Quantitative structure-activity relationships of pine weevil antifeedants, a multivariate approach

Article abstract of DOI:10.1021/jf070014p

Antifeedant activity of mainly phenylpropanoic, cinnamic, and benzoic acids esters was tested on the pine weevil, Hylobius abietis (L.). Of 105 compounds screened for activity, 9 phenylpropanoates, 3 cinnamates, and 4 benzoates were found to be highly active antifeedants. To understand the structure-activity relationships of these compounds, a multivariate analysis study was performed. A number of molecular and substituent descriptors were calculated and correlated to results from two-choice feeding tests with H. abietis. Three local models were developed that had good internal predictive ability. External test sets showed moderate predictivity. In general, low polarity, small size, and high lipophilicity were characteristics for compounds having good antifeedant activity.

Full text of DOI:10.1021/jf070014p

J. Agric. Food Chem. 2007, 55, 9365–9372 9365
Quantitative Structure–Activity Relationships of Pine
Weevil Antifeedants, a Multivariate Approach
,
†
‡
§
KERSTIN SUNNERHEIM,* ANNELI NORDQVIST, GÖRAN NORDLANDER,
#
⊥
⊥
ANNA-KARIN BORG-KARLSON, C. RICKARD UNELIUS, BJÖRN BOHMAN,
§
§
‡
HENRIK NORDENHEM, CLAES HELLQVIST, AND ANDERS KARLÉN
Department of Chemistry, Uppsala University, Box 599, SE-751 24 Uppsala, Sweden; Department of
Medicinal Chemistry, Uppsala University, Box 574, SE-751 23 Uppsala, Sweden; Department of
Ecology, Swedish University of Agricultural Sciences, Box 7044, SE-750 07 Uppsala, Sweden;
Department of Chemistry, Ecological Chemistry Group, Royal Institute of Technology,
SE-100 44 Stockholm, Sweden; and Department of Chemistry and Biomedical Sciences,
University of Kalmar, SE-391 82 Kalmar, Sweden
Antifeedant activity of mainly phenylpropanoic, cinnamic, and benzoic acids esters was tested on
the pine weevil, Hylobius abietis (L.). Of 105 compounds screened for activity, 9 phenylpropanoates,
3
cinnamates, and 4 benzoates were found to be highly active antifeedants. To understand the
structure–activity relationships of these compounds, a multivariate analysis study was performed. A
number of molecular and substituent descriptors were calculated and correlated to results from two-
choice feeding tests with H. abietis. Three local models were developed that had good internal
predictive ability. External test sets showed moderate predictivity. In general, low polarity, small size,
and high lipophilicity were characteristics for compounds having good antifeedant activity.
KEYWORDS: Conifer seedling; feeding deterrent; large pine weevil; Hylobius abietis; Curculionidae;
benzoates; phenylpropanoates; cinnamates; phenylpropenoates; phenylacrylates; multivariate analysis;
PLS; QSAR
INTRODUCTION
ment in Sweden (5, 6). Efforts are also made to reduce damage
levels by improvement of silvicultural measures (4, 7).
Weevils of the genus Hylobius are important pests of managed
conifer forests in Europe, Asia, and North America (1). Several
species injure healthy trees by larval feeding in the root collar
region (2). In other species, the larvae develop in already dead
or dying roots, and the economic damage is made by the adult
weevils feeding on the stem bark of conifer seedlings (1). Severe
damage is common in regions where replanting of harvested
conifer forests is the prevalent forestry practice. In large parts
of Europe, the pine weevil, H. abietis, is the most destructive
pest of conifer regenerations, causing almost total mortality
among seedlings if no countermeasures are taken (3, 4).
Prophylactic treatment of seedlings with relatively persistent
insecticides is therefore regularly carried out before planting
An alternative way of protecting conifer seedlings may be to
treat the stem with a chemical substance being a feeding
deterrent for H. abietis (8). Many substances with antifeedant
effect have been identified for H. abietis in laboratory
bioassays (8–14), and a few also for the closely related H. pales
in North America (15). Much of the recent work with H. abietis
has investigated categories of compounds that are present in
pine weevil feces or in the bark of trees. Antifeedant compounds
in H. abietis feces are of particular interest, because some of
them may function as semiochemicals for this species (13).
Feeding on root bark is to some extent avoided near oviposition
sites (16), and this is apparently related to the fact that the female
adds feces onto eggs laid in gnawed cavities in the bark (13).
In that study, fractions from extracts of H. abietis feces were
tested in feeding bioassays with H. abietis. From active fractions
a number of compounds apparently originating from lignin were
identified, including several benzoates. In a subsequent study
(14), 55 commercially available or synthesized benzoic acid
derivatives were bioassayed. The analogues with the highest
antifeedant activity were esters of 2,4- and 3,5-dimethoxyben-
zoic acid. Two compounds with antifeedant effect on H. abietis
have also been isolated from the bark of lodgepole pine (Pinus
(
5). With the aim to abandon this use of insecticides, new
methods for physical protection of seedlings are under develop-
*
Address correspondence to this author at the Department of Natural
Science, Mid Sweden University, SE-851 70 Sundsvall, Sweden (e-
mail kerstin.sunnerheim@miun.se; telephone +46 60 148709; fax +46
6
0 148802).
†
Department of Chemistry, Uppsala University.
‡
§
#
⊥
Department of Medicinal Chemistry, Uppsala University.
Swedish University of Agricultural Sciences.
Royal Institute of Technology.
University of Kalmar.
1
0.1021/jf070014p CCC: $37.00  2007 American Chemical Society
Published on Web 10/10/2007

9366 J. Agric. Food Chem., Vol. 55, No. 23, 2007
Sunnerheim et al.
contorta) (8). These were ethyl 2,3-dibromophenyl propanoate
and ethyl cinnamate, that is, both with nonsubstituted aromatic
rings.
from previous work by H. Erdtman and T. Norin at the Department of
Organic Chemistry, KTH, Stockholm.
