Low Frequency Epithelial Cells in Bone MarrowAspirates from Prostate Carcinoma Patients AreCytogenetically Aberrant

Peter Mueller, M.D.1

Peter Carroll, M.D.2

Erin Bowers, B.S.1

Dan Moore II, Ph.D.1

Michael Cher, M.D.1,2

Joseph Presti, M.D.2

Maija Wessman, Ph.D.1

Maria G. Pallavicini, Ph.D.1,3

1 Cancer Genetics Group, University of California-San Francisco Cancer Center, San Francisco, Cal-ifornia.

2 Department of Urology, University of California,San Francisco, California.

3 Program in Urologic Oncology, University of Cal-ifornia-San Francisco Cancer Center, San Fran-cisco, California.

Supported by grants from the American CancerSociety and the Academy of Finland (M.W.).

Address for reprints: Maria Pallavicini, Ph.D., 2340Sutter Street S429, UCSF Cancer Center, Univer-sity of California, San Francisco, CA 94115.

Received May 8, 1997; revision received Septem-ber 26, 1997; accepted October 23, 1997.

BACKGROUND. Low frequency epithelial cells occur in bone marrow aspirates of

25-50% of patients with locally confined prostate carcinoma. It is assumed that

bone marrow epithelial cells derive from the primary tumor; however, it has not

been established unequivocally that they are tumor cells. Immunofluorescence

approaches were used to quantify the frequency of epithelial cells in bone

marrow aspirates from prostate carcinoma patients and genotypic analyses

were used to determine whether they contained numeric aberrations of chro-

mosomes 1, 7, and 8.

METHODS. Epithelial cells in bone marrow aspirates collected after radical pros-

tatectomy were visualized using fluorescence microscopy and fluorophore-linked

antibodies against cytokeratin 8,18 (CK) and prostate specific antigen (PSA). Anti-

bodies specific for proliferating nuclear cell antigen (PCNA) were used to evaluate

the cycling status of discriminated cells. Copies of chromosomes 1, 7, and 8 in the

discriminated epithelial cells were quantified using fluorescence in situ hybridiza-

tion.

RESULTS. CK1 cells were present in bone marrow aspirates from 30 of 66 patients

(approximately 45%) at a median frequency of 1.4 CK1 cells/105 mononuclear cells.

Few CK1 epithelial cells in the bone marrow aspirates coexpressed PSA and none

of the CK1 cells expressed PCNA. Approximately 70-75% of the CK1 cells contained

7 and 8 aneusomies. Gains of chromosome 1 occurred in 42% of the CK1 cells.

CONCLUSIONS. The majority of CK1 cells in bone marrow aspirates collected after

surgery are cytogenetically aberrant, which is consistent with a primary tumor

origin. The prevalence and frequency of CK1 cells is independent of tumor stage/

grade and androgen treatment. [See editorial on pages 394-8, this issue.] Cancer

1998;83:538 – 46. © 1998 American Cancer Society.

KEYWORDS: prostate carcinoma, micrometastases, chromosome abnormalities,bone marrow.

Tumor progression and treatment outcome in patients with pros-tate carcinoma are variable.1 Tumor stage and grade and serum

prostate specific antigen (PSA) levels are useful prognosticators.2– 4

However, even among patients with similarly staged tumors, diseaseprogression and treatment response are not uniform. Continued dis-ease progression after surgical removal of the primary tumor maysuggest that some patients are understaged at diagnosis; occult dis-ease at either local or distant sites may contribute to disease progres-sion. Rare epithelial cells are present in bone marrow aspirates ofsubsets of patients with solid tumors, including those with prostatecarcinoma. These epithelial cells are presumed to be micrometastasesthat seeded into the bone marrow. Bone marrow micrometastases

538

© 1998 American Cancer Society

have been associated with disease progression insome studies.5–9 However, the tumor origin of epithe-lial cells in the bone marrow has not been establishedunequivocally. Indeed, bone marrow aspirates fromsome patients without cancer also contain epithelialcells.10,11 Furthermore, approximately 25% of cancerpatients with epithelial cells in bone marrow aspiratesdo not show disease progression.8 Thus, it is impor-tant to determine whether the discriminated epithelialcells possess genomic or functional features consis-tent with tumor derivation and metastatic capability.

We used a combined immunophenotype/genotypeapproach11 to detect and quantify epithelial cells in bonemarrow aspirates from patients with prostate carcinomaand to assess the molecular cytogenetic features of thediscriminated cells. Cells labeled immunofluorescentlywith antibodies against cytokeratin (CK) 8/18 werequantified in bone marrow aspirates from 66 patientswith prostate carcinoma. Aneusomies of chromosomes 7and 8 occur in primary prostate tumor cells12–19 and areassociated with aggressive disease and metastases.12,17,20

Numeric aberrations of chromosome 1 also occur, butless frequently. Thus, we used fluorescence in situ hy-bridization (FISH) to determine whether epithelial cellsin bone marrow aspirates carry chromosome 1, 7, and 8aneusomies. In addition, the proliferative status of CK1

cells was measured using antibodies against proliferat-ing cell nuclear antigen (PCNA). Our data indicate thatthe majority of epithelial cells in bone marrow aspiratesfrom prostate carcinoma patients contain chromosome7 and 8 aneusomies. The relation between the presenceand frequency of epithelial cells in bone marrow aspi-rates, clinical stage, tumor grade, and prior treatment isdiscussed.

