A population-based cohort was collected from 1999 to 2005 from patients diagnosed with PRCA in the Pirkanmaa Hospital District that serves a population of a half a million inhabitants. The mean age at diagnosis for the 760 unselected patients was 62 (range 43-77), The median primary PSA and Gleason score were 8.9 and 6.4, respectively. The controls (n = 760) consisted of DNA samples from anonymous male blood donors obtained from the Finnish Red Cross.

A detailed description of the collection of the PRCA families has been previously described elsewhere 22. In this study, we analysed 178 families (youngest affected male from each family) with two or more affected first-degree relatives. The average number of affected family members was three. Familial cancer data was attained from the Finnish Cancer Registry and detailed clinical information, including Gleason score, WHO grade, PSA at the time of diagnosis, TNM stage and primary treatment from hospital records.

To examine the association between the variants and the disease aggressiveness a subgroup of 380 with clinically aggressive disease was analyzed. Consequently clinically aggressive disease patients were selected according to Gleason score over seven and an age at diagnosis less than 61 years.

All familial samples had been previously screened for the 11 Finnish BRCA1 and seven BRCA2 founder mutations 16. DNA samples from each family (n = 178), 285 unselected cases and 470 control samples were initially used for direct sequencing of the entire coding region and splice sites of PALB2. Unselected cases for screening the whole gene were early onset cases with Gleason score over seven. Additional analyses were carried out on four variants that showed a trend for association (PALB2 c.1592delT, 1674A>G, 2993G>A and 3300T>G). Unselected cases for additional analysis included 95 early onsets, aggressive cases and 368 non-aggressive cases with average age at diagnosis 67 years (range 63-77). Control samples in both primary and additional analysis were anonymous male blood donors from the Finnish Red Cross.

Mutation detection was performed through resequencing using the ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction kit with the ABI 3130xl sequencer (Applied Biosystems). Primers and the naming of sequence variants were produced to correspond to the GenBank reference sequences for PALB2 (NM_024675.3). Primer sequences are available on request. Variants were identified using Sequencher software 4.7 (Gene Codes Corporation, Ann Arbor, Mi). Loss of heterozygosity (LOH) analysis was carried out on two available tumor samples from the family with the segregating PALB2 1592delT. Selected sections of formalin fixed paraffin embedded tumors were obtained according to pathologist declaration. After tumor deparaffinization, DNA was extracted using a standard proteinase K protocol. Sequence from the tumor samples was compared to the patients' DNA sample from their peripheral blood.

Peripheral blood leucocytes from two cancerous individuals carrying PALB2 1592delT mutation were immortalized with Epstein-Barr virus (EBV) and sensitized to DNA crosslinking agent mitomycin C (MMC). Cells were treated with 0 nM, 1 nM, 5 nM, 10 nM and 50 nM MMC for 96 h and their viability was assayed by CellTiter-Blue^® Cell Viability Assay (Promega Corporation, Madison, WI, USA).

To validate significance of variants, the odds ratios and 95% confidence intervals were calculated and association analyses were made using the Fisher's exact test. Analyses were done using SPSS 12.0 statistical software. Pairwise linkage disequilibrium between three most frequently detected variants (1674A>G, 2993G>A and 3300T>G) in PALB2 were analysed with Haploview 4.1. To define the haplotypes of cases familial and unselected cases were combined.

Prompted by the previous observation that the Finnish truncating founder mutation, PALB2 c.1592delT, was identified segregating in one family with four PRCA cases 21, we wanted to study the contribution of other possible PALB2 mutations. Probands from 178 Finnish HPC (hereditary prostate cancer) families and 285 unselected cases were screened for the entire coding region of PALB2. Total of six variants were detected in the coding regions and the exon-intron boundaries of PALB2 (Table 1). All the variants have been previously described and only the truncating mutation (c.1592delT) has been shown to have functional consequences on DNA damage response. Association of four detected variants (c.1592delT, 1674A>G, 2993G>A and 3300T>G) were assessed in more detail in a larger set of unselected cases and controls. The truncating c.1592delT mutation was detected in two sporadic cases in addition to the previously described family 310 and in one control sample 21. Interestingly, family 310 has now been found to have also stomach and skin cancer, in addition to earlier described breast cancer, as indicated in Figure 1 and Table 2. Three of the detected variants, 1674A>G in exon 4, 2993G>A in exon 9 and 3300T>G in exon 12, co-existed in six patients out of 178 familial samples (OR 1.52 CI 95% 0.6-3.9), 20/748 unselected cases (OR 1.20 CI 95% 0.6-2.3) and in 17/760 of the controls. To assess question whether these three variants have a stronger joint effect that would strengthen the trend seen with singletons a haplotype analysis was performed. The strongest linkage was between 1674A>G and 2993G>A (D'= 0.975; r² = 0.14, LOD 34.7) when 1886 case chromosomes and 1714 control chromosomes were analysed. Although these two variants appeared to have a frequency in PRCA cases that is higher than the Finnish population control frequency, their combined haplotype gave no improved results, i.e. none of the haplotype combinations showed significant association with the disease. Variant 2993G>A is now reported for the first time in the Finnish population since it did not come up in the study of Finnish breast cancer patients 21. As reported by Rahman and colleagues 23 the 2993G>A variant is possibly damaging to protein function, based on SIFT (sorting intolerant from tolerant) analysis.

