SHORT REPORT Open Access
Applying neutral drift to the directed molecular
evolution of a b-glucuronidase into a b-
galactosidase: Two different evolutionary
pathways lead to the same variant
Wendy S Smith1, Jennifer R Hale1 and Cameron Neylon2*
Abstract
Background: Directed protein evolution has been used to modify protein activity and research has been carried
out to enhance the production of high quality mutant libraries. Many theoretical approaches suggest that allowing
a population to undergo neutral selection may be valuable in directed evolution experiments.
Findings: Here we report on an investigation into the value of neutral selection in a classical model system for
directed evolution, the conversion of the E. coli b-glucuronidase to a b-galactosidase activity. We find that neutral
selection, i.e. selection for retaining glucuronidase activity, can efficiently identify the majority of sites of mutation
that have been identified as beneficial for galactosidase activity in previous experiments. Each variant
demonstrating increased galactosidase activity identified by our neutral drift experiments contained a mutation at
one of four sites, T509, S557, N566 or W529. All of these sites have previously been identified using direct selection
for beta galactosidase activity.
Conclusions: Our results are consistent with others that show that a neutral selection approach can be effective in
selecting improved variants. However, we interpret our results to show that neutral selection is, in this case, not a
more efficient approach than conventional directed evolution approaches. However, the neutral approach is likely
to be beneficial when the resulting library can be screened for a range of related activities. More detailed statistical
studies to resolve the apparent differences between this system and others are likely to be a fruitful avenue for
future research.
Background
The development of experimental approaches for direc-
ted evolution has mirrored, to a certain extent, the
growing sophistication of our understanding of natural
evolution. Early experiments applied mutagenesis at ran-
dom along the length of a protein coding gene followed
by the application of a direct selection pressure [1,2], an
approach that Darwin would have readily understood.
Directed protein evolution has been used successfully to
tailor protein properties and to advance the understand-
ing of structure function relationships. However, despite
advances in screening technologies only a small fraction
of theoretical protein sequence can be sampled in
experiments. Recently new theoretical concepts and
computational programmes have been developed with
the aim of producing high quality mutant libraries [3].
Advances in random and focussed mutagenic techniques
such as the trinucleotide exchange method (TriNEx) [4],
transversion enriched sequence saturation mutagenesis
(SeSaM-Tv) [5] and iterative saturation mutagenesis
(ISM) [6] may aid the generation of libraries with func-
tionally enriched diversity.
In natural systems neutral drift is also likely to play a
role in priming a population for evolution. Neutral drift
is the incorporation of mutations which have little or no
effect on the protein in its current environment. Neutral
mutations however, may have adaptive potential and
allow adaptive events to occur more frequently.
* Correspondence: cameron.neylon@stfc.ac.uk
2Science and Technologies Facilities Council, Rutherford Appleton
Laboratory, Didcot, OX11 0QX, UK
Full list of author information is available at the end of the article
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
© 2011 Neylon et al; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons
Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in
any medium, provided the original work is properly cited.

Adaptive mutations in isolation are rare [7,8] as many
mutations are potentially deleterious, undermining pro-
tein stability or folding. These adaptive mutations which
are deleterious in isolation may however be able to be
included if other neutral mutations are first introduced
into the gene; or through the introduction of stabilising
mutations, which can bring residues into line with resi-
dues shared by the common ancestor or the consensus
residue across a protein family [9,10]. It is thought that
acquisition of new protein functions occurs in steps pro-
duced by single beneficial mutations, with the potential
for the initial steps to occur before selection for the new
function [8,10-12].
This poses the question; is it possible to modify a pro-
tein’s activity using neutral drift and produce variants
with higher activity towards an alternative substrate
than previously demonstrated using direct selection
experiments?
Bloom et al. [8] investigated how cytochrome P450
enzymes that have evolved neutrally with respect to
activity on a single substrate changed in their abilities to
catalyze reactions on other substrates. They concluded
that neutral genetic drift can lead to substantial changes
in protein functions that are not currently under selec-
tion, preparing the proteins to more readily undergo
functional evolution should selection favour new func-
tions in the future.
In a classic directed evolution experiment Matsumura
and Ellington [13] converted the E. Coli b-glucuronidase
towards b-galactosidase activity in three rounds of
mutagenesis and recombination. Their best variant con-
tained four mutations and further extensive screening of
libraries generated by random mutagenesis failed to
identify any variants with further improvements in activ-
ity. We have carried out an equivalent directed evolu-
tion experiment introducing neutral selection in an
attempt to identify other variants with further improve-
ments to activity through random mutagenesis.
