In this study we have shown it is possible to apply relatively simple image analysis methods to static images of developing axons, quantify key properties relevant to the study of axon guidance and combine data from multiple embryos to make statistical comparisons between groups sharing a particular property (here their genotype). Since it is very difficult to trace individual axons in dense axonal tracts in vivo
, our approach was to determine the average positions, directions and curvatures of populations of axons. Data were combined from multiple embryos so as to allow statistical comparisons that took into account variations between individual embryos or introduced by technical factors such as the degree of fixation, the amount of tracer injected or the precise plane of section in each individual. Using this approach, we detected the previously-described defect of chiasm development in Slit1−/−Slit2−/−
]. The real value of additional objectivity was demonstrated by our detection of hitherto unreported defects in Slit2−/−
Based on their qualitative examination of mutant chiasms, Plump et al.
] stated that they were unable to detect defects at the chiasm of Slit2−/−
mutants, with the caveat that they could not exclude defects that were beyond the sensitivity of their experiments. In fact, in the example that they show of labelled axons at the chiasm of Slit2−/−
of their paper) the anteroposterior width of the tract at the midline is about double the width of the wild-type. Our quantitative results indicate that there is indeed a defect of the Slit2−/−
chiasm with many axons mislocated abnormally anteriorly. Our results agree with the conclusion of Plump et al.
] that there are no major defects at the chiasm of Slit1−/−
mutants. Regarding the comparison of Slit2−/−
embryos, we found significant misorientation of some of the posterior axons contralateral to the injected eye in double mutants, but otherwise they were similar to single mutants. Such misorientation of posterior axons was not found in Slit1−/−
mutants, indicating that there are abnormalities of double mutants that are not present in either mutant alone. Overall, however, our results indicate that, early in the formation of the chiasm, Slit2 plays a more powerful role than Slit1 in constraining the growth of axons to their correct location across the ventral midline.
While our results suggest a greater similarity between the effects of loss of Slit2 alone and loss of both Slit1 and Slit2 than was suggested by Plump et al.
], the results of the two studies are in fact not strictly comparable. Plump et al.
] reported a major difference between the Slit2−/−
genotypes at E15.5, which is two days later than our findings. Although they reported defects of the double mutants at earlier ages, including E13.5, that look very similar to those found here, they did not provide data on Slit2−/−
embryos at earlier ages. This raises the interesting possibility that the effects of losing both Slit1 and Slit2 become progressively more severe than those of losing Slit2 alone as the chiasm develops from E13.5 to E15.5.
phenotypes reported here have a striking correlation to the ectopic projection of Slit−/−
retinal axons to the contralateral eye that we reported previously. In wild-type and Slit1−/−
embryos the inter-retinal projection is relatively small indicating that Slit1 is dispensable for keeping retinal axons out of the opposite eye. In contrast, loss of Slit2
function results in a dramatic increase in the size of the inter-retinal misprojection and a further increase occurs in Slit1−/−; Slit2−/−
]. It is easy to imagine that these phenotypes are causally linked: those retinal axons that cross the midline in aberrant positions in Slit
mutants are liable to end up on a track which predisposes them to misproject to the opposite eye. This provides an example of how the novel analysis techniques described here can shed extra light on axon guidance phenotypes previously observed using more traditional techniques.
The effect of Slit2 is likely explained by its spatial pattern of expression, which was studied here by in situ hybridization due to the lack of suitable antibodies. Clearly, it would be preferable to examine the distribution of Slit proteins, and our analysis of this issue is based on the assumption that the protein distribution would approximate to the mRNA distribution at a tissue level. Slit2 mRNA is expressed in the ventral midline in a position anterior to the normal chiasm. It is straightforward to understand how loss of this expression might be a critical factor allowing axons to cross in an abnormally anterior position. The effects of Slit1 loss are intriguing and less easily explained. Slit1 is normally expressed both anterior and posterior to the point of entry of retinal axons. Its loss in combination with that of Slit2 causes some misorientation of posterior axons after they have crossed the midline, but why this defect is not detected in single Slit1−/− embryos is not clear. It appears that the presence of other factors provides sufficient guidance even in the absence of Slit1 and that Slit2 might be one of these other factors. How Slit1 and Slit2 cooperate to prevent contralateral posterior misguidance is not clear. It is possible that Slit2 prevents axons approaching the midline from acquiring abnormalities that predispose some of them to require Slit1 repulsion from posterior contralateral territory. In this scenario Slit1 repulsion would only be required if Slit2 is lost.