Balancing Specificity and Promiscuity in Enzyme Evolution: Multidimensional Activity Transitions in the Alkaline Phosphatase Superfamily

Article abstract of DOI:10.1021/jacs.8b10290

Highly proficient, promiscuous enzymes can be springboards for functional evolution, able to avoid loss of function during adaptation by their capacity to promote multiple reactions. We employ a systematic comparative study of structure, sequence, and sub

Full text of DOI:10.1021/jacs.8b10290

Subscriber access provided by University of Winnipeg Library
Article
Balancing Specificity and Promiscuity in Enzyme Evolution: Multidimensional
Activity Transitions in the Alkaline Phosphatase Superfamily
Bert van Loo, Christopher D Bayer, Gerhard Fischer, Stefanie Jonas, Eugene Valkov, Mark
F. Mohamed, Anastassia Vorobieva, Celine Dutruel, Marko Hyvonen, and Florian Hollfelder
J. Am. Chem. Soc., Just Accepted Manuscript • DOI: 10.1021/jacs.8b10290 • Publication Date (Web): 30 Nov 2018
Downloaded from http://pubs.acs.org on December 1, 2018
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Balancing Specificity and Promiscuity in Enzyme Evolution:
Multidimensional Activity Transitions in the Alkaline Phosphatase
Superfamily
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Bert van Loo, Christopher D. Bayer, Gerhard Fischer, Stefanie Jonas, Eugene Valkov, Mark F.
Mohamed, Anastassia Vorobieva, Celine Dutruel, Marko Hyvönen* and Florian Hollfelder*
Department of Biochemistry, University of Cambridge, Cambridge, United Kingdom
*For correspondence: fh111@cam.ac.uk; mh256@cam.ac.uk
Abbreviations: AS: arylsulfatase; PMH: phosphonate monoester hydrolase, NPP: nucleotide
phoshodiesterase/pyrophosphatase; AP: alkaline phosphatase; PAS: Pseudomonas aeruginosa
arylsulfatase.
1
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 2 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
ABSTRACT
Highly proficient, promiscuous enzymes can be springboards for functional evolution, able to
avoid loss of function during adaptation by their capacity to promote multiple reactions. We
employ systematic comparative study of structure, sequence and substrate specificity to track the
evolution of specificity and reactivity between promiscuous members of clades of the alkaline
phosphatase (AP) superfamily. Construction of a phylogenetic tree of protein sequences maps out
the likely transition zone between arylsulfatases (ASs) and phosphonate monoester hydrolases
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(PMHs). Kinetic analysis shows that all enzymes characterized have four chemically distinct
1
1
17
phospho- and sulfoesterase activities, with rate accelerations ranging from 10 -10 -fold for their
9
12
primary and 10 -10 -fold for their promiscuous reactions, suggesting that catalytic promiscuity is
widespread in the AP-superfamily. This functional characterization and crystallography reveal a
novel class of ASs that is so similar in sequence to known PMHs that it had not been recognized
as having diverged in function. Based on analysis of snapshots of catalytic promiscuity ‘in
transition’ we develop possible models that would allow functional evolution and determine
scenarios for trade-off between multiple activities. For the new ASs we observe largely invariant
substrate specificity that would facilitate the transition from ASs to PMHs via trade-off-free
molecular exaptation, i.e. evolution without initial loss of primary activity and specificity toward
the original substrate. This ability to bypass low activity generalists provides a molecular solution
to avoid adaptive conflict.
2
ACS Paragon Plus Environment

Page 3 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
INTRODUCTION
The view that enzymes are specific for one substrate has been profoundly revised in recent years
by the realization that many enzymes are catalytically promiscuous.1-20 In some cases the
observation of promiscuous activities can be attributed to the modest catalytic requirements of the
promiscuous reactions, but more and more studies show promiscuous activities with large rate
enhancements for chemically demanding reactions requiring substantial stabilization of the
transition state.14,18,21-25 This observation raises the question how enzymes can satisfy the stringent
and, at the level of molecular recognition of ground and transition states, often divergent
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
requirements for catalysis of multiple reactions. Promiscuity has been postulated to play a role in
the evolution of new enzymatic functions, by providing the organism with an initially moderate
activity, which shortens the distance to the point at which this activity confers a selective
advantage.1,9,20,26,27 A promiscuous enzyme may thus serve as a starting point for the evolution of
9
a new enzyme with a different activity (possibly even prior to gene duplication), so that catalysts
e.g. for the breakdown of newly encountered compounds (e.g. xenobiotics)28 can be created
relatively rapidly.
1
O'Brien and Herschlag (1999) were the first to link data on enzyme promiscuity to its
mechanistic underpinnings. However, despite the increasing evidence for the generality of enzyme
promiscuity,3,4,8 a systematic, quantitative analysis of this historically underreported phenomenon
and its role in evolution is only starting to emerge. In this work, we studied a set of phylogenetically
related phosphonate monoester hydrolases (PMHs) and newly identified class of arylsulfatases
(ASs) that are part of the alkaline phosphatase (AP) superfamily in order to explore transitions
between multiple enzymatic activities characteristic for this superfamily. This approach enables
3
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 4 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
us to probe to what extent the phylogenetic annotation and the experimentally determined activities
coincide. The diagnostic substrates 1-4 (top line, Figure 1) represent four distinct hydrolytic
reaction types catalyzed by AP-superfamily members and differ in their reactive functional groups,
substrate charges, sizes and hydrophobicities, and the nature of the transition states involved in
their uncatalyzed hydrolyses. The reactions catalyzed by these enzymes are thermodynamically
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
demanding, with half-lives ranging from 200 days to 10 years. The AP superfamily is
characterized by crosswise catalytic promiscuity, i.e. the ‘primary’ reactions of one enzyme are
found to be promiscuous reactions catalyzed by other family members13,14,18,21,23-26,29,30 (Figure 1),
as well as additional activities.18,31 We use substrates 1-4 to assign a function to sub-groups of the
superfamily (e.g. potentially as phosphate monoesterases, phosphodiesterases,30 phosphonate
monoester hydrolases,32 or arylsulfatases,33-36), leading to classification of newly identified
enzymes according to their respective top activity, even though the natural substrates may not be
known.
Studies with such representative substrates have been used to characterize trajectories of directed
evolution experiments37-39 and to infer specificity changes in evolution1,24,40-42 and epistatic
constraints.43,44 This matrix of substrate classes and new superfamily members allows us to address
the following questions for these hydrolases: (i) how is protein sequence phylogeny correlated to
enzyme activity and the specificity determinants, (ii) how will multiple activities of these
promiscuous enzymes trade-off against each other or will promiscuity be maintained in all
members of this superfamily?
Structurally characterized members of the AP superfamily (Figure 1) highlight similarities
in fold and active site residues.29,45,46 However, understanding the specificity determinants within
this common fold is hampered by considerable sequence variation outside the active site. This
4
ACS Paragon Plus Environment

