Molecular recognition by multiple metal coordination inside wavy-stacked macrocycles

Most biological and synthetic receptors for small organic molecules employ a combination of relatively weak intermolecular interactions such as hydrogen bonds. A host compound that utilizes stronger yet reversible bonding in a synergistic manner could realize precise recognition, but the regulation and spatial arrangement of such reactive interaction moieties have been a challenge. Here, we show a multinuclear zinc complex synthesized from a macrocyclic ligand hexapap, which inwardly arranges labile metal coordination sites for external molecules. The metallomacrocycle forms a unique wavy-stacked structure upon binding a suitable length of dicarboxylic acids via multipoint coordination bonding. The saddle-shaped deformation and dimerization realize the differentiation of the interaction moieties, and change of guest-binding modes at specific metal coordination sites among the many present have been achieved utilizing acid/base as external stimuli.


Responses to the reviewers' comments
Following the comments from the reviewers, we have revised our manuscript. The reviewers' comments are cited in italics and in blue letters. Word files of the manuscript and the supplementary information with "Track Changes" records have been separately uploaded.

To reviewer 1
First, we would like to thank the present reviewer for helpful suggestions in revising the manuscript. We carefully revised the manuscript and the cif files. According to the changes described below, Fig. 3  Detailed replies and changes we have made are as follows: Crystallographic Review It appears that the two structures have been worked on by different people, and there is considerable inconsistency between the structural refinements. The treatment of the two structures should be harmonized. For example, it should be noted that different target values have been used for restraints of equivalent bonds in the different structures -these should be consistent across the two structures.
The two structures have been reanalyzed by the same crystallographer. Target values for restraints of equivalent bonds in the different structure are now consistent across the two structures. Molecular structures are similar for the two structures, but the space group is different. The number of independent atomic sites in crystallographic asymmetric unit of one structure (P4 1 ) is approximately four times larger than that of the other one (I4 1 /a). So several different treatments were required for each analysis.
We would like to describe the scale of current crystallographic procedures. The molecular structures of the present study have approximately 300 atoms excluding hydrogens. Size and scale of the molecule are fairly large for small-molecule crystallography. The size is comparable to that of small-size protein with less than 100 residues. In addition, there are large solvent accessible voids in the crystal structure similar to macromolecular crystallography.
As the structures currently stand, they should not be published, however they could be brought up to publication standard .

Structure 187a
Although this structure is essentially correct, there is a large number of issues with the structure refinement which need to be fixed before it is suitable for publication.

Firstly, ALL A-and B-level alerts should be explained through the use of appropriate _vrf's.
We added appropriate _vrf's for all B-level alerts as suggested by the reviewer. All A-level alerts have been resolved after the reanalysis. After the checkCIF (ver. 26/Feb/2017), the responses in the vrf form that were intended for Alert level B also appear for the Alert level C tests under the same test number, but we suppose that this would be an error of the checkCIF program.
It is unclear whether PLATON SQUEEZE has been applied. If it has not, it probably should be.
We applied PLATON SQUEEZE in the revised refinement as suggested.
A large number of reflections have been individually omitted from the refinement. Only the few worst matching reflections should actually be omitted.
We have not omitted reflections in the revised refinement. There is a PLAT934 B-level alert in the checkcif report. All of these are low-order reflections. The data were measured using a diffractometer for small molecule crystallography. In addition, there are large void space in the crystal structure. These outliers mainly come from the difficulties of measurement and refinement.
There must be an explanation of the disorder present, how it was modeled, and a further explanation of any other restraints used in the _refine_special_details section of the CIF. The authors should note that the ISOR restraint is intended for use only on isolated atoms. It should not be used for bonded atoms, in which case RIGU, DELU and SIMU are the appropriate restraints to use.
There is no disorder in the current structure. The explanations on applied restraints are given in the _refine_special_details section of the revised cif file. Detailed information on the refinement can be referred in the _shelx_res_file section. No ISOR restraint was applied in the current refinement.
Atom O10 should be attached to C66 in the asymmetric unit, as this forms a complete molecule.
This point was corrected as suggested.
The tetrachloroethane molecule will probably be better refined as half-occupied.
We did refinements using full-and half-occupancy models for the tetrachloroethane molecule.
Judging from the refinements, we selected the half-occupied model as suggested reviewer.

