Sequence alignment and predicted structures of ϕC31, ϕBT1, and TG1 integrases and their recombination directionality factors (RDFs)
Sequence alignments (Figs. 2 and 3) show that the three RDFs are more closely related than the corresponding RDF-binding domains on the integrases. TG1 integrase is only 24–25% identical to ϕBT1 and ϕC31 integrases, which are about 43% identical to one another. Similarly, among the RDFs, TG1 RDF is the outlier, with 61–63% identity to the other two, which are 85% identical to one another. These observations suggest likely RDF cross-reactivity between ϕBT1 and ϕC31 integrases.
To understand the nature of integrase-RDF interactions, we used AlphaFold2-multimer to model the structures of the three integrases in complex with their cognate RDFs. As expected, the predicted structures of the two DNA-binding domains are very similar to the experimental structure of the DNA binding domains of LI-Int bound to half an attP site23, which was used to model binding of our integrases to DNA. (Fig. 4). All three complex models are quite similar and predict that the RDF uses a set of loops to clamp onto a hinge region between the integrases’ second DNA-binding domain (DBD2; sometimes called the zinc-binding domain or ZD) and the coiled coil that is inserted within it. (Fig. 4). These models are in good general agreement with prior experimental data24,25,26. The coiled coil is known to mediate synaptic contacts between paired att sites18,23,24,26,27,28. Our models suggest that RDF binding partially but not fully restrains the mobility of the coiled coil, but further structural work is needed to fully understand how such partial restraint controls reaction directionality.
In vivo, all three integrases catalysed attP x attB recombination to near completion, and as expected did not act on their attR x attL substrate in the absence of the RDF (Fig. 5), showing the strict directionality as well as efficiency of the integration reactions. A similar pattern was observed in vitro, with recombination being generally efficient (Fig. 6). However, in vitro recombination did not go to completion after 2 h, with ϕC31, ϕBT1, and TG1 integrases converting 76%, 54%, and 92% of the substrate, respectively. We cannot differentiate from these data whether the lower amount of in vitro product for ϕC31 and ϕBT1 integrases can be ascribed to an intrinsically lower initial reaction rate or instability of the protein over 2 h under the conditions used. The higher completeness of recombination observed in vivo could be due to the continuous expression of the proteins over the 16-hour growth period. The high activity of TG1 integrase is particularly noticeable, outperforming ϕC31 integrase, a recombinase that has been used in several in vitro and in vivo applications. Overall, the activities of the three integrases are consistent with our findings reported in an earlier study where we compared the activities of 10 different integrases29.
Effects of RDFs on the recombination activities of ϕC31, ϕBT1, and TG1 integrases
To study the specificity of integrase-RDF interactions across the three integrases, we studied the activities of each integrase in the presence of the three different RDFs both in vivo and in vitro. We did this in two ways: First, by using the integrase and the RDF as separate proteins, and secondly by constructing integrase-RDF fusions21 to account for effects due to differential binding affinities of integrases for non-cognate RDFs. Use of fusions also avoids potential effects on activity due to differences in expression levels of the integrase and RDF proteins. In addition to activating attR x attL recombination, RDFs inhibit recombination of attP x attB by their respective integrases18,30,31. To see if there is a correlation between the degree of RDF-mediated activation of excisive recombination and inhibition of integrative recombination, we also investigated the inhibition of attP x attB recombination by the cognate RDF for the three integrases.
In vivo, the ϕBT1 RDF, when fused to ϕBT1 integrase (and to a lesser extent ϕC31 integrase), was the most effective in both activating excisive recombination and inhibiting integrative recombination (Fig. 5). Furthermore, plotting the reaction endpoints for all of the pairwise tests shown in Fig. 5 gives a strong anti-correlation between the endpoints of the attP x attB and the attR x attL reactions: the data plotted in Fig. 7a have a correlation coefficient of -0.99. This confirms that equilibrium was reached in these in vivo assays, and that the effectiveness of a particular RDF in promoting attR x attL reactions directly correlates with its effectiveness in inhibiting attP x attB reactions: each pair tested reached its particular “set point” regardless of the starting conditions.
