PIN-like ductal carcinoma of the prostate has frequent activating RAS/RAF mutations

Abstract

Aims:

Prostatic intraepithelial neoplasia-like (PIN-like) ductal carcinoma is a rare tumor characterized by often cystically dilated glands architecturally resembling high-grade PIN, but lacking basal cells. These tumors are frequently accompanied by Grade Group 1 acinar cancer and behave relatively indolently. In contrast, conventional ductal adenocarcinoma of the prostate is an aggressive variant comparable to Grade Group 4 acinar cancer. Here, we used targeted next generation sequencing to molecularly profile PIN-like ductal carcinoma cases at radical prostatectomy.

Methods and results:

Five PIN-like ductal carcinoma samples at radical prostatectomy with sufficient tumor tissue available were analyzed for genomic alterations by targeted next generation sequencing using the JHU solid tumor panel. DNA was captured using SureSelect for 640 genes and sequenced on the Illumina HiSeq platform. Three of five (60%) of the PIN-like ductal carcinomas showed activating mutations in the RAS/RAF pathways which are extraordinarily rare in conventional primary prostate carcinoma (<3% of cases), including an activating hotspot BRAF mutation (p.K601E), an activating hotspot mutation in HRAS (p.Q61K) and an in-frame activating deletion in BRAF (p.T488_Q493delinsK). An additional two cases lacked BRAF or HRAS mutations but harbored in-frame insertions of uncertain significance in MAP2K4 and MAP3K6. One case had sufficient acinar tumor for sequencing, and showed similar molecular profile as the concurrent PIN-like ductal carcinoma, suggesting a clonal relationship between the two components.

Conclusions:

PIN-like ductal carcinoma represents a molecularly unique tumor, enriched for potentially targetable oncogenic driver mutations in the RAS/RAF/MAPK pathway. This molecular profile contrasts with that of conventional ductal adenocarcinoma, which is typically enriched for pathogenic mutations in the mismatch repair (MMR) and homologous recombination (HR) DNA repair pathways.

Introduction

Prostatic ductal adenocarcinoma is an aggressive morphologic subtype characterized by pseudostratified, tall columnar epithelium with variable architectural patterns. Conventional ductal carcinomas composed of papillary and cribriform patterns behave clinically like Grade Group 4 acinar tumors and are frequently associated with high grade acinar carcinomas ¹. Genomic analyses have revealed that ductal and acinar adenocarcinoma likely have a common origin when both morphologies are present ^{2, 3}. Ductal foci accrue additional alterations leading to frequent activation of targetable pathways such as PI3K- or WNT-signaling, based on early or late divergence from coincident acinar foci ^{3, 4}. We have previously shown that TMPRSS2-ERG gene fusions and PTEN loss are less common in ductal carcinoma compared to acinar adenocarcinoma ^{2, 5}. Ductal carcinomas also have frequent pathogenic germline alterations of DNA repair genes, congruent with their aggressive clinical behavior and which suggests that some may respond to immunotherapy ^{6, 7}.

Prostatic intraepithelial neoplasia (PIN)-like ductal carcinoma (also known as PIN-like carcinoma) is a rare tumor characterized most commonly by cystically dilated glands lined by pseudostratified columnar epithelium with flat and tufted architecture resembling high-grade prostatic intraepithelial neoplasia (PIN) but lacking basal cells. This entity was first described by Hameed and Humphrey with an incidence of 1.3% in biopsies ⁸. In seminal early studies, these tumors were termed “PIN-like ductal” due to the overlapping cytologic features with conventional ductal adenocarcinoma ⁹. Though molecular characterization has been limited, our institution reported that PIN-like ductal carcinomas have lower rates of ERG expression and PTEN loss, similar to what is seen in conventional ductal carcinoma ^{5, 10}. Clinically, however, PIN-like ductal carcinomas behave relatively indolently, akin to Grade Group 1 acinar tumors, in stark contrast to the aggressive clinical course described for most conventional ductal carcinomas ^{9, 10}. This suggests that the molecular features of PIN-like ductal carcinoma and conventional ductal adenocarcinomas may be divergent. Here, we perform the first molecular profiling study of PIN-like ductal carcinoma.

