|Home | About | Journals | Submit | Contact Us | Français|
In multicellular organisms, cells are polarised in the plane of the epithelial sheet, demonstrated in some cell types by oriented hairs or cilia. Many of the genes involved have been identified in Drosophila and are conserved in vertebrates. Here we dissect the logic of planar cell polarity (PCP). We review studies of genetic mosaics in adult flies. Marked cells of different genotypes are confronted, the aim being to understand how polarising information is generated and how it passes from one cell to another. We argue that the prevailing opinion that planar polarity depends on a single genetic pathway is wrong and conclude there are (at least) two independently acting processes. This conclusion has major consequences for the PCP field.
Animals are largely built of epithelia whose cells are specified by both scalars and vectors. The scalars are in the form of positional information that tells cells where they are located on the axes of the body; they use this information to decide where to differentiate, one from another, in order to build a pattern. But, to construct a part of London, or to find one’s way in a desert, plans or maps are not enough; one needs a compass or the sun for orientation. Likewise, to build a limb, individual cells need vectors to tell them in which direction to move, to divide as well as how to orient extensions, such as cilia, bristles or axons2-5. Multicellular organisms could not be built without vectors.
Over the last 110 years6, many embryologists have clarified the mechanisms of positional information and defined morphogens, molecules that are released from localised sources to form gradients of concentration. The local concentration of the morphogen (the scalar) tells each cell its distance from the source. By contrast, relatively few researchers have studied vectors, partly because polarity is often hidden and imperceptible. Although a latent polarity can sometimes be revealed by experiments7, a minority of cell types openly display their polarity by the orientation of hairs or cilia, a property called planar cell polarity (PCP) 8-11.
PCP is being intensively studied — but there are so many genes, experiments and contrasting models that the field is perplexing, even for insiders. Our purpose here is to reach the outsider by looking for a common logic of mechanism, rather than emphasising diverse outcomes. Because the insect integument is fundamentally a monolayer of cells that form oriented structures such as bristles, and because of 100 years of investment in its genetics, Drosophila is the best model system. However, the results from vertebrates, particularly from the molecular genetics of convergent extension, the stereocilia of the ear, mammalian hairs and the orientation of axon growth12-16 argue that the mechanisms of PCP are strongly conserved, at least between flies and vertebrates. The PCP field has become dominated by the view that planar polarity is the outcome of one genetic pathway. Our recent analysis of the adult abdomen of the fly challenges this view and here we try to explain why this is so and what the consequences are. We argue for a new way of looking at PCP.
A vector is not a simple product of some biochemical pathway (see17 for a discerning definition of polarity), it must be seen in an anatomical context — what matters is where a bristle points with respect to the body axis, for example towards or away from the head. During development, localised determinants and oriented morphogen gradients determine the scale and orientation of body axes and PCP appears to be set up as a downstream consequence. For example, in the fly wing, a clone of cells overexpressing only the morphogen Dpp makes a new peak in the concentration gradient and this induces a perfect winglet of the appropriate size, pattern and PCP18. Mutations in the signalling molecules Wnt11 and Wnt 5 affect the orientation of cell movements in the zebra fish19, 20. Even though these molecules can produce changes in PCP, the experiments do not establish that Dpp and/or the Wnts are components of the PCP machinery itself, a fact that is often forgotten. To understand that machinery one needs to define its components and work out what they do; the history of developmental biology argues that the best way to do this is via genetics21, 22, the “master science of biology”1.
Most of the genes so far known to act in the mechanism of PCP were identified in Drosophila and they fall into two groups: Mutations in genes of the first group not only change polarity but also alter the shapes of wings and legs and disturb growth. We limit discussion to fat (ft), dachsous (ds) and four-jointed (fj). The second group includes mutations that disturb cell polarity but have little if any effect on pattern. To simplify we discuss only genes that are central to the process: these are dishevelled (dsh), frizzled (fz) (Fig 1), prickle (pk), Van Gogh/strabismus (Vang/stbm) and starry night/flamingo (stan/fmi) (Table 1).
