Performance of DAXX Immunohistochemistry as a Screen for DAXX Mutations in Pancreatic Neuroendocrine Tumors

Abstract

Objectives:

DAXX immunohistochemistry (IHC) is often used as a surrogate for sequencing. We aimed to elucidate the sensitivity of IHC for DAXX mutation.

Methods:

All pancreatic neuroendocrine tumors (PanNETs) with DAXX mutations detected by sequencing and a subset of DAXX wild type PanNETs were analyzed for DAXX expression by IHC.

Results:

Of 154 PanNETs with MSK-IMPACT testing, 36 (30%) harbored DAXX mutations. DAXX mutations were associated with TSC2 mutations (46% vs. 10%, P < 0.0001), tended to co-occur with MEN1 mutations (63% vs 49%, P = 0.11), and tended to be mutually exclusive with ATRX mutations (11% vs 25%, P = 0.053). Of 27 available DAXX mutant PanNETs, 23 lost DAXX expression (85.2%). All 4 DAXX mutants with retained expression harbored DAXX mutations within the SUMO-interacting motif of the last exon. Telomere-specific fluorescence in situ hybridization demonstrated alternative lengthening of telomeres in all 4 cases. Of 20 PanNETs with wild type DAXX, 19 retained DAXX IHC expression (95%).

Conclusions:

The sensitivity and specificity of IHC for DAXX mutation is 85% and 95%, respectively. Last exon DAXX mutant PanNETs often show alternative lengthening of telomeres despite retained DAXX expression, likely due to escape of non-mediated decay.

INTRODUCTION

Pancreatic neuroendocrine tumors (PanNETs) are the second most common primary tumor of the pancreas and occur in approximately 1 in 100,000 persons annually.¹ These tumors have widely variable behavior ranging from incidental and indolent, to high grade metastatic neoplasms. While mitotic rate and proliferation index are commonly reported and used to manage patients, the status of two tumor suppressors has recently been reported to be important prognostically.^2–5

Along with a-Thalassemia/ x linked mental retardation (ATRX) gene, Death domain associated protein (DAXX) gene encodes a nuclear protein that regulates the deposition of histone H3.3 during the assembly of telomeric chromatin. DAXX function is paramount to various nuclear functions including transcription and cell cycle regulation (including apoptosis).^2–4 Recently, it has been shown that over 40% of PanNETs harbor mutations in either DAXX or ATRX, and that these mutations correspond with the presence of alternative lengthening of telomeres (ALT).⁵ DAXX and ATRX IHC have since been used in several studies as surrogates for mutation analysis, but 3 studies have found that ALT is present in 6–21% of PanNETs with retained DAXX and ATRX IHC.^5–7 Here, we investigate the relationship of DAXX mutations with DAXX protein expression via immunohistochemical analysis and show that the sensitivity of DAXX IHC depends on the site of DAXX mutation.

MATERIALS AND METHODS

Retrospective analysis of DAXX (transcript NM_001141970) mutations was performed on a cohort of PanNETs that had undergone molecular sequencing with MSK-IMPACT, a hybridization-based capture assay that identifies somatic mutations, copy number changes, microsatellite instability, and select structural rearrangements in over 400 genes against a patient’s matched normal sample.⁸ Domain location for mutant proteins was identified using the sequence based predicted modular organization of DAXX.³ P values for mutation associations were calculated using two tailed chi square tests with statistically significant findings of P < 0.05/(number of variables) to include Bonferroni correction.

