Comprehensive study of sansalvamide A derivatives and their structure-activity relationships against drug-resistant colon cancer cell lines

Article abstract of DOI:10.1021/jm070731a

We report an extensive structure-activity relationship (SAR) of 62 compounds active against two drug-resistant colon cancer cell lines. Our comprehensive evaluation of two generations of compounds utilizes SAR, NMR, and molecular modeling to evaluate the

Full text of DOI:10.1021/jm070731a

530
J. Med. Chem. 2008, 51, 530–544
Comprehensive Study of Sansalvamide A Derivatives and their Structure–Activity Relationships
against Drug-Resistant Colon Cancer Cell Lines
Katerina Otrubova, Gerald Lushington,† David Vander Velde,† Kathleen L. McGuire, and Shelli R. McAlpine*
Department of Medicinal Chemistry, 1251 Wescoe Hall Rd, The UniVersity of Kansas, Lawrence, Kansas, 66045-7582
ReceiVed June 21, 2007
We report an extensive structure–activity relationship (SAR) of 62 compounds active against two drug-
resistant colon cancer cell lines. Our comprehensive evaluation of two generations of compounds utilizes
SAR, NMR, and molecular modeling to evaluate the key 3D features of potent compounds. Of the seven
most potent compounds reported here, five are second-generation, emphasizing our ability to incorporate
potent features found in the first generation and utilize their structures to design potency into the second
generation. These analogs share no structural homology to current colon cancer drugs, are cytotoxic at
levels on par with existing drugs treating other cancers, and demonstrate selectivity for drug-resistant colon
cancer cell lines over noncancerous cell lines. Thus, we have established sansalvamide A as an excellent
lead for treating multiple drug-resistant colon cancers.
Introduction
Natural products often provide lead structures for new drugs.
These novel structures are critical for development of original
therapeutic small molecules that target new biological pathways.
Sansalvamide A (San A) is one such natural product (Figure
1). San A, which was isolated from a marine fungus (Fusarium
ssp.), exhibits antitumor activity.1–3 To date the synthesis of 86
analogs have been reported, 75 by us4,5 and 11 by Silverman et
al.6 The natural product is a depsipeptide and, as such, is prone
to deactivation by ring opening enzymes. To avoid this, most
of the 86 analogs synthesized, including all 75 of those reported
by our laboratory, were synthesized as the San A peptide
Figure 1. Retrosynthetic approach.
derivatives (Figure 1, amino acid 4), which are from here on
referred to as “San A-amide” derivatives. Cytotoxicity of San
of the seven most potent compounds reported here, five are
A-amide derivatives against pancreatic,7,6,8 colon,3,4,9,10 breast,
second-generation compounds, emphasizing our ability to
prostate, and melanoma cancers6 clearly indicate San A-amide’s
incorporate potent features found in the first generation to
excellent potential as a new therapeutic agent for the treatment
enhance the potency of the second generation of molecules.
of various cancers and support further exploration of this class
Our work reported below shows that there are two factors
of compounds. All 11 of the San A-amide derivates prepared
important for potency: a single N-methyl and a single D-amino
by Silverman and co-workers contain L-amino acids (L-aas),6,8,11,12
acid (D-aa). This work is validated by several current examples
and these demonstrate reasonable potency against one colon
in the recent literature where cyclic peptides and, specifically,
cancer cell line HCT-116. They attribute potency to the
pentapeptides with a single D-aa lock the macrocycle into a
single conformation.14,15 Further, it is well established that these
placement of multiple N-methyl moieties on the macrocycle.
However, our extensive studies show that the SAR in HCT-
cyclic pentapeptides mimic beta and gamma turns and serve as
116 is not due to the placement of multiple N-methyl moieties.
templates for appropriately positioning suitable binding motifs
for proteins.16,17 Given that we have discovered seven small
For the first time, we report here an extensive structure–ac-
tivity relationship (SAR) of 62 compounds against two drug-
molecules that target drug-resistant colon cancer cell lines at
resistant colon cancer cell lines (HCT-116 and HCT-15).13
potencies that are significantly greater than those previously
Herein we describe a comprehensive evaluation of our two
reported, that our derivatives share no homology with other
generations of compounds from a global view where we outline
classes of chemotherapeutic agents, that they are more potent
the three-dimensional (3D) features using NMR and molecular
than a current drug on the market (gemcitabine), and they have
reasonable ClogP values (0.2–3.3 range)18 and molecular
modeling, as well as a complete discussion of all 62 derivatives
structure–activity relationships. Further, these compounds are
weights (500–600), we feel that this class of compounds
described using data from two drug-resistant cell lines to clearly
provides potential structures for further chemotherapuetic
demonstrate a common mode of action against this cancer. Out
development.
Peptides are sometimes considered poor drugs for two
reasons: solubility and rapid degradation within cells. For linear
peptides to achieve 3D structures that will bind appropriately
to their protein targets, they are often composed of extended
sequences of amino acids (aas), which are insoluble. Cyclic
peptides, like San A, often perform better than linear peptides
* To whom correspondence should be addressed. Shelli R. McAlpine,
Department of Chemistry and Biochemistry, 5500 Campanile Road, San
Diego State University, San Diego, CA 92182-1030. Tel.: 619-594-5580.
Fax: 619-594-4634. E-mail: mcalpine@chemistry.sdsu.edu.
†
The University of Kansas.
aAbbreviations: MSS, microsatellite stable; MSI, microsatellite unstable.
10.1021/jm070731a CCC: $40.75
2008 American Chemical Society
Published on Web 01/11/2008

