Karsentj: You mentioned that CCD patients have a mild short stature. We know that in both mice and humans, Cbfa1 expression and function is extremely superimposable. You showed that the effect on chondrocytes is mostly on the humerus. How can this explain the short stature phenotype?
Mundlos: It doesn't.
Karsentj: We showed that in Cbfa1 mice overexpressing a dominant-negative form of Cbfa1 after birth there is a normal number of osteoblasts, although there is less bone, as you showed in the patients. Do you have any histology or any other indication that CCD patients have a normal number of osteoblasts?
Mundlos: We have no histology. To me, that patient is somewhat in between the 100% and 50% gene loss. It looks as if he has 70—80% gene loss. This would correlate with your knockout mice. Perhaps there is some sort of threshold beyond that, and then there is an effect on the number of osteoblasts. This patient was born with an osteoporotic skeleton and has maintained this. He probably just started off with a lower number of osteoblasts.
Bard: You didn't touch on what I thought was the most extraordinary feature of the human phenotype — the increase in number of teeth. And yet the gene seemed work through a completely different system (exoskeleton, neural crest cells). Does the mouse model also have abnormal numbers of teeth?
Mundlos: No, but that can quite easily be explained, because mice have only one set of teeth. In humans, it's only the secondary teeth — the permanent teeth — that are affected. Why this happens remains a mystery.
Meikle: Most of those supernumerary teeth are premolars. In mice the whole premolar and canine series has been silenced.
Karsenty: One hypothesis could be that bone tissue prevents tooth formation, and by having less bone you have a permissive effect on tooth formation. This is a great hypothesis because it is not testable!
OrnitIs there any effect on fibroblast growth factor (FGF) receptor 3 expression?
Mundlos: I don't know. We looked at proliferation, and that does not seem to be the reason why they are short. In that sense it is likely that FGF receptor 3 will not be affected.
Wilkins: In genetics courses one is told that most loss-of-function mutations are recessive. It is my impression that many transcription factor loci are haplo-insufficient. Is there a compilation of data on this? Has anyone actually looked at the databases to see if this is a general property of transcription factor loci?
Mundlos: That is probably true. Interestingly, in mice this is not a general assumption. Mice frequently don't have a phenotype in the heterozygote, and you need to have the homozygous to see an effect. In this respect, Cbfa1 is really quite an exception.
D. Cohn: Certainly for many structural proteins, haploinsufficiency in humans produces a phenotype.
Wilkins: I wasn't saying it would be diagnostic for or true only of transcription factor genes. But quite often it seems to be true for these genes; in Drosophila too, it's not just humans.
Mundlos: There seems to be a very important dosage effect.
Morriss-Kay: Why do you think the clavicle is singled out to be affected in the heterozygotes? Is it anything to do with the peculiar way in which the clavicle is made, being both a membrane bone and an endochondral bone, or is there some genetic precedent?
Mundlos: It is difficult to say. We have looked at the clavicles in these mice. It seems to be a bone which, from the developmental point of view, is very different to all the others. The precursor cells are unlike osteoblasts or chondrogenic precursor cells, and express either type II or type I collagen. Also, a condensation forms, but the precursor cells do not differentiate.
Morriss-Kay: The other special feature of the clavicle is its evolutionary origin from the dermal skeleton, which was added secondarily to the endoskeletal pectoral girdle (Goodrich 1958).
Hall: Therefore it has closer links to the teeth than it does to the endoskeleton, which is interesting in terms of the supernumerary teeth phenotype.
The cartilage on the mouse clavicle is secondary: it arises from the periosteum after bone formation. If you are losing the membrane bone first, but still getting the cartilage, is it coming from the periosteum?
Mundlos: It's not really secondary. In early condensations, there are large, peculiar precursor cells that build up in the middle, and around these cells, there are condensed cells. Some of these precursor cells express type I collagen, so they are more in the osteoblastic lineage, whereas the others express type II collagen, so they are more in the chondrogenic lineage. Then, at the same time there is bone formation and cartilage formation, and at the border between the two they form a growth plate.
Hall: There must be big differences between the human clavicle and the mouse clavicle.
Mundlos: A little bit of a compact bone is built at a later stage.
