Classification of auxin related compounds based on similarity of their interaction fields: Extension to a new set of compounds

Tomic, S.^1,2, Gabdoulline, R.R.¹, Kojic-Prodic, B.² and Wade, R.C.¹ ¹European Molecular Biology Laboratory, 69012 Heidelberg, Germany ²Institute Rudjer Boskovic, HR-10000 Zagreb, Croatia Among the five known types of plant hormone, auxins are the most important. In conjunction with other hormones, mostly cytokinins, they control the growth of stems, roots, flowers and fruits. Although auxins were the first type of plant hormone to be identified, the molecular mechanism of their action is still unknown .
We previously developed and applied a procedure employing similarity indices to classify auxin compounds according to their interaction properties [1]. The compounds were divided into four classes, according to their molecular interaction energy fields (MIFs): active, weakly active with weak antiauxin behaviour, inactive and inhibitory. The MIFs of different compounds were compared by computing similarity indices. Here, using this method, the activities of a new set of auxin-related compounds are predicted and a new 3D library of about 70 auxin-related compounds is presented using Chemscape CHIME. We derive a distribution of the compounds in 3D space according to their similarity indices and visualize it with the MAGE program. Analysis of this distribution enables rationalization of the correlations between the classes.

Keywords:QSAR, auxins, molecular modeling, molecular alignment,
interaction similarity, n-alkyl-IAA, CHIME, MAGE

Abbreviations: MIF: molecular interaction field; IAA: indole-3-acetic acid; SI: similarity index; Me: methyl; Et: ethyl; Pr: propyl; Bu: butyl

CONTENTS

1.Introduction

2.Outline of the classification method

3. Classification of new compounds

4. 3D library of auxins

5. Distribution of compounds in 3D space defined by similarity of their interaction properties

6. Summary

7. References

INTRODUCTION

We previously performed a classification of a set of about 50 auxin-related compounds with measured biological activities [2, 28] according to their interaction properties [1]. In this work, we used our classification procedure to predict the activities of 12 additional compounds, most of which are n-alkyl-IAAs. n-Alkylated-IAAs are interesting to study because the relations between their biological activities are expected to be correlated with their binding affinity and this study should provide some information about the possible shape and size of the auxin binding site. For systems of compound homologues and positional isomers, such as n-alkyl-IAAs, metabolic stability, partition between different cell compartments and uptake can be expected to differ less than for chemically more different compounds. For some of the compounds studied here, activity measurements have recently been performed [29, 30] and thus these provide a good test of the methodology. For others, activity measurements will be made [31] but are not yet available, and thus we present these classifications as predictions.
In addition, we have utilized the advantage of the World Wide Web to present: 1) a new 3D library of the auxins, consisting of the compounds classified previously [1] and those classified in this work, using Chemscape CHIME [32]; and 2) a new analysis of the compound classification by projecting similarity indices onto 3D space and displaying their distribution using the MAGE [33] program.

OUTLINE OF THE CLASSIFICATION METHOD[1]

The steps followed are:

(1) Model molecular structures

(2) Align conformers either by using atomic properties (program SEAL [39])
or by using interaction properties on the probe accessible surface of the molecule

(3) Calculate molecular interaction fields (MIF) using the GRID program [40]

(4) Divide compounds into classes according to similarity indices [41, 42] computed for their MIFs.
For auxin-related compounds, the classes are defined as follows. Class 1: active compounds-auxins, class 2: very low activity compounds with antiauxin characteristics, class 3: inactive compounds, class 4: antiauxins, i.e. compounds with inhibitory characteristics.

(5) Classify "new" compounds according to the similarity between their MIFs and the MIFs of "class kernels". A "kernel" is a reduced set of compounds defining the class.

Modeling of compounds:

From ab initio calculations performed on indole auxins [34, 35], there are two main conformations of these compounds differing in the mutual orientation of the indole ring and the main side chain with the carboxyl group: the planar conformation has the side chain in the plane of the indole ring and the tilted conformation has the side chain approximately perpendicular to this plane.
For all other non-indole auxins, two low energy conformations, "P" and "T", that could be aligned best to the planar and tilted IAA conformers, respectively, were modeled.

