Protein Engineering vol.4 no.2 pp.155-161. 1990
Correlation between stability of a protein and its dipeptide
composition: a novel approach for predicting in vivo stability of a
protein from its primary sequence
Kunchur Guruprasad, B.V.Bhasker Reddy and
Madhusudan W.Pandit1
Centre for Cellular and Molecular Biology. Hyderabad - 500 007, India
'To whom correspondence should be addressed
Statistical analysis of 12 unstable and 32 stable proteins
revealed that there are certain dipeptides, the occurrence of
which is significantly different in the unstable proteins
compared with those in the stable ones. Based on the impact
of these dipeptides on the unstable proteins over the stable
ones, a weight value of instability is assigned to each of the
dipeptides. For a given protein the summation of these weight
values normalized to the length of its sequence helps to
distinguish between unstable and stable proteins. Results
suggest that the in vivo instability of proteins is possibly
determined by the order of certain amino acids in its
sequence. An attempt is made to correlate metabolic stability
of proteins with features of their primary sequence where
weight values of instability for a protein of known sequence
could thus be used as an index for predicting its stability
characteristics.
Key words: instability index/primary structure/protein instability/
stability prediction/turnover rate
Introduction
Proteins are known to degrade rapidly when their conformations
are altered by mutations, incorporation of amino acid analogs,
denaturation or premature chain terminations, etc. (Goldberg and
St John, 1976). All such modifications are likely to prevent proper
folding or to disrupt tertiary structure which can make the
resulting aberrant polypeptide prone to degradation. However,
normal cellular proteins also display a wide range of half-lives;
turnover rates of individual proteins can differ as much as
1000-fold (Rechsteiner et al., 1987). These observations have
led to many speculations about the features of proteins that elicit
proteolysis. Several hypotheses have been proposed to explain
the intracellular stability (or instability) of proteins; these
hypotheses have been reviewed recently (Rechsteiner et al.,
1987). Current status suggests that among several factors,
sequence specific properties (Bachmair etai, 1986; Rogers
et al., 1986), global features and the location of a protein in the
cell are important in deciding the intracellular stability of a
protein. In this communication, we report on an interesting
observation that has emerged from the analysis of primary
sequences of a set of proteins that are known to degrade rapidly
(Rogers et al., 1986), and the comparison of these data with those
obtained from a set of stable proteins (i.e. those with
comparatively high in vivo half-lives). The proteins which are
known to degrade very rapidly are found to contain certain
dipeptides in relatively large proportions; these might be directly
or indirectly involved in the rapid degradation of proteins. Our
analysis also points out that there is yet another set of dipeptides,
the existence of which could be equally important for the stable
proteins. The statistical analytical procedure leading to the
dipeptide instability weight value reported here indicates the
possibility of its use in predicting whether a particular protein
would be unstable (or stable) in vivo.
Materials and methods
The strategy and the development of the instability index
As classified by Rogers et al. (1986), a set of 32 proteins with
an in vivo half-life of > 16 h was taken as a class of stable proteins
and a set of 12 proteins with an in vivo half-life of <5 h was
taken as an unstable class of proteins (Table I). The sequences
of these two sets of proteins were analysed separately.
The frequency of occurrence of each of the 20 amino acids
was calculated in the unstable as well as stable class of proteins
and was compared with the frequency of occurrence of various
amino acids in the total Protein Sequence Database of the PIR
(Release 12.0). These data (Figure 1) explicitly bring out certain
notable differences between the frequency of occurrence of
various amino acids in unstable and stable proteins. It can be
seen from Figure 1, that in the case of unstable proteins, amino
acids Met(M), Gln(Q), Pro(P), Glu(E) and Ser(S) are found to
occur with a relatively high frequency. The PEST hypothesis
proposed earlier by Rogers et al. (1986) reports the presence of
regions consisting of amino acids (Pro)P, (Glu)E, (Ser)S and
(Thr)T in the unstable proteins. Our observation indicates that,
unlike the PEST hypothesis, Thr(T) does not occur more
frequently in unstable proteins compared with stable proteins.
In fact, the presence of amino acids such as Met(M) is more
frequent in unstable proteins. Similarly it was noted that amino
acids Asn(N), Lys(K) and Gly(G) occur with relatively higher
frequencies in stable proteins when compared with the unstable
ones. Therefore, it appears that the PEST hypothesis could partly
be a reflection of the consequence of differential occurrence of
certain amino acids. However, the PEST hypothesis talks about
regions of negatively charged amino acids that tend to be clustered
and generally flanked by positively charged amino acids.
Therefore, it appears that instability or stability characteristics
are probably governed by an arrangement of certain amino acids
in a specific order. Hence the smallest unit that defines order
in a sequence being a dipeptide, the occurrence of an amino acid
in juxtaposition to the other might be a significant factor in
determining the stability of the protein. Therefore we chose to
search for the occurrence of all 400 possible dipeptides in the
two classes of proteins.
