Vol. 6, 849-855, October 1997 Cancer Epidemiology, Biomarkers & Prevention 849

DNA Image Cytometric Measurement as a Surrogate End Point

Biomarker in a Phase I Trial of a-Difluoromethylornithine

for Cervical Intraepithelial I

Iouri V. Boiko, Michele Follen Mitchell,2Dibip K. Pandey, R. Allen White, Wei Hu,Anais Malpica,3 Kenji Nishioka,4 Charles W. Boone,E. Neeby Atkinson, and Walter N. Hittelman

Departments of Clinical Investigation (I. V. B.. W. H., W. N. H.], Gynecologic

Oncology [M. F. M., D. K. P.], Biomathematics [R. A. W., E. N. Al, Pathology

IA. M.], and Surgical Oncology [K. NI, The University of Texas M. D.Anderson Cancer Center. Houston, Texas, 77030; and Chemoprevention

Branch, National Cancer Institute, Division of Cancer Prevention and Control.

Bethesda, Maryland 20892 [C. W. B.l

Abstract

Cervical intraepithelial neoplasia grade 3 (CIN 3) isconsidered a high-risk precursor of invasive cervicalcancer. a-Difluoromethybornithine (DFMO) is apromising antiproliferative chemopreventive agent. Thepurpose of this study was to evaluate image cytometricmeasurement of nuclear DNA (1CM-DNA) as a surrogateend point biomarker (SEB) in a Phase I trial of DFMOfor CIN. Thirty patients with CIN 3 were treated withDFMO at five doses, ranging from 0.0625 to 1.0 g/m2/day,for 1 month. Half of the patients had histologicalresponses. Twenty-five pre- and posttreatment cervicalbiopsy specimens (from 11 responders and 14nonresponders) were available for this analysis. 1CM-DNA was performed on 4-�.tm sections cut from formalin-fixed tissue blocks and stained with a thionin-S02Feulgen reaction. ICM-DNAs for each case wereexpressed as normalized measurements (against thenuclear modal absorbance of lymphocytes) of theabsorbance of each cell of interest and were presented inbar histograms. The mean normalized summedabsorbance (�ODn) was obtained as a mean histogram ofthe cell population of interest. Nineteen (76%) of 25

patients had a significant decrease in �OD� after DFMOtreatment. Posttreatment values were significantly bower

than pretreatment values in a paired analysis, andresponders had significantly lower values thannonresponders. Analyses of different 1CM-DNA

Received 12/10/96; accepted 6/27/97.

The costs of publication of this article were defrayed in part by the payment of

page charges. This article must therefore be hereby marked advertisement in

accordance with I 8 U.S.C. Section 1734 solely to indicate this fact.

‘ This work was supported by National Cancer Institute Contract CN-25433-02.

2 To whom requests for reprints should be addressed, at Department of Gyneco-

logic Oncology, Box 67, The University ofTexas M. D. Anderson Cancer Center,

1515 Holcombe Boulevard, Houston. TX 77030. Phone: (713) 745-2564: Fax:(713) 792-7586.

3 Present address: Brown and Associates Medical Laboratories. 8076 Del Rio.

Houston. TX 77054.

4 Present address: 7820 Kendalla Drive, Houston, TX 77036.

references, including percentile values of �OD�distribution, DNA malignancy grade, and 5c exceedingrate, showed a decrease of mean �OD� during DFMOtreatment. In addition, the summed posttreatment �OD�histograms also showed progressively shorter rightshoulders compared with pretreatment histograms inboth responders and nonresponders. We concluded that

the modulation of �OD� reflected the chemopreventioneffect of DFMO even before morphological changesappeared, and thus, 1CM-DNA may be useful as a SEB inchemoprevention trials of DFMO. Additional reasons forusing 1CM-DNA as a SEB are the relative simplicity ofits use, the high accuracy of the results, the low cost of

the reagents, the ability to use small tissue samples, andthe objectivity and reproducibility of the procedure.

Introduction

Cervical cancer remains an important health problem; it is thesecond most common cancer in women worldwide ( I ). CIN 3,�

which includes severe dysplasia and carcinoma in situ. is ahigh-risk precursor of invasive cervical cancer and is thus

suitable for the study of chemoprevention of cancer (2, 3). Onechemopreventive agent that is promising in the cervix is

DFMO, a potent antiproliferative agent that was used in a PhaseI chemoprevention trial in CIN 3 (3). DFMO will soon be testedin a Phase II placebo-controlled trial in CIN.