Materials and methods

Tissue samples:

After Johns Hopkins Institutional Review Board approval, cases of PIN-like ductal carcinoma with available tissue for molecular analysis were retrospectively identified by electronic search of the Johns Hopkins Hospital in-house and consult cases used in previous studies ¹⁰. Due to the cystic nature of many lesions and the relatively low abundance of tumor epithelium, only tumors occurring in radical prostatectomy specimens were selected for this study. We excluded any PIN-like ductal carcinoma identified on biopsies, cases with minimal PIN-like ductal component or admixed PIN-like ductal and acinar carcinoma (where the acinar and PIN-like ductal components were physically intermingled to avoid DNA contamination of the acinar component in the final eluate). Using these inclusion criteria, a total of 5 cases of PIN-like carcinoma had sufficient tumor tissue for inclusion in the study, including 3 cases with a geographically separate nodule of acinar carcinoma (Supplementary Figure S1). Their clinical-pathologic features are described in Table 1.

Table 1:

Clinical-pathologic and molecular characteristics of patients with PIN-like ductal prostatic carcinoma

PLD#	Age	Gleason grade	Tumor stage	Margins	Acinar carcinoma	Gene altered	Amino acid alteration	VAF	Mutation status
PLD1	66	3+4=7	pT2	Negative	Not identified	BRAF	p.T488_Q493delinsK	10.8	Pathogenic
PLD2	45	3+3=6	pT2	Negative	acinar in right PL mid; see Supp. Table S2 (PLD in left PL apex)	MAP2K4* MAP2K4* NCOR2* LRP1B ARID1A*	p.G11dup p.S26Y p.Q507_Q510del p.D1763V p.Q1334del	29.2 48 30.9 16.8 3.5	VUS VUS VUS VUS Pathogenic
PLD3	57	3+4=7	pT2	Negative	Acinar in same block as PLD, too small for sequencing	MAP3K6	p.F507_L508insWLHF	17.5	VUS
PLD4	72	3+3=6	pT2	Negative	Not identified	BRAF CREBBP	p.K601E p.S893L	33.8 45.8	Pathogenic Pathogenic
PLD5	60	3+3=6	pT2	Negative	Acinar in right apex (PLD in left mid), too small for sequencing	HRAS BRAF EP300	p.Q61K p.G32_A33del p.S1726R	15.2 34.6 55.8	Pathogenic VUS VUS

Note: All pathogenic alterations are listed for each case, with selected VUS. For comprehensive VUS list, please see Supplementary Table S1.

^*Denotes mutations seen in both the acinar and PIN-like ductal components.

DNA extraction:

The PIN- like ductal carcinoma components and the acinar carcinoma components were annotated on H&E slides and used to guide macrodissection of tumor from unstained slides. DNA was extracted using the Qiagen FFPE DNA extraction kit (Hilden, Germany) according to manufacturer’s directions. DNA concentrations were quantified using Quant-iT dsDNA High Sensitivity Assay Kit (Invitrogen, Carlsbad, CA, USA).

Next generation sequencing (NGS):

NGS was performed as previously reported ¹¹ using Illumina chemistry and a HiSeq 2500 (Illumina, San Diego, CA). Briefly, DNA was sheared using a Covaris LE220-plus (Woburn, MA). Coding exons of 640 genes were captured using the SureSelect-XT Target Enrichment System (Agilent Technologies Santa Clara, CA). Chips were loaded followed by cluster formation and 2 × 100 paired end sequencing to a 500–1000x average read depth. Reads from FASTQ files were aligned to Hg19 (GRCh27) using the Burrows-Wheeler alignment (BWA) algorithm and variant calling performed using MLDVC v5.0 (molecular diagnostics laboratory variant caller) and Haplotype Caller (Genome analysis tool kit 3.3). All variants from the variant call file were confirmed using the Integrated Genome Viewer (2.3.78, Broad Institute, Cambridge, MA) and annotated using the Catalog of Somatic Mutations in Cancer (COSMIC, v82, Sanger Institute, Hixton, UK) and dbSNP (v150, NCBI, NIH, Rockville, MD).

Results

The average age of the five patients included in this study was 60 (range: 45 to 72 years). The overall Gleason grade at radical prostatectomy, based on the nodule with the highest grade, was GG1 in 60% of cases and GG2 in 40% of cases. Three cases had a separate nodule of acinar carcinoma that was geographically distinct from the PIN-like ductal component (Supplementary Figure S1). The PIN-like ductal component was sequenced in all cases (Figure 1), while only a single case (#2) had sufficient acinar tumor present for sequencing of both components. Pathologic stage in all cases was pT2 (Table 1).

