Cytometry 6:234-237 (1985)

0 1985 Alan R. Liss, Inc.

A Method to Estimate the DNA Content of Whole Nuclei

From Measurements Made on Thin Tissue Sections
M. Bins and F. Takens
Division of Hematology, Department of Medicine and Department of Mathematics,
University of Groningen, The Netherlands
Received for publication May 4, 1984; accepted December 12, 1984

A microdensitometry method that allows

estimation of the distribution of the DNA content of nuclei in thin tissue sections is described. The method is based on a theoretical
model of Feulgen-stained spherical nuclei, of
different sizes, in each of which the DNA is
present as a homogeneous solution. In thin
sections of nuclei of different sizes, the fraction of DNA per section is inversely proportional to the radius of the nucleus. Histograms of the product of DNA content and
The interpretation of DNA determinations in tissue
sections is complicated, since the amount of DNA per
slice of nucleus depends on the position of this slice
within the nucleus (7). This problem can be circumvented to some extent by taking thick sections (5-8 pm)
(6) and correcting for thickness (5),but in most tissues it
is impossible then to analyze single nuclei due to overlap. In thin sections the absence of overlap allows one to
perform DNA measurements on slices of single nuclei,
but then the relation between the amount of DNA per
slice and the amount of DNA per whole nucleus is still
unclear. A method that allows the relation of the DNA
in collections of nuclear sections to the DNA in whole
nuclei is described.

radius per nuclear section are independent of

nuclear size but depend on total DNA content.
The distribution of the total DNA content of
nuclei in a section can be estimated from such
a histogram. The results of the measurements
of a Feulgen-stained rat liver section are
described.
Key terms:
sitometry

DNA, Feulgen stain, microden-

Probability of the DNA Content of Slices of Nuclei

With Fixed Amount, N, of DNA and Variable Radius, R
Since all distances between planes and centres of
spheres are equally probable, the probability of a cut
through a cell a t a distance between r and (r f dr) is
P(r)dr = (1R)dr
If we reparametrize the same slice by its DNA content,
D, we have

or

P(D)
MATHEMATICAL BACKGROUND
For the calculations a n idealized tissue model is used;
it is composed of cells with spherical nuclei homogeneously filled with a DNA solution and cut in infinitesimally thin sections.
DNA Content of Sections of Spherical Nuclei
For a plane at a distance, r < R, from the centre of
the sphere with radius, R, the intersection of plane and
sphere is a circle with radius
and hence with
a surface area A(r) = a (R2 - r2).When N is the total
amount of DNA, then the DNA content of a slice of a
(small) thickness, t, taken a t a distance, r, from the
centre of the sphere will be:
D = (volume density of DNA) x (volume of slice)
N
* a ( R 2 - r2)t
413 a R3
3Nt
- -.
(R2 - r2)

m,

4R3

dr
dD

= -

1
R

. P(r) = - .

l%l'

Solving for r and computing dr/dD, we have:

and
dr _
_ -

dD

2R2

3Nt.

JT'

Address reprint requests to M. Bins, Department of Medicine, University Hospital, P.O. Box 30.001, 9700RB Groninpen. The
Netherlands.

235

DNA CONTENTS OF NUCLEI FROM THIN TISSUE SECTIONS

so

The resulting curves for radii R, 2R, and 3R are given

in Figure 1.The location of these curves depends on R.
Probability of the DNA Content of Sections of Nuclei
With Variable Radius and Nuclear Content
In order to obtain a quantity from which the probability is independent of the radius, R, in spheres with equal
DNA content, we consider the product, Q, of DNA content and radius of the slice.

(volume density of DNA) x (volume of slice) x r

01

0.33

1.0

0.50
DNA

FIG.1. Probability piD) of the DNA contents of infinitesimally thin

sections of sphere-shaped nuclei with radii R, 2R, and 3R, containing
equal total amounts of DNA.

3Nt
- __
4R3

. (R2

3
-

.
(N)

As above, we can reparametrize the slice in terms of

Q and set:

or

0 1

Solving for r and computing dr/dQ, we have:

0.5

1.0
D N A x radius

FIG.2. Probability p(Q) of DNA x radius of infinitesimally thin sections of spheres of different sizes containing equal total amounts (N)
of
DNA, and twice this amount (2N). In practice the probability a t low Q
values is different due to the finite thickness of the sections.

dr
dQ

. (A)
.'
Q-i,
3Nt

containingtwice the amount of DNA yield a distribution

of the product of DNA and radius per nuclear section
distinct from the distributionof spheres with single DNA
contents, since similar slices will contain twice the
amount of DNA. The distribution is peaked twice as far
from the origin.

EXAMPLE
To demonstrate that this theory is valid in tissue sections, we measured a 2 pm section of rat liver. The
This function is independent of R. It has poles at Q = 0 integrated light absorption and the surface area of Feuland Q = 3Ntl4, which means that the latter pole de- gen-stained nuclei (4) were measured using a Texture
pends linearly on N, the total amount of DNA of the Analysis System (Leitz, Wetzlar) which was found to
have a coefficient of variation for DNA determinations
sphere. The minimum is a t about 0.35 Qmax.
The behaviour of this curve is shown in Figure 2 for a of Feulgen-stained monoclonal lymphoid cells in a blood
fixed amount of nuclear DNA (N) and twice this amount smear of 2.6% (2). In 388 nuclei the integrated light
(2N). As is indicated above, spheres of undeterminedsize absorbance (DNA in Arbitrary Units) multiplied by the