Although classes were on average distinguishable in both sets of conformations, better differentiation was achieved for the "T" than the "P" conformers [1]. So for the classification of any new compound, low energy conformers similar to the tilted conformation of IAA are used. The "T" conformers of all analyzed compounds are shown in the 3D library of auxins. The conformers may be obtained using any available ab initio, semiempirical or molecular mechanics method that enables adequate modeling of the desired compound. Although none of the molecular mechanics force fields that we tested was able to accurately reproduce results obtained by ab initio calculations (RHF 6-31G*) [34, 35, 36], our experience [36] shows that the CFF91 force field [37] used in the DISCOVER program [38] gives sufficiently good results.

Calculation of MIFs:

Molecular interaction fields were computed for each compound using the GRID [40] program. Calculations were performed either on points on a 1 spacing grid in a volume around molecules or on points defining a probe accessible surface. The most suitable probes were found to be H₂O, NH₂⁺, CH₃ and carbonyl O [1].

CLASSIFICATION OF THE NEW COMPOUNDS

Biological activity

The majority of the 12 new compounds tested in this work are n-alkyl-IAAs. In our previous work [1], we analyzed a few alkyl derivatives of IAA: 2-, 4- and 5- Me-IAA and 4-Et-IAA. In this work, we present the results of similarity analysis on 6-, 7-Me-IAA, 2-, 4-, 5-, 6-, 7-Et-IAA, 5-n-Bu-IAA and 5-n-Pr-IAA.

In addition, we analyzed 4,5-Cl₂-IAA, 5-Br-IAA and 5-methoxy-IAA.

There is only a limited number of biological activity

measurements for these compounds [3, 13, 22, 29, 30, 43]. Binding measurements were performed for 4-Et-IAA and 2-Me-IAA [13] only and there are no measurements on roots, which might point to possible antiauxin properties for the other compounds.

In order to achieve better insight into the auxin characteristics of these compounds, we subjected them to similarity analysis to assign each of them to one of the 4 classes we defined previously [1].

Recently a group of n-alkylated-IAAs have been systematically analysed [29] to study the influence of bulky systems at various positions of the indole ring on growth promoting activity. Although the experiments have not been completed yet, preliminary data on all mono methyl-IAAs (2-, 4-, 5-, 6-, 7-) are available [29].

For the series of 5-n-alkyl-IAAs with the alkyl chain being Me, Et, Pr and Bu, biological activities were measured before [3].

For 2-Et-IAA, the only report on its activity and synthesis is from 1935 [44].

For 4,5-Cl₂-IAA and the other di-chlorinated IAAs, a set of experiments has recently been completed [30].

The biological activities of 5-methoxy-IAA and 5-Br-IAA were measured by Findlay and Dougherty [43] and

Bttger et al. [22], respectively, and new measurements are planned [31].

Modeling

Ab initio RHF/6-31G* conformational analysis on 4-, 5- and 6-Et-IAA has been performed [45] and calculations on 7-Et-IAA are in progress [46]. The purpose of these calculations is to compare the effects of alkyl substituents on the potential energy surface and electron distribution with those of chloro substitutents. In this work, we used the results of the quantum mechanical calculations to model low energy "T" conformers [1] to be subjected to similarity analysis. For 4-, 5- and 6-Et-IAA, the two low energy conformers with the main side chain (side chain with the carboxyl group at the end) approximately perpendicular to the indole ring plane and the ethyl side chain in (p) and tilted (t) to the plane, but oppositely oriented to the main one, were analyzed. The second conformer is similar to the "T" conformer of 4-Et-IAA analyzed before [1]. Both conformers are local minima on the potential energy surface obtained by ab initio (6-31G*) analysis of 4-, 5- and 6-Et-IAA [45]. The latter conformer has lower energy and it is similar to the global minimum for 4-Et-IAA and the second (E=0.194 kJ/mol) and the third (E=0.247 kJ/mol) lowest local minima of 6-Et-IAA and 5-Et-IAA, respectively.

For 7-Et-IAA, the conformer with both side chains tilted to the indole ring plane (oriented opposite to each other) was used in the analysis.

The compounds were aligned using atomic properties and by maximization of similarity indices on the solvent accessible surface, "skin".

Classification procedures

Alignment

For the alignment based on atomic properties, partial charges from the CVFF parameter set [47] were used as before [1] and conformers were aligned using the SEAL [40] program with default weights for both steric and electrostatic terms (alignment referred to as type E in ref [1]).