Results
Screening system for neutral selection
It was necessary to identify an E. coli strain with no
endogenous glucuronidase or galactosidase activity so
that both activities of interest could be screened in the
same cells. We selected BW25141 cells [14] (gusA-,
lacZ-) for this purpose. We found that this strain was
much more efficiently transformed with purified plasmid
than with the products of ligation reactions and elected
to use a two stage approach. The wild type gusA gene
was subjected to random mutagenesis using error-prone
PCR or the Stratagene Genemorph system. The library
of PCR products was ligated between the NcoI and
EcoRI sites of pBAD/His (Invitrogen) and transformed
into super competent XL1-Blue cells (Stratagene) to
generate a large number of colonies (6 000 - 15 000 in
experiment A). The colonies were re-suspended in LB
medium and plasmid DNA was isolated. The purified
plasmid DNA was then used to transform BW25141
cells generating 800 - 3 000 colonies (experiment A).
Glucuronidase positive colonies, where glucuronidase
activity is defined as a colony displaying intense
blue colouration, were selected on plates containing
saturating concentrations of X-glu (5 bromo-4-chloro-
3-indolyl-b-Dgalactoside) and 2 mg.mL-1 arabinose.
See [Additional file 1: Supplementary Figure S1] for an
indication of colour variation.
Optimisation of mutation rate
The most common ‘variant’ in a library that has been
screened for maintenance of function will be the wild-
type protein. In initial experiments with a low mutagen-
esis rate we recovered entirely wild-type glucuronidase
genes from neutral screens. At much higher mutagenesis
rates the number of blue (neutral) variants dropped to
less than 10% which would lead to very low diversity in
the neutral library. The mutation rate was therefore
optimised by varying the concentration of input DNA in
the mutagenic PCR from 16 - 125 ng which gave per-
centages of white colonies varying from 50 - 30%
respectively. The concentration of target DNA used in
the preparation of the libraries was 31 ng in experiment
A and 16 ng in experiment B corresponding to muta-
genesis rates of one to two substitutions per gene per
round of screening.
Selection of galactosidase positive variants
Four rounds of neutral selection were carried out on the
gusA gene using 31 ng of template DNA in each round
of PCR based random mutagenesis (Experiment A). Six
glucuronidase positive variants from the fourth round
were chosen at random, and plasmid DNA was isolated
and sequenced. The neutral mutations identified were
E36K, D185G, P266S, H313Y, V484I, V536I, T589I, and
G594D. The positions of these mutations in a modelled
structure of the b-glucuronidase are shown in Figure 1
in yellow. After four rounds of neutral selection (see
[Additional file 2: Supplementary Table T1] for colony
numbers), in which a total of around 8000 colonies and
an estimated 24 000 mutations were screened, the
resulting neutral library was screened for b-galactosidase
activity. After 24 hours of incubation on plates contain-
ing arabinose and 5 bromo-4chloro-3-indolyl-b-D-galac-
toside (x-gal) those colonies appearing bluer than
control colonies expressing the wild type b-glucuroni-
dase were picked, and plasmid DNA was isolated and
sequenced. Three unique gene sequences were identified
and the substitutions found are detailed in Table 1.
In every galactosidase positive gene sequenced a
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
Page 2 of 10

substitution was found at one of two sites, T509 and
N566. These are sites previously identified by Matsu-
mura and Ellington [13] in their direct selection experi-
ments. These single substitutions were introduced into
wild type gusA using a Stratagene site directed mutagen-
esis kit and confirmed by DNA sequencing. Plasmids
containing genes for proteins with single residue substi-
tutions as well as the plasmids isolated from the screen-
ing experiment were transformed into BW25141 cells
and screened for b galactosidase activity. There was no
significant difference in intensity between N566S and
L510F/N566S. L510F alone was much paler. T509A
colonies produced a similar blue colour to colonies of
G278A/N308D/T509A. Thus the b-galactosidase activity
appears to be due to single key mutations, more specifi-
cally mutations that have been identified in previous
direct selection experiments. See [Additional file 2:
Supplementary Figure S1] for an indication of colour
variation.
Transformants from the fourth round of neutral muta-
genesis were also screened for glucuronidase activity,
952 deep blue colonies were selected, and subjected to
a further round of mutagenic PCR. The resulting
library was screened for b-galactosidase activity. Two
variants showing galactosidase activity were identified;
D57N/D508G/Y533C/S557A and G81V/A168S/W529L.
The D508 and S557 sites are two other mutations
identified by Matsumura and Ellington [13]. The
W529L mutation was identified in a direct selection
experiment, using a solution based screening system,
by Rowe et al. [15].
In experiment B 16 ng of target DNA was used in the
error prone PCR reactions to give a higher mutation
rate. After four rounds of neutral selection (see [Addi-
tional file 3: Supplementary Table T1] for colony num-
bers) and screening for galactosidase activity a single
unique galactosidase positive variant was identified. The
plasmid isolated coded for a glucuronidase enzyme with
T480A and N566S substitutions. The N566S substitu-
tion was identified in the first experiment and by Mat-
sumura and Ellington [13].