The expected (probable) occurrences of dipeptides were
calculated, assuming the constituents of dipeptides as independent
events, by equation:
= [A/°Cx)/7]
20 20
[N°(y)/T\ E E fl°(ry)
x = I y = I
(1)
where A^xy) and A^Cry) are the expected and the observed
occurrence respectively, of dipeptide xy\ and N°(x) and N°(y) are
the observed occurrences of amino acids x and v respectively.
T is the total number of amino acids in a particular class.
© Oxford University Press 155
atMasarykUniversityonJanuary29,2013http://peds.oxfordjournals.org/Downloadedfrom
K.Curuprasad, B.V.B.Reddy and M.W.Pandit
Table I. Proteins studied, their
Serial no. Proteina
(A) Stable proteins
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
(B) Unstable
33
34
35
36
37
38
39
40
41
42
43
44
ADK
ADH
AAT
CAI
CPA
CAT
CHY
CIS
CCC
AAD
DPA
DHF
ELA
FER
GPD
HEM
HEM
ABP
LDH
LYS
MYO
PAR
PGK
PLA
PYK
RNA
SOD
STI
SUB
THI
TPI
TRY
: proteins
EIA
a-Casein
/3-Casein
c-fos
c-myc
v-myb
HSP 70
HMG-CoA
ODC
P730
TAT
p53
half-life and II
Half-lifeb
(h)
133
139
66
72
137
80
50
128
26
118
80
97
46
61
84
209
209
65
171
16
127
41
207
78
181
61
41-200
40
84
155
122
44
0.5
2-5
2-5
0.5
0.5
0.5
1-2
1.5-3
0.5
1.0
0.5
0.5
PIR/EMBL*
Code
KIHUA
DEHUAB
XNPGDC
CRHU1
CPRTA
CSBO
KYBOA
YKPG
CCHU
OXPGDA
ADECD
RDBOD
ELPG
FRHUL
DEHUGL
HAHU
HBHU
JGECA
DEPGLH
LZCH
MYHO
PVRYC
KIBYG
PSKF2U
KICHPM
NRGPA
DSBOCZ
TISY
SUBSC
TXEC
ISBYT
TRBOTR
AD5001*
KABOSB
KBBOA2
HSCFOS*
HSMYCI*
GGCMYB1*
HHFF72
RDHYE
DCMSO
ASAP3R*
RNTATR*
MMP53*
II
29.4
17.4
36.3
25.1
21.5
27.5
15.3
20.8
11.4
34.4
14.2
27.4
23.0
23.8
16.1
7.0
6.1
23.1
25.6
19.9
9.1
22.8
22.2
15.1
21.2
52.2
7.0
27.4
12.0
5.6
19.7
20.6
100.3
57.7
96.5
78.8
92.2
74.5
44.1
54.5
52.3
50.4
53.9
70.0
"Abbreviations: ADK, adenylate kinase 1; ADH, alcohol dehydrogenase;
AAT, aspartate amino transferase; CAI, carbonic anhydrase I; CPA.
carboxypeptidase A; CAT, catalase; CHY, chymotrypsinogen; CIS, citrate
synthase; CCC, cytochrome C; AAD, D-amino acid oxidase; DPA,
deoxyribose phosphate aldolase; DHF, dihydrofolate reductase; ELA,
elastase; FER, ferritin: GPD, glyceraldehyde 3-phosphate dehydrogenase;
HEM, haemoglobin; ABP, L-arabinose binding protein; LDH, lactate
dehydrogenase; LYS. lysozyme c: MYO, myoglobin; PAR, parvaJbumin;
PGK, phosphoglycerate kinase; PLA, phospholipase A2; PYK, pyruvate
kinase; RNA, ribonuclease A; SOD. superoxide dismutase; STI. soybean
trypsin inhibitor; SUB. subtilisin: THI. thioredoxin; TPI. triosephosphate
isomerase; TRY. trypsinogen: EIA. adenovirus early protein; HSP70,
heatshock protein 70; HMG-CoA, hydroxymethyl glutaryl-CoA; ODC.
ornithine decarboxylase; P730. phytochrome: TAT. tyrosine amino
transferase.
b
References for half-life are taken from Rogers et al. (1986).
The chi-square values, x2
(*y) between observed and expected
occurrences of dipeptides (xy) for each class of proteins were
calculated by equation:
x\xy) = (2)
An average value of x2
for each class of protein was calculated
using equation:
400
X2
= (1/400) L x\xy)
x\=\
(3)
X2
was then used as the confidence limit to select significant
dipeptides for each class of protein. The condition used for
selecting significant dipeptides for unstable and stable classes of
proteins respectively are:
X usCry) ^ X"uS
and
> x2
s
X\s-s(xy) w a s
also calculated from values of observed
occurrences of dipeptides in unstable and stable classes of proteins
respectively, using the following equation:
(4)
where A^s(ry) and N^(xy), are the observed occurrences of
dipeptides in unstable and stable classes of proteins respectively.
