Raising genes from the Last Common Ancestor demonstrates its complexity

I do not why but papers about evolution always fascinate me (if they are not overwhelmed with population genetics and other hard core stats and math). This time I cam across a paper in JACS where authors used a computational technique called Ancestral sequence reconstruction (ASR) to rebuilt  ancient enzymes that nowadays form closely related bi-enzymatic complex. Basically, knowing the phylogeny (map of how species relates to each other) let you estimate the probability (maximum likelihood) that some mutation would occur back in its evolutionary history or vice versa: you can estimate the probability that current phylogeny would evolve given that ancestral protein sequence. Exactingly, you can go as back as you can (means - to the very Last Common organism Ancestor that we all evolved from, LUCA). In order to get that far the wider phylogeny tree you have the better: say from the most distant archaea to the bacterias. Authors asked a question whether LUCA back than 3.5 billion years ago had elaborate enzymatic networks. For that reason they rebuilt bieznyme complex (cyclase subunit HisF and the glutaminase subunit HisH) of imidazole glycerol phosphate synthase. Note that, without one another subunit would not work due to their close relationship in the synthesis reaction. They were able to show that the reconstructed proteins still retain almost the same specific activity and surprisingly able to tightly associate with each other. This work demonstrates that even LUCA had enzymes that were very closely associated and were able to perform such things as substrate tunneling (when product of one reaction is directly passed to another enzyme) or allosteric regulation (when product of one reaction regulates another enzyme). I wonder how would ribosome would look like given its high evolution conservativety.

The reason I decided to make a post about this paper is that ASR technique nowadays on the realm of very cheap gene synthesis let us 'play' with protein sequences such that we can go back in evolution and make probably more promiscuous enzyme that we can more easily 'teach' to perform reaction we want as these. Alternatively, we can improve folding properties of our protein by more targeted mutagenesis (since we have a good guess about its evolutionary history). Also, we might be able to produce an orthogonal protein/protein networks that still retains the specific activity while being not regulated by intracellular proteins. Any other ideas?