The Genemorph II™ random mutagenesis kit, com-
bining both Mutazyme and Taq, was used in experiment
C, thus eliminating the mutational bias encountered
using each enzyme alone. Screening for b-galactosidase
activity after four rounds of neutral selection provided
two variants; G245A/K370R/W529L and V67D/K370R/
W529L. The W529L mutation was again present. K370
has not been identified previously as advantageous for
increasing b-galactosidase activity but was identified
as a mutation site in a direct selection experiment to
increase the thermal stability of b-glucuronidase by
Flores and Ellington [16].
In experiments A, B and C the number of glucuroni-
dase positive variants taken through to the next round
ranged from 800 to 3262. An experiment was carried
Figure 1 Position of neutral mutations in the E. coli b-glucuronidase. The gusA gene was subjected to four rounds of neutral screening in
which 1815, 801, 1341, and 952 blue (gusA+) colonies were selected respectively. The fourth round neutral library was then screened for b-
galactosidase activity. The five residues where mutations contributed to galactosidase activity identified in this study and previously are shown in
red. Neutral mutations identified in the neutral screen are shown in yellow. Mutations that were identified in the galactosidase screen but did
not contribute significantly to galactosidase activity are shown in blue. The E. coli b-glucuronidase sequence was modelled on to the human
glucuronidase structure [26], using Swiss-Model [27,28].
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
Page 3 of 10

out using smaller library sizes of less than 600 variants
per round. Following screening for b-galactosidase activ-
ity no positive variants were found.
In summary each variant demonstrating increased
galactosidase activity identified by our neutral drift
experiments contained a mutation at one of four sites,
T509, S557, N566 or W529 (shown in red in Figure 1).
All of these sites have previously been identified in
direct selection experiments [13,16].
Characterisation of neutral and galactosidase positive
variants
Selected variants, along with the wild type b-glucuroni-
dase, were purified and their activities were characterised
in solution using p-nitrophenyl-b-D-galactopyranoside
(pNP gal) and p-nitrophenyl-b-D-glucuronide (pNP glu)
as substrates. The kinetic parameter kcat/Km is shown in
Figure 2 and the individual parameters are shown in
Table 2. The most active mutants produced from the
Table 1 Sequences of variants exhibiting b-galactosidase activity.
Wild Type A
4.1
A
4.3
A
4.5
A
5.1
A
5.2
B
4.1
C
4.1
C
4.2
Neutral 1 Neutral 2 Shuffled-1 Shuffled-2
AA substitutions 2 3 4 4 3 2 3 3 2 5
Silent Mutations 2 1 1 3 0 1
Position
36 E K
42 S N
57 D N
67 V D
81 G V V
185 D G
168 A S
217 R C
245 G A
266 P S
278 G A
308 N D
313 H Y
367 A P
370 K R R
378 E R
480 T A
484 V I
498 Q R
508 D G
509 T A A A A
510 L F
529 W L L L
533 Y C
536 V I
540 V D
557 S A P P
566 N S S S S
568 K Q Q
594 G D
Variants A4.1 - 4.5 were obtained from experiment A after four rounds of neutral screening. Variants A5.1 and A5.2 were obtained from the fifth round of neutral
screening. Variant B4.1 was obtained from experiment B using a high mutagenesis rate, after four rounds of neutral screening. Variants C4.1 and C4.2 were
obtained from experiment C after four rounds of neutral screening. Shuffled-1 and Shuffled-2 are the variants obtained by DNA shuffling of the variants with
galactosidase activity. The positions where mutations were previously identified by Matsumura and Ellington [13] are highlighted.
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
Page 4 of 10

neutral libraries were A4.1 (showing a 6 fold increase in
kcat/Km) and A5.1 (four fold increase in kcat/Km) in
terms of galactosidase activity. These mutants contain the
substitutions N566S and S557A respectively. The N566S
substitution alone produced a two fold increase in kcat/
Km using p nitrophenyl b-D-galactopyranoside as a sub-
strate, comparing well with the results obtained by Mat-
sumura and Ellington [13].
The mutant T509A and other mutants containing this
residue showed no significant increase in the kcat/Km
values obtained using p-nitrophenyl b-D galactopyrano-
side as a substrate. However, blue colonies were visible
on LB amp arabinose x-gal plates. Matsumura and
Ellington measured a four fold increase in activity for
the mutant T509A using p-nitrophenol galactoside as a
substrate [13]. The discrepancy with our results may be
due to the differences in the assays used, for example,
the difference in substrate leaving group. All of these
experiments used purified and re-transformed plasmid.
The T509A/S557P/N566S/K568Q variant, the best iden-
tified by Matsumura and Ellington, showed over a hun-
dred fold increase in galactosidase activity compared to
the wild type glucuronidase enzyme. The K568Q muta-
tion identified by Matsumura and Ellington was not
found in any of the galactosidase positive mutants
selected from the neutral selection libraries. This is
expected as K568Q has a deleterious effect on glucuro-
nidase activity. Plasmids containing a gene coding for
K568Q b-glucuronidase also produced white colonies
Figure 2 Kinetic parameters for selected variants. Names of the
variants are as given in Table 1. Glucuronidase activity is shown
with black bars, galactosidase activity with grey bars. Error bars are
the standard errors from fits of the data to a Michaelis-Menten
model. Variant glucuronidases were overexpressed from pBAD
derived plasmids in BW25141 cells. The cells were lysed and the
protein purified by Ni-NTA sepharose FF chromatography followed
by concentration and buffer exchange by ultra filtration. Protein
concentration was estimated by the method of Bradford [29].
