RETRACTED: The structure of SV40 large T hexameric helicase in complex with AT-rich origin DNA

[…] Required revisions:

1) The authors should make the interactions between the amino acids and the DNA more clear in Figure 4. It is not clear whether the authors are proposing specific interactions between the protein and the DNA or not in this portion of the structure. In Figure 3, the authors clearly illustrate the interactions they believe are occurring between the DNA and the protein. In contrast, in Figure 4 the relevant interactions are not delineated. Is this because of an absence of specific interactions with the DNA in this region? This question should be clarified by the authors.

In order to clarify the existence of specific interactions with the DNA in this region, we revised Figure 4 by adding detailed bonding interactions in Figure 4B, D, F in order to show the detailed specific contacts between the partially melted AT-rich origin DNA and the residues on the β-hairpin and DRF-loop.

2) The authors' model proposes that squeezing rather than a specific pushing or flipping of the bases out of the double helix is responsible for the partial denaturation. It is important to note, however, that specific interactions between the protein and the bases could have occurred prior to the structure captured. The best way to address this possibility would be to test mutants in DRF loops or K512/H513 amino acids and see how they effect initial DNA unwinding (e.g. using KMnO4 sensitivity). Without such mutants, the authors should discuss the possibility that there are more specific pushing or prying mechanisms that drive the partial melting but are not observed in the structure captured.

The K512/H513 amino acids of LT have been reported to be essential for origin melting (Kumar et al., 2007), and the equivalent His residue on the E1 helicase has also been shown to be critical for papillomavirus origin melting (Liu et al., M. Cell, 2007; Chunk et al., M. Cell, 2011). The fact that the K512/H513 mutant of LT failed to melt the origin suggests that specific interactions, in addition to the pure physical force of squeezing, may also be important for origin melting. We clarified this point in the subsection “The assembly path of LT hexamer/double hexamer on origin DNA” when discussing the model (new Figure 5, and the Figure 5 legend).

3) The authors should discuss their data and its implications for the mechanism of initiation more thoroughly. If melting occurs in the context of a planar SV40 LTag ring, then one of the resulting ssDNAs must exit this channel subsequently, as there is clear evidence that the final SV40 LTag helicase unwinds DNA while encircling ssDNA (Yardimci et al., 2012). Given the similarity between E1 and SV40 origin binding proteins, the authors should compare their findings to those made addressing DNA unwinding by the E1 helicase (e.g. Schnuck and Stenlund, 2010 and Liu et al., 2007). Finally, the authors should do a better job of explaining where initial DNA melting occurs at the SV40 origin and how this corresponds to their findings.

In the subsection “The assembly path of LT hexamer/double hexamer on origin DNA”, we added a discussion on how the hexamer/double hexamer assembly may generate local melting of the origin DNA for the initiation of replication. We used biochemistry to show that, upon adding Mg++, the pre-formed ring-shaped LT complex on origin DNA unwinds the origin with similar efficiency as the un-preformed complex on origin DNA. This suggests that the melted ssDNA region must become increasingly larger upon cyclic ATP binding/hydrolysis and that the melted ssDNA must exit the hexameric channel either fully or partially to achieve full origin melting and unwinding. We stated in the revision that the melted ssDNA must find a way to exit the hexameric channel, as LT has shown to preferentially translocate along one ssDNA during unwinding (Yardimci et al., 2012) and E1 has shown to interact with ssDNA (Enemark, 2006; and 2008). The exact details of how LT and other archaeal/eukaryotic replicative helicase interact and unwind fork DNA is still lacking. However, the action of LT unwinding can bypass a polypeptide road-block (similar to the size of a ssDNA) cross-linked to DNA substrates, suggesting that the LT hexamer has sufficient plasticity to alter its conformation around DNA and allow the passage of ssDNA without falling off the unwound substrate (Yardimci et al., 2012). Also, the mutational results for the potential role of the β-hairpin His residue in LT (Kumar, 2007) and E1 (Schnuck and Stenlund, 2010 and Liu et al., 2007) in origin melting are discussed here. We also stated that, at present, we have no data regarding the process of how this initial melting of oriDNA propagates into fully open replication forks, and that a more comprehensive understanding of this process will require future high-resolution information regarding LT interactions with different types of DNA substrates such as the origin dsDNA to partially melted origin to the final replication fork. These points are now discussed in this round of revision.

4) The experimental details of Figure 1D and E should be added to the Methods section or described in the figure legend. Based on typical helicase assays, the labeling of "ssDNA" and "dsDNA" appears to be reversed in these figures. In addition, labeling of "boil" and "-ATP" are reversed for the two panels (unless this is an unusual helicase assay, the data is mislabeled in Figure 1D).

