HOXB13 G84E-related Familial Prostate Cancers: A Clinical, Histologic, and Molecular Survey

Abstract

Recent genetic epidemiologic studies identified a germline mutation in the homeobox transcription factor, HOXB13 G84E, which is associated with markedly increased risk of prostate cancer, particularly early onset hereditary prostate cancer. The histomorphologic and molecular features of cancers arising in such carriers have not been studied. Here, we reviewed prostatectomy specimens from 23 HOXB13 G84E mutation carriers, mapping the total cancer burden by anatomically distinct cancer focus and evaluating morphologic features. We also assessed basic molecular subtypes for all cancer foci (ERG/SPINK1 status) by dual immunohistochemistry staining on full sections. The cohort showed median age 58 years, median serum PSA 5.7 ng/ml, and a median of 6 cancer foci (range 1–14) per case. Of evaluable cases, dominant foci were Gleason score 6 in 23%, 3+4=7 in 41%, 4+3=7 in 23%, and ≥8 in 14%; biochemical recurrence was observed in one case over a median of 36 months follow-up. Histologic review found a high prevalence of cases showing cancers with a spectrum of features previously described with pseudohyperplastic carcinomas, with 45% of cases showing a dominant focus with such features. Molecular subtyping revealed a strikingly low prevalence of ERG⁺ cancer with increased prevalence of SPINK1⁺ cancer (dominant focus ERG⁺ 17%, SPINK1⁺ 26%, ERG⁻/SPINK1⁻ 52%, single ERG⁺/SPINK1⁺ focus 4%). One ERG⁻/SPINK1⁻ dominant focus showed aberrant p63⁺ immunophenotype. In summary, HOXB13 G84E variant-related prostate cancers show frequent pseudohyperplastic-type features and markedly low prevalence of ERG⁺ cancers relative to unselected cases, and, especially, to early onset cohorts. These findings suggest that novel molecular pathways may drive disease in HOXB13 G84E carriers.

INTRODUCTION

Germline variations in >70 genes have been associated with the inherited predisposition to prostate cancer, including common variations in multiple genes that together may account for ~30% of familial prostate cancer risk, although these variants individually have small odds ratios (1.1–1.5)^1–6. Recently, through linkage analysis and next-generation sequencing, investigators at the University of Michigan and Johns Hopkins University identified a rare germline variation in HOXB13 (G84E substitution) in families with hereditary prostate cancer⁷. This rare variant (carrier frequency estimates of 0.1–1.5%) has been shown in multiple European-American cohorts to confer an ~3–10× odds ratio for the development of prostate cancer, particularly early onset and hereditary prostate cancer^8–19. HOXB13, a homeobox domain-containing gene, is strongly expressed in the prostate from embryonic stages to adulthood and has been shown to directly interact with the androgen receptor (AR), functioning to activate or repress distinct AR target genes^20–25. The molecular mechanism by which HOXB13 G84E variants promote prostate cancer development, however, is unknown.

Prostate cancer shows a variety of histologic patterns, most commonly acinar carcinoma, with variant histology observed in 5–10% of cases^{26, 27}. Numerous histologic variants of prostate cancer have been described, including ductal carcinoma, foamy gland carcinoma, pseudohyperplastic carcinoma, mucinous carcinoma and neuroendocrine/small cell carcinoma. Similarly, a variety of robust prostate cancer molecular subtypes have been identified, based primarily on the presence or absence of recurrent gene fusions involving 5’ regions of androgen regulated genes (most commonly TMPRSS2) fused to nearly full length members of the ETS transcription factor family (most commonly ERG)^28–30. In PSA-screened, Caucasian cohorts, approximately 50% of prostate cancer foci harbor ETS gene fusions (ETS⁺)³¹. The development of sensitive and specific antibodies against ERG, the 3’ partner in ~90% of all ETS fusions, has enabled rapid characterization of ETS fusion status across large cohorts by immunohistochemistry (IHC)^{32, 33}. Additionally, over-expression of SPINK1 was identified in ~5–10% of prostate cancers, which are nearly exclusively ETS fusion negative (ETS⁻)³⁴. ETS^+/−/SPINK1^+/− based subtyping of prostate cancer has been supported by recent reports of dual ERG/SPINK1 IHC protocols employed by independent groups, which have confirmed the mutually exclusive nature of these subtypes^{35, 36}. When present, ERG rearrangements and corresponding IHC staining are clonal in a given cancer focus, while SPINK1 often shows heterogeneous staining in a given focus. Given the multifocal nature of prostate cancer, men can harbor genetically distinct ERG⁺/SPINK1⁻, ERG⁻/SPINK1⁺ and ERG⁻/SPINK1⁻ foci, and ERG/SPINK1 status can be used to evaluate whether two foci of prostate cancer are genetically distinct (multiclonal).