SAR of SansalVamide A DeriVatiVes in Colon Cancers
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3 531
Figure 2. Synthesis of macrocycles. Conditions: (a) coupling agent,* DIPEA (3 equiv), DCM (0.1 M); (b) TFA (20%), anisole (2 equiv), DCM;
(c) LiOH (4 equiv), MeOH; (d) HCl in THF (0.05 M), anisole (2 equiv); (e) HATU (0.7 equiv), DEPBT (0.7 equiv), TBTU (0.7 equiv), DIPEA
(6 equiv), THF/CH3CN/DCM (2:2:1) 0.007 M. *TBTU (1.2 equiv) and/or HATU (0.75 equiv).
because a small number of aas define a 3D structure. San
A-amide analogs also have the advantage that they are lipophilic
and, therefore, they have rapid membrane absorption.19 Cyclic
peptides tend to have greater binding affinity for protein targets
than their linear counterparts or small molecules because they
have restricted bond rotation and are conformationally con-
strained.20 In addition, cyclic peptides degrade much slower than
linear peptides because proteases have difficulty cleaving amide
bonds located in a macrocycle.19 In summary, cyclic peptides
have commercially available chemical diversity (i.e., aas), are
efficiently synthesized, have defined 3D structures that have
good binding affinity for protein targets, are effective at
penetrating cell membranes, and are stable within cells. To date
there are 720 clinically used peptide drugs or candidates; 38%
of these are in clinical trials, 56% are in advanced preclinical
phases, and 5% are on the market.21,22 These peptide drugs are
used as prostate and breast cancer antitumor agents, HIV
protease inhibitors, osteoporosis-treating drugs, and immuno-
suppressants.23 Thus, there is outstanding precedence for treating
diseases with cyclic peptides.
There is an immediate need for new drugs that provide
alternatives for MSIa (drug-resistant) colon cancer patients.
Recently, several new drugs, specifically gemcitabine, oxalipl-
atin, bevucizumab, cetuximab, and the tyrosine kinases inhibi-
tors, have improved survival; however, current therapy is far
from acceptable. Although major efforts have been made, few
truly novel classes of compounds have been identified that have
activity against drug-resistant colon cancers (MSI) tumors. This
work reports our understanding of the complex structure–activity
relationship of 62 San A-amide derivatives against two drug-
resistant colon cancer cell lines, establishes a phenotype for
cytotoxicity in cell-based assays, and connects the 3D position-
ing of the active compounds’ side chains to growth inhibition.
Further, we clearly establish that our compounds are more potent
than Gemcitabine, a current drug treating these cancers.
Synthesis of San A Derivatives. The natural product, San
A, is a depsipeptide, however, macrocyclic peptides possess
increased stability compared to depsipeptides, and therefore, all
62 derivatives described here were constructed as the peptide
analogs (Figure 1). A succinct synthetic protocol has been
developed for the creation of these 62 derivatives.4,24 These
compounds provide valuable information with regards to
stereochemistry, amide bond geometry, and hydrophobic, hy-
drophilic, and aliphatic effects on potency. Further, these
derivatives have ClogP values between 0.18 and 3.3, thus
meeting Lipinski’s rules for solubility and effective diffusion
through cellular membranes.18 Our synthetic approach utilized
a convergent solution-phase strategy to establish a reliable and
inexpensive route for the large-scale production of compound
needed in additional biological studies (Figure 2). As such, our
outlined route provides access to the synthesis of gram quantities
of these compounds.
The synthesis of the San A-amide derivatives were completed
using the aas shown in Figure 3 via a synthetic strategy shown
(Figure 2). We prepared two generations of compounds, wherein
the first generation primarily involved the incorporation of all-
L- or all-D-aa as well as N-methyl moieties at each position
and the second generation included aas that involved the
exchange of a single N-methyl or D-aa at each position (both
generations are highlighted in Figures 4-9). The variations of
each generation allowed us to explore the role of stereochem-
istry, amide bond geometry, and hydrophobic, hydrophilic, and
aliphatic moieties. Overall, the first generation compounds
allowed us to elucidate key SAR involving compounds contain-
ing all D-aas and all L-aas, while the second-generation
compounds provided data on the specific role of each aa as it
relates to stereochemistry, transannular hydrogen bonding, and
polarity.
When 2(1H-benzotriazole-1-yl)-1,1,3-tetramethyl-uronium tet-
rafluoroborate (TBTU) and diisopropylethylamine (DIPEA)
were used, acid-protected residue 1a-f and N-Boc-protected
residue 2a-o were coupled to give the dipeptide 1-2-Boc
(80–94% yield). Deprotection of the amine on residue 2 using
TFA gave the free amine, 1 and 2 (∼quantitative yields).
Coupling of the dipeptide to monomer 3a-n gave the desired
tripeptide (fragment 1) in good yields (80%-95%).25 The
synthesis of fragment 2 was completed by coupling residue
4a-g to residue 5a-f to give dipeptide 4-5-Boc(90–95% yield).
The amine was deprotected on fragment 1 using TFA and the
acid was revealed in fragment 2 using lithium hydroxide.
Fragment 1 and fragment 2 were coupled using multiple
coupling agents,10,26–28 yielding 75 examples of linear pen-
tapeptides (66–90% yield).25 In the case of the di- and tripeptide
construction, acid/base workup removed excess reagents and
side products, and the NMR indicated compounds did not
require further purification. Only in the case of pentapeptides

532 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3
OtruboVa et al.
Figure 3. Amino acids used in the synthesis of 62 derivatives.
and macrocycles did we find it necessary to purify compounds
via silica chromotography, making the synthesis efficient. The
purity of all compounds was verified via NMR and LCMS. The
linear peptides were then cyclized using conditions developed
in our laboratory.24 Upon cyclization, the final compounds were
purified via flash chromatography and HPLC. When appropriate,
the side chains were deprotected (serines, lysines, and tyrosines).
The purity of all compounds was verified by NMR and LCMS.24
Structure–Activity Relationships (SAR). A total of 62 San
A-amide analogs have been tested for cytotoxicity against two
drug-resistant (MSI) colon carcinoma cell lines (HCT-116, HCT-
15). Two cell lines were utilized to ensure that our compounds
were consistently inhibiting drug-resistant colon cancers, to
clearly establish structure–activity relationships (SARs), and to
determine the key features necessary to inhibit growth of colon
cancer cells. Potency exhibited by the San A-amide peptide (1)
is shown so that comparisons can be made between the natural
product peptide and our synthetic analogs. The histogram in
Figure 4 shows the percent inhibition of growth produced by a
concentration of 1 µM compound with changes at position 1
for two MSI colon cell lines. Compounds 2 and 3 are first
generation compounds (shown in gray, Figure 4), while 4-9
are second-generation compounds (shown in black). The first
generation compounds involved incorporation of a single D-aa
in position 1 (2) or an N-methyl at 1 (3). It is clear from the
growth inhibition that the first generation compounds with
changes to position 1 are not potent (note: IC50s are exponential
curves therefore 2 and 3 have IC50s > 50 µM in both cell lines).
For the second-generation compounds in Figure 4, we incor-
porated a number of aas that are different from those of the
natural product primarily at position 1. Compound 4 includes
an L-tetrahydroisoquinoline aa, which places a rigid aromatic
moiety at 1. Given its lack of potency, this moiety is not a
positive influence on growth inhibition. Our first generation data
suggested that a D-aa with an H-bonding element at 1 would
potentially increase the potency, and compounds 5 and 6 were
designed to determine whether a H-bond element would improve
the cytotoxicity effect. Figure 4 shows that the potency of
compound 5 is greatly increased compared to that of 1, where
comparison of the potency of the compounds is done using
growth inhibition curves, which are exponential. The IC50 values
of 5 and 1 (Figure 10) are 1.5 µM versus 25 µM in HCT-116
and 1 µM versus 46 µM in HCT-15, respectively. Thus, 5 is up
to 50-fold more potent than 1. Interestingly, 6 is significantly
less potent than 1, possibly due to steric interactions with the
biological target. Interestingly, 7, which contains a D-tyrosine
like 5 but is also a polar element at position 4 (a lysine), has
greatly reduced potency. Compound 8 involved a significant
alteration in the structure, including a D-phenyl alanine at 1, an
N-methyl at 4, and a D-leucine generated a potent compound.
Comparison of IC50 values indicates that 8 is up to 70-fold more
potent than natural product peptide 1 and up to 100-fold more
potent than first generation compound 3 (Figure 10). Interest-
ingly, compound 9, which contains D-aas at 1, 3, and 4, has
limited growth inhibition against HCT-116 and no growth
inhibition effect against HCT-15. Thus, overall, two compounds
in the second-generation series demonstrated a clear and
significant improvement in growth inhibition, 5 and 8, where