236

BINS AND TAKENS

square root of the nuclear area (as a substitute for the

radius) was calculated. The measurements were performed using a 100 x oil immersion objective, and light
from a 12 V 100 W halogen light source filtered by a
wide-band pass filter centered at 548 nm. The resulting
histogram is shown in Figure 3a. Distinct regions originating from diploid ( 1,000) and tetraploid ( - 2,000)
nuclei are present, and possibly some measurements are
from octoploid ( - 4,000) nuclei. Although a n overlapping region is present between the peaks, a n approximation of the fractions of diploid and tetraploid nuclei
can be made.
The influence of the correction by the square root of
the area is shown in Figure 3b, representing the distribution of the uncorrected integrated absorbance values.
Figure 3c represents the distribution of the nuclear
areas. Although both Figures 3b and 3c show a bimodal
distribution, a linear arrangement of the distribution
adequate for interpretation is present only in Figure 3a.

n-

30-

EI

DNA x V E F C ( A U )

lo-

--

"j b
DISCUSSION
3 @,
When thin sections are made through spherical nuclei
of proliferating tissue, the distribution of the DNA content of the sections still reflects the distribution of the
nuclear DNA of the tissue, but the distribution histo- 20gram is obscured by the fact that its shape is influenced
by both the amount of nuclear DNA and the nuclear
D N A (AU)
volume. By multiplying the amount of DNA per nuclear
section by the radius of this section (or the square root 10of its surface area), a distribution histogram, which depends only on nuclear DNA and is independent from
nuclear volume, is obtained. This method allows one to
4,
1'
approximate the DNA distribution in thin tissue sec- 0
3500
tions or in selected cell classes within tissue sections.
The method clearly has its limitations, inherent in the
measuring of slices of nuclei instead of whole nuclei and " i
in the deviations, which occur in practice, from the
spherical shape assumed in the theoretical model. The
influence of elliptical, instead of spherical, nuclei on the Lo:
distribution of nuclear DNA was discussed by McCready
and Papadimitriou (6).Larger deviations from the spherical shape will cause results for which a n interpretation
will be very difficult or impossible, e.g., in spindleshaped, twisted, or lobulated nuclei.
In practice the theoretical shape of Figure 2 is modified by the thickness of the section, mainly at low values
of Q. When the section is through the outer border of
the nucleus and the slice of nucleus is smaller than the
thickness of the section (i.e., r < R < r + t), the shape
of the curve changes. This implies that, for Q values
obtained in sections where r < R < r + t, the interpretation of the curve becomes unreliable. For a 15relation
between thickness, t, and radius, R, the breaking point
of the curve can be calculated as 0.21 of the maximal Q
value; for a 1:3 relation the breaking point will be 0.41.
This should be kept in mind for the interpretation of the
FIG.3. Rat liver in a 2 prn section. a) Distribution of DNA x Jarea.
histograms obtained. Other practical deviations from b) Distribution of DNA without correction for G.c) Distribution of
the theoretical curve are caused by overestimation of nuclear area.

LldI JI Ill1I1IIII
c

DNA CONTENTS OF NUCLEI FROM THIN TISSUE SECTIONS

the area of cone-shaped nuclear sections, and by decreased accuracy in delineating area where poor contrast is present.
For DNA determinations flow cytometry, when applicable, has evident advantages over image analysis
methods. However, the loss of material during the preparation of a suspension from solid tissue may selectively
affect a cell class of interest whereas in tissue sections
no material is lost. Also, flow cytometry cannot possibly
give information about the relation between histological
appearance and DNA content. This can be provided by
image analysis methods; but DNA measurements in
tissue sections have the inherent problem that the relation between DNA per section and total nuclear DNA is
unclear. Nonetheless, DNA determinations in tissue sections yielded relevant correlations with clinical data in
renal adenomas and adenocarcinomas (1). By correcting
DNA measurements in nuclear sections for nuclear size
according to our method, the relation between DNA per
section and per nucleus is made more clear. This may
provide valuable information about the distribution of
nuclear DNA in tissue sections in relation to their microscopic appearance.

237

Preliminary results indicate that, in lymph nodes, the

microscopic appearance of sections of nuclei appears to
be related to the amount of nuclear DNA, a result that
could not be obtained by measurements on a suspension
of cells (3).

LITERATURE CITED
1. Bennington JL,Mayall BH. DNA cytometry on four micrometer
sections of paraffin embedded human renal adenocarcinomas and
adenomas. Cytometry 4:31-39, 1983.
2. Bins M. Feulgen densitometry in image analysis (letter). J Histochem Cytochern, in press.
3. Bins M, Poppema S DNA determinations in thin sections of lymphomas (abstract). Symposium on Morphometry in Morphological
Diagnosis, Delft, The Netherlands, 1984.
4. Duijndam WAL, van Duijn P. The dependence of the absorbance of
the final chromophore formed in the Feulgen-Schiff reaction on the
pH of the medium. Histochemie 35:373-375, 1973.
5. Kreicbergs A, Zetterherg A. Cytophotometric DNA measurements
of chondrosarcoma: Methodological aspects of measurements in tissue sections from old paraffin embedded specimens. Anal Quant
Cytol 284-92, 1980.
6. McCready R, Papadimitriou JM: An analysis of DNA cytophotometry on tissue sections in a rat liver model. Anal Quant Cytol 5:117123, 1983.
7. Weihel ER. Stereological methods. Vol. 2. Academic Press, London,
1980.