We have added a more detailed description of the methods for performing Figure 1D, E to the Methods section. The labeling of ssDNA and dsDNA in this unwinding assay is correct, because the unwound ssDNA product forms a very stable hairpin structure due to the palindrome origin sequence, causing the ssDNA product to migrate slower than the unwound dsDNA substrate. For clarity, we changed the labeling of “boil” and “–ATP” to “Boil” and “dsDNA” (dsDNA substrate control).

5) The manuscript formally reports one crystal structure (131-627), however, it describes observations for two distinct crystal structures (108-627 and 131-627). Either the authors should report both crystal structures (including the coordinates) or the discussion should be limited to only the 131-627 structure.

Given that both the LT131-627 and LT108-627 constructs yield essentially the same structure, we agree that it’s only meaningful to present one structure, and in this revision we only present the LT131-627 structure. Also for this revision, we obtained a 2.9 Å data set (originally 3.16Å) for the LT131-oriDNA complex structure, and we now report this 2.9Å structure using the LT131 construct. In our 3.16Å structure, we only built 24 bp out of the 32 bp oriDNA. With the 2.9Å data, we are able to build with confidence the entire 32 bp oriDNA, plus the two 3’-T overhang.

6) The authors should do a better job of introducing the AAA+ domain and its components (e.g. the DRF-loop) for a general audience.

We added more information on the AAA+ domain in the Introduction section (third paragraph), and introduced the β-hairpin and DRF loop of the LT helicase in the new Figure 1—figure supplement 2C and in the last paragraph of the subsection “Overall structure of the LT-dsDNA complex”.

7) In the subsection “Comparing the dsDNA-bound and DNA-free apo-structures of LTag hexamer” – "substantial concerted conformational changes" sounds contradictory to "conformations are quite similar" found later in the same sentence. Please clarify (does this sentence mean to say that the hairpins and DRF-loops move but the rest of the structure remain the same?).

Yes, the substantial conformational changes refer to changes in the six β-hairpins and DRF loops along the central channel, which result from amplified structural changes due to their special structural connections to the nucleotide pocket. The rest of the molecule and especially the hexamer as a whole have relatively very small changes. We revised the sentence (by adding “these hexamers”) to clarify the confusion: “Although these different DNA-free hexameric structures in different nt-bound states reveal substantial concerted conformational changes for the six β-hairpins and DRF loops within and along the central channel of LT hexamer, the overall conformations of these hexamers are quite similar.”

8) In the subsection “A mechanism for origin melting: Squeeze-to-Open” – "…diffraction quality was poor, indicating intrinsic differences…". This statement overinterprets the meaning of "poor diffraction quality". Lesser diffraction quality of the EP-half origin dsDNA in complex with LTag indicates that the crystals had different diffraction characteristics. Differences in interaction or remodeling are not required (though may be present). Please revise accordingly.

In light of this comment, we revised the statement to simply state the fact that the hexamer-EP-origin DNA co-crystal diffracted poorly without hypothesizing a possible cause for it.

9) The manuscript would benefit from an illustration of the authors' model discussed in the subsection “The assembly path of LTag hexamer/double hexamer on origin DNA”. This addition would help clarify the discussion presented. For example, in the last paragraph of the aforementioned subsection, it is unclear what exact interactions the authors are proposing "become sterically unfavorable".

We added Figure 5 to help illustrate the model we tried to present in the subsection “The assembly path of LT hexamer/double hexamer on origin DNA”. Thank you for the suggestion.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed before acceptance, as outlined below:

The manuscript presents a structure of the SV40 Large T-antigen hexamer bound to double-stranded (ds) DNA. Importantly, this is the first structure of a hexameric helicase bound to dsDNA and it reveals a narrow channel that is bound to distorted/melted DNA, suggesting a "squeezing" mechanism for the initial melting of origin DNA by hexameric helicases. The revised manuscript includes a higher resolution structure and addresses many of the criticisms of the previous presentations.

There is one area that remains problematic and needs to be addressed prior to publication. Although the squeezing model is intriguing and is a potentially powerful solution to how viral helicases solve the initial unwinding problem, the authors need to provide a more even-handed discussion of other potential models. There is substantial evidence that other mechanisms may contribute to the DNA melting event. Several lines of evidence in the literature indicate that the stepwise assembly – from a dimer to two hexamers – occurs sequentially via interaction of the OBDs at adjacent dsDNA recognition sites. This leads to a situation where the proteins spiral around the dsDNA helical axis. This is true for LT and also for E1 (for structures, see: 2ITL, 4FGN, 1KSX, 1KSY).

The current Discussion avoids discussing how the spiral assembly is converted to planar rings by implying that the subunits assemble at the origin as planar rings in the first place. This implication is provided without evidence and directly contradicts the structure of Bochkareva et al. (2006) (PDB 2ITL) where the LT OBDs are offset around the helical axis.