DFMO, a specific inhibitor of ornithine decarboxylase,inhibits polyamine synthesis. Despite extensive research ef-forts, the precise function of polyamines in cellular physiology

is not yet well understood. Basic pobyamines are capable ofnoncovalent interactions with nucleic acids and proteins (4-6),and these interactions may stabilize DNA structure and protect

DNA from nuclease digestion (5, 7), alter sequence-specificDNA-protein interactions, change the regulatory process in the

chromatin of nuclei (8, 9), and affect the attachment of DNA to

the nuclear matrix, possibly affecting replication and transcrip-

tion (5, 10, Il). Because DFMO inhibits ornithine decarboxy-base activity and pobyamine synthesis, it can inhibit or delayDNA synthesis both in vitro and in vivo ( 1 2, 1 3). This processhas been associated with decreased cell proliferation (12, 14),loss of the mitogenic response to growth factor stimulation( I 5), and reduction of proto-oncogene (c-fos and c-tnvc) ex-

pression (16).Here, we investigated two hypotheses. The first was that

DFMO decreases DNA synthesis or total DNA content in CIN

S The abbreviations used are: CIN 3, cervical intraepithelial neoplasia grade 3;

DFMO. a-difluoromethylomithine: SEB. surrogate end point biomarker: 1CM-

DNA, image cytometric measurement of nuclear DNA: � normalized

summed absorbance: DNA-MG, DNA malignancy grade: 5cER. Sc exceeding

rate: df, degrees of freedom.

on February 26, 2021. © 1997 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

850 DNA Image Cytometry in a Cervix Chemoprevention Trial

3 lesions. The second involved a potential SEB for evaluatingthe effects of DFMO.

The end point of interest in chemoprevention studies iscancer incidence reduction ( I 7). However, this end point re-

quires years of follow-up, and such studies are expensive. SEBsare intermediate markers of carcinogenesis that allow trials to

be of shorter duration, to require fewer subjects, to be lower in

cost, to use small tissue samples, and to aid in beaming moreabout the carcinogenic process ( I 7).

For several decades, flow cytometric measurement of nu-clear DNA and 1CM-DNA have been used for research and

diagnostic purposes as biomarkers of proliferation and neopbas-tic transformation (1 8-24). For example, an increase in nuclearDNA during the progression of CIN lesions has been shown

(21 , 22). 1CM-DNA has also been used for monitoring doseeffectiveness in some chemotherapeutic studies (23, 24).

Recently, 1CM-DNA has been proposed for purposes ofSEB development because it uses an intact tissue sample, and

thus, important information about tissue architecture and thedegree of neoplastic progression is retained (17). Despite rel-

ative limitations of 1CM-DNA in tissue sections (25, 26), well-standardized analysis (27) of nuclear summed absorbance canstill reflect the relative total amount of cell DNA. Thus, oursecond hypothesis was that DFMO decreases DNA synthesisand/or total DNA content in CIN 3 lesions and that 1CM-DNA

measures that reflect DFMO’s effect on DNA content mayserve as SEBs of the effects of chemoprevention. To test ourhypotheses, we used 1CM-DNA to examine pre- and posttreat-ment biopsies from a group of CIN 3 patients during a chemo-prevention trial of DFMO.

Materials and Methods

Patient Groups and Tissue Sampling. Thirty patients with

CIN 3 involving over one-third of the surface area of the cervix(i.e., areas - I .0-1 .5 cm in diameter) were treated in a Phase I

chemoprevention trial. Six patients were assigned to each offive dose levels of DFMO, ranging from 0.0625 to 1.0 g/m2/day, for I month. Pretreatment colposcopicalby directed biopsysamples (measuring 1 X 2 X 2 mm) were compared with

posttreatment colposcopicalby directed cone biopsy samples(measuring 3 cm in diameter at the cone base and 2 cm inheight). Fifteen (50%) of these patients had histological re-sponses after treatment, of which 5 were classified as complete

responses (a posttreatment diagnosis of metaplasia or reactiveand granulation tissue) and 10 were evaluated as partial re-sponses [CIN lesions of bower grade (i.e. , 1 or 2) than before

treatmentj. Samples from 25 of the 30 patients were availablefor 1CM-DNA analysis because of biopsy size and the need foradequate field size.

Preparation of Materials. The tissue blocks were randomlyordered, and the same microtome was used to cut all of thespecimens at the same time. Serial 4-tam-thick tissue sectionswere stained with H&E for light microscopic evaluation, andother sections were stained for the Feulgen reaction using theprotocol of Xibbix Technologies (Vancouver, British Columbia,

Canada; Ref. 28). All slides were stained at the same time,under the same conditions. To increase the accuracy of thisstudy, paired pre- and posttreatment samples from each patientwere cut and treated under the same conditions, at the same

time. Briefly, following deparaffinization, the samples wererehydrated and postfixed in Boehm-Sprenger fixative. After 45

mm of acid hydrolysis (5 N HCI; 23#{176}C),the sections werestained with thionin-SO2, washed, dehydrated, cleared in xy-lene, and mounted for image analysis. The pre- and posttreat-

ment H&E-stained slides were reviewed to identify the patho-logical areas of interest, including dysplastic and reactive areas

in the epithelium, by two pathologists (A. M. and I. V. B.) and

used as templates to map Feulgen-stained slides.