Kingsley: Cbfal is not specific to osteoblasts. In embryonic development it shows up earlier in both chondrogenic and osteoblast precursors. Is this the only known example of a phenotype in the cartilage-based element? Most of the phenotypes that are usually described are the dramatic loss of the alizarin red-stained bone throughout the skeleton.
Kronenberg: There are a lot of markers of differentiation that are missing, decreased or delayed in the chondrocytes. The chondrocytes do not appear to be normal. Nobody knows whether that's because the precursors which have reasonably high level Cbfal expression have a delayed effect, or whether it's the low level of Cbfal in those chondrocytes that none the less matters.
Mundlos: There is a cartilage phenotype in the heterozygotes, in the patients, because there are all these skeletal abnormalities on the X-ray. There is an abnormal metaphysis and epiphysis, and there is a small pelvis: this is all due to the abnormal growth of cartilage.
Wilkie: Have you considered the possibility that your patient with the CBFA1 mutation and atypical, severe phenotype could have a second mutation somewhere else? In the mouse, how much does the genetic background affect either the phenotype or the detailed timing and location of gene expression?
Mundlos: We have not looked at the timed expression. We have crossed our mouse in a few different backgrounds, and there is absolutely no difference. The phenotype is very strong and very penetrant.
Wilkie: Going back to your short-digit mouse, the obvious question is, have you looked for a sonic hedgehog (Shh) mutation in type Al brachydactyly? If it is not caused by an Shh mutation, do you think that the short-digit phenotype could be attributable to the deletion that you're seeing near to Shh? Is this deleting another gene as well as causing a position effect on Shh?
Mundlos: We have tested a family for linkage to the syntenic region, and it does not map there. If there was a trans effect, meaning there is a gene that maps close to it and which is either deleted or mutated or whatever, then the complementation study would have not worked. That is, if you were to cross them, then the double heterozygotes should not have the phenotype, but since they have exactly the same phenotype it is highly likely that we are actually talking about the same gene. It must therefore be something which regulates it, unless we have missed the mutation of the gene, which I hope we haven't.
Wilkie: How would you explain why a deletion in cis to Shh is associated with a more severe phenotype in the heterozygote than knocking out the gene? Is the gene 'knockout' not a true knockout, for instance?
Mundlos: That's a difficult one. It may be also that it is a factor of the level of expression. We know from some knockouts, for instance, that they're not complete knockouts: there is linkage of some normal allele which may account for that. I don't know. It could also be that within that division you have something which is regulatory plus another gene.
Kingslej: There could also be negative regulatory elements within the deletion. In this case not only would expression be reduced, but also ectopic expression might be produced by regulatory changes at the same time.
Mundlos: There was no expression.
Kingslej: Rob KrumlauPs group recently published a paper on a mouse mutant called Sasquatch (Sharpe et al 1999). They also proposed an effect on Shh regulatory elements. But, if I remember correctly, the postulated location of those elements was way on the other side of the genes. They have a transgene insertion that gives a limb phenotype that doesn't affect Shh coding region. He proposed it was affecting sonic regulatory region, but it was distal instead of proximal. This creates an unusual situation. Your allele gave essentially all the phenotypes that sonic did, which would suggest all the regulatory elements would have to be affected in your deletion, in order to recapitulate all of the phenotypes. Yet this other paper suggests that at least some of the regulatory elements are on the other side of gene. Have you any thoughts about how to reconcile these observations?
Mundlos: I recall the paper. Do they know where it is? It is a random insertion, and they have mapped this random insertion close to Sonic hedgehog but they can't really differentiate where it is.
Kingslej: I think there was an argument that it was distal.
Wilkie: As you know, in humans there are quite a lot of 7q translocations that are associated with human holoprosencephaly that don't actually disrupt the Shh gene itself (Belloni et al 1996). These led people slightly off-track at first during the search for the disease gene. They are all on the 5' (telomeric) side of the gene. This is consistent with important regulatory elements lying upstream of the gene.
Mundlos: These translocations are also quite far away.
Wilkie: Yes, up to 450 kb according to the paper by Belloni et al (1996).
Kingslej: Your map looked like it was drawn with deletions proximal to the sonic locus. The centimorgans were labelled from the centromere.
