Human Genetics

, Volume 121, Issue 6, pp 729–735

Genome-wide linkage scan of prostate cancer Gleason score and confirmation of chromosome 19q

Authors

  • Daniel J. Schaid
    • Division of BiostatisticsMayo Clinic
  • Janet L. Stanford
    • Division of Public Health SciencesFred Hutchinson Cancer Research Center
    • Department of Epidemiology, School of Public Health and Community MedicineUniversity of Washington
  • Shannon K. McDonnell
    • Division of BiostatisticsMayo Clinic
  • Miia Suuriniemi
    • Cancer Genetics Branch, National Human Genome Research InstituteNational Institutes of Health
  • Laura McIntosh
    • Division of Public Health SciencesFred Hutchinson Cancer Research Center
  • Danielle M. Karyadi
    • Cancer Genetics Branch, National Human Genome Research InstituteNational Institutes of Health
  • Erin E. Carlson
    • Division of BiostatisticsMayo Clinic
  • Kerry Deutsch
    • Institute for Systems Biology
  • Marta Janer
    • Institute for Systems Biology
  • Lee Hood
    • Institute for Systems Biology
    • Cancer Genetics Branch, National Human Genome Research InstituteNational Institutes of Health
Original Investigation

DOI: 10.1007/s00439-007-0368-5

Cite this article as:
Schaid, D.J., Stanford, J.L., McDonnell, S.K. et al. Hum Genet (2007) 121: 729. doi:10.1007/s00439-007-0368-5

Abstract

Despite evidence that prostate cancer has a genetic etiology, it has been extremely difficult to confirm genetic linkage results across studies, emphasizing the large extent of genetic heterogeneity associated with this disease. Because prostate cancer is common—approximately one in six men will be diagnosed with prostate cancer in their life—genetic linkage studies are likely plagued by phenocopies (i.e., men with prostate cancer due to environmental or lifestyle factors), weakly penetrant alleles, or a combination of both, making it difficult to replicate linkage findings. One way to account for heterogeneous causes is to use clinical information that is related to the aggressiveness of disease as an endpoint for linkage analyses. Gleason grade is a measure of prostate tumor differentiation, with higher grades associated with more aggressive disease. This semi-quantitative score has been used as a quantitative trait for linkage analysis in several prior studies. Our aim was to determine if prior linkage reports of Gleason grade to specific loci could be replicated, and to ascertain if new regions of linkage could be found. Gleason scores were available for 391 affected sib pairs from 183 hereditary prostate cancer pedigrees as part of the PROGRESS study. Analyzing Gleason score as a quantitative trait, and using microsatellite markers, suggestive evidence for linkage (P-value ≤ 0.001) was found on chromosomes 19q and 5q, with P-values ≤ 0.01 observed on chromosomes 3q, 7q, and 16q. Our results confirm reports of Gleason score linkage to chromosome 19q and suggest new loci for further investigation.

Introduction

Although considerable evidence suggests that prostate cancer, the most frequent of all non-cutaneous cancers in men (Jemal et al. 2006) has a genetic etiology, and a number of linkage studies have suggested numerous chromosomal regions, confirmatory linkage studies have been inconsistent. To date, at least 15 genome-wide linkage scans for prostate cancer have been performed, as reviewed elsewhere (Easton et al. 2003; Matsui et al. 2004; Schaid 2006; Camp et al. 2005). Overall, the cumulative data across all studies show some suggestive evidence for linkage to almost every chromosome. Although some results partially overlap between studies, there is no evidence for a single, or a common few, major susceptibility loci.

Several studies, both functional and epidemiological, provide strong support for putative loci, while other studies suggest that the roles of reported putative loci are small (Ostrander and Stanford 2000; Schaid 2006). In total, these studies illustrate the difficulty in finding consistent linkage results across different studies, and emphasize the etiological heterogeneity of the disease.