Image Analysis. At The University of Texas M. D. Anderson

Cancer Center (Houston, TX), the CytoSavant computer-

assisted image analysis system (Oncometrics Imaging Co.,

Vancouver, British Columbia, Canada) was used for 1CM-DNA

analysis (29). This system has a scientific-grade charge-coupled

device, which is able to give a small pixel size, 100% fill factor,high quantum efficiency, low readout noise, wide dynamic

range, good linearity, and geometric stability (Xillix Technol-ogies). The charge-coupled device transducer is positioned in

the primary image plane of a Nikon 20/0.75 Plan Apo objective.

This arrangement results in the acquisition of a chromatically

and geometrically correct image with square pixels. On aver-

age, each nucleus is covered by over 500 pixels. The device

uses algorithms for the automated detection of the nuclearboundaries along the highest local gradient between the nuclear

stain and unstained cytoplasmic background, which defines the

nuclear boundary in a precise, objective, and reproducible way.

Cell Selection. Lymphocytes were used as internal standardcontrols to normalize each slide and to correct for staining

variations. From each tissue section, nuclear images of 50 ± 15

(mean ± SD) lymphocytes and 130 ± 40 epithelial nuclei frompreviously mapped CIN 3 areas were collected by a pathologist

(I. V. B.) in a semi-interactive procedure and stored in thecomputer memory. The collection of nuclei in interactive mode

helped to control selection bias (only nonoverlapping nuclei

with easily detected boundaries and without “capping” were

chosen). Additional control of cell quality was performed byexamination of a computer-stored cell gallery, which allowed a

magnified display of the stored image of each nucleus. In this

way, the scored cells could be reviewed, and fragmented cellswere eliminated from the analysis. The number of selected

nuclei depended on how much of the pathological area of

interest was on the analyzed slide. The coefficient of variation

of �OD of lymphocytes was 5%.

Epithelial cells (about 100 per sample) from normal-ap-

pearing regions that were adjacent to CIN lesions were also

analyzed in available samples to determine the mean �OD ofnormal epithelial cells in relation to the modal �OD of lym-

phocytes. A correction factor of 1 . 1 5 was used to account for

the down-shift in the �OD distribution due to the difference inchromatin compaction in lymphocytes and epithelial cells. Af-

ter the correction factor was applied, the modal �OD of lym-

phocytes was summed as a normalized unit of measurement

(which corresponded to the unit “2c,” which has been used in

previous studies) and linearly extended to establish the normal-

ized �OD scale.

Interpretation of DNA Histograms. ICM-DNAs for eachcase were expressed as normalized measurements (against thenuclear modal absorbance of lymphocytes) of the �OD of each

cell of interest and were presented in bar histograms. The

following data were obtained and used as 1CM-DNA refer-

ences: (a) �ODn for each case, obtained as a mean �ODn

histogram of the cell population of interest (30); (b) DNA-MG,

based on the “2c deviation index” and presented as a continuousscale ranging from 0.01 to 3.0 (3 1 ); (c) 5cER, defined as the

percentage of cells having a “DNA content” of more than Sc

(32); and (d) percentile value of �ODn distribution, defined asrates of �ODn at the 5th, 25th, 50th, 75th, and 95th percentile

points of the �OD� histogram.

on February 26, 2021. © 1997 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

Cancer Epidemiology. Biomarkers & Prevention 851

Statistical Analysis. A nonparametric (Wilcoxon) test of sig-nificance was performed to assess the differences between

�ODn in pre- and post-DFMO treatment CIN 3 samples. The

SPSS statistical software was used for analysis.The modulation of �ODn during DFMO treatment was

analyzed also using a mixed effect linear model. Specifically,we assumed

Yjko+�tXt+13�2+13�X3+1.Li+Y�,+#{128},jk,

where: (a) Y�3k �5 the logarithm of the �OD� for patient i, group

j, and replication k; (b) group 1 is composed of pretreatmentsamples from nonresponders; group 2 is composed of pretreat-ment samples from responders; group 3 is composed of post-

treatment samples from nonresponders; and group 4 is com-posed of posttreatment samples from responders; (c) x1 is 0 for

pretreatment data and 1 for posttreatment data; (d) x2 is 0 for

data from nonresponders and 1 for responders; (e) �. whichrepresents the random effect for patient i, is normally distrib-uted with mean 0 and variance o�2; (j) �y which represents therandom effect of groupj within patient i, is normally distributedwith mean 0 and variance a’,2; and (g) �.jk is normally distrib-

uted with mean 0 and variance o�.Thus, the patient is considered a random factor, response/

nonresponse is a between-patient factor with both fixed andrandom components, and pretreatment/posttreatment is a with-in-patient factor with both fixed and random components. The

design is unbalanced with unequal replicates. The variance ofcase means around the group mean is permitted to differ amongthe four groups. The model was fit using the lme function of

S-PLUS, version 3.4. All significanttests were performed using

the likelihood ratio test.