How do these genes organise PCP? To help answer this question, we introduce the fly abdomen into the current picture of PCP (for this picture, see10). We rely on a functional assay that springs from the finding that clones of mutant cells alter the polarity of wildtype cells nearby23, 24. The beauty of this assay is that the clone and its surround can be given different genotypes by the experimenter and, in the best systems, the polarity and genotype can be monitored cell by cell. We call the cells of the clone sending cells because, for simplicity and for this review, we focus on the information that is being passed from the clone to the receiving cells surrounding it (of course, information may also go in the opposite direction). Now, take a small clone of cells lacking the fz gene: in both abdomen25 and wing24, the clone reverses the polarity of some nearby wildtype receiving cells so that all cells point inwards (Fig 1). Clones overexpressing fz reverse polarity of some receiving cells so that all cells point outwards, away from the clone (Fig 2a). It follows that information from sending cells makes receiving cells turn to point their hairs towards cells with a lower level of Fz and away from those with a higher level26; typically, this effect spreads several cells into the surround.
In the current literature there is a consensus that the main genes (Table 1) act in a single pathway to build PCP e.g.10, 27-29. An upper tier of proteins, encoded by the Ds, Ft and Fj genes (which we call the “Ds system”) is thought to polarise and regulate the activity of a lower tier, consisting of the Fz receptor and associated proteins such as Vang and Stan (which we call the “Stan system”). The lower tier is then thought to interact with executive proteins involved in making the polarised structures30 (such as actin). This single-pathway hypothesis is not fact but has been reiterated so often that it is becoming perceived and presented as such e.g.31. We now offer four pieces of evidence that it is incorrect, at the very least in the abdomen where we have done our experiments:
First, the most persuasive piece of evidence: in the functional assay, excess Ds, Ft or Fj in the sending cells can repolarise receiving cells even when all the cells, sending and receiving, lack Fz, or Stan, or both32 (Figs (Figs2b,2b, ,3).3). Thus the genes of the Ds system can drive PCP in the complete absence of the Stan system.
Second, if there were two independent systems, blocking either should have a weaker effect than blocking both. In fact, in stan− or fz− flies (in which the Stan system is broken) hair polarity in the abdomen is only slightly disturbed. Similarly, when ds is removed (to break the Ds system), polarity is, again, little damaged. Yet, if both systems are broken at once (ds− stan− flies) the orientation of both hairs and bristles is mostly randomised32 (Fig 4).
Third, ds− cells provide a sensitised assay for activity of the Stan system: in the absence of Ds, clones that lack or overexpress Fz repolarise receiving cells over a longer range than similar clones in the wildtype; also sending cells with modest alterations in the level of Fz that would normally have no visible consequence, now change the polarity of ds− (or ft−) receiving cells28, 32, 33. Therefore, if raising the level of Ds (or Ft) in the sending cells were to alter Fz activity in those cells, as the single pathway model might predict, then ds− (or ft−) receiving cells should be very responsive. In fact, sending cells that express Ds (or Ft) have no effect at all on the polarity of ds− (or ft−) receiving cells32. It seems that neither Ds nor Ft affect the Stan system of the sending cells.
Fourth, when manipulated in clones, the two systems are fundamentally different, they can even have effects of opposite sign. Assays deploying the Ds system (for example, sending cells that overexpress ft) behave differently in the two compartments of the abdominal segment; reversing the polarity of receiving cells in front of the clone in the anterior compartments and behind in the posterior compartments34. By contrast, clones affecting the Stan system behave in the same way in the two compartments — clones lacking fz always reverse the polarity of cells behind the clones25.
Some of these arguments are more persuasive than others, but together they make a strong case that the Ds and Stan systems are separate pathways contributing to PCP by different mechanisms34.