From a clinical perspective, it is important to understand the predisposition to an aggressive form of the disease; such cancers are life-threatening and should be the focus of therapeutic interventions and development. Gleason grade indicates prostate tumor differentiation (ranging from 1 for well-differentiated to 5 for poorly differentiated cellular architecture) and is a measure of disease aggression (Epstein et al. 1993; Lerner et al. 1996). The two predominant Gleason grades in the tumor are added together to yield the final Gleason score. Although the score can range from 2 to 10, most tumors are scored in a much more narrow range, most commonly six and seven (Sengupta et al. 2006). Using Gleason score as a quantitative trait in linkage analyses, Witte et al. (Witte et al. 2000) found strong evidence for linkage on chromosomes 5q, 7q, and 19q. The findings on chromosomes 7q and 19q were further strengthened in subsequent analyses of fine-mapping markers using the same sib-pairs (Neville et al. 2002, 2003). The findings for chromosome 19q were further confirmed by an independent study (LOD = 3.9 for chromosome 19q) (Slager et al. 2003). This latter study also found a weak linkage signal on chromosome 5q (LOD = 1.4), but did not find evidence for linkage on chromosome 7. In addition, chromosomes 4p and 15q were found to have LOD scores near 2.0 or larger (Slager et al. 2003). In a second independent replicate dataset, Witte et al. (Witte et al. 2003) reported weak linkage for chromosomes 7q (LOD = 2.1), 5p (LOD = 1.6), and 9q (LOD = 1.2). A recent report from the University of Michigan found weak evidence for linkage to chromosome 6q (LOD = 2.1), as well as to chromosomes 1p (LOD = 1.8) and 5p (LOD = 2.1) (Slager et al. 2006). Recently, Pal et al. (Pal et al. 2006) further examined the role of chromosome 19q by evaluating the association of single nucleotide polymorphisms (SNPs) in the HEPSIN gene with both the risk of prostate cancer and Gleason score among affected men. This gene is located on 19q11-13.2, and its 13 exons encode a type II transmembrane cell surface serine protease that is over-expressed in prostate cancer, and so makes an ideal candidate to explain linkage of chromosome 19q with prostate cancer. After evaluating 11 SNPs, they found four that had allele frequencies different between prostate cancer cases and controls with P-value < 0.05, and a major haplotype that was significantly more frequent among cases versus controls.

Because of the somewhat consistent findings of Gleason score linkage to chromosome 19q, and suggestions for other chromosomes, we performed a genome-wide linkage analysis of Gleason score as a quantitative trait using microsatellite markers.

Materials and methods

Pedigree collection

The 254 hereditary prostate cancer (HPC) families participating in the Prostate Cancer Genetic Research Study (PROGRESS) and who were included in an initial genome-wide scan (Janer et al. 2003) formed the basis of this analysis; six of the above were excluded because of missing clinical or genetic data. Because of concerns about genetic differences across ethnic groups, such as marker allele frequencies and genetic susceptibilities, and too few pedigrees with non-Caucasian ancestry, we excluded an additional 11 pedigrees: African American (n = 5) or other (n = 6) heritage. After these exclusions, there were 237 pedigrees of Caucasian ancestry which were characterized by either three or more men affected with prostate cancer, three generations of men affected by prostate cancer, or two first-degree relatives diagnosed with prostate cancer before age 65. All prostate cancer survivors, unaffected male relatives over age 40, and selected female relatives were invited to join the study. Participants were asked to complete a baseline survey and provide a blood sample, and permission to access medical records was requested from all men with prostate cancer. Study procedures and materials were approved by the Institutional Review Board of the Fred Hutchinson Cancer Research Center.

The average age at diagnosis per pedigree ranged from 47 to 78 years, with 35 pedigrees having an average age of 58 years or younger. The number of affected men per pedigree ranged from 2 to 11, with an average of 3.2 per pedigree.

Clinical data

Medical records were used to validate self-reported prostate cancers and to abstract detailed clinical data. Of the 784 medical records received, 100% confirmed the prostate cancer diagnosis. Data on Gleason score (biopsy, TURP, and/or radical prostatectomy), stage of disease (localized, regional, or distant; clinical and pathological if available), level of prostate-specific antigen (PSA) at diagnosis, and cause of death (prostate cancer-specific or other) and age at death were abstracted onto a standardized form by an abstractor trained to code prostate cancer medical records from the Seattle-Puget Sound SEER registry. Clinical data were abstracted for all men with prostate cancer; surgical pathology data were additionally abstracted for men who underwent radical prostatectomy. Data was doubly entered into the database.

Genotypes

Genotypes measured on 441 microsatellite markers were available from a completed genomic scan as described previously (Janer et al. 2003). The validity of the genotypes was evaluated by checking Mendelian inheritance, using Pedcheck (O’Connell and Weeks 1998) and locally written procedures. As an additional check, we paired all markers, to examine the percentage of identical genotypes across subjects, to ensure that no markers were erroneously duplicated through mislabeling. Likewise, we paired all subjects to examine the percentage of identical genotypes across all markers, and to be sure that all monozygotic twins were accurately identified. Additional genotypes that were likely erroneous, determined by the default double-recombinant error-detection option of Merlin (Abecasis et al. 2002) were removed. This step resulted in the removal of 2,246 (0.3%) of the 667,530 genotypes available for analysis.

