HOW MULTIPLE AUXIN RESPONSIVE ELEMENTS MAY INTERACT IN PLANT PROMOTERS: A REVERSE PROBLEM SOLUTION

HOW MULTIPLE AUXIN RESPONSIVE ELEMENTS MAY

INTERACT IN PLANT PROMOTERS: A REVERSE

PROBLEM SOLUTION

VICTORIA V. MIRONOVA

Institute of Cytology and Genetics SB RAS

10 Lavrentyev Ave., Novosibirsk 630090, Russia

Novosibirsk State University, 2 Pirogov Str., Novosibirsk 630090, Russia

[email protected]

NADYA A. OMELYANCHUK*, MARIA S. SAVINA*,†,PETR M. PONOMARENKO*, MIKHAIL P. PONOMARENKO*,

VITALY A. LIKHOSHVAI*,† and NIKOLAY A. KOLCHANOV*,†

*Institute of Cytology and Genetics10 Lavrentyev Ave., Novosibirsk 630090, Russia

†Novosibirsk State University

2 Pirogov Str., Novosibirsk 630090, Russia

Received 15 October 2012

Revised 12 December 2012Accepted 13 December 2012

Published 8 February 2013

Plant hormone auxin is a key regulator of growth and development. Auxin a®ects gene ex-

pression through ARF transcription factors, which bind speci¯cally auxin responsive elements(AuxREs). Auxin responsive genes usually have more than one AuxRE, for example, a widely

used auxin sensor DR5 contains seven AuxREs. Auxin responsive regions of several plant genes

have been studied using sets of transgenic constructions in which the activity of one or severalAuxREs were abolished.

Here we present the method for analysis of the datasets on promoter activity assays having

promoter sequences, namely, number and sequences of AuxREs, altogether with their measured

auxin induction level. The method for a reverse problem solution considers two extreme modelsof AuxRE cooperation. Additive model describes auxin induction level of a gene as a sum of the

individual AuxREs impacts. Multiplicative model considers pure cooperation between the

AuxREs, where the combined e®ect is the multiplication of the individual AuxRE impacts.

The reverse problem solution allows estimating the impact of an individual AuxRE into theinduction level and the model for their cooperation. For promoters of three genes belonging to

di®erent plant species we showed that the multiplicative model ¯ts better than additive. The

reverse problem solution also suggests repressive state of auxin responsive promoters before

auxin induction. The developed method provides possibility to investigate AuxRE structure-activity relationship and may be used as the basis for a novel approach for AuxRE recognition.

Keywords: Auxin; Auxin Responsive Element (AuxRE); primary auxin response; plant; reverseproblem; quantitative structure�activity relationship.

Journal of Bioinformatics and Computational BiologyVol. 11, No. 1 (2013) 1340011 (21 pages)

#.c Imperial College Press

DOI: 10.1142/S0219720013400118

1340011-1

1. Introduction

Plant hormone auxin regulates growth and development by activating or inhibiting

expression of many genes. Among them, genes of early auxin response are the most

well studied at present. The primary/early response to auxin means changes in gene

expression within 30min after exogenous auxin treatment.1 Comprehensive tran-

scriptome analysis over a wide range of auxin treatments and in various tissues

revealed that the primary auxin response was almost entirely restricted to up-reg-

ulated genes.2 There are three main types of early auxin response genes��� Aux/IAA,

SAUR and GH3.3 Auxin responsive elements (AuxREs) in the promoters of some of

these genes have been analyzed using a variety of genetic manipulations (deletion

analysis, linker scanning, site-directed mutagenesis and others).4�6 The current

model for auxin response mechanism is the following: at low auxin levels a hetero-

dimer of an ARF and an Aux/IAA proteins is bound to the promoter of an early

auxin response gene, thus keeping the promoter in a repressed state.2,7 Auxin Re-

sponse Factors (ARFs) are transcription factors, which bind speci¯cally to the

AuxREs and by this way activate or inhibit transcription from the target genes. The

Aux/IAA proteins are the repressors for ARF-activators.2 High auxin level promotes

proteasome-mediated degradation of Aux/IAA proteins and allows ARF-activators

to enhance or induce transcription of auxin responsive genes.

In auxin responsive promoters usually there are multiple AuxRE sites. For ex-

ample, the promoter of widely used auxin sensor DR5 consists of seven AuxREs.6

Auxin induction level can widely vary depending on the number of AuxREs (and the

length of a spacer among them), their core sequences and °anks.1�3,6,8�11 Moreover,

both ARF-activator and ARF-repressor may bind to AuxRE.6 Despite the wide

application of the DR5 reporter, there is a little data on the mechanisms underlying

the high e±ciency auxin response of this reporter and other genes with multiple

AuxREs in their promoters.

In the present work by the reverse problem solution we performed systematic

analysis on activity of multiple AuxREs in auxin responsive gene promoters, both

wild type and subjected to genetic manipulations. For three cases: (1) DR5 reporters

in arabidopsis6; (2) mutational analysis of soybean GH3 promoter5; (3) linker

scanning mutations in Pisum sativum IAA4/5 promoter,4 ��� we succeeded in linear-

additive approximation for the contributions of individual AuxREs to the magnitude

of auxin response. Our analysis on the data from three experiments provided inde-

pendent evidences for the following: (i) at low auxin level promoters of auxin in-

ducible genes are under repression; (ii) multiple AuxREs in a gene promoter

synergistically interact providing for the magnitude of auxin response where the

individual impacts multiplied together. In addition, quantitative structure�activity

relationship analysis of the multiple AuxREs from the Ps-IAA4/5 gene suggested

that the activity of an AuxRE site negatively correlates with the abundance of SnW

trinucleotides and an increase in the DNA inter base pair step parameter Slide in its

nucleotide sequence.

V. Mironova et al.

1340011-2

2. Materials and Methods

2.1. Datasets and data processing method

Promoter regions having di®erent numbers of AuxRE sites and driving various ac-

tivities of reporters in auxin response were taken from the published data.4�6,10 We

collected the sequences of TGTCnC-like AuxREs from these papers with [�16; þ 16]

°anks and the total length L¼ 32 nt. The DNA sequences f�ig1�i�L, � 2 fA;T;G;Cgwere centered to the proved, predicted or destroyed AuxRE core.

As genetic manipulations may di®erently a®ect AuxRE activity, we describe it

here using the remaining activity coe±cient �AuxRE. For a gene with multiple

AuxREs this coe±cient values can hardly be evaluated experimentally, so we

assigned �AuxRE value to each AuxRE depending on the distance of mutation posi-

tion from the AuxRE center. AuxRE was considered as normally functioning,

�AuxRE ¼ 1, if its sequence with the length L, had not been changed by genetic

manipulations. If any nucleotide in the L length AuxRE sequence was changed

within the 3 bp, 9 bp, 12 bp, 15 bp around the AuxRE center, we set the remaining

activity coe±cient �AuxRE as 0, 0.6, 0.85 and 0.9, respectively. If changes were made

in the both °anks, we multiply these coe±cients together. If a new TGTCnC-like site

was formed de novo due to genetic manipulations we consider this site as normal.

