Analysis of Colorectal Tumor Progression byMicrodissection and Comparative GenomicHybridization
Helen E. Alcock,1 Timothy J. Stephenson,2 Janice A. Royds,1 and David W. Hammond1*1Institute for Cancer Studies, Division of Genomic Medicine, University Medical School, Sheffield, United Kingdom2Department of Pathology, Royal Hallamshire Hospital, Sheffield, United Kingdom
In colorectal cancer (CRC), the progression fromadenoma to carcinoma has been associated with theaccumulation of a number of specific geneticchanges. Such abnormalities have been character-ized at both the whole-chromosome level and themolecular level (e.g., Fearon and Vogelstein, 1990;Bardi et al., 1995; Ried et al., 1996). However, littleinformation is available on the relevance of thesealterations in the advance toward a metastatic phe-notype.
Clinical staging systems attempt to identifyprognostic categories within a disease. In CRC,the most widely used is the Dukes scheme,which originally identified three groups of tu-mors: A, confined to the bowel wall; B, penetrat-ing through the muscularis propria into thesurrounding fat or adventitia; and C, the pres-ence of lymph node metastases. Five-year sur-vival rates for stages A, B, and C are estimated tobe approximately 90, 60, and 25%, respectively(Khankhanian et al., 1977). However, such sys-tems do not identify the subsets of patientswithin the Dukes categories who may have bet-ter or worse prognoses (Kern et al., 1989).
There is an urgent clinical need for more accu-rate prognostic indicators. This could be achievedby an improvement in the detection of clinicallyoccult disease, or the development of more accu-rate methods of assessment of the metastasizingpotential of the primary tumor itself.
The aim of the present investigation was to iden-tify different patterns of copy number change incolorectal tumor progression from adenoma to livermetastasis. To investigate genetic differences asso-ciated with different stages of the disease, we stud-ied a number of histologically distinct sub-regionsfrom a series of archival colorectal tumor samples,in an attempt to reflect the classification categoriescommonly used in staging colorectal cancer. Fur-ther, by the microdissection of discrete tumor sub-regions, we were able to directly compare areas
Supported by: Yorkshire Cancer Research.*Correspondence to: David W. Hammond, Institute for Cancer
Studies, University Medical School, Beech Hill Road, Sheffield, S102RX, UK. E-mail: [email protected]
Received 28 March 2002; Accepted 27 November 2002DOI 10.1002/gcc.10201
within a primary tumor with a corresponding me-tastasis. Genetic alterations could thus be associ-ated with histological sub-regions of a primary tu-mor, and it could be determined whetherhistologically more invasive parts of a primary tu-mor (such as the leading edge) were more similar tothe metastasis than to other primary tumor sub-regions.
MATERIALS AND METHODS
Clinical Material
The initial cohort used in the investigation con-sisted of 20 cases selected on the basis of a pairedprimary colorectal adenocarcinoma and a corre-sponding hepatic metastasis. In one case (LM18),only a biopsy of the primary tumor was available,but in all other cases sections from several archivalblocks of the primary tumor were used.
A system of breaking down the tumor samplesinto histopathological sub-regions was applied, inan attempt to sample tumor from different posi-tions within the bowel wall, within the constraintsof working with two-dimensional tissue sections.This involved microscopic examination of hema-toxylin and eosin–stained adjacent serial sectionsand discussion with a histopathologist before mi-crodissection. These sub-regions were distin-guished according to the histological structures sur-rounding the region of interest. Tumor that wasidentified as being at the bowel lumen and notbeing confluent with tumor that had invaded intothe muscularis layer was termed P; tumor that wassurrounded by the muscularis layers of the bowelwall, but not the serosal fatty tissue, was termed M;tumor microdissected from deposits within serosalfat was termed S. It was not always possible toidentify all these sub-regions. Additional materialin the form of adenomas (LM5, LM7, LM13,LM24, and LM25) and metastases to lymph nodes(LM5, LM7, LM13, LM21, LM23, and LM25) wasalso included in the investigation, where available.
Microdissection and Degenerate OligonucleotidePrimed (DOP)-PCR
Sample microdissection and subsequent PCRamplification were carried out on hematoxylin-stained sections, as previously described (Alcock etal., 1999). DOP-PCR was then carried out by use ofdegenerate primers according to the methods ofSpeicher and colleagues (1993), by use of 5 U ofTaq (Bioline, London, UK).