MATERIALS AND METHODSPatients and Specimen CollectionPatients with prostate adenocarcincoma were stagedclinically using digital rectal examination, serum PSAlevels, transrectal ultrasound, radionuclide bone scan,and cross-sectional imaging of the pelvis. Local tumorextent was assessed by transrectal ultrasound in pa-tients without surgical removal of the prostate and bypathologic evaluation in the radical prostatectomy pa-tients. The accuracy of transrectal ultrasound com-bined with serum PSA measurements to assess localtumor stage in patients at the University of CaliforniaSan Francisco (UCSF) approximates 75%.21 Lymphad-enectomy was performed in a subset of patients.Lymph nodes were assumed to be without metastaticdisease in patients who did not undergo lymph nodedissection. Radionuclide bone scans were negative inall patients, indicating the absence of macroscopicbone metastases. Tumor stage was assigned according

to the TNM classification.22 Tumor grade was assignedaccording to the Gleason system using prostatectomyspecimens from patients undergoing radical prosta-tectomy or biopsy specimens in the remaining pa-tients.23

Bone marrow aspirates (2– 4 mL) were collectedfrom 1 side of the posterior iliac crest in 66 patientsimmediately after surgical intervention for adenocar-cinoma of the prostate. Peripheral blood or bone mar-row from patients without known epithelial tumorsserved as controls. All specimens were obtained afterinformed consent was collected on protocols ap-proved by the Committee on Human Research atUCSF. Surgical intervention included radical prosta-tectomy without staging pelvic lymphadenectomy (2patients), radical prostatectomy with staging pelviclymphadenectomy (35 patients), lymphadenectomyalone (2 patients), cryosurgical ablation of the prostate(25 patients), and orchiectomy alone (2 patients). Pel-vic lymphadenectomy was performed in 11 of 25 pa-tients who underwent cryosurgical ablation. Lymph-adenectomy was not performed in 14 patients whoeither had a low estimated probability of metastases tothe lymph nodes (11 patients) or advanced age or highoperative risk (3 patients). A subset of patients (37patients) received androgen ablation by orchiectomyor hormonal therapy 3 months prior to bone marrowsampling.

Cell Processing and Immunofluorescence LabelingMononuclear cells from bone marrow aspirates frompatients with prostate carcinoma, hematologic disor-ders (controls), and healthy volunteers were processedfor immunostaining. Mononuclear cells were isolatedfrom bone marrow specimens using density gradientseparation (Histopaque-1077; Sigma Chemical Co., St.Louis, MO) and diluted to approximately 1.5 3 106

cells/mL for deposition on glass slides using cytocen-trifugation as described previously.11 Slides werestored at —20 °C until subsequent processing. A pros-tate cancer cell line, LNCaP (ATCC CRL-1740; Ameri-can TypeCulture Collection, Rockville, MD), was usedas a control for immunolabeling and to measure assaysensitivity.

Slides containing mononuclear bone marrow cellsfrom patients without prostate carcinoma or mixturesof LNCaP cells and peripheral blood mononuclearcells from healthy individuals were labeled immun-ofluorescently with antibodies directed against CKand PSA. Cells were fixed in 0.5% formalin in phos-phate-buffered saline (PBS) (0.1 M NaH2P04, 0.1 MNaHPO4, pH 7.4) for 75 minutes at room temperature(RT), rinsed twice in water, air-dried at RT, and fixed incold (—20 °C) methanol/acetone (3:1, volume:volume)

Epithelial Cells in Bone Marrow/Mueller et al. 539

for 5 minutes. After 2 washes in water, cells wereplaced in 43 standard saline citrate (SSC) (0.15 MNaCl, 0.015 M sodium citrate, pH 7.0) containing0.05% Triton-X-100 (Sigma Chemical Co.) for 5 min-utes. Fifty microliters of 4 3 SSC containing 1% bovineserum albumin (BSA) (Sigma Chemical Co.) was ap-plied to each slide for 20 minutes. Anti-PSA (rabbitantihuman PSA; DAKO Corp., Carpinteria, CA), 1:100in 4 3 SSC/1% BSA, was applied for 30 minutes at RT.Slides were rinsed twice in 4 3 SSC 1 0.05% Triton-X-100 for 10 minutes each followed by a 30-minuteincubation with rhodamine-conjugated swine anti-rabbit (DAKO Corp) (1:100 dilution in 4 3 SSC/1%BSA) and a fluorescein (FITC)-conjugated anti-CK an-tibody specific for CKs 8 and 18 (clone Cam 5.2; Bec-ton-Dickinson, San Jose, CA) (1:10 in 4 3 SSC/1%BSA). The slides were washed twice as mentionedearlier and counterstained with 4’, 6-diamidino-2-phenylindole (DAPI) (Sigma Chemical Co.), 0.4 mM, inantifade.24

