dimm, and eya, only Tv neurons express dac and have activated the BMP pathway. Could the mere absence of dac and/or BMP activation be sufficient to specify the Tvb/dAp fate? To test this, we analyzed expression of Nplp1 in dac and BMP mutants, but found no evidence of ectopic Nplp1 expression in Tv neurons (see below). Thus, specification of Tvb/dAp neurons likely requires additional factors restricted to this cell type.

Figure 2
col and Other Ap Neuron Determinants Regulate Both Nplp1 and DopR

The COE family of HLH regulators is highly evolutionary conserved [16], and is represented in Drosophila by a single member, col (Flybase, knot) [17]. COE genes play important roles during nervous system development in Caenorhabditis elegans and vertebrates [1824], and col is expressed in the developing Drosophila central nervous system (CNS) [17], although no function has yet been assigned to it there. The involvement of members of this gene family in nervous system development in other species, and the embryonic CNS expression of col, prompted us to investigate the possible role of col during Ap neuron specification. col has a dynamic expression pattern in the VNC (below), and we initially focused on its expression in mature Ap neurons, at 18 h AEL, and larval stages. We find that at these stages, col is expressed specifically in Tvb/dAp neurons (Figure 1F), and expression is maintained in these neurons at least into the third larval stage (see below). This raised the possibility that col plays a role in Tvb/dAp cell fate specification. This notion was supported by the complete loss of Nplp1 and DopR expression in col mutants (Figure 2A, A,2B,2B, B,2F,2F, F,2G,2G, and 2I).

col Acts Upstream of Certain ap Neuron Determinants

Previous studies have addressed the regulatory interactions between several of the Ap neuron determinants [912]. However, these had not been addressed in the case of dac, eya, and dimm. As expected from the late onset of dimm expression, we find no evidence of dimm regulation of Dac or Eya (Figure S2M and S2N). Similarly, as expected from the mild effect of dac upon FMRFa, we find that dac does not regulate Dimm (Figure S2L). In contrast, in eya mutants, we find a nearly complete loss of Dimm expression in the Tv, Tvb, and dAp neurons (Figure S2K and S2O).

With a more complete picture of how previously identified Ap neuron determinants interact genetically, we next addressed whether col acts upstream, downstream, or in parallel to other Ap neuron determinants. We first analyzed the expression of Col in embryos mutant for these other regulators. In general, we found no severe effects on Col expression (Figures S2A–S2I). The one exception was in sqz, in which we found a reproducible increase in the number of Col cells (Figure S2B, S2C, and S2J). This was, however, expected, since sqz affects the composition of Ap cluster cells, with an increase both in the number of Ap cluster cells (specifically in T1) and an increase in Tvb cells at the expense of Tv cells (in T1–T3) [9,10]. In line with this, we also find an increase in Nplp1 cells in sqz mutants (Figure S2B, S2C, and S2J).

Does col act upstream of other Ap neuron determinants instead? To address this, we analyzed the expression of these other regulators in col mutant embryos. This analysis was facilitated by the fact that col2 and col3 mutants, which are both genetically strong alleles, are not protein null, thus allowing for detection of Col in col mutants. In addition, Col is expressed by all four Ap cluster neurons at stage 15 (see below). In col mutants (Figure 3), we find that although sqz and Dac are largely unaffected (Figure 3A, A,3B,3B, and 3J), ap, Eya, and Dimm are all completely absent from Ap neurons (Figure 3C–3H and 3J). Because ap, eya, and dimm all regulate FMRFa, the loss of these regulators prompted us to look at the expression of FMRFa as well, and as expected, in col mutants, we find a complete loss of FMRFa in the Tv neurons (Figure 3I and and3J).3J). However, FMRFa is still expressed in the SE2 neurons, a feature common to all identified FMRFa regulators except dimm (Figure 3I). These results suggest that Ap cluster neurons are generated in col mutants, but are incompletely specified because they express part of their normal specification code such as dac and sqz, but not other elements of the code such as ap, Eya, Dimm, FMRFa, Nplp1, and DopR.

Figure 3
col Regulates ap, eya, dimm, and FMRFa, but Not sqz and dac

The Ap Cluster Is Generated by Thoracic Neuroblast 5–6 and col Is Expressed at a Late Stage in This Lineage

The severe effect of col upon ap, eya, and dimm within all Ap cluster neurons, and upon FMRFa within the Tv neuron, is at odds with the restricted expression of Col in Tvb/dAp neurons at late embryonic stages. We therefore analyzed whether Col is more widely expressed at earlier embryonic stages, focusing on the Ap cluster neurons. This revealed that although col is restricted to Tvb neurons at stages 17 and 16 (Figure S3K and S3L), we observe expression in all four Ap cluster neurons at stage 15, the stage at which these neurons are first identifiable using ap, Eya, Dac, and sqz as specific markers (Figure S3E).

