2001 Wiley-Liss, Inc.
Cytometry 47:17 (2002)
DOI 10.1002/cyto.10030
Lack of Correlation Between AT Frequency and
Genome Size in Higher Plants and the Effect of
Nonrandomness of Base Sequences on Dye Binding
Martin Barow and Armin Meister
Institut fur Pflanzengenetik und Kulturpflanzenforschung (IPK), Department of Cytogenetics, Gatersleben, Germany
Received 11 December 2000; Revision Received 3 September 2001; Accepted 22 September 2001
Background: Different plant species vary as to the ratio
of nucleotide base pairs of genomic DNA. A correlation
between genome size and base pair ratio has been
claimed. Base composition can be analyzed by base-spe-
cific dyes.
Methods: Genome size is determined by flow cytometry
of suspensions of nuclei stained by the base independent
dye, PI. For estimation of the AT frequency, the AT-
specific dyes 4,6-diamidino-2-phenylindole, dihydrochlo-
ride (DAPI) and Hoechst 33342 (HO) were used. We
define a dye factor (DF) as the ratio of the two estimates
(peak ratios) of nuclear fluorescence intensities of sample
relative to reference plant nuclei using a given dye and an
intercalating fluorochrome.
Results: No significant correlation between genome size
and the DF for DAPI was found when 54 plant species
were investigated. However, similarities within and differ-
Different species do not vary as to their genomic DNA
content only. The ratio of the complementary nucleic
bases adenine and thymine (AT) and guanine and cytosine
(GC) also differs (1). In animals, there is a tendency of
larger genomes to have a higher GC frequency (2,3). A
similar relation seems to exist in higher plants (2), but this
assumption is based on the data of only six species, com-
piled from different sources (4,5).
According to Vinogradov (3), the positive correlation of
GC frequency and genome size may be explained by the
greater physical and chemical stability of large genomes
containing a relatively high GC frequency. It is possible to
check the claimed correlation between genome size and
AT frequency by means of flow cytometry after staining
with intercalating and base-specific fluorochromes, re-
spectively. Several papers report on base pair ratios of
some species using this method (6 8). However, to our
knowledge, no comprehensive investigation of base pair
ratios has been done for plant species in relation to their
DNA content and phylogenetic position.
Beside this general aim, the relation between AT-spe-
cific fluorescence and the AT frequency of the genome is
investigated. Dolezel et al. (6) showed that base-specific
ences among the plant families were shown. The compar-
ison of DAPI and HO DFs gave no consistent differences as
would be predicted from the model of different binding
site length of dyes. This result may be explained by the
nonrandom distribution of base pairs.
Conclusions: There is no general correlation between
genome size and AT/GC ratio in higher plants. Similar
AT/GC ratios within a plant family result from the general
similarity of the DNA sequences within a family. The
fluorescence of base-specific dyes is influenced by the
nonrandom distribution of bases in the DNA molecule.
Cytometry 47:17, 2002. 2001 Wiley-Liss, Inc.
Key terms: AT/GC ratio; DNA content; flow cytometry;
base-specific dyes; dye binding; Hoechst; PI; DAPI; angio-
sperms
fluorescence is not proportional to the base content, but
rather follows a curvilinear relation. Provided that the
bases in the genome are distributed randomly, a formula is
given for the relation between base frequency and fluo-
rescence intensity (8,9). It is based on the condition that
only a certain number, n, of consecutive bases of the same
type (AT or GC) is able to bind a dye molecule. For
Hoechst 33342 (HO), this number has been found to be
n 5 (8,9), whereas a range of n 3 4 is proposed for
4,6-diamidino-2-phenylindole, dihydrochloride (DAPI; 10).
Using this formula, a comparison between DAPI and HO
should give an unambiguous value of n for DAPI.
The results do not support this idea. A possible reason
is the deviation of base distribution from randomness.
Therefore, by evaluating known base sequences from sev-
eral organisms, the effect of nonrandom base distribution
on fluorescence intensity is determined and discussed.
Grant sponsors: Deutsche Forschungsgemeinschaft; Grant number: ME
1083/2-1; Grant sponsor: Fond der Chemischen Industrie.
*Correspondence to: Armin Meister, Institut fur Pflanzengenetik und
Kulturpflanzenforschung (IPK), D-06466 Gatersleben, Germany.
E-mail: meister@ipk-gatersleben.de
2 BAROW AND MEISTER