Z has been reported as 16, when is should be 4 (Z' is 1/4, rather than 1).
We corrected Z value as suggested.
The structure could probably benefit from a global rigid bond restraint, rather than the few selectively applied ones.
Thank you for good suggestion. We cannot find more global rigid bond for this molecule.

Structure shelx
This structure, as supplied, is chemically incorrect. The disorder of the diacid molecules has not been modeled in any sensible fashion These is a number of incorrect bonds in the model, and the scattergun approach to addition of hydrogens is inappropriate. The authors should make use of appropriate PART commands to correctly model this disorder, and should competitively refine the occupancies of the disorder components. Additionally, one disorder component of the diacid forms a covalent bond to an adjacent tetrachloroethane molecule. The tetrachloroethane molecule should be placed in an appropriate PART, and its occupancy refined complementary to that of the atom to which it is adjacent.
Total number of individual atomic sites is approximately 300, which four times more than that of the other structure due to low crystallographic symmetry. The difficulty of the refinement is similar to that of small-sized macromolecular crystallography.
To determine chemically correct structures, we applied many kinds of constraints and restraints in the refinement. Many of the six-membered rings are constructed using AFIX 66 constraints.
Almost all the hydrogen atoms were removed to achieve sufficient parameter shift to su ratio.
Anti-bumping restraints were applied as suggested by the reviewer.
We carefully checked the converged structure, and confirmed that it did not contain any errors on structures and occupancies.
It should be noted that SQUEEZE has been correctly implemented in the refinement of this structure.
We also applied PLATON SQUEEZE in the revised refinement.

Again, all A-and B-level CIFCHECK alerts should be answered with _vrf's in the CIF. The following
ALERTs, however, should be fixed, rather than explained: All A-level alerts were fixed in the revised refinement. We also answered all B-level alerts as suggested by the reviewer. After the checkCIF (ver. 26/Feb/2017), the responses in the vrf form that were intended for Alert level B also appear for the Alert level C tests under the same test number, but we suppose that this would be an error of the checkCIF program. We still use global application of ISOR and SIMU restraints like macromolecular crystallography. Without these restraints, it is very difficult to determine a reliable and chemically correct structure due to low data quality and many parameters. To determine a chemically possible solution, we also applied XNPD restraints.

PLAT780_ALERT_1_B Coordinates do not Form a Properly Connected Set Please Do !
-There is no reason not refine the structure as a single connected set.
This problem was solved in the present structure.
Again, this structure needs an explanation of the disorder modeling, and application of restraints used in the _refine_special_details section of the CIF.
The explanations on applied restraints are given in the _refine_special_details section of the revised .cif file. Detailed information on the refinement can be referred in the _shelx_res_file section.

To Reviewer 2
In this manuscript, Nabeshima et al described the one-pot synthesis and binding properties of an imine macrocycle that contains multiple metal chelation sites inwardly. The structures of the imine macrocycle and its complexes with dicarboxylates were unambiguously characterized by 1H NMR spectroscopy and single crystal X-ray diffraction, together with absorption and emission spectra.
The synthesis of the imine macrocycle is very efficient thanks to the judicious synthetic design. All binding sites are positioned inward and convergent to the center, thus affording high specificity and selectivity towards pimelic acid. The formation of a dimeric structure upon guest binding is very unique and nicely demonstrates the power and advantage of self-assembly or dynamic combinatorial library. Authors also showed that the binding mode could be reversibly switched by acid-base chemistry. I think this work is very creative and sufficiently novel for the publication in the Nature.
Commun., and I recommend this manuscript published in this Journal essentially as it stands.
Thank you very much for your kind comments on our manuscript. We appreciate your recommendation for publication of our work. We believe this paper contributes to the design strategy of functional supramolecular complexes, as well as the development of selective molecular sensors or allosteric multinuclear metal catalysts.
In the SI, p S27, line 12, there is a typo in the crystal data: ...(4e)4 …should be (4e)2 Thank you for your comments. We corrected the corresponding part accordingly.

