p16 Immunohistochemistry as a Screening Tool for Homozygous CDKN2A Deletions in CNS Tumors

Abstract

The 2021 World Health Organization classification of tumors of the central nervous system emphasizes the significance of molecular parameters for an integrated diagnosis. Homozygous deletion of cyclin-dependent kinase inhibitor 2a (CDKN2A) has been associated with an adverse prognosis in IDH-mutant gliomas, supratentorial ependymomas, meningiomas, and MPNST. In this study, we examined the value of p16 protein immunohistochemistry as a rapid and cost-effective screening tool for a homozygous CDKN2A deletion. Genetic analyses for CDKN2A in 30 pleomorphic xanthoastrocytomas, 32 IDH-wild-type high-grade gliomas, 40 supratentorial ependymomas with ZFTA-RELA gene fusion, 21 IDH-mutant astrocytomas, and 24 meningiomas were performed mainly by a molecular inversion probe assay, a high-resolution, quantitative technology for the assessment of chromosomal copy number alterations. Immunohistochemistry for p16 proved to have a high positive predictive value (range 90% to 100%) and an overall low negative predictive value (range 22% to 93%) for a homozygous CDKN2A deletion. In a setting where molecular testing is limited for cost and time reasons, p16 immunohistochemistry serves as a useful and rapid screening tool for identifying cases that should be subjected to further molecular testing for CDKN2A deletions.

The identification of genetic alterations as diagnostic or prognostic criteria has revolutionized the modern approach to tumor pathology. With the introduction of the current 2021 World Health Organization (WHO) classification of tumors of the central nervous system (CNS), even more tumor types are now being defined by underlying (epi) genetic events.¹

The protein p16^INK4a (p16) is encoded by the cyclin-dependent kinase inhibitor 2a (CDKN2A) gene located at 9p21. p16 plays a crucial role in the maintenance of the hyperphosphorylated state of the Rb family proteins, which are important for inducting G1 cell cycle arrest.^2–6 p16 negatively regulates the CDK4-Cyclin D complex, which in turn serves as a negative regulator of the Rb protein.⁷ There is the presumption that decreased expression of p16 or pRb or increased activity of CDK4 can trigger cell proliferation.⁸

CDKN2A is frequently inactivated in cancer, often by homozygous deletion (HD), but also by promoter hypermethylation or loss of heterozygosity and concurrent point mutations.⁹ Rates of p16 inactivation of >85% have been reported in pancreatic carcinoma, 60% in leukemia and melanoma, 50% to 70% in squamous cell carcinoma of the head and neck, 30% in colorectal carcinoma, and 20% in breast carcinoma.⁹

In Isocitrate dehydrogenase (IDH)-mutant gliomas, CDKN2A/B deletions have been associated with a poor prognosis,^10,11 and the detection of a HD of CDKN2A and/or CDKN2B is now being recognized as a criterion for the diagnosis of a CNS WHO grade 4 astrocytoma, IDH-mutant.¹ Moreover, CDKN2A HDs were found to indicate an adverse outcome in pediatric supratentorial ependymoma (ST-EP) with RELA-fusion.¹² In meningioma, CDKN2A/B HD has been associated with a higher risk of malignant progression or early recurrence^13,14 and has now been defined as a criterion for CNS WHO grade 3 anaplastic meningioma.¹ Deletions of the CDKN2A gene at 9p21 represent an early event in the malignant transformation of neurofibromas into malignant peripheral nerve sheath tumor.¹⁵ HDs of the CDKN2A region are encountered in >80% of pleomorphic xanthoastrocytoma (PXA) and are central to the diagnostics of PXA.^16,17

Usually, p16 protein expression is not seen in normal brain tissue, which is reflected by a negative immunostain.⁸ To be immunohistochemically detectable, the expression of p16 has to be upregulated.¹⁸ It has been assumed that p16 immunopositivity can be explained by other alterations of the p16-CDK4-pRb pathway (eg, CDK4 amplification, RB loss of heterozygosity).⁸