What is the counterevidence? What are the results supporting the single pathway model? One argument came from the eye, where it was claimed that ft− cells can bias the polarity of ommatidia in wild type but not in fz− flies29 — however the sample size was insufficient to draw this conclusion and, also, appropriate controls were not provided (ft+ clones in fz− eyes). But, in the wing, it has been stated that ft− clones do not repolarise cells in fz− flies28 and also, in fz− abdominal pleura, the hairs are randomised and do not respond to clones overexpressing an active form of Ds32. Put with the contrary results on the dorsal abdomen, these findings might suggest that Ds acts through Fz as part of a single pathway in some organs but not in others. However we judge this unlikely, mainly because fundamental processes are normally conserved and used again and again, not only in different organs of one species but also between species. We prefer a simpler explanation, an example of a “don’t worry hypothesis”1: perhaps the PCP of eyes, wings and pleura of fz− flies is too disturbed for cells to be able to respond to the Ds signal? An explanation that fits, because, in fz− flies, the eyes, wings and pleura are much more depolarised than the tergites. Under this particular “don’t worry hypothesis”, the Ds signal would be trying to impose a polarity on cells in disarray; it might be like looking for ripples caused by throwing a stone into a rough sea.
There is another way of regarding this central issue: our experiments on the abdomen show that the Ds system has an inherent capacity to change polarity without the Stan system. Thus, in different organs, even if the Ds and Stan systems make contributions of different weight and in different ways, we would argue that both systems must have independent (and probably qualitatively distinct) inputs into the cell biology of what we all pay attention to — the orientation of hairs and other indicators of PCP.
If we accept there are two pathways, two new questions stand up and shout for answers: First, how does the Ds system polarise cells? Second, what polarises the Stan system?
In a field of cells in the wildtype, everyone agrees that there needs to be a biasing input to orient the Ds system and, most likely, this is done by morphogen gradients driving ds and/or fj transcription. Gradients of both Ds and Fj have been inferred and/or seen in the eye, abdomen and, perhaps, in the wing29, 34-37 and the orientation of both gradients shown to influence PCP38. Simon’s work in the eye shows nicely that the two gradients are redundant; flattening of one is insufficient to disturb PCP, while flattening of both produces randomisation of polarity. Reversing the Fj gradient can even turn the ommatidial polarity around38. The functional assay in the abdomen suggests that Fj acts mainly on Ft (Fig 2c) and therefore, both in the wildtype and in Simon’s experiments, we imagine that the Fj gradient generates a Ft gradient of activity. Then the mutually opposing gradients of Ft and Ds must orient individual cells — but how?
Previously, this last question was addressed by asking how Ft might feed into Fz, which we now think is the wrong question. Instead, we have shown that the cadherin-family proteins, Ds and Ft can polarise cells without Fz, so the right question is: how do they do it on their own? Important experiments28, 39, 40, using antibodies against the two proteins, suggested that in vivo, Ds and Ft make trans-heterodimers that form bridges from one cell to another (Fig 5e,f). Also, in vitro, Ds and Ft stabilise each other across intercellular boundaries and promote adhesion between cells41, 42. These papers suggest that Ds-Ft heterodimers are agents in PCP and we have therefore built a speculative model that employs Ds-Ft heterodimers and is based on the functional assays (Fig 6). When applied to the Ds system, these assays show that, in order to change polarity of the receiving cells, either Ds or Ft is sufficient in the sending cells, but both proteins are essential in the receiving cells (Fig 2d,e). The findings are clear and simple but the interpretation is not.
To build a model (Fig 6) we imagined that, in any cell, the numbers of Ds molecules that are engaged in transheterodimers (with Ft molecules in adjacent cells) might differ between the anterior and posterior faces; an intracellular asymmetry that could orient the cell. If so, an altered ratio of Ds and Ft in the sending cell could affect the number or distribution of Ds-Ft trans-heterodimers in the receiving cells. Thus, when a sending cell contains, say, excess Ft, these Ft molecules draw Ds molecules to the adjoining proximal face of the nearest receiving cell to form trans-heterodimers. Ds then becomes redistributed within the receiving cell, taking molecules away from its distal face. That face would have relatively more Ft and this would, in turn, draw excess Ds to the facing membrane of the next cell, thereby propagating the original signal. Note that the sending cell need only contain too much of either Ds or Ft, but Ds and Ft are both needed in the receiving cell, for polarisation of the first receiving cell and for propagation to the next32 (Fig 6).
There are many ideas about how Fz might mediate PCP11. To choose between these, we again rely on the functional assay in the abdomen. Looking at the main genes implicated in the Stan system (Table 1), we ask are they needed for function in the sending cell, the receiving cell or in both?