Is Col expressed even in the progenitor cells generating Ap cluster neurons? Because ap, Eya, Dac, and sqz are not expressed in Ap cluster neurons prior to stage 15, resolving this issue required the identification of the neuroblast lineage generating the Ap cluster. Extensive work during the last two decades has provided a detailed lineage map of most, if not all, of the 30 neuroblasts found in each hemisegment, and has identified regulatory genes expressed by different neuroblasts [2529]. These studies, together with the extreme lateral positioning of early Ap cluster neurons and their appearance only in thoracic hemisegments, allowed us to use a series of specific markers and determine that the Ap cluster is generated by NB 5–6—a lateral-most neuroblast that has been shown to generate larger lineages in the thoracic segments [28,29] (see Figure S3 and Text S1 for a full description of the mapping of Col expression in NB 5–6). These results, combined with previous detailed studies of NB 5–6 development, allow us to outline a tentative model for the thoracic NB 5–6 lineage, and to place Col expression within this lineage (see below).

Down-Regulation of col Is Critical for Proper ap Cluster Differentiation

Col is expressed by all four newly born Ap cluster neurons and is essential for Ap cluster specification, as evident from the complete loss of ap and Eya expression in col mutants. Col is rapidly down-regulated from three Ap cluster cells and maintained only in Tvb. Is the down-regulation of col important for proper Ap cluster differentiation? To test this, we misexpressed col using the apGAL4 driver, which is not expressed until stage 16, thus maintaining col expression in all four Ap cluster neurons at the time when Col is normally down-regulated. This experiment led to frequent activation of Nplp1 in one additional Ap cluster neuron (Figure 4A and and4B),4B), and staining for Dimm reveals that this cell is indeed the Tv neuron (Figure 4C). FMRFa expression is frequently down-regulated in Tv, but we do observe Tv cells that co-express FMRFa and Nplp1 (Figure 4B). The finding that col misexpression in the Ap cluster only leads to one ectopic Nplp1 cell and no ectopic Dimm cells indicates that col cannot induce a peptidergic cell fate, at least not with this late driver. However, because the two unaffected cells already are expressing ap and Eya, we predicted that co-misexpression of dimm, together with col, should trigger Nplp1 expression in all four Ap cluster neurons. This is indeed what we find (Figure 4D and and4E).4E). In summary, down-regulation of col in three Ap cluster neurons is essential for proper Ap cluster specification.

Figure 4
col Misexpression in the Ap Cluster Triggers Ectopic Nplp1 Expression

col Can Act in a Context-Dependent Manner to Generate Additional Ap Cluster Neurons

Col is expressed prior to ap and Eya in the Ap cluster neurons, and it is essential for ap and Eya expression within these cells. To address whether col is also sufficient to activate ap and Eya, we misexpressed col in all neurons, using the elav-GAL4 driver (Figure 5) [30]. This led to ectopic activation of both ap and Eya (Figure 5A–5E). In addition, we found some activation of Dimm, Nplp1, and DopR expression (Figure 5F–5L and 5O–5Q). Although ectopic activation of ap or Eya alone was found in several regions, co-activation was largely confined to neuroblast row 5—the anterior region of Gsbn expression (Figure 5M and and5N).5N). We typically observe six to ten ap/Eya co-expressing cells in the lateral-most part of row 5 (Figure 5M and and5N).5N). Ectopic activation of ap/Eya, together with Dimm, Nplp1, and DopR, was also confined to lateral-most row 5, i.e., the posterior-most part of gsblacZ cells, and further confined to thoracic segments (Figure 5B–5D, 5J, 5L, and 5O–5Q). Ectopic Nplp1/DopR expression is not overlapping with FMRFa, and there is clear evidence of ectopic Dimm expression (Figure 5P and and5Q),5Q), indicating that additional peptidergic neurons are being generated. Ectopic generation of ap/Eya double-expressing cells, i.e., ectopic “Ap cluster” neurons, was observed already at stage 13, i.e., prior to when Ap cluster neurons are normally born (unpublished data).

Figure 5
Misexpression of col Activates ap, Eya, Dimm, Nplp1, and DopR, but in a Context-Dependent Manner

These results show that col can activate ap and Eya in a number of neurons, but can act to generate bona fide Ap cluster neurons only in a highly context-dependent manner: in lateral, thoracic, row 5 neurons. The appearance of six to ten Eya-expressing cells, but only three to five Nplp1/DopR-expressing cells, and no evidence of ectopic FMRFa expression, suggests that the generation of ectopic Ap cluster neurons is biased toward Tvb (Nplp1/DopR expressing) as opposed to Tv (FMRFa expressing) cell fate. In contrast, although col function depends upon these three positional cues, our results indicate that col is able to override the temporal coding within lateral row 5, and activate Nplp1 and DopR in earlier-born neurons.