MATERIALS AND METHODS
Plant Material
Fifty-four species of 17 families have been studied (Ta-
ble 1), 6 of them for the first time with respect to DNA
content. The plants were cultivated in pots on garden
mulch in a greenhouse under standard conditions. In
some cases, especially for gymnosperms, Fagaceae and
Rosaceae, material was taken from plants growing out-
side. Young leaves were used for analysis. For gymno-
sperms, Fagus sylvatica, and the Rosaceae Physocarpus
opulifolius and Cydonia oblongaflower and leaf buds
were analyzed instead.
Some of the standards proposed by Dolezel et al. (11)
were used as references: Allium cepa, Glycine max, Pi-
sum sativum, Raphanus sativus, Secale cereale, and Vi-
cia faba. The calculation of reference values is described
in the Results and Discussion section.
of instrument settings, which might possibly influence the
results.
Statistical Analysis
The statistical analyses were performed using Statistica
for MacIntosh (Statsoft, Tulsa, OK).
RESULTS AND DISCUSSION
Estimation of Genome Size and AT Frequency
For the base-independent intercalating dye PI, the fluo-
rescence intensity F_PI is proportional to the genome size
2C with a proportionality constant k 1:
F_PIsample k1 2Csample (1)
The genome size of an unknown sample can be deter-
mined by comparing with a reference plant (11) as

Preparation of Nuclear Suspensions 2Csample 2Creference R_PIsample (2)

The analysis is based on the mean of four measurements
where R_PIsample is the ratio of sample and reference
of material from different individuals. For preparation of
fluorescence intensity with dye PI ( peak ratio in the
suspensions of nuclei, 30 100 mg tissue from each spe-
flow cytometric histogram):
cies was chopped with a razor blade together with mate-
rial from a reference plant in 1 ml ice-cold staining buffer
in a Petri dish according to Galbraith et al. (4) and filtered R_PIsample F_PIsample/F_PIreference (3)
through a 35-m mesh (Falcon 12 75 mm tube with a
For a base-specific dye such as AT-specific DAPI with a
35-m strainer cap).
frequency AT of AT bases of the genome, one would
The Galbraith buffer (4) was supplemented either
expect proportionality to the total number of the corre-
with 50 g/ml propidium iodide (PI; Molecular Probes,
sponding bases, i.e., an equation similar to eq. (1), but
Eugene, OR) 50 g/ml DNase-free RNase (Boehringer
with AT 2C instead of 2C. However, this seems not to be
Ingelheim Bioproducts Partnership, Heidelberg, Ger-
the case (6). Rather, a curvilinear relation exists in the
many), with 1 g/ml DAPI (Molecular Probes), or with
form
1 g/ml HO (Molecular Probes). In some cases, 0.51%
Triton (Sigma, Steinheim, Germany) for reducing adhe-
sion of cellular debris, 5% polyvinylpyrrolidone 25 F_DAPIsample k2 fATsample 2Csample (4)
(PVP; Serva, Heidelberg/New York) for binding phe-
where f(AT) represents the relation between the fre-
nolic compounds, and/or 50 100 mM potassium met-
quency of AT bases and the AT-specific fluorescence and
abisulfite (Riedel-de Haen, Seelze, Germany) as an anti-
k 2 is a proportionality constant. f(AT) can be considered
oxidant were added in order to obtain acceptable
the binding probability of a dye molecule to the DNA
coefficient of variation (CV) values. Different buffers
molecule with a given AT frequency. The ratio
result only in minor variation of peak ratios in experi-
R_DAPIsample between fluorescence of sample and refer-
ments with internal standardization (11).
ence is then analogous to eq. (3)
The analysis was done with a FACStarPLUS flow cytom-
eter (Becton Dickinson, San Jose, CA) equipped with two
argon lasers INNOVA 90-5 (Coherent, Palo Alto, CA) using fATsample 2Csample
R_DAPIsample (5)
the analysis program CellQuest. PI fluorescence was ex- fATreference 2Creference
cited with 500 mW at 514 nm and measured in the FL1
channel using a 630-nm band pass filter. DAPI and HO In order to obtain a parameter that characterizes the
fluorescence was excited with 200 mW in the UV range AT-specific fluorescence independently of the genome
and measured in the FL1 channel using a 450-nm band size, Zonneveld and Van Iren (12) used the PI peak ratio
pass filter. To reduce counts from fluorescent debris, gates relative to the DAPI peak ratio as an experimental param-
were set in the FL1/SSC dot plot. Usually 10,000 nuclei eter correlated with base composition. Similarly, we de-
within the gate were measured. fine a dye factor (DF) for a certain base-specific dye (in this
case, DAPI) in the following way:
Comparison of DAPI and HO
Seven combinations of species were measured after DF_DAPIsample R_DAPIsample/R_PIsample
staining with DAPI and HO. The corresponding measure-
ments were carried out on the same day to avoid a change fATsample/fATreference (6)
 CORRELATION OF AT FREQUENCY AND GENOME SIZE 3
Table 1
DNA Content, DAPI Factor, and AT Frequency of the Investigated Species*