Results

Data for patient ages, doses of DFMO, �OD� of pre- andposttreatment samples, and histological response are presentedin Table 1 . The mean age was the same (29 years) in responders

and nonresponders. Nineteen (76%) of the 25 patients studiedshowed a decrease in �ODn after DFMO treatment. Threepatients had no �ODn changes, and three showed an increase of

�ODn after treatment. Although no significant dose-dependentresponse in �ODn was detected as a fraction of dose, perhaps

due to the small number of cases, all patients who had increasesof �ODn following treatment had received bow (0.062�-0.12S

g/m2/day) doses of DFMO.According to histological diagnosis, there were 1 1 re-

sponders and 14 nonresponders among the 25 patients who hadevaluabbe pre- and posttreatment samples available for the

study. Only 1 of 1 1 responders had no �OD� change afterDFMO treatment (all other responders had decreases in �ODn

5 of 14 nonresponders had either no change or an increase inposttreatment �ODn).

The group mean values of �OD� according to histologicalresponse and treatment category are shown in Table 2. There

was a trend for lower pretreatment �ODn in responders than innonresponders; however, this trend was not significant. Al-

though both the responders and nonresponders showed an over-

all significant decrease in �ODn after DFMO treatment, theresponders’ values dropped more than the nonresponders’ val-ues, which resulted in a significant difference between �ODn

values for responders and nonresponders in posttreatment sam-pies ( I .25 versus 1 .5, respectively).

The values obtained from DNA-MG analysis are pre-sented in Table 3. DNA-MG, which is based on a 2c deviationindex distribution, is defined as the sum of the squares of the

Table I Distribution of age, pre- and posttreatment �OD� and histologicalresponse by DFMO dose level

Dose level Age (�OD�) Histological

response(g/m2/day) (yr) Pretreatment Posttreatment

1.00 41 1.19 1.03 PR”

30 1.72 1.05 PR

23 1.51 1.38 NR

24 1.53 1.42 NR

35 1.94 1.85 NR

0.50 27 1.27 1.16 PR

40 2.11 1.41 NR

28 2.01 1.19 NR

20 1.34 1.31 NR

0.25 23 1.19 1.18 PR

29 1.48 1.11 CR

29 1.63 1.26 NR

25 1.60 1.28 NR

0.125 39 1.75 1.81 NR

31 1.30 1.55 NR

40 1.75 1.11 PR

25 2.21 1.14 CR

26 1.37 1.07 CR

22 1.33 1.19 PR

0.06 22 1.69 1.33 NR

23 1.38 1.41 NR

36 1.88 1.76 NR

26 1.43 1.51 NR

22 1.46 1.13 CR

40 1.30 1.20 CR

Mean ± SE 1 .58 ± 0.06” 1 .3 1 ± 0.05”

“ PR, partial response; NR, no response; CR, complete response.

“P < 0.01.

Table 2 Mean 10 D� by DFM 0 treatment and h istological respon. cc category

Histologicalresponse

No. ofPatients

Mean

(�OD,)

pretreatment

Mean

(�OD�)

posttreatment

Nonresponders 14 1.79 1.50 0.04

Responders 10 1.69 1.25 <0.01

P 0.52 0.02

Table 3 DNA-MG values by DFMO treatment and histologi cal response

category

Histological

response

DNA-MG P

Pretreatment Posttreatment

Responders 2.50 2.05 <0.01

Nonresponders 2.63 2.43 0.04

Total 2.58 2.26 <0.01

differences between the �OD� of single cells and the 2c value,divided by the number of measured cells (31, 32). In our study,2c was evaluated as the modal �OD of lymphocytes multiplied

by a correction factor. Mean DNA-MG in this study was

significantly lower in posttreatment than in pretreatment sam-pies for both responders (P < 0.01 ) and nonresponders (P <

0.04).Values for 5cER, the number of cells with �ODn exceed-

ing Sc, were also lower in posttreatment samples than in pre-treatment samples (Table 4). There was a posttreatment de-crease in both responders and nonresponders.

on February 26, 2021. © 1997 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

852 DNA Image Cytometry in a Cervix Chemoprevention Trial

Table 4 SeER by DFMO treatment and histological response category

Histological

response

No. of

pattents

No. of patients with SeER Cells Mean 5 cER (%)