Enzyme activity was quantified using the substrates p-nitrophenyl-b-
D-galactopyranoside (pNP-gal) and p-nitrophenyl-b-D-glucuronide
(pNP-glu) in 50 mM Tris HCl buffer pH 7.4. The absorbance at 405
nm was monitored using a Safire2 microplate reader (Tecan). An
extinction coefficient of 17 400 M-1 cm-1 was used to calculate p-
nitrophenyl-phosphate concentration. The concentration of protein
used ranged from 0 to 200 nM in the glucuronidase assays and
from 500 nM to 5 μM in the galactosidase assays. The concentration
of pNP-gal ranged from 100 μM to 15 mM and that of pNP-glu
varied between 10 μM and 5 mM.
Table 2 Kinetic parameters for pNP-glycoside hydrolysis by selected variant glucuronidases.
b-glucuronidase activity b-galactosidase activity
Variant kcat (s-1) KM (μM) kcat/KM (M-1.s-1 × 103) kcat (s-1 × 10-3) KM (mM) kcat/KM (M-1.s-1)
Wild-type 109 ± 42 260 ± 120 410 ± 20 6.0 ± 0.2 2.7 ± 0.7 2.2 ± 0.6
N566S 128 ± 36 7000 ± 2000 18 ± 4 28 ± 2 7.2 ± 2 3.9 ± 0.8
T509A 161 ± 30 420 ± 60 380 ± 30 10 ± 2 5.4 ± 2.9 1.8 ± 0.5
T/S/N/K - - - 99 ± 18 0.35 ± 0.2 280 ± 140
A4.1 55 ± 19 2900 ± 500 19 ± 6 21 ± 5 1.8 ± 0.8 12 ± 2
A4.3 129 ± 13 200 ± 40 660 ± 180 15 ± 8 6.0 ± 0.3 2.5 ± 1.2
A4.5 58 ± 6 240 ± 30 240 ± 60 2 ± 1 2.4 ± 1.3 0.8 ± 0.3
B4.1 46 ± 8 1400 ± 300 32 ± 5 15 ± 5 2.7 ± 1.1 5.5 ± 1.3
A5.1 2.7 ± 0.8 89 ± 38 30 ± 16 148 ± 43 22 ± 5 6.8 ± 1.4
A5.2 83 ± 22 280 ± 110 300 ± 90 63 ± 12 7.3 ± 2.6 8.6 ± 1.2
Shuffled-1 - - - 46 ± 9 0.3 ± 0.1 166 ± 80
Shuffled-2 - - - 96 ± 17 0.5 ± 0.3 200 ± 134
Neutral-1 77 ± 21 35 ± 5 220 ± 90 - - -
Neutral-2
C4.1
83 ± 9
21.8
440 ± 80
73.6
190 ± 30
296
-
9
-
9.21
-
0.98
The parameters were estimated as described in Materials and Methods. The parameters and errors for each parameter are derived from an independent fit of the
data to a Michaelis-Menten model. Glucuronidase activity for Shuffled-1, Shuffled-2, and the T/S/N/K variant, and galactosidase activity for Neutral-1 and Neutral-2
were not detectable under the conditions of our assay.
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
Page 5 of 10

when screened for glucuronidase activity. Two variants
from the fourth round neutral library obtained during
experiment A (neutral 1 and neutral 2) retained high b
glucuronidase activity but no b galactosidase activity
was measurable at the protein and substrate concentra-
tions used, indicating less galactosidase activity than the
wild type glucuronidase itself showed.
The mutants N566S, A4.1 and B4.1 produced twenty
fold lower Kcat/km values than wild type b glucuroni-
dase when p-nitrophenyl-b-D-glucuronide was used as
a substrate. This may indicate that the selection level
is relaxed. However, no significant colour variation
between colonies was apparent from the × glu screening
when blue colonies were selected. The x-glu/gal screen
is a qualitative assay as opposed to the quantitative
assay used to determine the kinetic data.