SAR of SansalVamide A DeriVatiVes in Colon Cancers
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3 533
Figure 4. Compounds with changes to position 1. Grey compounds 2 and 3 are first generation compounds, while black compounds 4-9 are
second-generation. Each data point is an average of three wells run in three assays. Error ) (5%.
both show greater than 45% improvement against both cell lines.
These two compounds showed up to 70-fold improved potency
over that seen with the natural product peptide 1 and up to 100-
fold improved potency over first generation compounds.
The histogram in Figure 5 shows the percent inhibition of
growth by compounds with changes in position 2. Compounds
10-12 are first generation compounds (shown in gray, Figure
5), while 13-21 are second-generation compounds (shown in
black). The first generation compounds involved incorporation
of an N-methyl in position 2 (10), a D-aa (11), or both (12). It
is clear from the growth inhibition that the first generation
compounds in this series do not have improved potency over
the parent peptide (1) (note: 10, 11, and 12 have IC50s > 50
µM in both cell lines). For the second-generation compounds,
compound 13 incorporates a D-phenyl alanine at position 2
and is not potent, yet 14 contains an N-methyl D-phenyl
alanine at 2 and shows nanomolar IC50 values (Figure 10).
Both compounds 15 and 16 feature the polar residue D-serine
in position 2, where 16 also contains valine resides in
positions 4 and 5 (versus leucine residues at these positions
in both 15 and 1). Both compounds are inactive compared
to 14, that is, they have IC50s in the high micromolar range.
Further, a more hydrophobic moiety at position 2, for
example, a benzyl oxy group that is seen in 17, also
demonstrated low potency. Additional compounds were then
synthesized to investigate the impact of multiple changes to

534 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3
OtruboVa et al.
Figure 5. Compounds with changes to position 2. Grey compounds 10-12 are first generation compounds, while black compounds 13-21 are
second-generation.
the macrocyclic backbone. This included a compound with
two D-aas in positions 2 and 3 (18) and one with N-methyl
D-aas in positions 2 and 3 (19). Both compounds were not
potent. Compound 20, which contains a D-aa at 2 and an
L-ethyl moiety at 3, was also not potent. Finally, compound
21, which contains an N-methyl D-phenyl alanine and is
similar to 14, but also contains a cyclohexyl moiety at 4,
similar to potent compound 45 (comprehensively discussed
in Figure 7), did not have significant potency. Thus, combined
features from two potent compounds (e.g., 14 and 45) do
not generate synergistic cytotoxic effects. Overall, one
compound in the second-generation series demonstrated a
clear and significant improvement in growth inhibition (14),
which maintains greater than 45% against both cell lines.
Comparison of the IC50 values of 14 and 1 (Figure 10), are
0.8 µM versus 25 µM in HCT-116 and 0.7 µM versus 46
µM in HCT-15, respectively, thus, 14 is up to 70-fold more
potent than 1 and up to 100-fold more potent than first
generation compound 12, which contains an N-methyl
D-leucine.

SAR of SansalVamide A DeriVatiVes in Colon Cancers
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3 535
Figure 6. Compounds with changes to position 3. Grey compounds 22-27 are first generation compounds, while black compounds 9, 18, 19,
28-34 are second-generation.
Compounds 22-27 are first generation compounds (shown
in gray, Figure 6), while compounds 9, 18, 19, and 28-34 are
second-generation (shown in black). The first generation
compounds involved incorporation of an N-methyl in position
3 (22), a D-aa (23), or both (24). Of these three compounds, 23
was the most active. Indeed, it shows an IC50 in the nanomolar
range (Figure 10). In contrast to 1, which contains an isopropyl
moiety at 3, compounds containing a methyl (25) or an ethyl
(26) moiety at position 3, were synthesized and tested. Both of
these compounds showed potency of less than 45% inhibition
at 1 µM against our cell lines. For the second-generation
compounds, 25, 26, and 20 involved alterations at both position
2 and 3 and their potency is discussed earlier. Both compounds
28 and 29 feature the polar residue D-serine in position 3, where
29 also contains valine resides in positions 4 and 5 (versus
leucine residues at these positions in both 28 and 1). Both
compounds are inactive and have IC50s in the high micromolar
range. Compound 30 is similar to potent compound 23, which
has a D-leucine at 3, but rather it contains a D-phenylalanine at
3. In parallel, the cytotoxicity of 31 was tested, where 31 has a
D-benzyl protected serine at 3 and valines substituted into
positions 2, 4, and 5. Both 30 and 31 were inactive relative to
23. Analogously, compounds 32 and 33, which contained a
D-methylalanine at 3 were also both inactive. Interestingly, 23
with a D-valine at 3 is very potent with a nanomolar IC50 value,
yet 34, which contains valines substituted at 2, 4, and 5, in
addition to a D-valine at 3, is not. Additional second generation
compounds included substitutions at both the 3 and 4 positions
or the 3 and 5 positions to explore if specific features that were
critical to potency in the first generation could be utilized