As we were interested in picking up factors which contribute to
the instability, in calculating x^is-sC*)7
). e a c n
squared difference
in the occurrence of dipeptides in unstable and stable classes was
weighted inversely by the occurrence of dipeptides in the stable
class. A third set of dipeptides was selected from these
X2
us-a(x
y) values satisfying the condition:
^ X2
us-5
The potential occurrence, P(xy) for each of the dipeptides, in
all the three sets was calculated by equation:
P(xy) = N°(xy) I PF(xy) (5)
Each of the above three sets of dipeptides was further classified
into two subsets depending upon P(xy) being significantly
different from unity (>1.5 or <0.64). The corresponding
potential values for dipeptides from all the three sets were used
to formulate the conditions given in Table II. We classified the
dipeptides that satisfy each of the conditions given in Table II
based on the chi-square value and its Pixy) of every dipeptide
for each class separately. Weight value of instability, corresponding
to each of the conditions, was obtained by summing up
yV°(.ry) for all dipeptides (xy) satisfying that condition. The
impact factor for /th condition (IF,) was estimated by equation:
IF, = (Vus, - Vs,) / Vs, (6)
where Vus, and Vs, are the normalized values of occurrence of
dipeptides satisfying /th condition in the unstable and stable class
of proteins respectively. The impact factor for the /th condition
156
atMasarykUniversityonJanuary29,2013http://peds.oxfordjournals.org/Downloadedfrom
Stability characteristics of protein in vivo
a
o
<
o
z
o
o
o
UJ
o
1 0 0
8 0
6 0
3
O
4 0
UJ
3
O
UJ
oc. 20
Ul
TOTAL PIR DATABASE
UNSTABLE PROTEINS
STABLE PROTEINS
W C M H Y F Q N I R D P T K E V S G A L
AMINO ACID
Fig. 1. Relative frequency of occurrence of individual amino acids per 1000 amino acids in various sets of proteins; PIR—Protein Sequence Database is taken
from Release 12.
was operated on, to bring it into a positive range, and was termed
as the instability weight value (IWV,) given by equation:
IWV,- = 2 + (IF, / |LIF|) (7)
where LIF was the lowest impact factor observed. The
contribution of each of the dipeptides towards instability was
obtained by summing the instability weight values corresponding
to the condition(s) satisfied by the dipeptide and termed as the
dipeptide instability weight value (DIWV). The DIWVs for all
400 combinations are presented as a matrix (GRP matrix) in
Table III. The instability index (II) for a protein was then
computed using the DIWV by equation:
Table II. Rule-based weightages
Conditions
If the
1.
2.
3.
4.
5.
6.
7.
dipeptide satisfies:
Pus > 1.50 and Ps
/>us < 1.50 and Ps
Pus < 0.64 and Ps
Pas > 0.64 and Ps
Pus.s > 1.50
PUM < 0.64
None of the above
< 1.50
> 1.50
> 0.64
< 0.64
conditions
Instability
weight value
+ 13.34
- 1.88
- 6.54
+24.68
+20.26
- 7.49
+ 1.0
II = (10/L) E DIWV(jciy,
i = i
(8)
where jc^y,-+1 is a dipeptide, L is the length of the sequence and
10 is a scaling factor. Instability indices for various proteins in
the stable as well as in the unstable class are given in column
5 of Table I.
Results and discussion
Dipeptides involved in instability of proteins
The observed and the expected frequency of occurrence of
dipeptides in the sets of both stable and unstable proteins helped
to identify the dipeptides which are predominant in either of these
sets. GRP-matrix (Table W) indicates that out of the 400 possible
dipeptides 81 (i.e. 20%) with DIWV > 1 may be involved in
the instability of the protein, whereas -72 (i.e. 18%) with DIWV
< 1 may contribute to the stability. We do not know at present
how the presence (or the absence) of dipeptides such as these
render a protein stable or susceptible to degradation. However,
the net effect of their combined presence in a protein appears
to reflect in its instability index that serves as a useful indicator
in deciding the stability (or instability) characteristics of the
protein.
In vivo half-life of a protein versus its instability index
We have presented various proteins that are used in our analysis
along with the values of their half-lives in Table I. It can be seen
from the results that all the unstable proteins have an II > 40,
whereas all the stable proteins, with the only exception of RNase
A, have an II < 40. The segregation of the two classes of proteins
based on the II is illustrated in Figure 2. This figure clearly brings
out differences between the two classes of proteins. The relatively
high value of the II (= 52.2) for RNase A puts this protein into
the class of unstable proteins; however, it is known that RNase
A in its native form is held together by four disulphide bridges
between Cys groups. Such disulphide bridges are known to impart
significant stability to the protein making it resistant to degradation
157
atMasarykUniversityonJanuary29,2013http://peds.oxfordjournals.org/Downloadedfrom
K.Guruprasad, B.V.B.Reddy and M.W.Pandit
Table III.
