Evaluation of Immunohistochemical Markers, CK17 and SOX2, as Adjuncts to p53 for the Diagnosis of Differentiated Vulvar Intraepithelial Neoplasia (dVIN)

Histological diagnosis of differentiated vulvar intraepithelial neoplasia (dVIN), the precursor of human papillomavirus (HPV)-independent vulvar squamous cell carcinoma (VSCC), can be challenging, as features of dVIN may mimic those of non-dysplastic dermatoses. To aid the diagnosis, p53-immunohistochemistry (IHC) is commonly used, and mutant expression patterns are used to support a histological diagnosis of dVIN. However, a proportion of dVIN can show wild-type p53-expression, which is characteristic of non-dysplastic dermatoses. Furthermore, recent research has identified a novel precursor of HPV-independent VSCC—the p53-wild-type differentiated exophytic vulvar intraepithelial lesion (de-VIL). Currently, there are no established diagnostic IHC-markers for p53-wild-type dVIN or de-VIL. We evaluated IHC-markers, cytokeratin 17 (CK17), and SRY-box 2 (SOX2), as diagnostic adjuncts for dVIN. For this, IHC-expression of CK17, SOX2, and p53 was studied in dVIN (n = 56), de-VIL (n = 8), and non-dysplastic vulvar tissues (n = 46). For CK17 and SOX2, the percentage of cells showing expression, and the intensity and distribution of expression were recorded. We also performed next generation targeted sequencing (NGTS) on a subset of dVIN (n = 8) and de-VIL (n = 8). With p53-IHC, 74% of dVIN showed mutant patterns and 26% showed wild-type expression. Median percentage of cells expressing CK17 or SOX2 was significantly higher in dVIN (p53-mutant or p53-wild-type) and de-VIL than in non-dysplastic tissues (p < 0.01). Diffuse, moderate-to-strong, full epithelial expression of CK17 or SOX2 was highly specific for dVIN and de-VIL. With NGTS, TP53 mutations were detected in both dVIN and de-VIL. We infer that immunohistochemical markers CK17 and SOX2, when used along with p53, may help support the histological diagnosis of dVIN.

In this study, we evaluated immunohistochemcial markers CK17 and SOX2 as adjuncts to p53 for the diagnosis of dVIN. For this, we examined the expression of these markers in dVIN, de-VIL, high-grade squamous intraepithelial lesion (HSIL; precursor of HPV-related VSCC), and non-dysplastic vulvar tissue. In addition, to facilitate the identification of other potential diagnostic markers, we performed next generation targeted sequencing (NGTS) on a subset of dVIN and de-VIL.

Histology Review
A total of 137 lesions were included after histology review, comprising dVIN (n = 56), de-VIL (n = 8), HSIL (n = 27), and non-dysplastic vulvar tissue (n = 46). Non-dysplastic tissues comprised lichen sclerosus (n = 15), non-specific reactive changes/inflammation (n = 18), and histologically normal vulvar tissue (n = 13). Six of the histologically normal vulvar tissues were identified from resection specimens of VSCC, and 7 were residual tissues from vulvar surgeries conducted for benign pathologies in women with no history of VIN or VSCC. Clinico-pathological characteristics are presented in Table 1.

Immunohistochemistry (IHC)
Immunohistochemistry results are presented in Table 2, illustrated in Figures 1-4, and summarized below.

