Pro-metastatic collagen lysyl hydroxylase dimer assemblies stabilized by Fe2+-binding

Collagen lysyl hydroxylases (LH1-3) are Fe2+- and 2-oxoglutarate (2-OG)-dependent oxygenases that maintain extracellular matrix homeostasis. High LH2 levels cause stable collagen cross-link accumulations that promote fibrosis and cancer progression. However, developing LH antagonists will require structural insights. Here, we report a 2 Å crystal structure and X-ray scattering on dimer assemblies for the LH domain of L230 in Acanthamoeba polyphaga mimivirus. Loop residues in the double-stranded β-helix core generate a tail-to-tail dimer. A stabilizing hydrophobic leucine locks into an aromatic tyrosine-pocket on the opposite subunit. An active site triad coordinates Fe2+. The two active sites flank a deep surface cleft that suggest dimerization creates a collagen-binding site. Loss of Fe2+-binding disrupts the dimer. Dimer disruption and charge reversal in the cleft increase Km and reduce LH activity. Ectopic L230 expression in tumors promotes collagen cross-linking and metastasis. These insights suggest inhibitor targets for fibrosis and cancer.


Reviewer #2 (Remarks to the Author):
Very interesting work on the lysyl-hydroxylase (LH) domain derived from the giant virus Acanthamoeba polyphaga mimivirus. Surprisingly, this virus contains a set of coll agen-like molecules, which are lysyl-hydroxylated and further glycosylated by the same protein L230. The crystal structure of the LH domain of L230 is reported here. Although phylogenetically distant, the viral LH domain might serve as a model to invertebrate as well as vertebrate LHs as it has high level of conservation in secondary structure, and in Fe2+ and 2-oxoglutarate coordinating residues. Moreover, it might also represent a useful model for studying dimerization, as found in this study. A novel cla im of this study is that dimerization is driven by the LH domain itself, opposed to what was reported previously for human and mouse LHs, where dimerization was attributed to a sequence between the glycosyltransferase activity and the lysyl hydroxylase activity domains. The authors have to clearly address this discrepancy. Two other levels of complications (not even mentioned in the manuscript!) are role of FKBP65 in driving dimerization of LH2, as well as FKBP65-dependent activity of LH2 (previous work of the authors). All these somewhat diminishes significance of L230 LH domain as a model for studying human LH2. Furthermore, activity and substrate specificity of viral L230 are quite different from LH2, which makes experiments with ectopic L230 expression q uite limited. Other major points: 1. As the authors, I also found very surprising that dimerization was lost following mutation of the Fe2+-binding ligands or Fe2+ removal, although Fe2+ is not involved in the dimerization interface. Nevertheless, Fe2+ binding might be important for formation of the tertiary structure. I found intriguing that at least one mutant H825A (one of Fe2+ coordinating residues) showed abnormal elution profile from gel-filtration column though with normal monomer mass as determine d by light scattering. To better address role of Fe2+ in dimer formation (or even folding!) more experiments are required. These might include C D and fluorescence spectroscopy in order to assess secondary and tertiary structures with and without Fe2+ (and wt vs mutants). Finally, role of Fe2+ in regulating prometastatic lysyl hydroxylase dimer assemblies, as claimed in the manuscript title, has also to be confirmed for LH2. 3. Based on SAXS derived protein envelops the authors state that L230 LH domain fo rms tail-to-tail dimers similar to full-length human LH2. Very strong statement without any analysis and explanation. 4. L873D mutation leads to apparent decreased (but not loss of activity as stated in figure 2) activity for real collagen chain. Nevertheless, L873D activity becomes comparable to those for synthetic peptide, where the mutation has no effect. I would suggest to have the same scales for Y -axis in panels f and g (or combine the panels) to highlight it. This will support an idea that dimerizat ion increases binding of "lengthy" or oligomerized collagen chains, which subsequently leads to increased "apparent activity". 5. An idea that substrate-binding groove positions collagen between subunit active sites is not obvious as active sites are located in an anti-parallel mode. Is there a possibility to load both sites in right orientation by forming a loop in a single collagen chain? Or maybe it is more biologically relevant to model binding of two adjacent chains (out of three chains forming a trip le helix in a zipper fashion)? Guo et al have provided the first crystal structure information for an important family of collagen modifying enzymes in their manuscript. This work will influence the drug disc overy efforts to target this enzyme which is known to play an important role in the metastasis of multiple cancer types. While the work is exciting and the crystal structure dataset is of relatively high quality based on the statistics presented in supplementary tables, there are several key points to address before publication.
1. In figure 1 the authors indicate the the Fe binding site and the 2-OG binding sites are both required for enzymatic activity. However, in Figure 3 they show that only the Fe bi nding site is required for dimerization. First, it would be useful if the authors could comment on the role of the 2 -OG site since its clearly important for activity. It would also be appropriate to perform the collagen crosslink assays in figure 4 as well as the in vivo lung colonization assay in figure 5 with both the Fe and 2 -OG binding site mutants . Doing so would clearly show that dimerization and the specific Fe and 2 -OG binding sites that have been called out in this manuscript have functional impor tance. Figure 5A specifically details that only tumors on the surface of the lung are quantified. This approach seems flawed. A more thorough investigation would require sectioning of the lungs and assessing tumor foci found throughout the organ. Additionally the authors could stain for expression of L230 to be sure that protein expression can verified in vivo. L230 expression should also be validated in the migration, invasion, colony forming, and proliferation studies included in this figure by Western blot, rather than PC R included in the supplemental information. However, that may be difficult due to the viral origin of the protein (see below).

