The nontranscribed spacer varies widely in length between and (sometimes) within species. In yeast there is a short nontranscribed spacer, relatively constant in length. In D. melanogaster, there is almost a twofold variation in the length of the nontranscribed spacer between different copies of the repeating unit. A similar situation is seen in X. laevis. In each of these cases, all of the repeating units are present as a single tandem cluster on one particular chromosome. (In the example of D. melanogaster, this happens to be the sex chromosome. The cluster on the X chromosome is larger than that on the Y chromosome, so female flies have more copies of the rRNA genes than male flies.)
In mammals the repeating unit is very much larger, comprising the transcription unit of ~13 kb and a nontranscribed spacer of ~30 kb. Usually, the genes lie in several dispersed clusters?in the case of man and mouse residing on five and six chromosomes, respectively. One interesting (but unanswered) question is how the corrective mechanisms that presumably function within a single cluster to ensure constancy of rRNA sequence are able to work when there are several clusters.
The variation in length of the nontranscribed spacer in a single gene cluster contrasts with the conservation of sequence of the transcription unit. In spite of this variation, the sequences of longer nontranscribed spacers remain homologous with those of the shorter nontranscribed spacers. This implies that each nontranscribed spacer is internally repetitious, so that the variation in length results from changes in the number of repeats of some subunit.
The general nature of the nontranscribed spacer is illustrated by the example of X. laevis.Figure 4.17 illustrates the situation. Regions that are fixed in length alternate with regions that vary. Each of the three repetitious regions comprises a variable number of repeats of a rather short sequence. One type of repetitious region has repeats of a 97 bp sequence; the other, which occurs in two locations, has a repeating unit found in two forms, 60 bp and 81 bp long. The variation in the number of repeating units in the repetitious regions accounts for the overall variation in spacer length. The repetitious regions are separated by shorter constant sequences called Bam islands. (This description takes its name from their isolation via the use of the BamHI restriction enzyme.) From this type of organization, we see that the cluster has evolved by duplications involving the promoter region.
We need to explain the lack of variation in the expressed copies of the repeated genes. One model would suppose that there is a quantitative demand for a certain number of "good" sequences. But this would enable mutated sequences to accumulate up to a point at which their proportion of the cluster is great enough for selective pressure to be exerted. We can exclude such models because of the lack of such variation in the cluster.
The lack of variation implies the existence of selective pressure in some form that is sensitive to individual variations. One model would suppose that the entire cluster is regenerated periodically from one or from a very few members. As a practical matter any mechanism would need to involve regeneration every generation. We can exclude such models because a regenerated cluster would not show variation in the nontranscribed regions of the individual repeats.
We are left with a dilemma. Variation in the nontranscribed regions suggests that there is frequent unequal crossing over. This will change the size of the cluster, but will not otherwise change the properties of the individual repeats. So how are mutations prevented from accumulating? We see in 4.10 Crossover fixation could maintain identical repeats that continuous contraction and expansion of a cluster may provide a mechanism for homogenizing its copies.