The following esters were prepared by acid-catalyzed esterification
of their corresponding commercial benzoic, propanoic, or E-propenoic
acids: methyl 3,4,5-trihydroxybenzoate (B42), isopropyl 4-hydroxy-
benzoate (B47), methyl 2-methylpropanoate (P01), methyl 3-(3-
methoxyphenyl)propanoate (P02), methyl 3-methylpropanoate (P03),
methyl 3-(4-hydroxy-3-methoxyphenyl)propanoate (P04), methyl 3-(3-
bromo-4-methoxyphenyl)propanoate (P05), methyl 3-(3,4-dichlorophe-
nyl)propanoate (P06), methyl 3-(3,4,5-trimethoxyphenyl)propanoate
(P07), methyl 3-(4-methylphenyl)propanoate (P12), methyl 3-(4-
fluorophenyl)propanoate (P14), methyl 3-phenylpropanoate (P20),
methyl 3-(2-methoxyphenyl)propanoate (P21), methyl 3-(2,4-dimeth-
ylphenyl)propanoate (P23), methyl 3-(4-isopropylphenyl)propanoate
The aim of the present study was to find the structural criteria/
chemical features required for an active pine weevil antifeedant
and what structural changes can be made to improve the activity.
For that reason, a number of analogues to the earlier isolated
antifeedants (8, 13) were bought or synthesized and evaluated
for antifeedant activity. The results from the previous study of
the benzoates (14) were used as guidance in the selection of
test compounds. We have mainly focused on esters of 3-phen-
ylpropanoic, cinnamic, and benzoic acids. Complex relationships
between structure and the activity of, among others, insect
antifeedant and bird repellents have earlier successfully been
studied with multivariate methods (17–20). In this study
multivariate models based on a number of molecular and
substituent descriptors were performed, correlating the anti-
feedant activity to chemical structure.
(
P26), methyl 3-(4-chlorophenyl)propanoate (P27), methyl 3-(4-bro-
mophenyl)propanoate (P28), ethyl 3-phenylpropanoate (P29), methyl
-methyl-3-phenylpropanoate (P31), methyl 3-(2,3-dimethoxyphenyl)-
2
propenoate (C01), methyl 3-(2,4-dimethoxyphenyl)propenoate (C02),
methyl 3-(3,4-dimethoxyphenyl)propenoate (C03), methyl 3-(3,5-
dimethoxyphenyl)propenoate (C04), methyl 3-(4-methylphenyl)prope-
noate (C05), methyl 3-(4-isopropylphenyl)propenoate (C06), methyl
3-(4-trifluoromethylphenyl)propenoate (C07), methyl 3-(4-nitrophe-
nyl)propenoate (C09), ethyl phenylpropenoate (C10), propyl phenyl-
propenoate (C11), isopropyl phenylpropenoate (C12), butyl phenyl-
propenoate (C13), and 2-butyl phenylpropenoate (C14).
MATERIALS AND METHODS
1
13
General. H NMR (400 MHz) and C NMR (100.5 MHz) spectra
were recorded on Varian Unity 400, Bruker 400, or Bruker 250
apparatus by using the solvent signals (CDCl or CD OD) as internal
3 3
Typical Esterification Procedure: Methyl (E)-3-(4-Methylphenyl)-
propenoate (C05). (E)-3-(4-Methylphenyl)propenoic acid (1.2 mmol)
was dissolved in methanol (11 mL). Sulfuric acid (3 drops) was added.
The mixture was refluxed for 4 h. The reaction mixture was concen-
trated using a rotary evaporator, diluted with water, and extracted with
standards. For TLC, silica gel plates with fluorescent indicator were
used (Merck silica gel 60 F254, 0.25 mm).
Chemicals. The following compounds were purchased from Sigma-
Aldrich Co.: 2-hydroxy-4-methoxybenzoic acid (B01), 3,5-dihydroxy-
benzoic acid (B02), 4-hydroxy-3,5-dimethoxybenzoic acid (B04),
ethyl ether (30 mL). The ether phase was washed with Na
2 3
CO (aq) (5
4
-hydroxy-3-methoxybenzoic acid (B06), 4-hydroxybenzoic acid (B07),
methyl 3-hydroxy-4-methoxybenzoate (B14), methyl 4-hydroxybenzoate
B18), 3-hydroxy-4-methoxybenzoic acid (B23), 3,4-methylenedioxy-
mL), dried over MgSO , and evaporated to give white crystals (204
4
mg, 1.15 mmol): yield, 93%.
(
Methyl 3-(4-trifluoromethylphenyl)propanoate (P13), methyl 3-(2,3-
dimethoxyphenyl)propanoate (P22), and methyl 3-(3,5-dimethoxyphen-
yl)propanoate (P08) were prepared by Pd/C-catalyzed hydrogenation
of their corresponding propenoates according to the standard procedure
benzoic acid (B25), methyl 2-hydroxy-3-methoxybenzoate (B27),
methyl 2,4-dimethoxybenzoate (B32), methyl 2,4,6-trimethoxybenzoate
(
B33), methyl 2,6-dimethoxybenzoate (B35), methyl 4-hydroxy-3-
methoxybenzoate (B37), and methyl 3-(4-hydroxyphenyl)propanoate
P25).
The following compounds were purchased from Lancaster Synthesis
Co.: 3,5-dimethoxybenzoic acid (B03), methyl 2,4-dihydroxybenzoate
(
21).