536 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3
OtruboVa et al.
Figure 7. Compounds with changes to position 4. Grey compounds 35-37 are first generation compounds, while black compounds 8, 38-45 are
second-generation.
synergistically to improve activity. These compounds will be
discussed in Figures 7 and 8, respectively. Comparison of the
IC50 values of first generation compound 23 and 1 (Figure 10),
are 1.3 µM versus 25 µM in HCT-116 and 0.8 µM versus 46
µM in HCT-15, respectively. Thus, 23 is up to 60-fold more
potent than 1. Overall, one compound in the first-generation
series, 23, demonstrated a clear and significant improvement
in growth inhibition maintaining greater than 45% against both
cell lines.
Compounds 35, 36, and 37 are first generation compounds,
(shown in gray, Figure 7), while compounds 8 and 38-45 are
second-generation (shown in black). The first generation
compounds involved incorporation of a D-aa at 4 (35), N-methyl
at 4 and a D-aa at 5 (36), or two D-aas in positions 4 and 5 (37).
None of these three compounds were active above 40%
inhibition. For the second-generation compounds in Figure 7,
we incorporated a number of aas that are different from those
of the natural product primarily at position 4. Potent compound
8 involved alterations at positions 1, 4, and 5 and its nanomolar/
low micromolar IC50 values were discussed earlier. Compound
38 involved the incorporation of an N-methyl moiety at 4, and
although it showed some promise in HCT-15 cell lines (i.e., in
the nanomolar potency), it did not show potential for growth
inhibition against HCT-116. Compound 39 incorporates a lysine
at position 4, which greatly diminishes its potency relative to
1. Interestingly, hydrophobic compound 40, which contains a
carboxybenzyl (Cbz)-protected lysine at 4, is relatively active
(averaging 43% against both cell lines). Yet, compound 41,
which incorporates a lysine at 4 and a D-valine at 3 and parallels
39 in structure is not potent, and neither is 42, which contains
a Cbz-protected lysine at 4 and a D-valine at 3, paralleling 40.
It is interesting to note that both 41 and 42 contain the key
element (a D-valine) that makes 23 so active, yet both of these
compounds are not potent despite containing that residue at
position 3. Finally, neither polar compound 43 nor hydropho-
bically related derivative 44 are active (note: both contain three

SAR of SansalVamide A DeriVatiVes in Colon Cancers
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3 537
Figure 8. Compounds with changes to position 5. Grey compounds 46 and 47 are first generation compounds, while black compounds 8, 48-54
are second-generation.
D-aas at 1, 2, and 3). Thus, polar moieties do not enhance the
cytoxicity of compounds nor does the random incorporation of
hydrophobicity (i.e., a Cbz-protected group). Rather, potency
appears to be directly related to the specific conformation of
the individual molecules. Finally, the incorporation of the potent
element from position 3, an N-methyl D-valine and a cyclohexyl
moiety at 4, generated compound 45. This molecule is relatively
potent, with IC50s in the low micromolar and nanomolar range
for the two cell lines. Thus, only two compounds, both second
generation, had growth inhibition percentages greater than 45%
in both cell lines: 8 and 45. Both of these showed a significant
increase in cytotoxicity over 1 and the first generation com-
pounds. It is interesting to note that compound 40, a second-
generation compound, has a 43% average growth inhibition,
yet structurally similar compounds 42 and 44 are inactive. This
difference in potency is believed to be due to constraints
imposed by the other residues within the macrocycle that lead
to a specific conformation, ultimately allowing positive interac-
tions and presentation of residues with its protein target for 40
or negative interactions in the case of 42 and 44.
Compounds 46 and 47 are first generation compounds with
substitutions focusing on position 5 (shown in gray, Figure 8),
while compounds 8, 48-54 are second-generation (shown in
black). The first generation compounds involved incorporation
of a D-aa at 5 (47), or an N-methyl (46). Both of these two
compounds showed growth inhibition at g45% for HCT-116,
but both showed low inhibition of HCT-15. For the second-
generation compounds, potent compound 8 involved alterations
at positions 1, 4, and 5 and its nanomolar/low micromolar IC50
values were discussed earlier. In addition, compound 48

538 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3
OtruboVa et al.
Figure 9. Compounds with N-methyl moieties and their enantiomers. All compounds are first generation.
involved the incorporation of an N-methyl D-aa at 5, and it
demonstrated reasonable potency with an average of potency
of 45% for both cell lines (42% for HCT-116 and 48% for HCT-
15). Thus, it is IC50 values were in the very low micromolar
range (Figure 10). Compounds 49, 50, and 51, which all
contained the potent residue from 23 (a D-valine at position 3),
and substitutions at position 5 had reduced potency compared
to the parent compounds 1, 46, 47, or 23. Thus, combined
features do not generate synergistic cytotoxic effects. Compound
52, which contained an active N-methyl residue at 3 and a
D-leucine at 5 [based on 47] showed no growth inhibition in
either cell line. Finally, compounds 53 and 54, which incorpo-
rated the potent residue from 14, an N-methyl D-phenylalanine
at 2, and a Cbz-protected lysine or lysine at 4, respectively,
both exhibited lower activity than 14. Thus, although two first
generation compounds, 46 and 47, both showed potency against
HCT-116, only one compound (8), which is a second-generation
molecule, had growth inhibition percentages greater than 45%
in both cell lines.
Interestingly, comparing compounds with N-methyls at vari-
ous positions to their enantiomers indicates that two compounds
are relatively active (Figure 9). One is compound 46; as
discussed earlier, it contains an N-methyl at position 5. In
addition, compound 62, which has all D-aas and an N-methyl
on both positions 3 and 5, appears to be relatively active.
Intriguingly, compounds 58 and 61, which are enantiomers of
46 and 62, respectively, are not active, indicating their biological
target is chiral.
The IC50 values for the most potent compounds are shown
in Figure 10. Compound 14 has nanomolar IC50 values for both
colon cancer lines, it is ∼70-fold more active than the parent
natural product peptide 1, and it shows greater than 250-fold
differential selectivity for colon cancer cells over normal cells
(WS1 skin fiber blasts). Gemcitabine was used as a control
compound, as it is currently used for treating a number of
cancers, including pancreatic and colon cancers. Importantly,
compound 14 shows ∼15-fold greater potency than gemcitabine
for inhibiting growth against these cancer cell lines. Thus, these

SAR of SansalVamide A DeriVatiVes in Colon Cancers
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3 539
Figure 10. IC50 values of compounds run in three cell lines: HCT-116, HCT-15, and WS-1 (normal cells, skin fibroblasts). Data represents results
from a concentration curve29 taken from four concentrations, where each concentration data point is from three separate experiments performed in
quadruplicate. Margin of error ) (5%; 200 µM is the outside limit of our detection.
San A-amide derivatives have greater potency than a current
drug on the market, indicating their potential for use as a
therapeutic agent.
N-methyl at 5 and a D-valine at 3, demonstrates lower
cytotoxicity than the “parent” compounds 23 or 46. One very
important aspect shown by this SAR is that polar compounds,
regardless of the positional placement of the polar aa, are
ineffective: 7, 15, 16, 28, 29, 39, 41, 43, and 54. Yet, the
additional substitution of relatively hydrophobic elements with
correct placement are potent: compounds 14, 40, and 45.
Interestingly, compounds with elements that are more hydro-
phobic than the parent compound 1 and do not have the
hydrophobic element correctly oriented, that is, 17, 21, 30, 31,
42, 44, and 53, are not potent.
The key connection between potency and structure involves
constraining the macrocycle into its active conformation and
thus binding to its target in that position. Recent publications
highlighted that a single N-methyl D-aa was the central
structural component required to maintain a dominant
Summary of SAR Results. After testing 62 compounds, we
have come to the conclusion that no single feature or position
is critical to potency but rather, as is typical in complex systems,
there are several determining factors. The most important
features to emerge from this SAR study involve the incorpora-
tion of a D-phenylalanine or D-tyrosine at position 1 (as observed
for 8 and 5, respectively), an N-methyl D-phenylalanine at 2
[seen in 14], a D-valine at 3 [i.e., 23], a hydrophobic moiety at
4 [as denoted by structures 45 and 40], and finally an N-methyl
L-leucine or an N-methyl D-leucine at 5[contained in active
compounds 8, 46, 47, and 48]. However, the combination of
these structural elements does not provide synergistically active
compounds. For example, compound 51, which contains an