Dual labeling of cells with antibodies againstPCNA and CK required modification of the proceduredescribed earlier to retain the PCNA antigen. Cellswere fixed in 0.5% formalin in PBS (pH 7.4) for 20minutes at 4 °C, rinsed twice in water, and, without airdrying, fixed in cold (-20 °C) methanol (100%) for 10minutes. After 2 washes in water and a 5-minute PBSrinse, cells were incubated with biotinylated anti-PCNA (clone PC 10; Caltag Laboratories, South SanFrancisco, CA) (diluted 1:25 in PBS) for 2 hours at RT.Slides were washed twice in 4 3 SSC 1 0.05% Triton-X-100 for 10 minutes each and then incubated with 50mL of 4 3 SSC/1% BSA for 20 minutes. Texas Redconjugated avidin (Vector Laboratories, Burlingame,CA), diluted 1:400 in 4 3 SSC/1% BSA, was applied for30 minutes at RT, and slides washed consecutivelywith 4 3 SSC, 4 3 SSC 1 0.05% Triton-X-100, and 4 3SSC for 10 minutes each and incubated (5 minutes atRT) in PNM (5% nonfat powdered milk [Sigma Chem-ical Co.], 0.1% NaN3 in 0.1 M monobasic sodium phos-phate, 0.1 M dibasic sodium phosphate/0.05% Non-idet P-40 [Sigma Chemical Co.], pH 8.0). Thereaftersamples were incubated with biotinylated goat anti-avidin antibody (Vector Laboratories) (diluted 1:100 inPNM) for 30 minutes. Subsequent washes were donewith PN 3 times, 10 minutes each, and a second layerof Texas Red conjugated avidin was applied withwashes in PN as described earlier. Anti-CK-FITC label-ing was performed as described earlier.

MicroscopyImmunofluorescently labeled epithelial cells in bonemarrow aspirates were quantified using fluorescencemicroscopy. The number of CK1 cells, on 12 slides/

patient, that displayed cytoplasmic FITC signals with acytoskeleton web-like staining pattern were enumer-ated using a Zeiss Axioscope with 3630 magnification(Zeiss, Inc., Thornwood, NY). The total number ofhematopoietic cells screened was estimated by multi-plying the average number of cells contained in fivemicroscopic fields with the total number of micro-scopic fields analyzed/slide. Between 1 3 104 and 6 3105 (mean, 2.0 3 105) mononuclear bone marrow cellswere scored for each prostate carcinoma patient. Thelocations of CK1 cells on the slides, as well as an imageof the surrounding field, was recorded using comput-erized image microscopy so that cells that lost immu-nophenotypic markers during the FISH processingprocedure could be relocated accurately. Details of theimage acquisition and relocation procedure have beendescribed previously.11

FISHMolecular cytogenetic characterization of the discrim-inated epithelial cells (CK1 and PSA1/2) and sur-rounding hematopoietic cells was performed usingFISH. Cells on slides were hybridized with pericentro-meric probes specific for a-satellite repeated se-quences on chromosomes 1, 7, and 8. Cell loss afterhybridization of cells labeled with anti-CK and anti-PCNA antibodies precluded molecular cytogeneticanalysis of all CK1 cells. FISH was performed as de-scribed by Pinkel et al.25 Briefly, cells on slides weresoaked in PN buffer for 20 minutes, dehydrated in anethanol series, and denatured for 3 minutes at 72 °C in70% formamide/2 3 SSC. DNA probes (1 mL each ofchromosome 7- spectrum green and chromosome 8-spectrum orange or chromosome 1-spectrum orange;Vysis, Downer’s Grove, IL) were denatured for 5 min-utes at 72 °C in 55% formamide, 1 3 SSC, and 1 gdextran sulfate (total volume 10 mL). Denaturedprobes were applied to the air-dried slides, which weresealed with a coverslip and rubber cement. Cells werehybridized in a moisture chamber at 37 °C overnightand then washed in 1 3 SSC at 72 °C, 2 3 SSC at 37 °C,2 3 SSC at RT, and H2O at RT for 5 minutes each. DAPIin antifade solution was applied to facilitate visualiza-tion of the nucleus.

Hybridization domains in the discriminated epi-thelial cells (relocated using the stored coordinatesand images), as well as surrounding cells in the samefield, were enumerated. The number of hybridizationdomains/nucleus is considered to be proportional tochromosome copy number. The maximum back-ground trisomy and monosomy frequency in bloodand bone marrow specimens from patients withoutcancer was 1% and 5%, respectively, for all probesutilized.

540 CANCER August 1, 1998 / Volume 83 / Number 3

Statistical MethodsStatistical methods were used to evaluate the relation-ship between CK1 cell frequency and clinical param-eters and to establish confidence intervals for CK1 cellfrequencies for individual patients. “Prevalence” is de-fined as the proportion of bone marrow aspirates fromthe cohort study group that contained at least oneCK

1

cell. “Frequency” refers to the number of CK1

cells among the total number of cells screened in eachbone marrow aspirate. A generalized linear model26,27

was used to assess the effects of patient age, tumorstage, preoperative serum PSA level, Gleason score,and presence or absence of androgen ablation on CK1

cell frequencies and prevalence. To visually assess het-erogeneity among different samples, 95% confidenceintervals (95% CI) for CK1 cell frequencies were cal-culated for each specimen. Each 95% CI is based onthe assumption that the number of CK1 cells in anindividual specimen follows a binomial distribution.Upper and lower bounds for each CK1 frequency werecalculated using a method proposed by Blyth andColin.28