col Can Be Partially Rescued by Expression of ap and eya

The loss- and gain-of-function studies place col clearly upstream of ap and eya. Does col act merely to regulate ap and eya in early postmitotic Ap cluster neurons, or does it play additional roles during Ap cluster formation? To further address this issue, we attempted to “cross-rescue” col with ap and eya, by expressing ap and eya in a col mutant background (Figure 6). First, as a positive control, we attempted to rescue col by providing col activity using elav-GAL4/UAS-col. This led to a robust rescue, both of Ap cluster determinants (Eya, ap, and Dimm) and of terminal differentiation genes (Nplp1, DopR, and FMRFa) (Figure 6A–6H, 6L, and 6M). Similar to the col misexpression experiments, we find a clear increase in “Ap cluster” neurons, primarily of the Tvb type, as evident from the finding of six to ten ap/Eya- and three to five Nplp1/DopR-expressing neurons per hemisegment (Figure 6E–6H and 6M). Next, we attempted to “cross-rescue” col mutants with ap and eya, and could indeed find a significant degree of rescue of Ap cluster formation, as evident both from Dimm and FMRFa expression (Figure 6J–6M). In contrast, we found no evidence of rescue of Nplp1 or DopR in these embryos (Figure 6I and and6L).6L). Because we can detect Col in the col2 and col3 mutant backgrounds, we could identify a Dimm/Col-expressing cell adjacent to the Tv/FMRFa neuron (Figure 6K and and6N).6N). This indicates that ap/eya can partially rescue Tvb cell fate, but in the absence of col activity, these “Tvb” neurons do not activate Nplp1. In summary, the finding that in col mutants, ap and eya can partially rescue the Tv cell fate, but not Tvb cell fate, suggests additional roles for col in Tvb specification.

Figure 6
col Rescues Both Tv and Tvb Cell Fate in col Mutants, whereas ap and eya Only Rescue Tv Cell Fate

col Acts at Multiple Steps of Tvb Specification

Our results indicate that Tvb cell fate is not specified by a linear colap/eyadimmNplp1/DopR genetic cascade. To further address this issue, we tested the sufficiency of col, ap, and eya to activate Dimm when misexpressed both alone and in combination. These experiments reveal that although col can trigger some ectopic activation of Dimm, there is little effect upon Dimm when misexpressing ap, eya, or ap/eya (Figure 7A). In contrast, co-misexpression of col with either ap or eya, and in particular, co-misexpression of all three genes, leads to striking ectopic Dimm expression (Figure 7A–7E).

Figure 7
col Can Act in a Combinatorial Manner with ap and eya to Activate Dimm, and with All Three Regulators to Activate Nplp1 and DopR

Does col play a role even at the final step of Tvb differentiation, i.e., in the activation of Nplp1? We attempted to address the possible late role of col by misexpressing it alone and together with other Ap neuron determinants, and then assay its potency in activating Nplp1. Importantly, if there is a simple linear colap/eyadimmNplp1/DopR genetic cascade at work, the effect of triple co-misexpression of ap/eya/dimm should not be enhanced by addition of col to this code. However, we find a striking enhancement of ectopic Nplp1 expression when adding col to this code (Figure 7F–7J). One particular double co-misexpression combination, col/ap, was more potent than others in activating both Dimm and Nplp1 (Figure 7E and and7J).7J). A likely explanation for this effect is that co-misexpression of col/ap activates significant ectopic eya expression (unpublished data).

To further address the late role of col, we generated a transgenic RNA-interference (RNAi) line (UAS-col-dsRNA) and attempted to suppress col gene activity by crossing this line to apGAL4. Because apGAL4 also drives expression in the developing wing disc, we first analyzed the efficiency of this novel tool in suppressing col gene activity in this tissue. This phenocopied the effects of col mutants on wing development [31], with a clear L3–L4 wing vein fusion (unpublished data), indicating that this RNAi transgene specifically blocks col gene activity. However, upon analyzing late larval (third instar) CNSs, we found no effect upon Col expression in Tvb neurons, and as expected, no effect upon Nplp1 expression (unpublished data). Recent studies reveal that RNAi can be efficiently enhanced by overexpression of components of the RNAi pathway, in particular of the Dicer-2 (Dcr-2) gene (G. Dietzl and B. Dickson, personal communication). We therefore co-expressed Dcr-2 with col dsRNA (UAS-Dcr-2/+; apGAL4/UAS-col-dsRNA), and found a clear effect not only upon Col, but importantly, also upon Nplp1 expression. We found no obvious effect in the first instar larvae (unpublished data), but in third instar larvae, Col expression is specifically and completely lost from all Tvb/dAp cells (Figures 8A, A,8B,8B, and and8E).8E). This leads to a complete loss of Nplp1 in 44% of Tvb/dAp cells (Figure 8C–8E), and strongly reduced expression in the remaining expressing cells (Figure 8B). Strikingly, this strong effect upon Nplp1 is not an indirect effect of down-regulation of ap, Eya, or Dimm (Figure 8A–8E). As anticipated, col RNAi has no effect upon FMRFa expression in the third instar larvae (n = 5 VNC; unpublished data).

Figure 8
Postmitotic RNA Interference of col Leads to Loss of Nplp1 Expression, without Loss of ap, Eya, or Dimm

Together, the finding that col can act potently with ap and eya to activate Dimm, that col can act potently with ap, eya, and dimm to activate Nplp1, and that postmitotic RNAi of col strongly affects Nplp1 expression without affecting ap, Eya, or Dimm, does not fit with a simple linear cascade model for Tvb specification. Rather, it strongly suggests that col is acting at several levels during Tvb specification and differentiation.