540 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3
OtruboVa et al.
Figure 11. The NOEs (row a) and the 3D structures for five compounds, as determined by NMR (row b is equatorial view and row c is top view).
Compounds 1 and 26 are inactive and 5, 14, and 45 are potent compounds.
conformation in macrocycles with five aas.14,15,30 Thus, the
inclusion of a single N-methyl D-aa in our active structures
locks the ring into its low energy conformation, and this
conformation appropriately presents the side chains to the
biological target. Indeed, there is extensive evidence that
cyclic peptides and, specifically, pentapeptides with a single
Spectra were acquired on a Bruker AV-800 instrument with
a TCI cryoprobe operating at 800.23 MHz for H. Samples
1
were dissolved in DMSO-d6 or 75% DMSO-d6/25% H2O at
concentrations of 2–3 mg/mL. Shown in Figure 11 are the
3D structures and the NOEs observed between the amide
protons of five compounds: 1, 26, 5, 14, and 45. These five
compounds were chosen because 1 represents the natural
product peptide structure as a comparison molecule. Com-
pound 26 is a nonactive compound and, therefore, represented
a good negative control, whereas compounds 5, 14, and 45
are all active compounds and, therefore, could be used to
determine a common structural motif.
Formally, a cyclic peptide must reverse the direction of
the peptide chain twice relative to a linear peptide simply as
a consequence of being cyclic. Four of the five compounds
(1, 26, 5, and 14) are observed to have two separated
amide-amide NOEs. Compound 45 contains an N-methyl
residue in the location where an amide-amide NOE might
be expected. The NOE marker of a type II beta turn is a
cross-peak between the amide protons of the third and fourth
residue of the turn. However, this cross-peak appears for other
turn types as well, and our previous experience has shown
that small cyclic peptides may have turn types that deviate
significantly from any of the standard turns.32 The NMR data
indicates that a structure has a turn-like structure centered
on residues 1 and 2 correlates with the potency of the
compound. Compounds 1 and 26 are the control structures
as they are relatively nonpotent compared to 45, 5, and 14.
Both 1 and 26 show weak NOEs between positions 1 and 5,
whereas the potent compounds 45, 5, and 14 all show strong
14,15
D-aa lock the macrocycle into a single conformation.
Further, it is well established that these cyclic pentapeptides
mimic beta and gamma turns and serve as templates for
appropriately positioning suitable binding motifs for pro-
teins.16,17 Thus, out of the seven potent compounds described
in Figure 10, six compounds follow this rule of a single
N-methyl and D-aa: 5, 14, 23, 45, 46, and 48. Further, because
all compounds are cyclic peptides, they have restricted bond
rotations and are conformationally constrained, suggesting
that their binding affinity for their target is greater than any
small molecule.20 In addition, all seven active compounds
are lipophilic, thus, they are rapidly absorbed through
membranes.19 Indeed, any hydrophilic element that was
substituted into the macrocycle greatly diminished their
potency. Finally, these macrocycles are relatively straight-
forward to synthesize and can be made on multigram scale
for further biological studies.
NMR Structural Studies. To explore the proposed hy-
pothesis regarding a single N-methyl and D-aa locking the
compound into an active position, we evaluated our com-
pounds using NMR data and computational data. We
determined the 3D structure of our compounds using TOCSY
and NOESY experiments.31 Note that the NOESY had
stronger cross-peaks for these compounds than the ROESY.

SAR of SansalVamide A DeriVatiVes in Colon Cancers
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3 541
Figure 12. CoMFA steric features for the HCT116 (a) and HCT15 (b) in QSAR models are shown. Features correspond to isosurfaces of the
product of the mean probe-ligand van der Waals interactions and the CoMFA coefficients, with yellow surfaces representing the most favorable
20% of the interaction space, and cyan surfaces representing the least favorable 20%. CoMFA pharmacophore map for the five most potent compounds
(listed in order of most potent to least potent; 8, 14, 23, 5, and 48 in HCT-116 and in HCT-15 compared to nonpotent compound 26 are overlaid
according to the following color scheme (listed most potent to least potent): 8 ) gray, 14 ) yellow, 23 ) green, 5 ) orange, 48 ) magenta, and
26 ) black. The cycle begins at the lower left, and residues 1-5 run counterclockwise around the cycle.
NOEs between 1 and 5. In addition, compounds 1 and 26
show a strong NOE between positions 2 and 3, yet, this is a
weak NOE for 5. Compound 14 also has a weak NOE but it
is between 3 and 4 (because it has an N-methyl at 2, it is
unable to have an NOE between 2 and 3).
From the 3D structures shown in Figure 11, it is apparent
that nonpotent compounds 1 and 26 have very similar side
chain positioning, whereas active compounds 5, 14, and 45
are structurally different from these control compounds.
Although all three active compounds are not identical to each
other in 3D structure, 5 and 14 have similar side chain
orientations that can be seen via the top view (row c).
Compound 45 has a distinctive 3D structure, presumably
because of the high percentage of cis-amides present. This
may also be responsible for the cytotoxicity of 45 against
normal cell lines.
Molecular Modeling Studies Using CoMFA. All 62
compounds were analyzed by a CoMFA model33 to under-
stand their SAR and how it relates to structure. CoMFA
modeling provides a projection of a pharmacophore map that
is generated by fitting the experimental inhibition data to a
partial least-squares fit (Figure 12). Although NMR data can
be utilized using the CoMFA model, CoMFA is most
frequently used to generate 3D structures using growth
inhibition data. Our CoMFA structures were generated from
the growth inhibition data on the HCT-116 and HCT-15
cancer cell lines, and we show that these structures generated
using CoMFA are consistent with the 3D structures generated
by NMR. The pharmacophore maps for each cell line were
almost identical, indicating that the compounds were most
likely reaching the same target in both cell lines. The aligned
ligand manifolds were based on a conformer for the unmodi-
fied San A-amide pentapeptide (1). The conformer was
derived from the linear chain analog of 1 (constructed in
SYBYL35 in a default ꢀ-sheet conformation using the
biopolymer peptide builder) and using molecular dynamics
simulation (Tripos force field).36 The resulting structure was
then optimized in SYBYL to within default molecular
mechanics (MM) convergence criteria using the Tripos force
field. This cyclization procedure was repeated five times using
different molecular dynamic random seeds to ensure con-
sensus in the predicted structure. Although these data are
only a computational representative of the structural con-
former, placing all five amide protons oriented above the
pseudoplane of the ring systematically places all side chains
in distinct regions of space, which generates a relatively
reasonable model for qualitative SAR assessment. The full
manifold of ligands were then constructed by initially
conserving this backbone conformation, then allowing each
San A-amide analog to relax via an unconstrained MM
optimization (same force field and charges as above; default
convergence settings). From the CoMFA analysis, it was
observed that steric/van der Waals interactions accounted for
86% and 84% of the variable dependency for the HCT-116