RESULTSImmunofluorescence-Based Epithelial Cell DetectionThe sensitivity and specificity of the immunofluores-cence-based assay to discriminate epithelial cellsamong a background of hematopoietic cells were es-tablished using LNCaP tumor cells, which expressCK8/18 and PSA. LNCap cells labeled with FITC-con-jugated anti-CK8/18 show a web-like filamentous pat-tern characteristic of cytoskeleton networks (Fig. 1a).Surface PSA expression on LNCaP cells appears asdiffuse red fluorescence (Fig. 1b). LNCaP tumor cellswere seeded at frequencies between 10-1 to 10-5

among mononuclear cells isolated from peripheralblood from healthy donors and quantified using fluo-rescence microscopy. The measured frequency ofseeded tumor cells was similar to the expected fre-quency in three independent experiments (Fig. 2). Theassay sensitivity was sufficient to detect a prostatetumor cell present at a frequency of 10-5. The speci-ficity of the immunofluorescence-based assay was as-sessed by screening 1.43 3 106 mononuclear cellsfrom peripheral blood or bone marrow aspirates col-lected from 7 different healthy individuals. A singleCK1 cell was detected; thus our estimate of the back-ground frequency of CK1 cells is 0.7 per million cells(95% CI, 2 3 1028 to 3.9 3 1026).

Patient Demographics and Clinical FeaturesBone marrow aspirates were collected from 66 pa-tients with prostate carcinoma and 3 patients without

prostate carcinoma. The median age of the prostatecarcinoma patient population was 64 years (range,47– 80 years). Fifty-five of 66 tumors were GleasonGrade 5–7. There was no lymph node involvement in62 patients. The number of patients with T1, T2, T3,and T4 tumors was 1, 34, 26, and 6, respectively.

Ck1 Cells in Bone Marrow AspiratesThe relation between the prevalence of CK1 bonemarrow aspirates and clinical parameters was evalu-ated. Approximately 45% of bone marrow aspirates (30of 67 patients) contained CK1 cells. CK1 bone marrowaspirates occurred in 13 of 34 patients (38%) withT2N0 tumors, 11 of 26 patients (42%) with T3N0 tu-mors, and 4 of 5 patients (80%) with T4N1 tumors(Fig. 3). Overall, CK1 bone marrow aspirates weredetected in 25 of 61 patients (41%) without lymphnode metastases. The prevalence of CK1 bone marrowaspirates was independent of serum PSA levels at di-agnosis, tumor classification, and androgen ablationprior to surgery.

The frequency of CK1 cells in 30 CK1 bone mar-row aspirates varied among prostate carcinoma pa-tients. The frequency of CK1 cells in bone marrowspecimens that contained at least 1 CK1 cell rangedfrom 0.27–73.2 CK1 cells (mean 6 95% CI, 6.1 6 15.3)per 105 mononuclear cells. The median CK1 frequencyin bone marrow aspirates categorized according totumor classification (T1N0, T2N0, T3N0, and T4N0/TxN1) is summarized in Table 1. The CK1 cell fre-quency was statistically similar at all stages. Further-more, the median CK1 cell frequency was dissociatedfrom androgen treatment and independent of patientage and serum PSA levels at diagnosis (data notshown).

We next investigated whether CK1 cells in thebone marrow aspirates expressed PSA and nuclearproteins associated with cell proliferation (e.g., PCNA).CK1 cells were labeled with rhodamine-conjugatedantibodies directed against PSA and FITC-linked anti-bodies specific for PCNA. Using this dual stainingprocedure, normal hematopoietic cells appear asdimly stained objects (arrows in Figs. 1c and 1e) withlow FITC and rhodamine-linked fluorescence. CK1

PSA1 epithelial cells from two patients are shown inFigures 1c and 1d. Less than 12% of the CK1 cells in 30bone marrow aspirates coexpressed PSA. CK-PSA1

cells were not detected in any bone marrow specimen.PCNA expression in CK1 cells in bone marrow aspi-rates from 11 patients was measured to indicate pro-liferative status. None of the CK1 cells expressedPCNA, although 12–25% of the surrounding hemato-poietic cells were PCNA1.

Epithelial Cells in Bone Marrow/Mueller et al. 541

Aneusomy Analyses of CK1 Cells in Bone MarrowAspiratesNumeric abnormalities involving chromosomes 1, 7,and 8 in CK1 cells were measured using FISH. Chro-mosome 7 and 8 hybridization domains were quanti-fied in CK1 cells from 24 patients. Fourteen bonemarrow aspirates had 1 CK1 cell and 10 bone marrowaspirates contained . 1 CK1 cells, some of whichappeared to be present in clumps. Photomicrographsof a single CK1 cell and a clump of CK1 cells in bonemarrow aspirates from Stage T3aN0 and T2bN0 pa-tients are shown in Figures 1e and 1g, respectively. Inthese same cells, multiple hybridization domains forchromosomes 7 (red) and 8 (green) clearly are visible

(Figs. 1f and 1h). Chromosome 7 and 8 hybridizationdomains were quantified in 105 CK1 cells; 80 CK1 cellsfrom these same patients also were examined forchromosome 1 aneusomies. Approximately 70 –75% ofCK1 cells displayed . 2 copies of chromosome 7and/or 8 (Table 2). Approximately 66% (69 of 105) ofthe cells carried aneusomies of both chromosomes 7plus 8. In contrast, 42% of the specimens were aneu-somic for chromosome 1. The frequency of CK1 cellswith normal copy numbers of chromosomes 1, 7, or 8was 58%, 23%, and 25%, respectively. Thus, it is ap-parent that the majority of CK1 cells are aneusomic.Only 1 of 79 cells was tetrasomic for chromosomes 1,7, and 8. Chromosome 7 or 8 aneusomies occurred