Immunohistochemistry (IHC)
Immunohistochemistry results are presented in Table 2, illustrated in Figures 1-4, and summarized below.
MIB1: Increased MIB1 expression was noted in 52% of dVIN, 88% of de-VIL, all HSILs, and 39% of non-dysplastic vulvar tissues. For dVIN, de-VIL, and non-dysplastic lesions, increased MIB1 expression was most frequently observed in the basal and/or parabasal layers, whereas, for HSIL, this was across the full epithelial thickness. Difference in MIB1 expression in dVIN and non-dysplastic vulvar was not significant (p = 0.08). CK17: For dVIN and de-VIL, the median percentage of cells expressing CK17 was significantly higher (p < 0.01) compared to non-dysplastic vulvar tissues (Figures 1-4). Increased CK17-expression was recorded for dVIN that showed mutant pattern or wildtype p53-expression, with no significant difference among these sub-groups, in comparison with non-dysplastic vulvar tissue.
Of dVIN, 80% showed diffuse, moderate-to-strong expression, either across full epithelial thickness or in the suprabasal layers. This pattern of expression was also noted in 88% of de-VIL and 63% of HSIL. In dVIN, de-VIL, and HSIL, expression of CK17 followed the distribution of dysplastic cells. Of non-dysplastic vulvar tissue, 80% showed either no expression, or patchy, weak, expression in the suprabasal layers.
SOX2: For dVIN and de-VIL, the median percentage of cells expressing SOX2 was significantly higher (p < 0.01) compared to non-dysplastic vulvar tissue (Figures 1-4). Increased SOX2-expression was recorded for dVIN that showed mutant pattern or wildtype p53-expression, with no significant difference among these sub-groups, in comparison with non-dysplastic vulvar tissue.
Of dVIN, 86% showed diffuse, moderate-to-strong expression, either across full epithelial thickness or in the basal and suprabasal layers. This pattern of expression was also noted in 88% of de-VIL and 88% of HSIL. In dVIN, de-VIL, and HSIL, the expression of SOX2 followed the distribution of dysplastic cells. Of non-dysplastic vulvar tissue, 81% showed either no expression, or scattered, weak, expression in the basal (predominantly) and suprabasal layers.

Next Generation Targeted Sequencing
In the subset of dVIN (n = 8) and de-VIL (n = 8) that was studied, a total of 106 pathogenic mutations in 44 genes were identified (Table S2). The median number of mutations detected in dVIN was 3 (range: 1-6), and in de-VIL was 5 (range: 1-31). Pathogenic mutations detected in at least 3 samples of dVIN or de-VIL are depicted in Figure 6. The most commonly detected mutations for both diagnoses are described below.

Discussion
Since pathological diagnoses inform treatment decisions, accurate distinction of dVIN from reactive/non-dysplastic lesions is crucial. We observed that the majority of dVIN show diffuse, moderate-to-strong expression of CK17 and SOX2, across the full epithelial thickness or in the suprabasal layers. In contrast, the majority of non-dysplastic lesions show either no expression, or patchy/scattered weak expression of CK17 and SOX2. We therefore infer that IHC with CK17 and SOX2 can help distinguish dVIN from non-dysplastic lesions.