Hall: The supernumerary teeth really are very striking. Whilst less bone may be a trigger, presumably you need factors other than just less bone to get supernumerary teeth from the permanent dentition. I assume that there is no basis normally in the permanent dentition for making additional teeth, so what else do you need to get them?
Blair: In osteogenesis imperfecta there are not supernumerary teeth but there is microdontia.
Hall: I'm presuming you need something that will continue to maintain a proliferative population at the base of these permanent teeth.
Meikle: You mentioned the importance of bone in this. We might well be looking at neural crest migration in the number of cells that are available for making teeth. In patients with hypodontia, they are missing teeth and they also have small jaws and relatively little dentoalveolar bone.
Hall: If it was such an early event, you would expect it to affect both the primary and secondary teeth. This seems to come in very late and only affects the permanent dentition.
Mundlos: The mechanisms of formation of primary and secondary teeth are quite different, though. The secondary tooth formation happens at an early stage but these are quiescent. The number of condensations that are giving rise to the primary teeth is something that the gene must affect. They are built after the primary teeth have been formed and stay quiescent.
Hall: So the gene is affecting the number of centres that are being formed.
Burger: I seem to remember that odontoblasts also express Cbfal.
"Newman: At the beginning of you paper you said that although we know the gene, we still don't understand the syndrome. Different parts of the body that one might expect to behave similarly behave differently; different individuals behave differently. One conclusion from this is that the gene may be the wrong level of description: instead, we should concentrate on processes and recognize that processes involve cooperation among many different gene products. But any given gene product won't necessarily be doing the exact same thing each time that process is occurring.
Mundlos: That is right. We see quite frequently that we have similar phenotypes, when we have effects in the same pathways. For instance if you have defects in somite differentiation, then there's a whole bunch of genes that regulate that. The phenotype will always be similar in that they have these segmentation defects. I think that's how it has to be defined. Similar diseases can be caused by defects in quite different genes, but they are phenotypically similar because these genes act on the same pathway or target. This makes life difficult for us.
Bard: Stuart Newman's last point has touched on something that has been going through my mind as I've been listening. The process that seems the most important in bone formation, at least in limbs, is size delimitation: how does the embryo know, for example, that the humerus is the right size, for example? This isn't merely the action of one gene. There is some process which is determining size here. It happens in the digits as the condensations fragment into the phalanges. In a sense, nothing I've heard today has touched on this process of spatial determination. We have some sense of the mechanisms of segmentation size for the vertebrae, but the equivalent processes in the limb remain obscure.
"Newman: With analogy to other types of systems, if you have a container of water and you agitate it, this will cause waves of a certain size no matter whether you agitate it with a pencil or with a spoon. It is a characteristic of the system. Similarly with a string of a violin, the vibrations depend on the thickness, length and tension of the string, not how you excite it. In the end its characteristics will come through, even though they might not instantly, after the transients have subsided you get waves of a particular size.
Bard: I just don't see how we can bring size into this, even though it seems so integral to what's going on.
Newman: Don't these reaction—diffusion systems that we both study give rise to structures with characteristic sizes?
Bard: No one has looked at 3D simulations of reaction—diffusion models and they're much harder to interpret than 2D ones, because it's not clear how diffusion will affect the patterns.
Kingsley: In the old Oster et al (1988) paper that used essentially the formalism of the reaction—diffusion equations (although not in exactly that form), a combination of a local activator and a lateral inhibitor with the right constraints will generate a repeating pattern. The claim was made that the same sets of equations could be used either to branch the condensations or to segment the condensations.
Kingsley: The other point that was made in those papers is that although the equations do a good job of predicting the behaviour, the equations themselves are compatible with many different physical realizations that could either be mechanical or chemical or diffusing. Therefore I do still think there's a huge need to do the difficult embryology and gene hunting to try to put real molecules and processes on the limbs. Even if the behaviour fundamentally is based upon those properties, since the equations are compatible with different physical models, without some embryology and gene hunting we won't ever know whether any of them will match up.