FIGURE 1. Photomicrographs of epi-

thelial cells. LNCaP cells labeled with (a)

fluorescein conjugated anti-cytokeratin

(CK) and (b) rhodamine-linked anti-pros-

tate specific antigen (PSA). Panels c and

d show CK1PSA1 and CK1PSA2 epite-

lial cells, respectively, in bone marrow

aspirates from individual patients. Nor-

mal hematopoietic cells appear as dimly

stained objects surrounding the epithe-

lial cell. (e) Epithelial cell in bone marrow

aspirate expressing CK and PSA and (f)

fluorescence in situ hybridization (FISH)

analysis. (g) A cluster of CK1 epithelial

cells is visible. (h) shows FISH on the

same cells. Chromosome 7 hybridization

domains appear green, whereas those

corresponding to chromosome 8 are red.

542 CANCER August 1, 1998 / Volume 83 / Number 3

more frequently than those involving chromosome 1(P , 0.001). Two patients contributed 56 CK1 cells tothe study group. Exclusion of these patients to reducepossible bias introduced by the relatively large contri-bution of measurements on cells from two patientsdid not affect the statistical significance of preferential7 and 8 aneusomies compared with chromosome 1.Molecular cytogenetic analyses of multiple CK1 cells

present in the same bone marrow aspirate revealedheterogeneous aneusomy patterns. In three of ten as-pirates that contained multiple CK1 cells, the discrim-inated cells displayed varying combinations of two,three, and four copies of chromosome 7 in conjunc-tion with similar variability in chromosome 8 copynumber. Thus, the cells do not appear to be cytoge-netically homogeneous.

FIGURE 3. Incidence and frequency of

cytokeratin positive (CK1) cells in bone

marrow aspirates. The frequency of CK1

cells in each bone marrow aspirate is

shown on the ordinate for patients with

(■) and without (E) hormonal treatment

prior to surgery. The lower (') and up-

per (Á) bounds of the 95% confidence

intervals (95% CI) are shown by dark

horizontal lines for each bone marrow

specimen with CK1 cells. The upper

bound of the 95% CI for negative bone

marrow aspirates from androgen-

treated patients is represented by light

gray horizontal lines. In each case, the

95% CI is based on the number of cells

scored and is useful for estimating the

certainty with which a specimen is des-

ignated as a ‘‘positive or negative’’ bone

marrow aspirate.

FIGURE 2. Frequency of cytokeratin

positive, prostate specific antigen posi-

tive epithelial cells admixed with hema-

topoietic cells. The expected frequency

of epithelial cells based on the number

of LNCaP cells seeded is shown on the

abscissa. The measured frequency of

epithelial cells and the lower (') and

upper (Á) bounds of the 95% confi-

dence intervals are shown for each

measurement.

Epithelial Cells in Bone Marrow/Mueller et al. 543

DISCUSSIONDissemination of cells from the primary tumor toother organs is a prerequisite to develop metastases. Itis assumed that assays for micrometastases that detectrare epithelial cells in bone marrow aspirates fromcancer patients are measuring disseminated cells thathave seeded from the primary tumor. However, lowfrequency epithelial cells have been detected in bonemarrow aspirates from patients without cancer,10

which suggests caution against assuming that the CK1

cells in bone marrow aspirates uniformly originatefrom the tumor. It is interesting to note that CK1 cellsalso are present in bone marrow aspirates from pa-tients bearing tumors that typically do not metastasizeto the bone marrow (e.g., colorectal carcinoma8). Ifmicrometastatic assays are to be useful for prognosti-cation, increased understanding of the features of rareCK1 cells that may contribute to metastases is impor-tant. Thus, we used molecular and functional ap-proaches to determine whether CK1 epithelial cells inbone marrow aspirates exhibited features consistentwith a tumor origin (e.g., cytogenetically aberrant) andmetastatic potential (e.g., proliferation).

The sensitivity and specificity of the immunoflu-orescence-based assay to detect prostate epithelialcells among hematopoietic cells was established. As-say specificity approximated 1025. One CK1 cell wasdetected in a specimen from a healthy individual,indicating that the overall background frequency ofCK-labeled cells approximates 1026. The prevalence ofCK1 bone marrow specimens from healthy donors inour study is 14%, which is somewhat higher thanpublished reports (2–5%).7,10 Although it is possible

that the immunofluorescent detection assay generatesa higher incidence of false-positive results than themore conventional immunocytochemical approaches,the relatively small numbers of patients without epi-thelial tumors that were analyzed in our study pre-cludes rigorous comparison of the two approaches.However, the specificity of the immunofluorescenceassay is likely to be high because at least 75% of thediscriminated CK1 cells in bone marrow aspiratesfrom prostate carcinoma patients are cytogeneticallyaberrant and thus likely to be tumor cells.