Negative controls were taken of the buffer cov-ering each section before microdissection, and onlyone microdissected area was sampled per slide.
Comparative Genomic Hybridization
DOP-PCR products were labeled (green) andprepared for comparative genomic hybridization(CGH) analysis, as previously described (Alcock etal., 1999). Reference probes (red) were the samesex as the case from which the sample was takenand were size matched to the labeled DOP-PCRproducts by adjusting the nick translation reaction.A 500-ng sample of reference DNA was hybridizedin each experiment. Hybridization was carried outfor 4 days. A normal male/female experiment (500ng each) was included in each batch of CGH ex-periments. Copy number gains were scored if themean ratio was above or equal to a threshold of 1.15and deletions, if the ratio was below or equal to0.85. When these values were used in analysis ofthe normal male/female controls, no gains or losseswere detected (data not shown). The Y chromo-some, the p arms of acrocentric chromosomes, andcentromeric, telomeric, and heterochromatic re-gions were excluded from analysis because theseare known to be areas in which hybridization isunreliable (Kallioniemi et al., 1994).
RESULTS
Microdissection and DOP-PCR
DOP-PCR product suitable for use as a CGHtest probe was successfully generated in 65 out of79 microdissected samples. Successful CGH wasachieved in 53 out of 65 experiments. In total, datawere obtained for material from 17 cases; adenomasamples (“A”) in 5 cases, primary tumor sub-region(P) in 11 cases, (M) in 7 cases, (S) in 10 cases,lymph node metastasis (LN) in 4 cases, and livermetastasis (L) in 15 cases. Results were obtainedfor a primary tumor sample from case LM18. How-ever, it was not possible to assign a histopatholog-ical sub-region in this case because only a biopsysample was available and the surrounding tissuearchitecture could not be examined. The details foreach case successfully analyzed are shown in Table1. A representative CGH profile from case 14 isshown in Figure 1.
Comparative Genomic Hybridization
Copy number changes involving every chromo-some arm (excluding acrocentric short arms) wereidentified in all of the 53 samples. Thirty-five sam-ples had more deletions than amplifications, 15 had
370 ALCOCK ET AL.
more amplifications than deletions, and three hadequal numbers of each. All samples had at least oneunique change, compared to other samples fromthe same case. The mean number of copy numberchanges per sample was 20. Table 2 illustrates thecopy number changes found in each case. Therewere no significant differences in either the num-ber of copy number changes or the proportion ofthese being unique to a particular sample in thedifferent site categories (Table 3).
The most frequently observed gains in the pri-mary tumors were of Xq (100%), Xp (82%), 4q(65%), 12q (59%), and 5q (53%). The most fre-quently observed losses were of 8p (88%); 10q,15q, and 18q (all 71%); 5p and 22q (both 65%); and4p and 16p (both 59%).
In metastatic material, the most frequent gainswere of Xq (100%), Xp (88%), and 6p and 12q
(both 50%). The most frequently observed losseswere of 8p (94%); 16p (75%); 9p, 10q, 18q, and 22q(all 63%); 15q (56%); and 4p and 4q (both 50%).Adenoma was the only category not to show loss of17p.
In a hypothesis-generating exercise, Fisher’s ex-act test was used to establish the significance ofassociations between specific regions of copy num-ber change and the different histopathological cat-egories. The regions listed below were nominallysignificant before correcting for multiple testing.
Loss of 15q24 was associated with adenoma sam-ples compared to all other samples (P � 0.043).Loss of 12q24.1 was associated with metastases(liver and lymph node) compared to primary tumorand adenoma samples (P � 0.019). Loss of10p13–14 was associated with lymph node metas-tases compared to all other samples (P � 0.026).
