NHase in halophilic archaeon DL31

I notice that both an alpha chain and a beta chain of a cobalt centred NHase have been recently annotated (14 June 2012) in the genome of the rather unprepossessingly named "halophilic archaeon DL31". It must be quite an elusive beast because Google Images doesn't deliver a picture of anything suitable.
Halophilicity quite often comes with tolerance to alkaline conditions, which can mean all sorts of utility in the world of synthetic biocatalysis. As I mentioned in the "No extremes" post we published with Roger Sheldon's group on the synthetic potential of one of these halophilic NHases, but this is a good example of a genome-sourced potential halophilic NHase. BLASTing both the alpha and beta chains against prokaryotic genomes shows that the beta chain is most similar to beta chain of Caldalkalibacillus thermarum TA2.A1 (I think we can guess the proclivities of this microbe) with 3 of the next 4 best hits being the sequences from the crystal structures 3HHT, 2DPP and 1V29 all of which are moderately thermophilic.

Interestingly there is a slightly different aspect to the alpha chain BLAST: Pseudonocardia dioxanivorans and Rhodopseudomonas palustris make up the top hits before Caldalkalibacillus thermarum TA2.A1 turns up.

Aligning cobalt centre NHases

I have been using the FATCAT protocol to look at structural alignments of differing cobalt centred NHases. This is available as a tool here.
Here is a picture of the alpha chains of 1V29 and 1 IRE aligned.

And here is a picture of the alignment of the beta chains. A bit more variation there.

More automated modelling

I had a go at trying to get SWISSmodel to produce a structure prediction for the 3D structure of the single subunit NHase of Monosiga brevicollis. It had a fair stab at it, as I would expect since there are existing PDB files of alpha and beta subunits which match the relevant portions of the chain, but still it isnt right. The placement of the metal ion is the immediate indicator for me that it has gone wrong- I would want the active site in the middle not on an exposed edge.



Considering the similarity of the sequence to the alpha subunit to that in the PDB file 1IRE (left below), and the beta subunit in PDB file 1V29 (middle below), I was expecting something a bit like those to thrown together, with some guess at the histidine-heavy chain linking the two ends (poking out to the right on my guesstimate shown below right).



Salt bridges and hydrogen bonding

I have been playing with PISA tool as it offered insight into size and nature of the interface between the two subunits in NHases. I stuck in the PDB data for the most distinct cobalt-centred NHases- 1UGP, 1V29 and 3HHT, and PISA suggested that these have 61, 67 and 65 hydrogen bonds (respectively) between the alpha and beta subunits, and between 26 and 29 salt bridges across this interface. Compare these numbers with that for 2QDY (the iron-centred NHase from AJ270)... 51 hydrogen bonds and 8 salt bridges. That looks quite a difference... sadly things might not be so clear-cut in grouping these enzymes by metal centre- the newest PDB for 3QXE gives an average of 53 hydrogen bonds and 15 salt bridges, and that is cobalt-centred. Furthermore I do wonder, with such a small set of structures, whether we are seeing a distinction based on the fact that the first three cobalt-centred PDBs are of enzymes specifically tagged as "thermophilic", and AJ270 and Pseudomonas putida aren't. [I also looked at 2DPP and got 62 hydrogen bonds and 27 salt bridges].

PDBs on a phylogenetic tree

There is a figure of a phylogenetic tree in the Prasad Balla review from 2010 which gives a separation of the different currently known sequences. I have very roughly stitched the tree together from the two parts given on two pages of the review and then I have dropped on top the rough positions of the three most distinct cobalt centred PDB files (1V29, 1UGP and 3QXE) and one of the rhodococcus iron centred files (1AHJ). It is interesting that workers seem to be producing PDB files solely from the right hand half of this tree.

3QXE- the new NHase pdb file- from Pseudomonas putida

I thought I'd have a look to see how different the new pdb file was from those already in the public domain.
First of all a ClustalW alignment of the alpha chains shows that it has a score of 52 with 1V29 and a score of 54 with 1UGP. (For reference, 1V29 and 1UGP have a score of 59 between them). The numbers might not be completely robust because I note that the modified cysteines are not dealt with the same in the three amino acid sequences. The central portions of the sequence are very similar, with most of the diversity at the ends (see below: 1V29 then 1UGP and finally 3QXE)

Using DeepView's RMS measure of 3D similarity after an Iterative Magic Fit of  the new pdb with the files for 1V29 and 1UGP enzymes you get two numbers of approximately 1-1.05 (for reference between 1V29 and 1UGP the value is 1.15). This all suggests that this new file describes a new distinct structural arrangement.

Looking at the NHase from Monosiga brevicollis

I have been looking at the sequence for the single subunit NHase from the eukaryote Monosiga brevicollis. There are four things that spring to mind of interest:
1. The thing has joined the two prokaryotic subunits together in the fashion B to A.
2. The usual signature bit of the sequence says it is a cobalt centred NHase, and the alpha subunit is very similar to that shown by the two cobalt centred NHases with pdb files 1V29 and 1UGP.

 3. The joint between the alpha and beta subunits has a LOT of histidines in it, which explains why it took a long time to come off the nickel affinity column!

4. The surface loop which is nice differentiating feature between cobalt-centred and iron centred NHase (which dont have it) is very different with this beast. The picture below is of 1V29, with the loop indicated.

How many clearly different structures of NHase are reported as pdb files?

Another helpful thing that the Iterative Magic Fit function within DeepView can do is provide a numerical value (based on a RMS calculation) for the similarity of the three dimensionality of the enzymes being overlaid. After concentrating on enzymes which are clearly NHases, discarding structures which are clearly mutants of an existing structure (which is either obvious from the title or the combination of author and date of submission), you get left with 13 distinct pdb files [1AHJ, 1IRE, 1UGP, 1V29, 2AHJ, 2CYZ, 2CZ6, 2D0Q, 2DPP, 2QDY, 2ZPB, 3A8O and 3HHT]. I then IMFed every one of these against each other to get a matrix of values for similarity. Basically if they had a value of less than 0.5, they look pretty much the same, greater than 1 than they look noticeably different.
The results of this showed that ALL the iron centred enzymes have RMS of <0.5 with each other which is not entirely surprising since they are all from Rhodococcus species. They all have RMS of greater than 1 for all the cobalt centred enzymes. There are two distinctly different cobalt centred arrangements: the Pseudonocardia thermophila pair of 1IRE and IUGP, and the trio of Bacillus structures of 1V29, 2DPP and 3HHT. Within these groupings they have RMS of less than 0.5, and between the groupings RMS of greater than 1.
Effectively this says to me that there are three basic structures for NHase known currently- one iron based and two cobalt based. Below is an excerpt from a table holding all the results from these calculations which shows this clustering.

PDB ID	1AHJ	1IRE
1AHJ		1.06
1IRE	1.06
1UGP	1.06	0.07
1V29	1.19	1.14
2AHJ	0.45	1.09
2CYZ	0.44	1.07
2CZ6	0.47	1.09
2D0Q	0.45	1.09
2DPP	1.22	1.16
2QDY	0.46	1.09
2ZPB	0.44	1.08
3A8O	0.47	1.11
3HHT	1.22	1.15