The histologic characteristics and molecular subtypes of prostate cancer developing in HOXB13 G84E carriers have not been explored. Here we comprehensively characterized 23 prostate cancers arising in men harboring this variant through complete mapping and morphologic evaluation of all cancer foci at the time of prostatectomy specimens. Importantly, for each patient, we also evaluated the basic molecular subtype of all cancer foci in each case for ERG/SPINK1 status by IHC, rather than the more typical strategy of just evaluating the dominant focus (usually defined by Gleason score and/or size).

MATERIALS AND METHODS

Cohort and Histologic Evaluation

With prior IRB approval, archival tissue blocks or archival blocks and slides from a total of 24 prostatectomy cases of patients previously identified as germline carriers of HOXB13 G84E mutation^{7, 17} were identified and retrieved from the University of Michigan Department of Pathology slide library. These prostatectomy cases were from either radical retropubic prostatectomies or robotic-assisted laparoscopic prostatectomies, as indicated in Table 1. The University of Michigan prostatectomy tissue submission protocol is a partial submission of the parenchyma with full submission of the capsule, enabling IRB-approved frozen tissue procurement but full microscopic evaluation of the capsule for staging and margins. It has been detailed previously and validated to demonstrate its diagnostic equivalence to full submission of complete submissions³⁷. Briefly, prostatectomy specimens were weighed, measured, and inked. Seminal vesicles, apex, and base were amputated, and the remaining prostate was serially sectioned at 3-mm to 5-mm intervals perpendicular to the long axis of the gland from the base to apex. Two intervening sections were snap-frozen for tissue procurement after removal of the circumferential capsules of the procurement sections. Capsule sections were submitted for fixation in 10% neutral buffered formalin with the unprocured sections before routine tissue processing and sectioning.

Table 1.

Clinical, pathological and molecular features of prostate cancers arising in HOXB13 G84E variant carriers

						Dominant Cancer Focus						All Cancer Foci
Subject #	Age (years)	PSA (ng/mL)	Surgery1	Stage2	Follow up3	Gleason Score	Size4	Location5	PH cyto6	PH arch7	Molecular Subtype8	Neg (#)	ERG+ (#)	SPINK1+ (#)	Total (#)
21	59	4.7	RRP	pT2c	52	treated (3+4)	1.1	TZ	treated	treated	Neg	3	0	0	3
12	40	4.9	RRP	pT2c	39	3+3	1.2	PZ	1	1	Neg	8	0	0	8
22	40	3.1	RRP	pT2c	177	3+3	1.5	PZ	1	1	Neg	3	0	0	3
11	60	4.9	RRP	pT2c+	131	3+4	1.1	PZ	1	1	Neg	8	3	2	13
20	68	3.2	RALP	pT2c	46	3+4	2.2	PZ	1	1	Neg	3	3	0	6
13	58	7.1	RRP	pT2c	12	4+3	0.8	PZ	1	1	Neg	12	0	1	13
15	54	7.4	RRP	pT2c	10	4+3	1.2	PZ	1	1	Neg	7	4	3	14
4	51	4.5	RALP	pT2c	38	3+4	1.2	PZ	1	1	SPINK1+	1	2	3	6
8	58	4.6	RALP	pT2c	0	3+4	0.4	PZ	1	1	SPINK1+	6	0	5	11
14	61	7.3	RRP	pT2c	137	4+3	2.2	PZ	1	1	SPINK1+	5	2	2	9
17	57	5.6	RALP	pT2c	2	3+4	3.2	PZ	1	0	Neg	6	1	0	7
1	64	6.4	RALP	pT2a–b	7	3+3	1.4	TZ	0	0	ERG+	0	1	0	1
23	64	8.4	RRP	pT2c	115	3+3	1	PZ	0	0	ERG+	5	1	0	6
9	42	7.4	RRP	pT2c	132	3+4	1.5	PZ	0	0	ERG+	3	3	1	7
10	64	5.2	RALP	pT2c	27	3+4	1.1	PZ	0	0	ERG+	3	2	0	5
2	53	6.5	RRP	pT2c	36	4+3	2.2	PZ	0	0	ERG+/SPINK1+	1	0	1	3
7	51	5.7	RRP	pT2c	17	3+3	0.3	PZ	0	0	Neg	7	0	0	7
16	58	18.2	RRP	pT2c	18	3+4	1.1	PZ	0	0	Neg	5	0	0	5
19	43	5.1	RALP	pT3a	37	3+4	1.4	PZ	0	0	Neg	3	1	0	4
5	59	4.6	RALP	pT3a	22	4+58	1.4	PZ	0	0	Neg, p63+	2	1	1	4
6	53	14.0	RALP	pT2c+	16	4+3	1.1	PZ	0	0	SPINK1+	3	1	2	6
3	63	12.6	RALP	pT2a–b	34	4+4	1.9	PZ	0	0	SPINK1+	1	0	2	3
18	57	30.2	RRP	pT3b+	55*	4+5	1.4	TZ	0	0	SPINK1+	0	0	1	1

