Background
Prostate cancer is heterogeneous regarding its pathologic features, genetic background and clinical outcome. Clinical-decision making mostly depends upon serum Prostate Specific Antigen (PSA) level, clinical tumour stage, and pathologic biopsy Gleason score (GS) – a grading system based on architectural tumour patterns [
1]. While patients with the lowest GS ≤6 (WHO/ISUP group 1) have an excellent patient outcome, those with the highest GS 9–10 (WHO/ISUP group 5) have the worst [
1,
2]. The clinical outcome of GS 3 + 4 = 7 (WHO/ISUP group 2) prostate cancer patients is variable. Improving risk assessment in this subgroup of patients is of clinical relevance as biopsy GS 3 + 4 = 7 is an important threshold for active treatment. Recent studies have indicated that, among Gleason grade 4 growth patterns, cribriform growth is associated with worse clinical outcome [
3‐
6].
In recent years the clinical relevance of intraductal carcinoma of the prostate (IDC) – a malignant epithelial proliferation filling and extending pre-existent glands – has been acknowledged. Although not included in the Gleason grading system, IDC has been associated with high GS, advanced tumour stage, biochemical relapse and distant metastasis [
7‐
12]. IDC often mimics invasive cribriform carcinoma, requiring basal cell immunohistochemistry for their distinction. Recently, our group has shown that patients with cribriform and/or intraductal carcinoma (CR/IDC), have significantly worse disease-specific survival probabilities than those without, regardless of GS [
13]. Furthermore, patients with focal CR/IDC have similar outcome as men with extensive CR/IDC, indicating that the mere presence of this growth pattern is an adverse feature [
13,
14].
Although the number of mutational events in prostate cancer is relatively low, copy number alterations (CNAs) are significantly more frequent [
15‐
24]. Several studies have developed molecular prognostic signatures, showing that indolent tumours have relatively few CNAs in contrast to large-scale CNAs in high-grade or metastatic tumours [
16,
17,
25,
26]. However, both the intra- and inter-tumour heterogeneity pose significant challenges for personalizing treatment in patients with prostate cancer [
27‐
29]. For instance, GS 7 prostate cancers harbour a wide range of CNA burden varying between <1% to 50% [
26].
Since presence of CR/IDC growth pattern is an independent, adverse clinico-pathologic parameter, we hypothesize that CR/IDC represents a morphological substrate of genomic alterations associated with aggressive disease [
13]. The objective of this study was to determine the CNAs and single nucleotide variants (SNVs) associated with CR/IDC using bioinformatics analyses of datasets from The Cancer Genome Atlas Project (TCGA) and the Canadian Prostate Cancer Genome Network (CPC-GENE).
Discussion
Recent studies have indicated the clinical importance of both invasive cribriform and intraductal carcinoma of the prostate [
6,
13,
14]. In the current study, we hypothesized that CR/IDC represents a morphologic substrate of genomic alterations associated with aggressive disease. We found that CR/IDC was associated with increased genomic instability together with chromosomal deletions of 3p13, 6q15, 8p21–23, 10q23, 13q14, 16q21–24, 18q21–23, and amplification of 8q24. The genetic losses and amplifications included several genes related to aggressive prostate cancer such as loss of
PTEN,
RB1,
TP53 and amplification of
MYC. Altogether, these findings support our hypothesis that CR/IDC is a specific morphologic substrate of genomic alterations associated with aggressive disease.