Page 5 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
ambiguity is highlighted by two pairwise comparisons: e.g. by BcPMH and PAS (pairwise
sequence identity: 27%). These two AP-superfamily members have different primary activities,
but the majority (65%) of the amino acids align structurally and key active site residues (in
particular the crucial formylglycine (fGly) nucleophile and the leaving group stabilizing residues)
are in virtually identical positions (Figure 2, Table S2-S3). Conversely, for two superfamily
members with the same primary activity (e.g. the sulfatases PAS vs human arylsulfatase C
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(
HsASC)), similar pairwise structural alignment scores arise (67% alignment, 28% identity; Table
S2-S3). These comparisons suggest that enzymes with different primary functions can be as
closely related in sequence and structure as enzymes that share the same primary activity and raise
the question which sequence, fold or active-site features are indicative for specificity and
promiscuity. Beyond assignment of function, the identification of specificity determinants would
also give insight into the transition of function within the superfamily. We focus on the sequence
space where a transition between activities of the AP superfamily might occur. This space is
populated by enzymes that are more closely related to PMHs (in terms of protein sequence
phylogeny and structure) than to any other ASs of the AP superfamily, but their substrate
specificity profiles show that they are ASs, correcting misannotation, and locating the point of
functional divergence between the two activities.
MATERIALS AND METHODS
Materials. Phosphate monoester 1, phosphonate monoester 3c and sulfate monoester 4
were purchased from Sigma. Phosphate diesters 2a-2c and phosphonate monoesters 3a and 3b
were synthesized in a similar fashion as described previously47-49 (for details see supporting
5
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 6 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
information, SI). All restriction enzymes and T4 DNA ligase were from Fermentas. Vector
pASKIBA5plus and Streptactin resin were purchased from Stratech scientific. Pfu turbo was from
Agilent. Bacterial strains Advenella kashmirensis WT001, Ralstonia metallidurans CH34, Stappia
aggregata IAM 12614 and Silicibacter pomeroyi DSS-3 were purchased from the DSMZ. Cell
material of Agrobacterium radiobacter K84 and purified DNA from Rhodopseudomonas palustris
CGA009 were obtained from ATCC.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Sequence retrieval, multiple sequence alignment and phylogenetic analysis. The
sequences included in the phylogenetic tree of all arylsulfatases (ASs) and phosphonate monoester
hydrolases (PMHs) of known activity (Figure 3A, Figure S1-S5), either from published data or
found in this study (see Table S5 for details on the sequences included), were aligned using T-
coffee (expresso mode), Clustal, MAFFT and MUSCLE (all with default settings). These
alignments were used as input to build a maximum likelihood phylogenetic tree with RAxML
5
0
8
.2.10, using various amino acid substitution matrices with a -model for rate heterogeneity,
estimate proportion of invariable sites and empirical base frequencies (substitution matrix (LG,
VT or WAG) +G+I+F, see legends to Figure S1-S4 for details) running on the CIPRES Science
5
1
Gateway (www.phylo.org/portal2/). The optimal amino acid substitution matrix and parameter
5
2
configuration for tree-building was calculated using ProtTest 3.4.2. . The T-coffee alignment
(expresso mode) was also used to build a Bayesian maximum likelihood tree using MrBayes
(Figure S5, see SI for details). The tree in Figure 3A is based on the T-coffee alignment (expresso
mode), with Figure S1 serving as a legend to Figure 3A.
In order to obtain the genes encoding putative arylsulfatases (ASs) and phosphonate
monoester hydrolases (PMHs), existing sequenced bacterial genomes were subjected to a BLAST-
search, initially using the amino acid sequences of B. caryophili18,53 and R. leguminosarum32
6
ACS Paragon Plus Environment

Page 7 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
PMHs as search sequences. During the course of this study sequences of the enzymes for which
the primary putative activities were confirmed by experimental data were also used as search
sequences. The protein sequences of the resulting hits were aligned using ClustalX and screened
for the presence of putative active site residues. Enzymes in which any of the putative active site
residues were missing were discarded. Based on the nature of the active site residues (Figure 2)
the putative sequences were designated to be either arylsulfatases (ASs) or PMHs. Initial data
indicated that the majority of the hits (>90%) originated from - and -proteobacteria. For
simplicity, we restricted the BLAST-search to these two subgroups of bacteria for the data that
were included in the final alignment. The guide tree from the ClustalX alignment was used to
identify obvious outliers that were subsequently discarded. The final alignment was done with 85
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(putative) PMHs, 95 (putative) ASs (for details see Table S6-S7) and 87 (putative) CSs using the
5
4
3
D-coffee mode of T-coffee as described previously. These alignments were used as input to
5
0
build a maximum likelihood phylogenetic tree with RAxML BlackBox 8.0.24, using the Le &
5
5
Gascuel amino acid substitution matrix (LG) with a -model for rate heterogeneity, estimate
proportion of invariable sites and empirical base frequencies (LG+G+I+F) running on the CIPRES
51
Science Gateway (www.phylo.org/portal2/). Several additional methods (Clustal , MAFFT or
MUSCLE) were employed to create alternative phylogenetic trees, based on the same data and
using the same settings (Figure S8). The 3D-coffee alignment was also used to build a Bayesian
maximum likelihood tree (Figure S9B, SI). The tree shown in Figure S10 was built with RAxML
8
.0.2450 running on the CIPRES Science Gateway51 (www.phylo.org/portal2/) based on 67
positions (as found in the complete alignment) representing all residues within a distance of 4.0 Å
of the 11 putative active site residues in the X-ray structure (and included these 11 positions, see
5
5
Figure S13 for details), using the Le & Gascuel amino acid substitution matrix (LG) with a -
7
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 8 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
model for rate heterogeneity and estimate proportion of invariable sites (LG+G+I). The optimal
amino acid substitution matrices and parameter configuration for tree-building were calculated
using ProtTest 3.4.2.52
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Cloning of arylsulfatase (AS) and phosphonate monoester hydrolase (PMH) encoding
genes. The genes encoding the various PMHs and ASs (Table 1) were amplified by PCR using the
primer pairs described in Table S8. We used either commercially available genomic DNA
(
Rhodopseudomonas palustris CGA900) or whole cell material (all other organisms) of the
respective organisms as a template for the PCR reactions. Primers were used at 0.4 nM in a reaction
-1
®
with 0.2 mM dNTPs and 0.05 U µL Pfu-Turbo DNA polymerase. The temperature program
used for SaAS was 15 min at 95 ºC without polymerase, followed by 30 cycles of 60 s at 95 ºC,
4
5 s at 58 ºC, 180 s at 72 ºC, and finished with 10 min at 72 ºC. For all other sequences, the
temperature program consisted of 15 min at 95 ºC without polymerase, followed by 30 cycles of
-1
60 s at 95 ºC, 45 s at 68 ºC - 0.5 ºC cycle (each cycle the temperature of this segment was lowered
by 0.5 ºC), 180 s at 72 ºC, and finished with 10 min at 72 ºC. The PCR products were digested
with various restriction enzymes (Table S8) and subsequently ligated into appropriately digested
pASKIBA5plus plasmid DNA using T4 DNA ligase. The ligation mixture was transformed into
Escherichia coli TOP10 using electroporation. Cells were plated on LB medium containing
ampicillin and resulting colonies were checked for insert using a PCR reaction with Taq
polymerase and colony material as the template. Plasmid DNA was extracted from positive clones
and confirmed by sequencing.
Protein production and purification. Expression of the respective genes in the
pASKIBA5plus vector results in a translational fusion with an N-terminal Strep-tag. Strep-tagged
BcMPH and RlPMH were produced and purified as described previously.18,32 The other PMHs and
8
ACS Paragon Plus Environment