¹RRP, radical retropubic prostatectomy; RALP, robot-assisted laparoscopic prostatectomy.

²All patients were N0Mx or NxMx.

³Follow-up in months.

^*Patient experienced biochemical recurrence.

⁴Greatest linear dimension of the dominant focus.

⁵Primary location of tumor in peripheral zone (PZ) or transition zone (TZ).

⁶Cytologic features of pseudohyperplastic (PH) carcinoma were present (1) or absent (0).

⁷Architectural features of pseudohyperplastic (PH) carcinoma were present (1) or absent (0).

⁸Molecular subtype by immunohistochemistry for ERG and SPINK1. Foci negative (Neg) for both ERG and SPINK1 are indicated.

⁸Gleason score included for descriptive purposes only, as appropriateness of Gleason score for p63+ cancers remains controversial.

Blocks from the 24 HOXB13 G84E cases were recut to produce H&E slides by standard means if original archival slides were not available for review. The total cancer burden for each case was mapped in three dimensions as we have detailed recently³⁸ and validated previously by others³⁹. Briefly, by careful reconstruction of each quartered serial section of the prostatectomy as if it were a whole mount, anatomically distinct foci were mapped as independent if they were greater than 3mm from adjacent focus within the same (or adjoining) section and 4mm or more from the closest cancer on an adjacent section. Each focus was reviewed and individually graded by 2005 ISUP consensus guidelines⁴⁰ by the study pathologists (S.C.S. and S.A.T.). The dominant focus was designated as the focus with the highest Gleason Score, and the greatest linear dimension was measured. Any focus with equivocal morphology for carcinoma was resolved by use of basal cell cocktail immunostain (detailed below) and diagnostic consensus (S.C.S., L.P.K. and S.A.T.). Clinical data were assembled by review of the institutional electronic medical record.

After mapping the entire cancer burden by focus, all foci were systematically re-reviewed to identify and tabulate the prevalence of each of selected histologic variants of prostatic adenocarcinoma. These included conventional acinar adenocarcinoma, pseudohyperplastic carcinoma, small cell carcinoma, mucinous adenocarcinoma, signet ring cell carcinoma, atrophic carcinoma, oncocytic carcinoma, ductal adenocarcinoma, carcinoma with Paneth cell-like neuroendocrine differentiation, squamous cell carcinoma, and giant cell carcinoma, as described previously^{26, 27}. A focus was considered variant if the variant morphology was the predominant pattern of the focus. Initial review identified a high prevalence of cancer foci showing features previously described within the spectrum of pseudohyperplastic carcinomas. Given the range of pseudohyperplastic-type features exhibited in these cases, each case was classified as to whether it harbored carcinoma demonstrating cytologic and/or architectural features of pseudohyperplastic carcinoma, and whether these features were present in the dominant cancer focus^{26, 41, 42}. The cytologic features of pseudohyperplastic carcinoma (less specific to this variant), included tall/columnar epithelium, pale to granular foamy cytoplasm, and basally oriented nuclei. The architectural features of pseudohyperplastic carcinoma evaluated were presence of an irregular luminal border and cystically dilated glands with papillary infoldings. Presence of luminal pink amorphous secretions and pink crystalloids, features of both pseudohyperplastic and non-pseudohyerplastic carcinoma (and occasional benign glands), were also tabulated for each cancer focus. Finally, each focus was evaluated for features previously associated with ERG⁺ prostate cancer, including prominent cribriform architecture, intraductal growth, and blue mucin⁴³.