Figure 1:

Representative H&E images of primary prostate cancer cases with PIN-like ductal carcinoma component demonstrating cystically dilated glands lined with pseudostratified columnar epithelium resembling high grade PIN. A, Case #1 with a pathogenic activating in-frame deletion in BRAF (p.T488_Q493delinsK). B and C, Cases #2 and #3 with no obvious pathogenic alterations, but showing in-frame insertions in MAP2K4 and MAP3K6, respectively. D, Case #4 with an activating hotspot BRAF mutation (p.K601E). E, Case #5 with an activating hotspot mutation in HRAS (p.Q61K).

Targeted sequencing revealed that three of five (60%) of the PIN-like ductal carcinomas harbored activating mutations in the RAS/RAF pathway (Table 1 and Supplementary Table S1). Case #4 had an activating hotspot BRAF mutation (p.K601E), which may confer at least partial sensitivity to BRAF inhibitor vemurafenib ¹². Case #5 showed an activating hotspot mutation in HRAS (p.Q61K), most commonly reported in follicular thyroid carcinomas. This case had a co-occurring in-frame deletion of BRAF (p.G32_A33del). Though currently classified as a variant of uncertain significance (VUS), this BRAF alteration has been reported in the germline of a single patient with a Rasopathy (ClinVar RCV000544255.1), suggesting potential functional significance. Case #1 showed an in-frame deletion in BRAF (p.T488_Q493delinsK). This deletion has been previously reported to shorten and lock the β3/αC-helix loop into a conformation that favors dimer formation and BRAF activation and may be sensitive to RAF dimer inhibitors ¹³. Cases #2 and #3 lacked BRAF or HRAS mutations but showed in-frame insertions in MAP2K4 and MAP3K6, respectively, which were classified as variants of uncertain significance and were likely germline based on allelic fraction (Table 1).

Interestingly, case #2 had a pathogenic mutation in ARID1A (p.Q1334del), which encodes for a protein involved in chromatin remodeling, at a low variant allele fraction suggesting it was somatic (3.5%). Of note, Case #4 had a missense mutation in the histone acetyltransferase CREBBP (p.S893L) that is classified as pathogenic, albeit with only mild predicted effects on protein function ¹⁴. Though only tumor DNA was sequenced, the variant allele fraction of this alteration (45%) could be consistent with a germline alteration (Supplementary Table S1). Case #5 was noted to have a variant of unknown significance (VUS) in the closely related histone acetyltransferase EP300 (p.S1726R) which has some overlapping functions with CREBBP at a similar variant allele fraction suggesting germline origin.

Case #2 had a geographically separate focus of acinar carcinoma that was also sequenced to examine the clonal relationship between PIN-Like ductal carcinoma and acinar carcinoma (Supplementary Table S2). The acinar carcinoma from this case showed identical MAP2K4 mutations as the PIN-like ductal component (with variant allele fractions likely consistent with germline alterations), as well as the same NCOR2 variant of unknown significance (likely germline) seen in both components. Importantly, the same pathogenic (and likely somatic) ARID1A mutation as seen in the PIN-like ductal component was also present in the acinar component, suggesting a clonal relationship between the two, though an LRP1B variant of unknown significance seen in the PIN-like ductal component was not detected in the acinar component.

Discussion

In this first sequencing of PIN-like ductal carcinoma, we demonstrate that they are enriched with potentially targetable oncogenic driver mutations in RAS/RAF signaling and distinct from conventional ductal carcinoma which is enriched for DNA repair gene alterations ^{6, 7}. BRAF or HRAS alterations are extraordinarily rare in conventional primary prostate carcinoma (<3% of cases in TCGA cohort) ¹⁵. Here, PIN-like ductal carcinoma cases showed multiple alterations in the BRAF gene (p.K601E, p.T488_Q493delinsK and p.G32_A33del). Notably, the most prevalent alteration observed in BRAF among human tumors – p.V600E - was not identified in our study set. While tumors with V600 mutations (class 1 BRAF mutations) are RAS-independent monomers and are sensitive to BRAF inhibitors, the K601E mutation (class 2 BRAF mutation) is RAS-independent, signaling as constitutive dimer and responding only partially to vemurafenib ¹². Our data are in line with the overall rarity of the V600E mutation in prostate cancers sequenced to date (0.024%), where the p.K601E mutation appears to be ten times more common, occurring in 0.24% of cases ¹⁶. Rearrangements of the BRAF or RAF1 locus may also drive tumorigenesis in a minority of prostate cancer cases ¹⁷, though we did not assess for these in the current study. Finally, Case #5 showed an activating hotspot mutation in HRAS (p.Q61K), another unusual mutation for prostate cancer. Head and neck squamous cell carcinoma harboring oncogenic HRAS mutations have shown promising response to farnesyl transferase inhibitor tipifarnib ¹⁸ (NCT02383927).