The Combinatorial Codes Are Highly Potent and Highly Specific

Previous studies have identified several regulators acting to specify Tv fate and to control FMRFa expression. Although co-misexpression of parts of this code had been previously tested, all possible combinations had not. Similar to the combinatorial activation of Nplp1 and DopR, we find that whereas co-misexpression of ap/sqz, ap/dac, or ap/dimm has limited effect upon FMRFa expression (Figure 9) [9,10,12], triple co-misexpression of these regulators, and in particular of ap/dimm/dac, leads to a dramatic ectopic activation of FMRFa (Figure 9G–9K and and99M).

Figure 9
The Two Combinatorial Codes Act in a Highly Potent and Highly Selective Manner

The identification of two partly overlapping and highly potent combinatorial codes allowed us to ask an important question: Does combinatorial misexpression of these regulators merely lead to a general confusion with a mixed neuronal identity, or are these codes truly instructive and specific? To address this issue, we analyzed the expression of Nplp1 and FMRFa in the various misexpression backgrounds. Not surprisingly, when common and partial components of these codes are misexpressed, such as ap/dimm (common to both Tv and Tvb/dAp neurons), we find ectopic activation of both Nplp1 and FMRFa in different subsets of cells (Figure 9B, B,9H,9H, and and9M).9M). However, as a third, and cell-type specific, regulator is added, not only does the amount of ectopic, terminal differentiation gene expression increase dramatically, but we find less evidence of cross-activation of the inappropriate downstream gene (Figure 9C–9F and and99I–9M). This surprising finding reveals that combinatorial misexpression may act in a highly specific and instructive manner, and that these combinatorial codes may be viewed as potent binary switches for cell fate specification. In addition, for both codes, ectopic activation of FMRFa and Nplp1 is observed in neurons throughout the VNC and brain, and traverses many developmental boundaries, such as anteroposterior, dorsoventral, and mediolateral boundaries.

Discussion

We have identified a sequential regulatory cascade of combinatorial coding that acts to specify two unique neuronal cell fates during Drosophila CNS development. Combined with previous studies [6,7,914], the findings described in this study provide the following model for Ap cluster generation and specification (Figure 10). Neuroblast 5–6 forms in the first wave of neuroblast delamination, at late stage 8, and generates a mixed lineage of glia and neurons [28,29]. At stage 13, Col expression is turned on specifically in thoracic NB 5–6, in two subsequent ganglion mother cells (GMCs) at stages 13/14, and in the four Ap cluster neurons generated from these GMCs, during stages 14/15. We have not resolved the birth order of the four Ap cluster neurons, and the sibling relationship of Tv and Tvb is thus unclear. When Ap neurons are born, col activates ap and eya, whereas sqz and dac are activated by unknown regulator(s). sqz appears to play an early postmitotic role, apparently acting in the Notch pathway, to ensure proper Ap cluster composition, and sqz mutants display both additional Ap cluster cells (in T1) and additional Tvb cells (in T1–T3). col is rapidly down-regulated from three Ap neurons, but remains expressed in the Tvb, where it acts with ap and eya to activate dimm at stage 16. At late embryogenesis, col acts with ap, eya, and dimm to activate Nplp1 and DopR in Tvb. In the Tv neuron, ap and eya act, apparently independently of col, to activate dimm expression. In the Tv neuron, eya furthermore plays a role in setting up competence to respond to the BMP signal. At stage 17, the Tv axon reaches the dorsal neurohemal organ (DNH) and receives the BMP ligand Gbb that activates the Wit receptor, and then triggers activation of the BMP pathway in the Tv neuron. At 18 h AEL, ap, eya, sqz, dac, dimm, and BMP signaling cooperate to activate FMRFa in the Tv neuron. In addition to their roles in neuropeptide regulation and BMP signaling, both ap and eya act to ensure proper axon pathfinding of Tv neurons (eya), as well as Tvb and dAp neurons (ap and eya). The role that col may play more directly in axon pathfinding has not been resolved due to the fact that the expression of the appropriate axonal markers (apGAL4, Nplp1, DopR, and FMRFa) is completely absent in col mutants. Given the complexity of axon pathfinding, we would anticipate that several other regulators are yet to be identified before a combinatorial code for “Tv-type” or “Tvb-type” axonal pathfinding is deciphered. Indeed, we find no evidence that combinatorial misexpression of the abovementioned regulators can dictate axonal projections, because ectopic Nplp1 or FMRFa axons are following many different routes in the VNC (Figures 7H and and9I).9I). Finally, dimm also plays additional roles to those described above and is necessary for the expression of neuropeptide-processing enzymes in peptidergic neurons. Importantly, dimm acts independently to control expression of the neuroamidase gene PHM in the Tv and Tvb neurons, and dimm is sufficient to activate PHM in most, if not all, VNC neurons. Thus, during the specification and differentiation of the Ap neurons, there exists a remarkable diversity in the division of labor between the identified regulators, with most of them participating in more than one, but never all, of the identified events.