For epithelial tumors associated with papillomavirus infection, p16 immunohistochemistry has already been established as an accurate diagnostic tool.^5,19,20 For mesothelioma, it has been reported that immunohistochemical assessment of p16 expression does not correlate well with CDKN2A HD determined by fluorescence in situ hybridization (FISH).²¹ To date, only a few studies regarding the value of p16 immunohistochemistry in the assessment of CDKN2A deletions in brain tumors are available in the literature.^10,22 The CDKN2A copy number status in glial tumors has often been evaluated by deduction from methylation profiles.^10,14 However, methylation profiles typically do not allow an exact quantitative assessment of gene copy number status. In this study, the CDKN2A status was determined by a molecular inversion probe assay (MIP), which facilitates precise identification of homozygous CDKN2A deletions in tumors. Here, we explore the validity of p16 immunohistochemistry as a surrogate marker for a CDKN2A HD in diffuse and nondiffuse glial tumors and meningiomas.

METHODS

Tumor Samples

We examined 147 tumors arising in the CNS, including 30 PXA, 32 high-grade gliomas, IDH-wild-type (HGG), 21 IDH-mutant astrocytomas (A), 40 suprapentorial ependymoma, and 24 meningiomas. Formalin fixed paraffin embedded tissue samples were retrieved from the archives of the Institute of Neuropathology, University of Bonn Medical Center and the DGNN German Brain Tumor Reference Center. Tumor diagnosis was made according to the 2021 WHO classification of tumors of the CNS¹ by consensus of at least 2 experienced neuropathologists (T.P., F.A., and V.Z.). All specimens were examined on the basis and according to the legal requirements of the revised Declaration of Helsinki of the World Medical Association in 1983. A proportion of the patients had either been included in the HIT2000 ependymoma trial, the SIOP-LGG-trial/register, or the HIT-HGG study.

Immunohistochemistry

Immunohistochemistry with antibodies against p16^INK4a (mouse monoclonal, E6H4, Roche Diagnostics) was performed on 4-µm thick formalin fixed paraffin embedded tissue sections on a Ventana Benchmark XT Immunostainer (Roche Ventana). Immunohistochemical slides were considered positive when tumor cells showed either nuclear or both nuclear and cytoplasmic staining. At least 2 neuropathologists (T.P., F.A., and V.Z.) independently evaluated stained slides without knowing the CDKN2A status.

DNA Extraction and MIP Assay

Hematoxylin-eosin–stained sections of each case were reviewed carefully for sufficient tumor content (>80%) before DNA extraction. Using the QIAmp DNA Mini Tissue Kit (Quiagen GmbH), DNA was extracted from all cases according to the manufacturer’s protocol. Genomic copy number losses and gains were assessed by the MIP array (OncoScan CNV Plus Array, Affymetrix). The MIP array includes 335,000 inversion probes (version V2.0) with a median probe spacing of 2.4 kb. Thirty-four probes cover the CDKN2A gene locus. MIP analysis was performed using at least 80 ng of tumor DNA, as previously described.^23,24 In brief, all probes contain 2 genomic homology regions, each flanking a single nucleotide polymorphism site. Gaps are filled and ligated after annealing to the DNA, followed by digestion of the remaining noncircularized probes by exonucleases. After cleavage, the now inverted probes are amplified by polymerase chain reaction using universal primers, followed by labeling with fluorescent molecules and hybridization to oligonucleotide chip arrays. Raw data were analyzed using the Nexus Copy Number 8.0 Discovery Edition software (BioDiscovery). To obtain copy number and loss of heterozygosity estimations, we used the manufacturer’s single nucleotide polymorphism-FASST2 segmentation algorithm.