2.1.1. Processing the data on DR5 reporters6

We took the sequences from the paper by Ulmasov and colleagues.6 The min-35S

without DR5 sequence (5'-tgtgtgagtagttcccAgataagggaattaggg-3') was used as a

control and the base for making DR5 constructions. DR5 sequences themselves were

inserted before the capitalized \A" in the min-35S. mDR5-DR5 is 5'-

ccttttgGctccctttTGTCTC-3'; DR5(Nx) is (cctttTGTCTC)N where N ¼ 2 . . . 8, DR5

(8x)rev is (GAGACAaaagg)8.

The activity coe±cients of the individual AuxREs were estimated by the rule

described above in 2.1 and listed in the Table 1. The wild type AuxRE sequence in

the DR5 promoter was actually taken from GmGH3 promoter5 and has an activity

Table 1. AuxRE sequences from the DR5 constructions6 with their estimated activity coe±cients.

n DR5 variant AuxRE sequence, TGTCnC� 13bp*

Substitutions in 5'/3'

AuxRE °anks �AuxRE

(i) (ii) (iii) (iv) (v)

1 min35S tgtgtgagtagttcccagataagggaattagg noAuxRE 0

2 mDR5-DR5 tttggctccctttTGTCTCagataagggaatt 5 bp/wt 0.63 DR5(Nx), 5'-copy gtagttcccctttTGTCTCccttttgtctcag wt/1bp 0.6

4 DR5(Nx), internal copy tttgtctccctttTGTCTCccttttgtctccc 5 bp/1bp 0.36

5 DR5(Nx), 3'-copy tttgtctccctttTGTCTCagataagggaatt 5 bp/wt 0.6

6 DR5(8x)rev, 5'-copy tttgtctccctttTGTCTCgggaactactcac 5 bp/wt 0.67 DR5(8x)rev, 3'-copy ccttatctcctttTGTCTCccttttgtctccc wt/1bp 0.6

*modi¯ed sequences are underlined

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-3

(impact to auxin induction) - ’DR5. The scheme of the regulatory regions in the DR5

constructions and their response to auxin treatment6 is shown in Fig. 1.

Using the experimental data6 we calculated the auxin induction level for each

genetic construction as ratio of the reporter activity after and before auxin treat-

ment, normalized to the control (Fig. 1).

2.1.2. Processing the data on GH3 promoter variants5

Similarly, we processed the data on the genetic constructions with auxin responsive

soybean GH3 promoter.5 From the genetic constructions we took the sequences of

TGTCnC-like AuxREs (Table 2, Fig. 2). Three AuxREs at the positions �228, �176

and �129 above the transcription start site were considered to have the activities

’�228GH3, ’�176GH3 and ’�129GH3, respectively.

2.1.3. Processing the data on Ps-IAA4/5 linker scanning mutations4

Ballas and co-authors created 20 reporter lines under the part of Ps-IAA4/5 pro-

moter with inserted through the whole auxin responsive promoter region 10 bp linker

scanning (LS) substitutions (AAGCTAGCAA).4 As a result of the analysis of these

lines the authors found out in Ps-IAA4/5 promoter two extended auxin responsive

islands, called A- and B-boxes. They noted two signi¯cant for auxin response AuxRE

sites at the positions �176ðþÞ and �187ðþÞ in the A-box with the TGTC(C/A)C

consensus. However, the pro¯le of auxin response along the position of LS sub-

stitutions in the analyzed part of the promoter suggests that this region contains

more AuxREs. Carefully analyzing the pro¯le we predicted the presence of 8 addi-

tional potential AuxREs (see Table 3 for the details). To explain the increase in

auxin induction caused by the LS mutations we predict de novo formation of three

AuxREs at positions �224ð�Þ, �225ð�Þ and �304ð�Þ in LS-226, LS-216, and LS-

306 reporter lines, respectively.

As the number of potential AuxREs exceeds the number of variables that can be

reliably estimated from the linear regression for 20 independent experiments, we

classi¯ed the potential AuxREs according to their position. All potential AuxREs

within the A-box and the B-box were considered to have the activities ’A-box and

Fig. 1. The scheme for the regulatory regions in the DR5 reporters with the remaining activity coe±cients

of the individual AuxREs. The auxin induction levels for the various DR5 variants were taken from.6

V. Mironova et al.

1340011-4

’B-box, respectively. All other AuxREs were united to the \C-box" and considered to

have an activity ’C-box.

2.1.4. Processing the data on P3(4X) auxin sensor10

As an independent data for testing the AuxRE features found by quantitative

structure�activity relationship analysis (see below) we used the data on mutation

analysis of the P3(4X) reporter.10 The authors made 12 point mutations in the

Table 2. Normal and mutated AuxREs in the variants of the GmGH3 promoter5 considered here.

n

GH3 reporter,

AuxRE position, (chain) AuxRE sequence TGTCnC� 13bp*

Nearest substitutions

in 5'/3' °anks �AuxRE

(i) (ii) (iii) (iv) (v)

1 WT, �228, (�) gtgggtccagatgTGTCgCcacgtcagcaaaa wt/wt 1

2 M1, �228, (�) gtgggtccagatgTGTCgCcacgaaagcaaaa wt/5bp 0.6

3 M11, �228, (�) gtgggtccagatgTGTCgCctttaaagcaaaa wt/2bp 0.64 [�229;�110], �228, (�) gtgggtccagatgTGTCgCcgggaactactca wt/2bp 0.6

5 WT, �176, (þ) tccctggccctcgTGTCTCctcaataagctac wt/wt 1

6 [�183; 0], �176, (þ) gttcccgccctcgTGTCTCctcaataagctac 8 bp/wt 0.85

7 [�179; 0], �176, (þ) agtagttccctcgTGTCTCctcaataagctac 7 bp/wt 0.858 [�173; 0], �176, (þ) tgtgtgagtagttcccCTCctcaataagctac core/wt 0

9 L1, �176, (þ) tccctggaagatcttcCTCctcaataagctac core/wt 0

10 D1, �176, (þ) gttcccgccctcgTGTCTCctcaataagctaa 8 bp/wt 0.8511 [�183; �159], �176, (þ) gttcccgccctcgTGTCTCctcaataagctaa 8 bp/wt 0.85

12 WT, �129, (þ) acacgcaatccttTGTCTCaataagttccact wt/wt 1

13 L5, �129, (þ) aagatcttcccttTGTCTCaataagttccact 5 bp/wt 0.6

14 min35S tgtgtgagtagttcccagataagggaattagg noneAuxRE 0

*modi¯ed sequences are underlined

Fig. 2. The scheme for the regulatory regions in the GmGH3 reporter lines.6 The auxin induction level of

the reporters5 as well as the remaining activity coe±cients of an individual AuxREs are shown for each

line.