TABLE 1. Clinical Details of Analyzed Cases of Colorectal Cancer
Case details Primary tumora Secondary tumor(s)b
Number Agec Sex Site Graded Site(s) Timee
LM1 63 F rectum M liver 3 yearsLM4 54 F cecum M 1 regional LN 0
liver 0LM8 52 F rectum M liver 7 monthsLM9 52 M colon W 1 regional LN 0
liver 0LM10 71 F ascending/transverse colon W liver 0LM11 69 M cecum M-P liver (biopsy) 0LM12 72 M colon M liver 0LM13 65 M colon M 1 regional LN 0
liver 3 monthsLM14 72 M colon M 1 regional LN 0
liver (biopsy) 7 monthsLM15 64 M rectum M 1 regional LN 0
liver 0
LM16 69 F rectosigmoid Movary (1block) 0liver (biopsy) 0
LM17 49 M sigmoid colon/rectum M liver (biopsy) 0LM18 67 M transverse colon (biopsy) M liver (biopsy) 0LM19 54 M colon M-W liver 0LM20 82 F cecum P 1 regional LN 0
liver (biopsy) 0
LM21 47 F sigmoid colon M2 regionalLNs 0liver (biopsy) 0
LM24 44 M rectum M regional LN 0liver (biopsy) 0
aWhere possible, the anatomical site of the tumor has been recorded as accurately as possible but in some cases less precise information was available.“biopsy” indicates that only a biopsy of the lesion was available.bSecondary tumors are listed by site. LN, lymph node. Regional lymph node metastases were defined in the histopathology report. Where the lymphnode metastasis is distant from the primary tumor, the site is recorded in the table.cAge at initial presentation with the primary tumor.dGrade refers to the differentiation of the primary tumor; M, moderately differentiated; P, poorly differentiated; W, well-differentiated.eTime between resection of the primary and secondary tumors.
371CGH OF COLORECTAL TUMOR PROGRESSION
Figu
re1.
CG
Hpr
ofile
ofth
ely
mph
node
met
asta
sis
from
case
LM14
.The
mea
nflu
ores
cenc
era
tiois
show
nby
the
blue
line
and
the
95%
limits
byth
eor
ange
lines
.The
vert
ical
red
and
gree
nlin
esre
pres
ent
cuto
ffva
lues
of0.
85an
d1.
15,r
espe
ctiv
ely.
Reg
ions
ofga
inar
ein
dica
ted
bygr
een
bars
toth
eri
ght,
and
regi
ons
oflo
ssby
red
bars
toth
ele
ft,o
fea
chch
rom
osom
eid
eogr
am.
372 ALCOCK ET AL.
TABLE 2. Results of CGH Analysis of Colorectal Carcinoma*
*A, area of adenoma; P, area of tumor that had not penetrated the muscularis layers; M, area of tumor in which an invasive edge had penetrated themuscularis layers of the bowel wall but not invaded the serosal fatty tissue; S, area of tumor within serosal fat; LN, lymph node metastasis; L, livermetastasis. Regions of gain or loss that are common to or contain sub-regions common to another sample from the same case are shown in bold. Thus,any region in regular text is unique to that sample, among all samples from that particular case. Regions of amplification or deletion in italics are commonto sub-regions in the lymph node metastasis of that case. Regions that are underlined are common to sub-regions in the liver metastasis of that case.
375CGH OF COLORECTAL TUMOR PROGRESSION
Loss of 6q24–25 (P � 0.025) or gain of 4q11–13(P � 0.022) was associated with serosal fat samplescompared to all other samples. Loss of 4q32–34(P � 0.0002) or 22q11–12 (P � 0.045) was associ-ated with S, L, and LN samples compared to A, P,and M samples. Loss of 8q24.1 was associated withnon-metastatic tumor samples compared to LNand L samples (P � 0.037). After correction, loss of4q32–34 remained highly significantly associatedwith S, L, and LN samples.
Analysis of Clonal Relationships Between Samples
A final aim of the investigation was to examinethe results of CGH analysis of different sub-regionsof tumor, to examine the possible clonal relation-ships between samples. Comparing one samplewith another revealed that no two samples wereidentical, although common changes were identi-fied between all samples from a case. Every samplehad unique changes not shared by other samplesfrom the same case, suggesting a high degree oftumor heterogeneity. The model described byKuukasjarvi and colleagues (1997) was used to ex-amine the closeness of the clonal relationships be-tween samples. The calculation estimates theprobability of shared alterations between two sam-ples occurring by chance, by taking into accountthe frequency of each alteration throughout theentire cohort of experiments. The likelihood of twosamples being clonally related thus increases withdecreasing values of P. Probabilities were thereforecalculated for each possible pair of samples fromeach case. These are shown in Table 4.