850k Methylation Analysis

Tumor DNA was converted by bisulfite treatment and analyzed on an Infinium Human Methylation EPIC (850 k) (Illumina) according to the manufacturer’s protocol. Methylation data obtained from the array were analyzed with the Heidelberg methylation brain tumor classifier, version v11b4 (www.molecularneuropathology.org).²⁵

Multiplex Ligation-dependent Probe Amplification Analysis

Multiplex ligation-dependent probe amplification analysis (MLPA) was performed using Salsa Probe Mix P088 (MRC Holland) according to the manufacturer’s instructions.

Statistical Analysis

Statistical analysis was performed using the GraphPad Prism tool, version 9 (GraphPad Prism 9.0.0, GraphPad Software). Results were analyzed with a 2-tailed Fisher exact test and considered significant when P <0.05.

RESULTS

Case Cohort

We explored 30 PXAs (18 female, 12 male), with patients’ mean age being 20 years (17 pediatric cases < 18 years). Clinical diagnosis in 27 tumors was PXA CNS WHO grade 2 according to the 2021 WHO brain tumor classification,¹ and 3 cases fulfilled the criteria for a grade 3 tumor (Fig. 1). A BRAF V600E mutation was found in 25 PXAs (83%). One tumor was a recurrent PXA. Patients with high-grade IDH-wild-type gliomas (n=32) were older (mean age 34.8 years, 14 female and 16 male patients). Among glioblastomas, 17 showed features of a giant cell glioblastoma and 3 of an epithelioid glioblastoma. Three cases qualified as diffuse hemispheric glioma, H3 G34-mutant and one case as diffuse midline glioma, H3 K27-altered. All ST-EP (n=40, 17 female, 23 male) occurred in pediatric patients (mean age 8 years). ST-EPs were graded as CNS WHO grade 3 tumors, and ZFTA (C11orf95)-RELA fusions were detected in all 40 cases (Fig. 1, Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/PAS/B665). Our cohort also comprised 21 astrocytomas, IDH-mutant (mean age 39 years, 12 male and 9 female patients). Astrocytoma grading was based on morphologic features with 2 cases being CNS WHO grade 2, 16 cases CNS WHO grade 3, and 3 cases CNS WHO grade 4. Among all 24 meningioma, the mean age was 61 years, with 11 female and 13 male patients. Four meningioma corresponded to CNS WHO grade 1, 8 meningioma were atypical (CNS WHO grade 2), and the remaining 12 cases qualified for the diagnosis of anaplastic meningioma (CNS WHO grade 3).

FIGURE 1.

Case series (WHO diagnosis, location, p16 protein expression, and CDKN2A mutational status). CPA indicates cerebellopontine angle; DHG, diffuse hemispheric glioma; DMG, diffuse midline glioma; GBM, glioblastoma; IHC immunohistochemistry; pHGG, pediatric high-grade glioma.

Genetic Analysis of CDKN2A Deletion

High-resolution genomic copy number analysis was performed by MIP in 134 cases. The estimation of the CDKN2A status was performed by manual review of its gene locus on chromosome 9. HD of the CDKN2A gene could be detected in 21 (70%) of PXA cases, whereas 9 cases (30%) harbored either a hemizygous CDKN2A deletion or no copy number loss. In the high-grade IDH-wild-type glioma cohort, 7 tumors (22%) harbored a HD and 25 tumors showed no or a hemizygous CDKN2A gene loss (78%). HDs of the CDKN2A gene were identified in 10 (25%) of all ST-EP, whereas 30 tumors (75%) showed no alterations in the CDKN2A gene or a hemizygous loss (n=13). Three IDH-mutant astrocytomas (cases #108, #109, #110) showed a homozygous CDKN2A deletion by MIP analysis, and 4 cases showed a hemizygous CDKN2A deletion (Fig. 1). We evaluated the copy number status of CDKN2A in the meningioma cohort by MIP (n= 16) or MLPA (n=8). By MIP analysis, a homozygous CDKN2A deletion was found in 6 morphologically anaplastic meningioma (cases #132, #134, #138, #139, 140, and #144). One case presented with a hemizygous deletion (Fig. 1).