DNA shuffling
Five variants, selected from both the fourth and fifth
round neutral libraries (A4.1, A4.3, A5.1, A5.2, B4.1)
obtained from experiments A and B, were pooled and
subjected to DNA shuffling [17,18]. Six blue colonies
were obtained when the transformants resulting from the
shuffled, pooled mutants were screened for beta galacto-
sidase activity, using x-gal, and these yielded two unique
variants on sequencing; S42N/E378A/T509A/S557P/
N566S/K568Q (Shuffled 1) and G81V/Q498R/T509A/
S557P/N566S/K568Q (Shuffled 2). Both of these variants
contain the same four substitutions as the best variant
identified in the direct selection experiment. All but one
of the substitutions that were present in the pooled var-
iants but identified as unimportant above (D57N, A168S,
R217C, G278A, N308D, T480A, D508G, L510F, Y533C,
V540D) have been lost from the shuffled variants. The
single exception to this is the G81V substitution that has
survived in the Shuffled 2 variant. S557 has been con-
verted to proline in both cases, again the best substitu-
tion identified at this position. Two further substitutions
have been introduced to each of the shuffled variants.
Both have an additional arginine (E378R for Shuffled 1,
Q498R for Shuffled 2) and interestingly both have
obtained the K568Q substitution, the final substitution
identified by Matsumura and Ellington but missing from
our neutral libraries as it is deleterious for glucuronidase
activity. The kcat/Km values obtained for both of these
clones using p nitrophenyl b-D-galactopyranoside as a
substrate were similar to that obtained for the mutant
T509A/S557P/N566S/K568Q. Shuffling of the neutral
libraries themselves failed to generate colonies exhibiting
more galactosidase activity than negative controls.
Additional Screening
The purified plasmid libraries obtained from rounds four
and five of neutral drift experiment A and round four of
neutral drift experiment B were transformed into
BW25141 cells and the transformants were applied to LB
amp plates containing 100 μg.mL-1 ampicillin, 2 mg.mL-1
arabinose and one of the substrates; 5-bromo-4-chloro-3-
indolyl-b-Dfucopyranoside (X-fucopyranoside), 5-bromo-
4-chloro-3-indolyl-b-Dxylanopyranoside (X-xylopyrano-
side), 5-bromo-4-chloro-3-indolyl-N-acetyl-b-Dglucosa-
minide (X-glucosaminide), 5-bromo-4-chloro-3-indolyl-
N-acetyl-b-Dgalactosaminide (X-galactosaminide) or 5-
bromo-4-chloro-3-indolyl-b-D11 glucopyranoside (X-
glucopyranoside). Screening wild type b-glucuronidase
against these substrates showed that it was active towards
glucopyranoside and weakly active towards xylopyrano-
side but inactive towards fucopyranoside, glucosaminide
and galactosaminide. Three colonies in the round four
library from neutral drift experiment A appeared blue
compared to the other colonies in the screened library
and compared to wild type b glucuronidase colonies
when screened against the substrate X-xylopyranoside.
Three unique gene sequences were identified; F88S/
I170V/A323T, Q385L/F448H/T509A and Q40R/A560P/
N574S/G594R/E595D. Geddie and Matsumura [19] used
site saturation mutagenesis of the active site loop resi-
dues 557, 566 and 568 to produce clones exhibiting
increased xylosidase activity. T509A is the only mutation
the two studies have in common. Therefore, our neutral
selection experiment has identified other residues that
may contribute to xylosidase activity.
Discussion
A central problem with many directed evolution experi-
ments based on random mutation is the tendency for the
selection process to become trapped in a local minimum.
The original experiments of Matsumura and Ellington
[13] on beta-glucuronidase exemplify this with their final
selected variant, while performing orders of magnitude
better in the target activity than the original protein, still
performing an order of magnitude worse than that of a
naturally evolved beta-galactosidase. Much of the techni-
cal development of new methods for creating libraries
for directed evolution has been focussed on develop-
ing methods and protocols to overcome this problem
via saturation mutagenesis, recombination, or other
approaches. The use of neutral selection appears to offer
a complementary approach to avoiding the problem of
local minima by increasing the spread of starting points
in the “fitness landscape”. If the problem of the conven-
tional approach is that it only allows variants to find the
top of the hill on which the researcher begins, then neu-
tral selection offers a way of increasing the range of those
starting points while remaining in “functional areas”. This
avoids the problems that saturation mutagenesis can
cause by generating a large proportion of non-functional
variants for screening.
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
Page 6 of 10

Thus the aim of our experiments was to use neutral
selection to identify new variants that can provide
improvements to the desired function that are not acces-
sible by a conventional direct selection experiment based
on random selection. From this perspective the results
are disappointing as we have identified only variants and
positions of mutation that have been described in other
direct selection experiments aimed at converting b-glu-
curonidase into a galactosidase. However, it is worth not-
ing that by performing neutral selection we obtained
improved variants with modifications at four out of the
five sites previously identified. The fifth mutation which
is beneficial to galactosidase activity but deleterious to
glucuronidase activity was not obtained from the neutral
selection, as expected, but was selected in a single round
when five neutral variants were shuffled together.