Discussion
Since pathological diagnoses inform treatment decisions, accurate distinction of dVIN from reactive/non-dysplastic lesions is crucial. We observed that the majority of dVIN show diffuse, moderate-to-strong expression of CK17 and SOX2, across the full epithelial thickness or in the suprabasal layers. In contrast, the majority of non-dysplastic lesions show either no expression, or patchy/scattered weak expression of CK17 and SOX2. We therefore infer that IHC with CK17 and SOX2 can help distinguish dVIN from non-dysplastic lesions.
Recent studies have increasingly recognized that accurate diagnosis of dVIN may not be achieved on the basis of histological assessment alone [38][39][40][41]. The spectrum of histological features of dVIN can overlap with that of non-dysplastic lesions, and occasionally, with that of HPV-related HSILs as well [40,41]. Therefore, to improve diagnostic accuracy, an IHC panel comprising p16 and p53 is being recommended for lesions with a histological suspicion of dVIN [2,4,17,39]. Block-type p16-expression is a reliable surrogate of high-risk HPV-infection, and helps discriminate HSIL from dVIN [40,41]. In addition, mutant pat-terns of p53-expression, i.e., basal and/or parabasal over-expression, or complete absence of expression, can help distinguish dVIN from non-dysplastic lesions [42,43]. Furthermore, the mid-epithelial wild-type p53-expression, which is considered characteristic of HSIL (seen in 96% of HSIL in the current study), can also help discriminate HSIL from dVIN [44].
However, similarly to most IHC-markers, p53 is not without limitations. Increased p53expression may not only result from the mutated protein, but also from the accumulation of the wild-type protein in response to DNA-damage in inflammatory conditions [42,45]. Therefore, increased p53-expression may be observed in non-dysplastic lesions as well [45]. Moreover, a proportion of dVIN (26% in our series), and the recently recognized VSCC precursor, de-VIL, show wild-type p53-expression [2,10,23,24]. In addition to p53 and p16, the proliferation marker MIB1 is commonly used to aid the diagnosis of dVIN. However, increased MIB1 expression can be seen both in dVIN and in reactive, non-dysplastic lesions [46,47].
Although over the past years, a number of novel diagnostic IHC-markers have been studied for dVIN, none have so far been translated to the clinics [15]. This is largely because most of these markers were evaluated in small cohorts and/or in single studies [15].
For dVIN, CK17 is the only IHC-marker that has been evaluated in more than one study, which includes one of our previous studies [20,27]. CK17 is a basal/myo-epithelial cell-associated protein [48]. In normal skin (genital/non-genital), CK17 is expressed only in the appendages, and not in the squamous epithelial lining [33,49]. Increased CK17expression across full-epithelial thickness or in the suprabasal layers has been observed in SCC, and precursors of SCC, of the cervix [49], vulva [20,27], anus [33], larynx [34], and oral cavity [30].
SOX2 is another IHC-marker that has been widely studied in SCC of both genital and non-genital locations [28,29,32]. A recent meta-analysis identified SOX2 as a promising biomarker for tongue SCC, which is also an HPV-independent SCC [35]. SOX2 is a transcription factor postulated to promote cancer development and progression, and increased SOX2-expression has been observed in SCC, and precursors of SCC, of the cervix [50], vulva [28,29], esophagus [51], and oral cavity [52].
To evaluate CK17 and SOX2 as diagnostic adjuncts for dVIN, we systematically examined their expression in a carefully selected set of dVIN, de-VIL, HSIL, and nondysplastic vulvar tissues. Increased expression of CK17 was observed in dVIN showing mutant patterns or wild-type p53-expression, compared to non-dysplastic vulvar tissue, in line with previous reports [20,27]. CK17-expression was also increased in de-VIL and HSIL, compared to non-dysplastic tissue. However, similarly to Podoll et al. [27], we also observed that CK17-expression can be higher in dVIN than in HSIL. This could be due to the paradoxical maturation exhibited by the dysplastic cells of dVIN, as it has been reported that CK17-expression is most prominent in dysplastic cells with some degree of phenotypic maturation [53].
For SOX2, increased expression was observed in dVIN showing mutant patterns or wild-type p53-expression, compared to non-dysplastic vulvar tissue, in line with the previous report [29]. Increased SOX2-expression was also observed in de-VIL and HSIL, compared to non-dysplastic tissue. However, unlike CK17, SOX2-expression levels were fairly similar in dVIN and HSIL, as was also reported by Brustmann et al. [29].
For both CK17 and SOX2, we sought to identify lesion-specific expression patterns, as pattern-based IHC-interpretation can be more reproducible and easier to implement in research or practice, than manual counting of stained cells [30]. Diffuse, moderate-to-strong expression of CK17 or SOX2, across full epithelial thickness, or in the basal and suprabasal layers showed high sensitivity and specificity for the diagnosis of dVIN. Among p53, CK17, and SOX2, the highest sensitivity was obtained for SOX2 (86%), whereas, the highest specificity was obtained for p53 (83%) (Table S1). We therefore believe that CK17 and SOX2, when used in a panel along with p53, can complement the pathological assessment of dVIN. However, CK17 and SOX2, cannot be used to discriminate dVIN from HSIL. Moreover, for both CK17 and SOX2, increased expression was observed in a number of Pharmaceuticals 2021, 14, 324 9 of 14 non-dysplastic lesions, whereas, a number of dVIN showed minimal expression. Therefore, for discriminating dVIN from non-dysplastic lesions, CK17 and SOX2 should be interpreted in the context of histology and clinical information, and should not be used as the sole diagnostic criterion. Further studies may help determine whether non-dysplastic lesions that show CK17 and SOX2 expression harbor molecular perturbations associated with early pre-neoplastic changes that precede histological manifestation.
In this study, we also investigated the molecular profile of a subset of dVIN and de-VIL, with a view to facilitate the identification of other potential diagnostic markers. In dVIN, TP53 and CDKN2A mutations were detected most frequently, in line with previous reports [47,[54][55][56][57][58][59]. Interestingly, TP53 mutations were also detected in de-VIL, and dVIN that showed wild-type-expression on p53-IHC. Discordant results of TP53 mutations and p53-IHC have been previously reported for dVIN [22]. Akbari et al. also detected TP53 mutations of uncertain significance in some of the de-VIL in their cohort [24]. The higher frequency of TP53 mutations in our cohort could be explained by the greater sequencing depth of our NGTS panel, which covers all exonic sites of TP53. In de-VILs, we could detect PIK3CA mutations, which have been reported to characterize these lesions [23,24]. However, these were not present in all de-VILs. Other interesting findings from the NGTS were: (i) frequent detection of RNF43 mutations in both dVIN and de-VIL, which have been previously associated with endometrial carcinoma [60], and (ii) detection of mutations in KEAP1, CDH1, PIK3R1, and POLE exclusively in de-VIL. We also observed that a large number of the mutated genes in both dVIN and de-VIL could be mapped to the PI3K/AKT/mTOR pathway [61], indicating a potential involvement of this pathway in HPV-independent VSCC carcinogenesis. Nevertheless, to determine the clinical significance of these preliminary findings, further investigation in larger cohorts is necessary.
There are several limitations associated with the current study. As this is a retrospective, single-center study, a selection bias cannot be ruled out. Our findings need to be validated in prospective, multi-center cohorts. The IHC-markers were assessed by visual estimation, and manual assessment is always associated with a certain degree of subjectivity. Computerized image analyses may offer higher accuracy and help identify more reliable cut-offs to improve the interpretation of CK17 and SOX2 IHC. Furthermore, our NGTS panel comprised a limited number of amplicons, and the detections of mutations can be influenced by the platform used.

