3D QSAR Analysis of Colchicine Analogs

J. R. Dahl, M. E. Staretz, J. L. Potenziano and S. B. Hastie, Department of Chemistry, State University of New York, Binghamton, NY 13902-6000


1. To obtain a predictive QSAR model to aid in the synthesis of colchicine analogues to help interpret colchicine-tubulin interactions in the binding site.

2. Use Conformational Molecular Field Analysis as a visual tool to assess potential regions of a theoretical binding pocket for steric and electrostatic interactions with colchicine analogues.


The colchicine binding site on tubulin is a major receptor site for antimitotic drugs. Drug binding to this site alters the normal assembly of tubulin into microtubules. This plays an integral part in cell division and other microtubule-mediated processes.[1] Since the three dimentional structure of tubulin is unknown, the nature of the colchicine binding site on tubulin as yet is undetermined. We explored the tubulin-drug interaction at the colchicine site by the use of Conformational Molecular Field Analysis (CoMFA).[2] This relatively new approach to QSAR allows the visualization as well as quantitation of electrostatic and steric fields involved in the tropone region of the binding site. In this work, two series of colchicine analogues were synthesized and their biological activities assessed by the concentration at which they inhibited 50 % of the in vitro assembly of tubulin into microtubules (I50). In the first series the C-10 position of colchicine (1) was altered (Figure 1 and Table 1). The second series was comprised of a substituted aromatic ring which replaced the tropone ring. This group of compounds represents the allocolchicine series (Figure 1 and Table 2). Prediction of I50 values using COMFA was attempted for colchicine analogues that were not part of the training set or had not yet been tested for biological activity.



Bovine microtubular protein (MTP) was used for all I50 measurements. The protein was extracted from fresh bovine brain tissue and purified by successive cycles of polymerization and depolymerization, before a final separation using ion exchange chromatography.[3] The I50 values were obtained on 20 (M MTP containing 1 mM GTP in PME buffer ( Pipes 100 mM, MgSO4 1mM, EGTA 1mM), pH 6.9 at 37( C, after 20 minutes of incubation. Six different concentrations of drug were tested in duplicate on a 20 minute polymerization monitored every 5 seconds at 350 nm. The experiment was performed on a Hewlett-Packard 8451 UV-visible spectrophotometer containing a thermostated cell holder. The drug concentration at which 50% of the MTP polymerization inhibition occurred, was determined by a linear extrapolation of the data. All the molecules used in this study were synthesized by our laboratory except (1) and (36).


QSAR analysis was performed on a Silicon Graphics Workstation (4D/35, Irix 5.2), using Sybyl 6.1 molecular modelling software.[4] CoMFA was accomplished using 33 colchicine analogues (19 from the colchicine series and 14 from the allocolchicine series) as a training set for development of the model. The crystal structures for colchicine[5], allocolchicine[6], and isocolchicine[7], were entered into a database. The subsequent derivatives were contructed based on the crystal structure coordinates of these three molecules. Charges were calculated by two methods, Gasteiger-Huckel and the semiempirical MNDO method. The AM1 method could not be used due to the presence of two sulfur containing molecules and the MOPAC version used. The substituent sidechains of all analogues were minimized and the trimethoxy aromatic rings of all molecules were aligned using the FIT subroutine in Sybyl 6.1. Alignments were optimized by a SYSTEMATIC SEARCH of all sidechains.

The CoMFA subroutine placed the molecules in a 3-D grid which contained 2 angstrom spacings and a +1 charge at intersecting points. All default parameters were used in our analysis. Partial least squares (PLS) was then used in the CoMFA to obtain a linear equation from the data. An initial analysis was performed to determine the crossvalidated r-squared (a value between 0 (no relationship) and 1 (perfect model). This r-squared term determines the validity of the non-crossvalidated model. The partition coefficients used in this analysis were calculated using the method of Bodor et.al.[8]


Tables 1 and 2 list the I50 values of the training set of molecules used to develop the model. The CoMFA results are depicted visually in the photographs by the red, blue, green, and yellow contours around the substituted region of the molecule. Colchicine(1) is depicted as the prototype molecule in the photographs. The green area represents regions where steric bulk is allowed, while yellow depicts regions unfavorable to steric interactions. The blue area is a region where positive charge interactions are favored, while the red region represent more favorable interaction with negative charges.