Page 9 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
ASs were produced in E. coli BL21 (DE3) by co-expressing the formyl glycine generating enzyme
5
6
18,32
(
FGE) from Mycobacterium tuberculosis H37v (MtbFGE) from pRSFDuetMtbFGE plasmid.
E. coli BL21(DE3) were typically grown in 750 mL of 2×YT medium with ampicillin (100 mg L-
1)
8oC. Once the culture had reached the desired temperature, expression of MtbFGE, was induced
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
-1
o
and kanamycin (50 mg L ) at 37 C to an A600nm ~ 0.4, at which point the culture was cooled to
2
by the addition of 1 mM IPTG, approximately 30 minutes prior to the induction of expression of
-1
the strep-tagged PMH/AS by adding up to 200 g L anhydrotetracycline followed by overnight
growth at 28°C. Cells were harvested by centrifugation and resuspended in 50 mM Tris-HCl pH
8.0. The resuspended cells were lysed either by an Emulsiflex-C5 high pressure homogenizer or
by sonication and cell-free extract was obtained by centrifugation at 40,000 × g for 90 minutes.
The cell-free extract was subsequently loaded onto a Q-Sepharose anion exchange column and
protein was eluted with a gradient of 0-1 M NaCl. Activity was tested towards either phosphonate
monoester 3c (PMHs) or sulfate monoester 4 (ASs). The overexpressed proteins typically eluted
at 0.3-0.5 M NaCl. The active fractions were pooled, 1/10 of the pooled volume of 1 M Tris-HCl
pH 8.0 + 1.5 M NaCl was added to the sample, which was subsequently loaded onto 1 mL Strep-
Tactin column equilibrated with 100 mM Tris-HCl pH 8.0 + 150 mM NaCl. The column was
washed with 100 mM Tris-HCl pH 8.0 + 150 mM NaCl to remove unbound protein and the tagged
proteins were eluted with 2.5 mM d-desthiobiotin in 100 mM Tris-HCl pH 8.0 + 150 mM NaCl.
-1
The active protein containing fractions were pooled and concentrated to 10-15 mg mL protein
and loaded onto a Superdex 200 prep grade size exclusion column (GE Life Sciences) running in
1
00 mM Tris-HCl pH 8.0 + 150 mM NaCl. Active protein eluted at a molecular weight
corresponding to either a dimeric (ASs) or tetrameric (PMHs) protein (see Figure S14, SI). Protein
9
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 10 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
containing fractions were concentrated down to 100-200 µM, aliquoted, flash frozen in liquid
nitrogen and stored at -20 °C.
The possibility that contaminants contribute to the observed activities was ruled out by the
following considerations: the absolute levels of some of the promiscuous reactions are similar to
previously described ‘native’ functions,19,30,42,57,58 making it unlikely that they are caused by a
contaminating protein that co-purifies from the expression host (E. coli). Precautions taken to
prevent and rule out cross-contaminations include protein purification with an affinity-tag (Strep-
tag binding to Strep-Tactin), using fresh affinity resin for each variant and extensive cleaning of
the other columns with 0.5 M sodium hydroxide in between purifications to remove and inactivate
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
M
any AS or PMH from previous purifications. A wide range of K -values (5-100 fold differences
for the same substrate) was measured, indicating that the promiscuous activities are not caused by
a contaminant (in which case the promiscuous conversions would show similar K -values
M
characteristic of that contaminant).
Catalytic effects of metals and metal content determination using microPIXE. All
2
+
2+
enzymes were activated most strongly by Mn (compared to other divalent metal ions, Ca or
2
+
2+
Mg ), and kinetic experiments were carried out under conditions of saturating Mn (see SI).
The metal content of the various ASs and PMHs (Table S9) was determined using
microPIXE (carried out at the University of Surrey Ion Beam Centre) as described previously,18,32
using sulfur atoms as the internal quantitative standard to calculate the metal ion occupancy.
Enzyme assays. Enzyme assays were performed in either 20 mM sodium acetate, 100 mM
NaCl (phosphonate monoester 3a-c, pH 4.8-5.6), 20 mM bis-tris propane, 100 mM NaCl
(phosphonate monoesters 3b and 3c, pH 5.5-9.0) or SID buffer (44 mM succinic acid, 33 mM
imidazole, 33 mM diethanolamine, all other substrates, pH 5.5-9.6). Ionic strength was maintained
10
ACS Paragon Plus Environment