With the "skin" alignment, two different sets of aligned compounds are defined: one obtained by maximizing similarity indices between each new compound and 2-Me,4-Cl-phenoxy acetic acid, a compound from the auxin (class 1) kernel and the other utilizing 5,7-Cl₂-indole-3-isobutyric acid, a compound from the antiauxin (class 4) kernel. In this way, the influence of the choice of the leading compound on classification was analysed.

Calculation of MIFs

The MIFs were determined on four sets of points: skin obtained by pair wise optimization of MIFs (independent of alignment) and three different grids (for three different alignments), for four different probes.

1) on 3D grids

For all three different alignments, interaction energies with H₂O, NH₂⁺, CH₃ and O probes were calculated on identical grids using the GRID program [40]. For each "new" compound, i, the kernel similarity index KSI_j' (j'=1- 4) was calculated.

where SI_x are Carbo (c) [41] or Hodgkin (h) similarity indices [42] and nkj'= number of compounds in the kernel of class j'.

For the new compounds analysed, the difference between the KSI_c and KSI_h is at most 2 %.

2) on skins

The same procedure was utilized for the MIFs on the compound skins. SIs calculated on skins were maximized for each pair of compounds: a new compound, and a 'kernel compound'. This procedure is independent of alignment.

Results and discussion

The various classification procedures yield similar results, predicting at least weak auxin activity for all compounds analyzed.

Differences between the classification procedures

For the majority of molecules, the classification does not depend on probe, type of alignment or choice of MIFs. However, there are small differences between the results obtained by the different procedures. In the analysis of MIFs on grids, the highest discrimination between classes is achieved with the H₂O probe followed, in order, by the NH₂⁺, CH₃and O probes. In the analysis of MIFs on the skin, the largest differentiation of classes is obtained with the methyl probe, and the lowest with the H₂O probe.

When MIFs are calculated on grids, most of the discrepancies in classification of differently aligned molecules are observed for the NH₂⁺probe. The greatest differences between the classifications with MIFs computed on skins and MIFs computed on grids is obtained for the H₂O and NH₂⁺ probes.

Classification of compounds is based on the results of the majority of predictions. For each molecule, 16 classifications was performed based on 16 different sets of MIFs.

The compounds for which the classification obtained with different procedures differ most are: 7-Et-IAA, 2-Et-IAA and 5-methoxy-IAA.

7-Et-IAA is predicted to be a weak auxin with antiauxin characteristics as it was mostly classified to class 2. However, in a few cases it was classified as class 1: in the set of the compounds aligned by SEAL with the NH₂⁺ probe, in the sets of compounds aligned by skin,

with the H₂O and NH₂⁺ probes and when MIFs on the skin were used with the H₂O probe.

2-Et-IAA and 5-methoxy-IAA are predicted to belong to class 1. However, in sets of compounds aligned by skin (2-Et-IAA in both, and 5-methoxy-IAA in the set aligned to 5,7-Cl₂-indole-3-isobutyric acid) they were put in class 3 with the NH₂⁺ probe. When MIFs on the skin calculated for the H₂O probe were used, 2-Et-IAA and 5-methoxy-IAA were put into classes 4 and 2, respectively.

Calculated vs measured activities

n-Alkyl-IAAs

According to the calculated KSIs, both conformers of 4-, 5- and 6-Et-IAA, both methyl substituted IAAs, 6- and 7- Me-IAA as well as 5-Pr-IAA and 5-Bu-IAA belong to class 1. The same classification was determined before for the "T" and "P" conformers of 4-Et-IAA and 4- and 5-Me-IAA [1]. The results obtained are in agreement with the available experimental data [2, 3, 6, 13, 19, 20, 28, 29]. In Avena coleoptile straight growth tests [2, 3, 19, 29, 42], the ring substituted Me-IAAs, 4-Et-IAA, 5- Et-IAA, 5-n-Pr-IAA and 5-n-Bu-IAA enhance growth. In addition the binding measurements [13] show that 4-Et-IAA and 2-Me-IAA bind to auxin binding protein 1, a putative auxin receptor [48-50], although the binding affinity of 2-Me-IAA is very low. 2-Et-IAA is predicted to be an auxin, as it was mostly classified in class 1. However, its KSI₁ is lower than those of the majority of the other compounds and in three cases it was classified differently: once in class 4, and twice in class 3. Koegl and Kostermans [44] claimed 2-Et-IAA to be inactive. However, this experiment was done in 1935 and the compound might not have been correctly identified when synthesized as a later attempt to synthesize this compound by the procedure given failed [31]. A new experiment for this compound is planned for the near future [31].
By classification procedures, 7-Et-IAA is predicted to have weak antiauxin characteristics. The biological activity of this compound has not been measured yet.