To address this concern, we added more discussion on other possibilities discussed in previous publications regarding spiral assembly of helicases, with the relevant references (Enemark et al., EMBOJ, 2002; Bochkareva et al., EMBOJ, 2006; Meinke et al., PlosBiol, 2007; Meinke et al., JV, 2013) and the related data sets cited in this revision. Hopefully, such revision will make our paper to be presented with a more even-handed discussion of other possibilities.

At the same time, in order to give readers a clear view of the previous structural studies in the above cited PDBs (2ITL, 4FGN, 1KSX, 1KSY, all are OBD alone bound to the PENs (LT) and Hexa (E1) nucleotide recognition sequences), as well as the advancement of our studies of LT OBD-helicase domain in complex with oriDNA, we added Figure 5—figure supplement 1. Panel A shows the structure of 2ITL above as a representative of structures of OBD alone (LT131-258) with four PENs. These studies all reveal how OBD interacts with PEN sequence (GAGGC) in LT or Hexa seqeunce in E1. Panel B shows the co-crystal struture of LT131-627 (OBD-helicase domain) bound to a full half origin (two PENs, plus AT-rich). Purified LT131-627 is capable of initiating hexamer/double hexamer assembly and origin unwinding in vitro (see Cuesta et al., JMB, 2010; Chang et al., Cell Reports, 2013; and this study in Figure 1). This structure clearly reveal that the two OBDs of the dimeric OBD-helicase LT interact with the origin DNA differently from that when OBD acts a separate domain. Like in the OBD alone case, the two OBDs bound to DNA do not interact with each other. Dimerization is through the helicase domain. Panel C shows the co-crystal structure of LT131-627 hexamer bound to a full half origin DNA, in which hexamerization is through the helicase domain, whereas the six OBDs are partially disordered (not interacting with each other to hexamerize or bound tightly to the PENs/Pseudo-PEN).

The detailed side-side structural comparison above indicate it is the helicase domain that oligomerizes around origin DNA and the OBD does not hexamerize, which is also consistent with our prior EM studies showing flexible OBD domain (Cuesta, JMB, 2010; Valle, JMB, 2006) and X-ray studies showing hexamerization of helicase domain alone (Li et al., Nature, 2003; Gai et al., Cell, 2004; Lilystrome, G&D, 2006; Zhou et al., JBC, 2012). OBD clearly is important for binding to PEN and the Pseudo-PEN, which may be critical for the correct initial assembly of the helicase dimer intermediate on the AT-rich origin, but OBD may not play a role in oligomerization. Therefore, the structure of Bochkareva et al. (2006) (PDB 2ITL) where the individual LT OBDs are offset around the helical axis because of the space of PENs should not be able to form a hexamer, and thus we don’t think our hexamerization model and origin melting in the helicase domain is comparable the structures of OBD domain alone used by Bochkareva et al. (2006) (PDB 2ITL) and other related structures cited here (i.e. 2ITL, 4FGN, 1KSX, 1KSY).

The conversion of a spiral assembly to a planar one is shown in the work of Stenlund with E1, where an extended footprint is subsequently converted to a compact footprint. This conversion is associated with the conversion of a double-trimer to a double-hexamer and also – most importantly – with DNA melting. By describing LT assembly as planar at all points in time, the authors avoid making a significant comparison to the work of Stenlund on DNA melting. Although it is possible that E1 and SV40 LT function by different mechanisms, because the authors are only looking at the final outcome of the assembly, it is not possible to eliminate a role for a switch between a spiral and planar conformation as contributing to the distorted DNA structure observed. A more even-handed discussion of other possibilities should be included in a revised Discussion.

We discussed and cited the trimer assembly/melting and two other relevant studies from Stelund group in the manuscript. We think that the authors presented solid biochemistry data showing origin melting during the assembly progress of E1 that is likely performed at the trimer/double trimer stage, which is also implied to be through spiral to planar conversion possibility. For LT, we fully appreciate this suggestion about discussing the possibility of conversion from sprial to planar for LT. In this round of revision, we added more such discussions in a few places. One example is in the Discussion section: “This simple explanation (for the observed planar vs spiral) would then imply an ability of switching between the planar and staircase conformations for LT and E1, even though only one conformation has been consistently observed in either protein structures determined to date. In this consideration, it is possible that the crystallographic packing somehow selectively traps the β-hairpins in the planar conformation for LT and the staircase conformation for E1, despite the crystallization buffer condition.” In this revision, we also added a new paragraph discussing this possibility: “While the hexamer-dsDNA complex structure reveals a near-planar ring of six subunits, the possibility of changing from spiral intermediate assembly to the final planar hexamer also exists, which may provide extra energy to melt the origin in addition to the squeezing force from the narrower channel. While no spiral assembly of hexamer helicases on dsDNA has ever been reported, the spiral assembly of hexameric motors on ssDNA has been reported for E1, DnaB, and Rho…”. We hope this revision has achieved a more even-handed discussion of other possibilities, as suggested by this comment.