Reference 2C DNA DAPI AT frequency

Family Code Species speciesb content (pg) factor (%)
Ginkgoaceae A Ginkgo biloba S.c. 21.6 1.20 65.3
Pinaceae B Abies concolora V.f. 36.1 0.97 60.9
B Larix decidua S.c. 25.7 0.96 60.7
B Picea abies V.f. 38.6 0.93 60.1
B Pinus sylvestris V.f. 44.2 0.97 60.9
Ranunculaceae C Anemone ranunculoidesa V.f. 36.8 0.91 59.7
C Anemone sylvestris V.f. 17.0 0.93 60.1
C Aquilegia vulgarisa R.s. 1.00 0.94 60.3
Chenopodiaceae D Atriplex rosea R.s. 2.11 1.00 61.5
D Beta vulgaris R.s. 1.84 1.06 62.7
D Spinacia oleracea R.s. 2.68 0.99 61.3
Urticaceae E Urtica dioica P.s. 2.33 1.01 61.7
E Urtica urens G.m. 1.07 1.08 63.1
Fagaceae F Castanea sativaa R.s. 1.95 1.06 62.7
F Fagus sylvatica G.m. 1.30 1.09 63.3
F Quercus robur G.m. 2.18 1.06 62.7
Rosaceae G Cydonia oblonga G.m. 1.98 0.89 59.2
G Duchesnea indica G.m. 4.18 0.96 60.7
G Physocarpus opulifolius R.s. 0.710 0.98 61.1
Fabaceae H Glycine max Std. 2.73 1.11 63.6
H Lathyrus articulatus P.s. 11.5 1.05 62.5
H Phaseolus vulgaris G.m. 1.57 1.05 62.5
H Pisum sativum Std. 9.07 1.00 61.5
H Trifolium pratense G.m. 1.06 1.17 64.8
H Vicia faba Std. 26.2 1.02 61.9
Brassicaceae I Alliaria petiolata R.s. 2.69 0.97 60.9
I Arabidopsis thaliana R.s. 0.424 0.97 60.9
I Brassica napus R.s. 2.94 0.98 61.1
I Raphanus sativus Std. 1.38 0.97 60.9
I Sinapis arvensis G.m. 1.35 0.96 60.7
Cucurbitaceae J Cucumis sativus R.s. 1.02 1.06 62.7
J Cucurbita moschata R.s. 0.968 0.99 61.3
J Cucurbita pepo G.m. 1.17 0.99 61.3
J Momordica charantia R.s. 1.42 1.10 63.5
Solanaceae K Capsicum frutescens P.s. 7.34 1.13 64.0
K Lycopersicon pimpinellifolium G.m. 2.28 1.07 62.9
K Nicotiana tabacum V.f. 9.77 1.02 61.9
Lamiaceae L Hyssopus officinalis R.s. 1.12 1.05 62.5
L Stachys grandifloraa P.s. 12.4 0.94 60.3
L Teucrium scorodonia R.s. 2.85 0.95 60.5
Asteraceae M Chrysanthemum multicolor V.f. 32.5 1.10 63.5
M Haplopappus gracilis G.m. 2.38 1.01 61.7
M Lactuca sativa G.m. 6.59 0.94 60.3
Asparagaceae N Asparagus officinalis R.s. 3.63 0.90 59.4
Alliaceae O Allium cepa Std. 33.7 1.20 65.3
O Allium ledebourianum P.s. 17.4 1.11 63.6
O Allium porrum A.c. 65.5 1.12 63.8
O Allium ursinum A.c. 62.1 1.17 64.8
Liliaceae P Fritillaria uva-vulpisa A.p./A.c. 165 1.15 64.4
Poaceae Q Hordeum vulgare S.c. 10.3 0.72 55.3
Q Oryza sativa R.s. 1.17 0.79 57.0
Q Secale cereale Std. 16.0 0.72 55.4
Q Triticum aestivum V.f. 32.8 0.67 54.1
Q Zea mays G.m. 5.90 0.62 52.8
*The genome size was calculated with Pisum sativum Viktoria, Kifejto Borso as the primary standard using the intercalating dye PI.
The AT frequency was calculated from the DAPI factor with Eqs. (6) and (7). For the primary standard Pisum sativum, a value of 2C
9.07 pg and an AT frequency of 61.5% were used (see Genome Size and AT Frequency of the Standards). The 2C values and AT
frequencies of the secondary standards were calculated on the base of the primary standard and the results obtained from Dolezel et al.
(11), Tables 4 and 7 and laboratory L1. The family-attached code is used for the identification of the families in Figure 1.
a
Species tested for the first time for DNA content.
b
The species used as secondary standards beside the primary standard Pisum sativum (P.s.): A.c., Allium cepa; A.p., Allium porrum;
G.m., Glycine max; R.s., Raphanus sativus; S.c., Secale cereale; V.f., Vicia faba; Std., standard itself.
4 BAROW AND MEISTER