The lung colonization study in
3. Lastly, the authors use the viral LH (L230) for their biological valid ation studies in figure 5. It is unclear why they did not create the relevant mutants in the conserved sites of human LH1 -3 and express those constructs in lung cancer cells to assign the necessary physiological relevance to their studies. Furthermore, it would be appropriate to silence LHs in these lung cancer cells to ask whether they could reduce the number of metastatic foci #s from those found in the control cells. Overexpression assays are generally not thought reliable in isolation. The corresponding gene silencing assays would add depth and validity to those observations.

Reviewer #1
No critiques to address.

Reviewer #2
A novel claim of this study is that dimerization is driven by the LH domain itself, opposed to what was reported previously for human and mouse LHs, where dimerization was attributed to a sequence between the glycosyltransferase activity and the lysyl hydroxylase activity domains. The authors have to clearly address this discrepancy.
On the basis of homology modeling, sequences between the glycosyltransferase and LH domains previously shown to be required for LH dimerization (Heikkinen et al Matrix Biol. 30:27-33, 2011) are not located on the dimerization interface of L230. This conclusion is supported by evidence from LH protein cross-linking experiments (Heikkinen et al Matrix Biol. 30:27-33, 2011). How these sequences regulate dimerization remains unclear, but it is possible that they contain binding sites for the peptidyl prolyl isomerase FKBP65 or other factors that promote LH2 dimerization. Unlike full-length LH2, which requires FKBP65 to form dimers, the L230 catalytic domain formed dimers in the absence of FKBP65, raising the possibility that FKBP65 is necessary for dimerization of the full-length LHs but not isolated catalytic subunits. We have added a paragraph to the Discussion section to more fully address this point.
Two other levels of complications (not even mentioned in the manuscript!) are role of FKBP65 in driving dimerization of LH2, as well as FKBP65-dependent activity of LH2 (previous work of the authors). All these somewhat diminishes significance of L230 LH domain as a model for studying human LH2.
As discussed above, the reviewer has raised an important point. We have included a statement in the Discussion to address it.
Furthermore, activity and substrate specificity of viral L230 are quite different from LH2, which makes experiments with ectopic L230 expression quite limited.
The limited available comparative data do show that L230 and human LHs have distinct substrate specificities (Rutschmann et al. Appl Microbiol Biotechnol 98:4445-4455, 2014), which does limit the value of L230 as a model of human LHs. We have added a statement to the Discussion section to make this point.

As the authors, I also found very surprising that dimerization was lost following mutation of the Fe2+-binding ligands or Fe2+ removal, although Fe2+ is not involved in the dimerization interface. Nevertheless, Fe2+ binding might be important for formation of the tertiary structure. I found intriguing that at least one mutant H825A (one of Fe2+ coordinating residues) showed abnormal elution profile from gel-filtration column though with normal monomer mass as determined by light scattering. To better address role of Fe2+ in dimer formation (or even folding!) more experiments are required. These might include CD and fluorescence spectroscopy in order to assess secondary and tertiary structures with and without Fe2+ (and wt vs mutants).
We performed circular dichroism spectroscopy and found no evidence that L230 secondary structure was altered by mutations in the Fe 2+ -binding ligands or by histidine protonation to deplete Fe 2+ . We have added these findings to the Results section. Why the elution profile was abnormal for one Fe 2+ -binding ligand (H825) but not the others is unclear. One possibility is that the mutation caused a minor change in conformation that resulted in a solvation change.
Finally, role of Fe2+ in regulating pro-metastatic lysyl hydroxylase dimer assemblies, as claimed in the manuscript title, has also to be confirmed for LH2.
We generated full-length recombinant human LH2 protein that contains a mutation (D689A) in the amino acid triad and showed that this mutation leads to loss of dimerization. We have added these findings to the Results section.