Figure 10
Model for the Thoracic NB 5–6 Lineage and for the Genetic Pathways Specifying Tv and Tvb Neurons

Sequential Combinatorial Coding: Feedforward Loops Providing High-Fidelity Control of Neuronal Specification

We have identified a multistep process for specifying the Tv and Tvb cell fates. What would be the purpose of this type of sequential combinatorial coding? In other model systems with higher genetic resolution, such as Escherichia coli and yeast, extensive genetic analysis has revealed that this type of sequential gene regulation is quite common [32,33], and a recent study in C. elegans suggests it may also function during neuronal specification [34]. These regulatory nodes, also known as feedforward loops (FFL), have been shown to ensure fidelity in gene regulation [35,36]. For instance, in a simple FFL in which gene A regulates gene B, and A/B then co-operate to regulate C, activation of C depends upon prolonged A expression such that A/B will have time to activate C. If A is only active in a short burst, B may be activated, but C is not, because A/B never co-express for a sufficiently prolonged period of time. The role of col during Ap cluster specification provides an excellent example of a FFL used during neuronal cell fate specification. col is expressed in all four Ap cluster cells and plays an early role in activating ap and Eya, but is only maintained in Tvb, where it plays a later role in activating first dimm, then Nplp1. Maintained expression of col in all four Ap cluster neurons, by driving col expression from apGAL4, leads to activation of the Tvb program also in the Tv neuron. Thus, a burst of col expression has a different informational value than persistent col expression—general Ap cluster specification versus Tvb specification.

Combinatorial Coding and Global Cues: Context Dependence versus Independence

Misexpression of each of the two identified combinatorial codes leads to striking ectopic activation of the Nplp1 and FMRFa genes, and we are surprised by two particular aspects of these findings. First, the global potency of these codes—co-misexpression triggers ectopic FMRFa of Nplp1 in a number of neurons, located in many different anteroposterior, dorsoventral, and mediolateral positions. Thus, it would appear that early regulators mainly act to ensure proper combinatorial coding in each neuron, and play a minor role in restricting cell fate by limiting the cell's competence—once the proper code is in place, the cell fate specification program will be carried out irrespective of the history of the cell. Second, the striking binary effect of these codes is noteworthy—the change of one single player in a code completely alters target gene choice. For instance, misexpression of ap/dimm/dac leads almost exclusively to strong FMRFa activation, but simply replacing dac with col leads to almost exclusively Nplp1 activation. Thus, it appears that more complete codes not only have great potency, but also have great specificity.

The Maintenance of Col, Ap, and Eya Expression

Col shows a very dynamic expression pattern in the VNC—exemplified in NB 5–6 by the expression in the neuroblast, in two GMCs, in all four Ap cluster neurons, and finally only in Tvb. This poses three obvious questions: what activates col in the neuroblast, what shuts it down in three of the Ap cluster neurons, and finally, what maintains col in Tvb? As for the activation of col in the late 5–6 neuroblast, row 5 neuroblast determinants, thoracic determinants, and late temporal determinants are obvious candidates. Indeed, our current work has identified input from a number of such upstream regulators (D. Karlsson, M. Baumgardt, and S. Thor, unpublished data). It is less clear why col expression is lost from three Ap cluster cells and maintained in Tvb. It is possible that the initial expression in all four Ap cluster cells merely reflects residual expression, as an effect of the activation by earlier determinants acting in the neuroblast. But why is col then maintained in Tvb, and similarly, what maintains eya and ap in all four Ap cluster cells? One simple solution would be autoregulation of each gene. But surprisingly, there is no evidence of autoregulation of col (this study) or ap [6,9] (eya has not been addressed). In addition, we find no clear evidence of cross-regulation between col, ap, or eya, at least not during embryonic stages. Thus, it seems likely that other mechanisms, either unidentified regulators or, perhaps, epigenetic mechanisms, act to ensure the continual expression of these regulators during larval (and perhaps adult) life.

Materials and Methods

Fly stocks.

The following stocks were used: apmd544 (referred to as apGAL4), apP44, aprK568 (referred to as aplacZ), sqzDF, sqzGAL4, sqzie, UAS-sqz, UAS-ap, witA12, witB11, UAS-myc-EGFP–farnesylation (referred to as UAS-mEGFPF) [9]; dimmp1, dimmrev4, UAS-dimm [11]; elav-GAL4, [30]; dac3, dac4 [37]; UAS-dac [38]; eyaCli-IID [39]; Df(2L)eya10 (referred to as eya10), P{UAS-eya.B.II} (referred to as UAS-eya) [40]; UAS-nls-myc-EGFP (referred to as UAS-nmEGFP) [41]; ladybird fragment K driving lacZ (referred to as lb-lacZ) [42]; col1, col2, col3 [43]; UAS-col [31]; P{lacZ-un4}hkb5953 (referred to as hkblacZ) [44]; gsb01155 (referred to as gsblacZ) [45]; and wg758 (referred to as wgGAL4) provided by K. Basler. Mutants were kept over CyO, Act-GFP, or TM3, Ser, Act-GFP balancer chromosomes. w1118 was often used as wild type.

RNAi of col.

A 497–base pair (bp) fragment of the col cDNA (position 1,584–2,080 of GenBank sequence NM 080074) was amplified by PCR, sequenced, and then inserted as an inverted repeat into the pWIZ vector [46]. Transgenes, denoted UAS-col-dsRNA, were generated by standard methods. Six different UAS-col-dsRNA transgenic lines were driven by apGAL4, and showed varying degrees of col wing effects (L3–L4 wing vein fusion). Two strong lines were used for CNS analysis. To enhance the col RNAi effect, a UAS-Dcr-2 transgene (on the X chromosome) was used (a gift from G. Dietzl and B. Dickson).