Pretreatment Posttreatment Pretreatment Posttreatment

Nonresponders I I 9 2 5.5 0.3 0.03

Responders 14 13 9 9.4 4.6 0.05

Total 25 22 11 8.3 3.1 0.01

Table 5 Percentile values of �OD,, distribution by DFMO treatm

histological response category

ent and

Treatment and Percentile of �OD�

histological response 5th 25th Sorb 75th 95th

Responders

Pretreatment 0.96 1.25 1.49 2.01 2.92

Posttreatment 0.80 1 .03 1.21 1.40 1.82

Nonresponders

Pretreatment 0.97 1.29 1.62 2.16 3.04

Posttreatment 0.80 1.13 1.37 1.76 2.64

Post/Pretreatment ratio

Nonresponders t).83 0.88 0.85 0.81 0.87

Responders 0.83 0.82 0.81 0.70 0.62

After DFMO treatment, both responders and nonre-sponders had a left shift in summed posttreatment �ODn his-

tograms compared with pretreatment histograms. However, therates of this shift were different and resulted in a prominent

shape difference between the posttreatment �ODn histogramsof responders and nonresponders. This difference can be ex-

plained by changes in individual �OD� histograms after DFMOtreatment. Ten of 1 1 responders but only 9 of 14 nonrespondershad a posttreatment decrease of � So, the summed post-treatment �OD� histogram of nonresponders included two

components. one incorporating histograms of patients who hada decrease (as did almost all of the responders) and one incor-porating histograms of patients who had no change or an

increase. The 5th, 25th, 50th, 75th, and 95th percentile points of

the �ODn distribution histograms give a quantitative descrip-tion of the data presented in the �ODn histograms (Table 5).

To show that the distribution of slide thickness was ran-dom and did not affect �ODn measurements, we checked area

size and �OD of lymphocytes in pre- and posttreatment sam-pIes. The chromatin structure features of lymphocytes were also

compared to prove that DFMO did not affect chromatin struc-ture and that �ODn changes were, therefore, not an aftereffectof acid hydrolysis from the Feulgen reaction. The data for meanarea, SOD, and chromatin structure of pre- and posttreatment

lymphocytes show there was no significant DFMO effect on

any feature in the lymphocytes (Table 6).Additional information about the relationship between in-

dividual �ODn value distribution within and between differentgroups of patients before and after DFMO treatment was ob-tamed by carrying out a mixed-effect linear model analysis. Thevariances of groups 1 , 2, and 3 did not differ significantly (f= 0.336, df = 2, P = 0.85); however, the variance of group 4(posttreatment, responders) was significantly different from theremaining variances (� = 18.610, df 3, P < 0.001). Theinteraction between treatment and response was not significant

Uv�t 1 .094, df = I , P = 0.30), but treatment and responsewere individually significant (treatment, � = 19.828, df 1,

P < 0.001; response. � = 7.726, df = I, P = 0.005). Esti-mated values are: f3� - 0.5447, f3� = -0.2293, �2 = -0.12 19,

Table 6 Mean area, IOD,. and chromatin structure values of lymphocytes inpretreatment and posttreatment samples

Area IOD,, Entropy Energy Correlation Contrast

Pretreatment 199 91.7 3.2096 0.073 220.6 56

Posttreatment 205 92 3.2088 0.073 2 17.7

P 0.10 0.78 0.88 1 0.37 0.41

o’i = #{176}‘2= O�3 = 0.1816, O�4 = 0.0298, o’,� 0.0035, and o� =

0.2987. Thus, posttreatment values were bower than pretreat-

ment values, and responders had lower values than did nonre-sponders. These results of the mixed modal analysis support thehypothesis that DFMO significantly affected the �

Fig. I shows the individual �OD� histograms for biopsy

specimens from three patients: a nonresponder, a partial re-sponder, and a complete responder. These histograms demon-strate a shift in the DNA index for the patients who responded

and no shift for the patients who did not respond.

For each patient, two histograms were constructed, one forpretreatment �ODn and one for posttreatment �ODn. The his-

tograms represented the proportion of cells falling in a bin,based on rank percentile of optical densities. The individual

histograms were then averaged, giving them equal weight, toget the composite histogram. A composite histogram of eachgroup was then created by averaging the groups’ individualhistograms, thus giving equal weight to each case.

As shown in Fig. 2, the pretreatment biopsies from re-sponders and nonresponders showed similar �OD� distribu-

tions; however, there was a tendency toward a more normalizedpattern in the responders. Following treatment, there was a leftshift in distribution for both responders and nonresponders.

Discussion

CIN 3 is considered a high-risk precursor to invasive cervicalcancer (2, 3). Both genomic instability and proliferative dys-regulation play important roles in the progression of CIN 3lesions and could result in increased mean DNA content (17).Cells with high DNA content have been found to have a

particular proliferative behavior and could play an importantrole in tumor evolution (33) and in the progression of CIN 3 to

invasive cervical carcinoma. It has been shown that “high-risk”CIN 3 lesions, those lesions residing beside invasive cancers,

have a higher percentage of cells with high DNA content thando CIN 3 lesions that are not associated with cancer (22).Therefore, the opposite process, decreasing the number of cellswith high DNA content in CIN 3, may reflect an abolition of

“malignant” behavior of CIN 3 and a reversal of precancerouslesions.

