N-terminal extension of the yeast IA
3
aspartic proteinase
inhibitor relaxes the strict intrinsic selectivity
Tim J. Winterburn
1
, Lowri H. Phylip
1
, Daniel Bur
2
, David M. Wyatt
1
, Colin Berry
1
and John Kay
1
1 School of Biosciences, Cardiff University, UK
2 Actelion Pharmaceuticals Ltd, Allschwil, Switzerland
Gene-encoded inhibitors of aspartic proteinases are
rather rare in nature. Thus, there is a need to under-
stand the mechanisms of action of the few that are
known, in order to exploit their therapeutic potential
[1]. We have described previously one such inhibitor:
the IA
3
protein from Saccharomyces cerevisiae [1–4].
This remarkable polypeptide not only is a highly
potent inhibitor of its target enzyme, saccharopepsin,
but also appears to be completely specific for this sole
target proteinase [1,2]. Crystal structures solved for
complexes of IA

(Received 30 March 2007, revised 23 May
2007, accepted 25 May 2007)
doi:10.1111/j.1742-4658.2007.05901.x
Yeast IA
3
aspartic proteinase inhibitor operates through an unprecedented
mechanism and exhibits a remarkable specificity for one target enzyme, sac-
charopepsin. Even aspartic proteinases that are very closely similar to
saccharopepsin (e.g. the vacuolar enzyme from Pichia pastoris) are not sus-
ceptible to significant inhibition. The Pichia proteinase was selected as the
target for initial attempts to engineer IA
3
to re-design the specificity. The
IA
3
polypeptides from Saccharomyces cerevisiae and Saccharomyces castellii
differ considerably in sequence. Alterations made by deletion or exchange
of the residues in the C-terminal segment of these polypeptides had only
minor effects. By contrast, extension of each of these wild-type and chimaer-
ic polypeptides at its N-terminus by an MK(H)
7
MQ sequence generated
inhibitors that displayed subnanomolar potency towards the Pichia enzyme.
This gain-in-function was completely reversed upon removal of the exten-
sion sequence by exopeptidase trimming. Capture of the potentially posi-
tively charged aromatic histidine residues of the extension by remote,
negatively charged side-chains, which were identified in the Pichia enzyme
by modelling, may increase the local IA
3
concentration and create an

3
inhibitor are identical in PpPr. PpPr also
has the crucial Ala residue present at position 213 in
its sequence. Despite this close relatedness, the two
enzymes differ drastically in their susceptibility to inhi-
bition by IA
3
. Accordingly, it was of considerable
interest to examine whether IA
3
could be adapted to
loosen its stringent specificity and, in this way, begin
the process of engineering it to target aspartic protein-
ase(s) other than saccharopepsin. Since PpPr is not
inhibited effectively by IA
3
yet is so closely related to
saccharopepsin, it was an obvious choice as the initial
new target enzyme. In the present study, we show that,
inter alia, inhibitors with subnanomolar potency
against PpPr, can be generated by simply attaching a
histidine-rich extension at the N-terminus of the IA
3
polypeptide. This dramatic alteration in behaviour
may be explained by the positively ionisable histidine
residues initiating additional contacts outside the active
site that promote occupation of the active site of the
target proteinase by the inhibitory segment.
For ease of interpretation, residues in the inhibitors
are denoted by single letter abbreviations while those

Identity Residue number
M
M
M
M
(H)
ZQ
(H)
ZQ
N K D E
34242218 681 81
K (nM)
55 ± 11
15 ± 3
100 ± 20
NI
3 ± 0.5
4 ± 0.5
15 ± 5
280 ± 30
2 ± 0.2
10 ± 1
SMK H
E
N K D
H
N K D
SMK
SZK
SZK