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-5

Table 3. Normal and mutated AuxREs in the LS reporter lines of Ps-IAA4/5.4

n Box Position Chain Line

AuxRE, sequence TGTCTC� 13bp,

linker

Substitution,

distance �AuxRE

Solution

’AuxRE

1 �127 (�) WT aggcaagctaagtTGTCGGtagcaacatcaac wt 1 1.30

2 LS-140 aggcaagctaagtTGTCGGtagcaacttgcta 8 bp 0.85 1.25

3 C �135 (þ) WT ttaaatctgttgaTGTTGCtaccgacaactta wt 1 1.30

4 LS-140 ttaaaagctagcaaGTTGCtaccgacaactta core 0 1.00

5 LS-151 caaaatctgttgaTGTTGCtaccgacaactta 12 bp 0.95 1.28

6 �141 (þ) WT cattctttaaatcTGTTGAtgttgctaccgac wt 1 1.30

7 LS-140 cattctttaaaagctagcAagttgctaccgac core 0 1.00

8 LS-151 agctagcaaaatcTGTTGAtgttgctaccgac 6 bp 0.6 1.17

9 LS-159 aattctttaaatcTGTTGAtgttgctaccgac 13 bp 0.95 1.28

10 �161 (�) WT atttaaagaatggTGTCTCcttataggggtga wt 1 1.79

11 LS-140 ttttaaagaatggTGTCTCcttataggggtga 13 bp 0.95 1.73

12 LS-151 attttgctagcttTGTCTCcttataggggtga 1 bp 0.6 1.42

13 LS-159 atttaaagaatttgcTagCtttataggggtga core 0 1.00

14 LS-170 atttaaagaatggTGTCTCcttttgctagctt 4 bp 0.6 1.42

15 A �176 (þ) WT tatgtcccattctTGTCACccctataaggaga wt 1 1.79

16 LS-159 tatgtcccattctTGTCACccctataaagcta 9 bp 0.85 1.63

17 LS-170 tatgtcccattctTGaagCtagcaaaaggaga Core 0 1.00

18 LS-181 tatgaagctagcaaGTCACccctataaggaga Core 0 1.00

19 LS-191 gcaatcccattctTGTCACccctataaggaga 10 bp 0.85 1.63

20 �187 (þ) WT tggtaggtgaataTGTCCCattcttgtcaccc wt 1 1.79

21 LS-170 tggtaggtgaataTGTCCCattcttgaagcta 8 bp 0.85 1.63

22 LS-181 tggtaggtgaataTGaagCtagcaagtcaccc Core 0 1.00

23 LS-191 tggtaaagctagcaaTCCCattcttgtcaccc Core 0 1.00

24 LS-201 agcaaggtgaataTGTCCCattcttgtcaccc 10 bp 0.85 1.63

25 C �224 (�) LS-226 tctggaatttggaTGTTGCtagcttttttgag de novo 1 1.30

26 �225 (�) LS-216 ctggttgctagctTGTTGAgattgttttgaga de novo 1 1.30

27 �270 (þ) WT attcacatgctcaTGTTTCctcaaaatcaacg wt 1 1.19

28 LS-251 attcacatgctcaTGTTTCctcaaaatcaagc 12 bp 0.95 1.17

29 LS-261 attcacatgctcaTGTTTaagctagcaaaacg Core 0 1.00

30 LS-271 attcacataagctaGcaaCctcaaaatcaacg Core 0 1.00

31 LS-279 aagctagcaatcaTGTTTCctcaaaatcaacg 4 bp 0.6 1.11

32 LS-285 gcaaacatgctcaTGTTTCctcaaaatcaacg 10 bp 0.85 1.16

33 B �282 (�) WT ggaaacatgagcaTGTGAAtggagggtccctt wt 1 1.19

34 LS-261 ttaaacatgagcaTGTGAAtggagggtccctt 12 bp 0.95 1.17

35 LS-271 ggttgctagcttaTGTGAAtggagggtccctt 2 bp 0.6 1.11

36 LS-279 ggaaacatgattgctaGcttggagggtccctt Core 0 1.00

37 LS-285 ggaaacatgagcaTGTttgctagctttccctt Core 0 1.00

38 LS-295 ggaaacatgagcaTGTGAAtggagggttgcta 9 bp 0.85 1.16

39 �293 (�) WT catgtgaatggagGGTCCCttatgtgagttgg wt 1 1.19

40 LS-271 tatgtgaatggagGGTCCCttatgtgagttgg 13 bp 0.95 1.17

41 LS-279 tgctagcttggagGGTCCCttatgtgagttgg 6 bp 0.6 1.11

42 LS-285 catgtttgctagcttTCCCttatgtgagttgg Core 0 1.00

43 LS-295 catgtgaatggagGGTtgCtagcttgagttgg Core 0 1.00

44 LS-306 catgtgaatggagGGTCCCttatgtgttgcta 10 bp 0.85 1.16

45 C �302 (�) WT ggagggtcccttaTGTGAGttggttatgggaa wt 1 1.30

46 LS-285 tagctttcccttaTGTGAGttggttatgggaa 8 bp 0.85 1.25

47 LS-295 ggagggttgctagctTGAGttggttatgggaa Core 0 1.00

48 LS-306 ggagggtcccttaTGTGttgctagcttgggaa core 0 1.00

49 LS-315 ggagggtcccttaTGTGAGttggttattgcta 9 bp 0.85 1.25

50 �304 (�) LS-306 agggtcccttatgTGTTGCtagcttgggaaag de novo 1 1.30

V. Mironova et al.

1340011-6

AuxRE site ft�4 ! a�4, T�3 ! a�3, G�2 ! a�2, T�1 ! g�1, C1 ! a1, t2 ! g2,

C3 ! a3, c4 ! a4, c5 ! a5, a6 ! g6, t2 ! c2, t2C1 ! c2t1g. The data were processed

similarly to the described above.