DISCUSSION
Patterns of Genomic Change
The majority of the samples showed more de-creases than increases in copy number, a findingparalleled in the literature (Paredes-Zaglul et al.,1998; Korn et al., 1999). In this study, the reportedcases showed a mean of 20 copy number changesper case, a relatively high level compared to thatreported by some other authors. For example,means of 12.6 and 11.6 were found in metastaticsamples (Paredes-Zaglul et al., 1998; Al-Mulla etal., 1999) and a mean of 10.2 in primary material(Korn et al., 1999). There are several possible ex-planations for these discrepancies; for example, dif-ferences in the cutoff values used may contributeto variation. The 1.15/0.85 values used in the cur-rent study have been used by some laboratories(De Angelis et al., 1999; Rooney et al., 2001), butother values (e.g., 1.2/0.8 or 1.25/0.75) have beenused elsewhere (Ried et al., 1996; Paredes-Zaglulet al., 1998; Al-Mulla et al., 1999; Korn et al., 1999).However, these sets of values are usually deter-mined by inspection of CGH profiles from normal/normal experiments in each laboratory and, as such,are more likely to reflect local experimental proto-cols than significant differences in findings. Addi-tionally, discrepancies may have been introducedby the amplification of DNA extracted from fixedmaterial. We previously established that whenCGH experiments between DNA from fresh tissueand amplified DNA from the same material (pre-viously paraffin embedded) were carried out, con-cordant results were obtained (Alcock et al., 1999).There has also been some debate in the literatureas to whether DOP-amplified CGH test probesshould be paired with similarly amplified referenceDNA. Huang and colleagues (2000) found thatamplified reference DNA gave a significantly moreaccurate CGH profile, whereas Larsen and col-leagues (2001) obtained better results by use ofamplified test DNA vs. un-amplified reference. Inthe current study, size-matched un-amplified ref-erence DNA was used. On balance, it is unlikelythat serious inaccuracies were introduced by thisapproach.
The DNA from most colorectal cancer samplesin the literature has been extracted from relativelylarge tissue samples, which contain a number ofdifferent cell types. It is therefore possible that anumber of individual abnormalities might be ob-scured, either by being part of a clone below thelevel of detection, or by complementation of anabnormality by another clone. The DNA with
TABLE 3. Mean Copy Number Changes and MeanPercentage of Unique Changes According to the Site of the
Microdissected Sample
SiteaMean total number
of changesMean percentageunique changesb
A 22 42P 18 43M 21 50S 21 52LN 25 45L 19 41
aA, area of adenoma; P, area of tumor that had not penetrated themuscularis layers; M, area of tumor in which an invasive edge hadpenetrated the muscularis layers of the bowel wall but not invaded theserosal fatty tissue; S, area of tumor within serosal fat; LN, lymph nodemetastasis; L, liver metastasis.bCopy number changes that were not shared to any degree by anyother sample from the same case.
376 ALCOCK ET AL.
which CGH was carried out in this study was ob-tained from a limited number of cells. This wouldprovide greater opportunities for the detection ofabnormalities present in small clones within a het-erogeneous population. Striking confirmation ofthis intra-tumor heterogeneity was previouslyshown by Klein and colleagues (1999), who carriedout CGH on individual cells from a single tumor.
Differences in copy number abnormalities wereobserved from cell to cell. Another possibility isthat this series of advanced tumors is actually morecomplex because of the length of time they havebeen growing. Conventional models suggest thatolder tumors will have acquired more genetic ab-normalities than younger ones. It is also possiblethat in some CGH experiments with higher than
TABLE 4. Likelihood of a Clonal Relationship for Different Sample Combinations Within Each Case
Case Samples CRa Case Samples CRa
LM1 P and S P � 0.05 LM15 All (M, S, and L) —M and S P � 0.05M and L P � 0.05S and L P � 0.05
LM4 All — LM16 All (P, M, and L) —A and P P � 0.001 P and M P � 0.05A and L P � 0.001 P and L P � 0.05P and L P � 0.001 M and L P � 0.05
LM8 All (A, P, S, and L) — LM17 All (A, P, S, and L) —A and P P � 0.05 A and P P � 0.05A and S P � 0.05 A and S P � 0.001A and L P � 0.05 A and L P � 0.05P and S P � 0.05 P and S P � 0.05P and L P � 0.05 P and L P � 0.05S and L P � 0.05 S and L P � 0.05
LM9 S and L P � 0.05 LM18 Primary tumor and L P � 0.001LM10 P, M, and L — LM19 All (P, M, and L) —
P and M P � 0.05 P and M P � 0.001P and L P � 0.001 P and L P � 0.001M and L P � 0.05 M and L P � 0.001
LM11 All (M, S, and LN) — LM20 All (P, S, LN, and L) —M and S P � 0.05 P and S P � 0.05M and LN P � 0.05 P and LN P � 0.05S and LN P � 0.05 P and L P � 0.05
S and L P � 0.05S and L P � 0.05LN and L P � 0.05
LM12 All (P, S, and L) — LM21 All (M, LN, and L) —P and S P � 0.05 M and LN P � 0.05P and L P � 0.05 M and L P � 0.05S and L P � 0.05 LN and L P � 0.05
LM13 All (A, M, S, andL)
— LM24 All (A, P, and L) —
A and M P � 0.05 A and P P � 0.001A and S P � 0.05 A and L P � 0.05A and L P � 0.05 P and L P � 0.001M and S P � 0.05M and L P � 0.05S and L P � 0.05
LM14 All (P, S, LN, andL)
—
P and S P � 0.05P and LN P � 0.001P and L P � 0.05S and LN P � 0.001S and L P � 0.05LN and L P � 0.001
aCR, probability that the respective sample pairs are related by chance alone.