Epithelial cells were present in bone marrow as-pirates in 45% of patients with prostate carcinoma.The prevalence of CK1 bone marrow aspirates wassimilar in T2N0 and T3N0 tumors. Although there wasa trend toward a higher prevalence of CK1 bone mar-row aspirates from patients with T4N0 tumors, thesmall number of samples accrued at this stage limitsstatistical significance. The prevalence of CK1 speci-mens in our patient series is similar to that reported inother studies9,29 –31 when bone marrow aspirates werecollected at diagnosis. For example, Oberneder et al.29

detected CK1 cells in 33% of patients with prostatecarcinoma. Pantel et al.30 found CK1 cells in 54% ofStage C patients by harvesting bone marrow aspira-tions from $2 sites. Bretton et al.31 found that CK1

cells were present in bone marrow from 30% of pa-tients with varying stages of prostate carcinoma. Thefrequency of CK1 cells in our study ranged between0.27–73.2 epithelial cells per 105 mononuclear bonemarrow cells, which is within the range of previouslypublished reports of specimens collected at diagno-sis.29 –31 Neither the prevalence of CK1 bone marrowcells or the frequency of CK1 cells correlated withother risk factors such as tumor stage, grade, or initialPSA value or with androgen ablation.

Characterization of the proliferative status of CK1

cells in the bone marrow is important to determinewhether features of the discriminated epithelial cells

TABLE 2Chromosome 1, 7 and 8 Aneusomies in CK1 Cells in Bone MarrowAspirates

Hybridization domains inCK1 cells

Chromosome

1 7 8 7 and 8

2 copies 46 24 26 13.2 copies 34 79 75 69,2 copies 0 2 4 36Total cells 80 105 105 105.% CK1 cells with .2 copies 42.5 75.2 71.4 65

CK: cytokeratin; 1: positive.

Values are the number of CK1 cells displaying hybridization domains.

TABLE 1Median Frequency of CK1 Cells in Positive Bone Marrow Aspirates

Tumor classificationHormonaltherapy

Prevalenceof CK1

specimen

Medianfrequency ofCK1 cellsa

T1N0 No 100% 0.51(1/1)

T2N0 No 35% 1.42(7/20)

T2N0 Yes 43% 1.05(6/14)

T3N0 No 56% 2.15(5/9)

T3N0 Yes 35% 2.84(6/17)

T4N0/TxN1 Yes 80% 0.73(4/5)

CK: cytokeratin; 1: positive.a Cytokeratin positive (CIC1) cells/100,000 mononuclear cells in CK1 specimens.b Significantly different from patients who did not receive hormonal therapy (P , 0.01).

544 CANCER August 1, 1998 / Volume 83 / Number 3

are consistent with metastatic development. In ourstudy, the proliferative status of CK1 cells in bonemarrow aspirates was measured using antibodiesagainst PCNA, a cell cycle-related protein. None of theCK1 cells expressed PCNA. Earlier studies32 demon-strated a low incidence of proliferating CK1 cells inbone marrow aspirates from patients with breast, gas-tric, and colorectal carcinomas. CK1 cells that ex-pressed Ki-67, a cell cycle-related protein, werepresent in only 1 of 33 patients. Only a minority ofCK1 cells in bone marrow aspirates in a subset ofpatients with colorectal carcinoma express p120, aprotein that may be involved in cell cycling.32 Dataobtained using antibodies against PCNA, Ki-67, andp120 may not be comparable directly because theirability to label the same population of proliferatingcells is controversial.33 Nevertheless, the fraction ofproliferating cells measured using immunohisto-chemical markers in the majority of studies appears tobe low. Thus, it is unclear whether the CK1 cells inbone marrow aspirates have the potential to contrib-ute to metastases. It is possible that their presence is areflection of body tumor burden and/or widespreadtumor cell dissemination, and they may not contrib-ute directly to the development of overt bone marrowmetastases. Additional studies are needed to establishtheir role in the formation of metastases.

Molecular cytogenetic analysis of the CK1 popu-lation in bone marrow aspirates from prostate carci-noma patients was accomplished using FISH. A simi-lar approach has been used to investigate molecularchanges in epithelial cells in breast carcinoma34 andcolorectal carcinoma.35 Gains of chromosome 7 or 8more frequently occur in metastatic tumors than pri-mary prostate tumors12,13 and are associated with anincreased risk of developing distant metastases.17

Gains of chromosome 1 in primary prostate carci-noma have been reported.16 We evaluated whetherCK1 cells that disseminate to bone marrow carry chro-mosome 7 and 8 aneusomies by hybridizing the CK1

cells with chromosome 1, 7, and 8 pericentromericDNA probes. The number of hybridization domains inCK1 interphase cells hybridized with chromosomespecific pericentromeric probes is considered to beproportional to chromosome copy number. Hybrid-ization artifacts due to low hybridization efficiencyand/or somatic pairing12,36 complicate inferences thatsingle hybridization domains reflect monosomy; thuswe excluded cells that displayed single hybridizationdomains of all chromosomes measured. The fre-quency of trisomic normal hematopoietic cells rangesfrom 0.3–1.0%.37 Our data demonstrate that at least70 –74% of the CK1 cells in bone marrow aspiratesfrom prostate carcinoma patients display gains of

chromosomes 7 and/or 8. Approximately 42% of theCK1 tumor cells showed chromosome 1 trisomy/tet-rasomy. Thus, CK1 cells in the bone marrow containmore numeric aberrations of chromosomes 7 and 8than chromosome 1. Our finding that the majority ofCK1 cells in bone marrow aspirates from prostatecarcinoma patients are cytogenetically aberrant areconsistent with CK1 cells originating from a primarytumor, even in specimens containing only one CK1

cell. Concordant molecular cytogenetic analyses ofmatched primary tumor and bone marrow epithelialcells in patients with prostate carcinoma will substan-tiate these findings further.