First amino
acid of
dipeptide
W
C
M
H
Y
F
0
N
1
R
D
P
T
K
E
V
s
G
A
L
GRP matrix of
Second
W
1.0
24.68
1.0
-1.88
-9.37
1.0
1.0
-9.37
1.0
58.28
1.0
-1.88
-14.03
1.0
-14.03
1.0
1.0
13.34
1.0
24.68
condition-basec
amino acid of dipeptide
C
1.0
1.0
1.0
1.0
1.0
1.0
-6.54
-1.88
1.0
1.0
1.0
-6.54
1.0
1.0
44.94
1.0
33.6
1.0
44.94
1.0
M
24.68
33.6
-1.88
1.0
44.94
1.0
1.0
1.0
1.0
1.0
1.0
-6.54
1.0
33.6
1.0
1.0
1.0
1.0
1.0
1.0
H
24.68
33.6
58.28
1.0
13.34
1.0
1.0
1.0
13.34
20.26
1.0
1.0
1.0
1.0
-6.54
1.0
1.0
1.0
-7.49
1.0
instability values for
Y
1.0
1.0
24.68
44.94
13.34
33.6
-6.54
1.0
1.0
-6.54
1.0
1.0
1.0
1.0
1.0
-6.54
1.0
-7.49
1.0
1.0
F
1.0
1.0
1.0
-9.37
1.0
1.0
-6.54
-14.03
1.0
1.0
-6.54
20.26
13.34
.0
.0
.0
.0
.0
.0
.0
0
1.0
-6.54
-6.54
1.0
1.0
1.0
20.26
-6.54
1.0
20.26
1.0
20.26
-6.54
24.68
20.26
1.0
20.26
1.0
1.0
33.6
400 possible dipeptides
N
13.34
1.0
1.0
24.68
1.0
1.0
1.0
1.0
1.0
13.34
1.0
1.0
-14.03
1.0
1.0
1.0
1.0
-7.49
1.0
1.0
I
1.0
1.0
1.0
44.94
1.0
1.0
1.0
44.94
1.0
1.0
1.0
1.0
1.0
-7.49
20.26
1.0
1.0
-7.49
1.0
1.0
R
1.0
1.0
-6.54
1.0
-15.91
1.0
1.0
1.0
1.0
58.28
-6.54
-6.54
1.0
33.6
1.0
1.0 -
20.26
1.0
1.0
20.26
D
1.0
20.26
1.0
1.0
24.68
13.34
20.26
1.0
1.0
1.0
1.0
-6.54
1.0
1.0
20.26
14.03
1.0
1.0
-7.49
1.0
P
1.0
20.26
44.94
-1.88
13.34
20.26
20.26
-1.88
-1.88
20.26
1.0
20.26
1.0
-6.54
20.26
20.26
44.94
1.0
20.26
20.26
T
-14.03
33.6
-1.88
-6.54
-7.49
1.0
1.0
-7.49
1.0
1.0
-14.03
1.0
1.0
1.0
1.0
-7.49
1.0
-7.49
1.0
1.0
K
1.0
1.0
1.0
24.68
1.0
-14.03
1.0
24.68
-7.49
1.0
-7.49
1.0
1.0
1.0
1.0
-1.88
1.0
-7.49
1.0
-7.49
E
1.0
1.0
1.0
1.0
-6.54
1.0
20.26
1.0
44.94
1.0
1.0
18.38
20.26
1.0
33.6
1.0
20.26
-6.54
1.0
1.0
V
-7.49
-6.54
1.0
1.0
1.0
1.0
-6.54
1.0
-7.49
1.0
1.0
20.26
1.0
-7.49
1.0
1.0
1.0
1.0
1.0
1.0
S
1.0
1.0
44.94
1.0
1.0
1.0
44.94
1.0
1.0
44.94
20.26
20.26
1.0
1.0
20.26
1.0
20.26
1.0
1.0
1.0
G
-9.37
1.0
1.0
-9.37
-7.49
1.0
1.0
-14.03
1.0
-7.49
1.0
1.0
-7.49
-7.49
1.0
-7.49
1.0
13.34
1.0
1.0
A
-14.03
1.0
13.34
1.0
24.68
1.0
1.0
1.0
1.0
1.0
1.0
20.26
1.0
1.0
1.0
1.0
1.0
-7.49
1.0
1.0
L
13.34
20.26
2(
-
.0
.0
.0
.0
.0
.0
).26
.0
.0
.0
.0
7.49
.0
.0
.0
.0
.0
.0
(Creighton, 1988). Therefore it appears that cross-linking between
Cys groups may override the impact of dipeptides. In our
approach we have not taken into account such higher-order
elements, which could modify the intrinsic stability characteristics
of the protein. It is also interesting to note that none of the unstable
proteins which we have studied has such a high degree of crosslinking
(disulphide bridges per 100 amino acids) between Cys
groups. Thus there is a direct correlation between the half-lives
of various proteins and the II arrived at by the method described.
