Grant emerged from the cabin frowning, carrying her notepad.
‘More news from São Paulo,’ she said. ‘They’ve refined their model.’
‘And?’
She propped the notepad on the guard rail in front of Prabir; it was displaying a graphic of a couple of large molecules bound to strands of DNA. ‘I don’t know what to make of this. I was hoping they’d find evidence that some part of the protein resembled a transcription factor, and recognised disabled promoters—’
Prabir stopped her. ‘I used to know all these terms when I was a kid, but I’m pretty hazy now. Can you—?’
Grant nodded apologetically. ‘Promoters are sequences of DNA that sit next to the coding region of a gene, which is the part that actually describes the protein. Transcription factors bind to promoters to initiate the copying of the gene into RNA, which is then used to make the protein: to “express the gene”.
‘If a gene is accidentally duplicated, mutations that build up in the promoter of one copy might eventually stop that copy from being expressed. To identify a gene that’s become inactive like that, you’d need something that was capable of binding to a damaged promoter—something roughly the same shape as a transcription factor, but a little less fussy. And then to reactivate the gene, there’d be a number of possible strategies, either working base by base to repair point mutations in the promoter, or snipping the whole thing out and splicing in an intact version.’
Prabir said, ‘OK, that all makes sense. Now what have the modellers found?’
Grant hit a button on the notepad and animated the graphic. ‘Thisbloody thing just crawls along causing havoc during DNA replication. What normally happens is: the double helix unwinds, the two strands separate, and DNA polymerase comes along and stitches together a new complementary strand for each of them, from free-floating bases. What the São Paulo protein does is slide along each single strand, cutting it up into individual bases, while splicing together a whole new strand of DNA to take its place. Then DNA polymerase comes along and duplicates that.’
Prabir took the notepad from her and slowed down the animation so he could follow the steps. ‘But what’s the relationship between the old sequence and the new sequence?’
‘Basically, the new one is the old one, plus noise. SPP changes shape as it binds to each base of the original strand—it assumes a different conformation depending on whether it’s cutting out adenine, guanine, cytosine, or thymine—and that in turn determines the base it adds to the new strand. But the correlation isn’t perfect; there are some random errors introduced.’
Prabir laughed, disbelieving. ‘So it’s just an elaborate, self-inflicted mutagen? These creatures might as well be bathing their gonads in radiation or pesticide?’
Grant replied dejectedly, ‘That’s what they’re claiming.’
He replayed the animation. ‘No. This is crazy. If you wanted to add a few extra random errors to your offspring’s DNA, would you take the easy way out and just alter your DNA polymerase slightly, so it made occasional mistakes—or would you invent a whole new system like this for making deliberately flawed single-strand copies?’
‘Well, exactly,’ Grant said. ‘And even if you had a good reason to take that approach, the whole protein’s vastly over-engineered. There are commercial enzymes that do something similar, and they’re about one-hundredth the molecular weight.’
‘Maybe there’s a bug in their software. Or maybe there’s some logic to the changes, some pattern that they simply haven’t noticed.’
Grant shrugged morosely. ‘They’ve synthesised some of the protein for real now; they’re doing test-tube experiments as we speak, to try to confirm all this.’
She seemed to be taking it all too much to heart. Prabir said, ‘You knowthat the things we’ve seen here can’t be explained by random mutations. Maybe there’s still some way this can be compatible with your theory. But whatever’s going on, at least we’re closing in on it.’
‘That’s true.’ She smiled. ‘They have synthesised protein in São Paulo, and I have cultured fruit pigeon spermatocytes. By morning they’ll know what happens in a test tube, and we’ll know what happens in a living cell.’
When Prabir woke, this prediction had been fulfilled. Grant had been up since three o’clock trying to make sense of the results.
The experiments in São Paulo had confirmed the computer modeclass="underline" fed a few hundred different test strands of DNA, the protein had chopped them up and synthesised new strands of exactly the same length, copying the original sequence but introducing random errors. Another group, in Lausanne, had repeated the work and found the same thing.
Grant had detected RNA transcripts for the São Paulo protein in the pigeon spermatocytes, which implied that the protein itself was being made in the cells; she had no direct test for it. But when she compared sequence data for cells before and after meiosis, the error rate was about a thousand times less than for the two experiments in vitro.
She said, ‘There has to be a second protein, some kind of helper molecule that modifies the whole process.’
‘So they need to look harder at the sequence data?’ Prabir suggested. ‘The gene for it must be in there somewhere.’
‘They’re looking. SPP alone is a bit like a pantograph with a whole lot of superfluous hinges. So maybe this is something that binds to it and stabilises it—not enough to produce perfect copies, but enough to allow its internal state to reflect the last few dozen bases to which it was bound.’
Prabir opened his mouth to say, Turing machine, but he stopped himself. Most processes in molecular biology had analogies in computing, but it was rarely helpful to push them too far. ‘So it could recognise a sequence of something like a promoter, even though it only binds to one base at a time?’
‘Maybe,’ Grant agreed cautiously. ‘They’ve also got hold of samples from the Ambon fruit pigeon, and they’re going to see what pure, synthetic SPP does to an entire chromosome, in the presence of nothing but a supply of individual bases.’
As they waded ashore, Prabir looked down into the warm, clear water where he’d swum with Madhusree, then across the dazzling white beach where they’d played. He wasn’t just cheating her out of a role in the study of the butterflies, he was depriving her of the chance to demystify the island, to purge it of its horrors the way he was doing for himself.
But he could never have brought her back here. He could never have undone the one good thing.
Grant wanted to collect specimens of the butterfly’s other stages, so they spent the morning doing nothing but searching appropriately succulent leaves for the spiked larvae, and the branches of the same trees for pupae. The original versions would not have been hard to spot: both had been covered in bright-orange patches, warning colours to signal their toxicity. Grant found signs of leaf damage that looked promising, but there were no culprits nearby. If the larvae had switched strategy and opted for camouflage that was as efficient as the adults’, their movements would be far too subtle for the image-processing software to detect.
They stopped and ate lunch in the middle of the forest, in a rare spot where the ground was rocky enough to keep the shrubs at bay. Prabir still didn’t feel safe sitting down until he’d sprayed a cordon of insect repellent on the ground; the ants were never content to stay inside the orchids, waiting for easy prey. He wasn’t sure why they didn’t swarm up the trees and take nestlings; maybe they were lacking some crucial adaptation for the task, or maybe it just wasn’t worth the energy.
Grant said, ‘So your whole family was here for three years, from 2010? Was your sister born on the island?’
He laughed. ‘We weren’t quite that isolated. We went to Ambon on the ferry four times a year. And we flew down to Darwin for the birth.’