Immunohistochemical analysis of DAXX in situ protein expression was performed on all DAXX mutant PanNETs with available tissue, and on 20 PanNETs with wild type DAXX. For the immunohistochemical detection of DAXX, rabbit monoclonal antibody (mAb) clone E94 (ab32140; Abcam, Cambridge, Mass) was used on all cases tested. Select DAXX mutant PanNETs were also tested with a second immunohistochemical antibody for DAXX: rabbit mAb clone25C12 (#4553; Cell Signaling Technology, Danvers, Mass). While the immunogen for mAb E94 was not further specified, mAb 25C12 was generated with a synthetic peptide surrounding aa 259 of the DAXX protein. All assays were performed on a Leica Bond-III automated staining platform (Leica, Buffalo Grove, Ill). E94 was optimized on panels of normal tissue and mutated PanNETs. Best staining was achieved at a dilution of 1:300 (0.8 ug/ml) using a 30 minutes heat-based antigen retrieval step employing a high pH buffer solution (ER2, Leica). A polymer-based secondary kit (Refine, Leica) was used for the detection of the primary. Immunohistochemical interpretation was based nuclear staining: complete loss of nuclear staining was considered ‘lost’ or ‘negative’. Non-neoplastic lymphocytes served as positive internal controls for each sample.

In cases with DAXX mutation and retained IHC, alternative lengthening of telomeres was assessed by Telomere-specific FISH. Cy3-labeled peptide nucleic acid (PNA) probe (TTAGGGTTAGGGTTAGGG 3’) from Applied Biosystems was used and samples hybridized according to manufacturer’s protocol with few modifications. Briefly, paraffin sections were dewaxed, microwaved in 10 mM sodium citrate buffer, pretreated with pepsin, washed, and dehydrated in cold ethanol series. The PNA Probe and target DNA were co-denatured for 8 minutes, incubated overnight at 37°C, washed twice in 2XSSC+0.1% Tween20, rinsed with sterile distilled water and counterstained with antifade medium containing DAPI (Vectashield, Vector Laboratories). Slides were scanned using a Zeiss Axioplan 2i epifluorescence microscope equipped with a megapixel CCD camera (CV-M4+CL, JAI) controlled by Isis 5.5.9 imaging software (MetaSystems Group Inc, Waltham, Mass). Metafer and VSlide module within MetaSystems were used to generate virtual image of the tissue sections (DAPI and H&E); in all, corresponding H&E section was used to identify tumor regions.

The entire section was hybridized and scanned through 63X for the presence of ALT-positive cells. Cells exhibiting the characteristic large, ultrabright, intranuclear signal (telomere sequence aggregates) were considered ALT-positive. A case was considered ALT-positive if five or more 63X fields contained ALT-positive cell(s). In all, normal stromal elements/regions were also analyzed to confirm the restriction of ALT-signals to tumor cells.

RESULTS

DAXX Mutations: Molecular Profiling

Out of 154 PanNETs with MSK-IMPACT results, 46 (30%) had DAXX mutation(s) (Fig. 1). DAXX mutations were significantly correlated with MEN1 and TSC2 mutations and trended to be mutually exclusive with ATRX mutations. Consistent with its role as a tumor suppressor, no recurrent hotspots were identified. Sixty-four mutations were identified in 46 PanNETs, including 25 nonsense mutations, 20 frameshift deletions, 17 missense mutations, 1 splice site mutation, and 1 in frame deletion.

FIGURE 1.

Distibution of DAXX mutations across the gene. DAXX has 6 regions of conserved sequences. The first and last correspond to SUMO-interaction motifs (SIMs), then there are 2 helical segments including the DAXX helix bundle (DHB) domain. Towards the C terminal, the “acidic” region is present, which contains 80% Glu/Asp residues, followed by segments rich in Ser/Pro/Glu (SPE) and Ser/Pro/Thr (SPT) residues. Mutation were distributed throughout the gene without hotspots. Black dots represent truncating mutations, green dots represent missense mutations, and the brown dot represents an in frame indel.

Of the 46 PanNETs with DAXX mutations, 5 had ATRX mutations, 29 had MEN1 mutations or deletions, and 21 had TSC2 mutations or deletions. Of the 108 PanNETs with wild type (WT) DAXX, 27 had ATRX mutations, 53 had MEN1 mutations or deletions, and 11 had TSC2 mutations or deletions (Table 1). TSC2 mutations were significantly enriched in DAXX mutant PanNETs (P = 0.0001) while ATRX mutations trended to be mutually exclusive with DAXX mutations without reaching statistical significance (P = 0.053).