which depends only on the AT-specific fluorescence and

is independent of the genome size, as can be seen from
the equation.
Because the actual value of DF_DAPIsample depends on
the reference used, we define P. sativum as a primary
standard that results in DF_DAPIPisum 1 because the
sample agrees with the reference. This value corresponds
to an AT frequency of 61.5% (see next section).
According to references (8) and (9), the relation f(AT)
has the form

fAT 1 AT ATn /1 ATn (7)

It is based on the following assumptions: (1) a certain

number, n, of consecutive AT bases is necessary for bind-
ing one dye molecule and (2) the AT bases are distributed
randomly within the DNA molecule. If n and the ATreference
are known and DF_DAPIsample is measured, the AT fre-
quency of the sample can be calculated in the following
way. According to eq. (6), f(ATsample) results in

fATsample DF_DAPIsample fATreference (8)

By means of a numeric approximation method (regula
falsi, carried out by a VisualBasic program), the AT fre-
quency can be calculated from f(ATsample) using eq. (7).
For the sake of simplicity, the above formalism was
established for DAPI. However, it is valid for any other
dye, including GC-specific dyes, analogously. Most authors
used these equations to calculate the AT/GC ratio from
their flow cytometric results. However, for various rea-
sons, eq. (7) may give incorrect results: (1) the AT bases
may be distributed nonrandomly and (2) the value of n is
not known exactly [e.g., 3 or 4 for DAPI according to
Portugal and Waring (10)].
The DF for DAPI, DF_DAPI may be a better parameter
for characterizing the AT frequency. It depends only on
the AT frequency itself and does not require any assump-
tion as to the fluorescence dependence on AT.
Genome Size and AT Frequency of the Standards
P. sativum Viktoria, Kifejto Borso was used as the
primary standard. The values for the genome size and AT
frequency are based on reference (6). For the genome
size, a value of 2C 9.07 pg is used. The 2C values of the
other references were calculated on the basis of this
primary standard and the results obtained by Dolezel et al.
(11), Table 4, laboratory L1.
The AT frequency of P. sativum was calculated from
eqs. (6) and (7) using a value of 59.5% for human leuko-
cytes (6,13) and an assumed binding site length for DAPI
of n 4. According to Table 1 in Dolezel et al. (6), the PI
peak ratio of Homo sapiens leukocytes relative to the
primary standard P. sativum is R_PIleuco 1/1.295
0.772. Analogously, the DAPI peak ratio is R_DAPIleuco
1/1.432 0.698. Therefore, the DAPI DF of human leu-
kocytes relative to P. sativum is DF_DAPIleuco 0.698/
0.772 0.904. By combining eqs. (6) and (7), the follow-
ing equation results:
FIG. 1. DAPI factor (correlated with the AT frequency) in relation to the
genome size. The letters indicate the family according to Table 1. No
correlation between genome size and DAPI factor is found, neither in the
total amount of data nor within the families (same letter). However,
significant differences exist among different families (see Table 2).
DF_DAPIleuco 1 ATleuco ATleuco
n
/1 ATleuco
n
/1 ATPisum ATPisum
n
/1 ATPisum
n
(9)
With the corresponding values for DF_DAPIleuco, ATleuco,
and n, it results in:
0.904 1 0.595 0.595 4/1 0.595 4
/1 ATPisum ATPisum
4
/1 ATPisum
4
(10)
From this equation, ATPisum is calculated by a numerical
method as described above as 0.615 or 61.5%. Based on
this primary standard, the AT frequency of the secondary
standards was calculated by eqs. (6) and (7) in a similar
way.
Correlation Between Genome Size
and AT Frequency
The investigated species, their genome sizes, DAPI fac-
tors, and AT frequencies (calculated from the DAPI factor
on the basis of eq. (7) with n DAPI 4) are listed in Table
1. The relationship between genome size and DAPI factor
is shown in Figure 1. No trend is discernible. Correspond-
ingly, the coefficient of correlation, r 0.19 (P 0.16),
is clearly below the significance level. It even has an
opposite sign when compared with the results of Vino-