542 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3
OtruboVa et al.
and HCT-15 models, respectively. This agrees well with our
qualitative assessment that polar residues afford no improve-
ment in bioactivity relative to the hydrophobic aas.
seven small molecules target these drug-resistant colon cancer
cell lines at potencies that are significantly greater than those
previously reported, share no homology with other classes
of chemotherapeutic agents, and have reasonable ClogP
values and molecular weights (450–600). Out of the seven
most potent compounds reported here, five are second-
generation compounds, emphasizing our ability to incorporate
potent features found in the first-generation into the develop-
ment of effective second-generation molecules. Thus, this
class of compounds provides potential structures for further
chemotherapuetic development. Importantly, the most potent
compounds show 70-fold improved potency over the natural
product peptide, 250-fold differential selectivity for colon
cancer cells over normal cells, and 15-fold greater potency
than a current drug on the market used to treat cancers.
Assays determining the active compounds’ mechanism of
action are ongoing and will be reported in the near future.
For our QSAR analysis, we have thus focused variations
in the steric profile of ligands, with the primary features,
that is, pharmacophore maps for the HCT-116 and HCT-15
models being shown in Figure 12. Based on the above
CoMFA analysis, we determined that the top scoring
compounds are 8 and 14, with compounds 23, 5, and 48
ranking close behind.33 The CoMFA data strongly suggests
that potency is enhanced when aas 1 and 5 have strong
interactions (indicated by the yellow shape on the molecule).
Importantly, strong interactions between residues 5 and 1 are
also observed in the NOE data for the potent compounds 5,
14, and 45 (Figure 11),whereas the less potent compounds,
1 and 26, both show weak NOEs between 1 and 5. These
strong interactions appear to be most enhanced when a D-aa
is in positions 1 and 5 (8). Compound 5 may have scored
well because of the hydrogen-bonding element on the
D-tyrosine at 1. A combination of a D-aa in 5 and a D-tyrosine
at 1 is being synthesized. Positive interactions are also
predicted when compounds contain a D-valine at position 3
and an L-aa at position 4 (the yellow shape indicates a positive
interaction between the derivative and the protein target).
These predictions are identical to those gathered from a visual
inspection of our SAR data.
A number of interesting trends are evident in the CoMFA
models. Placement of a D-phenylalanine at position 1 is
generally unfavorable (indicated by cyan feature), except
where the D-Phe at position 1 is capable of interacting with
a D-leucine at position 5 (yellow feature). In HCT116, the
L-Phe is also considered unfavorable (cyan feature on the
bottom left) unless it interacts tangibly with another D-aa on
position 2 (small yellow feature). A methylated amide at 2
(yellow feature) seems to be consistently favorable. Placement
of an L-valine at position 3 is consistently considered
unfavorable. Coupling between L-Leu side chains in positions
4 and 5 is a favorable feature for both cell lines. Collectively,
this reveals complex interdependencies between the various
diversity points in this scaffold. While this preliminary model
provides useful insight in decoupling the various interde-
pendencies, key enhancements are required to derive solid
pharmacophore insight that extends beyond the confines of
the San A-amide based scaffold. First of all, some ambiguities
exist in the relative alignment of the various ligands in that
compounds such as 14 and 13 have two phenylalanine
residues, of which either one could plausibly align with the
Phe-1 residue of San A-amide. Furthermore, a better under-
standing is required of the true bioactive conformer, in that
the structural diversity inherent in the ligand manifold could
dictate the presence of multiple distinct bioactive backbone
conformers throughout the set of San A-amide analogs.
Experimental Section
Thymidine Incorporation/Growth Inhibition Assay. Prolifera-
tion of the HCT-116, HCT-15 colon cancer, and WS-1 normal
cell lines were tested in the presence and absence of the
compounds using 3H-thymidine uptake assays. Cells treated with
the compounds were compared to DMSO-treated controls for
their ability to proliferate as indicated by the incorporation of
3H-thymidine into their DNA. Cells were cultured in 96-well
plates at a concentration of 3000 cells/well. The media was
IMDM for HCT-116, RPMI 1640 for HCT-15, and MEM for
WS-1 with L-glutamine, 10% fetal bovine serum, and 0.1%
penicillin-streptomycin antibiotics. After incubation for ap-
proximately 6 h, the compounds were added. The compounds
were dissolved in DMSO at a final concentration of 2 mM and
tested at the concentrations indicated in the manuscript. The
DMSO concentration was held constant in all wells at 1%. After
the cells had been incubated with the compounds for 56 h, 1
mCi 3H-thymidine per well was added and the cells were cultured
for an additional 16 h (for the cells to have a total of 72 h with
the drug), at which time the cells were harvested using a PHD
cell harvester from Cambridge Technology Incorporated. The
samples were then counted in a scintillation counter for 2 min
each using ScintiVerse universal scintillation fluid from Fisher.
Decreases in 3H-thymidine incorporation, as compared to
controls, are an indication that the cells are no longer progressing
through the cell cycle or synthesizing DNA, as is shown in the
studies presented.
Synthesis. For synthesis details of the first-generation com-
pounds, see ref 4. For second-generation compounds, see ref 5.
The final characterization data of the potent compound are listed
below.
NMR Structural Experiments. Spectra were acquired on a
Bruker AV-800 instrument with a TCI cryoprobe operating at
800.23 MHz for 1H. Samples were dissolved in DMSO-d6 or 75%
DMSO-d6/25% H2O at concentrations of 2–3 mg/mL. The 2D
experiments were done with the standard Bruker parameters and
pulse programs dipsi2esgpph and noesyesgpph, with water sup-
pression provided by excitation sculpting.
Molecular Modeling Studies Using CoMFA. In these prelimi-
nary models, the aligned ligand manifolds were based on a
conformer for the unmodified San A-amide pentapeptide (1). The
conformer was derived from the linear chain analog of 1 (con-
structed in SYBYL34 in a default ꢀ-sheet conformation using the
biopolymer peptide builder) and using molecular dynamics simula-
tion (Tripos force field).35 The resulting structure was then
optimized in SYBYL to within default MM convergence criteria
using the Tripos force field. This cyclization procedure was repeated
five times using different molecular dynamic random seeds to ensure
consensus in the predicted structure. Although these data are only
a computational representative of the structural conformer, placing
all five amide protons oriented above the pseudoplane of the ring
Conclusion
We report here for the first time a comprehensive evalu-
ation of our two generations of compounds against several
drug-resistant colon cancer cell lines. This global evaluation
included the determination of specific features that generate
active compounds in these two generations of derivatives,
that is, placement of a single N-methyl and D-aa, which is
responsible for the primary activity of these macrocyclic
pentapeptides. In addition, it involves an extensive discussion
on the 3D structures using NMR and CoMFA of the active
compounds compared to inactive derivatives. Our most potent