The “slicing problem” makes the precise determination ofDNA content and modal stem-line ploidy difficult on tissue

sections (25, 26). The method of correcting DNA pboidy meas-urements in tissue sections (25) can be used to address thisproblem only for special cell populations with “spherical nucleiand uniform DNA concentration throughout the nucleus.”

on February 26, 2021. © 1997 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

8

A

N

0

fC

eIIS

B

N

0

fC

C

IIS

1.00 1.00 2.00

EOD�

3.00 4.00

11

Na

0

I �.

C

C

l

S 2�

0

EOD�

I0.00

Ii.#{243}o 2.#{243}o

EOD�3.00 4.00

7,

5,

2

‘I I1.00

Cancer Epidemiology, Biomarkers & Prevention 853

C

N0

fC

e1

IS

N0

fC

eIIS

Ill L�I�2.00 3.� 4.00

�oDn

Fig. 1. Individual �OD, histograms from three CIN 3 patients before and after DFMO treatment by histological response. A. a nonresponsive patient: B, a partially

responsive patient: C, a complete responsive patient.

Because the CIN 3 cell populations in our study were quite

heterogenic, we did not use this method. However, we believethat the slicing problem did not affect our data, because we

measured �OD� in pairs: pre- and posttreatment samples from

each patient. Under these conditions, paired comparison was

achieved in genetically similar (especially for nonresponders)cell populations before and after DFMO treatment. Uniformconditions of slide preparation and measurement of pre- and

on February 26, 2021. © 1997 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

0.1

%

0

I

A0.1

04

. 08

.02

. 06

B

0

I

C

e

IS

.04

I

0.1

%0�

I

EOD� EOD�

2

C D0.1

0 .08

I

.06

C

e

I .02

S

.04

I 2 3 4

�OD�

Fig. 2. The summed pre- and posttreatment �OD,, histograms for responders (A and B) and nonresponders (C and D).

854 DNA ImageCytometry in a Cervix Chemoprevention Trial

C

e

IS

C

C.

S

posttreatment lymphocytes were used as controls to demon-strate that no factors other than DFMO affected � Underthese circumstances, we believe that �OD� changes reflect amodulation of DNA content during DFMO treatment.

Our data show that all patients with CIN 3 lesions had highlevels of � and nonresponders showed a trend to havemore high pretreatment �OD� than responders. These datamean that both responders and, especially, nonresponders had

in pretreatment samples high numbers of cells with high DNA

content that contribute to the malignant behavior of CIN 3 (22).

After I month of DFMO administration, reduction of

�ODn was achieved in 10 of 1 1 responders and in 9 of 14nonresponders. If a posttreatment �ODn decrease in responderscan be associated with a change in the histological grade of CINlesions (21 , 22), the reduction of �OD� in nonresponders sug-

gests that DFMO could decrease “DNA amount” in CIN 3lesions even before the appearance of morphological changes.Whether this change is due to a decrease in proliferation or

elimination of particular clones is unknown. We believe thatthese results suggest antiproliferative activity of DFMO be-

cause, as was shown in an animal model, DFMO can inhibit ordelay DNA synthesis and decrease proliferation (12-14). These

data agree with our findings that DFMO can decrease the level

of cell proliferation, which has been detected by proliferating

cell nuclear antigen and MPM-2 modulation (34, 35).To study the antiprobiferative activity of DFMO, we ana-

lyzed the �ODn histograms with different descriptors. The

visual left shift of posttreatment �ODn histograms, which was

quantitatively detected by analysis of percentile values of

�ODn distributions and a decrease of DNA-MG and ScER,

shows a depletion in cell number and in the right shoulder of the

histogram. These changes probably correspond to depletion in

cells in the S-phase and G,-M of the cell cycle. These data

suggest that DFMO has an antiprobiferative and cytostatic ef-

fect. The evidence that DFMO did not affect �OD in lympho-

cytes also confirms the cytostatic, rather than cytotoxic, effect

of DFMO; it selectively affected DNA content in the probifer-

ative cell population (CIN 3) but not the quiescent one (lym-

phocytes).To be sure that this result was not dependent on the quality

of the image analysis system or on the operator, additional

measurements of �OD� in the same samples were performed in

the Cancer Imaging Department of the British Columbia Cancer

Agency (Vancouver, British Columbia, Canada) on their

CytoSavant system by another operator. This independent study

on February 26, 2021. © 1997 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

Cancer Epidemiology, Biomarkers & Prevention 855

also detected a significantly lower �OD� after DFMO treatmentthan that in pretreatment samples (data not shown).

In conclusion, our study suggests that: (a) the �OD� in

CIN 3 tissues significantly decreased during the DFMO che-moprevention trial, and this decrease was due to a depletion of

cells with high DNA content; and (b) the modulation of �ODn

reflected the chemoprevention effect of DFMO before the ap-

pearance of morphological changes, providing a rationale forthe possible use of 1CM-DNA as a SEB in chemopreventiontrials with DFMO. Additional reasons for using 1CM-DNA as

a SEB are its relative simplicity, low cost of reagents, ability to

use small tissue samples, objectivity, and reproducibility.