Heterogeneity in the aneusomy patterns amongCK1 cells within individual patients was observed inthree of ten patients. These data may suggest thatmultiple, cytogenetically distinct populations of tumorcells seeded into the bone marrow or that the genomicinstability typically associated with tumor cells mayalter the aneusomy features of the seeded cells. How-ever, in other patients the CK1 cells uniformly showedthe same hybridization pattern, which may indicateseeding of a clonal tumor population. Analysis of ad-ditional cells will be required to confirm the possibilityof molecular heterogeneity and/or genomic evolutionamong CK1 cells in bone marrow aspirates.

Epithelial cells in bone marrow aspirates fromprostate carcinoma patients are cytogenetically aber-rant with preferential aneusomies involving chromo-somes 7 and 8. Our data suggest that the majority ofCK1 cells in bone marrow aspirates from prostatecarcinoma patients are likely to be tumor cells. How-ever, it still is unclear whether CK1 cells in bone mar-row aspirates contribute to the generation of metas-tases. According to Paget’s seed and soil hypothesis,only a minority of disseminated tumor cells are able toproliferate at distant sites.38 Increased understandingof the molecular cytogenetic changes and functionalfeatures associated with metastatic development willfacilitate assessment of the potential of low frequencytumor cells in bone marrow aspirates to form metas-tases.

REFERENCES1. Gittes RF. Carcinoma of the prostate [see comments]. N Engl

J Med 1991;324(4):236 – 45.2. Ohori M, Wheeler TM, Scardino PT. The New American

Joint Committee on Cancer and International Union AgainstCancer TNM classification of prostate cancer. Clinicopath-ologic correlations. Cancer 1994;74:104 –14.

3. Grignon DJ, Hammond EH. Report of the Prostate CancerWorking Group. Arch Pathol Lab Med 1995;119(12):1122– 6.

4. Rosen MA. Impact of prostate-specific antigen screening onthe natural history of prostate cancer. Urology 1995;46(6):757– 68.

Epithelial Cells in Bone Marrow/Mueller et al. 545

5. Cote RJ, Rosen PP, Lesser ML, Old LJ, Osborne MP. Predic-tion of early relapse in patients with operable breast cancerby detection of occult bone marrow micrometastases. J ClinOncol 1991;9(10):1749 –56.

6. Diel IJ, Kaufmann M, Goerner R, Costa SD, Kaul S, Bastert G.Detection of tumor cells in bone marrow of patients withprimary breast cancer: a prognostic factor for distant me-tastasis. J Clin Oncol 1992;10(10):1534 –9.

7. Jauch KW, Heiss MH, Gruetzner U, Funke I, Pantel K, Rieth-mueller G, et al. Prognostic significance of bone marrowmicrometastases in patients with gastric cancer. J Clin Oncol1996;14:1810 –7.

8. Lindemann F, Schlimok G, Dirschedl P, Witte J, RiethmullerG. Prognostic significance of micrometastatic tumour cellsin bone marrow of colorectal cancer patients. Lancet 1992;340(8821):685–9.

9. Pantel K, Izbicki J, Passlick B, Angstwurm M, Haussinger K,Thetter O, et al. Frequency and prognostic significance ofisolated tumour cells in bone marrow of patients with non-small-cell lung cancer without overt metastases. Lancet1996;347(9002):649 –53.

10. Pantel K, Schlimok G, Angstwurm M, Weckermann D,Schmaus W, Gath H, et al. Methodological analysis of im-munocytochemical screening for disseminated epithelial tu-mor cells in bone marrow. J Hematother 1994;3(3):165–73.

11. Litle VR, Lockett SJ, Pallavicin MG. Genotype/phenotypeanalyses of low frequency tumor cells using computerizedimage microscopy. Cytometry 1996;23:344 –9.

12. Bandyk MG, Zhao L, Troncoso P, Pisters LL, Palmer JL, vonEschenbach AC, et al. Trisomy 7: a potential cytogeneticmarker of human prostate cancer progression. Genes Chro-mosom Cancer 1994;9:19 –27.

13. Baretton GB, Valina C, Vogt T, Schneiderbanger K, DieboldJ, Lohrs U. Interphase cytogenetic analysis of prostatic car-cinomas by use of nonisotopic in situ hybridization. CancerRes 1994;54(16):4472– 80.

14. Cher ML, MacGrogan D, Bookstein R, Brown JA, Jenkins RB,Jensen RH. Comparative genomic hybridization, allelic im-balance, and fluorescence in situ hybridization on chromo-some 8 in prostate cancer. Genes Chromosom Cancer 1994;11(3):153– 62.

15. Jones E, Zhu XL, Rohr LR, Stephenson RA, Brothman AR.Aneusomy of chromosomes 7 and 17 detected by FISH inprostate cancer and the effects of selection in vitro. GenesChromosom Cancer 1994;11(3):163–70.