produced in recombinant form in Escherichia coli
and purified to homogeneity as described in the
Experimental procedures. The S. castellii IA
3
was,
however, only marginally more effective as an inhibitor
of PpPr than its S. cerevisiae counterpart (cf. 2 and 1;
Fig. 1).
The effect of C-terminal segment residues on the
inhibitory activity of the N-terminal segment
The sequences of IA
3
from S. castellii and S. cerevisiae
are aligned in Fig. 2. These show only 45% identity in
the N-terminal ‘segment’ (residues 2–32) that has been
demonstrated previously to contain the inhibitory
activity towards saccharopepsin [1–4]. Residues 33–35
are identical in both sequences and form a link
between the inhibitory N-terminal ‘segment’ and resi-
dues of the C-terminal ‘segment’. The C-terminal seg-
ment (residues 36–81; Fig. 2) from S. castellii IA
3
is
considerably longer than its counterpart (residues
36–68) in the S. cerevisiae polypeptide and differs sub-
stantially in sequence (Fig. 2). To establish whether the
respective C-terminal segments might have an influence
(beneficial or detrimental) on any inhibitory activity
that might be intrinsic to the N-terminal segments,
chimaeric proteins were engineered in which residues

influence on inhibition of saccharopepsin.
The reciprocal chimaera, which consisted of residues
1–34 from S. castellii IA
3
fused to residues 35–68 from
the S. cerevisiae polypeptide, was slightly more effect-
ive as an inhibitor of PpPr than the wild-type
S. castellii IA
3
(cf. 5 and 2; Fig. 1), with the measured
K
i
falling into the single digit nanomolar range. Since
these two polypeptides differ only in the nature of
their C-terminal segments, it would appear that the
C-terminal segment (residues 35–81) from S. castellii
IA
3
has a slight detrimental effect on the inhibitory
activity against PpPr that is intrinsic to its own N-ter-
minal segment. This interpretation was examined by
producing a shorter variant of the S. castellii sequence
that lacked any C-terminal segment and so consisted
only of residues 2–34. This had a comparable inhibi-
tory potency to that of the chimaera (cf. 6 and 5;
Fig. 1). The detrimental effect of S. castellii residues
35–81 may result from adverse interaction(s) occurring
either within the full-length S. castellii polypeptide
(residues 1–81) itself or between the C-terminal seg-
ment of the polypeptide and PpPr at a remote site far

(inhibitor 4) is even more effective
than inhibitor 6 against saccharopepsin (K
i
< 0.1 nm)
[1–4]. This behaviour stands in stark contrast to that
observed against PpPr where inhibitor 6 was > 500-
fold more effective than inhibitor 4 (Fig. 1). Conse-
quently, the effect of exchanging residues within the
inhibitory sequence of 6 was examined. Replacement
of the S. castellii residues 24–34 by the corresponding
residues from S. cerevisiae IA
3
had only a small (three-
to four-fold) adverse effect on inhibitory potency
against PpPr (cf. 7 and 6; Fig. 1). However, when the
key residues K18 and D22 that have been shown to be
so important in restricting the activity of S. cerevisiae
IA
3
to saccharopepsin as its sole target proteinase were
introduced into the S. castellii sequence in place of the
intrinsic M18 ⁄ K22 pair, the inhibitory activity against
PpPr was essentially destroyed (cf. 8 and 6; Fig. 1).
Thus, it would appear that the residues at positions 18
and 22 again play a decisive role, allowing effective
inhibition of PpPr by the S. castellii polypeptide.
Changes in other locations, including the ‘remote’
attachment of residues 35–81 from its own C-terminal
segment, cause only minor perturbation of the inhibi-
tory potency intrinsic to the N-terminal segment.

X5–X11, Fig. 3B) would be long enough to make
some of the predicted contacts with the side-chains of
residues such as Asp161, Asp164 and Glu17 on the
surface of PpPr; and an extension of nine amino
acids (residues X3–X11) would exploit the potential
binding site offered by this patch to the full (Fig. 3B).
Consequently, IA
3
variants with four (HHZQ) and
seven (HHHHHZQ) residue extension sequences,
respectively, were designed initially to introduce the
appropriate number of potentially positively charged
(at the experimental pH of 4.7) histidine residues (at
positions X8–X9 or positions X5–X9, respectively)
followed by a norleucine residue (indicated by Z, at
position X10) and a glutamine (residue X11) in place
of the naturally occurring N-terminal (methionine)
A
B
Fig. 3. Representation of PpPr and the extension residues of IA
3
.
(A) Negatively ionisable surface residues (red) adjacent to the edge
of the active site of PpPr; the active site is occupied by a putative
helical IA
3
inhibitor with the residue at its N-terminus serving as a
potential attachment point for an extension; (B) potential interac-
tions of the indicated negatively ionisable surface residues (red)
of PpPr with several positively ionisable amino acids of the