Acknowledgments

We are grateful for helpful comments from Branko Palcic, Ph.D., Calum

MacAuley, Ph.D., and James Bacus, Ph.D. Editorial assistance from Sunita

Patterson and manuscript preparation by Pat Williams are appreciated.

References

I . Parkin, D. M., Pisani, P., and Ferlay, J. Estimates of the worldwide incidence

of eighteen major cancers in 1985. tnt. J. Cancer, 54: 594-606, 1993.

2. Mitchell, M. F., Hittelman, W. K., Lotan, R., Nishioka, K., Tortolero-Luna, G.,

Richards-Kortum, R., Wharton, J. T., and Hong, W. K. Chemoprevention trials

and surrogate endpoint biomarkers in the cervix. Cancer (Phila.), 76: 1956-1977,

1995.

3. Mitchell, M. F., Tortolero-Luna, G., Lee, J, Hittelman, W. K., Lotan, R.,Pandey, D., Wharton, J. T., Hong, W. K., and Nishioka, K. Results of a Phase I

trial of a-difluoromethylomithine in patients with cervical intraepithelial neopla-

sia. Gynecol. Oncol., 60: 101, 1996.

4. Pegg, A. E. Polyamine metabolism and its importance in neoplastic growth

and as a target for chemotherapy. Cancer Res., 48: 759-774, 1988.

5. Marton, L. J., and Pegg, A. E. Polyamines as targets for therapeutic interven-

tion. Annu. Rev. Phamacol. Toxicol., 35: 55-91, 1995.

6. Tabor, H., and Tabor, C. W. Polyamines. Annu. Rev. Biochem.. 53: 749-790,

1984.

7. Basu, H. S., Sturkenboom, M. C. J., Delcros, J. G., Csokan, P. P., and Szollosi,

J. Effect of polyamine depletion on chromatin structure in U-87 MG human brain

tumour cells. Biochem. J., 282: 723-727, 1992.

8. Panagiotidis. C. A., Artandi, S., Calame, K., and Silverstein, S. J. Polyamines

alter sequence-specific DNA-protein interactions. Nucleic Acids Res. 23: 1800-1809, 1995.

9. Tarn, L. W., and Secco, A. S. Base-pair opening and spermine binding. B-DNA

features displayed in the crystal structure of a gal operon fragment: implicationsfor protein-DNA recognition. Nucleic Acids Res.. 23: 2065-2073,1995.

10. Basu, H. S., Wright, W. D., Deen, D. F., Roti-Roti, J., and Marton, L. J.

Treatment with a polyamine analog alters DNA-matrix association in HeLa cell

nuclei: a nucleoid halo assay. Biochemistry, 32: 4073-4076, 1993.

1 1. Koza, R. A., and Herbst, E. J. Deficiencies in DNA replication and cell-cycle

progression in polyamine-depleted HeLa cells. Biochem. J., 281: 87-93, 1992.

12. Logsdon, C. D., Alves, F., and Rosewicz, S. Role of polyamines in glucocor-

ticoid effects on pancreatic acinar AR42J cell growth and differentiation. Am. J.

Physiol., 262: G285-G290, I 992.

13. Sanbom, G., Niederkom, J., Kan-Mitchell. I., and Albert, D. Prevention of

metastasis of intraocular melanoma in mice treated with difluoromethylomithine.

Graefes Arch. Clin. Exp. Opthalmol., 230: 72-77, 1992.

14. Elitsur, Y., Strom, J., and Luk, G. D. Inhibition of omithine decarboxylase

activity decreases polyamines and suppresses DNA synthesis in human colonic

lamina propria lymphocytes. Immunopharmacology. 25: 253-260, 1993.

15. Huber. M., and Poulin, R. Permissive role of polyamines in the cooperative

action of estrogens and insulin or insulin-like growth factor I on human breast

cancer cell growth. I. Clin. Endocrinol. Metab.. 81: 1 13-123, 1996.

16. Wang, J. Y., Wang, H. L., and Johnson, L. R. Gastrin stimulates expression

of protooncogene c-myc through a process involving polyamines in IEC-6 cells.

Am. J. Physiol., 38: C1474-C1481, 1995.

17. Boone, C. W., and Kelloff, G. J. Intraepithelial neoplasia. surrogate endpoint

biomarkers, and cancer chemoprevention. J. Cell. Biochem. Suppl.. 17F: 37-48.

I993.

18. Barlogie, B., Drewinko, B., Schumann. J., Gonde, W., Dosik, G.. Latreille. J.,

Johnston, D. A., and Freireich, E. I. Cellular DNA content as a marker of

neoplasia in man. Am. J. Med., 69: 195-203, 1980.