The other IAA derivatives

5-Methoxy-IAA was mostly classified into class 1, except once, in class 3 and once, in class 2. Findlay and Dougherty [43] determined moderate auxin activity for it.
4,5-Cl₂-IAA is predicted to be an auxin in agreement with the recently performed Avena coleoptile straight growth tests on di-chloro substituted IAAs [30], by which 4,5-Cl₂-IAA is one of the most active auxins.
For 5-Br-IAA, Bttgeret al. [22] measured moderate auxin activity in Avena coleoptile, and by our method it is classified in class 1.

Summary

The results of similarity analysis performed on the three sets of differently aligned compounds may be summarized as follows; the final classification of the compounds does not depend on alignment, probe and choice of MIFs, although slight discrepancies for individual probes are possible.

The results on n-alkyl-IAAs are as follows.

Based on the similarity between their molecular interaction fields (MIFs) and MIFs of kernel compounds defining the four different classes of auxins, 2-, 4-, 5-, 6- and 7-Me-IAA, 2-, 4-, 5- and 6-Et-IAA, 5-n-Pr-IAA and 5-n-Bu-IAA belong to class 1: compounds with auxin activity. However, for 2-Et-IAA, the differences between KSI_i (i=1- 4) are low. 7-Et-IAA was assigned to class 2: weak auxins with antiauxin characteristics. It seems that the longer alkyl chain at position 7 of the indole ring introduces antiauxin characteristics and at position 2 decreases the binding affinity of the IAA derivatives. The length of the alkyl chain at the other positions, 4, 5 and 6, does not seem to produce significant changes although KSI₁ for 5-n-Bu-IAA is systematically lower than for the other 5-n-alkyl IAAs.

4,5-Cl₂-IAA, 5-Br-IAA and 5-methoxy-IAA are auxins with activity decreasing from 4,5-Cl₂-IAA to 5-methoxy-IAA.

Figure caption

The mean KSI_i (i=1- 4) values, for the H₂O (#1) and NH₂⁺ (#2) probes, calculated from three sets of differently aligned MIFs on the grid are represented as histograms with their standard deviations given by bars.
The compounds 1-12 are ordered as follows: 4-Et-IAA (1), 5-Et-IAA (2), 6-Et-IAA (3), 7-Et-IAA (4), 7-Me-IAA (5), 6-Me-IAA (6), 5-Br-IAA (7), 5-n-Bu-IAA (8), 4,5-Cl₂-IAA (9), 5-methoxy-IAA (10), 5-n-Pr-IAA (11) and 2-Et-IAA (12). The histograms for 7-Et-IAA and 2-Et-IAA are colored differently.

Figure 2

The results on the probe accessible surface obtained with MIFs for the CH₃probe.

3D LIBRARY of 70 auxin related compounds displayed with CHIME.

class1 strong (and moderately active) auxins,

class2 weak auxins with antiauxin behaviour,

class3 auxin-related inactive compounds,

class4 antiauxins.

The compounds labeled by (*) are utilized as class-kernels

additional set (including compounds analysed before [1] and in this work)
Compounds from this set were classified using the proposed

model

and classes are indicated.

The classification achieved is in accord with the available measured biological activities.