The II of a protein, therefore, could be used directly to predict
whether a given protein is stable or unstable based on the II being
either <40 or >40 respectively, provided the sequence of the
protein is known.
Validity of the II of a protein in determining its stability
In order to test the merits of our method in predicting the stability
characteristics of proteins, we used another set of proteins which
was not a part of our database used in developing the II but for
which the data on in vivo half-life became available. These
proteins are listed in Table IV along with their IIs. It can be seen
from the last two columns of Table IV that, in every case, the
prediction based on the II of the protein is in complete agreement
with the conclusion based on the in vivo half-life of the protein.
This shows clearly that it is possible to predict the stability
characteristics of a protein with reasonable success by using its II.
Comparison with earlier hypotheses
The results of our analysis indicate that, although there is a
significant number of proteins which shows overlap between the
predictions made by the II described here and those governed
by PEST hypothesis, there exists a considerable number of
proteins in which there is no correlation between the PEST
hypothesis and the experimental observations. For example,
protein P730 (phytochrome) which contains PEST regions with
a negative PEST-FIND (PF) score between -2.21 and -30.47
is expected to be stable according to the PEST hypothesis.
However, this protein has been shown to be unstable (Pratt et al..
1974). The II calculated by our method for this protein is 50.4.
and categorizes it as an unstable protein in agreement with
S.No. OF PROTEIN (UNSTABLE, t )
36 40 44
X
ill
o
z
0 0
9 0
8 0
7 0
6 0
5 0
4 0
3 0
20
1 0
[
•
•
•
g
1 1
•
m
O
-o r
o
o
1 1 1
V
•
o
o (
o o
o
1 1 1
1 1 1 1 1
-
-
-
_
o
>°
o ° OOoo oo °
o o
o
°o o o
1 1 1 1 1 1 1 1 1 1
4 8 12 16 20
S.No. OF PROTEIN (STABLE, 0)
2 4 28 3 2
Fig. 2. Us of various proteins used in the analysis. The proteins could be
identified by their serial (S) nos. given in Table I. O. Protein with S. nos.
1—32; • . proteins with S. nos. 33—44.
experimental observations. There are a few other proteins, e.g.
AAT (aspartate amino transferase). DHF (dihydrofolate
reductase) and PYK (pyruvate kinase). which are reported to
contain at least one PEST region with a positive PF score and
158
atMasarykUniversityonJanuary29,2013http://peds.oxfordjournals.org/Downloadedfrom
Stability characteristics of protein in vivo
Table IV. Prediction of
known
Protein
Metallothionein (gold)
Protein kinase C
Phosphoenol pyruvate
carboxylase
Glucose-6-phosphate
dehydrogenase
Ornithine amino
transferase
Frustose-1,6-
diphosphatase
Cytochrome P450
Cytochrome b5
Fatty acid synthetase
Malate dehydrogenase
Arginase
Cytochrome b
Ubiquitin
Catalase
Na+
/K+
-ATPase
Transferrin receptor
Cytochrome c oxidase
/3-Galactosidase
Actin
c-AMP protein kinase
c-GMP protein kinase
Calmodulin
the stability
PIR
Code
SMRT2
KJRTC2
QYEC
DESHGC
XNRTO
PAPGF
O4RTP2
CBRT5
FZRTI
DEPGMM
WZBYR
CBRT
UQBY
CSRT
PWSHNA
JXHU
OBMS2
GBECE
ATHU
OKBO1R
OKBOG
MCHU
of a few proteins whose
Half-life
(h)
<2a
2-8"
5"
15"
19b
36b
48b
55b
71b
96b
96b
130b
9-40"
9^tt)a
9-40"
9^0a
41-200"
41-200"
41-200"
2-8"
2-8"
9-40a
U
66.82
40.29
46.8C
27.18C
34.21
29.16C
35.89
32.86
12.37
27.8C
28.78C
32.86
3O.33c
33.82
31.67C
22.06°
36.25
37.62C
36.19
49.3C
41.4C
27.53
half-life is
Whether stable
(S) or unstable
(US)
11
US
us
us
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
s
us
us
s
from
Half-life
US
US
US
S
S
S
S
S
S
S
S
S
S
S
S
S
S
s
s
us
us
s
a
Rechsteiner et al. (1987).
b
Dice and Goldberg (1975).
c
The II was calculated from the available sequence as the source of the
original protein was not mentioned in the cited reference.
therefore are expected to be unstable according to the proponents
of the PEST hypothesis. Experimental results, however, are not
in agreement with these expectations as these proteins have been
reported to be stable in vivo (Rogers et al., 1986). It is interesting
to note that Us for these proteins predict them to be stable
(Table I) in keeping with the experimental results. These
examples indicate that instability of a protein is governed by
factors other than those suggested by the PEST hypothesis.