Immunohistochemistry.

Full-length col (in pET17b, provided by M. Crozatier and A. Vincent) and C-terminal part of dimm (amino acid residues 210–390 in pGEX-4T) were expressed in bacteria, and antibodies were raised in rabbits and guinea pigs (AgriSera, Vännäs, Sweden). Antibodies to IPNamide and proNplp1 were raised in hens against the peptide sequences NVGTLARDFQLPIPNamide and FLGRVLPPTRATASTHRSRL, respectively (AgriSera). Antibodies to proFMRFa were raised in rabbits against the C-terminal peptide sequence GAQATTTQDGSVEQDQFFGQ (Covance, Princeton, New Jersey, United States). Peptide antibodies were affinity purified, and all polyclonal sera were pre-absorbed against pools of early embryos. Other antibodies used were: guinea pig α-Deadpan (1:1,000) (provided by J. Skeath), rat monoclonal α-Gsbn (1:10) (provided by R. Holmgren), mouse monoclonal antibody (mAb) α-Col (1:250) (provided by M. Crozatier and A. Vincent), mAb α-ELAV (1:10), mAb α-Repo (1:10), mAb α-c-Myc mAb 9E10 (1:30), concentrated mAb α-β-gal mAb 40–1a (1:20), mAb α-Dac dac2–3 (1:25), mAb α-Eya 10H6 (1:250) (all from Developmental Studies Hybridoma Bank, Iowa City, Iowa, United States), rabbit α-β-gal (1:5,000; ICN-Cappel, Aurora, Ohio, United States), rabbit α-GFP (1:500; Molecular Probes, Eugene, Oregon, United States), and rabbit α-phospho-histone H3-Ser10 (pH 3) (1:2,500; Upstate/Millepore, Billerica, Massachusetts, United States). Immunolabeling was carried out as previously described [10].

Confocal imaging and data acquisition.

Zeiss LSM 5 Confocal microscope was used to collect data for all fluorescent images (Zeiss, Oberkochen, Germany); confocal stacks were merged using LSM 5 software or Adobe Photoshop (Adobe Systems, San Jose, California, United States). Where immunolabeling was compared for levels of expression, wild-type and mutant tissue was stained and analyzed on the same slide. Bright-field images were collected on a Nikon E400 microscope with a SPOT-RT digital camera (Nikon, Tokyo, Japan). Statistical analysis was performed using Microsoft Excel (Redmond, Washington, United States). Where appropriate, images were false colored to facilitate for color-blind readers.

Supporting Information

Figure S1

A Specific Subset of Ap Neurons, the Tvb and dAp Cells, Expresses the Nplp1 Neuropeptide Gene:

(1.7 MB AI).

Figure S2

col Is Not Regulated by Other Ap Neuron Determinants, whereas eya Regulates Dimm:

(8.2 MB AI).

Figure S3

Thoracic Neuroblast 5–6 Generates the Ap Cluster, and col Is Expressed in Late-Stage Thoracic NB 5–6:

(8.0 MB PDF)

Text S1

Thoracic Neuroblast 5–6 Generates the Ap Cluster, and col Is Expressed in Late-Stage Thoracic NB 5–6:

(52 KB DOC)

Accession Numbers

The FlyBase (http://flybase.bio.indiana.edu/search) accession numbers for the genes and gene products discussed in this paper are Ap (CG8376), col (Flybase knot) (CG10197), dac (CG4952), dimm (CG8667), DopR (CG9652), eya (CG9554), FMRFa, (CG2346), Gbb (CG5562), Nplp1 (CG3441), PHM (CG3832), sqz (CG5557), and Wit (CG10776).

The GenBank (http://www.ncbi.nlm.nih.gov/Genbank) accession number for the col cDNA from which a 497-bp fragment was taken is NM 080074.

Acknowledgments

We are grateful to M. Crozatier, A. Vincent, R. Holmgren, P.H. Taghert, J. Skeath, A. DiAntonio, G. Mardon, K. Jagla, K. Basler, G. Dietzl, B. Dickson, the Developmental Studies Hybridoma Bank at the University of Iowa, and the Bloomington Stock Center for sharing antibodies, fly lines, and DNAs. We thank M. Landgraf, D. Van Meyel, G. Technau, and J. B. Thomas for critically reading the manuscript. We are grateful to A. Garces for initiating the studies of collier in our lab. O. Ahlstrom provided excellent technical assistance.

Abbreviations

AEL
after egg laying
CNS
central nervous system
dAp
dorsal Ap
mAb
monoclonal antibody
RNAi
RNA-interference
VNC
ventral nerve cord

Footnotes

¤ Current address: Division of Developmental Neuroscience, Medical Research Council National Institute for Medical Research, London, United Kingdom

Competing interests. The authors have declared that no competing interests exist.

Author contributions. MB, IMA, DK, and ST conceived and designed the experiments and analyzed the data. MB, IMA, and DK performed the experiments. HE contributed reagents/materials/analysis tools. ST wrote the paper.

Funding. This work was supported by the Swedish Research Council, by the Swedish Strategic Research Foundation, by the Knut and Alice Wallenberg foundation, and by the Swedish Royal Academy of Sciences to ST.