SAR of SansalVamide A DeriVatiVes in Colon Cancers
Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3 543
(6) Liu, S.; Gu, W.; D, L.; Ding, X.-Z.; Ujiki, M.; Adrian, T. E.; Soff,
G. A.; Silverman, R. B. N-Methylsansalvamide A peptide analogues.
Potent new antitumor agents. J. Med. Chem. 2005, 48, 3630–3638.
(7) Pan, P.-S.; McGuire, K. L.; McAlpine, S. R. Identification of
compounds potent against pancreatic cancer cell lines. Bioorganic and
Med. Chem. Lett. 2007, 17, 5072–5077.
(8) Ujiki, M.; Milam, B.; Ding, X.-Z.; Roginsky, A. B.; Salabat, M. R.;
Talamonti, M. S.; Bell, R. H.; Gu, W.; Silverman, R. B.; Adrian, T. E.
A novel peptide sansalvamide A analogue inhibits pancreatic cancer
cell growth through G0/G1 cell-cycle arrest. Biochem. Biophys. Res.
Commun. 2006, 340, 1224–1228.
(9) Otrubova, K.; Styers, T. J.; Pan, P.-S.; Rodriguez, R.; McGuire, K. L.;
McAlpine, S. R. Synthesis and novel structure-activity relationships
of potent Sansalvamide A derivatives Chem. Commun. 2006, 1033–
1034.
(10) Carroll, C. L.; Johnston, J. V. C.; Kekec, A.; Brown, J. D.; Parry, E.;
Cajica, J.; Medina, I.; Cook, K. M.; Corral, R.; Pan, P.-S.; McAlpine,
S. R. Synthesis and cytotoxicity of novel Sansalvamide A derivatives.
Org. Lett. 2005, 7, 3481–3484.
systematically places all side chains in distinct regions of space,
which generates a relatively reasonable model for qualitative SAR
assessment. The full manifold of ligands were then constructed by
initially conserving this backbone conformation, then allowing each
San A-amide analog to relax via an unconstrained MM optimization
(same force field and charges as above; default convergence
settings).
For the HCT116 assay (57 molecules in the original set), a
strongly correlating (R2 ) 0.92; Q2LOO ) 0.74) four-component
model was obtained after discarding four inactive members of
the training set (compounds 13, 25, 28, and 52, respectively).
The 5-fold (FF) cross-validation experiments attest to exceptional
stability and predictivity in the model, with Q2 ) 0.72. For
FF
HCT15 (55 molecules), a solid five-component representation
(R2 ) 0.95; Q2
) 0.71) was attained after discarding five
LOO
inactive ligands (compounds 5, 13, 15, 9, and 52). The model
also exhibits good stability and internal predictivity (Q2
)
(11) Lee, Y.; Silverman, R. B. Rapid, high-yield, solid-phase synthesis of
the antitumor antibiotic Sansalvamide A using a side-chain-tethered
phenylalanine building block. Org. Lett. 2000, 2, 3743–3746.
(12) Gu, W.; Liu, S.; Silverman, R. B. Solid-phase, Pd-catalyzed silicon-
aryl carbon bond formation. Synthesis of Sansalvamide A peptide.
Org. Lett. 2002, 4, 4171–4174.
FF
0.64). From the CoMFA analysis, it was observed that steric/
van der Waals interactions accounted for 86 and 84% of the
variable dependency for the HCT116 and HCT15 models,
respectively. This agrees well with our qualitative assessment
that polar residues afford no improvement in bioactivity relative
to the hydrophobic Phe, Leu, and Val aas. We trained our
CoMFA models to the natural logarithmic quantity log[inh. %].
In an attempt to capture tangible information from inactive
compounds, we circumvented the undefined nature of log[0] by
arbitrarily specifying small nonzero inhibition values of 0.001%
(i.e., log[inh. %] ) -6.91) for inactive species. CoMFA
interaction terms for the ligand manifold were computed using
the Tripos Standard Field probe, including both steric and
electrostatic field components, with the latter incorporating a
distance-based dielectric modulation. Field cut-offs and smooth-
ing protocols were set according to default values. CoMFA
model training and evaluation were then utilized via the
SAMPLS algorithm,36 which reports leave-one-out cross-
(13) A total of 80–85% of colon cancers are considered treatable using
current chemotherapeutic drugs. The remaining 15–20% of colon
cancers are drug-resistant colon cancers. The two types of colon
cancers are referred to as MSS for the drug-sensitive colon cancers
and MSI for the drug-resistant colon cancers. MSS (drug-sensitive)
have an IC50 ∼ 5 µM.
(14) Chatterjee, J.; Mierke, D. F.; Kessler, H. N-methylated cyclic
pentaalanine peptides as template structures. J. Am. Chem. Soc. 2006,
128, 15164–15172.
(15) Heller, M.; Sukopp, M.; Tsomaia, N.; John, M.; Mierke, D. F.; Reif,
B.; Kessler, H. The conformation of cyclo(-D-pro-ala-) as a model for
cyclic pentapeptides of the DL type. J. Am. Chem. Soc. 2006, 128,
13806–13814.
(16) Tyndall, J. D.; Pfieiffer, B.; Abbenante, G.; Fairlie, D. P. Over one
hundred peptide-activated G protein-coupled receptros recognized
ligands with turn structure. Chem. ReV. 2005, 105, 793–826.
(17) Viles, J. H.; Mitchell, J. B.; L., G. S.; Doyle, P. M.; harris, C. J.;
Sadler, P. J.; Thornton, J. M. Multiple solution conformations of the
integrin-binding cyclic pentapeptide cyclo(Ser-D-Leu-Asp-Val-Pro)
analysis of the (phi, psi) space available to the cyclic pentapeptides.
Eur. J. Biochem. 1996, 242, 352–362.
(18) The ClogP values were calculated using an algorithm. The logP value
of a compound, which is the logarithm of its partition coefficient
between n-octanol and water log(coctanol/cwater), is a well established
measure of the compound’s hydrophilicity. Low hydrophilicities and,
therefore, high logP values cause poor absorption or permeation. It
has been shown for compounds to have a reasonable probability of
being well absorbed, their logP value must not be greater than 5.0.
The distribution of calculated logP values of more than 3000 drugs
on the market underlines this fact.
(19) Amidon, G. L.; Lee, H. J. Absorption of peptide and peptidimimetic
drugs. Annu. ReV. Pharmacol. Toxicol. 1994, 34, 321–341.
(20) Wenger, R. M. Synthesis of cyclosporin and analogues: Structural and
conformational requirements for immunosuppressive activity. Prog.
Allergy 1986, 38, 46–64.
validated Q2
values for the PLS-fitted model. To ensure a
LOO
robust, correlative, predictive, and statistically significant model,
we iteratively discarded the maximally outlying molecule in each
training set until a core manifold was obtained whose Q2
LOO
value was greater than 0.60 for a number of PLS components
not exceeding 1/7th of the total number of molecules under
consideration.
Acknowledgment. We thank San Diego State University
for financial support.
Supporting Information Available: A table that includes the
description of all changes made to compound structures in text
format and IC50 data curves are supplied. This material is available
free of charge via the Internet at http://pubs.acs.org.
(21) Marx, V. Watching peptide drugs grow up. Chem. Eng. News 2005,
83, 17–24.
(22) Marx, V. Giving biotechnology a chemical push. Chem. Eng. News
2005, 83, 17–24.
(23) Loffet, A. Peptides as drugs: Is there a market. Eur. Pept. Soc. 2002,
8, 1–7.
References
(1) Cueto, M.; Jensen, P. R.; Fenical, W. N-Methylsansalvamide, a
cytotoxic cyclic depsipeptide from a marine fungus of the genus
Fusarium. Phytochemistry 2000, 55, 223–226.
(2) Hwang, Y.; Rowley, D.; Rhodes, D.; Gertsch, J.; Fenical, W.;
Bushman, F. Mechanism of inhibition of a poxvirus topoisomerase
by the marine natural product Sansalvamide A. Mol. Pharmacol. 1999,
55, 1049–1053.
(24) Styers, T. J.; Rodriguez, R.; Pan, P.-S.; McAlpine, S. R. High-yielding
macrocyclization conditions used in the synthesis of novel Sansalva-
mide A derivatives. Tetrahedron Lett. 2006, 47, 515–517.
(25) Dipeptide and tripeptide structures were confirmed using 1H NMR.
(3) Belofsky, G. N.; Jensen, P. R.; Fenical, W. Sansalvamide: A new
cytotoxic cyclic depsipeptide produced by a marine fungus of the genus
Fusarium. Tetrahedron Lett. 1999, 40, 2913–2916.
(4) Styers, T. J.; Kekec, A.; Rodriguez, R.; Brown, J. D.; Cajica, J.; Pan,
P.-S.; Parry, E.; Carroll, C. L.; Medina, I.; Corral, R.; Lapera, S.;
Otrubova, K.; Pan, C.-M.; McGuire, K. L.; McAlpine, S. R. Synthesis
of Sansalvamide A derivatives and their cytotoxicity in the colon
cancer cell line HT-29. Bioorg. Med. Chem. 2006, 14, 5625–5631.
(5) Rodriguez, R.; Pan, P.-S.; Pan, C.-M.; Ravula, S.; Lapera, S.; Singh,
E.; Styers, T. J.; Brown, J. D.; Cajica, J.; Parry, E.; Otrubova, K.;
McAlpine, S. R. Synthesis of second generation Sansalvamide A
derivatives: Novel templates as potent antitumor agents. J. Org. Chem.
2007, 72, 1980–2002.
1
All linear pentapeptides were confirmed using LCMS and H NMR.
(26) Unpublished results from the Guy lab at the Department of Chemical
Biology and Therapeutics, St Jude Children’s Research Hospital,
Memphis, TN 38103, and published results from our lab show that
the use of several coupling reagents facilitates formation of the peptide
bond in high yields.
(27) Bolla, M. L.; Azevedo, E. V.; Smith, J. M.; Taylor, R. E.; Ranjit,
D. K.; Segall, A. M.; McAlpine, S. R. Novel antibiotics: Macrocyclic
peptides designed to trap Holliday junctions. Org. Lett. 2003, 5, 109–
112.
(28) Liotta, L. A.; Medina, I.; Robinson, J. L.; Carroll, C. L.; Pan, P.-S.;
Corral, R.; Johnston, J. V. C.; Cook, K. M.; Curtis, F. A.; Sharples,
G. J.; McAlpine, S. R. Novel antibiotics: Second generation macro-