When we examined PF scores for the PEST regions for the
set of proteins given in Table IV, it was observed that predictions
based on the PEST hypothesis and II did not tally in two cases.
Sodium potassium ATPase (PWSHNA) has three PEST regions
with positive PF scores of 6.93, 5.78 and 5.05 and therefore
should be unstable. However, the prediction on the II suggests
that the protein is stable and is in accordance with the
experimental observation (Rechsteiner etai, 1987). Metallothionein
(SMRT2) completely lacks PEST regions with a positive
PF score. But this protein has an II, which predicts that the protein
is unstable—the conclusion is again in agreement with
experimental observation (Rechsteiner etai, 1987). The II
appears therefore to pick some features in the sequence which
are overlooked by the PEST hypothesis. It may be pointed out
that out of the 81 dipeptides significant for the unstable class
(DIWV > 1 in Table HI) only 39 are originating from dipeptides
containing either P, E, S or T. More than half of the dipeptides
are exclusive of any of these amino acids. These facts establish
the significance of dipeptides devoid of P, E, S or T. In addition,
it is interesting to note that there are 25 dipeptides containing
either P, E, S or T which have DIWV < 0 indicating that in
fact, they contribute significantly to the stability of the protein,
in contrast to the proponents of the PEST hypothesis. It is possible
to carry out rigorous experimental checks by selective clipping
of residues, which contribute significantly to the high PF score,
and following the fate of the residual protein in terms of stability.
Rogers et al. (1986) have reported that /3-casein at its N-terminal
end (residues 1 -25) has only one PEST region with a positive
PF score. The clipping of this region should lead to a residual
protein which should be stable; however, the II suggests that the
residual protein will remain unstable. Similarly these workers
have reported another protein, v-myb, where residues 1 —16 at
the N-terminal end contain a PEST region with a positive PF
score (1.85). As proposed by them, removal of this region should
make the protein stable. The II for the residual v-myb, however,
suggests that it would be unstable. Experimental checks such as
these will help in arriving at discrete features which play a crucial
role in conferring stability on the protein.
Another hypothesis proposed by Bachmair et al. (1986), based
on the studies on (3-galactosidase and termed as the N-end rule,
suggests that a specific amino acid at the N-terminus can serve
as one of the determinants for in vivo half-life of a protein. As
the N-end rule is derived from studies on a single protein, one
cannot be certain about its applicability to all the proteins in
general. In addition, it has already been suggested that the Nend
rule may be primarily cotranslational (Rechsteiner et al.,
1987), and therefore has a limited use in determining the stability
characteristics of a protein in general. We have, therefore, not
exhaustively compared our data with that given by the N-end
rule. However, preliminary examination of such data indicates
that predictions based on the N-end rule are not in anyway better
than those based on the PEST hypothesis.
Analysis was carried out for all the proteins whose sequences
are known from the SWISS-PROT Protein Sequence Database
(Release 13) in order to pick up proteins that have very low II
values (< 10). Similarly, proteins that have very high II values
(> 90) were also picked up. For the reasons discussed earlier,
proteins that have a cysteine content higher than RNase A were
eliminated. Out of these, unique proteins showing extremely high
as well as low II values are listed in Table V. These proteins
would prove to be good candidates to provide the experimental
validity to the predictions in regard to their instability based on II.
The results allow us to make certain plausible conjectures about
point mutations altering the stability of a protein. The GRP-matrix
consists of certain dipeptides with high positive as well as negative
values. Point mutations in an unstable protein may lead to a
change from one dipeptide with a high positive value to another
with a negative value, consequently making a protein stable.
A single residue change in a protein can affect the contribution
of instability weight values of two dipeptides, which arise
due to residues on either side of it. Therefore all possible
tripeptides were first generated where the dipeptides comprising
them had DIWV values either > 1 or <0 (refer to Table HT).