Immunohistochemistry

Dual immunohistochemistry (IHC) for ERG and SPINK1 was performed at the Michigan Center for Translational Pathology as detailed recently³⁵. Briefly, primary antibodies used were rabbit monoclonal anti-ERG (EPR3864, 1:100 dilution, Cat no. 790–4576, Ventana Medical Systems-Roche) and a mouse monoclonal anti-SPINK1 (1:100; H00006690-M01, clone 4D4, Abnova). IHC was performed using an automated protocol on the Discovery XT autostainer (Ventana Medical Systems-Roche), with secondary antibodies and detection using brown/DAB chromogen for ERG and Red for SPINK1 (hematoxylin counterstain). Vascular endothelium was used as an internal positive control for ERG; normal pancreas was the positive control for SPINK1. Basal cell markers were assessed using an automated protocol on the Discovery XT autostainer using a cocktail (Ventana Medical Systems-Roche) containing primary antibodies against high molecular weight cytokeratin HMWCK (34βE12) and p63 (clone 4A4) and secondary ultraMap anti-msHRP antibody. Hematoxylin II was used as a counterstain. All immunostains were evaluated by S.C.S., L.P.K. and S.A.T.

Criteria used for assessment of the molecular status of an adenocarcinoma focus were as reported before^{35, 38}. Briefly, given the usual dichotomous ERG staining in prostate cancer, a focus was considered ERG⁺ if diffuse, strong nuclear staining was observed in all cancer cells (weaker staining was noted if present). A focus was considered SPINK1⁺ if >5% of the cancerous cells showed strong cytoplasmic positivity. A focus was considered dual ERG⁺/SPINK1⁺ if cancerous cells demonstrated diffuse positivity for ERG and ≥5% of these cells showed concomitant SPINK1⁺ cytoplasmic staining. Lack of basal cell markers (p63 and high molecular weight cytokeratin) was used to support a diagnosis of carcinoma in challenging cases. In cases where the morphologic dominant focus was found by ERG/SPINK1 dual stain to consist of two different clones, the focus with higher Gleason score was designated dominant, while the focus with larger maximum linear dimension was designated dominant among foci showing the same Gleason score.

Statistical Analysis

Descriptive statistics and associations among clinicopathological parameters from different cohorts were assessed using MedCalc 12. Two-tailed non-parametric two sample tests or Fisher exact tests were used for all comparisons and P values less than 0.05 were considered statistically significant.

RESULTS

We evaluated archival prostatectomy material for 24 cases of prostatic adenocarcinoma arising in HOXB13 G84E germline mutation carriers treated by prostatectomy at the University of Michigan. One case showed no carcinoma remaining on deeper sections and was excluded from further study. The 23 case final cohort showed a median age of 58.0 years, median preoperative PSA of 5.7 ng/mL, a median prostatectomy weight of 41.4g; all patients were of European descent. Pathologic staging showed that in 87% (20/23) cases the adenocarcinoma was organ confined (pT2a–b 9% (2/23); pT2c 78% (18/23); pT3a 9% (2/23); pT3b 4% (1/23)). No case showed lymphatic metastasis among the 18/23 cases where lymph nodes were dissected. Surgical margins were positive in 13% (3/23) of cases. Median follow-up time for the cohort was 36.0 months; a single patient experienced biochemical recurrence at 29.2 months and expired from metastatic hormone refractory carcinoma at 56.4 months. One subject had been treated by hormone therapy prior to prostatectomy, and was not Gleason scored. Of the remaining 22 cases, dominant foci were Gleason score 3+3=6 in 23% (5/22), 3+4=7 in 41% (9/22), 4+3=7 in 23% (5/22), and ≥8 in 14% (3/22). Per case, the median greatest linear dimension of the dominant focus was 1.2 cm. The dominant focus was based in the transition zone in 13% (3/23) of cases. A median of 6 cancer foci (range, 1–14) were identified per case. Table 1 lists clinicopathologic parameters per case for all evaluated patients.

Morphologic evaluation of the entire cancer burden of these cases was performed, identifying typical acinar prostatic adenocarcinoma in all cases, with no foci of ductal adenocarcinoma, urothelial carcinoma, or other neoplasm. Additionally, no foci of neuroendocrine, mucinous, signet ring, giant cell, sarcmatoid, lymphoepithelioma-like, oncocytic, or Paneth cell-like variants of acinar adenocarcinoma were identified. In contrast, morphologic evaluation of the 22 untreated cases showed prevalent pseudohyperplastic-type features (Figure 1 and Table 1)^{27, 41, 42}. Cytologic features of pseudohyperplastic carcinoma (tall/columnar epithelium with pale to finely granular or vacuolated cytoplasm and basally oriented nucleoli, Figure 1A–B) were present in 45% of dominant foci, while the architectural features (dilated glands with papillary infoldings, Figure 1C–D) were present in 41% of dominant foci (Table 1). Across all foci, the cytologic features were more common, with 82% of cases showing at least one such focus, than were architectural features, with only 55% of cases showing such a focus. Other features, associated with but not specific to pseudohyperplastic carcinoma, prevalent among this cohort were intraluminal pink amorphous secretions (at least one focus in 91% of cases, 20/22) and pink crystalloids (86%, 19/22), see Figure 1A–B. Additionally, three cases showed at least one focus of with features of foamy gland carcinoma (14%, 3/22), one case showed a focus of atrophic carcinoma (5%, 1/22), while another case showed a focus of acinar adenocarcinoma with a number of glands showing a PIN-like morphology, including stratified epithelium reminiscent of tufted PIN (5%, 1/22).