Based on SAXS derived protein envelops the authors state that L230 LH domain forms tail-to-tail dimers similar to full-length human LH2. Very strong statement without any analysis and explanation.
We agree and now simply state that, on the basis of findings from SAXS analysis, L230 forms dimeric complexes.
L873D mutation leads to apparent decreased (but not loss of activity as stated in figure 2) activity for real collagen chain. Nevertheless, L873D activity becomes comparable to those for synthetic peptide, where the mutation has no effect. I would suggest to have the same scales for Y-axis in panels f and g (or combine the panels) to highlight it. This will support an idea that dimerization increases binding of "lengthy" or oligomerized collagen chains, which subsequently leads to increased "apparent activity".
We appreciate these helpful suggestions and have modified the text and Figure 2 accordingly.
An idea that substrate-binding groove positions collagen between subunit active sites is not obvious as active sites are located in an anti-parallel mode. Is there a possibility to load both sites in right orientation by forming a loop in a single collagen chain? Or maybe it is more biologically relevant to model binding of two adjacent chains (out of three chains forming a triple helix in a zipper fashion)?
The reviewer's observation that the active sites are in anti-parallel mode does indeed raise the suggested possibilities. We have modified the text accordingly.

Both structures were deposited with Fe3+ ions (FE (III) ION), should be Fe2+ (FE2 (II) ION).
We appreciate the reviewer noting the typographical error in the structure deposits. This has been corrected.

Line 135. Not only in human!
The word "human" has been removed.

Reviewer #3
In figure 1 the authors indicate the the Fe binding site and the 2-OG binding sites are both required for enzymatic activity. However, in Figure 3 they show that only the Fe binding site is required for dimerization. First, it would be useful if the authors could comment on the role of the 2-OG site since its clearly important for activity.
We agree that our findings raise the possibility that 2-OG-binding ligands regulate dimerization. However, in Figure 3b, L230 dimerization was preserved following mutation of the 2-OGbinding ligand (R887). Thus, our findings suggest a specific role for Fe 2+ -binding ligands in dimerization.
It would also be appropriate to perform the collagen crosslink assays in figure 4 as well as the in vivo lung colonization assay in figure 5 with both the Fe and 2-OG binding site mutants . Doing so would clearly show that dimerization and the specific Fe and 2-OG binding sites that have been called out in this manuscript have functional importance.
To prove that the Fe 2+ -and 2-OG-binding site mutants, which are enzymatically inactive, lack biological activity, we stably transfected the Fe 2+ -and 2-OG-binding site L230 mutants into lung cancer cells and showed that they do not increase tumor cell migration or invasion, which are key pro-metastatic properties. These findings have been added to the Results section. Figure 5A specifically details that only tumors on the surface of the lung are quantified. This approach seems flawed. A more thorough investigation would require sectioning of the lungs and assessing tumor foci found throughout the organ.

The lung colonization study in
In the orthotopic lung tumor model reported here, tumor cells that metastasize to the contralateral lung lodge in alveolar capillary beds in peripheral lung tissues, generating metastatic deposits that are visible on the lung surface. We agree that counting tumors on the lung surface does not accurately assess total numbers of metastatic deposits in the contralateral lung. However, when used to assess experimental and control cohorts in the same manner, this approach does accurately reflect relative metastatic activity. We have reported this quantitative approach widely ( Additionally the authors could stain for expression of L230 to be sure that protein expression can verified in vivo. L230 expression should also be validated in the migration, invasion, colony forming, and proliferation studies included in this figure by Western blot, rather than PCR included in the supplemental information. However, that may be difficult due to the viral origin of the protein (see below).
Unfortunately, there are currently no anti-L230 antibodies available through commercial or proprietary sources.
Lastly, the authors use the viral LH (L230) for their biological validation studies in figure 5. It is unclear why they did not create the relevant mutants in the conserved sites of human LH1-3 and express those constructs in lung cancer cells to assign the necessary physiological relevance to their studies.
We have already reported that a mutation in an Fe 2+ -binding ligand (D689) in human LH2 leads to loss of enzymatic activity (Guo et al Arch Biochem Biophys 618:45-51, 2017), and the Simon lab showed that the same LH2 mutant is unable to increase tumor cell migration (Eisenger-Mathason et al, Cancer Discovery 3:1190-1205, 2013. We have stated these points in the Discussion section.
Furthermore, it would be appropriate to silence LHs in these lung cancer cells to ask whether they could reduce the number of metastatic foci #s from those found in the control cells. Overexpression assays are generally not thought reliable in isolation. The corresponding gene silencing assays would add depth and validity to those observations.
We have already reported that short hairpin RNA-mediated silencing of LH2 in human and murine lung cancer cells leads to reduced tumor cell migratory, invasive, and metastatic activities (Chen et al. J Clin Investig 125:11471162, 2015).