To Reviewer 3
Novelty: Moderate Impact: The work will be of great interest to the general chemistry community, but not to the general science community. We recommend publication to a high impact general chemistry journal.
A multinuclear complex capable of undergoing size specific guest-host interaction is described. 2D NMR experiments in conjunction with structural characterization via x-ray diffraction of crystals of the complexes show that the authors indeed made the complex and that it bound 2 substrates in a site specific manner. A unique recognition profile could be isolated using acidic conditions as stimuli.
These unique profiles do not perform any chemical function nor do they contain novel binding motifs.
The work is very interesting but ultimately not immediately of general utility.
Thank you very much for your great interest in our paper. We appreciate your valuable comment. As we stated in the introduction, most biological and synthetic receptors employ a combination of weak intermolecular interactions such as hydrogen bonds, p-p interactions, and van der Waals interactions. Meanwhile, multiple coordination bonds are employed for molecular binding in this supramolecular system. This design strategy provides a fresh viewpoint not only to general chemists, but also to specialists with various academic disciplines including physics, biology, and medicines. Furthermore, although the current paper does not include immediate applications, the macrocycle capturing small molecules utilizing multiple coordination bonds leads to the next-generation allosteric metal catalysts and/or specific molecular sensors. Thus, we believe that this paper is suitable for publication in Nature Communications.
General observations: • ROESY experiments showed cross peaks that supported both the binding of the substrate and the binding of 2 macrocycles to make the "wavy-dimer" • The crystal structure of the precursor and the guest bound complexes are consistent with the cross peaks observed by ROESY showing specific atoms brought near each other through space • Characterization of synthesized materials is pristine.
Thank you very much for supporting our experimental results and analyses.
• Manuscript contains a lot of run on sentences.
According to your pointers, the following sentences have been revised. • Authors claim that the bimolecular recognition mode is incredibly stable since an excess of the ligand did not lead to changes observed by 1H NMR. This would also be observed if rapid exchange was taking place. The best way to support this claim is to calculate a binding constant using concentration dependent studies. Absent that, the authors could add excess acetic acid or some other small mono acid and show that the dimer does not dissociate.
We appreciate your valuable suggestions. The host-guest complex of the bimolecular recognition mode [1 2 Zn 12 4e 2 X n ] has a C 2 point-group symmetry (6 different Zn-pap units), because the top and bottom macrocycles are in a different environment as the result of the binding of 4e 2-.
Meanwhile, the guest-free [1 2 Zn 12 X n ] dimeric framework and the tetramolecular recognition mode [1 2 Zn 12 4e 4 X n ] have a S 4 point-group symmetry (3 different Znpap units). As seen from the Supplementary Fig. 17, the 1 H NMR spectrum showed that the host-guest complex has a C 2 symmetry, thus it was confirmed that the Zn-complex existed as the bimolecular recognition mode [1 2 Zn 12 4e 2 X n ], and the two 4e 2molecules were fixed in the cavity on the NMR timescale.
After careful reinvestigations of the 1 H NMR spectra, however, we noticed that the possibility of the binding of additional 4e 2onto the bimolecular recognition mode [1 2 Zn 12 4e 2 X n ] with rapid exchange compared to the NMR timescale cannot be excluded. In other words, X, a labile ligand exchanging faster than the NMR timescale, can be an additional 4e 2if excess amount of H 2 4e were added.
Based on the discussion above, the following descriptions have been revised.
(Revised) the addition of more than 2 molar amounts of the pimelic acids H 2 4e against the wavystacked dimer [1 2 Zn 12 X n ] did not change a binding mode, but the host-guest complex stably existed in a bimolecular recognition mode [1 2 Zn 12 4e 2 X n ] (see Supplementary Fig. 17). and [1 2 Zn 12 4e 4 X n ] have been also modified in the following two points: 1) The structure has been slightly tilted to emphasize its dimeric structure; 2) ROH or ROligands, depicted as green circular cylinders coordinating to the Zn atoms of the dimer, have been omitted.
(Original) (Revised) · Page S17, in the caption of Supplementary Fig. 17 (Original) Addition of excess amount of H 2 4e did not result in further binding to [1 2 Zn 12 4e 2 X n ].
(Revised) (deleted) · Page S18, in the caption of Supplementary Fig. 18 (Original) Dimer of Zn-hexapap with two pimelates, (Revised) Dimer of Zn-hexapap in bimolecular recognition mode, (Original) Dimer of Zn-hexapap with four pimelates, (Revised) Dimer of Zn-hexapap in tetramolecular recognition mode, • The authors claim that binding is size specific in the regime of 4 to 5 carbon diacids noting an increase in fluorescence when those guests were introduced; this is not well supported as the NMR with the larger diacids showed multiple species forming likely because of oligomerization which would lead to self-quenching which is consistent with what the authors observed.