The difference in the II value (AII^) for a pair of tripeptides
'axb' and 'ayb' obtained by changing the central residue and
keeping the neighbouring residues the same was estimated by
the relation:
AIItri = [DIWV(ax) + DIWV(xfc)] [DlWV(ay)
+ DIWVO*)] (9)
We have analysed the data on substitutions in the central
159
atMasarykUniversityonJanuary29,2013http://peds.oxfordjournals.org/Downloadedfrom
K.Guruprasad, B.V.B.Reddy and M.W.Pandit
Table V. Unique proteins with extreme values of Us
Protein name
(A) Proteins with II > 90 and length > 50
Amelogenin
Calspermin
Beta casein precursor
Core antigen
Contiguous repeat polypeptide (CRP) precursor
Early El A 11 kd protein
Early El A 32 kd and 26 kd proteins
Early El A 6 kd protein
EBNA-2 Nuclear protein
Extensin precursor
Alpha/beta-gliadin precursor (Prolamin)
Gamma/gliadin precursor
low mol. wt glutenin precursor
Bl-Hordein
Sperm histone (Protamine)
Involucrin
Myc proto-concogene protein
Sperm-specific protein PHI-0
Salivary proline-rich protein precursor
Rev protein anti-repression transactivator
Ribosomal protein S18
40S Ribosomal protein S26
Spermatid-specific protein
Transition protein 2
Female-specific transformer protein
Tequment phosphoprotein
11 kd protein
Probable E3 protein
(B) Proteins with II < 10 and length >50
Acylphosphatase, erythrocyte
Antifreeze protein 1IA7 precursor
ATP synthase lipid-binding protein
Azurin iso-1
Cytochrome C556
Cytochrome C
DNA-binding protein (7 kd)
Fatty-acid-binding protein
Type-1 fimbrial protein
Nitrogen regulatory protein PII
Haemoglobin alpha-A chain
Hisactophilin
Small histidine-alanine-rich protein precursor
Histidine-rich-protein precursor
Acrosin inhibitor II
Insulin
Light-harvesting protein, alpha chain (LH-2)
Mammary-derived growth inhibitor
Retinoic-acid-induced differentiation factor
Major outer membrane lipoprotein precursor
Myoglobin
Outer membrane protein precursor (porin)
Outer mitochondrial membrane protein (porin)
Profilin
Retinol-binding protein II. cellular (CRBP-II)
50S Ribosomal protein LI
30S Ribosomal protein S14
S-100 protein, alpha chain
S-Antigen protein precursor
Probable early protein GP5
Gene 5 protein
Early protein GP5B
Gene 15 Protein
Lysis protein
SWISS-PROT Code
A JVI CXJJ D U V UN
CALSSRAT
CASBSBOVIN
CORASHPBV4
CRPPSRAT
E111SADEM1
E1ASADEN2
EIA6SADEN5
EBN2SEBV
EXTNSDAUCA
GDAOSWHEAT
GDB2SWHEAT
GLTASWHEAT
HOR1SH0RVU
HSPSCOTJA
INVOSLEMCA
MYCSFELCA
PHI0SHOLTU
PRPlSHUMAN
REVSHIV13
RS18SMARP0
RS264DROME
SSS1SSCYCA
STP2SMOUSE
TRSF4DROME
USO9SHSV11
V11KSPRV
VE3SBPV2
ACYESHUMAN
ANPXSSEAM
ATPHSSOYBN
AZU1SMETJ
C556 SAGRTA
CYCSANAPL
DN7ASSULAC
FABHSRAT
FM1ASECOLI
GLNBSECOLI
HBASAEGHO
HIAPSDICDI
HRPSPLAFF
HRP1SPLAFA
IAC2SBOVIN
INSSBALBO
LHA2SRHOCA
MDGISBOVIN
MK1SMOUSE
MULISERWAM
MYGSHORSE
OMPCSSALTY
PORISNUECR
PROFSACACA
RET2SRAT
RL1SBACST
RS14SMYCCA
S10ASBOVIN
SANTSPLAFW
VG5SBPPH2
VG5SSPV4
VG5BSBPPZA
VI5SVACCV
VLYSSBPT7
Table VI. Substitutions of central amino acid
likely to) alter All by > 75 as
of All are given in brackets)
a
Trp
Cys
Met
His
Tyr
Gin
x—y
Met-Ala(105)
His-Val(84)
His-Gly(87)
His-Thr(77)
His-Gly(87)
Asn-Gly(75)
His-Gln(92)
His-Val(92)
His-Met(80)
His-Gln(116)
His-Arg(116)
His-Arg(75)
His-Thr(99)
Pro-Gln(78)
Pro-Gln(78)
Ser-Gln(92)
Ser-Gln(75)
Ser-Arg(75)
Tyr- Pro(98)
Tyr-Gly(75)
Tyr- Pro(78)
Tyr-Trp(86)
Tyr-Gly(87)
Asn-Gly(87)
Ile-Thr(75)
Ile-Gly(106)
Met-Trp(87)
Met-Tyr(75)
Met-Arg(99)
Met-Glu(116)
Met- Ala(86)
Met-Arg(92)
Met-Gly(84)
Met-His(78)
Met-Arg(86)
Met-Glu(75)
Met-Glu(75)
Met-Trp(81)
His-Arg(80)
Ser-Cys(78)
Ser-Tyr(87)
Ser-Cys(75)
Ser-Tyr(83)
Ser-Phe(75)
Ser-Val(75)
Ser-Tyr(78)
in a tripeptide (axb) that are
a result of the replacement of x by v (values
b a
His
Tyr Asn
Tyr
Asn
lie lie
lie
Tyr
Tyr
Tyr Arg
Tyr
Tyr
Asn
Asn
Phe
Val
Cys
Pro
Pro
Met
Try
Asp
Ala
Ala
He
Glu
Glu Asp
His
His Thr
His
His Lys
His
Tyr
Tyr
Pro
Pro
Pro
Ser Glu
Ser
Tyr
Gin Ala
Arg
Pro
Pro
Pro
Pro
Glu
.t— v
Ile-Gln(75)