Figure 1. Prostate adenocarcinoma with pseudohyperplastic features occurs frequently in cancers arising in HOXB13 G84E carriers.

Morphologic assessment of all cancer foci arising in our cohort of HOXB13 G84E cases showed prevalent features previously associated with the pseudohyperplastic variant of prostatic adenocarcinoma. A subset of cases showed the (more specific) architectural features of this variant, including irregular/tufted luminal borders and cystically dilated glands with papillary infoldings, which involved the dominant focus in over half of cases (representative focus pictured in A., 4X, H&E, and B. 4X, basal cell cocktail). The less specific cytologic features of pseudohyperplastic carcinoma, including tall/columnar epithelium, pale to granular cytoplasm, and basally oriented nucleoli were even more prevalent. Pink amorphous luminal secretions were frequent (C., 40X, H&E), as were pink crystalloids (D., 40X, H&E), each present in a cancer focus in >85% of cases.

Several morphologic features, previously associated with the ERG⁺ prostatic adenocarcinoma molecular subtype⁴³ and not typical of pseudohyperlastic carcinoma, appeared infrequent during initial mapping studies. These included blue mucin, prominent cribriform growth pattern, and intraductal growth. Systematic re-review of all foci demonstrated that 23% (5/22) of cases showed at least one focus of carcinoma showing luminal blue mucin. Prominent cribriform growth and intraductal growth were also infrequent, with only 18% (4/22) and 5% (1/22) of cases showing such a focus, respectively. The benign prostatic parenchyma was anatomically and histologically unremarkable in these cases.

To assess basic molecular subtypes of prostate carcinoma arising in HOXB13 G84E carriers, we assessed ERG and SPINK1 status by dual IHC. Dual ERG/SPINK1 IHC was performed on full tissue sections representing the dominant foci and all unique mapped foci from all 23 cases to ascertain the prevalence of these two essentially mutually exclusive prostate cancer molecular subtypes^34–36. In HOXB13 G84E carriers, if positive, ERG showed strong, diffuse expression across all cancer cell nuclei in a cancer focus (Figure 2A–B), while SPINK1 IHC showed heterogeneous expression, particularly at the periphery and invasive front of foci (Figure 2C–D). Both of these staining patterns are similar to previous observations using the same antibodies in unselected prostate cancer cohorts^{33, 35, 38, 44}. Importantly, the ability to assess clonality of a focus through dual ERG/SPINK1 IHC revealed that the morphologic dominant focus was composed of two different clones in 5 of the 23 cases (22%), as shown in Figure 3A–D.

Figure 2. ERG and SPINK1 expression patterns are similar in HOXB13 G84E cases to genetically unselected cases.

In in ERG⁺ foci of adenocarcinoma (A., 10x, H&E), the staining pattern was of strong, diffuse, uniform nuclear positivity (B., 10X, ERG immunostain with DAB/brown chromogen), similar to previous reports using this clone/staining protocol^{35, 38, 44}. An asterisk shows an endothelial-lined lymphovascular space, with staining of internal positive control endothelial cells. In SPINK1⁺ foci (C., 10X, H&E), the staining pattern was of cytoplasmic positivity, frequently decorating glands at the periphery or invasive front, and heterogeneous (D., 10X, SPINK1 immunostain, red chromogen).

Figure 3. Dual ERG/SPINK1 immunostaining on whole sections identifies clonally independent cancer foci that are morphologically indistinguishable.

Several dominant foci of HOXB13 G84E cases were identified as “collision tumors” of clonally independent adenocarcinomas using ERG/SPINK1 dual immunostain for molecular subtyping (5 of 23 cases, 22%). The polyclonal nature of these foci was not otherwise distinguishable on morphologic grounds (representative examples: A.& C., 4X, H&E). As is evident on dual ERG/SPINK1 immunostain, such cases included collisions of ERG⁺/SPINK1⁻ and ERG⁻/SPINK⁺ clones (B., 4X) and collisions of ERG⁺/SPINK1⁻ and ERG⁻/SPINK1⁻ clones (D., 4X). ERG immunostain with DAB/brown chromogen; SPINK1 immunostain with red chromogen.