Page 11 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
constant within ±10% in the pH range 4.8-9.6. The catalytic parameters towards the promiscuous
substrates for the new PMHs and ASs were determined at their respective pH-optima of their
primary activity18,32 (Table S10-S17 and Figure S8 and S9).
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2+
All enzymes were activated most strongly by Mn (compared to other divalent metal ions,
2
+
2+
2+
Ca or Mg ), and kinetic experiments were carried out under conditions of saturating Mn (see
supporting information and Table S10-S28 for details). Enzymatic hydrolysis was followed by
monitoring the release of 4-nitrophenolate at 400 nm using a SpectraMax Plus multiwell reader.
The extinction coefficients of 4-nitrophenol at varying pH were determined for pH 5.5-9.6,
-1
-1
increasing from 1,000 to ~20,000 M cm with increasing pH.
-1
Initial rates (Vobs in M s ) were plotted against the substrate concentration ([S]) and kinetic
parameters kcat, K and/or kcat/K were determined using equations 1-3. For fitting with equation
, kcat/K is treated as a single parameter. In order to obtain a reliable fit to equations 1-3, a
minimum of 16 different initial rate measurements were performed in such a way that it included
substrate concentrations ranging from at least 5-fold below the estimated K -value to at least 5-
fold above the estimated K -value (with increments that increase with increasing substrate
M
M
2
M
M
M
concentration), wherever substrate solubility allowed for it. If the latter was not the case, at least
substrate 16 concentrations were chosen that range from 5-100% of the maximum solubility in 5-
1
0% increments. The K -value was estimated from several initial rate measurements covering a
M
wide range of substrate concentrations (varying by several orders of magnitude). Typical enzyme
concentrations used were 0.0016-0.74 M (for primary activities) or 0.16-10 M (for promiscuous
activities).
푉표푏푠 = 푘
×[퐸푛푧]×[푆]
eq. 1
푐푎푡
퐾
+[푆]
푀
11
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 12 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
푘
푐푎푡
푉표푏푠
=
× [ꢀꢁꢂ] × [ꢃ]
eq. 2
eq: 3
퐾
푀
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
푘
×[퐸푛푧]×[푆]
푐푎푡
푉
=
표푏푠
2
퐾
+[푆]+푆 ⁄퐾
푀
ꢄ퐼
kcat/K
M
refers to the first irreversible step of the enzymatic reaction, i.e. in this case to the formation
of the covalent intermediate between the formylglycine nucleophile and the respective P or S
center of the substrates. The phenolate product does not bind significantly to the enzyme, rendering
the reverse reaction to starting material unlikely. All specificity comparisons are based on kcat/K
M
and therefore refer to the same molecular event, the attack of the nucleophile, regardless of the
overall rate-limiting step.
The errors () indicated in Table S10-S29 were either obtained directly from the curve
fitting to equations 1-3 or calculated according to equation 4 (for kcat/K
M
if kcat and K
M
were
obtained by fitting to equation 1 or 3).
ꢆ
ꢆ
푘
푘
ꢅ푘
ꢅ퐾
푐푎푡
푐푎푡
푐푎푡
푀
훿
= | | × √(
) ꢇ (
)
eq. 4
퐾
퐾
푘
퐾
푀
푀
푐푎푡
푀
Crystallization, data collection, structure determination and analysis. All crystals were
-1
grown at 18 °C by hanging or sitting drop vapor diffusion with protein (~9-12 mg mL ) reservoir
ratios (v/v) of 1:1 (SpAS1 and SpAS2), 2:1 (ArPMH) or 1:3 (SpPMH). (i) Strep-SpAS1 crystals
grew in 0.1 M Tris-HCl pH 8.0, 20% (w/v) PEG 6000, 0.75 M LiCl. Crystals were soaked in the
mother liquor containing 10% (v/v) glycerol prior to flash cooling in liquid nitrogen. (ii) Strep-
12
ACS Paragon Plus Environment

Page 13 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
SpAS2 crystals grew in 20% (w/v) PEG 3350, 0.2 M ammonium formate. Crystals were soaked in
the mother liquor containing 20% (v/v) glycerol prior to flash cooling in liquid nitrogen. (iii) Strep-
ArPMH crystals grew in 0.1 M Tris-HCl pH 7.0, 10% (w/v) PEG 8000 and 0.2 M MgCl . Crystals
2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
were cryo-protected by adding 3 M trimethylamine-N-oxide (TMAO) to the mother liquor and
soaking the crystal for 1 minute before flash cooling in liquid nitrogen. (iv) Strep-SpPMH crystals
grew 0.1 M bis-tris-propane pH 7.0, 9% (w/v) PEG 8000, 0.2 M MgCl . They were soaked in the
2
mother liquor containing 20% (v/v) glycerol prior to flash cooling in liquid nitrogen.
Briefly, all crystallographic data were collected at ESRF or Diamond synchrotron sources
and processed using standard crystallographic software. For details on processing and refinement
see Table S32. Coordinates and experimental structure factors have been submitted to the Protein
Data Bank with accession codes 4UPH (ArPMH), 4UPI (SpAS1), 4UPK (SpPMH) and 4UPL
(
SpAS2).
RESULTS AND DISCUSSION
Classifying adjacent AP-Superfamily clades with phosphonate monoester hydrolase
and arylsulfatase primary activity. In order to assemble a test set of related members of the AP
superfamily populating the transition zone between sulfatases and phosphonate monoester
hydrolases, we ran a BLAST search using the amino acid sequences of the two previously
characterized PMHs18,32 as search sequences. The genomes of - and -proteobacteria yielded well
over 100 putative sequences that were originally annotated either as PMHs or arylsulfatases (ASs)
(
see Table S6-S7 for details). All of them contained the Cys-X-Pro/Ala-X-Arg recognition motif
essential for the post-translational formation of cysteine into the formylglycine nucleophile found
13
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 14 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
59
18,19
in ASs and PMHs
(Figure 2). Based on the level of similarity to the active site residues of
PMHs we divided the sequences into three groups. One group – termed T-PMHs (bearing a
threonine next to the catalytic nucleophile; Thr107 in RlPMH and BcPMH, homologous to position
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
A
His in Figure 2, Table S4) - had identical active site residues to the original two PMHs. The
A
second group, D-PMHs, characterized by an aspartate in place of a threonine (His in Figure 2).
The classification of these two groups as PMHs (Table S4) was supported by high overall pairwise
sequence similarity to the known PMHs (~50%). The third group of enzymes assumes an
intermediate position between PMHs and known ASs. These enzymes have histidine and lysine at
A
A
positions His and Lys ,(as in ASs), but two metal-coordinating residues are identical to the ones
B
A
in PMHs (Gln at position Asp ; His instead of Asn ). The active site of this group of enzymes is
therefore a hybrid between classical ASs and PMHs.
One T-PMH, two D-PMHs and six of the hybrid sulfatase-like proteins (Table 1) were
expressed in E. coli and purified to homogeneity. In the final step of protein purification by size
exclusion chromatography the PMHs eluted as tetramers, whereas the six sulfatase-like proteins
eluted as dimers (Figure S14). Activity assays showed turnover for all four substrate classes 1-4
that represent primary activities in AP-superfamily members (Figure 1). T-PMHs and D-PMHs
preferentially hydrolyzed phosphate diesters (2a-2c) and phosphonate monoesters (3a-3c). The
sulfatase-like proteins preferentially hydrolyzed arylsulfate monoesters (4). This experimental
result establishes them as a new class of ASs, even though they were identified from the databases
2
+
using PMHs as search sequences. However, they do prefer Mn as their active site metal ion (as
2+
2+
in PMHs), instead of the Ca /Mg -ions typical for previously identified ASs (see SI and Table
S4 for details). In these new ASs we observe high activities towards phosphoesters 1-3 (kcat/K
M
-
4
-1
-1
21
values as high as 10 s M ), similar to the bona fide sulfatase PAS, known to hydrolyze sulfate
14
ACS Paragon Plus Environment