Immunohistochemistry for p16 Protein

The p16 positive cell rate was assessed by at least 2 neuropathologists determining the amount of p16 positive cells. Overall, the following items were included in the interpretation of the p16 immunohistochemistry: rate (in %) of positively stained cells, staining pattern (nuclear/cytoplasmic) (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/PAS/B665). When possible, positively labeled lymphocytes, microglial cells, and other reactive cells were excluded from the scoring. We used 5% positive-stained cells as a cutoff level. Cases with >5% positive cells were defined as immunohistochemically positive p16 cases. Examples of tumors with and without homozygous CDKN2A deletion are depicted in Figure 2.

FIGURE 2.

p16 immunohistochemistry: PXA case #30 (p16 positive, A) and case #22 (p16 negative, B). HGG case #52 (p16 positive, C) and case #39 (p16 negative, D). EP case #99 (p16 positive, E) and case #70 (p16 negative, F). A, IDH-mutant case #112 (p16 positive, G) and case #109 (p16 negative, H) (scale bars in A–H: 50 µm).

Among all 30 PXAs, 27 tumors showed negativity for p16 protein by immunohistochemistry, including 20 (74%) carrying a homozygous CDKN2A deletion (sensitivity, 74%; specificity, 67%). Case #7 showed p16 positivity (15% nuclear and cytoplasmic) in a fraction of cells, although genetic analysis confirmed a CDKN2A HD (Fig. 1, Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/PAS/B665). Case #23 displayed about 20% p16 positive-stained cells (nuclear and cytoplasmic) and a hemizygous CDKN2A gene deletion. Overall, immunohistochemistry for p16 in PXA showed a high positive predictive value (95%; PPV) and low negative predictive value (22%) for an underlying homozygous CDKN2A deletion (Fig. 3).

FIGURE 3.

Copy number plots of cases with and without homozygous CDKN2A loss. Copy number profile of case #30 (A) and the 9p21 region (C) showing no copy number losses of CDKN2A. B and D, Copy number profile of case #22 with copy number losses in the chromosomal region 9p21 (presented in detail in D). E, Copy number profile without CDKN2A HD of case #112. F, Copy number profile of case #109 demonstrating homozygous deletion of CDKN2A.

All high-grade IDH-wild-type glioma cases with a homozygous CDKN2A deletion (n=7) showed an unequivocal loss of p16 protein expression (sensitivity 44%, specificity 100%). Of the cases without deletion or with only a hemizygous CDKN2A deletion, 16 (70%) displayed a positive p16 protein immunostaining (100% PPV, negative predictive value, NPV, 64%).

In the ependymoma cohort, 9 cases with homozygous CDKN2A deletion displayed a p16 protein loss. One case with a CDKN2A homozygous loss showed a partly positive p16 immunohistochemistry, probably due to spatial genetic heterogeneity in this case. All ST-EP with hemizygous deletion or no CDKN2A copy number alteration but two cases presented with a detectable p16 protein expression in immunohistochemistry (sensitivity 82%; specificity 97%; 90% PPV; 93% NPV). Case #87, which had no CDKN2A deletion and only <5% positively stained tumor cells, was considered p16 negative and is an example of a borderline case, which under certain circumstances, should be subjected to further molecular investigation.

To obtain accurate results for the evaluation of the p16 immunohistochemical status in IDH-mutant astrocytoma, we compared the p16 positively stained cells with the spatial distribution of IDH1 (R132) positive tumor cells on serial sections. IDH-mutant high-grade astrocytoma without CDKN2A HD revealed detectable p16 protein in varying fractions of tumor cells (range > 5% to 90% of stained cells). Cases #108, #109, and #110—harboring a CDKN2A-HD—exhibited a loss of p16 protein expression in the majority of tumor cells (range 0% to 5% positive cells). One case with a partial hemizygous CDKN2A deletion also exhibited <5% positively stained cells in the p16 immunohistochemistry. Overall, we determined a 60% sensitivity, 100% specificity, 100% PPV, and 88% NPV in this group.