From this data we produced a model (Fig 6) in which there is a gradient of Fz activity across the field and the cells interact so that the level of Fz activity of any cell becomes modified towards an average of the levels of its neighbouring cells. To become polarised, a cell then compares the levels of Fz activity in neighbouring cells, using Stan, and points its hair towards the neighbour with the lowest value. This mathematical model is built with Fz, Vang and Stan25.
“One can do things in a very sophisticated mathematical way… but there is a difference between theories being correct and theories being true. Many theoreticians don’t make that distinction, and, even though many theories are correct in the logical sense, they are untrue because they don’t relate to the natural thing we’re all interested in”1.
The currently most popular model of PCP, the “Tree/Amonlirdviman model”45, 46, is based largely on a different set of data. The direction of the field was abruptly diverted when it was discovered that some PCP proteins are distributed asymmetrically in wing cells, at least during a short period prior to formation of the cell hairs43, 47-49. For example, Stan accumulates on both the proximal and distal faces of cells (Fig 5a), Fz and Dsh accumulate on the distal membranes47, 48 (Fig 5c) while Vang and Pk accumulate on the proximal membranes46, 50. Using these facts, some assumptions, and, later, a mathematical simulation requiring optimisation of several parameters, the Tree/Amonlirdviman model was built to explain how localised protein interactions within and between cells might drive PCP45, 46 see Fig 6).
However, the functional assays raise serious objections to this model:
First, Pk is a central component of the Tree/Amonlirdviman model where it acts, with Vang, in an amplification step to localise Fz on one side of the cell. Moreover, cells lacking Pk lose the assymmetric localisation of Vang, Fz and Stan44, 46, 50 (Fig 5d), assymmetries that are essential to the model and were used to build it. Yet, pk− cells can receive and propagate the Fz-dependent signal as well as, or better than, wildtype cells25, 44, 51. These results argue that the assymmetric accumulation of proteins is not the primary engine by which a cell acquires polarity to accord with its neighbours. Instead it might be an outcome of the intercellular gradient of Fz activity, acting to stabilise nascent polarity. This conjecture fits with repolarisation occurring more freely in pk− cells.
Second, Stan is a key protein in PCP; the functional assays show Stan to be essential in both sending and receiving cells. Stan is also required for Fz to accumulate normally on the membrane; in its absence Fz is seen mainly in the cytoplasm. (Fig 5b)48. Yet the Tree/Amonlirdviman model ignores it.
Third, in the Tree/Amonlirdviman model, the polarity of a cell depends on and incorporates the asymmetric distribution of Fz within that cell. The model therefore might have difficulty in explaining how a cell lacking Fz can be repolarised as we have observed25.
Note that both models depend on interactions between neighbouring cells to consolidate initial, possibly small, differences in Fz activity. Both models posit local interactions between proteins, but with different elements and outcomes. The Tree/Amonlirdviman model has Fz and Vang interacting to change their distributions at or near the membrane. The result is a sharp differential of Fz in each cell, from one surface to the opposite surface, to make an intracellular gradient that orients the cell (arrows in Fig. 5a). Our model depends on interactions via intercellular homodimers of Stan that bring the level of Fz activity in one cell towards an average of the levels of its neighbours; this process initiates and propagates changes in polarity when the sending cells and receiving cells differ sufficiently in their levels of Fz activity. In the wildtype epithelium, we imagine a shallow intercellular gradient of Fz activity, with only small incremental differences in the scalar levels from one cell to the next, detected via the Stan bridges and polarising each cell.
The functional assays also argue that the Ds and Stan systems operate in logically distinct ways: In the Stan system, information about the level of Fz activity is conveyed by means of the Stan bridges, so that Fz in one cell behaves like a “ligand”, sending a message to Vang in the neighbouring cell, which acts like a “receptor”. However the Ds system acts through a two-way interaction between Ds and Ft, each one acting as both a ligand and as a receptor.