(
Methyl 3-(4-aminophenyl)propanoate (P15) and methyl 3-(4-N,N-
dimethylaminophenyl)propanoate (P16) were obtained by Pd/C-
catalyzed hydrogenation of methyl 3-(4-nitrophenyl)propenoate and
methyl 3-(4-N,N-dimethylaminophenyl)propenoate, respectively, ac-
cording to the standard procedure (21).
(
(
B08), methyl 2-methoxybenzoate (B10), methyl 3-methoxybenzoate
B15), 2-hydroxy-3-methoxybenzoic acid (B21), 2-hydroxy-5-meth-
oxybenzoic acid (B22), 3,4-dimethoxybenzoic acid (B24), 2-hydroxy-
benzoic acid (B26), methyl 2-hydroxy-5-methoxybenzoate (B28),
methyl 3,5-dihydroxybenzoate (B40), methyl 3,5-dimethoxybenzoate
Isopropyl 3-(4-methoxyphenyl)propanoate (P17) was obtained by
reacting isopropyl 3-(4-hydroxyphenyl)propanoate with methyl iodide
and potassium carbonate in acetone according to the standard procedure.
(
B45), methyl 4-methoxybenzoate (B48), and 3-(2-methoxyphenyl)-
propanoic acid (P18).
-(2-Methylphenyl)propanoic acid (P19) was purchased from Matrix
Methyl 3-(4-butyloxyphenyl)propanoate (P10) was obtained by
reacting methyl 3-(4-hydroxyphenyl)propanoate with potassium hy-
droxide and butyl iodide according to the standard procedure.
3
via Chemtronica, Stockholm, Sweden.
Methyl 3-(4-acetylphenyl)propanoate (P11) was obtained by reflux-
ing methyl 3-(4-hydroxyphenyl)propanoate with an excess of acetic
anhydride according to the standard procedure.
For syntheses of methyl 2,4-dihydroxy-3,6-dimethylbenzoate (B09),
methyl 2,3-dimethoxybenzoate (B11), methyl 2,3,4-trimethoxybenzoate
(
(
B12), methyl 3,5-dinitrobenzoate (B16), methyl 3,5-dibromobenzoate
B17), isopropyl 2,4-dimethoxybenzoate (B19), 2,2,2-trifluoroethyl 3,5-
Methyl 3-(3,4-dimethoxyphenyl)propanoate (P24) was obtained by
reacting methyl 3-(3,4-dihydroxyphenyl)propanoate with sodium hy-
dride and methyl iodide in THF according to the standard procedure.
dimethoxybenzoate (B20), methyl 2,5-dimethoxybenzoate (B34), methyl
,4-dihydroxybenzoate (B36), methyl 3-chloro-4-methoxybenzoate
B38), methyl 3,4-methylenedioxybenzoate (B39), methyl 4-hydroxy-
,5-dimethoxybenzoate (B43), methyl 3,4,5-trimethoxybenzoate (B44),
methyl 3,5-dimethylbenzoate (B46), S-ethyl 3,5-dimethoxybenzothioate
B49), N-ethyl 3,5-dimethoxybenzamide (B50), methyl 4-n-octylben-
3
(
3
Methyl 3-(4-N,N-dimethylaminophenyl)propenoate (C08) was pre-
pared from 4-(N,N-dimethylaminophenyl)propenoic acid by reaction
with DCC and DMAP and methanol in CH Cl (11).
2
2
(
Racemic mixtures (2R,3S and 2S,3R) of the esters methyl 2,3-
dibromo-3-phenylpropanoate (P34) and ethyl 2,3-dibromo-3-phenyl-
propanoate (P35) were obtained by bromination of their corresponding
E-phenylpropenoate, methyl phenylpropenoate and ethyl phenylpro-
penoate, respectively, according to the standard procedure (21). P35
was also isolated from bark of Pinus contorta; see ref 8.
zoate (B52), dodecyl 3,4-dimethoxybenzoate (B53), 3-(E)-hexenyl 3,5-
dimethoxybenzoate (B54), 2-methoxy-4-(2-propenyl)phenyl 3,5-dimethoxy-
benzoate
(B55),
and
3-(3,4-dimethoxyphenyl)propyl
3,5-
dimethoxybenzoate (B56) see (14).
For syntheses of methyl 5-hydroxy-2-methoxybenzoate (B13), methyl
-hydroxy-6-methoxybenzoate (B29), methyl 3-hydroxy-2-methoxy-
2
Ethyl 3-phenyl-3-hydroxypropanoate (P32) and ethyl 3-(2-bro-
mophenyl)-3-hydroxypropanoate (P33) were obtained by adding ethyl
bromoacetate to benzaldehyde and 2-bromobenzaldehyde, respectively,
zinc dust, and copper acetate in THF according to the standard
Reformatsky procedure (21).
benzoate (B30), methyl 4-hydroxy-2-methoxybenzoate (B31), and
methyl 3-hydroxy-5-methoxybenzoate (B41), see ref (11).
3
,4,5-Triacetoxybenzoic acid (B05), 3,4,5-trimethoxybenzamide
(B51), and methyl 3-(4-methoxyphenyl)propanoate (P09) were obtained

Pine Weevil Antifeedants
J. Agric. Food Chem., Vol. 55, No. 23, 2007 9367
(
S)-Methyl 2-amino-3-phenylpropanoate (P30) was prepared by
Table 1. Benzoic Acid Derivatives (When R Is Hydrogen, It Is Omitted
from the Table)
reacting L-phenylalanine with thionyl chloride followed by methanol
according to the standard procedure (21).
All reactions were monitored by TLC. The spectroscopic data of
the products were analyzed and compared with literature data.
Collection and Maintenance of Weevils. Both sexes of H. abietis
were collected during spring migration at a sawmill in southern Sweden,
where they landed in large numbers as a response to a massive emission
of attractive conifer volatiles. After collection, the weevils were stored
in darkness at 10 °C and provided with fresh Scots pine, Pinus sylVestris
L., branches or stems with tender bark as food. These storage conditions
interrupted the reproductive development, so that females did not begin
to oviposit until about a week after they had been transferred to the
experimental conditions, that is, a light regimen of L18 h /D6 h at 22
subst
no.