Dual IHC showed that the dominant focus was ERG⁻/SPINK1⁻ in 52% (12/23), ERG⁺/SPINK1⁻ in 17% (4/23), SPINK1⁺/ERG⁻ 26% (6/23), and ERG⁺/SPINK1⁺ in 4% (1/23) of cases. All but one case showed at least one ERG⁻/SPINK⁻ focus (96%, 22/23), while over half of cases showed at least one ERG⁺/SPINK⁻ or ERG⁻/SPINK1⁺ focus (57%, 13/23 and 52%, 12/23, respectively). Of the 145 independent foci mapped across all cases, 66% (95/145) were ERG⁻/SPINK1⁻, 17% (25/145) were ERG⁺/SPINK1⁻, 17% (24/145) were ERG⁻/SPINK1⁺, and a single focus, 1% (1/145) was ERG⁺/SPINK1⁺.

We noted several unique observations in these HOXB13 G84E-carrier cases. In addition to the single focus with a small area of ERG⁺/SPINK1⁺ prostate cancer embedded in a larger ERG⁺ focus (Figure 4A–B), we also observed a case with a collision between ERG⁺/SPINK1⁻ and ERG⁻/SPINK1⁺ foci. In this case, distinct ERG⁺/SPINK1⁻ and ERG⁻/SPINK1⁺ cells were present in the same gland (Figure 4C–D), which we interpret as glandular fusion between separate clones. Also of note, one case demonstrated aberrant diffuse expression of p63 among the carcinoma cells at the time of diagnostic needle biopsy workup, a rare aberrant immunophenotype reported previously^{45, 46}. To assess for the prevalence of this feature among HOXB13 G84E carriers, IHC for basal cell markers (HMWCK/p63) was performed on full sections containing the dominant focus for all cases (as well as on any additional foci equivocal for carcinoma, see Materials and Methods, 47 total sections stained). Review of the stained sections redemonstrated only this single case with dominant focus showing aberrant p63⁺ immunophenotype carcinoma (ERG⁻/SPINK1⁻) (Figure 4E–F), which showed an atrophic to basaloid morphology consistent with Gleason score 4+5=9, though grading remains controversial in these cases^{46, 47}. Intriguingly, this patient also harbored a distinct focus of acinar adenocarcinoma, with a number of glands showing a PIN-like morphology, which was ERG⁺/SPINK1⁻ (Figure 4G–I).

Figure 4. Unique immunomorphologic observations in HOXB13 G84E cases.

One case with conventional acinar adenocarcinoma morphology (A., 10X, H&E) showed a dominant focus that was ERG⁺/SPINK1⁺ (B., 10X, ERG/SPINK1 dual immunostain, ERG in DAB, SPINK1 red chromogen), a phenomenon previously identified in only one of nearly 4500 genetically unselected cases^{35, 36}. In another case, apparent glandular fusion between independent ERG⁺/SPINK1⁻ and ERG⁻/SPINK1⁺ clones was seen in a collision focus (C., 20X, H&E); individual cells within the same gland showed either nuclear positivity for ERG (DAB/brown) or cytoplasmic positivity for SPINK1 (red chromogen), without any dual positive cells (D., 20X, ERG/SPINK1 dual immunostain). In a single case, the dominant focus (ERG⁻/SPINK1⁻, E., 20X, H&E) showed aberrant diffuse nuclear expression of p63 (F., 20X, basal cell cocktail), a rare immunophenotype reported previously⁴⁵. Within the same section, this case also showed a separate focus of acinar adenocarcinoma with a number of glands showing a PIN-like morphology (G., H&E, H., basal cell cocktail, both 10X). This focus was ERG⁺/SPINK1⁻ (I., 10X, ERG immunostain).

DISCUSSION

Recently, Ewing et al. determined through epidemiologic studies and targeted next generation sequencing of over 90 multicase families that a germline variant in the homeobox gene, HOXB13 (G84E), was associated with predisposition to prostate cancer⁷. Subsequent studies have confirmed that the HOXB13 G84E variant is rare (carrier frequencies of 0.1–1.5%), most common among men of Western European descent, and, particularly amongst young men with a positive family history, associated with increased prostate cancer risk (odds ratios ~3–10)^8–19. However, the pathologic features of such cases have not been studied in detail. Thus, we performed a systematic clinicopathologic, histomorphologic, and molecular review of the University of Michigan subset of the Ewing et al. HOXB13 G84E cases.