Four HEPSIN SNPs (rs2451996, rs1688043, rs2305746, and rs2305747) were selected for analysis in the present study based on the findings of Pal and colleagues (Pal et al. 2006). Two (rs2305746 and rs2305747) were reportedly associated with prostate cancer, one (rs1688043) showed a specific association with Gleason score, and one (rs2451996) showed no association with either prostate cancer or Gleason grade (Pal et al. 2006). Genotyping of the four HEPSIN SNPs was performed using the TaqMan SNP Genotyping Assay (Applied Biosystems, Foster City, CA, USA). Genomic DNA previously extracted from peripheral blood lymphocytes was amplified using an Applied Biosystems GeneAmp 9700 thermocycler in the presence of PCR primers and fluorescently labeled probes obtained from Applied Biosystems (Foster City, CA, USA). The amplification conditions consisted of an initial denaturation for 10 min at 95°C, followed by 40 cycles of 15 s at 95°C and 1 min at 60°C. After the amplification the results were scored using the 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) and the alleles were automatically called using Sequence Detection System software.

Statistical analyses

Because the majority of relative pairs in our pedigrees were sib-pairs, and because the clinical information was likely to be most consistent among men from the same generation, we focused our linkage analyses of Gleason score using only affected sib-pairs. The statistical analyses were based on the Haseman–Elston (HE) method, which regresses a function of the traits for a pair of sibs on the estimated proportion of alleles shared identical-by-descent (IBD) for the pair. Choices of the trait function included the difference-squared (original HE) and the mean-corrected cross-product (revised HE) (Elston et al. 2000).

The sample mean of the Gleason score was used for the revised HE function. The dependence of multiple brother-pairs from the same family was accounted for in the analysis by weighted least squares, as implemented in the S.A.G.E. software (S.A.G.E. 2004). To ensure the most accurate estimates, multipoint IBD sharing probabilities, based on Merlin software (Abecasis et al. 2002), were computed at each marker position. Allele frequencies were estimated using all subjects in the full set of 237 HPC pedigrees. Each of the trait functions (original HE and revised HE) was analyzed separately at each marker position. A t-test was used to test the null hypothesis of no linkage. Any test that resulted in an asymptotic P-value of 0.05 or less was verified by calculating a permutation P-value. The number of permutations was chosen such that the empirical P-value was within 20% of its true P-value with 95% confidence. Permutation P-values are presented in the text and accompanying tables while the plots are based on nominal P-values (to avoid computing permutations for each position for each chromosome).

To evaluate the association of SNPs in the HEPSIN gene with Gleason score, we used linear regression to regress Gleason score on a count of the number of rare alleles for each SNP. To account for genotype correlations among cases from the same family, generalized estimating equations were used (Zeger et al. 1988) assuming an exchangeable working correlation matrix. To evaluate whether the HEPSIN SNPs explained our observed linkage on chromosome 19, we used the residuals from regressions of Gleason score on two genotype indicators (the most common genotype was used as a baseline) for each SNP as quantitative traits in our linkage analyses, for both the original and revised HE methods.

Results

Gleason scores were obtained from radical prostatectomy specimens for 346 (58%) men and from needle biopsies for the remaining 248 men. The mean Gleason score was 5.9 (range 2–10). The distribution of Gleason scores, presented in Table 1, illustrates that the majority of men (78%) had scores ranging from 5 to 7. The distribution of relative-pair types with available Gleason scores for the 237 hereditary PC pedigrees is shown in Table 2. Fifty-four HPC pedigrees were uninformative, having no affected brother pairs with available Gleason scores. The remaining 183 HPC pedigrees included 391 brother pairs available for analysis.
Table 1

Distribution of Gleason scores among men used in linkage analyses

Gleason score

%

2

1

3

3

4

9

5

28

6

25

7

25

8

7

9

3

10

0.2

Table 2

Affected relative pair types in the 237 HPC pedigrees

No. of affected mena

No. of pedigrees

No. of sibling pairs

No. of parent/offspring pairs

No. of cousin pairs

No. of Avuncular pairs

0

1

2

3

4

6

8

10

0

1

2

3

0

1

2–4

5–7

8+

0

1

3

5–8

<2

33

33

       

33

   

33

    

33

   

2

93

18

75

      

93

   

87

5

1

  

83

10

  

3

78

3

22

 

53

    

67

9

2

 

69

2

7

  

64

9

5

 

4

24

 

4

1

10

 

9

  

22

1

1

 

18

1

3

2

 

15

 

9

 

5

7

 