Ile-Thr(77)
Ile-Gly(lll)
Glu-Pro(98)
Glu —Lys(80)
Glu-Val(87)
Glu-Lys(78)
Trp-Try(75)
Trp-Gly(87)
His-Gly(80)
His-Gly(80)
Arg-His(99)
Arg-Tyr(133)
Arg —Asn(113)
Arg-Pro(99)
Arg-Gly(lll)
Arg-Gly(87)
Arg-Tyr(139)
Arg-Pro(l03)
Ser-Tyr(87)
Ser-Tyr(83)
Ser— Asn(78)
Ser-Tyr(78)
Ser-Gly(78)
Arg-Thr(80)
Ser— Lys(78)
Glu-Gln(78)
Glu-Asn(81)
Met-Ile(86)
Met-Ile(87)
Arg-Pro(99)
Arg-Gly(86)
Arg-Leu(75)
Arg-Pro(l 05)
Arg-Leu(78)
Cys-His(78)
Cys-Trp(107)
Cys-His(92)
Cys-His(78)
Cys-His(93)
Cys-Asp(lOl)
position of a tripeptide along with their associated AIItrj >
and listed only those amino acid replacements (Table VI)
the AE
can be
be:
where
before
[trj value changes significantly (
b
Glu
Glu
Glu
Cys
He
Asp
Pro
His
Asn
Tyr
lie
Trp
Trp
Trp
Trp
Trp
Asn
Arg
Arg
Arg
Pro
Pro
Glu
Glu
Trp
Pro
Cys
Cys
His
Pro
Trp
Trp
Trp
Arg
Arg
Trp
Thr
Thr
Trp
Thr
Thr
/alues,
where
>75). The formula that
used for estimating the change in the n of a protein
All
protein
= (10/L) *
AIIprotejn is the difference in II
and after the reiplacement of
Alltn
would
(10)
of a protein of length L
residue x bv residue v.
160
atMasarykUniversityonJanuary29,2013http://peds.oxfordjournals.org/Downloadedfrom
Stability characteristics of protein in vivo
Therefore depending upon the length of the protein and AIIproKin
value, one can choose effective substitutions from AIItri values
given in Table VI, so that the II of the protein assumes any value
<40 and hence becomes stable.
As AIIprolein is inversely proportional to the length of the
protein, a single substitution may not be sufficient to change II
to the expected level; hence, more than one substitutions may
be necessary. It should also be kept in mind that some of the
substitutions which suggest improvement in the stability may not
be desirable in terms of the functional requirements of the protein.
Only those substitutions which would enhance the stability of a
protein without affecting its function would be useful.
Conclusions
Our analysis and the method developed for predicting the in vivo
stability of a protein suggest that the primary determinants of
the stability of the protein probably reside in its primary structure
and is an intrinsic property of a protein. There appears to be a
correlation between the sensitivity of a protein to in vivo
degradation and the presence of certain dipeptides in it. The
overall influence of such dipeptides appears to contribute to
instability or stability characteristics of proteins. Apart from these
sequence-dependent characteristics several factors, e.g. structuredependent
features, the presence of disulphide bridges, ligand
binding, protease recognition mechanisms, etc., are known to
determine the in vivo protein stability. Therefore stability of a
protein as manifested in vivo could be a net effect of the
contributions made by several such factors. Our work indicates
that sequence-specific elements is one of the significant factors
which play an important role in determining the stability of a
protein. The observations presented here might help open new
vistas, where the knowledge about the dipeptides having specific
characteristics could be used in the modification of existing
proteins or in the design of novel protein molecules of a desired
stability.
References
Bachmair.A., Finley,D. and Varshavsky.A. (1986) Science, 234, 179-186.
Creighton.T.E. (1988) BioEssays, 8, 57-63.
Dice,J.F. and Goldberg.A.L. (1975) Arch. Biochem. Biophys., 170,213-219.
Goldberg,A.L. and St John.A.C. (1976) Annu. Rev. Biochem., 45, 747-803.
Pratt,L.H., Kidd.G. and Coleman.R. (1974) Biochim. Biophys. Aaa, 365,
93-107.
Rechsteiner.M., Rogers.S. and Rote,K. (1987) Trends Biochem. Sri., 12,
390-394.
Rogers.S.. Wells.R. and Rechsteiner.M. (1986) Science, 234, 364-368.
Received on May 10, 1990; accepted on August 30, 1990
161
atMasarykUniversityonJanuary29,2013http://peds.oxfordjournals.org/Downloadedfrom