Sensory Cell Fates: Four Defaults for the Price of One

Summary

The specification of different subtypes of olfactory sensilla, which harbor the olfactory receptor neurons (ORNs) in the Drosophila antennae, is poorly understood. Loss of the transcription factor Rotund (Rn) leads to a simultaneous mis-specification of several ORN classes, transforming them into different ‘default’ cell fates.

The presence of morphologically and molecularly distinct classes of receptor neurons with different excitation properties is a hallmark of all sensory systems, from worms to mammals. Great progress has been made in Drosophila towards understanding patterning mechanisms of sensory epithelia. In the olfactory system, much of the recent work has focused on the development of synaptic connections to second and third order neurons [1], as well as the functional characterization of the underlying circuits [2]. However, the specification of olfactory receptor neuron (ORN) fates remains incompletely understood. In this issue of Current Biology, Li et al. [3] present important insights into this process by describing how loss of the transcription factor Rotund (Rn) leads to a simultaneous mis-specification of several ORN classes into different ‘default’ cell fates.

The olfactory system of the fly comprises two sets of structures: the antennae and the maxillary palps. The antenna alone is covered with ~400 sensory sensilla exhibiting considerable morphological and molecular diversity. Patterning this olfactory epithelium is a major challenge: 50 ORN types need to be specified, and the genome encodes 80 olfactory receptor genes to choose from [4], with usually only one receptor expressed per ORN, although specific exceptions to this rule exist [5]. In the third antennal segment, between one and four ORNs that are the progeny of a single precursor are located within each olfactory sensillum, which belongs to one of three morphological types: basiconic, trichoid, and coeloconic [6]. The specification of sensilla results from two developmental programs: cell–cell contact signaling (lateral inhibition) to specify founder cells, which then recruit a cluster of secondary progenitors that undergo asymmetric cell divisions to give rise to the sensillum and its ORNs [6]. The proneural genes atonal (ato) and amos, together lozenge (lz) all encode transcription factors that define each of the three sensillum types [7–9] (Figure 1). Important progress has been made towards identifying factors regulating the final stage of ORN specification, i.e. the expression of their individual olfactory receptor gene [10,11]. However, our knowledge about what happens between these two stages of development remains incomplete: what are the factors regulating different fates within sensillum types?

Figure 1. Decision landscape of the three olfactory sensillum types.

The transcription factor Rotund (Rn) is necessary for the specification of at least seven sensillum subtypes (the Intermediate sensillum ai1 is not shown). Surprisingly, Rn acts in all three sensillum types: coeloconic sensilla defined by the expression of Atonal (Ato) in its founder cells, trichoid sensilla (low Amos+Lozenge), and basiconic sensilla (high Amos+Lozenge). Loss of Rn leads to a mis-specification of sensilla into the nearest Rn-negative fate (red arrows). Hence, Rn, in concert with other transcription factors (Engrailed, Dachshund), distinguishes discrete cell fates from their respective default state, in all three types (plain arrows). Concentric color circles symbolize the combinatorial transcription factor code in each sensillum. The situation was simplified for basiconic sensilla which require more decision steps, but factors governing early distinction of S and T&L are unknown. S, small basiconic sensilla; T&L, thin and large basiconic sensilla.

Li et al. show that mutations in the gene rotund (rn), which encodes a Krüppel-like transcription factor, lead to the mis-specification of eight sensilla, belonging to all three types, basiconic, trichoid, and coeloconic [3] (Figure 1). Lineage tracing experiments reveal that Rn is indeed expressed during development in 18 ORNs located specifically within these eight sensilla subtypes. Interestingly, in rn mutants, expression of olfactory receptor genes specific to the ORN’s located within these eight sensilla are switched to the patterns characteristic of different, rn-negative sensilla. The authors conclude that the expression of Rn is what distinguishes these sensilla from a rn-negative default. Hence, recruitment of the transcription factor Rn to a subset of developing cells induces alternative fates, thereby increasing the number of sensillum subtypes in the epithelium. In agreement with this hypothesis, the authors identify a conserved element common to the promoters of all olfactory receptor genes expressed in rn-positive ORN’s, and mutation of this element dramatically decreases expression of transgenic reporter constructs. However, Rn expression disappears at the onset of OR gene expression, and the identified element does not resemble canonical binding sites for Rn [13]. Hence, regulation of terminal cell fates by Rn is most likely indirect.