We found that the 2993G>A variant exhibits a borderline significant odds ratios in a subgroup of 380 unselected cases with a Gleason score over seven and an age at diagnosis less than 61 years (20/380; OR 1.9; 95% CI 1.0-3.5). Edwards and colleagues 24 have presented similar age related observation with BRCA2 mutations in prostate cancer. They reported protein truncating mutations in BRCA2 to associate with PRCA diagnosed at or before age of 55. Our subgroup analysis also revealed that the variant 1674A>G has an OR of 1.4 (86/380; 95% CI 1.0-1.9) in a group of 368 patients with an average age at diagnosis of 67 years. However, the 1674A>G variant is commonly found in the Finnish population and well tolerated according to SIFT analysis, indicating no clear contribution to cancer formation. Similarly, the Val932 Met (2794G>A) change in exon 8 is well tolerated and has no effect on protein function 23.

In order to obtain information about loss of heterozygosity in the tumors from c.1592delT mutation positive patients, two available paraffin-embedded tumors were analysed. Repeated PCR-analysis did not reveal existence of LOH, which argues against the role as tumor suppressor gene. However the sample size is very small and therefore more tumor samples from PALB2-associated tumors needs to be characterized to be able to reliable assess PALB2 function in tumor formation. To date there is only one report indicating LOH in PALB2 related tumors. Unfortunately also this study lacks a statistical power to make an appropriate conclusion about the role of PALB2 in tumorigenesis 25.

Previously, MMC sensitivity test to c.1592delT mutation has been done in a reporter cell line 21. To test whether the previously described PALB2 c.1592delT mutation has functional consequences also in a heterozygous form, as present in all our patients, we predisposed two lymphoblast cell lines with c.1592delT and a wild type carrying control cell line to DNA cross linking agent MMC (Figure 2). c.1592delT mutation carriers were patients (15 and 24) from family 310 earlier described by Erkko et al. 21. In both cell lines MMC-induced growth inhibition was observed suggesting PALB2 haploinsufficiency. It is also possible that dominant-negative effect is affecting in tumor formation in these patients.

Geographical and ethnical differences in the PRCA susceptibility alleles have been previously reported in the CHEK2 gene that acts in the same pathway as PALB2, and therefore, variation in the PALB2 mutation spectrum in different populations would not be surprising. In the present study and in the study by Erkko et al. 21, the number of mutations found in PALB2 is relatively small among both Finnish breast and PRCA patients, with nine and six variants detected, respectively, compared to the British breast cancer family study by Rahman and colleagues 23 where a total of 50 variants and five frameshift mutations were reported. Recently, Tischkowitz and colleagues 26 reported a study of 95 US PRCA patients from Michigan, all diagnosed at < 55 years of age. Eleven variants in PALB2 were found, but none of them were truncating and no association with the disease was observed. When comparing our patients to those very early-onset patients studied by Tischkowitz and colleagues 26 there is a clear difference in the mean age of onset between the two datasets. On the other hand, no variants were found in the 14 Ashkenazi-Jewish and 21 French-Canadian PRCA cases with family history of cancer 27. In Finland, the smaller number of variation found likely reflects the known genetic homogeneity and founder effect of the population 28.

Our findings indicate that no other deleterious PALB2 variants, except 1592delT mutation, contribute even marginally to PRCA risk in Finland. However, it remains possible that other genes from the BRCA1-PALB2-BRCA2 pathway have PRCA-predisposing alleles