In vitro, in the presence of their cognate RDFs, ϕC31, ϕBT1, and TG1 integrases recombined 76%, 29%, and 94% respectively of their attR x attL substrates. As for attP x attB recombination, TG1 integrase is the most active among the three, giving near complete conversion of the substrate plasmid (Fig. 6). The overall pattern emerging from this analysis is that TG1 integrase is the most active in both integrative and excisive reactions. Incomplete reactions in vitro could be due to the factors discussed above for the in vitro attP x attB reactions, as well as weak directionality for the attR x attL reaction: that is, the equilibrium constant for the attR x attL reaction in the presence of RDF may not lie as far in favour of products vs. substrates as it does for the attP x attB reaction in the absence of RDF. This is supported in relation to the ϕC31 integrase by the conversion of 35–44% of the attP x attB substrate to product in the presence of the RDF (Fig. 6). In contrast, the in vitro data for ϕBT1 integrase in the presence of its RDF suggests that it simply did not reach equilibrium under the conditions used.
Plotting the reaction extent for each in vitro experiment (Fig. 7b) highlights additional aspects of these reactions. Unlike the in vivo inversion assays, the in vitro assays monitor deletion of a plasmid segment. Therefore, the expected equilibrium of a given reaction is more complicated to predict. While the forward reaction depends on intramolecular synapsis of two att sites within the same plasmid, the reverse reaction requires intermolecular synapsis between att sites on separate DNA circles that may have diffused away from one another. If formation of the intermolecular synapse is too difficult under the conditions used, the endpoints would be expected to lie on the axes, as they do for ϕBT1 and TG1. In contrast, if the barrier to intermolecular synapsis is not significantly different from the barrier to intramolecular synapsis, the reaction endpoints would be expected to lie on the diagonal, similar to what is seen in Fig. 7a. That is indeed approximately the case for ϕC31, indicating that ϕC31 integrase may form intermolecular synaptic complexes more readily than ϕBT1 and TG1 integrases do.
In vitro, ϕBT1 and TG1 RDFs were strikingly effective at inhibiting attP x attB recombination by their respective integrases, giving less than 5% activity in both cases. Complete inhibition of attP x attB reaction by the RDF is a key feature necessary for use of integrase-RDF pairs in applications where integrases are used as binary genetic switches. TG1 integrase will be particularly suitable for building such devices since it shows near complete integrative and excisive activities. Figure 7a also shows that different integrase – RDF pairs could be used in applications where a tunable switch is required (e.g. a promoter inversion reaction that is partially rather than fully biased toward one outcome).
We used integrase-RDF fusions21 to further investigate the specificity of integrase/RDF interactions across the three integrases and their RDFs. Among the three integrases, ϕC31 integrase showed the least orthogonal behaviour, responding to attR x attL activation and attP x attB inhibition by all three RDFs in both in vivo and in vitro reactions. In all cases, ϕC31-RDF was not as effective as the other two at regulating the activities of the integrase (Figs. 5 and 6).
As expected, ϕBT1 integrase prefers its cognate RDF in regulation of attR x attL and attP x attB recombination (Fig. 6). However, there is a noticeable difference in the interaction of ϕBT1 integrase with the RDFs when the two proteins are supplied separately and when they are fused together. ϕBT1 integrase had limited affinity for ϕC31-RDF and TG1-RDF when the RDFs were used as separate proteins. However, when the non-cognate RDFs were fused to ϕBT1 integrase, they were more effective in activation of attR x attL recombination and inhibition of attP x attB recombination (Fig. 6).
In contrast to ϕC31 integrase and ϕBT1 integrase, TG1 integrase showed a high degree of selectivity for its cognate RDF, and insignificant effects on its activity by ϕC31-RDF and ϕBT1-RDF, either supplied as separate proteins or when fused to the integrase. This is especially noticeable in in vitro reactions, the exception being ϕC31-RDF showing attR x attL activation when fused to TG1 integrase (Fig. 6).…Read more by Biomolecular Sciences, Alexandria, School of Molecular Biosciences, Glasgow, USA, MacDonald, Department of Biochemistry, Rice, The University of Chicago, Molecular Biology, Chicago, Alasdair I., School of Pharmacy, University of Glasgow, Heewhan, Aron, W. Marshall, Stark, UK, Faculty of Science, Liverpool, Holland, Baksh, Femi J., Olorunniji, Phoebe A., Shin, Liverpool John Moores University