Our study is in line with previous studies on genetic abnormalities related to CR/IDC growth. Dawkins et al. [
62] and Bettendorf et al. [
63] observed more frequently loss of heterozygosity (LOH) in IDC than in the invasive prostate cancer component. Qian et al. showed gain of chromosomes 7, 12, and Y, loss of chromosome 8, and amplification of
c-MYC in cribriform cancer compared to other Gleason grade 3 and 4 patterns [
64]. In a meta-analysis on recurrent CNAs, Williams et al. [
33] compared 568 primary prostate cancer tumour samples from 8 previous studies [
16,
19,
20,
65‐
69] with 115 metastatic prostate cancer samples from 5 studies [
16,
22,
67,
70,
71]. Strikingly, the prevalence of recurrent CNAs in metastatic prostate cancers corresponded with several of the CNAs found enriched in CR/IDC, such as
PTEN and
NKX3–1. Recently, Chua et al. studied differences in RNA expression in prostate cancer with and without CR/IDC. They found that the long non-coding RNA SChLAP1, which has been associated with tumour progression, was significantly higher in CR/IDC, and that CR/IDC growth was associated with hypoxia [
72‐
74]. Together these findings further support a strong relation of CR/IDC with molecular tumour progression. On the other hand, we did not find a statistically significant difference between GS 3 + 4 = 7 without CR/IDC and GS 6 cases, which further supports the question whether it is clinically relevant to distinguish CR/IDC-negative GS 3 + 4 = 7 from GS 6 prostate cancer cases.
Although prostate cancer with CR/IDC showed increased genomic instability, it is not yet clear to what extent these molecular alterations are exclusively present in CR/IDC tumour glands or whether these alterations can also be found in surrounding non-cribriform tumour glands. Using RNA in situ hybridization, we previously found that SChLAP1 was not only over-expressed in CR/IDC structures but also in adjacent non-cribriform cancer glands suggesting that it represents a field effect during tumour progression and not a specific characteristic of CR/IDC growth [
72,
75]. In our study, CR/IDC was more frequently present in cases with higher GS. To exclude that genomic alterations were merely relating to higher GS and not to CR/IDC per se, we performed PGA subgroup analysis and logistic regression for CNAs, which indeed revealed an independent associated with CR/IDC in the TCGA cohort. Further comparisons of microdissected growth patterns within individual patients are mandatory to determine what events are specific for CR/IDC and which represent general effects of progression.
Elucidation of the molecular alterations associated to CR/IDC is not only of interest for molecular-biology, but might also have future impact for prostate cancer diagnosis and management. Prostate biopsies only sample a limited volume of the entire tumour and might be false-negative for CR/IDC due to sampling artefact. Since IDC represents an extensive proliferation of neoplastic cells within pre-existent acini which connect with the urethra, we postulate that these cells and/or their DNA can be shed into urine. Identification of molecular alterations associated with CR/IDC in voided urine could form the base of non-invasive tests for detection of aggressive CR/IDC.
The current study has several limitations. While we set out to validate our findings in an independent cohort, we noticed that many events originally found in the TCGA cohort could not be confirmed in the CPC-GENE dataset. This may be explained by differences in cohort composition, since the TCGA was enriched for tumours with adverse pathologic features. In addition, the statistical power of the CPC-GENE cohort was lower than of the TCGA, as its study population was smaller, included samples with lower and more variable tumour percentage, and was strongly enriched for CR/IDC in GS 8–10. Nevertheless, both datasets independently revealed the association of CR/IDC with increased genomic instability and the deletions of various specific genomic regions and genes. Furthermore, tumour heterogeneity and sampling artefacts may have also influenced the outcome of this study, as our current data was based on DNA derived from a freshly frozen section per patient. Hence, there may have been, for instance, CR/IDC growth in an adjacent region that was not sampled for genomic analysis that may have been detected due to a field effect. This might be the cause of the relatively small effect sizes in the current study. Lastly, we did not independently analyse CR/IDC growth in relation to adjacent tumour glands using, for instance, laser-capture microdissection or in situ hybridization.
Conclusion
We found that pathologic CR/IDC growth pattern is associated genomic instability including deletions of 8p, 10q23, 13q22, 16q22–24, 17p13 and 21q22, as well as smaller 8q24 amplification. These results indicate that CR/IDC is a histopathological substrate of molecular tumour progression and present a rationale for its aggressive clinical behaviour.