AFIa AFIa
obsd pred
R1
R2
R3
R4
R5
R6
training set
B01
B02
B03
B04
B05
B06
B07
B08
B09
B10
B11
B12
B13
B14
B15
B16
B17
B18
B19
B20
OH
OH
OH
OH
OH
OH
OH
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OiPr
OCH2CF3
OH
OMe
24
41
–4
8
31
11
29
16
8
OH
OH
OMe
OMe
OAc
OMe
OMe
OMe
OAc
OH
OAc
OH
OH
OH
OH
°
C. This transfer was made about 10 days before the insects were used
14
48
31
46
31
80
73
54
–3
65
89
34
50
34
96
86
in the following bioassay.
26
27
42
44
64
62
60
43
53
61
27
61
50
98
82
Feeding Bioassay. The compounds were tested for their antifeedant
effect on H. abietis by means of a two-choice laboratory bioassay (14),
used in several previous studies (8, 11, 13, 14). For each test, 40 pine
weevils (20 females + 20 males) were used. They were placed in
separate Petri dishes provided with a Scots pine twig prepared with
delimited treatment and control areas. These twigs were enveloped in
aluminum foil, and two holes with a diameter of 5 mm and separated
by 25 mm were punched in the foil with metal rings. After removal of
the aluminum foil inside the rings, one of the two surfaces was treated
with 100 µL of a 50 mM methanol or methyl acetate solution of the
compound tested, and the other surface was treated with the same
amount of solvent alone. The following day, after the solvent had
evaporated, the metal rings were removed and the test started. The
proportion of available bark area that had been eaten by the test weevil
on the treatment and control area of each test twig was recorded after
OH
OH
OMe
OMe
OMe
OMe
Me
Me
OMe
OMe
OMe
OMe
OH
OH
OMe
NO2
Br
NO2
Br
OH
OMe
OMe
OMe
OMe
OMe
OMe
test set
B21
B22
B23
B24
B25
B26
B27
B28
B29
B30
B31
B32
B33
B34
B35
B36
B37
B38
B39
B40
B41
B42
B43
B44
B45
B46
B47
B48
B49
B50
B51
B52
B53
B54
B55
B56
OH
OH
OH
OH
OH
OH
22
17
51
7
14
21
95
74
33
24
30
37
35
56
54
46
56
56
49
60
59
55
63
42
47
61
59
34
45
21
39
50
52
63
86
61
71
59
16
98
24 h. There was generally no significant difference in response between
the sexes, and the data presented were therefore pooled.
The antifeedant effect measured for each compound is expressed
by means of an antifeedant index (AFIa) based on feeding area (13, 14)
OH
OMe
OMe
OMe
OCH2O
OH
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OiPr
OMe
SEt
OH
OH
OH
OH
OMe
OMe
OMe
OMe
OMe
OMe
AFIa ) 100 × (C - T) ⁄ (C + T)
OMe
OH
OMe
OMe
where C represents the mean area of control surfaces consumed and T
the mean consumed area of treated surfaces. Positive values (up to a
maximum of 100) reflect an antifeedant effect, whereas negative values
OMe 74
75
OH
OMe
OMe
35
99
(
down to a minimum of -100) indicate a stimulant effect on
feeding.
Multivariate Analysis. All molecular descriptors were calculated
with Sybyl (22) and are described in Table 4. For the substituents on
the aromatic ring (R –R ) σ and π values were taken from the literature
23). Indicator variables were used to account for the substitution
pattern. Score values (t1_subst, t2_subst, and t3_subst) were also calculated
for the different aromatic substituents (R –R ) using principal compo-
nent analysis (PCA) and used as descriptors. The PCA was based on
, σ , π, molecular refractivity, and the five Verloop steric parameters
OMe 55
89
OMe 51
–7
53
OH
OMe
Cl
OH
OH
OMe
2
6
(
36
57
23
54
21
10
32
95
61
63
54
74
86
–2
35
OCH2O
2
6
OH
OH
OH
OMe
OMe
OMe
Me
OH
OMe
OH
OMe
OMe
OMe
Me
σ
m
p
OH
OH
OMe
2
(
23). This analysis gave an R (2) ) 0.75 and a Q ) 0.35 (N ) 3).
A similar characterization was made of the ester substituent, that is,
the whole substituent in position 1 on the ring [–COR
and –CHdCHCOR ]. In the case of missing σ and π values they were
calculated using the ACD/sigma program (24). Score values for the
p
ester substituent (t1_ester, t2_ester, and t3_ester) were obtained using σ , σ ,
1
, –(CH
2
)
2
COR
1
,
1
OH
OMe
m
2
OMe
OMe
OMe
OMe
OMe
OMe
π, and molecular refractivity, and a PCA was performed (R ) 0.96,
2
NHEt
NH2
OMe
O-C12H25
a
Q ) 0.85, N ) 3).
OMe
n-C8H17
OMe
Simca-P+ (25) with autoscaling was used for the multivariate
analysis studies. A PCA was first performed on the whole data set of
OMe
OMe
OMe
OMe
23 251
62 131
17 207
41 221
1
05 compounds (Tables 1-3) using 53 descriptors (Table 4). The data
set was later divided into three groups corresponding to benzoic acid,
-phenylpropanoic acid, and cinnamic acid derivatives. These three
OMe
OMe
OMe
b
c
3
groups were used to derive local PLS models. In the class of benzoic
acid derivatives, five compounds were removed (B52–B56), because
they were structurally very different compared to the others.
a
b
(E)-3-Hexenyl-. 2-Methoxy-(4-prop-2-enyl)phenyl-. 3-(3,4-Dimethoxyphen-
c
yl)propyl-.