2.2. The reverse problem under consideration

The auxin induction level ’ of a gene in which regulatory region there are N AuxRE

sites is a complex function on individual AuxRE activities f’ng1�n�N. A priori one

can expect in°uence of any of the (2N � 1) possible combinations to auxin response,

which, in a general case, can be calculated as:

’ ¼ ’0 1þ P1

X1�n1�N

’n1þ � � � þ Pk

X1 � n1 � N� k

n1 < n2 � N� 1

: : :

nk�1 < nk � N

’n1n2...nk�1nkþ � � � þ PN’1...N

0BBBBBBBBBB@

1CCCCCCCCCCA

ð1Þ

where: 1 < k < N, ’0 is a basal auxin induction level (for example, without auxin);

Pk, 1 < k < N; are the contribution coe±cients for the k-combination from N

AuxREs, �1�n�NPn � 1; ’��...& is an impact of a given set of �-th, �-th, … and &-th

AuxREs into the auxin induction level ’ of the whole promoter. Auxin response

levels ’ are determined above for all the experiments.4�6

As Eq. (1) contains 2N variables (’0; ’0P1’n1; . . .’0Pk’��...& ; . . . ; ’0Pn’1...N), the

optimal number of independent experimental data for its reliable statistical estima-

tion by the method of multiple linear regression is ð2NÞ2. For N ¼ 3 in the case of the

soybean GH3 promoter it is 23 ¼ 8 variables, which estimation needs 82 ¼ 64 inde-

pendent experiments instead of 16 ones provided in.5 For DR5 with N ¼ 8 and Ps-

IAA4/5 with N ¼ 10 AuxREs this number is even higher. Thus, solution of the

problem (Eq. (1)) using the experimental data on auxin responsive gene expression4�6

is impossible. This explains why systematic in silico analysis of the auxin induction

mediated by multiple AuxREs has not been done yet.

We consider here two extreme cases of the general model (Eq. (1)): the additive

(Eq. (2)) and multiplicative models (Eq. (3)). The ¯rst model fP1 ¼ 1; Pn>1 � 0g,which we call here additive, relates to the situation with a negligible impact of any

interaction between the AuxREs comparing to their additive impacts:

’ ¼ ’0 1þXNn¼1

’n

!; ð2Þ

An alternative multiplicative model fPn<N � 0; PN ¼ 1g suggests that under

auxin induction all the AuxREs in auxin responsive gene promoter purely cooperate

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-7

with each other providing for increase in gene expression:

’ ¼ ’0 1þYNn¼1

’n

!; ð3Þ

Using the STATISTICA package, we estimated the part of the variance in the

experimental data (Fig. 1�3), which can be described by the alternative models

(Eq. (2)) and (Eq. (3)). The following transcendent form of the multiplicative model

(Eq. (3)) was calculated in STATISTICA:

ln’� ’0

’0

� �¼XNn¼1

lnð’nÞ: ð4Þ

The common method for solution of (Eq. (4)) is based on the sandwich theorem. It

is approaching limð’� ! ’0Þ�!0 as ln((’�’ðkÞ� Þ=’ ðkÞ

� Þ ¼ �ðkÞ þ �1�n�N ln(’n) for a

limited number of k steps with the sandwich rule f’ ðkþ1Þ� ¼ ð’ ð1�k 0�k: maxð�ðk 0ÞÞ<0Þ

� þ’

ð1�k 00�k: minð�ðk 00ÞÞ>0Þ� Þ=2g at a given signi¯cance level �; or establishing absence of

the limit.

2.2.1. The reverse problem for DR5 reporters6

For each of the DR5 reporter variant we described the auxin-induction level (Y ) as:

Y ¼ X0 þXNn¼1

�AuxRE;nXAuxRE;n; ð5Þ

Fig. 3. The scheme for the regulatory region in the Ps-IAA4/5 reporter lines.4 The auxin induction level

of the reporters4 as well as the estimated activity coe±cients of an individual AuxREs are shown for each

line.

V. Mironova et al.

1340011-8

where: Y is equal ’ (Fig. 1), X0 is the basal induction level ’0;XAuxRE;i ¼ ’0’DR5 in

the case of (Eq. (2)); or Y ¼ ln½ð’� ’0Þ=’0�, X0 is equal � and XAuxRE;i ¼ lnð’DR5Þin the case of (Eq. (4)). We used the remaining activity coe±cients �AuxRE from the

Table 1 and got the following system of linear equations:

Y ¼ X0; For \min� 35S";

Y ¼ X0 þ 0:6XAuxRE; For \mDR5DR5";

Y ¼ X0 þ ð1:2þ 0:36ðN � 2ÞÞXAuxRE ; For \DR5ðNxÞ":

8>><>>: ð6Þ

Calculating the multiple linear regression on the basis of (Eq. (6)) in the STA-

TISTICA package we consider Y as \Dependent variables", XAuxRE and X0 as

\Independent variables".

2.2.2. The reverse problem for GH3 promoter5

For the data on GH3 genetic constructions (Fig. 2, Table 2), using four independent

variables, X0;X�228AuxRE ;X�176AuxRE and X�129AuxRE, we created the system of 16

linear equations of the following type:

Y ¼ X0 þ �AuxREX�228AuxRE þ �AuxREX�176AuxRE þ �AuxREX�129AuxRE : ð7Þ

In analogy to DR5, X�228AuxRE ¼ ’0’�228GH3, X�176AuxRE ¼ ’0’�176GH3,

X�129AuxRE ¼ ’0’�129GH3 in the case of (Eq. (2)) or X�228AuxRE ¼ lnð’�228GH3Þ,X�176AuxRE ¼ lnð’�176GH3Þ, X�129AuxRE ¼ lnð’�129GH3Þ in the case of (Eq. (4)).

2.2.3. The reverse problem for Ps-IAA4/5 promoter4

To describe the IAA-induction for Ps-IAA4/5 genetic constructions (Fig. 3) we

created the system of 21 linear equations (Eq. (8)) with 4 variables X0;XA-box,

XB-box, and XC-box. Each AuxRE in the boxes A, B, C has the activity, ’A-box, ’B-boxand ’C-box, respectively. Within the box AuxREs has the same activity but among

the boxes the activity of individual AuxRE di®ers. To estimate decrease in AuxRE

activity caused by genetic manipulation we used the remaining activity coe±cients

�AuxRE from the Table 3.

Y ¼ X0 þX

AuxRE 2A�box

�AuxRE

!XA�box þ

XAuxRE 2B�box

�AuxRE

!XB�box

þX

AuxRE 62A�box;B�boxf g�AuxRE

0@

1AXC�box: ð8Þ

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-9

2.3. Quantitative structure�activity relationship

analysis of AuxREs

The reverse problem solution allows quantitatively estimating the impact of indi-

vidual AuxRE into the auxin-induction of gene expression. Here, we show how these

data can be explored further. We used the tool ACTIVITY12 to ¯nd out the context

or physico-chemical features of the AuxRE, which can signi¯cantly contribute to its

activity and strength. For this case we used our estimates for Ps-IAA4/5 AuxREs

(Table 6).