In terms of morphology, we observed a prevalent and characteristic appearance of the carcinomas arising in a cohort of 23 HOXB13 G84E germline mutation carriers: a substantial majority of these cases showed foci of carcinoma with a bland appearance, reminiscent of benign hyperplasia. Cancers demonstrating this constellation of features have been previously described by Humphrey et al. as the pseudohyperplastic variant of prostate carcinoma, an infrequent variant^{27, 41, 42}. Review of all cancer foci in these HOXB13 G84E cases demonstrated frequent pseudohyperplastic-type features among the cancers, involving the dominant focus in 45% and at least one cancer focus in 82% of these cases. As has been described for pseudohyperplastic cancers, there was very frequent admixture of conventional acinar adenocarcinoma. Such nonspecific features should not serve as a screening tool to identify HOXB13 G84E-related cases (genotyping is required). However, awareness of this prevalent, “pseudobenign”^{41, 42} morphology may be of use in biopsy evaluation of individual cases from known HOXB13 G84E kindreds, especially as experts contemplate increased surveillance of genetically affected individuals^{9, 11}.

In this HOXB13 G84E-related cohort, we observed a median age of 58 years and a median serum PSA at diagnosis of 5.7 ng/mL; all patients were of European descent. A median of 6 cancer foci per case were identified and the dominant focus showed Gleason score 6, 3+4=7, 4+3=7 and ≥ 8 in 23%, 41%, 23% and 14% of cases, respectively. We have recently studied two additional cohorts suitable for comparison. We employed this same focus mapping and ERG molecular subtyping strategy in a genetically unselected prostatectomy cohort (N=41, also of predominantly European descent, 86%), referred hereafter as “Cohort 2”, with median age of 59 years and PSA of 5ng/mL³⁸. Cohort 2 showed a median of 3 cancer foci, and dominant foci of Gleason Score 6, 3+4=7, 4+3=7 and ≥ 8 in 17%, 61%, 15% and 7% of cases, respectively. We also recently applied our dual ERG/SPINK1 IHC in a tissue microarray study of dominant cancer foci of 228 genetically unselected men undergoing prostatectomy, referred hereafter as “Cohort 3”. Cohort 3 demonstrated a median age of 61 years, median serum PSA of 6.2ng/mL frequent European descent (94%)³⁵. Dominant foci showed Gleason Score 6, 3+4=7, 4+3=7 and ≥ 8 in 19%, 49%, 20% and 11% of cases, respectively. Compared to Cohorts 2 and 3, our current cohort of HOXB13 G84E carriers was significantly younger than Cohort 3 (Mann-Whitney U-test P=0.002); however, they were otherwise quite similar, suggesting valid comparisons may be made with respects to distributions of molecular subtypes of carcinoma.

Thus, in addition to evaluating histologic features of HOXB13 G84E-related carcinomas, we also assessed basic molecular subtypes by IHC. We and others have shown that approximately 50% of PSA-screened prostate cancers harbor recurrent gene rearrangements that fuse regulatory regions of androgen regulated genes to ETS transcription factor family members (most commonly TMPRSS2 fused to ERG)^{28, 30, 31}. ERG IHC has been extensively validated as a diagnostic biomarker for carcinoma^{30, 33, 44}; thus, determination of the prevalence of this molecular subtype in HOXB13 G84E cases would be of use for evaluation of future biopsies for such patients. Assessing all foci of cancer arising in HOXB13 G84E carriers for ERG/SPINK1 status by IHC, we observed a remarkably low prevalence of ERG⁺/SPINK1⁻ foci. Amongst dominant foci, only 22% were ERG⁺, with 17% of all foci ERG⁺ (including the ERG⁺/SPINK1⁺ focus). For comparison, in the similarly mapped and stained Cohort 2³⁸, we observed significantly higher ERG+ prevalence (59% of dominant foci and 49% of all foci, P=0.008 and P<0.0001, respectively vs. HOXB13 Cohort). The same trend was apparent in Cohort 3³⁵, where we found that 43% of dominant foci used for TMA construction were ERG⁺ (P=0.07 vs. HOXB13 Cohort). For clarity, Figure 5A plots these key differences in ERG prevalence graphically. Further contrasting the markedly low prevalence of ERG+ cancer in the HOXB13 G84E cases is comparison to a recent meta-analysis of >10,000 cases by Stephensen et al. who reported that 47% of prostate cancers were TMRPSS2:ERG^{+ 31}.

Figure 5. Prevalence of prostate cancer molecular subtypes in HOXB13 G84E carriers compared to other reported cohorts.