gradov (2). Because enlargement of a genome should be
connected with an increased GC frequency, the DAPI
factor should be negatively correlated with the genome
 CORRELATION OF AT FREQUENCY AND GENOME SIZE 5
Table 2 family effect. It results also in a very small, nonsignificant
Differences Between the DAPI DFs coefficient of correlation (r 0.09).
of the Investigated Families*
Therefore, the high correlation that Vinogradov (2) ob-
Average DAPI factor Significance tained using data from literature is rather astonishing: only
Family with SD codeb six species were included; for three species, only a range
Poaceae 0.70 0.06 a of possible DNA contents was listed (4); for two of the
Asparagaceae 0.90 0a b three species (Zea mays and Lycopersicon esculentum)
Ranunculaceae 0.93 0.02 bc the DNA content varied more than twofold. If the latest
Rosaceae 0.94 0.05 bc
Pinaceae 0.96 0.02 bc
available values of genome size for these six species are
Brassicaceae 0.97 0.01 bc used instead (website http://www.rbgkew.org.uk/cval/
Lamiaceae 0.98 0.06 bc homepage.html), the coefficient of correlation will be dimin-
Chenopodiaceae 1.02 0.04 bcd ished to r 0.549 (P 0.26). These facts might be ex-
Asteraceae 1.02 0.08 bcd plained by a random constellation of data, which led to an
Cucurbitaceae 1.04 0.05 bcd
Urticaceae 1.05 0.05 bcd apparent significance within the limits of error probability.
Solanaceae 1.07 0.06 cd
Fagaceae 1.07 0.02 cde Differences of AT Frequency Among Plant Families
Fabaceae 1.07 0.06 cde Figure 1 shows obvious differences among the families.
Liliaceae 1.15 0a de
Alliaceae 1.15 0.04 de For example, all analyzed Poaceae (letter Q) have very low
Ginkgoaceae 1.20 0a e DAPI factors and, thus, a relatively small AT frequency,
whereas the corresponding values are clearly above the
*Significant differences appear among several families. Note
that Poaceae with the lowest DF (lowest AT frequency) signifi- average for Alliaceae (letter O). This finding has been
cantly differ from all other families. confirmed by analysis of variance with subsequent New-
a
Only one species was analyzed. man-Keuls-test.
b
Same letters indicate a lack of significant differences among In Table 2, the mean values and SDs of the DAPI factor
the corresponding families. Only families that do not have the
same letter are significantly different.
for each family are listed with an indication of significance
to characterize the differences among the families. De-
spite the small number of species studied within each
family, a large number of significant differences is found.
size. Hence, the positive correlation between genome size The existence of such differences is not surprising. Be-
and GC frequency with r 0.90 according to Vinogradov cause of the phylogenetic relations and the corresponding
could not be verified by our analysis of 54 plant species. similarities of the DNA sequences, similar AT frequencies
Using the AT values calculated on the basis of eq. (7) and, therefore, similar DAPI factors within a family and
instead of DAPI factors gives rather similar results (r corresponding differences between the families were to
0.18). This is not surprising, because the AT range in be expected. If the evolutionary increase of a genome is
spermatophyta is rather narrow and the DAPI factor and caused mainly by amplification of sequences already
AT frequency are highly correlated within this range. present in the genome and not by the addition to a large
Figure 1 also shows that within the families no trends are extent of sequences of deviating AT ratios, this could
detectable although species characterized by a wide range provide an explanation for the observed phenomenon.
of genome sizes have been included. This is confirmed by
the calculation of the pooled within-groups correlations Comparison Among DAPI and HO
that determines the correlations after subtracting the From the literature, no unequivocal value for the
mean DAPI factor for each family, so eliminating the binding site length n of DAPI is available. It is assumed