Page 15 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
60
A
esters in vivo. We conclude that the identities of the active site residues at positions Lys and
A
His are better predictors for functional classification of ASs and PMHs than overall sequence
similarity.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Structure-guided alignment of the newly identified sulfatases with previously characterized
PMHs18,32,53 and sulfatases in the AP superfamily (see Table 1 and Table S5 for details on the
sequences included) shows that the new ASs are more closely related to PMHs than they are to
6
0
61-64
65
any other known ASs, such as PAS or several ASs of bacterial
A, Figure S1 and Table S5).
To identify additional sub-family members (see Table S6 and S7 for a list of these
and human origin (Figure
3
sequences) the five PMHs (BcPMH, RlPMH and the three newly identified ones) and the six newly
identified ASs were used as search sequences in BLAST, leading to identification of 169 related
enzymes. We discarded the annotations listed in the database for these enzymes (see Table S6-S7
for details) and re-classified them based on their active site residues (Figure 2), yielding 80 putative
PMHs and 89 putative ASs. The phylogenetic relationship shows that the functional re-
classification coincides with the genetic division between the new ASs and the PMHs (Figure 3B),
although the original division was based solely on the conservation of the 11 active site residues.
Structural comparisons of PMHs and ASs. In order to determine whether the
aforementioned phylogenetic division corresponds to structural features of the newly identified
AP superfamily members, two D-PMHs, ArPMH and SpPMH, and sulfatases, SpAS1 and SpAS2,
were crystallized and their structures were determined (Figure 4, Figure S17 and S18, Table S32).
All of the enzymes show the conserved /-fold29,45,46 (Figure 1), familiar from all AP-
6
6
30
67
18,32
superfamily members, such as AP, NPP, PAS and both previously described PMHs.
6
8
Pairwise comparisons between all the known crystal structures of ASs and PMHs show that
15
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 16 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
SpAS1 and SpAS2 are more closely related to the PMHs than they are to any of the classical ASs
(
Tables S1-S3), mirroring the phylogenetic relationship seen in Figure 3A. This is contradicted by
the substrate specificity profile, demonstrating the perils of assigning function based on sequence
and structural homology alone.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
All conserved active-site residues (Table S4) align structurally, confirming the mechanistic
relatedness between the various PMHs and ASs (Figure 5). With the structures of new enzymes
available, we rebuilt the phylogenetic tree (Figure 3B) using only 67 positions from the original
alignment that represent the conserved active site and its direct surroundings (see Figure S13). The
resulting tree shows the same global division as the original tree in Figure 3B, i.e. a clear distinction
between the PMHs and ASs (Figure S10), which means that the active site and its immediate
surroundings contain sufficient information to classify these enzymes. The fact that the global
division between both clades is the same in both trees therefore strongly suggests that the
phylogenetic division observed in Figure 3B is directly connected to the difference in substrate
specificity.
3
2
18
Comparison of the four new structures with RlPMH, BcPMH, and the monomeric
6
7
PAS (Figure 4A) reveals a sequential increase in the oligomeric state within the superfamily.
SpAS1 and SpAS2 are homodimeric enzymes with their helical C-terminal extensions facilitating
the dimerization (Figure 4B-C, Figure S17A-B). In contrast, ArPMH and SpPMH are tetrameric
3
2
enzymes with dihedral symmetry and very similar to their functional homologues RlPMH and
BcPMH18 (Figure 4D-E, Figure S17C-D). The dimeric structure of the novel sulfatases
corresponds to one half of the PMH tetramer and this dimerization is mediated by the interface
that is equivalent to the larger of the two protomer-protomer interfaces in PMHs (Figure 4). The
C-terminal regions mediating the dimerization of SpAS1 and SpAS2 are predominantly -helical
16
ACS Paragon Plus Environment

Page 17 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
with a highly conserved hydrophobic motif, in contrast to the short -hairpins that mediate the
homologous oligomerization interface of the PMHs (Figure 6A). Insertions and deletions (Figure
6
B) reshape the interface between protomers and C-terminal tails, and may – in agreement with
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
69
the analysis by Hashimoto et al., - have a role in governing oligomerization (see SI and Figures
S19-S21).
PMHs and ASs are highly promiscuous enzymes. The previously observed catalytic
promiscuity of members of the AP superfamily13,14,18,21,23-26,29,30 raises the question whether this
feature is reflected in the newly identified superfamily members and its quantification provides the
opportunity to assess the transition between chemical preferences. We thus tested the ability to
hydrolyze the four primary substrate classes for the AP superfamily, represented by phosphate
mono- (1) and diesters (2), phosphonate monoesters (3) and sulfate monoesters (4) and report
Michaelis-Menten parameters (Figure 7A-B, Table S18-S27) at the pH of maximal activity for the
primary substrate (Table S10-S17, Figure S15-S16,18,19) with the most activating divalent metal
2
+
(
Mn , present at saturating concentrations, see SI). In addition to changes in the reactive group,
the size of the non-reactive substituents (previously shown to influence catalytic rates in PAS21
40
(
Table S29) and NPP ) was also addressed in measurements with phosphodiesters 2a-2c and
phosphonate monoesters 3a-3c. All PMHs and new ASs are active towards all four substrate
classes and are thus truly promiscuous, further establishing 'crosswise' catalytic promiscuity as a
general trait of AP-superfamily enzymes. Catalysis for all four chemically distinct reactions is
1
1
17
efficient, with large second order rate accelerations ((kcat/K
M
)/k ) for both the primary (10 -10 -
2
9
12
fold) and the promiscuous (10 -10 -fold) activities (Table S18-S27).
17
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 18 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
The relative differences between the phosphodiesterase and phosphonate monoesterase
activities of each individual PMH fall within one order of magnitude (Table S18-S22). However,
we observe an increased propensity for the T-PMHs to accommodate phosphonate monoesters
with larger non-reactive substituents, i.e. the smallest phosphonate monoester (3a) is converted
with the lowest efficiency. The D-PMHs show the opposite effect, preferring smaller substrates
(e.g. favoring 3a vs. 3b and 3c). This observation is consistent with the defining feature that
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
A
separates the D- and T-clans: the threonine in position His (Figure 2) accommodates the non-
reactive hydrophobic substituent of the substrate more readily than a charged aspartate would. In
addition, for both D-PMHs the difference in catalytic efficiencies toward phosphate monoester 1,
phosphodiesters 2a-c and phosphonate monoesters 3a-c is much smaller than for the three T-PMHs
(
Figure 7A1-A4). This lack of specificity could be explained if the protonated monoanionic form
of phosphate monoester 1 reacts like a phosphate diester 2 (as postulated previously for
3
0
70
phosphodiesterases NPP and GpdQ ). Given the acidic pH at which the catalytic parameters for
both D-PMHs were determined, the monoanionic form of phosphate monoester 1 is likely to be
present at high levels. Therefore, the relative lack of specificity of the two D-PMHs toward these
types of substrates could be explained by a pH-rate optimum that causes the D-PMHs to employ
identical catalytic effects for the conversion of both phosphate mono- and diesters.
Trade-off between promiscuous activities. For ASs, the specificity between the primary
and the multiple promiscuous activities (i.e. the ratio between the primary activity and
promiscuous activity for a given promiscuous substrate) appeared to vary little between the five
dimeric ASs (Figure S22B1-B6). Direct correlation plots of the promiscuous activities vs. the
primary sulfatase activity (both expressed as log(kcat/K )), showed a linear correlation with a slope
M
18
ACS Paragon Plus Environment