For the 36 bp length AuxRE sequences (Table 6) and their estimated ’AuxRE

values, ACTIVITY12 has found their signi¯cant correlations with (i) 1 of 38 DNA

helix properties from the database13 (Eq. (9)):

Pk½a;b�½AuxREn � snj

� �� ¼ 1

b� a

Xb�1

j¼a

Pkðsnj snjþ1Þ; ð9Þ

or (ii) 1 of 153 ¼ 3375 weighted abundances for all the possible trinucleotides z1z2z3(Eq. (10)):

½Z1Z2Z3�F ðAuxREn � snj� �Þ ¼ Xþ16

j¼�16

F ðjÞY3k¼1

�ðsnjþk�1 ¼ ZkÞ; ð10Þ

where: [a; b] is the �13 bp neighborhood of the TGTCnC core, �16 � a < b � þ16;

1 � k � 38; Z 2 fA; T; G; C; W ¼ Aþ T; R ¼ AþG; M ¼ Aþ C; K ¼ TþG;

Y¼Tþ C; S ¼ Gþ C; B ¼ TþGþ C; V ¼ AþGþ C; H ¼ AþTþC; D ¼AþTþG; n ¼ Aþ TþGþ Cg, 15-symbol nomenclature IUPAC-IUB CBN14; 0 �FðjÞ � 1, is the weight function: the higher the F(j), the larger the impact to auxin-

induction of the trinucleotide z1z2z3 at the position j. ACTIVITY accounts 360

variants of the F(j) (Fig. 4).

Every of possible 32� 31� 38 ¼ 37696 variants of Pk;½a;b� (Eq. (9)) and every of

possible 360� 153 ¼ 1215000 variants of ½Z1Z2Z3�F (Eq. (10)) were processed by the

ACTIVITY12 uniformly and independently one of another for the sequences fsAuxREj gand taking into account the estimated ’AuxRE. The each pair [xfsAuxREj g; ’AuxRE],

x2 fPk;½a;b�, ½Z1Z2Z3�FÞ is checked by the ACTIVITY for reliability using (i) 5 types of

Fig. 4. Some examples of the F(j) weight function considered in the ACTIVITY tool (Eq. 10)). The bold

line F(j) was found signi¯cant for the SnW trinucleotide.

V. Mironova et al.

1340011-10

correlation tests and (ii) 6 requirements for regression analysis in the multiple boot-

strap15 for the subgroups of analyzing data. We did it to reduce the dependence of the

solution results on data heterogeneity. On the basis of Zadeh's fuzzy set16 and utility

theory for decision making,17 ACTIVITY12 assigns [xfsAuxREj g;’AuxRE] the value �½xfsAuxREj g;’AuxRE� from�1 toþ1. The higher the�, themore the reliable correlations

and the ful¯lled requirements were identi¯ed on the biggest number of the bootstrap-

subgroups from ½xfsAuxREj g;’AuxRE�.

3. Results and Discussion

Genetic engineering experiments on auxin responsive genes provide data about a

complex structure of auxin responsive regions with multiple AuxREs. Due to this

complexity, estimation of individual AuxRE contributions is a problem which can

not be solved in a general case (Eq. (1)). For the reverse problem (Eq. (1)) solution

the number of the independent experiments increases as an exponential function

depending on the number of AuxREs in a promoter. Nevertheless, intensive genetic

experiments for the analysis of auxin responsive regions4�6 can be formalized in silico

to get some hidden features in the structure and functioning of auxin inducible

promoters. Here we present a quantitative structure�activity relationship analysis

for the auxin responsive elements in DR5, GmGH3 and Ps-IAA4/5 promoters.

As a ¯rst step we formalized the data on the structure of auxin responsive regions

and auxin induction level of the reporter lines (Fig. 1�3) from studies.4�6 We col-

lected sequences of wild type and mutated AuxREs and estimated the decrease in

their activity level caused by genetic manipulations (Table 1�3). At the second step,

we chose the model for AuxREs interaction in the auxin responsive promoter. Two

alternative models of AuxREs interaction for auxin responsive gene expression which

are the particular cases of the general function (Eq. 1)) were considered. The additive

model (Eq. (2)) describes auxin induction level as a summation of the impacts of all

the individual AuxREs. In contrast, the multiplicative model (Eq. (3)�(4)) accounts

for auxin-induction level as a multiplication of all the AuxREs impacts. Below we

present the results on the reverse problem solution using the both models.

3.1. The reverse problem solution for DR5 reporters6 in Arabidopsis

The auxin induction levels of eight DR5 reporter lines6 were described as functions on

individual AuxRE activities (Eq. (5)). The linear regression for (Eq. (6)) with 2

variables was calculated in the STATISTICA package. We got statistically unreli-

able result (� > 0:1) using the additive model (Eq. (2)). The multiplicative model

(Eq. (4)) gave us the statistically reliable (� < 0:025) solution: ’0 ¼ 0:53� 0:34,

lnð’DR5Þ ¼ 1:30� 0:14 (Table 4). The simple correlation coe±cient r ¼ 0:867 be-

tween the experimental data6 and the in silico prognosis (Table 4) can be interpreted

that the prognosis de¯nes more than 87% of the variance in the experimental data.6

This suggests that the results of (Eq. (3)) solution can be considered as an approx-

imate solution of (Eq. (1)) for the DR5 case.

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-11

The obtained results allow us to conclude that the multiplicative model (Eq. (4))

describes well interaction of AuxREs in the DR5 promoter during auxin response.

The value of ’0 < 1 allow proposing a repressive state of the DR5 reporters before

auxin treatment.

3.2. The reverse problem solution for soybean GH3 promoter5

Genetic analysis of the soybeanGH3 promoter5 revealed three TGTCnC-like AuxREs

(Table 2). We described the auxin induction level in 17 (Fig. 2) reporter lines as a

function on the activities of these three AuxREs (Eq. (7)). Similarly to the DR5

reporters, the additive model (Eq. (2)) did not give us a reliable solution (� > 0:15).

However, the multiplicative model (Eq. (4)) solution was statistically reliable

(� < 0:01): ’0 ¼ 0:40� 0:23, lnð’�228AuxREÞ ¼ 1:04� 0:29, lnð’�176AuxREÞ ¼ 1:51�0:28, lnð’�129AuxREÞ ¼ 0:98� 0:29 (Table 5). Thus, analysis of this data provided an

independent support for ¯tness of themultiplicativemodel as a ¯rst approximation for

AuxRE interaction in auxin responsive gene expression. In addition, the obtained

results allow us to range and compare the activities of the individual AuxREs in the

promoter.

3.3. The reverse problem solution for Pisum sativum IAA4/5 reporters4

The auxin responsive region of Ps-IAA4/5 gene was investigated by linker scanning

mutagenesis in 20 reporter lines.4 In this region we predict the presence of 10 po-

tential AuxREs having similarity with TGTCnC consensus to explain the variation

in the auxin response level of all the reporter lines (Table 3). We also predict that

two AuxRE sites were formed de novo in the lines LS-226 and LS-216. The AuxREs

were classi¯ed into three groups (A-, B- and C-boxes) according to their position (see

Methods for details).

