Don't Make Me Mad, Bub!

Abstract

The history of Bub1, a spindle checkpoint component, reveals a spectacular case of parallel evolution. In this issue of Developmental Cell, Suijkerbuijk et al. (2012) provide evidence that Bub1 has duplicated and diverged many times during eukaryotic evolution, dividing the functions of its ancestor between the two duplicated copies.

Stephen J. Gould argued that even if we could replay the tape of evolution, we would not get the same outcome (Gould, 1989). At the largest scale, he was right. But at smaller scales, similar events have occurred many times during evolutionary history, ranging from multicellularity to the invention of proteins. In this issue of Developmental Cell, Suijkerbuijk et al. (2012) present a striking example of parallel evolution from the spindle checkpoint (Murray, 2011), the circuit that keeps eukaryotic cells from segregating their chromosomes before they are correctly aligned on the spindle.

The spindle checkpoint proteins were identified in budding yeast. Later, homologs were identified in most other eukaryotes. For many components, there was one-to-one correspondence between homologs in different organisms, but there was one exceptional pair, identified in yeast as Bub1 (budding uninhibited by benzimidazole) and Mad3 (mitotic arrest deficient). In yeast, the two proteins have regions of homology to each other but different functions. Bub1 binds to the kinetochore, which assembles on centromeric DNA and attaches chromosomes to microtubules, whereas Mad3 inhibits Cdc20, the ultimate target of the spindle checkpoint (Chao et al., 2012; Lau and Murray, 2012). Cdc20 activates the anaphase-promoting complex (APC), the ubiquitinating machine that ends mitosis by triggering chromosome separation and mitotic cyclin destruction. Outside their shared regions, yeast Bub1 and Mad3 differ: Bub1 has an additional C-terminal domain containing a protein kinase, and Mad3 has a short N-terminal sequence containing a lysine(K)-glutamate(E)-asparagine(N) (KEN) box. As in other proteins, the KEN box in Mad3 targets the protein's ubiquitination and destruction (King et al., 2007).

Fission yeast, which is very distantly related to budding yeast, has a Bub1 and Mad3 pair that resembles the two budding yeast proteins, but the story grew murky in animals. In humans and Drosophila, the two related proteins are less different from each other: BUB1 is the functional homolog of yeast Bub1, but BUBR1, unlike yeast Mad3, retains a protein kinase domain.

Suijkerbuijk et al. (2012) have now shown how Mad3 in yeasts is related to BUBR1 in flies and humans. Their phylogenetic analysis shows that an ancestral Bub1 has been repeatedly duplicated and repurposed at least five times during eukaryotic evolution. Comparing species with only Bub1 to relatives with Bub1 and Mad3/BUBR1 pairs produced by duplication and divergence revealed striking parallels between these events. In unduplicated species, every Bub1 has five features from N to C terminus: a KEN box, three tetratrico (34) peptide repeats (TPRs), another KEN box, a kinetochore-binding domain, and a kinase domain. In the duplicated and diverged species, Bub1 has lost both KEN boxes but retains the other domains, whereas Mad3/BUBR1 retains the KEN boxes and a Cdc20-binding version of the TPR repeat and the kinetochore-binding domains. In fungi and nematodes, the kinase domain was lost, producing Mad3-like proteins, whereas insects and vertebrates kept this domain, giving rise to BUBR1-like proteins.

These events are most clearly seen in the ascomycete fungi, which include budding and fission yeasts (Figure 1). Bub1 was duplicated and diverged twice, once in the lineage leading to fission yeast and once on the road to brewers yeast (Saccharomyces cerevisiae). S. cerevisiae and its relatives result from a whole-genome duplication between 20 and 50 million years ago (Scannell et al., 2007). Budding yeasts that diverged earlier have a single Bub1, whereas those that evolved afterward have taken two paths, even though all descend from an ancestor with two copies of Bub1. In the branch containing S. cerevisiae, the two copies diverged, with Bub1 losing Cdc20-interacting motifs and Mad3 preserving these motifs but losing the kinase domain. In the other branch, one copy was inactivated or lost before divergence could occur, and the single Bub1 gene contains all the elements found before the genome duplication. Interestingly, fission yeast Mad3 is even more reduced, having lost the kinetochore-binding region as well as the kinase domain.