16. Konig JJ, Teubel W, van Dongen JW, Romijn JC, HagemeijerA, Schroder FH. Loss and gain of chromosomes 1, 18, and Yin prostate cancer. Prostate 1994;25(6):281–91.

17. Takahashi S, Alcaraz A, Brown J, Borell T, Herath J, Bergs-tralh E, et al. Aneusomies of chromosome 8 and Y detectedby fluorescence in situ hybridization are prognostic markersfor pathological stage C (pT3N0M0) prostate carcinoma. ClinCancer Res 1996;2:137–5.

18. Visakorpi T, Hyytinen E, Kallioniemi A, Isola J, KallioniemiOP. Sensitive detection of chromosome copy number aber-rations in prostate cancer by fluorescence in situ hybridiza-tion. Am J Pathol 1994;145(3):624 –30.

19. Zitzelsberger H, Szucs S, Weier HU, Lehmann L, BraselmannH, Enders S, et al. Numerical abnormalities of chromosome7 in human prostate cancer detected by fluorescence in situhybridization (FISH) on paraffin-embedded tissue sectionswith centromere-specific DNA probes. J Pathol 1994;172(4):325–35.

20. Qian J, Bostwick DG, Takahashi S, Borell TJ, Herath JF,

Lieber MM, et al. Chromosomal anomalies in prostaticintraepithelial neoplasia and carcinoma detected by flu-orescence in situ hybridization. Cancer Res 1995;55(22):5408 –14.

21. Wolf J Jr., Shinohara K, Carroll PR, Narayan P. Combinedrole of transrectal ultrasonography, Gleason score, andprostate-specific antigen in predicting organ-confined pros-tate cancer. Urology 1993;42(2):131–7.

22. Schroder FH, Hermanek P, Denis L, Fair WR, Gospodarow-icz MK, Pavone-Macaluso M. The TNM classification ofprostate cancer. Prostate Suppl 1992;4:129 –38.

23. Gleason DF. Classification of prostatic carcinomas. CancerChemother Rep 1966;50:125– 8.

24. Johnson GD, Nogueria JGM. A simple method of reducingthe fading of immunofluorescence during microscopy. J Im-munol Methods 1981;43:349 –50.

25. Pinkel D, Straume T, Gray JW. Cytogenetic analysis usingquantitative, high-sensitivity, fluorescence hybridization.Proc Natl Acad Sci U S A 1986;83(9):2934 – 8.

26. McCullagh P, Nelder JA. Generalized linear models. London:Chapman and Hall, 1989.

27. Chambers JM, Hastie TJ. Statistical models in S. PacificGrove, CA: Brooks/Cole, 1992.

28. Blyth J, Colin R. Approximate binomial confidence limits.J Am Stat Assoc 1989;81:843–55.

29. Oberneder R, Riesenberg R, Kriegmair M, Bitzer U, Klam-mert R, Schneede P, et al. Immunocytochemical detectionand phenotypic characterization of micrometastatic tumourcells in bone marrow of patients with prostate cancer. UrolRes 1994;22(1):3– 8.

30. Pantel K, Aignherr C, Kollerman J, Caprano J, RiethmullerG, Kollerman MW. Immunocytochemical detection of iso-lated tumor cells in bone marrow of patients with un-treated Stage C prostatic cancer. Eur J Cancer 1995;31A(10):1627–32.

31. Bretton PR, Melamed MR, Fair WR, Cote RJ. Detection ofoccult micrometastases in the bone marrow of patients withprostate carcinoma. Prostate 1994;25(2):108 –14.

32. Pantel K, Schlimok G, Braun S, Kutter D, Lindemann F,Schaller G, et al. Differential expression of proliferation-associated molecules in individual micrometastatic carci-noma cells. J Natl Cancer Inst 1993;85(17):1419 –24.

33. Bolton WE, Freeman JW, Mikulka WR, Healy CG, Schmit-tling RJ, Kenyon NS. Expression of proliferation-associatedantigens (PCNA, p120, p145) during the reentry of G0 cellsinto the cell cycle. Cytometry 1994;17(1):66 –74.

34. Mueller P, Weckermann D, Riethmuller G, Schlimok G. De-tection of genetic alterations in micrometastatic cells inbone marrow of cancer patients by fluorescence in situhybridization. Cancer Genet Cytogenet 1996;88(1):8 –16.

35. Litle VR, Warren RS, Moore D II, Pallavicini MG. Molecularcytogenetic analysis of cytokeratin 20 (CK20)-labeled cells inprimary tumors and marrow aspirates from colorectal can-cer patients. Cancer 1997;79:1664 –70.

36. Williams BJ, Jones E, Brothman AR. Homologous centro-mere association of chromosomes 9 and 17 in prostatecancer. Cancer Genet Cytogenet 1995;85(2):143–51.

37. Quijano C, Moore D II, Arthur D, Feusner J, Winter SS,Pallavicini MG. Cytogenetically aberrant cells are present inthe CD34 (1)CD33–38-19- marrow compartment in chil-dren with acute lymphoblastic leukemia. Leukemia 1997;11:1508 –15.

38. Paget S. The distribution of secondary growths in cancer ofthe breast. Lancet 1889;1:571–3.

546 CANCER August 1, 1998 / Volume 83 / Number 3