A. Among the HOXB13 G84E-related cohort described in this study (first panel), we found significantly lower prevalence of ERG⁺ dominant foci (outer rings, pie charts) and total ERG⁺ foci among all mapped foci (inner rings, pie charts) than in our recently reported³⁸, unselected Cohort 2 (second panel). Lower prevalence of ERG⁺ dominant foci among HOXB13 cases was also noted as compared to our recently reported³⁵, unselected Cohort 3, where only the dominant nodule was evaluated (third panel, outer ring of pie chart). B. In contrast, among the HOXB13 G84E-related cohort (first panel), we found a significantly higher prevalence of SPINK1+ dominant foci (outer rings, pie charts) as compared to unselected Cohort 3 where only the dominant nodule was evaluated (second panel, outer ring of pie chart). All P-values are for two-tailed Fisher Exact tests comparing the HOXB13 G84E-related cohort to Cohorts 2 or 3.

Although ERG status has been evaluated in several variants of prostate cancer finding similar rates of ERG⁺ as typical acinar carcinoma^48–50, we are unaware of published reports of ERG status in the pseudohyperplastic variant. Thus, it is unclear whether the observed lower prevalence of ERG⁺ cancers is a cause, effect, or only associated with the morphology observed. Moreover, histologic features previously associated with the ERG⁺ subtype⁴³ were uncommon in our current cohort. Lower rates of ERG⁺ prostate cancer have been reported in non-Caucasian populations (~25–40%%) and cancers arising in the transition zone^{31, 51–56}, neither of which were prominent in our cohort. The disparity of ERG⁺ prevalence in this cohort is even more remarkable compared to higher rates of ERG⁺ cancer reported among early onset cancers and/or those with low serum PSA⁵⁷.

Similarly, approximately 5–10% of prostate cancers, which are almost exclusively ETS fusion negative, show marked overexpression of SPINK1^{34, 58–60}. Thus, SPINK1 expression defines a second molecular class of ETS fusion negative carcinomas. While the clinical significance of SPINK1 expression remains unknown, preclinical studies in animal models have proven the principle that it is an attractive therapeutic target³⁴. Opposite the observation of low ERG⁺ prevalence in this HOXB13 G84E-related cohort, we found that the SPINK1⁺ molecular subtype of prostate cancer was over-represented in HOXB13 G84E carriers, with 30% of men harboring SPINK1⁺ dominant foci and 17% of all foci SPINK1⁺ (including the ERG⁺/SPINK1⁺ focus). Although we are unaware of studies assessing SPINK1 status across all foci in multifocal prostate cancer, in our initial characterization of unselected cases, we found that ~10% of prostate cancers were SPINK1^{+ 34}, while recent tissue microarray studies on >12,000 cases suggests that the SPINK1⁺ prevalence is ~5–6%^{36, 61}. In Cohort 3, we found that amongst 228 unselected localized prostate cancers only 5% were SPINK1⁺ (P=0.003 vs. HOXB13 cohort, see Figure 5B)³⁵. Our findings of a case with an ERG⁺ dominant focus showing patchy SPINK1 expression and a case with a dominant focus (ERG⁻/SPINK1⁻) with aberrant expression of p63⁴⁵ again underscore the uniqueness of these cases.

Of more general relevance to the diagnosis and study of prostate cancer are observations regarding clonality of carcinoma. Given that ETS gene fusions are highly clonal in an individual focus^{29, 30}, IHC for ERG, either alone or in combination with SPINK1, can be used to aid in identifying genetically distinct foci in multifocal prostate cancer. Importantly, in five cases, we identified a dominant focus that was in fact a “collision tumor” of genetically distinct clones that would otherwise have been considered a single focus by histology. Importantly, our observation of frequent dominant foci composed of “collision tumors” that can be appreciated on whole sections may explain a recent report of frequent intrafocal ERG heterogeneity observed in large tumors sampled by multiple tissue microarray cores⁶². These observations call for rigorous whole-section-based studies of dominant focus clonality and whether molecular assessment could assist in dominant focus assignment and Gleason scoring.

In summary, in this first histologic and molecular survey of cancers arising in HOXB13 G84E carriers, we identified aberrant morphologic and molecular features. Such findings need to be confirmed in additional cohorts of cancers arising in HOXB13 G84E carriers but may be relevant to future diagnostic practice. Preliminarily, however, these findings imply that novel molecular pathways may drive carcinogenesis in these cases and support comprehensive assessment of molecular alterations in prostate cancers arising in HOXB13 G84E carriers.