The Results ends with the analysis of possible interactions between bases at different positions in the uptake signal sequence motif he derived. We motivate this analysis as a possible explanation of the discrepancy between his uptake sequence motif and the one I derived years ago for the uptake sequences in the H. influenzae genome. I'm reproducing the two motifs below and below them his figure of his interaction analysis.
The diagram is at the top of this post. The center four positions of the core (left segment) are greyed out, because their effect on uptake is so strong that we can't make confident inferences about their interactions with other positions. The black brackets above each segment indicates that all of the bracketed positions participate in interactions with all the positions in the other bracketed segments, as indicated by the blue arrows. However the positions within a single bracketed segments do not interact with each other, unlike the minor covariation interactions (figure below) we found long ago between adjacent positions in the genomic USS sequences (pdf of the paper here).
Anyway, back to the Discussion...
First we can explain how the interaction analysis nicely reinforces the hypothesis that the uptake sequences in the genome are there as a direct and unselected molecular-drive consequence of the bias of the uptake machinery. This is an 'exception that proves the rule' situation, where the initial finding that the simple uptake-bias motif didn't match the genomic USS motif created doubt about the hypothesis, and the subsequent demonstration that interactions explain the discrepancy increased our confidence in it.
Then we can say that this leaves only the uptake bias itself in need of an explanation, and that we propose that it exists as part of a solution to the mechanistic problem of getting stiff, highly charged DNA molecules through the narrow secretin pore. Because cells efficiently take up closed circular DNAs we know that uptake doesn't usually initiate at a fragment end, but must initiate internally on DNA fragments (see this very old post). We hypothesize that the uptake bias favours sequences that are readily kinked, and that this kinking occurs mainly as consequence of interactions between the uptake sequence and mutually-interacting proteins of the uptake machinery (the uptake motif is itself only slightly bent, at the T-tracts). One reason to think that proteins mediate the interactions is that adjacent positions don't interact with each other.
Perhaps we can here pose a specific model of what parts of the uptake sequence interact with what parts of the machinery... This should take into account that the T-tracts interact with the core positions but not with each other.
Finally we can discuss the known or possible uptake biases of other species. First the other Pasteurellaceae, then the Neisserias, and finally bacteria where uptake bias may have been overlooked.