1

 

1

1

3

1

 

5

1

 

1

3

 

2

1

1

1

1

3

2

6

2

 

1

     

1

1

1

  

1

  

1

   

2

 

Italicized cells indicate uninformative pedigrees for HE sib-pair regression analyses

aAffected men who had both Gleason score and genotypes available

The linkage results for the original and revised HE methods are presented in Fig. 1. Each chromosome’s plot has the −log10 (P-value), denoted lgP, on the vertical axis. The dotted horizontal line indicates where the P-value = 0.001 (lgP = 3.0). Results were consistent between the original and revised HE methods. The largest linkage signals were observed on chromosomes 5q (permutation P-value = 0.0010) and 19q (permutation P-value = 0.0014). Other interesting, yet weaker, linkage signals (P-value ≤ 0.01) were found on chromosomes 3q, 7q, and 16q. See Table 3 for details of the positions of the largest linkage signals and the flanking markers.
https://static-content.springer.com/image/art%3A10.1007%2Fs00439-007-0368-5/MediaObjects/439_2007_368_Fig1_HTML.gif
Fig. 1

Linkage results using Gleason score as a quantitative trait. Solid lines are for the original HE analysis (i.e., squared trait difference) and dotted lines are for the revised HE analysis (trait mean-corrected product). For the y-axis, lgP = −log10 (P-value)

Table 3

Interesting regions of linkage (asymptotic P-value ≤ 0.01)

Chromosome

Permutation P-value

Flanking markers

Original HE (cM)

Revised HE (cM)

3q24-q25.1

0.0094 (153.4)

0.0033 (156.3)

D3S1744 (153.4)

D3S1746 (160.6)

5q35.3

0.0010 (198.6)

0.0185 (205.1)

AAT013 (194.2)

D5S408 (205.1)

7q21.11-q21.3

0.0623 (97.9)

0.0107 (97.9)

D7S820 (97.9)

D7S821 (106.3)

16q24.1

0.0149 (115.4)

0.3466 (115.4)

D16S402 (110.8)

D16S539 (122.4)

19q13.31-q13.41

0.0060 (88.0)

0.0014 (75.7)

D19S178 (70.5)

D19S246 (83.4)

   

D19S246 (83.4)

D19S601 (91.1)

Because the scoring of Gleason grade by pathologists has tended to migrate toward higher scores over time (Albertsen et al. 2005; Thompson et al. 2005), we were concerned that Gleason scores of the same value could represent different levels of disease aggression. However, as illustrated in Fig. 2, our data do not show a linear trend of the average Gleason score changing over years of diagnosis. To further explore this, we examined the relationship of Gleason score with stage (regional/distant versus local) and PSA at diagnosis (PSA < 20 versus 20 + ng/ml). The distributions of Gleason scores according to stage and PSA illustrate a slight trend for higher Gleason scores among men with regional/distant stage (Fig. 3a) or PSA > 20 (Fig. 3b). Based on linear regression, stage explained about 8% of the variation of Gleason scores (P < 0.0001); PSA explained about 2% of the variation (P = 0.002) (14.5% of men with Gleason score had missing PSA values).
https://static-content.springer.com/image/art%3A10.1007%2Fs00439-007-0368-5/MediaObjects/439_2007_368_Fig2_HTML.gif
Fig. 2

Average Gleason score (±standard error) according to year of diagnosis (5-year intervals)

https://static-content.springer.com/image/art%3A10.1007%2Fs00439-007-0368-5/MediaObjects/439_2007_368_Fig3_HTML.gif
Fig. 3

a Boxplot of Gleason score according to presence (+) or absence (−) of regional/distant stage of disease at diagnosis. In the boxplots, the white line is the median, the entire box covers the 25th to 75th percentiles, and the whiskers go to the nearest value not beyond 1.5 times the inter-quartile range. b Boxplot of Gleason score according to PSA at diagnosis. See description of boxplots in the legend of Fig. 3a

Because one of our strongest linkage signals was on chromosome 19q13, we measured four SNPs in the HEPSIN gene: rs2451996, rs1688043, rs2305746, and rs2305747. The latter two SNPs are in intron-9, and were significantly associated with prostate cancer risk as reported by Pal et al. (Pal et al. 2006). Furthermore, Pal et al. (Pal et al. 2006) found SNP rs1688043 to be significantly associated with Gleason score among the cases. In our data, none of these four SNPs were significantly associated with Gleason grade, whether analyzing one SNP at a time, or all SNPs simultaneously (P-values > 0.25; regression models R2 < 0.002). Using the regression residuals as quantitative traits for linkage analyses, the linkage results differed very little from the unadjusted Gleason scores, suggesting that these SNPs in the HEPSIN gene did not explain any of our observed linkage signals on chromosome 19q.