Thank you very much for your pointers. We did not intend to claim that the binding of dicarboxylates is size-specific, but intended to claim that the "wavy-dimer" structure is formed as a single species only when a specific size of carboxylates binds to the Zn-hexapap. We agree that the dicarboxylic acids other than adipic acid H 2 4d and pimelic acid H 2 4e also bind to the Zn-hexapap; they do not lead to the selective formation of "wavy-dimer", but to the formation of multiple species ( 1 H NMR, Figs. 2a-2i). In the original manuscript, "recognition of specific dicarboxylic acids" was used to describe that only H 2 4d and H 2 4e were bound to Znhexapap in the bimolecular recognition mode [1 2 Zn 12 4X n ] as describe in Fig. 3. However, this phrase might be interpreted as that only the specific dicarboxylic acids can interact with Znhexapap. To avoid this confusion, we would like to revise the description in relation to "recognition of specific dicarboxylic acids" as below. (Revised) Zn-hexapap recognized Reviewers' comments: Reviewer #1 (Remarks to the Author): As they stand, these structures should not be published. Although the structural quality has improved somewhat, there are still significant structural issues (especially with the P41 structure). Structure 1 (P41 structure): There is a number of problems with this structure still.
The most significant problems, which must be addressed before this structure can be published: Hydrogens are missing from the large proportion of the structure. These should be included. If the authors are having problems with rotating hydrogen atoms, they should firstly refine with fixed methyl groups (AFIX 33 rather than AFIX 137), and if necessary refine the water molecules without hydrogen atoms -however if possible hydrogen bond interactions are considered, it may be possible to apply appropriate restraints to locate these as well.
More seriously, the disorder of the central bridging ligands has still not been modelled. This is so serious such that when a casual observer looks at the crystal structure, they cannot discern the nature of these bridging ligands. These bridging ligands (or large macrocycle as it is currently modelled), also form two bonds to the chlorine atom of an adjacent solvent molecule. If the authors are unsure how to correctly model disorder they should consult another crystallographer.
The authors have use a wide array of thermal parameter restraints, and attempted a very finegrained approach. The collection of SIMU, RIGU and ISOR restraints can all be replaced with simply RIGU SIMU 0.01 0.02 2 which will appropriately restrain the entire structure. Other, the minor problems: _cell_angle_alpha 90.000 (5)  We believe that the quality of crystal is good from a macromolecular point of view but slightly insufficient for the criteria of small molecule crystallography. ; The authors have not correctly interpreted what this ALERT is asking them about. Although the response could be considered vaguely correct, it does not address that the ALERT is asking about the data limit applied during the structure refinement. _vrf_PLAT242_shelx ; PROBLEM: Low 'MainMol' Ueq as Compared to Neighbors of C210 Check RESPONSE: C210 is located between three methyl groups and a six-membered ring. Three methyl groups are located at the surface of the molecule and they are affected by strong thermal motion. The thermal motion of the six-membered ring is much smaller than those of the methyl groups. ; Examining the three methyl groups, it is obvious that they are exhibiting string rotational thermal motion. The central atom to which they are attached would not be expected to exhibit this thermal motion.
Struture 2 (I41/a) This structure appears to be largely correct, although there are some improvements that should be made. Atom O10 should be connected to C66 in the asymmetric unit. The two parts of the outer ligand should be joined together in the asymmetric unit. The rotational disorder of the t-butyl groups could also be modelled. Manuscript has been revised and is much more clear and concise as well as grammatically correct. Suggested Experiments and comments: • Authors claim that the bimolecular recognition mode is incredibly stable since an excess of the ligand didn't lead to changes by 1H NMR. This would also be observed if rapid exchange was taking place. The best way to support this claim is to calculate a binding constant using concentration dependent studies. Absent that, the authors could add excess acetic acid or some other small mono acid and show that the dimer does not dissociate. The combination of the supporting figure17&18 along with the titrations of pimelic acid is sufficient to show that the binding mode is changed which speaks to the stability of the wavy stacked dimer. The revised Figure 4 is more concise.
• The authors claim that binding is size specific in the regime of 4 to 5 carbon diacids noting an increase in fluorescence when those guests were introduced; this is not well supported as the NMR with the larger diacids showed multiple species forming likely because of oligomerization which would lead to self-quenching which is consistent with what the authors observed. Authors revised this claim to clarify that 4 and 5 carbon diacids led to sophisticated and clearly defined species as opposed to the possible mixtures that other linker lengths made. The authors also clarified that the decrease in fluorescence may have been due to self-quenching, and that they did not do lifetime studies to route this out. This revision is acceptable as the 4 to 5 carbon diacid bound structures were well characterized and that is more interesting than whether the host is specific to substrate length.