377CGH OF COLORECTAL TUMOR PROGRESSION
usual background noise, multiple hits on a partic-ular chromosome arm in fact represent a singleevent. However, there is little evidence of thisoccurring in the control experiments carried out oncell line material.
Overall, the CGH results from this set of exper-iments were in general agreement with those foundin colorectal cancer by other investigators. Indeed,some copy number changes (e.g., gain of chromo-some 7) appear to be common to a range of differ-ent solid tumors (Knuutila et al., 1998, 1999). Thiscoincidence of cancer chromosome abnormalitieshas also been established in conventional cytoge-netic studies (Mertens et al., 1997). A number ofthe changes we report here have been found tooccur at high levels throughout the literature. Gainof Xq and Xp and loss of 4p, 4q, 8p, 15q, 18q, and22q are found frequently in CGH studies of colo-rectal cancer (Ried et al., 1996; Meijer et al., 1998;Paredes-Zaglul et al., 1998; Al-Mulla et al., 1999;De Angelis et al., 1999; Korn et al., 1999; Araganeet al., 2001; Nakao et al., 2001; Rooney et al., 2001).The regular detection of these changes in bothCGH and in conventional cytogenetic studies(Bardi et al., 1995) has led to a number of investi-gations aimed at elucidating the genes involved.The 18q21 locus, site of SMAD4 and other genes,has been studied in CRC for a number of years(Fearon et al., 1990; Moskaluk and Kern, 1996;Thiagalingam et al., 1996; Woodford-Richens et al.,2001). Allelic loss at different regions on chromo-some 4 has been found to correlate with the degreeof aggressiveness in colorectal tumors (Arribas etal., 1999; De Angelis et al., 2001; Shivapurkar et al.,2001).
A great deal of recent work has focused on theshort arm of chromosome 8. A small interval at8p21 has been identified as the site of one potentialtumor-suppressor gene (Lerebours et al., 1999).The FEZ1 gene at chromosome band 8p22 hasbeen implicated in a number of tumors includingcolorectal carcinoma (Croce, 1999), and two regionsof deletion at 8p11.2–p12 have been considered toconfer “different and independent roles in thepathogenesis of colorectal cancer” (Chughtai et al.,1999). Others have studied the THBS1 gene at15q21.1 (Park et al., 2000) and a region of allelicloss at 22q13 (Castells et al., 1999).
Some changes seen at high frequency in thisseries are found at much lower rates in the litera-ture. These include gains at 12q and 5q and lossesat 5p, 9p, 10q, and 16p. These differences may beattributed to the more “searching” techniques ap-plied here, especially in terms of multiple sam-
pling, which does not appear to have been appliedto CRC elsewhere. One abnormality repeatedlyreported at high levels in the literature (Ried et al.,1996; Al-Mulla et al., 1999; De Angelis et al., 1999;Korn et al., 1999; Aragane et al., 2001; Nakao et al.,2001; Rooney et al., 2001), but seen only infre-quently in this series, is a gain on the long arm ofchromosome 20. This variation, presumably predi-cated on differences in material and methodology,is not unique. For example, loss of 8p, commonlyfound at high incidence by many groups, is re-ported only rarely in some recent studies (Araganeet al., 2001; Nakao et al., 2001; Rooney et al., 2001).