When we examine other studies using neutral selec-
tion a very similar picture emerges. The neutral drift
study of Gupta and Tawfik [20] produced mutational
compositions similar to those obtained by directed evo-
lution of PON 1. The best variant in the neutral drift
study contained the active site mutations S193T and
T332S. Mutations at these sites were previously identi-
fied in traditional directed evolution experiments
[21,22]. One of the neutral mutations seen in variants
improved for DEPCyC, F222S (Gupta & Tawfik, [20])
under neutral selection was not seen in direct selection
for the same substrate [21]. However this was not one
of the most active variants produced. Bershtein et al. [9]
carried out an accelerated neutral drift experiment using
TEM-1 b lactamase producing a range of mutations that
increased protein stability. The most common enriched
mutations found in the study were previously identified
when a destabilized form of TEM-1 was evolved to
regain stability [23]. Mutations at the same sites as the
remaining enriched mutations and all but three of the
purging mutations were also previously generated by
Bershtein et al. [24] when they subjected TEM-1 to ran-
dom mutational drift and purifying selection (to purge
deleterious mutations). Therefore, the mutations found
in the neutral drift studies of Gupta and Tawfik [20],
and Bershtein et al. [9] produced variants with muta-
tions at the same positions as those produced from
more conventional evolution experiments on the same
proteins.
It is possible that clonal interference may prevent
certain mutations from appearing. Rowe et al. [15] com-
pared directed evolution approaches using b glucuroni-
dase and found that the beneficial mutations (F365S/
W529L) could drive N566S into extinction even though
the mutations were potentially synergistic in effect. They
demonstrated that the undesirable extinction events
could be prevented by DNA shuffling. The mutation
K568Q did not appear in our neutral drift library because
of the loss of original glucuronidase activity. However,
DNA shuffling of our neutral clones produced galactosi-
dase positive variants containing this mutation. Therefore
DNA shuffling following either directed evolution or
neutral drift experiments may overcome both clonal
interference and trade off restrictions.
Overall it appears that neutral selection does not
provide new and improved variants over those seen in
conventional directed evolution experiments based on
random mutagenesis and direct selection.
If neutral selection does not provide improved variants
then it may still be of value if it identifies those variants
more efficiently. This is the central claim made by
Gupta and Tawfik [20]. In the case of b-glucuronidase
Matsumura and Ellington undertook three rounds of
selection, screening 7 000, 20 000, and 7 500 colonies in
each round respectively [13]. In our experiments we
screened smaller numbers; a total of between 8 000 and
17 000 colonies for four rounds of the neutral screen
and 5000 to 8 500 for the galactosidase screen. Arguably
this makes the neutral approach more efficient but the
advantages are marginal. When we carried out an
experiment using very small library sizes of less than
600 variants per round no positive variants were found
when we screened for galactosidase activity (see supple-
mentary information for details). The lack of b-galacto-
sidase positive variants was attributed to using such a
small library size.
The added complication of developing a screen for a
second activity and optimising the mutation rate is
enough of a disincentive that the neutral approach is
unlikely to be appealing for most straightforward opti-
mization experiments.
Our experiment was limited to relatively small num-
bers as the aim was to investigate the value of neutral
selection in conditions similar to those carried out in
most directed evolution experiments. One important
question is whether the results would be different had
larger numbers of variants been screened? Experiments
using saturation mutagenesis of the b-glucuronidase
gene have identified variants with greater increases in b-
xylosidase activity than the increase in b-galactosidase
activity previously observed via random mutagenesis
[19]. However the only one of these characterised for
glucuronidase activity has a massive decrease in kcat. A
larger scale screen using assays with much higher
throughput, such as FACS analysis [25], would be
required to investigate this more fully. Again the studies
of Gupta and Tawfik using serum paraoxonase [20] and
Bloom et al. [8] using cytochrome P450 do not identify
new variants despite screening much larger numbers.
Our conclusion, based on the combination of these stu-
dies and are own, is that there is little to be gained by
increasing screening sizes from thousands to millions
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
Page 7 of 10

Figure 3 Structures of the alternative substrates. Structures of the substrates 5-bromo-4-chloro-3-indolyl-b-D-fucopyranoside (X-
fucopyranoside), 5-bromo-4-chloro-3-indolyl-b-D-xylanopyranoside, (X-xylopyranoside) 5-bromo-4-chloro-3-indolyl-N-acetyl-b-D-glucosaminide (X-
glucosaminide), 5-bromo-4-chloro-3-indolyl-N-acetyl-b-D-galactosaminide (X-galactosaminide) and 5-bromo-4-chloro-3-indolyl-b-D-
glucopyranoside (X-glucopyranoside) used in the additional screening of the fourth round neutral drift libraries.
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
Page 8 of 10

for neutral selection. If it is feasible to screen very large
numbers then direct selection is likely to generate the
same results with a similar amount of effort.