Figure 1. The History of Bub1 in Ascomycete Fungi.

This phylogenetic tree shows selected yeast species with the organization of their Bub1 and Mad3 proteins. Red branches of the tree reflect events after the whole-genome duplication: solid branches retained and diverged two copies of Bub1, and dashed branches lost one copy before divergence could occur. Species names and phylogeny are from the Yeast Gene Order Browser (http://wolfe.gen.tcd.ie/ygob/).

This story illuminates two questions in protein evolution: (1) how proteins duplicate and diverge and (2) how thermodynamically stable proteins need to be. Ohno (1970) suggested that an ancestral protein produced one duplicate that retained the original function and another that acquired a new function (neofunctionalization). Others (Force et al., 1999) suggested the reverse: an ancestral, promiscuous protein had two functions that give rise to functionally simpler duplicates, each retaining only one ancestral function (subfunctionalization). For Bub1, the evidence from Suijkerbuijk et al. (2012) favors slimming: more recently derived evolutionary branches acquired the paired Bub1 and Mad3/BUBR1 proteins, whereas the older branches have a single Bub1, and every duplicated protein pair has reciprocally lost functional domains.

What were the functions of ancestral Bub1, and why were they in conflict with each other? We can use the current function of Bub1 and Mad3 in budding yeast to speculate. Bub1 has two roles at the kinetochore: detecting the absence of microtubules and ensuring that sister chromatids attach to opposite spindle poles. In contrast, Mad3 aids another checkpoint component, Mad2, in turning off the APC. The structure of Mad2 and Mad3 bound to Cdc20 shows collaboration (Chao et al., 2012): Mad2 keeps the N terminus of Cdc20 from the APC and Mad3 inserts a KEN box and destruction (D) box (another feature of APC substrates) mimic into Cdc20 sites that normally bind substrates for ubiquitination. The ancestral Bub1 probably did both jobs, with some molecules at the kinetochore, monitoring chromosome behavior, and others cooperating with Mad2 to turn off the APC. Gene duplication allowed the two copies to specialize, one for the kinetochore and the other for the APC. Supporting this idea, Bub1 differs depending on the presence or absence of Mad3: in species with Mad3, Bub1 has lost both KEN boxes, while in those without Mad3, Bub1 retains these features.

Why was the kinase domain retained in BUBR1 and lost in Mad3? The surprising answer appears to be protein stability: in some lineages, the kinase domain is needed to prevent BUBR1 from unfolding, even though it lacks kinase activity. The kinase domain of fly and human BUBR1 is a pseudokinase: it has mutations in key conserved residues that identify eukaryotic protein kinases. Suijkerbuijk et al. (2012) could not detect protein kinase activity in BUBR1, but mutations in the kinase domain that block ATP binding weakened the spindle checkpoint by reducing BUBR1 levels. Why keep a domain that makes you susceptible to disaster? The simplest explanation is that once the functions of Bub1 and Mad3/BUBR1 were divided, Mad3/BUBR1's catalytic kinase domain became dispensable; how it decayed depended on which mutation killed it first. In the yeasts, a nonsense mutation lopped off the unwanted kinase domain, creating Mad3-like proteins. But in chordates and insects, the ancestral Bub1 acquired potentially destabilizing mutations: changes that left the full-length protein stable but would have led to the unfolding of a truncated, Mad3-like protein. The only way to prevent this fate was to retain the fold of a kinase domain, explaining why kinase activity was dispensable, even though the domain retained almost all of the defining features of a protein kinase. Evolutionary arguments suggest that many proteins will acquire destabilizing mutations until they lie on the verge of falling apart (Zeldovich et al., 2007). During evolution, it would be easy for a protein to end up with “hidden” destabilizing mutations that would make it unfold unless it possessed other, stabilizing features.

How often do proteins undergo parallel duplication and divergence? The reciprocal structural changes after Bub1 duplication make this a compelling example that should inspire systematic searches for other examples of what is likely to be a widespread phenomenon.