References

  • Hobert O. Specification of the nervous system. WormBook. The C. elegans Research Community. 2005. In: doi/ 10.1895/wormbook.1.12.1. Available: http://www.wormbook.org/chapters/www_specnervsys/specnervsys.html. Accessed 22 December 2006. [PubMed]
  • Landgraf M, Thor S. Development of Drosophila motoneurons: Specification and morphology. Semin Cell Dev Biol. 2006;17:3–11. [PubMed]
  • Shirasaki R, Pfaff SL. Transcriptional codes and the control of neuronal identity. Annu Rev Neurosci. 2002;25:251–281. [PubMed]
  • Pearson BJ, Doe CQ. Specification of temporal identity in the developing nervous system. Annu Rev Cell Dev Biol. 2004;20:619–647. [PubMed]
  • Jessell TM. Neuronal specification in the spinal cord: Inductive signals and transcriptional codes. Nat Rev Genet. 2000;1:20–29. [PubMed]
  • Lundgren SE, Callahan CA, Thor S, Thomas JB. Control of neuronal pathway selection by the Drosophila LIM homeodomain gene apterous. Development. 1995;121:1769–1773. [PubMed]
  • Benveniste RJ, Thor S, Thomas JB, Taghert PH. Cell type-specific regulation of the Drosophila FMRF-NH2 neuropeptide gene by Apterous, a LIM homeodomain transcription factor. Development. 1998;125:4757–4765. [PubMed]
  • Schneider LE, Sun ET, Garland DJ, Taghert PH. An immunocytochemical study of the FMRFamide neuropeptide gene products in Drosophila. J Comp Neurol. 1993;337:446–460. [PubMed]
  • Allan DW, Pierre SE, Miguel-Aliaga I, Thor S. Specification of neuropeptide cell identity by the integration of retrograde BMP signaling and a xombinatorial transcription factor code. Cell. 2003;113:73–86. [PubMed]
  • Allan DW, Park D, St Pierre SE, Taghert PH, Thor S. Regulators acting in combinatorial codes also act independently in single differentiating neurons. Neuron. 2005;45:689–700. [PubMed]
  • Hewes RS, Park D, Gauthier SA, Schaefer AM, Taghert PH. The bHLH protein Dimmed controls neuroendocrine cell differentiation in Drosophila. Development. 2003;130:1771–1781. [PubMed]
  • Miguel-Aliaga I, Allan DW, Thor S. Independent roles of the dachshund and eyes absent genes in BMP signaling, axon pathfinding and neuronal specification. Development. 2004;131:5837–5848. [PubMed]
  • Marques G, Haerry TE, Crotty ML, Xue M, Zhang B, et al. Retrograde Gbb signaling through the Bmp type 2 receptor wishful thinking regulates systemic FMRFa expression in Drosophila. Development. 2003;130:5457–5470. [PubMed]
  • Park D, Han M, Kim YC, Han KA, Taghert PH. Ap-let neurons—A peptidergic circuit potentially controlling ecdysial behavior in Drosophila. Dev Biol. 2004;269:95–108. [PubMed]
  • Verleyen P, Baggerman G, Wiehart U, Schoeters E, Van Lommel A, et al. Expression of a novel neuropeptide, NVGTLARDFQLPIPNamide, in the larval and adult brain of Drosophila melanogaster. J Neurochem. 2004;88:311–319. [PubMed]
  • Dubois L, Vincent A. The COE—Collier/Olf1/EBF—transcription factors: Structural conservation and diversity of developmental functions. Mech Dev. 2001;108:3–12. [PubMed]
  • Crozatier M, Valle D, Dubois L, Ibnsouda S, Vincent A. Collier, a novel regulator of Drosophila head development, is expressed in a single mitotic domain. Curr Biol. 1996;6:707–718. [PubMed]
  • Dubois L, Bally-Cuif L, Crozatier M, Moreau J, Paquereau L, et al. XCoe2, a transcription factor of the Col/Olf-1/EBF family involved in the specification of primary neurons in Xenopus. Curr Biol. 1998;8:199–209. [PubMed]
  • Garcia-Dominguez M, Poquet C, Garel S, Charnay P. Ebf gene function is required for coupling neuronal differentiation and cell cycle exit. Development. 2003;130:6013–6025. [PubMed]
  • Persson P, Manetopoulos C, Lagergren A, Nygren J, Gisler R, et al. Olf/EBF proteins are expressed in neuroblastoma cells: Potential regulators of the Chromogranin A and SCG10 promoters. Int J Cancer. 2004;110:22–30. [PubMed]
  • Wang SS, Lewcock JW, Feinstein P, Mombaerts P, Reed RR. Genetic disruptions of O/E2 and O/E3 genes reveal involvement in olfactory receptor neuron projection. Development. 2004;131:1377–1388. [PubMed]
  • Prasad BC, Ye B, Zackhary R, Schrader K, Seydoux G, et al. unc-3, a gene required for axonal guidance in Caenorhabditis elegans, encodes a member of the O/E family of transcription factors. Development. 1998;125:1561–1568. [PubMed]
  • Pozzoli O, Bosetti A, Croci L, Consalez GG, Vetter ML. Xebf3 is a regulator of neuronal differentiation during primary neurogenesis in Xenopus. Dev Biol. 2001;233:495–512. [PubMed]