TABLE 1.

Recurrent Mutations in Pancreatic Neuroendocrine Tumors and Their Relationship to DAXX Mutation Status

	MEN1 Mutant, n (%)	TSC2 Mutant, n (%)	ATRX Mutant, n (%)
DAXX mutant (n = 46)	29 (63)	21 (46)	5 (11)
DAXX WT (n = 108)	53 (49)	11 (10)	27 (25)
P	0.11	0.0001	0.053

DAXX Immunohistochemistry

Twenty-seven PanNET cases with known DAXX mutation were available for immunohistochemical analysis. Of these DAXX mutant PanNETs, 23 showed loss of expression while 4 had retained expression (Figs. 2A, 2B; Table 2). All 6 DAXX mutant PanNETs tested with the second immunohistochemical antibody (clone 25C12) were concordant. Each of the 4 cases with retained expression harbored DAXX mutations that occurred in the small ubiquitin like modifier (SUMO)-Interaction motifs (SIMs) in the last exon, 2 of which were nonsense mutations that would not result in transcription of the affected SIM sequence (Fig. 2C, Table 2). Consequently, the sensitivity of DAXX loss for DAXX mutations was 85.2% and dependent on the presence of mutations in other than the last exon.

FIGURE 2.

Retained expression of DAXX in a distal DAXX mutant pancreatic neuroendocrine tumor positive for alternating lengths of telomeres (ALT). A, Microscopic sections showed an intermediate grade pancreatic neuroendocrine tumor with solid architecture, vascularization, and moderate to abundant cytoplasm. B, DAXX IHC from the same tumor shows retention of expression in nuclei with strong, uniform intensity. C, On Integrated Genomics Viewer, the tumor from (2A) and (2B) a nonsense mutation is present on the minus strand at 72%, signifying loss of heterogeneity, within the distal SUMO-interaction motif. D, Large, ultrabright, intranuclear FISH signals (orange) indicative of ALT (arrows) (scale bar = 5um).

TABLE 2.

DAXX IHC Results According to DAXX Mutation

DAXX IHC (mAb E94)	AA	NT	Exon	DOMAIN
Negative/lost	Y71*	213C>G	2	DHB
Negative/lost*	Q89*	265C>T	3	DHB
Negative/lost	E116*	346G>T	3	DHB
Negative/lost*	C118*	354C>A	3	DHB
Negative/lost	L146P	437T>C	3	DHB
Negative/lost	H175fs	524_527delACCT	3
Negative/lost	Q205*	613C>T	3	Helical
Negative/lost	G262C; E466*	784G>T; 1396G>T	3, 5	Helical
Negative/lost	N280S	839A>G	3	Helical
Negative/lost	N280K	840C>G	3	Helical
Negative/lost*	Q320*; G531R	958C>T; 1591G>A	3, 6	Helical, Helical
Negative/lost	R330*	988C>T	3	Helical
Negative/lost	C350R	1048T>C	3	Helical
Negative/lost*	Q395*	1183C>T	4	Helical
Negative/lost	D419fs	1255delG	4
Negative/lost*	E466*	1396G>T	5
Negative/lost	M482_Q483delinsI*	1446_1447delinsAT	5	SPE
Negative/lost	R602Gfs*12	1803delC	6
Negative/lost	F625Lfs*31	1875_1878delCAAT	6
Negative/lost	Q650*	1948C>T	6
Negative/lost*	Y660*	1977–1_1979dupGCTA	6
Negative/lost	H723Ifs*55	2156delC	7	SPT
Medium/retained	E743*	2227G>T	8	SIM
Medium/retained	S749*	2246C>G	8	SIM
Medium/retained	D750fs	2247dupA	8	SIM
Medium/retained	D750V	2249A>T	8	SIM

^*Also tested with clone 25C12 with concordant results.