The benzoic acid and 3-phenylpropanoic acid derivatives were
divided into training and test sets on the basis of structural diversity
using a full factorial design in three variables. In this design each
variable is explored at two levels, low and high. The variables were
derived from a PCA on the benzoic acid and the 3-phenylpropanoic
acid derivatives separately (Table 5) using the 53 descriptors (Table
4). Two substances were selected at each level, and four centerpoints
were used. For the 3-phenylpropanoic acid derivatives the low–high–
high variable level contained only one compound, which was included

9368 J. Agric. Food Chem., Vol. 55, No. 23, 2007
Sunnerheim et al.
Table 2. Propanoic Acid Derivatives (When R Is Hydrogen, It Is Omitted
from the Table)
Table 3. Cinnamic Acid Derivatives (When R Is Hydrogen, It Is Omitted
from the Table)
subst
no.
AFIa AFIa
R1
R2
R3
R4
R5
R6 obsd pred
C01
C02
C03
C04
C05
C06
C07
C08
C09
C10
C11
C12
C13
C14
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OMe
OEt
OMe OMe
OMe
99
96
36
72
76
92
62
49
32
83
85
96
38
48
103
93
33
46
85
82
65
58
40
83
72
79
59
67
OMe
OMe
OMe
OMe
OMe
Me
iPr
CF3
NMe2
NO2
OPr
OiPr
OBu
O-2-Bu
uents had low activities (AFIa < 50). In another study of insect
feeding deterrents (26), it was also found that nonpolar
substituents on low molecular weight aromatic compounds
increased the activity. It has also been shown that other lipophilic
compounds including monoterpenoids (10, 15, 27, 28), nonanoic
acid (12), allylanisole (10), dihydropinidine (29), and a number
of substituted cinnamic aldehydes, esters, and benzaldehydes
(
17) are potent pine weevil antifeedants. Among the more
lipophilic substances in our data set, the activity varied
substantially. For example, compound P07, methyl 3-(3,4,5-
trimethoxyphenyl)propanoate, had low activity (AFIa ) 26),
whereas pine weevils were totally deterred by twigs treated with
P23, methyl 3-(2,4-dimethylphenyl)propanoate (AFIa ) 100).
Similar complex structure–activity relationships were also
observed by Ericsson (17).
Besides the apparent observations mentioned above, the
structure–activity relationships were not easy to interpret. Thus,
a multivariate analysis of the data was performed. In this study
each compound was described with 52 variables describing
general molecular properties such as lipophilicity, polarity,
electronic properties, size, and substitution pattern (Table 4),
and these were correlated to antifeedant activity (AFIa) (Tables
1-3).
a
b
Enantiomerically pure, synthesized from L-phenylalanine. R/S mixtures. R,S/
c
d
S,R mixture. Isolated from Pinus contorta.
A PCA was first performed in Simca using the whole data
set of 105 molecules and 53 molecular descriptors (including
AFIa). In this analysis three components were extracted that
explained 49% of the variance in the data set. Because the score
plot (Figure 1) showed a clear grouping in the first component
between the benzoic acid derivatives, on the one hand, and the
phenylpropanoic/cinnamic acid derivatives, on the other, we
created local models based on the three structural classes
separately. Efforts to derive PLS models including all com-
pounds were unsuccessful.
in the training set. The five racemic compounds (P31–P35) were placed
in the test set. Furthermore, no compounds from the high–low–high
level were chosen, because it contained only racemic compounds.
In the derivation of the final PLS models, variables with little
importance, as judged from the VIP plot and the coefficient plot, were
excluded.
RESULTS AND DISCUSSION
Antifeedant acitivity was measured for 35 3-phenylpropanoic,
1
4 cinnamic, and 56 benzoic acid derivatives in the bioassay
After variable reduction, final PLS models were derived that
included 11, 8, and 11 descriptors for the benzoic acid, the
3-phenylpropanoic acid, and the cinnamic acid derivatives,
respectively (Figure 2A,C,E). The observed versus predicted
values (Tables 13) for the three different models are shown in
with H. abietis. Nine phenylpropanoates, three cinnamates, and
four benzoates were very active antifeedants, having an AFIa
of 95–100 (Tables 1-4). None of the tested phenylpropanoids
stimulated feeding (i.e., negative AFIa values), as was the case
for some of the benzoic acid derivatives (Table 1) (14). All
acids tested (12 benzoic and 2 phenylpropanoic acids) had low
activities, which is in agreement with previous findings by
Ericsson (17). Also, esters with hydroxy or other polar substit-
2
Figure 2B,D,F. The rmsEP and R pred values for the test sets
are shown in Table 5. The external predictivity was moderate
2
with an R pred of 0.38 and 0.34 for the benzoic acid derivatives
and 3-phenylpropanoic acid derivatives, respectively. PLS

Pine Weevil Antifeedants
J. Agric. Food Chem., Vol. 55, No. 23, 2007 9369
Figure 1. Score plot of the first and second principal components for the PCA of all compounds. Local PLS models were based on the separation
observed between the benzoic acid derivatives, on the one hand, and the 3-phenylpropanoic/cinnamic acid derivatives, on the other, captured in the first
principal component. The large lipophilic benzoic acid derivatives (B52–B56) correspond to the five solid circles to the right.