Bard: The best the models can do is suggest plausibility, and then put the ball back into the court of the experimentalists. But, because it is extraordinarily difficult to devise experiments that might disprove such vaguely constructed models it is actually rather hard to show that the models are implausible. Although they are helpful metaphors, I remain dubious about them.
Kingsley: I agree.
Hall: Today we have heard that retinoic acid is distributed across the limb by a gene which is differentially distributed and which has an effect which relates to the specification of pattern. This results in digits 1,2,3 in the chick limb; I don't think it tells us anything about why digit 1 is a different size from digit 3. Clearly something like SHH or retinoic acid may be playing a role in specifying the size of those digits, as well as their polarity. This seems to be close to the sorts of things that Turing was looking for: the differential response to a single molecule.
OrnitSure, but there is also differential expression of Hox genes D12 and D13.
Mundlos: The other problem is that there is no diffusion gradient.
Hall: Is that true for retinoic acid as well? I know there's been argument about this in the literature.
Morriss-Kay: There isn't any good evidence for a retinoic acid gradient, or even for a major role of retinoic acid in limb development. All we know is that in the chick, if you put a bead in, you will get a change in gene expression which leads to extra digit formation.
Tickle: I'm not sure that's entirely true. Thaller & Eichele (1987) could detect more retinoic acid posteriorly in the chick. Malcolm Maden has looked at this again recently, and he substantiates the idea that retinoic acid is enriched posteriorly (Maden et al 1998). It depends what you mean by a 'gradient': how many points you would require for a gradient?
Morriss-Kay: The trouble is that it is not possible to distinguish between free and bound (unavailable) retinoic acid.
Hall: So what is SHH or retinoic acid doing when you put it anteriorly in the limb bud?
Morriss-Kay: Retinoic acid up-regulates other genes local to the bead. There is also up-regulation of retinoic acid receptor (RAR)^, which has a retinoic acid response element in its promoter.
Kronenberg: A recent paper shows that SHH affects FGF synthesis through a complicated cascade (Zufiiga et al 1999). SHH stimulates the production of formin in the developing limb. Formin then stimulates the production of gremlin, a bone morphogenetic protein (BMP) agonist. Blockade of BMP signalling leads to induction of FGF4 synthesis. SHH doesn't move far but triggers a cascade. The diffusion of SHH is biologically regulated in an elaborate fashion, and it doesn't go very far— no more than a few cell layers. This then kicks off other things. It seems to be a pattern for SHH and Indian hedgehog (IHH) that they stimulate the production of things that have the ability to move further and faster.
Hall: So you get your cascade by sequentially turning on genes which have quite local actions, but progressing across the system.
Newman: There seem to be two independent phenomena. Everything that causes extra digits, in addition to inducing these genes, also expands the mesoblast. There is more mesenchyme to be organized. If all you had was more mesenchyme, you might expect to get just more indeterminate digits. But since you also affect these gradients of Hox proteins and other things, you get a polarity that makes any digits that arise have a particular character rather than a general character. It therefore seems that these duplication experiments are looking at two kinds of overlapping but distinct phenomena: first, more digits because of more mesenchyme and, second, particular kinds of digits because of these gradients that are induced by the retinoids or SHH.
M. Cohn: Addition of mesenchyme alone to the limb bud will not result in additional digits. For example, when tissue from the anterior part of one limb bud is grafted to the anterior part of a host limb bud, this will result in broadening of the host limb but not in the formation of extra digits unless the grafted cells have potential polarizing activity.
Pipette: The reason it doesn't is because in that case it is regulated by the zones of cell death. However, in the talpid mutants, the zones of cell death are lost and this may be the reason why there is more mesenchyme, which then results in extra digits that lack a true identity. The mesenchyme mass is indeed critical in at least setting the number of elements, but not their identity.
Hall: Presumably therefore the mesenchyme mass has some role in setting the size of those elements?
Pipette: I have been thinking about this. Digit 2 is obviously shorter in the forelimb. If you look at the overall shape of the forelimb bud, it is striking that on the anterior side there is much less mesenchyme than in the hind limb even. Perhaps small differences in how the apical ectodermal ridge (AER) extends anteriorly and in how big the zones of necrosis on the anterior side are, dictates the amount of mesenchyme available to form cartilage elements. Maybe the zone of cell death is even bigger on the anterior side of the wing than it is on the anterior side of the leg, which then leaves less mesenchyme for wing digit 2 to form.