A number of loci were found to be associatedwith particular histopathological categories. Of par-ticular interest are the loss of 12q24.1, significantlyassociated with metastases, and the loss of 4q32–34, significantly associated with tumor-penetratingserosal fat and metastases. Although attempts havebeen made to identify regions of the genome spe-cifically associated with advanced disease and me-tastasis in CRC by CGH (Ried et al., 1996), morerecent work has failed to find such a connection(Al-Mulla et al., 1999; Korn et al., 1999). It may bethat the sampling of cells from a defined, restrictedlocus allows the detection of specific geneticchanges that would otherwise remain hidden.
Clonal Relationships Between MicrodissectedSamples From Individual Cases
There was wide variation in the data that re-vealed a highly complex set of relationships be-tween samples and the degree to which individualsamples within each case appeared to be clonallyrelated. In the majority of cases, clonal relation-ships were likely to be close. However, in somecases (e.g., LM12 and LM15), the majority ofclones appeared unlikely to be closely related.Even in those cases with apparently closely relatedclones, the establishment of a simple series of the-oretical common precursors proved problematical.This finding was also recently reported by anotherlaboratory (Al-Mulla et al., 1999). One possibleexplanation for these results is that there are earlypredisposing changes that then permit a range ofunrelated genetic abnormalities to occur at differ-ent tissue locations. This is an unlikely explana-tion, however, because in the majority of casesthere is a close clonal relationship between indi-vidual samples.
It is also possible that the complexity of theseresults is because CRCs with known liver metas-tases were selected for analysis. These tumors mayhave been undergoing waves of clonal expansion
378 ALCOCK ET AL.
before detection of the disease. Most of thechanges that have led to malignancy may haveoccurred in the distant ancestors of the cellspresent in the resected specimen. Additional ran-dom changes have then developed over time atdifferent rates in different parts of each tumor.
Certain changes observed in the current investi-gation might be associated with particular tumor/tissue behavior. These data do not provide strongevidence for a sequential pathway of copy numberchanges running from adenoma, through increas-ingly invasive tumor and on to metastasis, as pre-viously suggested by the CGH findings of Ried andcolleagues (1999). However, it is possible that se-quential changes had taken place but that evidencefor this has been masked by numerous other ge-netic changes over the evolutionary lifetime of thetumor.
The use of new technologies has made it possi-ble to study genetic changes in different parts of anindividual tumor and, overall, this study has re-vealed a high degree of genetic heterogeneity inCRC. Some of the lesions we have identified mayhave specific links to particular stages of the dis-ease. However, the ability to tap into this diversityleaves open the question of which changes, oncedetected, are significant in terms of prognosis. Inmost cases, only a small fraction of any tumor willbe analyzed, making comparisons across studieshighly problematical. An increased understandingof primary tumor heterogeneity in the context ofmetastatic markers could lead to improved use ofboth existing and future prognostic indicators.
Al-Mulla F, Keith WN, Pickford IR, Going JJ, Birnie GD. 1999.Comparative genomic hybridization analysis of primary colorectalcarcinomas and their synchronous metastases. Genes Chromo-somes Cancer 24:306–314.
Aragane H, Sakakura C, Nakanishi M, Yasuoka R, Fujita Y, Tan-iguchi H, Hagiwara A, Yamaguchi T, Abe T, Inazawa J, YamagishiH. 2001. Chromosomal aberrations in colorectal cancers and livermetastases analyzed by comparative genomic hybridization. Int JCancer 94:623–629.
Arribas R, Risques RA, GonzalezGarcia I, Masramon L, Aiza G,Ribas M, Capella G, Peinado MA. 1999. Tracking recurrent quan-titative genomic alterations in colorectal cancer: allelic losses inchromosome 4 correlate with tumor aggressiveness. Lab Invest79:111–122.
Bardi G, Sukhikh T, Pandis N, Fenger C, Kronborg O, Heim S.1995. Karyotypic characterization of colorectal adenocarcinomas.Genes Chromosomes Cancer 12:97–109.
Castells A, Ino Y, Louis DN, Ramesh V, Gusella JF, Rustgi AK.1999. Mapping of a target region of allelic loss to a 0.5-cM intervalon chromosome 22q13 in human colorectal cancer. Gastroenter-ology 117:831–837.
Chughtai SA, Crundwell MC, Cruickshank NRJ, Affie E, ArmstrongS, Knowles MA, Takle LA, Kuo M, Khan N, Phillips SMA,Neoptolemos JP, Morton DG. 1999. Two novel regions of inter-
stitial deletion on chromosome 8p in colorectal cancer. Oncogene18:657–665.