Table 4. Description of the Variables Used in the Modeling
Substituent Descriptors for R2–R6
sigma meta and sigma para (23) for the substituents R2–R6, which describe
σ2,3,4,5, or 6
electronic properties of the substituent in the meta or para position,
respectively
pi value (23) for the substituents R2–R6, which describe lipophilic properties
π2,3,4,5, or 6
of the substituent
indicator variable for the substituents R2–R6 (0 ) hydrogen, 1 ) any other
substituent)
I2,3,4,5, or 6
L, B1, B2, B3, B4
t1_subst, t2_subst, t3_subst
Verloop’s steric parameters for the substituents R2–R6 (23)
principal component values from a PCA of the different substituents R2–R6;
t1 describes size and lipophilicity, t2 describes electronic properties, and t3
describes electronic properties together with size
Substituent Descriptors for the Ester Substituent (–COR1, -(CH2)2COR1, and -CHdCHCOR1)
MR1
molecular refractivity for the ester substituent calculated with the ACD/sigma
program (24) describes the steric or “bulk” properties of this substituent
molecular weight for the ester substituent calculated with the ACD/sigma
MW1
program (24)
pi value for the ester substituent calculated with the ACD/sigma program
π1
(24), which describes lipophilic properties of this substituent
inductive sigma value calculated with the ACD/sigma program (24), which
σ(Ind)
describes electronic properties of the ester substituent
resonance sigma value calculated with the ACD/sigma program (24), which
σ(Res)
describes electronic properties of the ester substituent
principal component values from a PCA of the different ester substituents; t1
t1_ester, t2_ester, t3_ester
describes electronic properties, t2 describes lipophilic properties, and t3
describes the difference between s(Res) and s(Ind)
Molecular Descriptors, Calculated with Sybyl Version 6.9 (22)
calculated log partition coefficient octanol–water
CLOGP
CMR
2
2
calculated molecular refractivity: MR ) (MW/d) × [(η – 1)/(η + 2)], where
MW is molecular weight, d is density, and η is refractive index; CMR
describes the steric or “bulk” properties of the whole molecule
number of rings
number of atoms
number of bonds
number of rotatable bonds
number of hydrogen bond acceptors
number of hydrogen bond donors
sum of the number of hydrogen bond donors and acceptors: HB_ALL )
RingCount
AtomCount
BondCount
RotBonds
HB_ACC
HB_DON
HB_ALL
HB_ACC + HB_DON
molecular weight
molecular surface area
molecular volume
polar volume, i.e., the volume of all nitrogen, oxygen, and sulfur atoms as
MWtotal
AREA
VOLUME
PV
well as hydrogens covalently bonded to these atoms
polar surface area, i.e., the surface area of all nitrogen, oxygen, and sulfur
PSA
atoms as well as hydrogens covalently bonded to these atoms
Descriptor for Biological Activity
antifeedant index
AFIa

9370 J. Agric. Food Chem., Vol. 55, No. 23, 2007
Sunnerheim et al.
Figure 2. (A) Coefficient plot for PLS model for the benzoic acid derivatives. Positive columns mean that the variables are positively correlated to
antifeedant activity. Negative columns are variables in which high values diminish the antifeedant activity. (B) Observed versus predicted plot for the
training set and external test set for the benzoic acid derivatives. Similar plots are shown for the 3-phenylpropanoic acid derivatives (C, D) and the
cinnamates (E, F).
2
models with scrambled y values produced Q values that were
mostly negative or very low.
In general, for all three PLS models, an overall low polarity
and size and a high lipophilicity are characteristic for a
compound having good antifeedant activity. In the three PLS
models this trend is captured by different combinations of
descriptors such as polar volume, polar surface area, hydrogen
bond donors and acceptors, molecular weight, and calculated
octanol–water partition coefficient (CLOGP). For example,
MWtotal is negatively correlated to AFIa (Figure 2C,E). The
molecular descriptors PV and PSA, which partly describe size,
In the PCA plot in Figure 1, the five benzoic acid derivatives
B52–B56) appear as outliers in the first principal component
and were not considered in the modeling. These compounds
contain long alkyl chains (Table 1) and are therefore larger and
also have a higher lipophilicity as compared to the other benzoic
acid derivatives. All other compounds were included in the
modeling.
(

Pine Weevil Antifeedants
J. Agric. Food Chem., Vol. 55, No. 23, 2007 9371
Table 5. Results from Multivariate Modeling
In the series of cinnamates, consisting of only 14 compounds,
we could not obtain any models when the data set was divided
into training and test sets. The PLS models were therefore
derived using all of the cinnamates. The obtained PLS model
showed that nonsubstituted cinnamates (C10–C14), with few
rotatable bonds, had a higher antifeedant activity. The two
compounds with a methoxy substituent (C01 and C02) in the
ortho position are among the most active, which is also reflected
by the coefficients for all six R2 descriptors.
Due to the nature of the PLS model, predicted AFIa values
>100 are possible but are, in reality, impossible. Predicted AFIa
values slightly larger than 100 as in the case for P23, P26, and
C01 should be interpreted as very promising compounds seen
in relation to the root-mean-square error of prediction (rmsEP)
for the model (Table 5) as for all quantitative predictions. Very
unrealistic predictions as for B53–B56 can be identified, as these
compounds are already considered to be very different from
the training set of the model.
benzoic acid
derivatives
propanoic acid
cinnamic acid
derivatives
derivatives
PCA
2
a
b
R
0.52
0.23
3
0.50
0.04
3
0.72
0.36
3
2
Q
c
N
PLS
2
a
b
R
Q
N
0.61
0.55
1
20
31
0.63
0.56
1
17
13 (18)
0.34 (0.28)
28 (27)
0.70
0.53
2
2
c
d
ntraining set
14
d
ntest set
2
e
g
R
RMSEP
pred
0.38
24
f
g
a
b
Explained variation or goodness of fit. Predicted variation or goodness of
c
d
prediction calculated from the training set. Number of components. Number of
e
2
compounds in the training set and test set. Rpred ) 1 – PRESS/SD, where
PRESS stands for predictive sum of squares and SD for sum of squared differences.