M. Cohn: In limb buds ofthe short-digit mutant embryos, you showed that there is no detectable expression of Hoxd11—13; however it seemed like you were looking at relatively late stages of Hox expression, when posterior Hox expression domains are moving anteriorly through the autopod. What about the early stages of Hox gene expression in the limb buds? Have you detected Hoxd11—13 in early bud or pre-bud stages?
Mundlos: We looked at Day 9, and there is no Hox expression. There is strong expression of Fgf8 in AER at this stage.
Tickle: What about Hoxd9? This is expressed in the forelimb. It would be interesting to know if that was there.
Mundlos: At Day 11 there might be some weak expression.
Tickle: Martin Cohn, were you thinking of the limbless mutant, where there is early activation of Hox genes even in the absence of Shh?
M. Cohn: I was actually thinking of the early initiation of the limb bud, which we found involves expression of Hoxd9. That expression domain originates in the pre-limb lateral plate mesoderm at forelimb levels long before Shh expression. As the forelimb bud emerges, Hoxd9 expression is maintained and more posterior HoxD genes are activated. I was leading to the question of whether your mutants have arrested the unravelling of the posterior Hox genes after limb bud initiation, or do you think that HoxD genes are never expressed in mutant limb buds?
Mundlos: I can't tell you whether there is expression at a very early stage. We tested the most posterior Hox genes and found no expression. If you lose Hox expression completely then these parts of the limb are just gone. If sonic regulates it either by maintaining it or initiating it, they're not there when they're actually needed at that later stage and this part of the skeleton is gone.
Hall: So the notion of a potential alternate pathway to sonic is something one should keep an open mind on I guess, both from the python work and from the sorts of studies that you are doing. Martin Cohn, I wasn't quite sure when you put your python tissue into the anterior part of the limb of chick, was it the python cells that were producing this sonic?
M. Cohn: From looking at sections of those limbs, it is the graft of python cells that is expressing Shh. There is also no experimental evidence from work on chick limb buds to indicate that transplanting cells with polarizing activity induces Shh in neighboring cells. In our experiments, Shh expression is induced in python cells that are transplanted under the chick apical ectodermal ridge, which is a source of
Ferrin-Schmitt: I would like to propose a model. We are working on M-twist. In twist null heterozygous mice, there is a duplicated hallux (big toe), with three phalanges instead of two, all the other digits are normal. This is on a C57/Black 6 background. Thus we tried to build a model to understand what would happen at the gene expression level. When we look at M-twist expression by in situ hybridizations, we see a gradient in the forming limb bud. When the mesoderm is not differentiated, there is a high expression of M-twist, but as soon as mesoderm begins to differentiate, M-twist expression vanishes. At least at one stage there is a gradient with a higher M-twist expression in the posterior part of the limb, and later a higher expression in what will be the future anterior part. We are interested in what happens with the model of FGF expression, and have heard today that FGF and FGF receptor (FGFR) expression allows the limb bud to grow.
I would propose that the proximal/distal axis would be established through the FGF/FGFR signalling plus the Hoxa9-Hoxa13 expression gradient; then the anterior/posterior gradient would be controlled by the zone of polarizing activity/SHH/retinoic acid signalling plus the Hoxd9-Hoxd13 expression gradient.
We wonder whether the gradient of Hox gene expression is modified in twist null heterozygous mice so that instead of the correct differentiation of the hallux in the heterozygous mice, there is another type of differentiation at the opposite side from the Shh expression region (the M-twist gradient might also be affected). So we tried to test this hypothesis, using Hox mutants with a reduced number of digits, namely Hoxa13~/~: if the gradient of M-twist expression is important for the differentiation of the anterior digits, we should be able to rescue this phenotype in double mutants by crossing twist null heterozygotes (carrying an additional hallux) with the Hoxa strain lacking the anterior digit.
Hall: But you don't have the double heterozygotes yet?
Perrin-Schmitt: We are constructing them now.
Hall: This nicely emphasizes that a cell which is sitting in a particular position is encountering a number of different gradients coming from different directions.
Was this article helpful?