  • Croci L, Chung SH, Masserdotti G, Gianola S, Bizzoca A, et al. A key role for the HLH transcription factor EBF2COE2,O/E-3 in Purkinje neuron migration and cerebellar cortical topography. Development. 2006;133:2719–2729. [PubMed]
  • Bossing T, Udolph G, Doe CQ, Technau GM. The embryonic central nervous system lineages of Drosophila melanogaster. I. Neuroblast lineages derived from the ventral half of the neuroectoderm. Dev Biol. 1996;179:41–64. [PubMed]
  • Doe CQ. Molecular markers for identified neuroblasts and ganglion mother cells in the Drosophila central nervous system. Development. 1992;116:855–863. [PubMed]
  • Prokop A, Technau GM. The origin of postembryonic neuroblasts in the ventral nerve cord of Drosophila melanogaster. Development. 1991;111:79–88. [PubMed]
  • Schmid A, Chiba A, Doe CQ. Clonal analysis of Drosophila embryonic neuroblasts: Neural cell types, axon projections and muscle targets. Development. 1999;126:4653–4689. [PubMed]
  • Schmidt H, Rickert C, Bossing T, Vef O, Urban J, et al. The embryonic central nervous system lineages of Drosophila melanogaster. II. Neuroblast lineages derived from the dorsal part of the neuroectoderm. Dev Biol. 1997;189:186–204. [PubMed]
  • DiAntonio A, Haghighi AP, Portman SL, Lee JD, Amaranto AM, et al. Ubiquitination-dependent mechanisms regulate synaptic growth and function. Nature. 2001;412:449–452. [PubMed]
  • Vervoort M, Crozatier M, Valle D, Vincent A. The COE transcription factor Collier is a mediator of short-range Hedgehog-induced patterning of the Drosophila wing. Curr Biol. 1999;9:632–639. [PubMed]
  • Shen-Orr SS, Milo R, Mangan S, Alon U. Network motifs in the transcriptional regulation network of Escherichia coli. Nat Genet. 2002;31:64–68. [PubMed]
  • Milo R, Shen-Orr S, Itzkovitz S, Kashtan N, Chklovskii D, et al. Network motifs: Simple building blocks of complex networks. Science. 2002;298:824–827. [PubMed]
  • Johnston RJ, Jr., Copeland JW, Fasnacht M, Etchberger JF, Liu J, et al. An unusual Zn-finger/FH2 domain protein controls a left/right asymmetric neuronal fate decision in C. elegans. Development. 2006;133:3317–3328. [PubMed]
  • Mangan S, Zaslaver A, Alon U. The coherent feedforward loop serves as a sign-sensitive delay element in transcription networks. J Mol Biol. 2003;334:197–204. [PubMed]
  • Mangan S, Alon U. Structure and function of the feed-forward loop network motif. Proc Natl Acad Sci U S A. 2003;100:11980–11985. [PMC free article] [PubMed]
  • Mardon G, Solomon NM, Rubin GM. dachshund encodes a nuclear protein required for normal eye and leg development in Drosophila. Development. 1994;120:3473–3486. [PubMed]
  • Shen W, Mardon G. Ectopic eye development in Drosophila induced by directed dachshund expression. Development. 1997;124:45–52. [PubMed]
  • Pignoni F, Hu B, Zavitz KH, Xiao J, Garrity PA, et al. The eye-specification proteins So and Eya form a complex and regulate multiple steps in Drosophila eye development. Cell. 1997;91:881–891. [PubMed]
  • Bonini NM, Leiserson WM, Benzer S. Multiple roles of the eyes absent gene in Drosophila. Dev Biol. 1998;196:42–57. [PubMed]
  • Callahan CA, Yoshikawa S, Thomas JB. Tracing axons. Curr Opin Neurobiol. 1998;8:582–586. [PubMed]
  • De Graeve F, Jagla T, Daponte JP, Rickert C, Dastugue B, et al. The ladybird homeobox genes are essential for the specification of a subpopulation of neural cells. Dev Biol. 2004;270:122–134. [PubMed]
  • Crozatier M, Valle D, Dubois L, Ibnsouda S, Vincent A. Head versus trunk patterning in the Drosophila embryo; collier requirement for formation of the intercalary segment. Development. 1999;126:4385–4394. [PubMed]
  • Bhat KM. The patched signaling pathway mediates repression of gooseberry allowing neuroblast specification by wingless during Drosophila neurogenesis. Development. 1996;122:2921–2932. [PubMed]
  • Duman-Scheel M, Li X, Orlov I, Noll M, Patel NH. Genetic separation of the neural and cuticular patterning functions of gooseberry. Development. 1997;124:2855–2865. [PubMed]
  • Lee YS, Carthew RW. Making a better RNAi vector for Drosophila: use of intron spacers. Methods. 2003;30:322–329. [PubMed]

Articles from PLoS Biology are provided here courtesy of Public Library of Science
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Links

Recent Activity

    Your browsing activity is empty.

    Activity recording is turned off.

    Turn recording back on

    See more...