Croce CM. 1999. The FEZ1 gene at chromosome 8p22 encodes aleucine-zipper protein, and its expression is altered in multiplehuman tumors. Proc Natl Acad Sci USA 96:3928–3933.
De Angelis PM, Clausen OPF, Schjølberg A, Stokke T. 1999.Chromosomal gains and losses in primary colorectal carcinomasdetected by CGH and their associations with tumour DNAploidy, genotypes and phenotypes. Br J Cancer 80:526–535.
De Angelis PM, Stokke T, Beigi M, Mjaland O, Clausen OPF. 2001.Prognostic significance of recurrent chromosomal aberrations de-tected by comparative genomic hybridization in sporadic colorec-tal cancer. Int J Colorectal Dis 16:38–45.
Fearon ER, Vogelstein B. 1990. A genetic model for colorectaltumourigenesis. Cell 61:759–767.
Fearon ER, Cho KR, Nigro JM, Kern SE, Simons JW, Ruppert JM,Hamilton SR, Preisinger AC, Thomas G, Kinzler KW, VogelsteinB. 1990. Identification of a chromosome 18q gene that is altered incolorectal cancers. Science 247:49–56.
Huang Q, Schantz SP, Rao PH, Mo J, McCormick SA, ChagantiRSK. 2000. Improving degenerate oligonucleotide primed PCR-comparative genomic hybridization for analysis of DNA copynumber changes in tumors. Genes Chromosomes Cancer 28:395–403.
Kallioniemi OP, Kallioniemi A, Piper J, Waldman FM, Gray JW,Pinkel D. 1994. Optimizing comparative genomic hybridizationfor analysis of DNA sequence copy number changes in solidtumours. Genes Chromosomes Cancer 10:231–243.
Kern SE, Fearon ER, Kasper BA, Termette WF, Enterline JP,Leppert M, Nakamura Y, White R, Vogelstein B, Hamilton SR.1989. Allelic loss in colorectal carcinoma. J Am Med Assoc 261:3099–3103.
Khankhanian N, Mallight GM, Russel WO, Schimel M. 1977. Prog-nostic significance of vascular invasion in colorectal cancer ofDukes’ B class. Cancer 39:1195–1200.
Klein CA, Schmidt-Kittler O, Schardt JA, Pantel K, Speicher MR,Riethmuller G. 1999. Comparative genomic hybridization, loss ofheterozygosity, and DNA sequence analysis of single cells. ProcNatl Acad Sci USA 96:4494–4499.
Knuutila S, Bjorkqvist AM, Autio K, Tarkkanen M, Wolf M, MonniO, Szymanska J, Larramendy ML, Tapper J, Pere H, El-Rifai W,Hemmer S, Wasenius VM, Vidgren V, Zhu Y. 1998. DNA copynumber amplifications in human neoplasms: review of compara-tive genomic hybridization studies. Am J Pathol 152:1107–1123.
Knuutila S, Aalto Y, Autio K, Bjorkqvist AM, El-Rifai W, HemmerS, Huhta T, Kettunen E, Kiurukuhlefelt S, Larramendy ML,Lushnikova T, Monni O, Pere H, Tapper J, Tarkkanen M, VarisA, Wasenius VM, Wolf M, Zhu Y. 1999. DNA copy number lossesin human neoplasms. Am J Pathol 155:683–694.
Korn WM, Yasutake T, Kuo WL, Warren RS, Collins C, Tomita M,Gray J, Waldman FM. 1999. Chromosome arm 20q gains andother genomic alterations in colorectal cancer metastatic to liver,as analyzed by comparative genomic hybridization and fluores-cence in situ hybridization. Genes Chromosomes Cancer 25:82–90.
Kuukasjarvi T, Karhu R, Tanner M, Kahkonen M, Schaffer A,Nupponen N, Pennanen S, Kallioniemi A, Kallioniemi OP, IsolaJ. 1997. Genetic heterogeneity and clonal evolution underlyingdevelopment of asynchronous metastasis in human breast cancer.Cancer Res 57:1597–1604.
Larsen J, Ottensen AM, Lundsteen C, Leffers H, Larsen JK. 2001.Optimization of DOP-PCR amplification of DNA for high-reso-lution comparative genomic hybridization analysis. Cytometry 44:317–325.