Materials and Methods
This retrospective, observational study follows the Standards for Reporting Diagnostic Accuracy (STARD) guidelines [62]; the checklist is provided in Table S3.

Histology Review
Vulvar lesions from 2014-2017 were identified from the electronic records of the Department of Pathology, Erasmus MC, and their hematoxylin-eosin (HE) stained glass slides were retrieved from the archives. These slides were reviewed by two pathologists (SDG and PCEG), and the lesions were classified as VSCC, dVIN, HSIL, de-VIL, or non-dysplastic vulvar tissue, using diagnostic criteria from literature [3,9,[11][12][13]20,23,24,46,47,63]. Clinical data (age and treatment information) were gathered from the patient records. All patient data were anonymized and patient materials were handled following the guidelines of World Medical Association Declaration of Helsinki.

Immunohistochemistry (IHC)
For IHC, a set of dVIN, HSIL, de-VIL, and non-dysplastic vulvar cases having sufficient tissue to prepare serial sections was selected. To ensure a comprehensive study of the IHC-expression, the selection was prepared in a way to include-(i) lesions on vulvar skin with or without adnexal structures; (ii) standalone lesions, and lesions in tissue adjacent to VSCC; and (iii) lesions representing a range of histological appearances for each diagnosis. Cases with a history of chemo/radiotherapy were excluded.
For each lesion, the areas most representative of the diagnosis were marked on the HE-slides, so that identical areas could be studied on the IHC-slides. IHC was scored by SDG and PCEG by visual estimation, as described below.
MIB1: increased expression = increased nuclear expression in the basal and/or parabasal layers; not increased expression = sporadic nuclear expression in basal (minor component) layers and/or parabasal layers (major component) [30,39].
CK17 and SOX2: The following parameters were recorded-(i) percentage of cells showing uniform cytoplasmic (CK17) or nuclear (SOX2) expression, irrespective of the intensity (ii) intensity of expression (weak/moderate/strong), and (iii) distribution of expression within the epithelium. In case of variation in the intensity and/or distribution, the predominant pattern was scored.