In our hands, there was a strong correlation between loss of p16 protein expression and homozygous CDKN2A deletion in anaplastic meningioma (PPV 100% and specificity 100%). Our cohort included 6 anaplastic meningioma with an HD of CDKN2A. The p16 expression in 5 of these cases was under 5% of tumor cells. Case #144 showed about 5% focal p16 positivity in small tumor cell swirl-like islets. The interpretation of the p16 immunohistochemistry in low-grade meningioma was not conclusive, as tumor cells without CDKN2A HD did not display p16 expression (NPV 72% and sensitivity 54%). Thus, the lack of p16 protein in low-grade meningioma cells should be regarded as a normal finding and not as a loss due to gene deletion (Fig. 4).

FIGURE 4.

Examples of p16 immunostaining in meningioma. A, Case #125 (fibroblastic meningioma CNS WHO grade 1, p16 negative, no CDKN2A HD). B, Case #132 (anaplastic meningioma, p16 negative, with CDKN2A HD). C, Case #135 (anaplastic meningioma, p16 positive in about 80% of cells, genetic status: no deletion). D, Case #141—anaplastic meningioma, p16 positive in about 50% of the cells, genetics: no CDKN2A deletion (scale bars in A–D: 50 µm).

DISCUSSION

There has been growing evidence for the importance of CDKN2A as a prognostic marker for an adverse clinical outcome in several tumors arising in the CNS, including ST-EP with ZFTA fusion, high-grade meningioma, anaplastic IDH-mutant astrocytoma, and oligodendroglioma.^10,12,14 The 2021 WHO classification of the CNS tumors introduces CDKN2A deletions in IDH-mutant astrocytoma as an independent criterion for grading these tumors as CNS WHO grade 4.¹ As for other brain tumors such as PXA, CDKN2A deletions serve as a helpful tool toward a precise diagnosis.^16,17

So far, no guidelines for the implementation of p16 immunohistochemistry in routine neuropathology exist. For some tumors outside the CNS, however, p16 immunohistochemistry has already been incorporated into tumor diagnostics. In this study, we explored the diagnostic value of p16 immunohistochemistry as an indicator of a homozygous CDKN2A deletion in brain tumors. Genetic CDKN2A alterations in 134 CNS tumors were accessed by quantitative, high-resolution copy number analysis (MIP). The CDKN2A status of additional 13 brain tumors was determined using MLPA or methylation profile (850 k) deduced copy number information.

Interpretation of p16 immunohistochemistry is intricate because normal brain tissue does not express p16 at a level that is immunohistochemically detectable. It must be upregulated to become immunohistochemically visible. Alterations in the p16-CDK4-pRb pathway, such as CDK4 amplification or lack of pRb, can lead to p16 upregulation with subsequent positive-resulting immunohistochemistry.

In one previous study comparing FISH for CDKN2A and p16 immunohistochemistry in glioblastomas, 95% of cases with homozygous CDKN2A deletion and 57% of cases with hemizygous deletion were negative for p16 by immunohistochemistry, respectively, whereas 30% of cases with normal CDKN2A copy number status were immunonegative for p16.²² In a more recent study, a comparison between CDKN2A HD by array technology and immunohistochemistry in IDH-mutant astrocytic gliomas failed to show a high correlation, although the precise discrepancy rates were not reported. The authors used a p16 antibody clone G175-405 and found that the basal nuclear p16 expression detected by this antibody was often too low for detecting loss of expression.¹⁰