Thus, we show that the PIN-like ductal carcinomas harbor a unique molecular profile, and is enriched for potentially targetable alterations in the RAS/RAF/MAPK signaling pathway. This is distinct from the molecular profile seen in conventional prostatic ductal adenocarcinomas, which are usually enriched for alterations in the mismatch repair (MMR) and homologous recombination (HR) DNA repair pathways and which might explain their more aggressive clinical behavior in contrast to the PIN-like ductal carcinomas ^{4, 6}.

The relative rarity of RAF and RAS gene mutations in human prostate cancer is puzzling and may be related to the underlying relatively low mutation rate in most primary prostate cancers. Prostatic expression of BRAF p.V600E is sufficient to initiate prostatic tumorigenesis in murine models ¹⁹. Similar mouse models have been developed for activated KRAS p.K12D ²⁰. Notably, a previous small study by our group examining another rare entity in the prostate, isolated intraductal prostate cancer occurring without invasive cancer, found an unexpectedly high rate of RAS/RAF alterations among seven cases studied, including an activating KRAS (p.G13P) mutation, and another case with BRAF p.K601E ²¹. Preclinical mouse models have shown that while activation of the RAS pathway alone is not sufficient to induce prostate cancer, additional alteration in the PTEN/PI3K/AKT pathway promotes prostate cancer progression and early lethality ²². Other studies have also shown that Pb-H-Ras^G12V transgenic mice ²³ and Pb-Cre; K-Ras^LSL-V12/+ conditional activating mice ²⁴ models exhibited low grade PIN, whereas combinatorial oncogenic mutations of K-ras and beta-catenin drive rapid progression of prostate tumorigenesis to invasive carcinoma ²⁴. Similarly, cooperation of the PI3K and MAPK pathway activations in PCa progression has also been characterized where PCa in Pb-Cre; Pten^fl/fl; K-Ras^LSL-G12D/+ mice displayed epithelial to mesenchymal transition (EMT) and metastatic phenotypes ²⁵. Another study looking at the genetic progression of high grade prostatic intraepithelial neoplasia (HG-PIN) to prostate cancer showed that HRAS (p.Q61K) mutation was specific to the HG-PIN component, but was not identified in the matched prostate carcinoma component ²⁶. Taken together with the results of the current study, these data suggest that RAS/RAF mutations might be sufficient to initiate prostatic carcinogenesis, but lack the potential to drive significant progression, hence the association with rare and potentially relatively indolent tumor subtypes such as PIN-like ductal carcinoma or isolated intraductal carcinoma.

We acknowledge several limitations of our study. First, it should be noted that while PIN-like ductal carcinoma is a distinct entity with characteristic molecular features described in this study, PIN-like carcinomas that were originally described by Hameed and Humphrey also include cases that resemble acinar carcinoma with non-columnar cells and rounded nuclei. While we have focused on the PIN-like ductal carcinoma in this study, it will be interesting to study in future the molecular profile of PIN-like adenocarcinoma with acinar features and how they contrast with PIN-like ductal carcinomas. Second, our sample size is unavoidably small, due to the rarity and low cellularity of PIN-like ductal carcinomas of the prostate. Third, due to the low amounts of DNA available from these lesions, we performed targeted sequencing which may fail to detect copy number alterations and genomic rearrangements which are relatively common in prostate cancer. Despite our small sample size, the unexpectedly high number of cases with oncogenic RAS/RAF mutations among this rare tumor type is likely highly significant given the very low prevalence of these alterations among conventional prostate cancers. Future studies will be necessary to confirm these findings and to shed light on the mechanistic role of these alterations in prostatic tumor progression.