Discussion

Our finding of suggestive linkage of Gleason score with chromosome 19q (75.7 cM) is consistent with previous reports of linkage to this region (Witte et al. 2000; Neville et al. 2002, 2003; Slager et al. 2003). The linkage peaks identified by Witte et al. (Witte et al. 2000) and by Slager et al. (Slager et al. 2003) were both located on chromosome 19q, at markers D19S433 (50.82 cM) and D19S902 (76.15 cM), respectively. Further analysis refined the region to 19q12-19q13.1 (Witte et al. 2000). However, unlike Pal et al. (Pal et al. 2006), we did not find SNPs from the HEPSIN gene to be associated with Gleason score, nor did these SNPs explain any of the linkage signal we found for chromosome 19q. The microsatellite markers that flank our largest linkage signal at 19q13.31-41 span approximately 8.2 Mb, and the distance from the nearest microsatellite marker to the SNPs in the HEPSIN gene is approximately 8.8 Mb. Although our linkage data alone cannot rule out the HEPSIN gene, because of the limits of linkage resolution with quantitative traits, our regression analyses suggest that the HEPSIN gene is not associated with Gleason score in our data, and that Gleason scores adjusted for HEPSIN SNPs did not alter our linkage findings. Hence, the HEPSIN gene does not appear to be causative of the linkage we observed for Gleason score with chromosome 19q.

Our finding of suggestive linkage of Gleason score to chromosome 5q is consistent with that reported by Witte et al. (Witte et al. 2003); their strongest evidence for linkage was for chromosome 5q31-33 (P-value = 0.0002). Slager et al. (Slager et al. 2003) also observed modest evidence of linkage to this region (LOD = 1.4).

A small number of other studies also restricted their linkage analyses to only men with aggressive disease. A study from Germany subset their pedigrees to those with at least two men with grade III disease (based on a system nearly equivalent to the American Joint Committee on Cancer grading). Among ten pedigrees with high-grade disease, a LOD score of 1.4 was reported for chromosome 7 (Paiss et al. 2003). In a similar fashion, three recent studies reported interesting linkage signals from genome linkage scans restricted to men with clinically aggressive prostate cancer based on Gleason score as well as stage, PSA level at diagnosis and other clinical variables. A study of 166 pedigrees pooled in the International Consortium for Prostate Cancer Genetics (ICPCG) found suggestive evidence for linkage on chromosomes 6 (LOD = 3.0), 11 (LOD = 2.4) and 20 (LOD = 2.5) (Schaid 2006). A study by Chang et al. (Chang et al. 2005) found suggestive evidence for linkage on chromosome X (HLOD = 2.54) and on chromosome 22 (HLOD = 2.06), and a study by Stanford et al. (Stanford et al. 2006), using a subset of the same families that were used in the ICPCG study, found suggestive evidence for linkage on chromosome 22 (dominant HLOD = 2.18). The present study, however, did not find evidence for linkage of Gleason score for any of these chromosomes, suggesting that the set of families defined as aggressive using criteria including grade, stage, PSA, and death from metastatic disease highlight different loci than those suggested by focusing on Gleason score alone. This highlights the fact that the different measures of aggressiveness may be capturing different aspects of the disease.

In summary, our study replicates the linkage findings for Gleason score on chromosomes 19q and 5q. These findings are consistent with prior published results for aggressive prostate cancer. In total, our findings, and those of others, suggest that the degree of severity of prostate cancer may be controlled by multiple genes. Because men are often over-diagnosed with prostate cancer (15–37% of men), meaning that they have clinically insignificant prostate cancer that would not otherwise be detected in their lifetime (Etzioni et al. 2002), identifying the genetic factors related to the spectrum of aggression of prostate cancer has important clinical implications that can impact a man’s prognosis and choice of treatment. Our results offer guidance on chromosomal regions worth additional follow-up.

Acknowledgments

We especially thank the members of the PROGRESS families for participating in this research. This research was supported by National Institutes of Health Grants RO1 CA080122 (to J.L.S.), RO1 CA78836 and KO5 CA90754 (to E.A.O.), and the Department of Defense fellowship W81XWH-04-1-0083 (to D.M.K.), with additional support from the Prostate Cancer Foundation and the Fred Hutchinson Cancer Research Center. E.A.O was supported in part by the Intramural Program of the National Human Genome Research Institute.

Copyright information

© Springer-Verlag 2007