Page 19 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
near unity for all substrates tested, except for phosphonate monoester 3c (Figure 7B1-B7).
Although ASs clearly prefer their “native” substrate, the ratio between the primary and a given
promiscuous activity is conserved, irrespective of the absolute level of the primary activity. As the
primary activity increases, so does the promiscuous activity, demonstrating near-constant
specificity and lack of trade-off.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
For the phosphonate monoesters 3a-c the variation in specificities between the dimeric ASs
increases with increasing size of the non-reactive substituent (R-groups in Figure 1; Figure 7B5-
2
1
B7). PAS, which had already been shown to be highly proficient toward phosphodiesters, showed
the same behavior (Table S29), despite being phylogenetically much further removed from the
PMHs than the dimeric ASs (Figure 3A). When PAS was included in the analysis, the conclusions
that can be drawn from the promiscuity correlations remain the same, although the trend toward
increased variation in substrate specificities with size of the R-group observed across phosphonate
monoesters 3a-c is much stronger (Figure 7B5-B7). These observations are consistent with the
following consideration: the putative binding site for these promiscuous substrates is under
selective pressure to accommodate sulfate monoesters. Phosphonate monoesters 3a-c have the
same overall charge as sulfates, but bear an additional group (R in Figure 1) and would be likely
to experience additional steric constraints in an active site evolved toward accommodating the
smaller sulfate monoester substrate. However, the relative promiscuity patterns of ASs towards
phosphate diesters 2a-c (which also have the same charge as sulfate monoesters) showed no
particular trend with respect to the size of the non-reactive substituent. There is no simple
explanation for these observations, except that the mixed effect of steric fit (including rotational
degrees of freedom in the substrate) and interactions with the bridging oxygen of the non-reactive
substituent force substrates into unique orientations, in which beneficial and detrimental binding
19
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 20 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
contributions may counteract each other. Consistent with the trend to favor promiscuous substrates
similar in size and charge to the primary substrate, for four (out of the five) new ASs, the smallest
promiscuous substrate with the same total charge, methylphosphonate 3a, was also the compound
toward which the enzyme had the highest activity besides the primary substrate.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
PMH and AS differ in trade-off between efficiency and specificity. Promiscuous
catalysts, in particular generalists, i.e. enzymes that catalyze more than one reaction at levels high
enough to support a functional role, have been postulated to be particularly versatile in their
adaptive potential, raising the question to what extent the catalysts in the activity transition zone
discussed here somehow remain ‘generalists’ or if they become specialists as they are subjected to
selection pressure for one activity. Higher activity of PMHs toward their primary substrates
(phosphate diesters 2a-c/phosphonate monoesters 3a-c) coincides with an improved degree of
specialization for the primary activities over the promiscuous conversion of phosphate or sulfate
monoesters (Figure 7A1-A4 and Figure S22A1-A4). The evidence for this hypothesis in PMH is
compelling (correlation slope < 1, p < 0.1, see Table S34-S35). and supports the notion that high
catalytic efficiency requires a more specialized enzyme. Likewise this scenario has also been
3
8
observed during laboratory evolution campaigns, in which case improvement in specificity was
driven by the requirement for improved turnover.
However, the observations for ASs stand in contrast to PMH: specificity (measured by the
ratio primary activity vs any given promiscuous activity) does not increase with increasing
catalytic efficiency (kcat/K ) toward sulfate monoester 4. Apparently in ASs activity increases for
M
the primary substrate (sulfate monoester 4) are invariably accompanied by an increase in any of
the promiscuous activities by the same order of magnitude, suggesting that this effect is at the very
20
ACS Paragon Plus Environment

Page 21 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
2
least not selected against. Indeed, all ASs prefer sulfate monoester 4 by ~10 -fold over
phosphonate monoester 3a, irrespective of the absolute level of the primary activity. This apparent
lack of further specialization, previously suggested as a necessity for obtaining extreme catalytic
efficiency for the native reaction,71 seems not to hamper near-optimal catalytic performance
toward its primary substrate, as the catalytic efficiency for PAS-catalyzed sulfate monoester
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
7
-1
-1 25
hydrolysis is close to diffusion control (kcat/K = 5 × 10 s M ). These correlations also support
M
the idea that these promiscuous activities are not just evolutionary relics of past activities from a
9
generalist ancestor. The cross-correlation of sulfatase vs. phosphodiesterase activities for ASs and
PMHs together with AP and NPP confirms that this linear correlation is specific for enzymes that
are under selective pressure for sulfatase activity (Figure 8A), notwithstanding the longer genetic
distance between PAS and SpAS1/SpAS2 (sequence identity: ~22%, Table S3) compared to
SpAS1/SpAS2 vs. PMHs (~33%, Table S3). The remarkable ability to improve catalysis toward
multiple chemically distinct substrates is illustrated by the following examples: for ASs
72
improvements for substrates with identical charges (yet different transition states) correlate well,
e.g. the monoanions of sulfate 4, phosphate diesters 2a-c and phosphonate monoester 3a (Figure
7B2-B5). Conversely, dianionic phosphate monoester 1 and the monoanionic sulfate 4 correlate
similarly (Figure 7B1) despite different ground state charges, but here the uncatalyzed reactions
proceed through analogous dissociative (‘exploded’) transition states.72
4
The conservation of these specificities across a 10 -fold difference in sulfatase kcat/K
M
-
levels suggests that this correlation is general. It seems that escape from this constant specificity
is difficult (or at least not selected for), when these sulfatases are under selective pressure for
improved sulfatase activity. If improvements in phosphoesterase activity are selected for, escape
from this correlation appears to be possible, but not in all directions, as shown recently for
21
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 22 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
26
SpAS1. As a result, sulfatases with low phosphodiesterase activity (i.e. enzymes located in the
top left corner in Figure 8A) are notably absent. It is likely that the conserved protein fold or the
intrinsic catalytic features required for enzyme-catalyzed sulfate monoester hydrolysis inherently
also favor catalysis of phosphoester hydrolysis, which constrains the observed specificities. This
observation suggests that there is a maximum degree of specialization between the primary
sulfatase activity and the activity toward phosphate monoesters, diesters and phosphonate
monoesters, when an enzyme is under selective pressure for sulfate monoester hydrolysis. It
remains to be seen whether the observed inability to specialize a sulfatase to minimize phosphate
transfer activity (and uncouple it from sulfatase improvements) is a general feature of these (or
any) enzymes. If so, it might explain why biological phosphorylation and sulfation are often
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
73
occurring in different, physically separated subcellular locations in mammals, thus avoiding
complications of intrinsic cross-reactivity.
Reactivity Correlations Across Evolutionary Snapshots. The characterization of active
sites with kcat/K
provides a unique opportunity to systematically analyze enzyme reactivity across an evolutionary
transition zone. Figure 7 shows one-against-one correlations that plot the kcat/K values of sub-
M
values differing over 4 orders of magnitude in a structurally conserved context,
M
group members for each promiscuous activity against its name-giving activity. Two scenarios can
be envisaged: (i) If reactions are promoted by catalytic effects that affect the chemically distinct
promiscuous reactions differentially, scatter should emerge, due to idiosyncratic patterns of
enzyme-substrate interactions and to different extents of transition state stabilization. In an
alternative scenario (ii), the interactions are quantitatively the same for all substrates, leading to a
linear correlation with a slope of 1.
22
ACS Paragon Plus Environment