Lerebours F, Olschwang S, Thuille B, Schmitz A, Fouchet P, Lau-rent-Puig P, Boman F, Flejou JF, Monges G, Pavaf F, Bedossa P,Sabourin JC, Salmon RJ, Parc R, Thomas G. 1999. Deletionmapping of the tumor suppressor locus involved in colorectalcancer on chromosome band 8p21. Genes Chromosomes Cancer25:147–153.
Meijer GA, Hermsen MAJA, Bank JPA, van Diest PJV, MeuwissenSGM, Belien JAM, Hoovers JMN, Joenje H, Snijders PJF, Wal-boomers JMM. 1998. Progression from colorectal adenoma tocarcinoma is associated with non-random chromosomal gains asdetected by comparative genomic hybridisation. J Clin Pathol51:901–909.
Mertens F, Johansson B, Hoglund M, Mitelman F. 1997. Chromo-somal imbalance maps of malignant solid tumors: a cytogeneticsurvey of 3185 neoplasms. Cancer Res 57:2765–2780.
Moskaluk CA, Kern SE. 1996. Cancer gets Mad: DPC4 and other
379CGH OF COLORECTAL TUMOR PROGRESSION
TGF� pathway genes in human cancer. Biochim Biophys Acta1288:M31–M33.
Nakao K, Shibusawa M, Ishihara A, Yoshizawa H, Tsunoda A, KusanoM, Kurose A, Makita T, Sasaki K. 2001. Genetic changes in colorec-tal carcinoma tumors with liver metastases analyzed by comparativegenomic hybridization and DNA ploidy. Cancer 9:721–726.
Paredes-Zaglul A, Kang JJ, Essig Y-P, Mao W, Irby R, Wloch M,Yeatman TJ. 1998. Analysis of colorectal cancer by comparativegenomic hybridization: evidence for the induction of the meta-static phenotype by loss of tumour suppressor genes. Clin CancerRes 4:879–886.
Park WS, Park JY, Oh RR, Yoo NJ, Lee SH, Shin MS, Lee HK, HanSY, Kim S, Kim SY, Choi C, Kim PJ, Oh ST, Lee JY. 2000. Adistinct tumor suppressor gene locus on chromosome 15q21.1 insporadic form of colorectal cancer. Cancer Res 60:70–73.
Ried T, Knutzen R, Steinbeck R, Blegen H, Schrock E, HeselmeyerK, du Manoir S, Auer G. 1996. Comparative genomic hybridiza-tion reveals a specific pattern of chromosomal gains and lossesduring the genesis of colorectal tumors. Genes ChromosomesCancer 15:234–245.
Ried T, Heselmeyer K, Haddad K, Blegen H, Schrock E, Auer G.1999. Genomic changes defining the genesis, progression, andmalignancy potential in solid human tumors: a phenotype geno-type correlation. Genes Chromosomes Cancer 25:195–204.
Rooney PH, Boonsong A, McKay JA, Marsh S, Stevenson DA,Murray GI, Curran S, Haites NE, Cassidy J, McLeod HL. 2001.Colorectal cancer genomics: evidence for multiple genotypeswhich influence survival. Br J Cancer 85:1492–1498.
Shivapurkar N, Maitra A, Milchgrub S, Gazdar AF. 2001. Deletionsof chromosome 4 occur early during the pathogenesis of colorectalcarcinoma. Hum Pathol 32:169–177.
Speicher MR, du Manoir S, Schrock E, Holtgreve-Grez H, SchoellB, Lengauer C, Cremer T, Ried T. 1993. Molecular cytogeneticanalysis of formalin fixed, paraffin-embedded solid tumors bycomparative genomic hybridization after universal DNA-amplifi-cation. Hum Mol Genet 2:1907–1914.
Thiagalingam S, Lengauer C, Leach FS, Scutte M, Hahn SA, Over-hausrer J, Willson JKV, Markowitz S, Hamilton SR, Kern SE,Kinzler KW, Vogelstein B. 1996. Evaluation of candidate tumoursuppressor genes on chromosome 18 in colorectal cancers. NatGenet 13:343–346.
Woodford-Richens KL, Rowan AJ, Gorman P, Halford S, BicknellDC, Wasan HS, Roylance RR, Bodmer WF, Tomlinson IPM.2001. SMAD4 mutations in colorectal cancer probably occur be-fore chromosomal instability, but after divergence of the micro-satellite instability pathway. Proc Natl Acad Sci USA 98:9719–9723.