DISTRIBUTION OF SIs OF CLASSIFIED COMPOUNDS IN 3D SPACE VISUALIZED BY MAGE

The analysis of the distribution of similarity indices in 3D space by MAGE is a new approach to the classification problem. It enables better rationalization of correlations between the classes.
The distance between compounds is defined as D(i,j) = sqrt(1-SI(i,j)).
Four points defining the 3D space were chosen as follows:
a) The first two points (1 and 2) are chosen as those for which the distance D(1,2), is largest. This D(1,2) is then used for normalization of all other distances.
b) Point 3 is chosen so that the area of the triangle formed by points 1,2 and 3 is the largest.
c) Point 4 is chosen so that the volume of the tetrahedron formed by points 1,2,3 and 4 is the largest.
The 3D coordinates assigned to these 4 points are (0,0,0), (1,0,0), (a,b,0), (c,d,e) , where a, b, c, d and e are computed unambiguously from the 6 pairwise distances. The other compounds are projected onto this defined 3D space with coordinates determined from their SI(i,j) values.
These projections are visualized using the MAGE program. Distributions are based on SI(i,j) of MIFs determined for the probes H₂O (#1), NH₂⁺(#2), CH₃(#3) and carbonyl O(#4).
In these distributions, three main regions can be distinguished, those populated with auxins, those with inhibitors and those with inactive compounds. The region occupied with auxin-related compounds without growth promoting activity is clearly separated from the others for all probes. However, although most of the antiauxin compounds are far from auxins, the border between these two populations is not always very clear. The compounds from class 2 are situated around the border between classes 1 and 4.

Summary

Prediction of biological activity (and other characteristics) of compounds based on the similarity of their molecular interaction fields to those of characterized compounds is a new, robust QSAR method.

In our previous article [1] we described the method and used it for the classification of auxin related molecules. In this work we used the method to predict the auxin activity of an additional 12 IAA derivatives: 9 alkyl derivatives, 4,5-Cl₂-IAA, 5-Br-IAA and 5-methoxy-IAA. According to the similarity interaction properties, all the compounds analysed possess at least weak auxin activity. 7-Et-IAA is predicted to be a weak auxin with antiauxin characteristics, and 2-Et-IAA and 5-methoxy-IAA are predicted to be of low activity.

We combined the results of both works and made a 3D library of about 70 auxin related compounds.

The distribution of these compounds in a 3D space defined by their similarity indices enables a new analysis of compound classification.

The automatisation of the method enables fast classification of any new set of auxin related

compounds.

The method could be used for the classification and/or prediction of activity of other, non auxin

compounds as well.

Acknowledgements

We are grateful to Dr. V. Magnus for information about biological activity measurements and discussion. S.T. is the recipient of a postdoctoral fellowship from the Alexander von Humboldt Foundation. This work was supported in part by German-Croatian WTZ, Project KRO-001-95.

REFERENCES

2. Porter, W.L. and Thimann, K.V., Phytochemistry, 4 (1965) 229.

3. Kojic-Prodic, B., Nigovic, B., Tomic, S., Ilic, N., Magnus, V., Konjevic, R., Giba, Z. and
Duax, W.L., Acta Crystallogr., B47 (1991) 1010.

4. Nigovic, B., Kojic-Prodic, B., Antolic, S., Tomic, S., Puntarec, V. and Cohen, J.D.,
Acta Crystallogr., B52 (1996) 332.

5. Antolic, S., Kojic-Prodic, B., Tomic, S., Nigovic, B., Magnus, V. and Cohen, J.D.,
Acta Crystallogr., B52 (1996) 651.

6. Lutz, B. T. G., van der Windt, E., Kanters, J., Klmbt, D., Kojic-Prodic B. and Ramek, M.,
J. Mol. Struct., 382 (1996) 177.