g
Root-mean-square error of prediction for the test set. rmsEP and R pred are
f
2
The potential of using antifeedants to protect forest regenera-
tion against pine damage has previously been demonstrated in
field tests with methyl 3,5-dimethoxybenzoate (B45) (30) and
ethyl 2,3-dibromo-3-phenylpropanoate (P35).
given for the test set without the racemic compounds and in parentheses for the
whole test set.
are also negatively correlated to AFIa. These two descriptors
also describing polarity together with the hydrogen-bonding
descriptors (HB_DON and HB_ACC) capture the trend that a
high-polarity diminishes AFIa (Figure 2A,C,E) and a high
CLOGP value increases the AFIa (Figure 2C). The positive
effect of a high lipophilicity is also captured by the t2ester, π1, π
In this study, several highly active antifeedants have been
identified among the phenylpropanoates and cinnamates. The
multivariate models have given a better understanding not only
of which structural properties are important for high antifeedant
activity but also of how to optimize the activity within the three
compound classes.
4
, and π 5 descriptors in the different substituent positions (R1,
R4, and R5) A para substituent with a high π value (high
ACKNOWLEDGMENT
lipophilicity) gives a high antifeedant activity in all three models
We acknowledge Magnus Besev, Abtin Daghigi, Konstantin
Dubrovinski, Tonja Freisolt, Henning Henschel, Sacha Legrand,
Meral Sari, and Veronica Siljehav for the syntheses of test
compounds.
(
Figure 2A,C,E). In the benzoic acid derivatives this can be
exemplified by comparing compounds that differ only in the
para position, for example, in order of increasing activity B43
(
R4 ) OH), B44 (R4 ) OMe), and B45 (R4 ) H) with AFIa
values of 10, 32, and 95, respectively. In the propanoic acid
derivatives the para substituent is the most varied in the whole
data set. Seven of the 12 different para substituents tested gave
an antifeedant activity of >83, and the 3 compounds (P11, P15,
and P25) with the lowest AFIa also have very low π4 values
LITERATURE CITED
(
1) Day, K. R.; Nordlander, G.; Kenis, M.; Halldórsson, G. General
biology and life cycles of bark weevils. In Bark and Wood Boring
Insects in LiVing Trees in Europe; Lieutier, F., Day, K. R., Battisti,
A., Grégoire, J.-C., Evans, H. F., Eds.; Kluwer Academic
Publishers: Dordrecht, The Netherlands, 2004; pp 331–349.
2) Cerezke, H. F. Warren rootcollar weevil, Hylobius warreni Wood
(
–1.23 to –0.64).
For the benzoic acid derivatives the esters are generally better
(
antifeedants than the carboxylic acids (Table 1). This general
trend has been observed previously (17). This observation is
also supported in this study by the model descriptors MW1, π1
and t2_ester in Figure 2A. However, there appears to be an
optimum concerning the size or lipophilicity of the alcohol
moiety, because compounds (B53–B56) with very large alkyl
groups such as alcohol moieties do not exert high antifeedant
activity. In the analysis of the benzoic acid derivatives an R5
substituent corresponding to a methoxy group or a hydrogen
atom seems to be beneficial for good antifeedant activity.
(
Coleoptera: Curculionidae), in Canada: ecology, behaviour,
damage relationships, and management. Can. Entomol. 1994, 126,
383–1442.
1
(
3) Örlander, G.; Nilsson, U. Effect of reforestation methods on pine
weevil (Hylobius abietis) damage and seedling survival. Scand.
J. For. Res 1999, 14, 341–354.
(4) Petersson, M.; Örlander, G. Effectiveness of combinations of
shelterwood, scarification, and feeding barriers to reduce pine
weevil damage. Can. J. For. Res. 2003, 33, 64–73.
5) Långström, B.; Day, K. R. Damage, control and management of
weevil pests, especially Hylobius abietis. In Bark and Wood
Boring Insects in LiVing Trees in Europe; Lieutier, F., Day, K. R.,
Battisti, A., Grégoire, J.-C., Evans, H. F., Eds.; Kluwer Academic
Publishers: Dordrecht, The Netherlands, 2004; pp 415–444.
(
All five racemic propanoic acid derivatives (P31–P35) in the
test set were overpredicted (Figure 2D). Thus, when the five
2
racemic compounds in the test set were excluded, R pred was
improved (Table 5). In a previous study (8) ethyl 2,3-
dibromopropanoate (P35) was isolated from P. contorta. The
isolated compound had an AFIa of 75, whereas the synthetic
RS/SR racemate obtained from the E-cinnamate had AFIa of
(6) Petersson, M.; Örlander, G.; Nilsson, U. Feeding barriers to reduce
damage by pine weevil (Hylobius abietis). Scand. J. For. Res.
2
004, 19, 48–59.
(
7) Wallertz, K.; Örlander, G.; Luoranen, J. Damage by pine weevil
Hylobius abietis to conifer seedlings after shelterwood removal.
Scand. J. For. Res. 2005, 20, 412–420.
8) Bratt, K.; Sunnerheim, K.; Nordenhem, H.; Nordlander, G.;
Långström, B. Pine weevil (Hylobius abietis) antifeedants from
lodgepole pine (Pinus contorta). J. Chem. Ecol. 2001, 27, 2253–
2262.
4
5. Unfortunately, the optical rotation was not determined for
the isolated sample, but one explanation for these results might
be that all four stereoisomers are not equally active and that an
excess of the more active stereoisomer(s) is biosynthesized in
the pines.
(

9372 J. Agric. Food Chem., Vol. 55, No. 23, 2007
Sunnerheim et al.
(
9) Lindgren, B. S.; Nordlander, G.; Birgersson, G. Feeding deterrence
of verbenone to the pine weevil, Hylobius abietis (L.) (Col.,
Curculionidae). J. Appl. Entomol. 1996, 120, 397–403.
structure–activity relationships (QSAR) of cinnamic acid bird
repellents. J. Chem. Ecol. 1999, 25, 2825–2845.