Next Generation Targeted Sequencing (NGTS)
NGTS was performed on a subset of dVIN and all de-VIL. The methodology is detailed in Table S5. In brief, areas having minimum 50% lesional cells were selected, and microdissected manually from hematoxylin-stained slides into 5% Chelex 100 resin (BioRad Laboratories, Hercules, CA, USA) cell lysis solution (Promega, Madison, WI, USA). Tissue fragments were subjected to proteinase K digestion for 16 h at 56 • C. After inactivating proteinase K, and removing cell debris and Chelex resin by centrifugation, the extracted DNA was used without further purification. DNA-concentration was measured with Qubit 2.0 fluorometer (Thermo Fisher Scientific, Waltham, MA, USA).
For NGTS, a custom-made panel, designed using the AmpliSeq designer (Thermo Fisher Scientific, Waltham, MA, USA) was used. This panel comprises 1042 amplicons covering hotspot regions in 65 cancer-related genes, and single nucleotide polymorphisms (listed in Table S5). NGTS was performed using the Ion Torrent platform (Thermo Fisher Scientific) following manufacturer's protocols. Ion AmpliSeq Library Kitplus-384 LV was used for library preparation.
Sequence information was analyzed with Variant Caller v.5.10.0.18 (Thermo Fisher Scientifc), and variants were annotated in a local Galaxy pipeline using ANNOVAR (version 2014-11-12, Wang Genomics Lab, University of Southern California, Los Angeles, CA, USA). For reporting, variants existing in Exome Variant Server NHLBI GO Exome sequencing project (ESP), the 100,000 Genomes Project, Genome Aggregation Database (GnomAD), or Genome of the NL (GoNL) databases were filtered out, and those having a variant allele frequency > 10%, and located in the exons or splice sites were included (with the exception of TERT-promoter mutations). Pathogenicity of coding non-synonymous variants were predicted using Protein Variation Effect Analyzer (PROVEAN), Sorting Tolerant From Intolerant (SIFT), and UMD-Predictor algorithms, and the Catalogue of Somatic Mutations in Cancer (COSMIC).

Ethics Statement
This study follows the guidelines of the Dutch Federation of Biomedical Scientific Societies (www.federa.org/codes-conduct, accessed on 15 March 2021), which state that no separate ethical approval is required for the use of anonymized residual tissue procured during regular treatment.

Statistical Analyses
Data were analyzed using SPSS Statistics (Version 25.0: IBM Corp., Armonk, NY, USA). Independent sample's t-test was used to compute two-sided p-values and identify statistically significant differences. p-value < 0.05 was considered statistically significant. Receiver operating characteristic (ROC) curves were plotted for p53, CK17, and SOX2, and areas under the curve (AUC) with 95% confidence intervals (CI) were measured, to assess their individual sensitivity and specificity for the diagnosis of dVIN. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of p53, CK17, and SOX2 for the diagnosis of dVIN were computed, using the histological diagnoses as gold-standard. Expression levels of CK17 and SOX2 were visualized by constructing bar-charts and box-plots. NGTS data were analyzed descriptively.

Conclusions
CK17 and SOX2 show significantly higher expression in dVIN that show mutant pattern or wild-type p53-expression, compared to non-dysplastic vulvar tissues. Particularly for lesions where dVIN is suspected on histology, and p53-expression is wild-type, diffuse, strong expression of CK17 and SOX2 can help confirm the diagnosis of dVIN. Therefore, we infer that immunohistochemical markers CK17 and SOX2 can be useful adjuncts to p53 for the diagnosis of dVIN.

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki, and in accordance with the guidelines of the Dutch Federation of Biomedical Scientific Societies (www.federa.org/codes-conduct, accessed on 15 March 2021), which state that no separate ethical approval is required for the use of anonymized residual tissue procured during regular treatment. This is also part of the regular treatment agreement with the patients at Erasmus MC.

Informed Consent Statement: Not applicable.
Data Availability Statement: All relevant data has been provided in the Supplementary Materials.