Table 3
Comparison of HO and DAPI DFs (Means of Four Independent Measurements SD)
DF
Species combination HO DAPI Difference n DAPIa
1. Glycine max/Lactuca sativa 1.307 (0.013) 1.160 (0.007) 0.147* 2.77 (0.15)
2. Pisum sativum/Secale cereale 1.463 (0.016) 1.416 (0.008) 0.047* 4.57 (0.15)
3. Allium ledebourianum/Secale cereale 1.770 (0.026) 1.557 (0.022) 0.213* 3.88 (0.16)
4. Raphanus sativus/Oryza sativa 1.278 (0.005) 1.249 (0.009) 0.029* 4.53 (0.16)
5. Allium cepa/Vicia faba 1.236 (0.011) 1.235 (0.004) 0.001 4.98 (0.22)
6. Glycine max/Raphanus sativus 1.297 (0.020) 1.190 (0.015) 0.107* 3.34 (0.31)
7. Vicia faba/Secale cereale 1.378 (0.061) 1.420 (0.010) 0.042 5.47 (0.76)
Difference DF_HO DF_DAPI

*Significant at P 0.05 according to the Mann-Whitney test.

a
n DAPI is calculated using eq. 11 (in parentheses: SD according to the law of error propagation). Contrary to the expectation, it varies
over a wide range.
6 BAROW AND MEISTER

Table 4
Relative AT-Specific Fluorescence (DF)*

Genome size DF for the given DF for random Difference relative to

Organism (Mbp) AT (%) sequence distribution random distribution (%)
Arabidopsis thaliana, chromosome 2 19.6 64.1 1.081 1.136 4.8
1.104 1.174 6.0
Arabidopsis thaliana, chromosome 4 17.5 64.0 0.848 1.128 24.8
0.792 1.163 31.9
Archaeoglobus fulgidis 2.2 51.4 0.545 0.568 4.1
0.474 0.490 3.3
Borrelia burgdorferi 1.5 71.8 1.657 1.591 4.2
1.923 1.793 7.2
Caenorhabditis elegans, chromosome 1 13.0 64.1 1.210 1.136 6.5
1.330 1.714 13.3
Caenorhabditis elegans, chromosome 2 14.9 63.8 1.179 1.120 5.2
1.287 1.152 11.7
Caenorhabditis elegans, chromosome 3 13.0 64.3 1.229 1.144 7.5
1.362 1.185 15.0
Caenorhabditis elegans, chromosome 4 16.8 65.3 1.253 1.198 4.5
1.384 1.255 10.3
Caenorhabditis elegans, chromosome 5 21.2 64.6 1.210 1.159 4.4
1.314 1.206 8.9
Caenorhabditis elegans, chromosome X 17.4 64.8 1.202 1.175 2.3
1.292 1.222 5.7
Chlamydia pneumoniae 1.2 59.4 0.864 0.899 3.9
0.829 0.872 4.9
Chlamydia trachomatis 1.1 59.7 0.910 0.914 0.4
0.867 0.894 3.0
Deinococcus radiodurans 3.3 33.4 0.144 0.132 8.8
0.102 0.075 35.7
Escherichia coli 4.6 49.2 0.599 0.494 21.3
0.560 0.404 38.7
Haemophilus influenzae 1.8 61.8 1.132 1.019 11.1
1.212 1.023 18.4
Helicobacter pylori 1.7 61.1 1.144 0.980 16.7
1.217 0.975 24.9
Homo sapiens, chromosome 21 33.8 59.1 0.899 0.883 1.8
0.926 0.856 8.2
Homo sapiens, chromosome 22 33.6 52.2 0.611 0.595 2.6
0.598 0.517 15.6
Methanococcus jannaschii 1.7 68.7 1.431 1.396 2.5
1.616 1.524 6.0
Mycobacterium tuberculosis 4.4 34.4 0.121 0.144 16.2
0.059 0.086 31.3
Mycoplasma genitalium 0.6 68.3 1.326 1.373 3.4
1.416 1.492 5.1
Neisseria meningitidis 2.3 48.5 0.630 0.467 35.0
0.576 0.382 50.7
Pseudomonas aeruginosa 6.3 33.4 0.097 0.132 26.5
0.054 0.075 28.6
Synechocystis sp. 3.6 52.3 0.731 0.599 22.1
0.732 0.522 40.2
Thermotoga maritima 1.9 53.8 0.513 0.653 21.4
0.425 0.587 27.5
Treponema pallidum 1.1 47.2 0.451 0.428 5.5
0.388 0.339 14.3
Vibrio cholerae 4.0 52.5 0.622 0.607 2.6
0.560 0.533 5.1
*Binding site lengths n 4 (first line, regular characters) and n 5 (second line, italic characters) for different organisms calculated
from the base sequence compared with a random base distribution of the same AT frequency. A DF of 1 corresponds to an AT frequency
of 61.5% with random distribution of AT base pairs (standard Pisum sativum). The base sequence of Pseudomonas aeruginosa was
downloaded from the website of the Pseudomonas Genome Project http://www.pseudomonas.com/, all remaining sequences from the
website of The Institute of Genomic Research (TIGR) http://www.tigr.org/