Table 4. The reverse problem (Eq. (6)) solution for auxin induced expression of the DR5 reporters.6

DR5

reporters6AuxRE

number

Auxin induction,

’, in vitro6

The total

activity of

AuxREs,PNi �AuxRE;i

ln½ð’� ’0Þ=’0�,’0 ¼ 0:53� 0:34

ð� < 0:025Þ

Auxin induction,

’, prediction

(� < 0:0001),

lnð’DR5Þ ¼ 1:30� 0:14

(i) (ii) (iii) (iv) (v) (vi)

#min-35S 0 1.00 0.00 �0.13 1.06

mDR5-DR5 1 1.32 0.60 0.40 1.69

DR5(2x) 2 3.00 1.20 1.54 3.05

DR5(4x) 4 15.60 1.92 3.35 6.94

DR5(6x) 6 26.48 2.64 3.89 16.83DR5(7x) 7 28.90 3.00 3.98 26.52

DR5(8x) 8 28.45 3.36 3.96 41.99

DR5(8x)rev 8 27.55 3.36 3.93 41.99

Linear correlation coe±cient (signi¯cance), r ¼ 0:867ð� < 0:01Þwas used to normalize the auxin induction level

V. Mironova et al.

1340011-12

Tab

le5.

Thereverse

problem

(Eq.(7))

solution

forthesoybeanGH3reporters.5

GH3reporters

Number

ofAuxREs

Auxin

induction,

’,in

vitro5

Rem

ainingactivitycoe±

cient,�AuxRE

ln½ð’

�’0Þ=’0�,

’0¼

0:40

�0:23

(�<

0:01

)

Auxin

induction,’,

predicted(�

<0:005),

lnð’

�228AuxREÞ¼

1:04�0:29

lnð’

�176AuxREÞ¼

1:51�0:28

lnð’

�129AuxREÞ¼

0:98�0:29

�228

�176

�129

(i)

(ii)

(iii)

(iv-228

)(iv-176

)(iv-129

)(v)

(vi)

WT

39.30

11

13.11

13.95

M1

37.06

0.6

11

2.82

9.33

M11

38.50

0.6

11

3.02

9.33

[�20

6;0]

27.88

01

12.94

5.17

[�18

3;0]

27.20

00.85

12.84

4.20

[�17

9;0]

24.04

00.85

12.22

4.20

[�17

3;0]

11.67

00

11.16

1.45

#[¡

107;0]

01.00

00

00.42

0.79

min35

S0

0.54

00

0�1

.01

0.79

L1

22.64

10

11.73

3.40

L5

311

.50

11

0.6

3.33

9.56

�(�

180;�1

10)

13.13

10

01.93

1.53

[�22

9;�1

10]

37.70

0.6

11

2.91

9.33

D1

11.70

00.85

01.19

1.83

D2

00.70

00

0�0

.27

0.79

D3

11.42

00

10.94

1.45

D4

11.53

00

11.05

1.45

Linearcorrelationcoe±

cient(signi¯cance),r¼

0:89

6ð�

<10

�6Þ

#was

usedto

normalizetheau

xin

inductionlevel

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-13

The reverse problem (Eq. (8)) was calculated in the STATISTICA package for

the variables ’0; ’A-box; ’B-box and ’C-box (Eq. (2)) or their natural logarithms

(Eq. (4)). Again, the additive model (Eq. (2)) did not give a statistically reliable

solution (� > 0:33), while the multiplicative model (Eq. (4)) did. The statistically

reliable (� < 0:05) solution of (Eq. (4)) is: ’0 ¼ 0:23� 0:56, lnð’A-boxÞ ¼ 0:86� 0:11;

lnð’B-boxÞ ¼ 0:22� 0:08; lnð’C-boxÞ ¼ 0:33� 0:10 (Table 6).

3.4. How multiple AuxREs interact in plant promoters

We showed above that the multiplicative model of AuxREs interaction in auxin

responsive gene expression works well for three di®erent genes belonging to three

di®erent plant genomes. The multiplicative model (Eq. (3)�(4)) explains more than

86% of the variance in the experimental data for DR5, GmGH3 and Ps-IAA4/5 (see

the correlation coe±cients in the Tables 4�6). While in silico prognosis using the

additive model (Eq. (2)) has not found to correlate with any of the experimental data

sets. We also received statistically relevant results with the multiplicative model on

other data sets (data not shown). The ¯tness of the model on a wide range of

experimental sets suggests that in the general model (Eq. (1)) the main impact to the

auxin induction level has the multiplicative e®ect between all the AuxREs and the

additive e®ect is minor.

The multiplicative e®ect (Eq. (4)) means that the impacts of all the AuxREs are

multiplied in auxin response providing for increase in gene expression. This can be

the case when the preinitiation complex of transcription factors exists on the auxin

responsive gene promoter waiting for the stimulus to start transcription. Another

important fact from the reverse problem solution is the values of ’0, which are less

than 1 for all the three genes. This means that before auxin treatment the auxin

responsive genes are under repression. The repression level seems to correlate with

the number of AuxREs, it is the lowest for Ps-IAA4/5 with ten potential AuxREs.

This result leads to propose the following mechanism for communication of

regulatory information between di®erent AuxREs in promoters of auxin-responsive

genes. All AuxREs in a promoter due to DNA loops and twists may build a single

space unit where the preinitiation complex consisting of ARFs, Aux/IAAs and other

proteins is formed. Without auxin both dissociation of the complex and initiation of

transcription from the promoter are repressed. This mechanism is consistent with

the generally accepted model for expression regulation of auxin responsive genes.2

According to this model ARFs normally bind auxin responsive regions as an inactive

heterodimers ARF-Aux/IAA. Aux/IAA degradation rates signi¯cantly increase

with increasing auxin concentration in a cell. Thus, induction of auxin responsive

genes starts really not from the basal but from a repressed level. Aux/IAA proteins

are short-lived, and therefore ARF-Aux/IAA heterodimers may dissociate without

increasing auxin concentration. We assume that multiplicative AuxRE interactions

in this case may prevent expression of auxin responsive genes due to sporadic

changes in Aux/IAA proteins level. In this case transcription will start only when

V. Mironova et al.

1340011-14

Tab

le6.