19. Bask, J. P. A. Manual of Quantitative Pathology in Cancer Diagnosis and

Prognosis. pp. 1-616. New York: Springer-Verlag New York, Inc., 1991.

20. Ross, J. S. DNA ploidy and cell cycle analysis in cancer diagnosis and

prognosis. Oncology (Basel), 10: 867-887, 1996.

21. Bibbo, M., Bartels, P. H., Dytch, H. E., and Wied, G. L. Ploidy pattems incervical dysplasia. Anal. Quant. Cytol. Histol.. 7: 213-217. 1985.

22. Hanselaar, A. G. J. M., Vooijs, G. P., Mayall, B. H, Pahlplatz. M. M. M., and

Van’t Hof-Grootenboer, A. E. DNA changes in progressive cervical intraepithe-

hal neoplasia. Anal. Cell. Pathol., 4: 3 15-324, 1992.

23. Kropff. M.. Chatelain, R., Muller. C. P., Wagner. A.. Wenzler, T., Bohmer,

H., and Bocking, A. Monitoring DNA cytometric parameters during the course of

chronic myelogenous leukemia. Anal. Quant. Cytol. Histol., /3: 433-439. 1991.

24. Kuo, S. H., and Luh, K. T. Monitoring tumor cell kinetics in patients

receiving chemotherapy for small cell lung cancer. Acta Cytol.. 37: 353-357.

I 993.

25. Bacus, J. W., and Bacus J. V. A method of correcting DNA ploidy meas-

urements in tissue sections. Mod. Pathol., 7: 652-664, 1994.

26. Bocking. A., Biesterfeld, S., and Liu, S. DNA distribution in gastric cancer

and dysplasia. in: Y-C. Zhang and K. Kawai (eds.). Precancerous Conditions andLesions of the Stomach, pp. 103-120. Berlin: Springer-Verlag, 1993.

27. Bocking, A., Giroud, F., and Reith, A. Consensus report of the European

Society for Analytical Cellular Pathology task force on standardization of diag-

nostic DNA image cytometry. Anal. Quant. Cytol. Histol., 17: 1-7, 1995.

28. Xillix Technology Company. Staining Procedure for the Xillix Cyto-savant

Automated Image Cytometer (AIC) and Automated Cervical Cell Screening

System (ACCESS), pp. 4-5. Vancouver: Xillix Technology Company, 1994.

29. Jaggi, B., Poon, S. S. S., MacAulay, C., and Palcic, B. Imaging system formorphometric assessment of conventionally and fluorescently stained cells. Cy-

tometry, 9: 566-572, 1988.

30. Ispizua, A., Baroja, A., and de Ia Hoz, C. Correlation of DNA content oflaryngeal epithelial lesions to degree of malignancy. J. Clin. Oncol., 12: 1600-1606, 1994.

31. Bocking, A., and Auffermann, W. Algorithm for DNA-cytophotometric

diagnosis and grading of malignancy (Letter). Anal. Quant. Cytol. Histol.. 6: 363,

1986.

32. Bocking, A., Adler, C. P., Common, H. H., Hilgarth, M., and Granzen.

Algorithm for DNA-cytophotometric diagnosis and grading of malignancy. Anal.

Quant. Cytol. Histol., 6: 1-8, 1984.

33. de la Hoz, C., and Baroja, A. Proliferative behaviour of high-ploidy cells in

two murine tumour lines. J. Cell. Sci., 104: 31-36, 1993.

34. Hu, W., Mitchell, M. F., Boiko, I., Malpica, A., Linares, A., and Hittelman,

W. N. Decreased PCNA expression in cervical premalignant lesions after che-moprevention by cs-difluoromethylomithine (DFMO). Proc. Am. Assoc. Cancer

Res., 37: 185, 1996.

35. Hu, W., Boiko, I, Mitchell, M. F., and Hittelman, W. N. Sequential dysregu-

lation of proliferation during cervical carcinogenesis as measured by MPM-2

antibody staining. Proc. Am. Assoc. Cancer Res., 36: 588, 1995.

on February 26, 2021. © 1997 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from

1997;6:849-855. Cancer Epidemiol Biomarkers Prev   I V Boiko, M F Mitchell, D K Pandey, et al.   cervical intraepithelial neoplasia.biomarker in a phase I trial of alpha-difluoromethylornithine for DNA image cytometric measurement as a surrogate end point

  Updated version

  http://cebp.aacrjournals.org/content/6/10/849

Access the most recent version of this article at:

   

   

   

  E-mail alerts related to this article or journal.Sign up to receive free email-alerts

  Subscriptions

Reprints and

  .pubs@aacr.orgDepartment at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

  Permissions

  Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cebp.aacrjournals.org/content/6/10/849To request permission to re-use all or part of this article, use this link

on February 26, 2021. © 1997 American Association for Cancer Research. cebp.aacrjournals.org Downloaded from