However, the longer (H)
5
ZQ-extended variant showed
a seven-fold improvement in potency against PpPr
(cf. 10 and 7; Fig. 1).
Since this seven-residue extension was already suffi-
cient to engender an improvement of inhibitory
potency against PpPr, the extension sequence was
lengthened further to include all seven histidine resi-
dues indicated by the model. The additional two histi-
dine residues (at positions X3 and X4; Fig. 3B) were
introduced downstream from a methionine and a
lysine residue (at positions X1 and X2, respectively),
the logic for which will be substantiated below. These
four MKHH residues were thus introduced upstream
from the (H)
5
-containing extension described above to
generate the sequence MK(H)
7
MQ (Fig. 3B). Coinci-
dentally, this extension contains sufficient histidine res-
idues to enable it to be used as an affinity tag for
purification purposes. In all of our previous studies
with IA
3
[1–4], recombinant protein versions such as
inhibitors 1–3 and 5 (Fig. 1) were purified to homogen-
eity from E. coli lysates by nickel-chelate chromatogra-
phy, facilitated by a LE(H)

(35–81; Fig. 2) comprising the C-terminal segment
were systematically deleted, in blocks of 12 ⁄ 13 residues
at a time. Truncation of the N-terminally tagged
S. castellii polypeptide (inhibitor 11) at residue Q68
generated inhibitor 12 which corresponded in overall
length to S. cerevisiae IA
3
. Although this resulted in a
seven-fold weakening in potency against PpPr (cf. 12
with 11; Fig. 4), a subnanomolar K
i
value was still
recorded for inhibitor 12. Further truncation at resi-
dues Y57 and K45, respectively (inhibitors 13 and 14;
Fig. 4) did not cause any further significant loss of
inhibitory potency against PpPr. Thus, in contrast to
the detrimental effect that was described above
when residues 35–81 were attached in the full-length,
C-terminally tagged inhibitor 2, the presence of
residues 69–81 at the C-terminus of the N-terminally
tagged S. castellii polypeptide appears to confer a
benefit to the inhibition of PpPr (cf. inhibitors 11 and
12; Fig. 4). This was substantiated by the data
obtained for the chimaeric inhibitor 15 (Fig. 4) which
was identical in length to inhibitor 12 but contained
residues 35–68 from S. cerevisiae IA
3
as the C-terminal
segment in place of the counterpart S. castellii residues
of inhibitor 12. Both inhibitors had comparable K

desM(X1)-extended IA
3
would still contain its His-tag
and so could be removed by nickel-chelate chromato-
graphy. Unlike an N-terminal lysine residue, glutamine
(at position X11; Fig. 3) does not in itself constitute a
stop point for cleavage by cathepsin C. However, if
dipeptide removal by cathepsin C is performed in the
presence of an excess of glutamine cyclotransferase,
once an N-terminal glutamine residue is newly exposed,
it is rapidly converted into pyroglutamic acid. Further
digestion by cathepsin C is thus prevented, leaving the
cyclised Q as the N-terminal residue (replacing the nat-
urally occurring Met1) of each IA
3
polypeptide. Appli-
cation of this trimming treatment to the longest and
shortest variants with the wild-type S. castellii sequence
(inhibitors 11 and 14) and to the chimaeric inhib-
itor 15, generated polypeptides 11T, 14T and 15T,
respectively, each with a pyrrolidone carboxylic acid
residue (cyclised Q) at its N-terminal end (Fig. 4). Each
trimmed polypeptide was purified as described in the
Experimental Procedures section by passage through a
nickel-chelate column to remove any residual parent
IA
3
with its intact histidine tag together with the two
enzymes used in the trimming procedure which are also
both C-terminally His-tagged. The purity, identity and

16
16T
Identity Residue number
*Q
MK(H)
MQ
34
68
1 81
K
i
(nM)
0.1 ± 0.1
30 ± 4
0.7 ± 0.1
K
Saccharopepsin
PpPr
MK(H)
MQ
MK(H)
MQ
0.3 ± 0.2
4 ± 0.3
Q
0.8 ± 0.1
45
57
S
Y