The various models may illustrate how the cells interact by means of the Stan system but they do not tell us how the Stan system becomes oriented in situ; there needs to be some input, aligned with the body axis, that would feed into Fz activity and orient PCP. In the past, the consensus was that the Ds system provides that input10, 17, 25, 28, 29. Indeed, Axelrod, Simon and colleagues28, 38, 45 believe that the Stan system is oriented by vectors “imposed through the agency” of the pervasive gradients of Fj and/or Ds45 (Fig 6), a view we judge to be unsupported. Moreover, evidence against this view is given in the points of argument against a single pathway presented earlier. Instead of an effect via the Ds system, we suggest that the morphogen gradients affect Fz activity more directly; in the abdomen there is even evidence suggesting that Hedgehog might act on Fz via the receptor protein Patched32. If this were true, Hedgehog would have at least two inputs into PCP, one via its effects on the transcription of both Fj and Ds and a separate one, via Fz.
If our views are correct and generalisable there are far-reaching consequences for the PCP field. Obviously, the question of one or two pathways is central and needs further tests in different organisms. Unfortunately, partly because we fly people have placed so much emphasis on the Stan system, particularly on Fz, little work has been done on the Ds system in vertebrates. For example, there appear to be four Fat genes in mammals, of which fat-j is the closest homologue of the Drosophila ft52. There are two homologues of ds but little is known of their functions and whether they are involved in PCP. If, in vertebrates, both systems were broken, would the PCP phenotypes in the stereocilia, in hair orientation and in convergent extension, be stronger? Another big question: if the two inputs from the two systems affect cell polarity independently, as we argue, then how are they integrated in the cell to fix the orientation of structures? And another: flies lacking both the Ds and Stan systems develop well and almost emerge as adults from the pupal case. They even have some residual and consistent polarity in hairs and bristles, suggesting that there are yet other inputs into vectors and into PCP.
The excessive growth shown by ft− clones has suggested that PCP and the regulation of cell division might be linked. There certainly needs to be feedback from a growing organ to tell all cells when the final size has been reached to stop mitosis. In each axis, this feedback should depend on the dimension of the organ in that axis. But how could dimension be encoded and transmitted to single cells? Scientists investigating the control of size have evidence that morphogen gradients are instrumental. But morphogens are generated from localised sources and spread out in decreasing concentration; it is not easy to see how they could directly control a pattern of growth that, typically, is evenly distributed over the tissue. However, as we have seen, morphogens do establish and orient the Ds system. That system might therefore translate the uneven slope of a morphogen gradient into an even and possibly linear gradient, providing a constant differential between the faces of each cell (or between neighbouring cells). If so, a cell could get a measure of dimension (in the relevant axis) by comparing the difference in the scalar (perhaps the number of Ds-Ft heterodimers) across an individual cell or between cells. In this way, PCP gradients could encode information of dimension that would tell the cells when to stop dividing53, 54. If these speculations were even partly true they would have many repercussions. For example, they could focus attention on how morphogens affect growth via the machinery of the Ds system, perhaps through the action of Ft in the Hippo pathway55-58— and thus help us find out why hippopotami are so short in stature and so broad in girth.
The models shown in Fig 6 are molecular and demand molecular tests. For the Stan system it is important to know how and to monitor when Fz activity is distributed across the cell and across the tissue (Fig. 6). Resolving this may require molecular probes to assay Fz PCP activity, as distinct from Fz protein accumulation or its involvement in transducing Wingless. Also, we need to know more about Ft and Ds, especially their interactions with each other and how their activities depend on Fj. The structures of Ft and Ds, massive proteins with many domains need further analysis. Too little is known about their routes through the cell, their distributions on and off the plasma membrane and their binding partners inside and outside the cell. We do not understand how differences in the distribution of Ds/Ft heterodimers could orient cells and point the outgrowing hairs. There is much to do.
In tribute to Sydney Brenner who was 80 on January 13th, 2007
At that time (ca 1964) “we tried to decompose the complexity of higher organisms into a set of subsidiary problems….. There’d be problems of how cells move… There’d be problems of how cells grow. There’d be problems of the polarity of the cells. Which in my mind is still the essential problem; in the sense that cells move in one direction and not in another, grow in one direction, or face the world from one side of themselves and not the other. How was all this polarity established?1.