7. Ray P. M., Dohrmann U. and Hartel R., Plant Physiol., 60 (1977) 585.

8. Hatano T., Katayama M. and Marumo S., Experientia, 43 (1987) 1237.

9. Hatano T., Kato Y., Katayama M. and Marumo S., Experientia, 45 (1989) 400.

10. Katayama M., Kato Y., Kimoto H., Fuji S., Experientia, 51 (1995) 721.

11. Katekar G. F. and Geissler A. E., Phytochemistry, 21 (1982) 257.

12. Katekar G. F. and Geissler A. E., Phytochemistry, 22 (1983) 27.

13. Rescher U., Walther A. Schiebl C. and Klmbt D., J. Plant Growth Regul., 15 (1996) 1.

14. Reinecke D. M., Ozga J. A. and Magnus V., Phytochemistry, 40 (1995) 1361.

15. Wain R. L. and Wightman F., Ann. Appl. Biol.,40 (1953) 244.

16. Fawcett C. H., Wain R. L. and Wightman F., Ann. Appl. Biol., 43 (1955) 342.

17. Veldstra H. and Van der Westeringh C., Recl. Trav. Chim. Pays. Bas., 70 (1951) 1113.

18. Brustrm H., Physiol. Plant., 3 (1950) 277.

19. Nitsch, J. P. and Nitsch, C., Plant Physiol., 3 (1956) 94.

20. Muir R. M. and Hansch C., Physiol. Plant., 28 (1953) 218.

21. Stenlid G. and Engvild K. C., Physiol. Plantarum, 70 (1987) 109.

22. Bttger, M. Engvild, K. C. and Soll, H., Planta ,140 (1978) 89.

23. Toothill J. R., Wain R. L. and Wightman F., Ann. Appl. Biol., 44 (1956) 547.

24. Hansen B., I. Bot. Nat. (Lond.), (1954) 230.

25. Veldstra H., Rec. Trav. Chim. Pays-Bas, 71 (1952) 15.

26. Pybus M. F., Wain R. L. and Wightman F., Nature, 182 (1958) 1094.

27. Smith G., Kennard, C. H. L. and White, A. H., Acta Crystallogr.,B34 (1978) 2885.

28. Hoffmann, A. L., Fox, S. W. and Bullock, M.W., J. Biol. Chem., 196 (1952) 437.

29. Antolic S., Kojic-Prodic B., Magnus V., Salopek B., Tomic S.:
Plant Physiology 114 (3, Suppl.) 158 (1997).

30. Antolic S., Salopek B., Kojic-Prodic B., Magnus V., and Cohen J. D., Structural
characterization and auxin properties of dichlorinated indole-3-acetic acids.
Plant Growth Regulation (1998), accepted for publication.

31. Magnus V., private communication

32. Chemscape CHIME, http://www.mdli.com/chemscape/chime

33. MAGE_3.1, http://www.faseb.org/protein/kinemages/MageSoftware.html

34. Ramek, M., Tomic, S. and Kojic-Prodic, B., Int. J. Quant. Chem.: Quant. Biol. Symp.,
22 (1995) 75.

35. Ramek, M., Tomic S. and Kojic-Prodic, B., Int. J. Quant. Chem., 60 (1996) 1727.

36. Tomic S., Ramek, M. and Kojic-Prodic, B., Combined ab initio SCF and molecular
mechanics studies of propionic and isobutyric acids and their indole derivatives related
to the phytohormone auxin (indole-3-acetic acid), accepted for publication in Croat.
Chem. Acta 3 (1998).

37. Maple J. R., Thacher T. S., Dinur U., Hagler A. T., Chem. Design Autom. News, 5(9)
(1990) 5.

38. BIOSYM (1995) DISCOVER, version 2.97, Biosym Technologies, 10065 Barnes
Canyon Rd., San Diego CA 92121, USA.

39. Kearsley S.K., Smith G.M., Tetrahedron Computer Methodology, 3 (1990) 615.

40. GRID user manual, edition 14, Molecular Discovery Ltd., West Way House, Elms
Parade, Oxford OX2 9LL, England

41. Carbo, R., and Calabuig B., Int. J. Quant. Chem., 42 (1992) 1681.

42. Hodgkin E. E. and Richards W. G., Int. J. Quant. Chem.: Quant. Biol. Symp., 14 (1987)
105.

43. Findlay, S.P. and Dougherty, G. J. Biol. Chem. 1883 (1950) 361.

44. Koegl F.and Kostermans D. G. F. R. Hoppe Seyler's Z. physiol. Chem. 235,(1935) 210.

45. Ramek M. and Tomic S. RHF Conformational Analysis of the Auxin Phytohormones
n-Ethyl-Indole-3-Acetic Acid (n=4,5,6), accepted for publication in Int. J. Quant. Chem. .

46. Ramek M., private communication
47. Dauber-Osguthorpe P., Roberts V. A., Osguthorpe D. J., Wolff J.,
Genest M., Hagler A. T., Proteins: Structure, Function and Genetics, 4 (1988) 31.

49. Tian, H., Klmbt, D., Jones, A. M., J. Biol. Chem., 270 (1995) 26962.

50. Lbler, M. Klmbt, D., J. Biol. Chem., 260 (1985) 9848.

Privacy Imprint