*Q
*Q
NE
NE
Fig. 4. Inhibition at pH 4.7 of PpPr and S. cerevisiae (saccharopepsin) by variant forms of IA
3
from S. castellii and S. cerevisiae. Sequences
of IA
3
from S. castellii and S. cerevisiae are depicted by open and dark-shaded boxes respectively, with residue 2 and the C-terminal residue
of each length variant identified individually. The MK(H)
7
MQ extension was positioned upstream from residue 2 at the N-terminus of inhibi-
tors 11-16. Inhibitors 11T, 14T, 15T & 16T were generated by removal of this extension by cathepsin C trimming to leave a cyclised Q resi-
due (= *Q) at the N-terminus of each polypeptide.
N-terminal extension of IA
3
T. J. Winterburn et al.
3690 FEBS Journal 274 (2007) 3685–3694 ª 2007 The Authors Journal compilation ª 2007 FEBS
values given in parentheses. Histidine was absent, indi-
cating the purity of the trimmed polypeptide that
resulted from the chromatographic procedures (see
Experimental procedures) and substantiating the
complete absence of His-tagged parent polypeptide or
any partially-processed intermediate. Directly compar-
able results were obtained for all of the other inhibitor
pairs described in Fig. 4 (data not shown for brevity).
For the chimaeric 15 ⁄ 15T pair and the shortest
14 ⁄ 14T pair, removal of the N-terminal extension in
this way resulted in an approximately 35-fold loss in

an improvement of approximately 100-fold in inhibitory
potency against PpPr relative to the C-terminally tagged
polypeptide (cf. 16; Fig. 4; 1, Fig. 1). This modification
thus transformed the ineffective polypeptide 1 into a
highly potent inhibitor with a subnanomolar K
i
value
against PpPr (16; Fig. 4). Once again, however, this gain
in potency was completely lost upon removal of the
N-terminal extension by treatment with cathepsin C.
The resultant, trimmed S. cerevisiae IA
3
reverted to
being as mediocre an inhibitor of PpPr as the original
construct with its C-terminal tag (cf. 16T; Fig. 4; 1,
Fig. 1).
Binding effects
An explanation for these effects may be advanced
based on remote interactions that occur outwith the
active site cleft of the target proteinase. Free IA
3
is
predominantly unstructured [5,6]. Neither S. cerevisiae
nor S. castellii IA
3
show any significant intrinsic affin-
ity for PpPr (inhibitors 1 and 16T and 2 and 11T) and
so the E + I « EI equilibrium lies well to the left.
When the extension with its multiple, positively ionisa-
ble histidine residues (at pH 4.7) is attached at the

purpose, five residues (X5–X9) resulted in an increase
in potency of almost an order of magnitude against
PpPr, with the X5 and X6 histidine residues potentially
establishing contacts with Asp164 and Glu17, respect-
ively, of the enzyme (Fig. 3B). Addition of a further
two histidine residues (X3 and X4) and a lysine at X2
consolidated this effect even more, resulting in a
further, more substantial gain in potency.
Since neither Glu17 nor Asp164 is conserved in the
sequence of saccharopepsin, the validity of this inter-
pretation was examined by determination of inhibition
constants for the interaction of the N-terminally exten-
ded inhibitors with saccharopepsin. The potencies of
inhibitors 9 (containing two histidines) and 10 (five his-
tidines) against this enzyme were closely similar and
comparable to that of the parent inhibitor 7 ( K
i
¼
0.4 ± 0.1, 0.3 ± 0.1 and 0.8 ± 0.1 nm, respectively).
Further lengthening to include all seven histidine resi-
dues of the MK(H)
7
MQ sequence resulted in extended
inhibitors that were only two- to ten-fold more potent
against saccharopepsin than their respective, trimmed
counterparts (cf. 15T and 15, 14T and 14 and 11T and
11; Fig. 4). Indeed, the trimmed S. castellii polypeptide
(inhibitor 11T; Fig. 4) had a potency against saccharo-
pepsin that was identical to that reported previously
[1] for the C-terminally histidine-tagged counterpart