But how does Rn become specifically expressed in the ORN’s of some, but not all sensillum subtypes, thereby making them acquire a different fate? In theory, both temporal and/or spatial regulation of Rn expression seem the most likely explanation. For instance, all rn-positive cells could be born within the same time window, exposed to a transcriptional environment that favors induction of Rn, whereas younger and/or older siblings encounter another transcriptional environment, and hence induce other factors, resulting in different ORN and sensillum fates [12]. Alternatively, all rn-positive cells might be born in close proximity to each other, within the same territory, which provides a unique combination of instructive signals [14]. In a second step, the precursors of different sensillum subtypes might then migrate to intermingle within the developing antenna, thereby forming the interspersed pattern observed in adults [15]. Finally, it cannot be excluded that a stochastic mechanism directly regulating Rn expression acts as a cell fate switch in developing ORN precursors. Such a situation would be similar to the specification of ommatidial subtypes in the fly eye, where the stochastic expression of a single transcription factor (called Spineless) in a subset of R7 photoreceptor cells leads to the formation of the retinal mosaic [16]. Whatever the mechanism, the authors conclude that unknown factors must exist in addition to Rn to further segregate the rn⁺ sensillum subtypes (for instance ab7 and ab10). Hence multiple transcription factors may be recruited combinatorially in a similar manner, many of which still remain to be identified.

What is truly surprising about the data presented here is that Rn acts in all three sensilla types (basiconic, trichoid, and coeloconic). Hence, since Rn is recruited to cells with different transcriptional environments (defined by factors like Atonal, Amos, and Lozenge), loss of Rn also leads to different ‘default states’ in each of the three types. It is of course tempting to speculate as to why the same factor is recruited across all sensilla. Often during development, successful regulatory modules are re-deployed at different time points, in different tissues, sometimes with variations [17]. In this case, using the same factor would be similar to ‘convergent evolution’, i.e. with the system selecting independently in each case a mechanism that works well. Multiple similar modules might act together, providing different branch-points within the decision landscape. Alternatively, the three types of sensilla could result from duplications of an evolutionary ancestor expressing rn only in some subtypes. Such expansion could be explained by the requirement to generate greater diversity in sensillum subtypes, as reflected by the diversity found between different insect species [18]. Ants, for instance, which rely extensively on pheromone communication, harbor hundreds of ORN classes. It is likely that the addition of a few factors like Rn, that act at multiple branch points in precursors of different sensillum types, are sufficient to achieve this hugely increased diversity. Future comparative studies will paint a clearer picture of the process.

The use of one transcription factor to distinguish multiple cell fates from different default fates seems a very elegant solution to generate diversity, since it reduces the complexity of the task at hand. However, defining default states is somewhat tricky, especially when the number of known factors within the network increases. In principle, a default state is defined using a loss-of-function phenotype, as in the case of Rn: absence of a transcription factor in cell A leads to loss of fate A, and transformation into another fate B. Alternatively, the absence of an instructive signal, or of the cell from which it emanates, can have the same effect of transforming A into B. However, loss-of-function phenotypes for two different genes sometimes lead to opposite phenotypes, i.e. loss of a factor transforms A into B, but loss of another factor leads to the transformation of B into A! In this case, which fate is the default state? This is the case in the Drosophila retina for R8 photoreceptors that can express Rhodopsin 5 or Rhodopsin 6 depending on whether they express one particular factor, or another [19]. In fact, the two factors are mutually exclusive and form a bi-stable loop that regulates cell fate decisions. If only one of the factors was known, the conclusion would be that it drives the cell away from the default state, yet this would only be half the truth. What if the fate of each rn-positive neuron and that of its default state neuron were regulated by a similar loop, and we simply have not found the second gene or phenotype yet? Although this might not be the case, additional factors that regulate the other cell fate decisions within the olfactory sensillum types must be identified.

The authors also demonstrate how genetic analysis is not our only hope: they present a computer model that can recapitulate the described fate decisions, using a combinatorial code of known transcription factors. However, on our way to understanding the whole process, important variables are still missing since factors governing additional branch points remain unknown. Filling these gaps will be the exciting task of the next few years, and now that the first step is taken, we can look forward to discovering how a compact network of transcription factors can lead to such astonishing neuronal diversity, which might one day allow us to understand the much greater diversity of mammalian ORNs.