DHB indicates DAXX helix bundle; SPE, domain rich in Ser/Pro/Glu residues; SPT, domain rich in Ser/Pro/Thr residues; SIM, sumo-interaction motif.

Twenty PanNETs with wild type DAXX also underwent immunohistochemical staining for DAXX. Nineteen of 20 (95%) PanNETs with WT DAXX displayed nuclear immunopositivity indicating retained expression, yielding a specificity of 95%. The single case with loss of DAXX expression and WT DAXX did not harbor any coding mutations, structural variants, deletions, or intronic mutations in DAXX. The DAXX staining on this case was technically satisfactory; positive DAXX labeling was observed in both internal (lymphocytes) and external controls.

Telomere-specific FISH

The only PanNETs with discordant DAXX mutation status and protein expression status were 4 PanNETs with DAXX mutations in the last exon, WT ATRX, and retained DAXX IHC. Telomere specific FISH was performed in all 4 cases, and all 4 cases were positive for ALT (Fig. 2D).

DISCUSSION

In this study, we show that DAXX IHC is specific (95%) for DAXX mutation in PanNETs, and sensitivity is approximately 85%, with retained DAXX expression present in all 4 of 4 PanNETs with mutations in the last exon.

Previous studies have shown the clinical importance of ALT and its association with DAXX or ATRX loss. Specifically, reduced survival and metastatic disease have been associated with DAXX loss.^5,7 Interestingly, VandenBussche et al, Marinoni et al and Singhi et al each separately found that 21%, 15% and 6%, respectively, of PanNETs positive for ALT had retained DAXX and ATRX expression on IHC.^5–7 These results suggest that DAXX/ ATRX IHC are not 100% sensitive for identifying PanNETs with ALT.

The DAXX gene is 8 exons and 752 amino acids in length. In our study, all cases harboring mutations in exons 1–7 were detected by immunohistochemical loss of expression. Interestingly, the 4 cases with retained DAXX protein expression all harbored mutations in the last exon. Consequently, the sensitivity of DAXX protein expression analysis depends on the site of the mutation. A likely reason for the retained expression of DAXX in the 4 cases of DAXX mutant PanNETs is escape of nonsense mediated decay. All 4 cases of DAXX mutant PanNET with retained DAXX expression harbored mutations in the last exon, and it is known that nonsense mediated decay is inefficient for nonsense mutations in the last exon.⁹

While DAXX expression was retained in all PanNETs with DAXX mutations in the last exon, all 4 of these mutations (including 2 nonsense mutations) occurred in an important area: the SUMO- interacting motif. This domain has important roles in DAXX function, including sensitizing cells to stress-induced apoptosis² and enhancing interactions between PML and DAXX which is important for apoptosis.¹⁰ The fact that ALT was present on telomere-specific FISH in all 4 cases with DAXX mutations in the last exon, retained DAXX, and WT ATRX also supports the likely scenario that presence of DAXX protein is not a fully sensitive marker of adequately functioning DAXX protein when DAXX mutations occur in the last exon.

There was a single case of DAXX WT PanNET with loss of DAXX on IHC. Lymphocytes served as an internal positive control for DAXX IHC in this case. While there was not an indication of a genetic DAXX abnormality on MSK-IMPACT, the gene likely was inactivated in another way: either by epigenetic silencing or an intronic event not covered by MSK-IMPACT.

Several limitations to this study exist, including a sample size too small for survival and response analysis according to DAXX mutation, as well as a relatively small number of DAXX mutations in the last exons.

In conclusion, we show that DAXX IHC has a sensitivity of approximately 85% for the presence of DAXX mutations in PanNETs. The discordant cases are largely a reflection of the site of mutation since mutations in the last exon are not concordant with IHC results.