Given that the neutral selection process creates a
library capable of generating useful variants for a modi-
fied function after one further round of direct selection,
an area where it may find useful application is where a
library is desired for generating multiple, related func-
tions. We demonstrated this by screening our libraries
against a range of x-glycosides and identifying specific
improved variants. The chosen glycosides (Figure 3) dis-
played a range of structural changes compared to glucur-
onides. As expected the neutral library had not
developed activity towards b-fucosides, b-glucosaminides
or b-galactosaminides. Three variants, F88S/I170V/
A323T, Q385L/F448H/T509A and Q40R/A560P/N574S/
G594R/E595D, were found to have increased activity
towards b-xylopyranosides compared to wild type b- glu-
curonidase. Again this was achieved in only one round of
selection. In this T509A is the only mutation in common
with the experiment carried out by Geddie and Matsu-
mura [19] in which saturation mutagenesis was used to
direct the evolution of GUS variants with increased xylo-
sidase activity. However we have no reason for thinking
that the other variants we identify would not be found in
a conventional directed evolution experiment.
Matsumura and Ellington [13] screened a selection of
mutants found during their experiment for glucosidase
and fucosidase activity. Their most efficient mutant for
b-galactosidase activity, T509A/S557P/N566S/K568Q,
was also found to have developed increased activity
towards fucosides and glucosides compared to wild type
b-glucuronidase. Likewise the single mutant N566S and
the quadruple mutant T509A/D531E/S557P/N566S had
increased fucosidase activity.
Neutral selection in the case reported here performs
no better than several rounds of direct selection but can
provide a library containing variants that perform
against a variety of substrates that can be selected in a
single round of direct selection. If multiple similar activ-
ities are desired then neutral selection can be an effi-
cient route to them. Overall the results we present here
do not provide evidence that supports the use of neutral
approaches as a more efficient approach to directed evo-
lution. At least for the system we have investigated the
additional diversity potentially provided by this approach
is either insufficient or of the wrong type to provide
access to local selection optima different from those
found in direct selection experiments. This may of
course vary from system to system and a detailed statis-
tical study of a wide range of systems where different
selection protocols to libraries of varying diversity seems
a promising course for future study.
Materials and methods
See [Additional file 3].
Additional material
Additional file 1: Materials and Methods. Full description of materials
and methods used.
Additional file 2: Supplementary figure S1. a set of three pictures
demonstrating the colour changes observed in bacterial colonies
exhibiting a positive reaction for X-glu and X-gal.
Additional file 3: Supplementary Table T1. a table detailing the
numbers of colonies counted at each stage in each library.
Acknowledgements
We thank the Biotechnology and Biological Sciences Research Council for
grants BB/D00652X/1 and E16590 which supported this work and the
Engineering and Physical Sciences Research Council for support through the
University of Southampton School of Chemistry Doctoral Training Account.
Author details
1School of Chemistry, University of Southampton, Highfield SO17 1BJ, UK.
2Science and Technologies Facilities Council, Rutherford Appleton
Laboratory, Didcot, OX11 0QX, UK.
Authors’ contributions
WSS carried out the initial neutral drift experiments, the characterisation of
neutral and positive variants, the DNA shuffling experiments and helped to
draft the manuscript. JRH carried out the later neutral drift experiments, the
additional substrate screening and contributed to the drafting of the
manuscript. CN conceived the study, participated in its design and
coordination and drafted the manuscript. All authors have read and approve
the final manuscript.
Competing interests
The authors declare that they have no competing interests.
Received: 18 May 2010 Accepted: 6 May 2011 Published: 6 May 2011
References
1. Oliphant AR, Struhl K: An efficient method for generating proteins with
altered enzymatic properties: application to beta-lactamase. Proc Natl
Acad Sci USA 1989, 86:9094-8.
2. Chen K, Arnold FH: Tuning the activity of an enzyme for unusual
environments: sequential random mutagenesis of subtilisin E for
catalysis in dimethylformamide. Proc Natl Acad Sci USA 1993, 90:5618-22.
3. Shivange AV, Marienhagen J, Mundhanda H, Schenk A, Schwaneberg U:
Advances in generating functional diversity for directed protein
evolution. Current Opinion in Chemical Biology 2009, 13:1-7.
4. Baldwin AJ, Busse K, Simm AM, Jones DD: Expanded molecular diversity
generation during directed evolution by trinucleotide exchange (TriNEx).
Nucleic Acids Res 2008, 36:e77.
5. Wong TS, Roccatano D, Loakes D, Te KL, Schenk A, Hauer B,
Schwaneberg U: Transversion enriched sequence saturation mutagenesis
(SeSaM-TV+): a random mutagenesis method with consecutive
nucleotide exchanges that complements the bias of error prone PCR.
Biotechnol J 2008, 3:74-82.
6. Reetz MT, Carballeira JD: Iterative saturation mutagenesis (ISM) for rapid
directed evolution of functional enzymes. Nat Protoc 2007, 2:891-903.
7. Amitai G, Gupta RD, Tawfik DS: Latent evolutionary potentials under the
neutral mutational drift of an enzyme. HFSP Journal 2007, 1:67-78.
8. Bloom JD, Romero PA, Lu Z, Arnold FH: Neutral genetic drift can alter
promiscuous protein functions, potentially aiding functional evolution.
Biology Direct 2007, 2:17.