Thereverse

problem

(Eq.(8))

solution

forPs-IA

A4/5reporters.4

LS-lines

ofPs-IA

A4/54

Number

ofAuxREs

Auxin

induction,

’,in

vivo

4

Thetotalactivityof

AuxREs,P N i

�AuxRE;i

ln½ð’

�’0Þ=’0�,

’0¼

0:23

�0:56ð�

<0:02

5Þ

Auxin

induction,

’,predicted(�

<0:05),

lnð’

A-boxÞ¼

0:86�0:11

lnð’

B-boxÞ¼

0:22�0:08

lnð’

C-boxÞ¼

0:33�0:10

A-box

B-box

C-box

(i)

(ii)

(iii)

(iv-A

)(iv-B

)(iv-C

)(v)

(vi)

WT

1020

.10

3.00

3.00

4.00

4.45

21.87

LS-315

1015

.85

3.00

3.00

3.85

4.21

20.84

LS-306

1030

.94

3.00

2.85

4.00

4.89

21.17

LS-295

911

.91

3.00

1.85

3.00

3.92

12.36

LS-285

814

.17

3.00

0.85

3.85

4.10

13.06

LS-279

918

.29

3.00

1.20

4.00

4.36

14.79

LS-271

911

.30

3.00

1.55

4.00

3.87

15.95

LS-261

917

.00

3.00

1.95

4.00

4.28

17.40

LS-251

1025

.25

3.00

2.85

4.00

4.68

21.17

LS-243

1022

.13

3.00

3.00

4.00

4.55

21.87

LS-234

1016

.31

3.00

3.00

4.00

4.24

21.87

LS-226

1145

.22

3.00

3.00

5.00

5.27

30.19

LS-216

1130

.00

3.00

3.00

5.00

4.86

30.19

LS-209

1025

.75

3.00

3.00

4.00

4.70

21.87

LS-201

1012

.92

2.85

3.00

4.00

4.00

19.26

LS-191

910

.42

1.85

3.00

4.00

3.79

8.29

LS-181

83.33

1.00

3.00

4.00

2.60

4.12

LS-170

97.50

1.45

3.00

4.00

3.45

5.95

LS-159

98.33

1.85

3.00

3.95

3.56

8.16

LS-151

99.91

2.60

3.00

3.55

3.73

13.49

LS-140

814

.11

2.95

3.00

1.85

4.09

10.53

Linearcorrelationcoe±

cient(signi¯cance),r¼

0:86

2ð�<

10�6Þ

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-15

the concentration of auxin really increase to the level providing dissociation of all

the heterodimers.

Note that the multiplicative model (Eq. (3)�(4)) is just the ¯rst approximation for

the AuxRE interaction mechanism in auxin responsive gene expression. The real

mechanisms are certainlymore complex andmay include intricate interaction between

additive and multiplicative mechanisms. Nevertheless, we showed the primary role of

multiplicative interactions of AuxREs during auxin induction of gene expression.

3.5. Quantitative structure�activity relationship analysis

of AuxREs sites

The reverse problem solution allows us to estimate and compare activities of the

individual AuxRE sites. To present further possible outputs of the method we per-

formed bioinformatic analysis of the wild type and mutated AuxRE sequences from

the Table 3 using ACTIVITY tool12 (See Methods for details).

We found that among other physicochemical AuxRE features (Eq. (9)), the most

signi¯cant linear-additive impact to auxin response level belongs to the DNA inter

base pair step helical parameter Slide at the region [�12; þ12] around the TGTCnC

consensus, �(Slide½�12;12�;’AuxREÞ ¼ 0:253. Slide means the displacement along an

axis in the plane of the base pair directed from one strand to the other. The

Slide½�12;12� values for the AuxRE sequences4 shown on x axis in Fig. 5(a) signi¯-

cantly correlate (r ¼ �0:397; � < 0:005) with our estimates for ’AuxRE. Thus, more

the Slide value around the AuxRE, less the impact of the AuxRE to the auxin

induction of a gene expression.

For signi¯cant context features (Eq. (10)) around the AuxRE, the ACTIVITY13

found that the SnW trinucleotide has the highest weight F(j) (Fig. 4, bold line) and

linear additive impact, �([SnW]\;’AuxREÞ ¼ 0:238, to auxin response level. The

values of [SnW]\ shown in the Fig. 5(b) signi¯cantly correlate (r ¼ �0:690; � < 10�7)

with our estimates for ’AuxRE of the individual AuxRE. Thus, we conclude that it has

to be a depletion of SnW trinucleotides around the functional AuxRE sites.

Using these statistically reliable correlations we constructed multiple linear

regression in the STATISTICA package (Eq. (11)):

’AuxREðAuxREn � fsnj gÞ ¼ 3:33� 0:26½SnW �\fsn

j g� 1:39Slide½�12;þ12�fsn

j g: ð11ÞIn silico prognosis (Eq. (11)) for ’AuxREfsAuxREj g shown on Fig. 5(c) signi¯cantly

correlates ðr ¼ 0:705; � < 10�9Þ with our estimates for ’AuxRE of the Ps-IAA4/5

AuxREs from the Table 5. As one can see, joint account of the signi¯cant context

and DNA conformation features improves 100 times the reliability of the in silico

prognosis. We veri¯ed the prognosis function (Eq. (11)) using the data from the

independent experiment10 with the P3(4X) reporters. The in silico prognosis for

estimated ’AuxREfsAuxREj g values of the P3(4X) reporter shown on Fig. 5(d) signif-

icantly correlates ðr ¼ 0:619; � < 0:025Þ with the experimental data.10

V. Mironova et al.

1340011-16

Thus, we obtained some evidences on context and conformation features of

AuxRE sites other than well known TGTCnC consensus. We showed here that these

features may also have a predictive value.

4. Conclusion

Transcription factors that mediate cellular responses to external and internal stimuli,

often have multiple binding sites in the promoters of target genes. The functional

contribution of the individual binding sites and the mechanisms of their interaction

for transcription initiation are mainly unknown. The indirect data on the mutational

analysis of the promoters with multiple binding sites may be used to study this aspect

of transcriptional regulation. Here we propose the method of the reverse problem

solution for the impacts of multiple binding sites into inductive gene expression on

the example of auxin responsive genes. We used the following published experimental

data on auxin induction of auxin responsive promoters with multiple AuxREs: (1)

di®erent DR5 reporters in Arabidopsis thaliana L.; (2) mutational analysis of

(a) (b)

(c) (d)

Fig. 5. The context and physicochemical features within �13 bp AuxRE neighbourhood found by theACTIVITY package12 as having the most signi¯cant impacts to auxin induction level of a gene.

(a) Correlation analysis for the Slide shift at [�12; þ12] region of the AuxRE. (b) Correlation analysis for

the distribution of the SnW trinucleotides around the AuxRE. (c) Correlation between the in silico prog-

nosis (Eq. (11)) and our estimates for the AuxRE activity in Ps-IAA4/5 (Eq. (4)). (d) Correlation betweenthe in silico prognosis (Eq. (11)) and our estimates for the AuxRE activity in the P3(4X) reporters.10

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-17

soybean GH3 promoter; (3) linker scanning mutations in Pisum sativum L. IAA4/5

promoter.