544 Journal of Medicinal Chemistry, 2008, Vol. 51, No. 3
OtruboVa et al.
cyclic peptides designed to trap Holliday junctions. Tetrahedron Lett.
2004, 45, 8447–8450.
(29) Concentration curves used to generate the IC50 data are shown in the
Supporting Information.
(30) Zhang, X.; Nikiforovich, G. V.; Marshall, G. R. Conformational
templates for rational drug design: Flexibility of cyclo(D-Pro1-Ala2-
Ala3-Ala4-Ala5) in DMSO solution. J. Med. Chem. 2007, 50,
2921-2925.
(31) NOESY data showed stronger cross peaks than ROESY data.
(32) Liederer, B. M.; Fuchs, T.; Vander Velde, D.; Siahaan, T. J.; Borchardt,
R. T. Effects of amino acid chirality and the chemical linker on the
cell permeation characteristics of cyclic prodrugs of opioid peptides.
J. Med. Chem. 2006, 49, 1261–1270.
for cytotoxicity. CoMFA is a modeling projection of a pharmacophore
map, as derived by partial least squares, which are fit to experimental
inhibition data by a weighted sum of spatial interaction terms between
a probe atom and a manifold of related ligand structures. These
interaction terms are carefully aligned according to conserved structural
features.
(34) SYBYL 7.2, The Tripos Associates, St. Louis, MO 20006.
(35) Clark, M.; Cramer, R. D. I.; Van Opdenbosch, N. Validation of the
general purpose Tripos 5.2 force field. J. Comput. Chem. 1989, 10,
982–1012.
(36) Bush, B. L.; Nachbar, R. B. Sample-distance partial least squares:
PLS optimized for many variables, with application to CoMFA.
J. Comput.-Aided Mol. Des. 1993, 7, 587–619.
(33) CoMFA determines the most likely conformation of the active
compounds and establishes specific interactions that appear important
JM070731A