9. Bershtein S, Goldin K, Tawfik DS: Intense Neutral Drifts Yield Robust and
Evolvable Consensus Proteins. J Mol Biol 2008, 379:1029-1044.
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
Page 9 of 10

10. Bloom JD, Arnold FH: In the light of directed evolution: Pathways of
adaptive protein evolution. Proc Natl Acad Sci USA 2009, 9995-10000.
11. Tracewell CA, Arnold FH: Directed enzyme evolution: climbing fitness
peaks one amino acid at a time. Curr Opin Chem Biol 2009, 13:3-9.
12. DePristo MA: The subtle benefits of being promiscuous: adaptive
evolution potentiated by enzyme promiscuity. HFSP Journal 2007, 1:94-98.
13. Matsumura I, Ellington AD: In vitro evolution of a beta-glucuronidase into
a beta-galactosidase proceeds through non-specific intermediates. J Mol
Biol 2001, 305:331-339.
14. Datsenko KA, Wanner BL: One-step inactivation of chromosomal genes in
Escherichia coli K-12 using PCR products. Proc Natl Acad Sci USA 2000,
97:6640-5.
15. Rowe LA, Geddie ML, Alexander OB, Matsumura I: A comparison of
directed evolution approaches using the beta-glucuronidase model
system. J Mol Biol 2003, 332:851-60.
16. Flores H, Ellington AD: Increasing the thermal stability of an oligomeric
protein, beta glucuronidase. J Mol Biol 2002, 315:325-337.
17. Stemmer WPC: DNA shuffling by random fragmentation and reassembly.
vitro recombination for molecular evolution Proc Natl Acad Sci USA 1993,
91:10747-10751.
18. Stemmer WPC: Rapid evolution of a protein in vitro by DNA shuffling.
Nature 1994, 370:389-391.
19. Geddie ML, Matsumura I: Rapid evolution of beta-glucuronidase
specificity by saturation mutagenesis of an active site loop. J Biol Chem
2004, 279:26462-8.
20. Gupta RD, Tawfik DS: Directed enzyme evolution via small and effective
neutral drift libraries. Nature Methods 2008, 5:939-941.
21. Aharoni A, Gaidukov L, Yagur S, Toker L, Silman I, Tawfik DS: Directed
evolution of mammalian paraoxonases PON1 and PON3 for bacterial
expression and catalytic specialization. Proc Natl Acad Sci USA 2004,
101(2):482-487.
22. Aharoni A, Amitai G, Bernath K, Magdassi S, Tawfik DS: High-Throughput
Screening of Enzyme Libraries: Thiolactonases Evolved by Fluorescence-
Activated Sorting of Single Cells in Emulsion Compartments. Chemistry &
Biology 2005, 12:1281-1289.
23. Hecky J, Muller KM: Structural perturbation and compensation at
physiological temperature leads to thermo-stabilization of beta-
lactamase. Biochemistry 2005, 44:12640-12654.
24. Bershtein S, Segal M, Bekerman R, Tokuriki N, Tawfik DS: Robustness-
epistasis link shapes the fitness landscape of a randomly drifting
protein. Nature 2006, 444:929-932.
25. Bergquist PL, Reeves RA, Gibbs MD: Degenerate oligonucleotide gene
shuffling (DOGS) and random drift mutagenesis (RNDM): Two
complementary techniques for enzyme evolution. Biomol Eng 2005,
22:63-72.
26. Jain S, Drendel WB, Chen ZW, Mathews FS, Sly WS, Grubb JH: Structure of
human beta-glucuronidase reveals candidate lysosomal targeting and
active-site motifs. Nat Struct Biol 1996, 3:375-81.
27. Guex N, Peitsch MC: SWISS-MODEL and the Swiss-PdbViewer: an
environment for comparative protein modeling. Electrophoresis 1997,
18:2714-23.
28. Schwede T, Kopp J, Guex N, Peitsch MC: SWISS-MODEL: An automated
protein homology-modelling server. Nucleic Acids Res 2003, 31:3381-5.
29. Bradford MM: A rapid and sensitive method for the quantitation of
microgram quantities of protein utilizing the principle of protein-dye
binding. Anal Biochem 1976, 72:248-54.
doi:10.1186/1756-0500-4-138
Cite this article as: Smith et al.: Applying neutral drift to the directed
molecular evolution of a b-glucuronidase into a b-galactosidase: Two
different evolutionary pathways lead to the same variant. BMC Research
Notes 2011 4:138.
Submit your next manuscript to BioMed Central
and take full advantage of:
• Convenient online submission
• Thorough peer review
• No space constraints or color figure charges
• Immediate publication on acceptance
• Inclusion in PubMed, CAS, Scopus and Google Scholar
• Research which is freely available for redistribution
Submit your manuscript at
www.biomedcentral.com/submit
Smith et al. BMC Research Notes 2011, 4:138
http://www.biomedcentral.com/1756-0500/4/138
Page 10 of 10