As a ¯rst approach, the data on promoter structure and activity a®ected by

genetic manipulations was formalized for all the promoter variants. Secondly, auxin

responsive promoter activity was described as a function on the impacts of individual

AuxREs. Di®erent promoter variants allow composing a system of equations per one

gene. As for the interaction of AuxREs in auxin responsive gene expression, two

alternative cases were considered, namely, additive and multiplicative e®ects.

By solution of the reverse problem for the systems of equation we found that the

multiplicative model of AuxRE cooperation ¯ts well to three independent experi-

ments on the genes from three di®erent plant species. The reverse problem solution

also suggests a repressive state of the promoters before auxin treatment. The pro-

posed method allows estimating relative contributions of the individual AuxREs that

are important for further analysis of AuxRE structure-activity relationships. Here we

demonstrated an example of such an analysis that allows hypothesizing additional to

TGTCnC consensus context and conformational AuxRE features.

Acknowledgments

This work was partially supported by The Dynasty Foundation grant for young

biologists, RFBR grants 11-04-01254-a, 11-04-01888-a and 12-04-33112, SS-

5278.2012.4, Integration SB RAS programs 80 and RAS Programs 6.8, B 26.29

and 28.

References

1. Paponov IA, Paponov M, Teale W, Menges M, Chakrabortee S, Murray JA, Palme K,Comprehensive transcriptome analysis of auxin responses in Arabidopsis. Mol Plant1:321�337, 2008.

2. Vernoux T, Brunoud G, Farcot E, Morin V, et al., The auxin signalling network trans-lates dynamic input into robust patterning at the shoot apex. Mol Syst Biol 7:508, 2011.

3. Hagen G, Guilfoyle T, Auxin-responsive gene expression: Genes, promoters and regula-tory factors. Plant Mol Biol 49:373�385, 2002.

4. Ballas N, Wong LM, Ke M, Theologis A, Two auxin-responsive domains interact posi-tively to induce expression of the early indoleacetic acid-inducible gene PS-IAA4/5. ProcNatl Acad Sci USA 92:3483�3487, 1995.

5. Liu ZB, Ulmasov T, Shi X, Hagen G, Guilfoyle TJ, Soybean GH3 promoter containsmultiple auxin-inducible elements. Plant Cell 6:645�657, 1994.

6. Ulmasov T, Murfett J, Hagen G, Guilfoyle TJ, Aux/IAA proteins repress expression ofreporter genes containing natural and highly active synthetic auxin response elements.Plant Cell 9:1963�1971, 1997.

7. Vernoux T, Besnard F, Traas J, Auxin at the shoot apical meristem. Cold Spring HarbPerspect Biol. 2(4): a001487, 2010.

8. Nakamura A, Higuchi K, Goda H, Fujiwara MT, et al., Brassinolide induces IAA5,IAA19, and DR5, a synthetic auxin response element in Arabidopsis, implying a crosstalkpoint of brassinosteroid and auxin signaling. Plant Physiol 133:1843�1853, 2003.

V. Mironova et al.

1340011-18

9. Guilfoyle T, Hagen G, Ulmasov T, Murfett J, How does auxin turn on genes? PlantPhysiol 118:341�347, 1998.

10. Ulmasov T, Hagen G, Guilfoyle TJ, ARF1, a transcription factor that binds to auxinresponse elements. Science 276:1865�1868, 1997.

11. Ulmasov T, Liu ZB, Hagen G, Guilfoyle TJ, Composite structure of auxin responseelements. Plant Cell 7:1611�1623, 1995.

12. Ponomarenko MP, Kolchanova AN, Kolchanov NA. Generating programs for predictingthe activity of functional sites. J Comput Biol 4:83�90, 1997.

13. Ponomarenko JV, Ponomarenko MP, Frolov AS, Vorobyev DG, Overton GC, KolchanovNA. Conformational and physicochemical DNA features speci¯c for transcription factorbinding sites. Bioinformatics 15:654�668, 1999.

14. IUPAC-IUB commission on biochemical nomenclature (CBN). Abbreviations and sym-bols for nucleic acids, polynucleotides and their constituents. J Mol Biol 55:299�310,1971.

15. Efron B. Bootstrap methods: Another look at the jackknife. Annals Statist 7:1�26, 1979.16. Zadeh LA. Fuzzi sets. Information Control 8:338�353, 1965.17. Fishburn PC. Utility Theory for Decision Making. John Wiley & Sons, New York, 1970.

Victoria Mironova received her Ph.D. in Bioinformtics and

mathematical modeling in 2010. She is currently Research Sta®

Scientist in the Laboratory for Molecular Genetic Systems of the

Department for Systems Biology, Institute of Cytology and Ge-

netics, Novosibirsk, Russia.

Nadya Omelyanchuk received her M.Sc. degree in Genetics

from Novosibirsk State University, Russia in 1970. She is currently

Senior Sta® Scientist in the Laboratory for Molecular Genetic

Systems of the Department for Systems Biology at the Institute of

Cytology and Genetics, Novosibirsk, Russia.

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-19

Maria Savina is a student of the Natural Sciences Faculty,

Novosibirsk State University, Russia. She is a laboratory assistant

in the Institute of Cytology and Genetics, Novosibirsk, Russia.

Petr Ponomarenko received his Bachelor's and Master's degrees

in Physics from Novosibirsk State University, Russia in 2008 and

in 2010, correspondently. He is a Ph.D. student in the Institute of

Cytology and Genetics, Novosibirsk, Russia.

Mikhail Ponomarenko received his M.Sc. degree in Physics

from Novosibirsk State University, Russia in 1985 and his Ph.D.

degree in Biology from Institute of Cytology and Genetics,

Novosibirsk, Russia in 1994. He is currently Senior Sta® Scientist

in the Department of System Biology at the Institute of Cytology

and Genetics, Novosibirsk, Russia. He is Laureate in Physics for

USSR National Graduate Students (1985), Belyaev's Prize Lau-

reate in Genetics (1995).

Vitaly A. Likhoshvai received his Ph.D. degree in Molecular

Biology from the Institute of Molecular Biology, Koltsovo, Russia,

in 1985, and Doctor of Sciences degree in Bioinformatics from the

Institute of Cytology and Genetics, Novosibirsk, Russia, in 2009.

Currently, he is a leading researcher of the Laboratory of Molec-

ular Genetic Systems of the IC&G.

V. Mironova et al.

1340011-20

Nikolay Kolchanov received his Ph.D., Dr.Sci. and Professor

degrees in Genetics from Institute of Cytology and Genetics,

Novosibirsk, Russia in 1975, 1989, 1992, respectively. He is cur-

rently academician of Russian Academy of Sciences, Director of

the Institute of Cytology and Genetics, Novosibirsk, Russia.

How Multiple Auxin Responsive Elements May Interact in Plant Promoters

1340011-21