3
upon encounter-
ing the active site of saccharopepsin, are already opti-
mized and so are sufficient by themselves to facilitate
tight, specific binding of this helical N-terminal seg-
ment of IA
3
. The E + I « EI balance thus lies far to
the right and the addition of further residues at the
N-terminus or beyond residue 34 of the inhibitory
segment is superfluous. However, in the case of PpPr,
the serendipitous positioning of negatively ionisable
residues in a patch adjacent to but remote from the
active site provides a capture site for positively ionisa-
ble residues in the N-terminal extension. By this
device, it is thus possible to transform IA
3
polypep-
tides with little intrinsic affinity for PpPr into inhibi-
tors with subnanomolar potency against this enzyme
as a target proteinase. For aspartic proteinases that do
not possess this fortuitous surface feature and which
are more distantly-related to saccharopepsin, including
those of clinical ⁄ agricultural relevance, it would appear
likely that changes will need to be made within the
inhibitory sequence of the N-terminal segment itself in
order to re-target the inhibitory activity of IA
3
.
Experimental procedures

bases encoding S. cerevisiae residues 1–34 whereas digestion
with NheI–XhoI permitted excision of the nucleotides enco-
ding residues 35–68. The respective excised fragments were
replaced with DNA encoding the corresponding residues
1–34 (inhibitor 5) or 35–81 (inhibitor 3) from S. castellii
IA
3
. Each relevant segment was amplified by PCR using
S. castellii IA
3
DNA as template and oligonucleotide pairs
containing the appropriate restriction enzyme sequence
(Table 1). The authenticity of each construct was confirmed
by sequencing. In this way, pET22b plasmids were gener-
N-terminal extension of IA
3
T. J. Winterburn et al.
3692 FEBS Journal 274 (2007) 3685–3694 ª 2007 The Authors Journal compilation ª 2007 FEBS
ated encoding chimaeric polypeptides which consisted,
respectively, of residues 1–34 from S. castellii followed by
residues 35–68 from S. cerevisiae IA
3
(inhibitor 5, Fig. 1)
or residues 1–34 from S. cerevisiae followed by residues
35–81 from S. castellii (inhibitor 3, Fig. 1).
To generate IA
3
polypeptides each extended at its N-ter-
minus and devoid of the LE(H)
6

of each desired length. Each forward primer
consisted of an NdeI consensus sequence followed by a Gln
codon before continuing inframe at the codon for residue 2
of the relevant IA
3
sequence. Each reverse primer encoded
stop codons in all three frames after the final desired IA
3
codon to ensure the appropriate target polypeptide length.
PCRs were performed with the high-fidelity PfuUltra
TM
polymerase (Stratagene). Following gel purification, each
amplicon was treated with NdeI and XhoI, prior to ligation
into the newpet-22b vector that had been similarly digested.
Sequencing confirmed the authenticity of each construct. In
this way, pET-22b plasmids encoding inhibitors 11–16 , each
with an N-terminal MK(H)
6
HMQ extension (Fig. 4) were
generated. The oligonucleotides used for each PCR
employed in this series are listed in Table 1.
Treatment to remove the N-terminal extension from each
extended IA
3
polypeptide was carried out using the TAG-
Zyme
TM
system, first described by Pedersen et al. [7], accord-
ing to the manufacturer’s instructions (Qiagen, Crawley,
UK). Briefly, this involved pretreatment of the DAPase

