U. Krug et al. / FEBS Letters 587 (2013) 460–466
465
45 and 36 sꢀ1, the turnover numbers of 5NT–NDom for ADP and
ATP, respectively, are also decreased compared to the full length
protein, but only by a factor of 13–20. These large differences in
the cleavage of AMP compared to ADP and ATP by 5NT–NDom
are likely to originate from the nature of the hydrolyzed bonds to
the terminal phosphate group. The phosphoanhydride bonds in
ADP and ATP are more labile than the ester bond cleaved in AMP.
The phosphate groups of ADP and AMP are good (thermodynami-
cally stable) leaving groups. In contrast, the alcoholate resulting
from AMP cleavage is thermodynamically unstable and hence a
weak leaving group. An activated phosphoester bond is also pres-
ent in pNPP. The aromatic ring with the nitro group withdraws
electrons from the O–P-bond and stabilizes the phenolic leaving
group. Thereby, it facilitates the nucleophilic attack on the phos-
phate group and expulsion of the p-nitrophenolate. The kcat of
5NT–NDom for pNPP is threefold higher than the kcat for ADP
and ATP (Table 1, Fig. 4).
Interestingly, the turnover of pNPP by 5NT–NDom is more than
threefold higher in comparison to the full-length enzyme. This
observation might be caused by a better accessibility of the active
site cavity in 5NT–NDom. In full-length 5NT the presence of the C-
terminal domain might hinder active site access for the small sub-
strate, which does not require the C-terminal domain for binding.
For 5NT–CDom no hydrolysis of the nucleotides or pNPP was
observed although Ca2+ and Co2+ were present in the assay buffer.
The C-terminal domain alone is catalytically inactive due to the
lack of the dimetal center and the Asp-His dyad. Binding of the
three arginines R375, R379 and R410 to the phosphate groups of
the substrates alone does not confer any measurable rate enhance-
ment of the hydrolysis reaction.
phosphate group by the arginines of the C-terminal domain and/
or the orientation of the AMP substrate for productive binding by
the ribose and the adenine base binding pocket of the C-terminal
domain are obviously essential for efficient hydrolysis of AMP.
For efficient expulsion of the leaving group, protonation of the
instable alcoholate is necessary. In the assumed Michaelis complex
based on the AMPCP cocrystal structure, there is no protein residue
positioned to fulfill this role (Fig. 1B). The leaving group might be
protonated by a water molecule. Alternatively, H117 might provide
a proton, but it is 5.0 Å away from the leaving group (in the AMPCP
model). This distance might be smaller in the transition state com-
plex. It is also possible that the AMPCP structure is no good model
for the productive AMP binding mode. In any way, H117 is also
present in 5NT–NDom and thus cannot explain the different reac-
tivity of full-length 5NT and 5NT–NDom towards AMP.
As no structural data of 5NT–NDom in complex with a substrate
are available up to now we can not rule out that new interactions
are formed to stabilize the binding of the substrates. However, the
kinetic data with Km values in the mM range imply that these
interactions would not be as strong as in full-length 5NT. In agree-
ment with the suggestion that in full-length 5NT the artificial sub-
strate pNPP does not bind to the C-terminal domain [7], the
specificity constant of full-length 5NT for pNPP is comparable to
those of 5NT–NDom for ATP and ADP.
4.2. Structure and function of the two 5NT domains in related proteins
5NT belongs to the calcineurin superfamily of phosphatases
containing a dimetal center in the active site [5,16]. However, only
the N-terminal domain is related to this protein superfamily. In
some of these metallophosphatases, the N-terminal domain is cat-
alytically active without the requirement of additional domains,
e.g., in the mammalian purple acid phosphatases [17–20]. This
demonstrates the ability of this domain to independently catalyze
the hydrolysis of phosphoesters or anhydrides. The dinuclear metal
ion center and the catalytic H117 are the conserved essential ele-
ments for catalysis [4].
Formation of the full-length protein in solution by mixing equi-
molar amounts of the two domains was neither observed in a na-
tive PAGE nor by any increase in the hydrolytic activity in the
activity assays (data not shown). These findings indicate that the
interactions between the two domains are too weak to reconstitute
the full-length protein from the two folded individual domains
without the presence of a covalent linker connecting both domains.
The structure of the C-terminal domain is almost unique to the
50-nucleotidases. A DALI search [21] for related domains revealed
that the C-terminal domain is currently only found in the 50-nucle-
4. Discussion
otidases and the thiosulfohydrolase SoxB (alignment of 169 Ca,
4.1. Implications for the catalytic mechanism
rmsd of 2.4 Å, PDB:2WDF). SoxB is an enzyme hydrolyzing a sul-
fur–sulfur bond. Both domains of SoxB are related to 5NT. Despite
the similarities and the presence of a dimetal center, SoxB shows
considerable differences. No domain motion was observed, the cat-
alytic dyad is missing and the active site residues in the C-terminal
domain are arranged differently [22]. As only the binding of the
substrate analog thiosulfate is shown, not many interactions with
the C-terminal domain are characterized. Nevertheless, it is very
likely that similarly to 5NT the C-terminal domain is responsible
for binding and positioning of the substrate and not for the core
catalytic steps.
Furthermore, independent expression of the two domains pro-
vides an effective strategy for NMR signal assignment. The expression
of each domain in a 2H/13C/15N background is possible with a yield
comparable to the 15N-labeled protein (data not shown). The binding
of AMPCP provided NMR spectra of comparable quality allowing
structural and dynamic studies at residue resolution for the holo
enzyme. This opens the way to NMR experiments to study the
dynamics of the unique domain motion of the enzyme in solution.
In summary, our data show that catalysis in 5NT depends
mainly on the N-terminal domain which is structurally related to
a variety of other dinuclear metallophosphatases. On the other
hand, the presence of the evolutionary not widely distributed C-
terminal domain enhances the catalytic efficiency for non-acti-
vated substrates. It also provides the specificity for nucleotide
The differences in the catalytic behavior of full-length 5NT and
5NT–NDom are related to the active site structure, which is formed
at the domain interface (Fig. 1B). The adenine base and the ribose
are bound by the C-terminal domain whereas the phosphate
groups form contacts to N116, H117 and metal ion 2 of the N-ter-
minal domain as well as R375, R379 and R410 from the C-terminal
domain [4]. Assuming a similar binding mode of the terminal phos-
phate group to 5NT–NDom the interactions of the phosphate
groups must be weakened compared to full-length 5NT as the argi-
nines are not present to coordinate the phosphate groups and to
provide electrostatic stabilization. An interesting result of the pres-
ent study is that in contrast to earlier suggestions [4] the data
determined for 5NT–NDom demonstrate that R375, R379 and
R410 are not necessary for binding of the substrate phosphate
groups or for transition state stabilization. This is indicated by
the high catalytic turnover numbers of 5NT–NDom for ATP, ADP
and pNPP. Based on the proposed mechanism [4], these results
demonstrate that in 5NT–NDom H117 and the metal ions are suf-
ficient for transition state stabilization of these substrates.
As mentioned in the previous section, the greatly reduced activ-
ity of the N-terminal domain to hydrolyze AMP (as opposed to ATP,
ADP and pNPP) is likely due to the less activated ester bond and
poor alcoholate leaving group in AMP. The coordination of the