The final PLS regression equation was:

I50 = -0.827 - (0.221) * LOGP

with a crossvalidated r-squared of 0.604 using the "leave one out" option for crossvalidation and 3 components. The non-crossvalidated r-squared was 0.84 (0.05 for this model. The standard error of the estimate was 0.348. A graph of actual vs predicted Log I50 is shown in figure 1, where the x-axis represents the actual log I50 and the y-axis contains the predicted log I50. Table 3 demonstrates the predicted and the actual I50 values for analogues synthesized but were not used to develop the model.

Photographs A, B and C show colchicine with the electrostatic and steric fields in proximity. Photograph D illustrates the alignment of all 33 colchicine analogues, in the presence of the CoMFA field. The final two photographs depict the same colchicine molecules and CoMFA fields. Photograph E is a Connolly surface of electrostatic potential, and photograph F is the same surface countour, but with the lipophilicity potential. The lipophilic character of a region is noted by the analog scale to the left of photograph F. Brown shows a more positive lipophilic region of the molecule and blue represents a more hydrophilic region. In photograph E the electrostatic potential is demonstrated by the most positive potential in blue regions and the most negative potential in orange or brown regions.


CoMFA has given insight into the interactions around the tropone ring of colchicine. Since the three dimensional structure of tubulin is unknown CoMFA is an indirect method of investigating the binding site. After alignment of the trimethoxy ring, it is apparent that the only difference between colchicine (1) and isocolchicine (19) is the inversion of the carbonyl oxygen and methoxy group at the C-9 and C-10 positions respectively. Since the I50 for these two molecules is a difference of 3 orders of magnitude analysis of this region was undertaken. When all of the molecules are aligned in the colchicine series, none of the substituents can be configured so they interfere with the C-9 position of (1) or (19). However, when the compounds in the allocolchicine series are added, the substituents are in an area between the C-9 and C-10 positions of (1). This is a consequence of the aromatic ring of (20) replacing the tropone ring of (1). Such an alignment allows the larger substituents of (24), (27), and (32) access to this area if they are conformationally unrestricted. In photographs A , B, and C the contour area in yellow represents the sterically disallowed region. The green contour represents a sterically allowed region near the C-10 position. This is consistent with the sterically bulkier groups of (6), (8), (9), (10), and (11). The electrostatic regions are more difficult to interpret, but appear consistent with the data. The red regions suggests an area where interactions with negative charges are more favorable. One of these regions lies to one side of the plane of the tropone ring as shown in photographs A, B, and C. A negative charge on the molecule in this area would be favorable to hydrogen bond donation. The blue region interacts faborably with a positive charge and thus would be an area to suspect hydrogen bond acceptor groups on the protein.


1. A linear model was obtained with a relatively good predictive ability to guide the synthesis of tubulin binding drugs structurally similar to colchicine.

2. The steric and electrostatic interactions must be further investigated. However, there appears to be a sterically allowed region near the C-10 position of colchicine and a sterically restrictive area near the C-9 position.


1. H. G. Capraro and A. Brossi. Tropolonic Colchicum Alkaloids. A.Brossi (ed.), The Alkaloids, Vol XXIII, Academic Press, New York, 1984, pp. 1-70.

2. R. D. Cramer III, D. E. Patterson, J. D. Bunce. Comparative molecular field analysis (CoMFA). 1. Effect of shape on binding of steroids to carrier proteins. J. Am. Chem. Soc. 110:5959-5967 (1988).

3. R. C. Williams and H. W. Detrich III. Separation of tubulin from microtubule-associated proteins on phosphocellulose. Accompanying alterations in concentration of buffer components. Biochemistry 18(12):2499-2503 (1979).

4. SYBYL Molecular Modeling Software, Version 6.1, TRIPOS, Inc., St. Louis, Missouri (1994).

5. L. Lessinger and T. N. Margulis. The crystal structure of colchicine. A new application of magic integers to multiple-solution directe methods. Acta. Cryst. B34:578-584 (1978).

6. M. F. Mackay, E. Lacey, and P. Burden. Structures of colchicine analogues. I. Allocolchicine. Acta. Cryst. C45:795-799 (1989).

7. L. Lessinger and T. N. Margulis. The crystal structure of isocolchicine, an inactive isomer of the mitotic spindle inhibitor colchicine. Acta. Cryst. B34:1556-1561 (1978).

8. N. Bodor, Z Gabanyi, and C. Wong. A new method for the estimation of partition coefficient. J. Am. Chem. Soc. 111:3783-3786 (1989).

Return to the Professor Bane's page