3
Table 1. Construction of mutant forms of IA
3
from S. cerevisiae and S. castellii. The indicated pairs of forward (F) and reverse (R) oligonucle-
otide primers were used to introduce the desired changes in S. castellii or S. cerevisiae IA
3
, thus generating each of the identified variants.
Identity Oligonucleotide sequences (5¢ to 3¢)
3 (F) CTAGCTAGCCCTGAAAGTAAGGAAAAAATGAAGAC
(R) CCGCTCGAGATGATCCATCAATTCATCTTTATCTTG
5 (F) GGAATTCCATATGAGTGATAAAAACGCTAACGTC
(R) CTAGCTAGCCATGTTTTTCATTCCTTCACTAGC
11 (F) GGAATTCCATATGCAGAGTGATAAAAACGCTAACGTCT
(R) CCGCTCGAGCGGCTATCTATCTAATGATCCATCAATTCATCTTTATC
12 (F) GGAATTCCATATGCAGAGTGATAAAAACGCTAACGTCT
(R) CCGCTCGAGCGGCTATCTATCTATTGTTCTTGCTTCCCAGCACC
13 (F) GGAATTCCATATGCAGAGTGATAAAAACGCTAACGTCT
(R) CCGCTCGAGCGGCTATCTATCTAATACGAATCTTGAGCTTTCTTTTC
14 (F) GGAATTCCATATGCAGAGTGATAAAAACGCTAACGTCT
(R) CCGCTCGAGCGGCTATCTATCTATTTTGTCTTCATTTTTTCCTTACTTTC
15 (F) GGAATTCCATATGCAGAGTGATAAAAACGCTAACGTCT
(R) CCGCTCGAGCGGCTATCTATCTACTCCTTCTTATGCCCCGCC
16 (F) GGAATTCCATATGCAGAATACAGACCAACAAAAAGTGAG
(R) CCGCTCGAGCGGCTATCTATCTACTCCTTCTTATGCCCCGCC
newpetTOP CTAGAAATAATTTTGTTTAACTTTAAGAAGGAGATAAGCTTATGAAACACCACCACCACCACCACCA
newpetBOT TATGGTGGTGGTGGTGGTGGTGTTTCATAAGCTTATCTCCTTCTTAAAGTTAAACAAAATTATTT
T. J. Winterburn et al. N-terminal extension of IA
3
FEBS Journal 274 (2007) 3685–3694 ª 2007 The Authors Journal compilation ª 2007 FEBS 3693
were identified by SDS ⁄ PAGE, pooled and concentrated, if

)1
solu-
tion of a-cyano-4-hydroxy-trans-cinnamic acid matrix (Sig-
ma, Poole, UK) plus 10 mm ammonium di-hydrogen
phosphate in 50% (v ⁄ v) acetonitrile in 0.1% (v ⁄ v) trifluoro-
acetic acid, mixed and allowed to air dry prior to analysis.
The mass spectrometer was internally calibrated using a
matrix ion at 568.13 Da and mass measurement accuracy
was typically ± 0.01%. The resultant data were analysed
using the massXpert computer program [8]. Modelling
calculations were carried out on an SGI Octane work-
station (Silicon Graphics, Geneva, Switzerland) with
dual R12000 processors, using the moloc program (Gerber
Molecular Design, Amden, Switzerland), as reported pre-
viously [1,4].
Acknowledgements
Supported by awards (to J.K.) from the UK Biotech-
nology and Biological Sciences Research Council
(grant numbers 72 ⁄ C13544 and 72 ⁄ 0014846). We are
very grateful to our colleagues Jakob Winther and
Anette Bruun (formerly of the Carlsberg Laboratory,
Copenhagen, Denmark) for help with production of
recombinant PpPr; to John Fox, Alta Biosciences, Bir-
mingham, for provision and analysis of synthetic pep-
tide variants of IA
3
; and to Doug Lamont and Kenny
Beattie, University of Dundee, for carrying out mul-
tiple mass spectrometry analyses of the IA
3

3
. Biochemistry 45, 13585–13596.
6 Green T, Ganesh O, Perry K, Smith L, Phylip LH,
Logan TM, Hagen SJ, Dunn BM & Edison AS (2004)
IA
3
, an aspartic proteinase inhibitor from Saccharomyces
cerevisiae, is intrinsically unstructured in solution. Bio-
chemistry 43, 4071–4081.
7 Pedersen J, Lauritzen C, Madsen MT & Dahl SW (1999)
Removal of N-terminal polyhistidine tags from recombi-
nant proteins using engineered aminopeptidases. Protein
Expr Purif 15, 389–400.
8 Rusconi F & Belghazi M (2002) Desktop prediction ⁄
analysis of mass spectrometric data in proteomic projects
by using massXpert. Bioinformatics 18, 644–645.
N-terminal extension of IA
3
T. J. Winterburn et al.
3694 FEBS Journal 274 (2007) 3685–3694 ª 2007 The Authors Journal compilation ª 2007 FEBS