Page 23 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
We explored these scenarios based on our new data for correlations of the main activity of
PMHs (Figure 7A1-A4) and for the new ASs (Figure 7B1-B7). For PMHs there is no discernible
trend: each substrate and each catalyst seem to be characterized by unique contributions to binding
and catalysis, indicative of particular arrangements for each pair. Scenario (ii) is observed for the
new ASs. The linearity of all but one of the correlations in Figure 7B1-B7 with a near identical
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
slope around unity suggest that within this group of enzymes, specificity for the primary substrate
4
over a given promiscuous one is constant across a 10 -fold variation in catalytic efficiency (kcat/K
M
)
towards its primary substrate. Indeed, this group of catalysts exhibits a correlation expected of a
chemical’ catalyst that is dominated by reactivity rather than substrate recognition. These
‘
correlations hold despite diverging sequences (even at the level of the conserved active site
positions, e.g. when comparing PAS vs. the new AS subfamily, Table S4). Only an increase in
substrate size - from methyl (3a) to ethyl (3b) and phenyl phosphonates (3c) (Figure 7B5-B7; p >
0.1, see Table S35 for details) - compromises the otherwise ideal correlation observed between
sulfate 4 and phosphonate 3a, suggesting that the availability of sufficient room for substrate
binding is a prerequisite for the correlated multi-reactivity improvements.
CONCLUSIONS AND IMPLICATIONS
Classification of new enzymes. Our data clearly show that a combination of sequence
analysis and thorough experimental verification of the catalytic function of putative enzymes is
7
4
necessary to avoid propagation of misannotations based on unverified functions. For example,
AkPMH was previously annotated as a sulfatase (Table 1), while our data classify the enzyme as a
23
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 24 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
1
phosphonate monoester hydrolase , based both on its phylogenetic relation to the known PMHs
and ASs, the nature of its active site residues (Table S4), and its substrate specificity profile. The
true substrates of the newly identified sulfo- and phosphoester hydrolases remain currently
unknown. However, the use of the type of model substrates (with hydrophobic leaving groups)
used in this study to assign the primary reaction types catalyzed by these enzymes has been used
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
7
5
historically to identify the substrate class they most likely act on in vivo. In no case has this initial
assignment been rescinded. Similar fluorogenic substrates continue to be used76,77 (e.g. as a
diagnostic, generally representative substrate indicating bacterial sulfatase activity61 or in
7
8
screening of metagenomic libraries for sulfatases ), even though sulfates sugars are more likely
7
9
candidates. One reason for the successful prediction of sulfatase activity with model substrates
may be that sulfate transfer is one of the thermodynamically most demanding natural reactions,80
so that a merely accidental activity can be ruled out. Taken together these arguments suggest that
the assignment to chemical classes of substrates is a widely used approximation and a basis for our
analysis of chemical reactivity at the level of reaction type, even if the precise identity of the
biological substrates remains to be elucidated.
Experimental verification of primary function is nevertheless complicated by the existence
of promiscuous side reactions and simple testing of the assumed primary activity could result in
incorrect functional annotation. The new class of sulfatases described in this study is a further
1
Note that we refer to enzymes that show its strongest activity towards substrates 3 as ‘phosphonate monoester
hydrolases (PMHs)’, because the first representative of this group was given this name. We have previously
5
3
3
2
suggested to name such enzymes ‘phosphonate monoester hydrolases/phosphodiesterases’, but stick here with the
simpler term PMH to group it with its historical antecedents. However, it is understood that name giving for these
hydrolases does not imply assignment of a biological role, given that no natural substrate is known and the absence of
phosphonates in the soil environment during the evolution of this enzyme makes an alternative assignment as a
phosphate diesterase (referring to the second best activity) possible, especially because more substrate candidates of
this type exist.
24
ACS Paragon Plus Environment

Page 25 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
example of this pitfall. The new sulfatases clearly have sulfate monoester hydrolysis as their
primary activity (Figure 7B), yet several of them catalyze phosphodiester/phosphonate monoester
hydrolysis at levels that are similar to those of the ‘native’ activity levels for PMHs (Table S18-
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
30
S22) and NPP. Since all of the previously uncharacterized enzymes (6 dimeric ASs, 3 new
PMHs) described in this study have either RlPMH or BcPMH as their closest homolog with
experimentally described function, their annotation as PMHs would have been plausible based on
their absolute activity levels for phosphonate monoester 3a. These considerations highlight that (i)
3
the large variations in absolute activity levels (>10 -fold, Figure 7) combined with the widespread
catalytic promiscuity in the AP-superfamily makes simply confirming activity by testing a single
substrate class unreliable for functional annotation, (ii) the specificities (i.e. the activity ratios of
primary vs. promiscuous substrates) are conserved and define the functional class the enzymes
belong to. Therefore, highly promiscuous enzymes should be tested on a panel of representative
substrates to establish the correct correlation between sequence (e.g. identity of active site residues,
phylogeny), structure and function.
Are there limitations to the evolution of enzyme specificity? For PMHs, higher rates
toward the primary phosphonate monoester substrate results in improved specificity toward its
primary substrate(s) over its promiscuous ones (Figure 7A1-A4 and Figure S15A1-A4). ASs
appear not to exhibit such an efficiency-specificity trade-off (Figure 7B1-B6 and Figure S15B1-
B6). This stands in contrast to the suggestion that evolution toward higher catalytic efficiency, i.e.
approaching levels close to diffusion control, will eventually result in an increase in specificity,71
7
-1
-1
4
since the near-diffusion controlled PAS (kcat/K
M
~10 s M ) is essentially as specific as the 10 -
fold less active SpAS1 (Figure 7B).
25
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 26 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
What may explain the apparent limit to specificity? If specificity is conserved, this means
that the many mutations separating the various sulfatases would need to affect all activities to the
same degree, and that mutations that maximize sulfatase activity while suppressing
phosphodiesterase activity have not been fixed during evolution. There are two features that
sulfatases could exploit to differentiate (with everything else being almost equal, i.e. leaving group,
smallest possible substituents for phosphodiesters) between sulfatase, phosphatase and
phosphodiesterase substrates: overall substrate charge and degree of necessity for leaving group
stabilization. Substrate specificity based on difference in charge of the substrate will inherently
favor phosphate monoesterase activity over all the others, whereas the high degree of leaving group
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
8
0
stabilization that will probably be required for sulfate monoester hydrolysis would benefit all
reactions, albeit to a different extent. Both properties are therefore extremely limited in providing
features that would exclusively benefit sulfatase activity. Furthermore, the low reactivity of likely
native sulfate ester substrates (e.g. sulfated sugars, kuncat ~10-26 s-1) must require sulfatases to
80
provide highly reactive functionalities in their active site.
The high levels of promiscuous phosphodiester hydrolysis that accompany high sulfatase
activity for our set of representative substrates, are unlikely to be relevant in vivo: natural aryl
phosphodiesters are rare and most naturally occurring phosphodiesters differ sufficiently in their
leaving group from naturally occurring sulfoesters, which may provide distinguishing features that
prevent cross-reactivity. Furthermore, phosphodiesters can have additional recognition anchors
(e.g. a functional group that can provide a specific interaction) in the non-reacting substituent (R
in Figure 1). The latter was observed for NPP, in which a large part of the specificity difference
(phosphodiesters 2 vs. phosphate monoester 1) was abolished when the non-reacting substituent in
the phosphodiester substrate was shortened to a methyl group.30,81
26
ACS Paragon Plus Environment