to be within the range of 3 4, whereas for HO n 5 is possible to calculate the unknown binding site length
found (10). According to eqs. (6) and (7), the DF for nDAPI from the known value nHO by the simplified
both dyes should be different. Moreover, it should be relation
 CORRELATION OF AT FREQUENCY AND GENOME SIZE
n DAPI nHO lnDF_DAPI/lnDF_HO (11) seria meningitidis and 31.9% for chromosome 4 of
which follows from eq. (10) in Godelle et al. (9).
In order to obtain reliable estimates from this equation,
the DF values should differ clearly from the value of 1;
otherwise, extremely imprecise values for nDAPI may result.
This can be reached in the following way. Pairs of species
were selected with DF values differing by at least 10%. The
DF values were not calculated relative to the primary stan-
dard P. sativum but relative to each other. The order of
species in the ratio is selected in such a way that DF 1.
Under this condition, the DF values are expected to be
higher for larger binding length n. This means, the DF values
should be always greater for HO compared with DAPI.
According to these conditions, seven pairs of species
were selected, the results for which are shown in Table 3.
With one exception (V. faba/S. cereale), all calculated
values for n DAPI are less than n HO 5. In the case of the
pair V. faba/S. cereale, the missing significance of differ-
ence and the high SD of n DAPI indicate that the values for
this combination do not really deviate from those for HO.
Nevertheless, the results of comparison of HO and DAPI
data are inconsistent. Two data pairs (V. faba/S. cereale
and A. cepa/V. faba) give an n DAPI value practically iden-
tical with n HO. The others vary between n DAPI 2.77 and
4.57. The mean of all calculated n DAPI is 4.22, so the
nearest integer value is 4.
What may be the reasons for the inconsistency of the
computed n DAPI values? If the assumption of a certain
number of consecutive base pairs of the same kind (AT in
this case) required for binding a dye molecule is correct,
the deviation may be explained by a nonrandom distribu-
tion of base pairs within the genome. The assumption of
random distribution of AT and GC base pairs as the con-
dition for eq. (7) may not always be fulfilled. This will be
the case especially for regions of repetitive sequences.
The extent to which such a nonrandomness of sequences
may cause the observed deviations can be investigated by
analyzing known sequences.
Effect of Nonrandomness of DNA Sequences on
Dye Binding
We have investigated 27 known sequences, published
in the internet (Table 4). The number of successive AT
base pairs necessary to bind one dye molecule is assumed
to be n 4 (corresponding to the most probable integer
value for DAPI as mentioned above) and n 5 (corre-
sponding to HO). The relative fluorescence intensities
were standardized in such way that a DF of 1 corresponds
to an AT frequency of 61.5% (the calculated AT frequency
of P. sativum) as used in Tables 1 and 2. The relative
fluorescence for a random distribution of AT base pairs was
calculated from the AT frequency using eq. (7) and the
relative fluorescence of the real distributions was calcu-
lated by scanning the sequences for groups of n 4 and
n 5 successive AT bases with the aid of a Visual Basic
computer program. The results are shown in Table 4.
Surprisingly, high differences in both directions result
from the calculations, up to extrema of 50.7% for Neis-
7
Arabidopsis thaliana (both with n 5) relative to the
expected random distribution. Moreover, chromosomes 2