Conclusion:
Overall the revised manuscript adequately addressed our concerns and is satisfactory for publication.
Following the comments from the reviewers, we have revised our manuscript. The reviewers' comments are cited in italics and in blue letters. Word files of the manuscript and the supplementary information with "Track Changes" records (from the 1st revision) have been separately uploaded.

To reviewer 1
As they stand, these structures should not be published. Although the structural quality has improved somewhat, there are still significant structural issues (especially with the P41 structure).

Structure 1 (P41 structure):
There is a number of problems with this structure still.
The most significant problems, which must be addressed before this structure can be published: We added the following two sentences.
There are no available reflections for analysis in d<0.9A region.
The signal to noise ratio is less than 1.5 in 0.85<d<0.9A in the data. We fully agree with your comment. We revised the response as follows; These atoms were located between three methyl groups of a tBu group and a six-membered ring.
Three methyl groups are exhibiting strong rotational thermal motion. The thermal motion of the central carbon atom is much smaller than those of the methyl groups.

Struture 2 (I41/a)
This structure appears to be largely correct, although there are some improvements that should be made. Atom O10 should be connected to C66 in the asymmetric unit. The two parts of the outer ligand should be joined together in the asymmetric unit. The rotational disorder of the t-butyl groups could also be modelled.
We changed an atomic coordinate of O10 to be connected to C66. The two parts of the outer ligand has been joined together. We also modelled the rotational disorder of two of the t-butyl groups.
There are additional minor errors: ; These reflections should be OMITted from the refinement.
We omitted these reflections from the refinement.
According to the changes described above, Fig. 3 and Fig. 4b,c in the manuscript, Supplementary   Figs. S14, S22, and S23 have been redrawn with the revised structures. Descriptions in the Methods section have been also revised.

Molecular recognition by multiple coordination inside wavy stacked macrocycles
Revised Manuscript: Ready for publication in Nature Comm.
Thank you very much. We appreciate your recommendation for publication of our work in Nature Communications.
From original review: • Manuscript does not contain but does contain a lot of run on sentences.
Manuscript has been revised and is much more clear and concise as well as grammatically correct.
We appreciate your advice to improve the manuscript. Overall the revised manuscript adequately addressed our concerns and is satisfactory for publication.

Suggested Experiments and comments: • Authors claim that the bimolecular recognition mode is incredibly stable since an excess of the
Thanks to your good advice, the manuscript has been revised to be more accurate and more concise. We really appreciate your thorough review. We hope our study contributes to the wide range of scientific communities.
Quote "The hydrogens at the methylene groups belonging to the "branched part" of the central dicarboxylates were not modelled due to the limitation of the SHELX program." -this is not true. These hydrogens can be modelled with SHELXL.
Quote "We tried to replace SIMU RIGU and ISOR restraints. The trials were not successful. So, we still use many restraints." -following the review is a copy of the converged structure where global RIGU and SIMU restraints have been applied, and the hydrogen atoms on the methylene atoms have been correctly modelled. Note that this refinement method removes the five B-level alerts relating to low Ueq values, and results in more sensible anisotropic thermal parameters over the whole model. It also removes the need for the XNPD restraint, and removes the need to DAMP the structure refinement.
Although the _vrf's now have more sensible comments, they are now no longer syntactically correct. For example, "_vrf_PLAT242_shelx" has become "_vrf_PLAT242". Omitting the _shelx identifier means the _vrf is no longer associated with the _shelx data block, and the _vrf no longer appears in the checkCIF report. This should be corrected for both structures.