Specific points:

1) In the third paragraph of the Introduction you use "LT131" and it has not been defined. In this introductory part of the paper the authors should describe the LT fragment rather than using this shorthanded name.

Now defined LT131, we used LT131-627.

2) In the first paragraph of the subsection “LT deletion constructs capable of unwinding origin containing dsDNA”, the authors should do a more complete job of explaining what LT131 is. Just saying LT131-627= LT131 seems a little cryptic. Would be better if the authors stated that the structural studies were performed on a fragment of LT extending from amino acids 131 to 627 that included both the OB-fold and AAA+ domains.

Clearly stated the domains of the used LT construct as suggested. “We used SV40 LT construct LT131-627 (containing residues 131-627, referred to as LT131 hereafter) that includes the OBD domain (residues 131-258) and the helicase domain (residues 260-627) (Figure 1A)”.

3) Although the authors have improved the documentation of the protein elements of the protein-DNA interactions the bases or base-pairs that the proteins are interacting with are still not well documented. In Figure 3 there is no information about the DNA sequences that are bound. In Figure 4 the sequences that are present in the structure are indicated for one panel (4B) but not for the remaining panels. Since it might clutter the figure, the best approach might be to make a table of all the protein-DNA interactions that could be provided as supplementary information.

In light of this suggestion, we revised Figure 3 to add more details, including the DNA residue numbers now and the amino acid residues of a particular LT subunit. In addition, we added two tables to show all the protein-DNA interactions: Supplementary file 1 to show the detailed interactions with DNA in the Zn-domain, and Supplementary file 2 to show the interactions in the AAA+ domain of the hexamer.

4) Last sentence of the Results: This sentence indicates that there are interactions "with the DNA from both the major and minor grooves". This is interesting but needs to be better described. I would suggest a full paragraph discussing the interactions that lead to this conclusion. Alternatively, the authors could summarize the minor and major groove interactions at the end of the discussion of the protein-DNA interactions. As there is no other place in the Results that does discuss major and minor groove interactions just referencing Figure 4 is not adequate. For example, how many of the interactions are in the major vs. the minor groove?

The information of the interactions from the major and minor groove of DNA are now included in Supplementary file 2.

5) The term "squeezing" suggests an active (energy-driven process)- triggers a narrowing of the ring to cause a "squeeze". This contradicts the discussion describing the highly consistent LT structure in the DNA-free and DNA-bound forms. Do the authors mean to suggest that the addition of a final subunit to close the ring causes an immediate squeeze (?). Is this what is intended by using the phrase "snapping force" in the last paragraph of the subsection “The assembly path of LT hexamer/double hexamer on origin DNA”? There does not seem to be a reason that the transition from a lock-washer to planar form could not provide a similar energetic drive for unwinding.

We don’t think the “squeezing” hypothesis contradicts with the highly consistent LT structure in the DNA-free and DNA-bound forms for the following reason. The DNA free hexamer forms in different nucleotide states (empty, ADP, or ATP bound) all have central channel smaller than the B-from dsDNA. When LT hexamer assembles on the origin dsDNA with ADP bound, the hexamer conformation and narrow central channel opening of the AAA+ domain is very similar to that of the ADP-bound hexamer without DNA in the central channel. Therefore, the assembly of the LT hexamer around the origin dsDNA with ADP binding will form a narrow channel on the AAA+ domain to provide “squeezing” on the dsDNA that may result in local melting. When binding to ATP, the AAA+ domain channel is about 3-5 Å narrower than that in the ADP-bound form (Gai et al., Cell, 2004), consistent with the much more extensive local origin melting when ATP is used instead of ADP (Borowiec and Hurwitz, EMBOJ, 1988).

However, we totally agree that the above evidence of squeezing by the narrower channel of the assembled hexamer on dsDNA when binding to ADP or ATP does not mutaully exclude the possibility that local origin melting could also be facilitated by the energetic force obtained from the conversion from a spiral conformation to a planar one upon completion of a hexamer assembly. We mentioned this possibility in last submission, and further emphasized this possibility in this revision (also please see our second response above on this point).

6) In the second paragraph of the subsection “The assembly path of LT hexamer/double hexamer on origin DNA” – an LT dimer is probably not directly comparable to an E1 trimer.

We agree on this point. We revised this statement to avoid the confusion: “Even though no stable trimer or tetramer intermediates of LT-oriDNA were observed, the formation of the stable LT dimer-oriDNA intermediate complex clearly does not cause origin DNA distortion or melting (Chang et al., 2013). Interestingly, previous biochemical evidence suggested that E1 helicase formed a trimer intermediate complex instead of a dimer on each half of the origin DNA to melt the origin DNA before hexamer/double hexamer formation (Liu et al., 2007).”