and 4 of the sole representative of the plant kingdom,
A. thaliana, although having nearly the same AT fre-
quency (64.0 and 64.1%), differ clearly from each other
(4.8/6.0% and 24.8/31.9%, respectively, relative to
the random value).
In order to be sure that the calculation made by the
computer program is correct, it was tested by analyzing
pseudorandom sequences, which gave very good agree-
ment with the results expected from eq. (7). These calcu-
lations support the idea of nonrandomness of base distri-
bution as the reason for deviation from the expected
(random) values. Nevertheless, the correlation coeffi-
cients r 0.975 for n 4 and r 0.951 for n 5
between random and real distribution are highly signifi-
cant. This means that, in general, a good approximation of
AT content can be computed on the basis of the DAPI
factor, but that important deviations are possible in some
cases. This may explain the inconsistency between DAPI
and HO values, which are based on different binding site
lengths, but not the lack of correlation between AT con-
tent and genome size. Therefore, an existent correlation
might become diminished, but not extinguished.
ACKNOWLEDGMENTS
We thank Ingo Schubert, Jaroslav Dolezel and Paul
Fransz for helpful discussions, Barbara Hildebrandt for
technical support, Katrin Menzel and her staff for the
qualified cultivation of the plants, and the team of the IPK
germplasm collection for providing seed material.
LITERATURE CITED
1. Bonner J. The nucleus. In: Bonner J, Varner JE, editors. Plant bio-
chemistry, 3rd edition. New York: Academic Press; 1976. p 37 64.
2. Vinogradov AE. Measurement by flow cytometry of genomic AT/GC
ratio and genome size. Cytometry 1994;16:34 40.
3. Vinogradov AE. Genome size and GC-percent in vertebrates as deter-
mined by flow cytometry: the triangular relationship. Cytometry
1998;31:100 109.
4. Galbraith DW, Harkins KR, Maddox JM, Ayres NM, Sharma DP,
Firoozabady E. Rapid flow cytometric analysis of the cell cycle in
intact plant tissues. Science 1983;220:1049 1051.
5. Bennett MD, Smith JB. Nuclear DNA amounts in angiosperms. Philos
Trans R Soc Lond [Biol] 1976;274:227274.
6. Dolezel J, Sgorbati S, Lucretti S. Comparison of three DNA fluoro-
chromes for flow cytometric estimation of nuclear DNA content in
plants. Physiol Plantarum 1992;85:625 631.
7. Marie D, Brown SC. A cytometric exercise in plant DNA histograms
with 2C values for 70 species. Biol Cell 1993;78:4151.
8. Langlois RG, Carrano AV, Gray JW, Van Dilla MA. Cytochemical
studies of metaphase chromosomes by flow cytometry. Chromosoma
1980;77:229 251.
9. Godelle B, Cartier D, Marie D, Brown SC, Siljak-Yakovlev S. Heterochro-
matin study demonstrating the non-linearity of fluorimetry useful for
calculating genomic base composition. Cytometry 1993;14:618626.
10. Portugal J, Waring MJ. Assignment of DNA binding sites for 4,6-di-
amidino-2-phenylindol and bisbenzimide (Hoechst 33258). A compara-
tive footprinting study. Biochim Biophys Acta 1988;949:158168.
11. Dolezel J, Greilhuber J, Lucretti S, Meister A, Lysak MA, Nardi L,
Obermayer R. Plant genome size estimation by flow cytometry: inter-
laboratory comparison. Ann Bot 1998;82 (supplement A):1726.
12. Zonneveld BJM, Van Iren F. Flow cytometric analysis of DNA content
in Hosta reveals ploidy chimera. Euphytica 2000;111:105115.
13. Shapiro HS. Distribution of purines and pyrimidines in deoxyribonu-
cleic acids. In: Fasman GD, editor. Handbook of biochemistry and
molecular biology, volume 2. Cleveland: CRC Press; 1976. p 241281.