(21) Vogel. Textbook of Practical Organic Chemistry, 5th ed.; Prentice
Hall: Engelwood Cliffs, NJ, 1989.
(
10) Klepzig, K. D.; Schlyter, F. Laboratory evaluation of plant-derived
antifeedants against the pine weevil Hylobius abietis (Coleoptera:
Curculionidae). J. Econ. Entomol. 1999, 92, 644–650.
11) Legrand, S.; Nordlander, G.; Nordenhem, H.; Borg-Karlson, A.-
K.; Unelius, C. R. Hydroxy-methoxybenzoic methyl esters:
synthesis and antifeedant activity on the pine weevil, Hylobius
abietis. Z. Naturforsch. 2004, 59B, 829–835.
12) Månsson, P. E.; Eriksson, C.; Sjödin, K. Antifeedants against
Hylobius abietis pine weevils: an active compound in extract of
bark of Tilia cordata linden. J. Chem. Ecol. 2005, 31, 989–1001.
13) Borg-Karlson, A. K.; Nordlander, G.; Mudalige, A.; Nordenhem,
H.; Unelius, C. R. Antifeedants in the feces of the pine weevil
Hylobius abietis: identification and biological activity. J. Chem.
Ecol. 2006, 32, 943–957.
14) Unelius, C. R.; Nordlander, G.; Nordenhem, H.; Hellqvist, C.;
Legrand, S.; Borg-Karlson, A. K. Structure–activity relationships
of benzoic acid derivatives as antifeedants for the pine weevil,
Hylobius abietis. J. Chem. Ecol. 2006, 32, 2191–2203.
15) Salom, S. M.; Carlson, J. A.; Ang, B. N.; Grossman, D. M.; Day,
E. R. Laboratory evaluation of biologically based compounds as
antifeedants for the pales weevil, Hylobius pales (Herbst) (Co-
leoptera: Curculionidae). J. Entomol. Sci. 1994, 29, 407–419.
16) Bylund, H.; Nordlander, G.; Nordenhem, H. Feeding and ovipo-
sition rates in the pine weevil Hylobius abietis (Coleoptera:
Curculionidae). Bull. Entomol. Res. 2004, 94, 307–317.
17) Eriksson, C. Isolation, synthesis and structure–activity relationships
of antifeedants against the pine weevil, Hylobius abietis. Doctoral
thesis in organic chemistry, KTH, Stockholm, Sweden, 2006.
18) Cao, S.; Wei, N.; Zhao, C.; Li, L.; Huang, Q.; Qian, X. Syntheses,
antifeedant activity, and QSAR analysis of new oxa(thia)diazoyl
(22) Sybyl 6.9, Tripos Inc., St. Louis, MO.
(23) Seydel, J. K.; Schaper, K.-J. Chemische Struktur und Biologische
AktiVität Von Wirkenstoffen; Verlag Chemie: Weinheim, Germany,
1979; pp 266–288.
(
(
24) ACD/ChemSketch& ACD/Sigma, version 4.55, Advanced Chem-
istry Development, Inc., Toronto, ON, Canada, 2000; www.
acdlabs.com.
(
(
25) Simca-P+, version 10.0, Umetrics, Umeå, Sweden, 2002;
www.umetrics.se.
(
26) Harmatha, J.; Nawrot, J. Insect feeding deterrent activity of lignans
and related phenylpropanoids with a methylenedioxyphenyl (pip-
eronyl) structure moiety. Entomol. Exp. Appl. 2002, 104, 51–60.
27) Salom, S. M.; Gray, J. A.; Alford, A. R.; Muleski, M.; Fettig,
C. J.; Woods, S. A. Evaluation of natural products as antifeedants
for the pales weevil (Coleoptera: Curculionidae) and as fungitoxins
for Leptographium procerum. J. Entomol. Sci. 1996, 31, 453–
(
(
(
(
4
65.
(
28) Schlyter, F.; Smitt, O.; Sjödin, K.; Högberg, H.-E.; Lö; fqvist, J.
Carvone and less volatile analogues as repellent and deterrent
antifeedants against the pine weevil, Hylobius abietis. J. Appl.
Entomol 2004, 128, 610–619.
(
29) Eriksson, C.; Sjödin, K.; Schlyter, F.; Högberg, H.-E. Synthesis
of (+)- and (–)-dihydropinidine by diastereoselective dimethylzinc
promoted allylation of 2-methyltetrahydropyridine-N-oxide with
an allylboronic ester. Tetrahedron Asymmetry 2006, 17, 1074–
(
(
(
(
1
080.
(
30) Nordlander, G.; Nordenhem, H.; Borg-Karlson, A.-K.; Unelius,
R. Use for conifer sapling protection. Swedish and PCT Patent
WO 0056152 A1, 2000.
3
(2H)-pyridazinones. J. Agric. Food Chem. 2005, 53, 3120–3125.
Received for review January 2, 2007. Revised manuscript received June
6, 2007. Accepted June 26, 2007. This study was supported by The
Swedish Research Council for Environment, Agricultural Sciences and
Spatial Planning, The Swedish Hylobius Research Program, and The
University of Kalmar.
19) Morimoto, M.; Tanimoto, K.; Nakano, S.; Ozaki, T.; Nakano, A.;
Komai, K. Insect antifeedant activity of flavones and chromones
against Spodoptera litura. J. Agric. Food Chem. 2003, 51, 389–
2
3
93.
(20) Watkins, R. W.; Lumley, J. A.; Gill, E. L.; Bishop, J. D.; Langton,
S. D.; MacNicoll, A. D.; Price, N. R.; Drew, M. G. B. Quantitative
JF070014P