In contrast, our data showed that the immunohistochemical loss of p16 protein expression detected with the E6H4 mouse monoclonal antibody serves as an effective screening tool for genetic CDKN2A deletions in CNS tumors. In our PXA cohort, only 1 case (case #7) that harbored a homozygous CDKN2A deletion showed about 15% positive tumor cells in the immunohistochemistry for p16 (PPV 95%). Among all high-grade IDH-wild-type gliomas that were included in our study, we found a high specificity (100%) and a high PPV (100%), suggesting that a negative p16 protein immunohistochemistry can be an indicator for a deletion of the CDKN2A gene. In the ependymoma cohort, immunohistochemical p16 loss served as a strong predictive marker for a homozygous CDKN2A deletion (PPV 90%, NPV 93%). Among the IDH-mutant astrocytomas, we observed a slightly inconsistent staining pattern that might be explained by the diffuse growth of these tumors. However, tumors with a molecularly confirmed homozygous CDKN2A deletion exhibited loss of p16 expression (cutoff value <5% of positive cells); we identified a high PPV (100%) for p16 immunohistochemistry. As homozygous CDKN2A deletions have an impact on astrocytoma grading and prognosis, the implication is that cases that show negativity for p16 should be subjected to further molecular testing. In our series, loss of p16 expression on immunohistochemistry correlated well with CDKN2A HDs particularly in higher-grade meningioma. These results are consistent with recent work from Tang and colleagues that screened 39 higher-grade meningioma for CDKN2A HDs or truncating mutations and concurrent immunohistochemical loss of p16 expression. The authors recommended p16 immunohistochemistry as an effective screening method for the CDKN2A gene status in higher-grade meningiomas.²⁶ The majority of low-grade meningeal tumors that were included in our series showed no significant expression for p16 protein, although there was no confirmed CDKN2A deletion in these cases. This might be a reflection of the fact that p16 is more often detected in high proliferative neoplasia. Hence, p16 is not a suitable screening tool for a CDKN2A deletion in lower-grade meningioma. These results are consistent with the observations made by Tang and colleagues.

Recent work identified and evaluated MTAP (methylthioadenosine phosphorylase) immunohistochemistry as an alternative surrogate marker for a CDKN2A homozygous loss in IDH-mutant gliomas.²⁷ MTAP, located closely to the CDKN2A locus, appears to be often co-deleted together with CDKN2A in glial tumors. It has been reported that 9% of cases with a CDKN2A deletion in malignant mesothelioma do not show evidence of MTAP codeletion in FISH.²⁸ Therefore, MTAP immunohistochemistry may miss CDKN2A-deleted cases. Satomi et al²⁷ showed that MTAP immunohistochemistry exhibited higher specificity compared with p16 immunohistochemistry with a G175-405 mouse monoclonal antibody in adult IDH-mutant astrocytoma. Nevertheless, the authors reported that the interpretation of MTAP immunohistochemistry was easy in only 61% of the successfully scored specimens, and 7% of the specimens were even uninterpretable. This, taken together with the lack of data about the exact frequency of simultaneous CDKN2A and MTAP codeletion in glial tumors questions the use of MTAP as a potential surrogate marker for homozygous CDKN2A deletion at this point of time.

Evaluation of p16 immunohistochemistry in the context of glial tumors can harbor potential pitfalls concerning the interpretation of the staining pattern. A frequent phenomenon that we observed was the intratumoral heterogeneity of the p16 immunoreaction. It can be hypothesized that tumor cell subpopulations are undergoing different regulatory stages involving the p16-CDK4-pRb pathway.

Assessment of the CDKN2A copy number status is gaining in importance. Our data support the opinion that a quantitative, high-resolution genomic copy number analysis (MIP) array represents a reliable method for identifying homozygous CDKN2A deletions. Nonetheless, immunohistochemical screening for p16 protein expression represents a useful, low-cost tool in the diagnostic workup of noninfiltrative gliomas, IDH-mutant gliomas, and high-grade meningiomas in particular as a rapid screening method before confirmatory genetic analysis.

Supplementary Material

pas-48-46-s001.docx ^{(20.3KB, docx)}