Minor correction -_diffrn_radiation_monochromator 'Synchrotron' -should read 'Si double crystal monochromator'
Regarding the second structure: Quote "We changed an atomic coordinate of O10 to be connected to C66. The two parts of the outer ligand has been joined together. We also modelled the rotational isorder of two of the t-butyl groups." The reviewer must apologize, there is also rotational disorder of the third t-butyl group, which needs to be modelled.
For this structure, again, refining with global RIGU and SIMU restraints (the SIMU slightly tightened to "SIMU 0.01 0.02 2") results in overall better quality thermal parameters, and removes the need entirely for any DAMP and XNPD restraints, and EADP constraints. An updated SHELX file has been appended to the report, with thermal paramter constraints and disorder modelled.
Additional points from the manuscript: Page 19, line 16: "The diffraction intensity data up to 1.00 Å was measured..." -in the structure refinement the applied data limit was 0.9 Å. It is also implied in the CIF that data was collected to a higher angle, however was cut down during structure refinement do to poor quality of diffraction at higher angles. The manuscript should be updated to reflect this.
Page 19, line 18: "The structure was determined and refined using SHELXL-2014" -SHELXL-2016 is the version used to refine the structure, and the reference should be updated to the 2015 paper (http://dx.doi.org/10.1107/S2053229614024218). SHELXL is not a structure solution program. The program used to solve the structure (SIR92) should be included and appropriately referenced.
Page 20, line 23: "Obtained data were processed using a Bruker APEX2 and refined using SHELXL-2014." -The SHELXL version should be correct, and the reference updated. The program used to solve the structure (SHELXS) should be mentioned and appropriately reference (in this case the 2008 paper is the correct reference -https://doi.org/10.1107/S0108767307043930).   Although the _vrf's now have more sensible comments, they are now no longer syntactically correct.
For example, "_vrf_PLAT242_shelx" has become "_vrf_PLAT242". Omitting the _shelx identifier means the _vrf is no longer associated with the _shelx data block, and the _vrf no longer appears in the checkCIF report. This should be corrected for both structures.
We corrected _vrf's. The _vrf's now appear in the checkCIF report.

Minor correction -_diffrn_radiation_monochromator 'Synchrotron' -should read 'Si double crystal monochromator'
We changed 'Synchrotron' to 'Si double crystal monochromator' Regarding the second structure: Quote "We changed an atomic coordinate of O10 to be connected to C66. The two parts of the outer ligand has been joined together. We also modelled the rotational isorder of two of the t-butyl groups." The reviewer must apologize, there is also rotational disorder of the third t-butyl group, which needs to be modelled.
For this structure, again, refining with global RIGU and SIMU restraints (the SIMU slightly tightened to "SIMU 0.01 0.02 2") results in overall better quality thermal parameters, and removes the need entirely for any DAMP and XNPD restraints, and EADP constraints. An updated SHELX file has been appended to the report, with thermal paramter constraints and disorder modelled.
As with the first structure, now we have understood what we should do from the attachment.
We did the final refinement based on the attachment. We additionally omitted one reflection using OMIT instruction. Many alerts as well as many restraints were successfully removed by this refinement method. We really appreciate the help from the reviewer.
Additional points from the manuscript: Page 19, line 16: "The diffraction intensity data up to 1.00 Å was measured..." -in the structure refinement the applied data limit was 0.9 Å. It is also implied in the CIF that data was collected to a higher angle, however was cut down during structure refinement do to poor quality of diffraction at higher angles. The manuscript should be updated to reflect this.
Thank you for your pointer. We changed the manuscript as follows.
· Page 17, line 16 (Original) The diffraction intensity data up to 1.00 Å was measured