Page 27 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
Direction of evolution. We interpolate data from extant enzymes related by protein
sequence phylogeny,21,25,78 which depict accessible functional space characterized by evolutionary
snapshots. The close neighborhood in sequence space in nearby clades of ASs and PMH suggest
a pathway for natural evolution, because a functional change can be brought with relatively few
mutations. However, this does not imply a real evolutionary history (as the enzymes’ substrates
are often unknown), but options based on the classification of primary and promiscuous reaction
types. With these caveats in mind we speculate on the possible evolutionary scenarios the
functional characteristics we observe. Given that sulfatases appear to be restricted in the maximal
specificity between their native sulfate monoester substrate and phosphoesters (Figure 7B1-B6),
the question of the consequences for evolution arises (Figure 8). Adaptation toward higher
sulfatase preference would locate a highly specialized enzyme closer to the top left corner, yet this
activity space is apparently ‘forbidden’. Yet none of the enzymes studied by us or others23,40 are
located in the activity space above the fitted curve shaded in light red in Figure 8A. The specificity
of enzymes far away from the specificity limit, such as AkPMH (marked A, Figure 8A) has no
apparent intrinsic restrictions to move in all directions of functional space. In response to the
selective pressure applied when e.g. equal or improved sulfatase activity is required, only half of
the trajectories will be accessible. The maximally specialized sulfatases such as AkAS (marked B,
Figure 8A) that have reached the border of the “forbidden” specificity region appear to be
intrinsically constrained to half of the accessible directions in functional space compared to
unconstrained movement. Since this restriction is more pronounced toward one particular activity,
it is more likely for a specialized phosphodiesterase to emerge from a sulfatase than the other way
around: 50% of all accessible trajectories starting from AkPMH (A in Figure 8A) result in
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
27
ACS Paragon Plus Environment

Journal of the American Chemical Society
Page 28 of 62
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
improved sulfatase activity, whereas for AkAS (B in Figure 8A), 75% of the accessible trajectories
result in improved phosphodiesterase activity, i.e. the local structure of the sequence-function
space restricts the accessible genotypic space asymmetrically, constraining evolution across a
fitness landscape in favor of one activity over the other. This asymmetry would agree with the
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
difference in uncatalyzed rate constants (kuncat) for both reaction types (kuncat natural sulfate
monoesters ~10-26 s ,
-1 80
k
uncat natural phosphodiesters ~10-16 s-1): it might be expected that
adaptation toward a specificity profile that requires less catalytic power (compared to its original
primary activity) is more likely than the other way around.
The invariable increase in phosphodiesterase activity accompanying increased sulfatase
activity can be classified as exaptation: adaption toward one trait (sulfatase activity) results in an
increased level of another trait (phosphodiesterase activity) for which there is no apparent
selection. In this scenario (Figure 8B), adaptation towards an improved promiscuous function
would follow neither of the frequently discussed evolutionary trajectories,4,82 involving strong or
weak initial negative trade-off, with the latter proceeding through a generalist intermediate. These
strong or weak negative trade-off scenarios were based on observed activities of mutants obtained
during directed evolution experiments.37,83. Here, by contrast, both the primary and promiscuous
functions would be enhanced to the same degree, which suggests an ‘indiscriminate improver’
scenario with no initial trade-off. This scenario is possible for the improvement of activity toward
phosphoesters 1-3a in ASs that maintain their existing sulfatase specificity, while improving their
sulfatase activity (Figure 7B1-B5). Key elements of overall structure and active site are conserved
(albeit across larger sequence variation than in laboratory evolution experiments), suggesting the
‘indiscriminate improver’ scenario is accessible. This scenario is likely to arise when higher
sulfatase activity is required, whereas selection pressure toward improved phosphodiesterase
28
ACS Paragon Plus Environment

Page 29 of 62
Journal of the American Chemical Society
1
2
3
4
5
6
7
8
9
activity appears unrestricted and can follow all possible pathways (Figure 8B). A major advantage
of the ‘indiscriminate improver’ scenario is that it allows for improvement of a new activity
without relying on neutral drift (weak initial negative trade-off) or the occurrence of intermediates
with poor function (strong initial negative trade-off). Instead, increased selection pressure for the
primary activity, e.g. sulfatase activity, lifts the level of the secondary activity, e.g.
phosphodiesterase activity, above the selective threshold (Figure 9). This molecular exaptation
scenario provides an additional adaptive route to previously suggested sub-functionalization
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
8
4
models for protein evolution, such as the Escape from Adaptive Conflict (EAC) model and the
8
5
Innovation, Amplification and Divergence (IAD) model. All three scenarios are based on the
assumption that both old and new activities have to be able to co-exist in the same enzyme.
However, for both the EAC and IAD models initial adaptation toward a new function is expected
to occur at the cost of the original activity and requires a generalist intermediate that is brought
about as a result of trade-off (and has low activity) and emerges through near-neutral evolution in
the original activity. By contrast, our Exaptive Sub-Functionalization model (ESF, Figure 9) can
start with and immediate adaptive phase for both original and new activity. It is valid even with a
specialist with improved primary activity and its constant specificity avoids non-functional or low-
activity intermediates during prior to adaptation towards a new activity. The ESF model requires
no trade-off between optimal function and adaptability and a selectable level of the secondary
activity is reached as a result of optimization toward the original primary activity.
Evolution of function in enzyme families. Prediction of structure-activity relationships
and manipulation of catalytic activity based on evolutionary implications are inherently difficult,
since only a small fraction of the sequence variation directly affects function. A larger fraction of
29
ACS Paragon Plus Environment