Prognostic importance of mitosis quantification and PHH3 expression in oral epithelial dysplasia

Oral epithelial dysplasia (OED) is diagnosed and graded using a range of histological features, making grading subjective and challenging. Mitotic counting and phosphohistone-H3 (PHH3) staining have been used for the prognostication of various malignancies; however, their importance in OED remains unexplored. This study conducts a quantitative analysis of mitotic activity in OED using both haematoxylin and eosin (H&E)-stained slides and immunohistochemical (IHC) staining for PHH3. Specifically, the diagnostic and prognostic importance of mitotic number, mitotic type and intra-epithelial location is evaluated. Whole slide images (WSI) of OED (n = 60) and non-dysplastic tissue (n = 8) were prepared for analysis. Five-year follow-up data was collected. The total number of mitosis (TNOM), mitosis type and intra-epithelial location was manually evaluated on H&E images and a digital mitotic count performed on PHH3-stained WSI. Statistical associations between these features and OED grade, malignant transformation and OED recurrence were determined. Mitosis count increased with grade severity (H&E: p < 0.005; IHC: p < 0.05), and grade-based differences were seen for mitosis type and location (p < 0.05). The ratio of normal-to-abnormal mitoses was higher in OED (1.61) than control (1.25) and reduced with grade severity. TNOM, type and location were better predictors when combined with histological grading, with the most prognostic models demonstrating an AUROC of 0.81 for transformation and 0.78 for recurrence, exceeding conventional grading. Mitosis quantification and PHH3 staining can be an adjunct to conventional H&E assessment and grading for the prediction of OED prognosis. Validation on larger multicentre cohorts is needed to establish these findings. Supplementary Information The online version contains supplementary material available at 10.1007/s00428-023-03668-6.


Introduction
Oral epithelial dysplasia (OED) describes a spectrum of histologically identified architectural and cytological disturbances involving the oral epithelium [1].These lesions may progress to oral squamous cell carcinoma (OSCC) [2].Higher grade lesions have higher risk of transformation, highlighting the need for an early and accurate diagnosis [1].OSCC is the most common malignant neoplasm of the oral cavity associated with a myriad of environmental aetiologies and genetic alterations [3][4][5].
Because of the direct relationship between OED and malignant transformation, the dysplasia grade is considered the most important prognosticator for malignant transformation [5].However, the current grading system (WHO, 2017) is associated with poor reproducibility, which can result in an inconsistent and unreliable diagnosis [6].Suggestions to mitigate these shortcomings include the use of clinical determinants and molecular markers [7].The binary grading system is an alternative criteria proposed to improve observer reproducibility by quantifying the minimum number of cytological and architectural features required for a diagnosis [8].However, this classification uses the same histological features listed in the WHO Classification, and there remains a lack of high-quality evidence to support the prognostic importance of many of these features [2].The recent update from the 5th Edition of the WHO Classification includes additional features, such as apoptotic mitoses and single cell keratinisation.However, the clinical relevance for inclusion of these features is unclear [9].A recent study explored histological feature-specific associations in OED with clinical outcomes.The predictive performance of the proposed models for OED progression exceeded conventional grading [10].However, a more detailed and prospective analysis of individual histological features is still needed to establish a more objective predictive/ grading system.
Mitotic figure counting is used for diagnosis and prognostication of various malignancies [11][12][13][14] including breast, gastric and neuroendocrine carcinomas [13,[15][16][17].However, its importance in precancer diagnosis and progression is yet to be explored.The main limitation of mitosis counting is the tediousness of the manual approach, in addition to interpretation differences due to variations in chromatin arrangements in the different mitotic stages, and the resemblance of apoptotic bodies and pyknotic nuclei with mitotic bodies (Fig. 1) [18].Many of these limitations can now be overcome by the increasing number of digital/computational tools which allow for automated quantification, providing more objective, efficient and reliable outputs [19].However, in the case of mitotic cell counting, attention also needs to be given to the presence of abnormal mitotic forms, characterised by mitotic asymmetry or an abnormal segregation of chromosomes [20].
Various biomarkers have been implicated in OED progression, but the evidence to support their routine use is still lacking [21].Phosphohistone-H3 (PHH3) is a specific protein phosphorylated during chromatin condensation in mitosis [22].It stains positively during the late G2 phase and M phase.Phosphorylation of the histone H3 starts to occur just before prophase which is not identifiable on haematoxylin and eosin (H&E) examination [18], lending to the role of PHH3 a useful marker.
The aims of this study were threefold: first, to conduct a quantitative analysis of mitotic activity in OED (including number, type and intra-epithelial location of mitoses) using digitised H&E sections and immunohistochemical (IHC)stained tissue with PHH3; second, to evaluate changes in mitotic activity relative to OED progression; and third, to develop and explore multivariable models using mitotic features for prediction of OED recurrence and malignant transformation, with comparison to conventional grading.

Case selection and tissue processing
Following ethical approval (reference 18/WM/0335), a retrospective sample of 68 H&E-stained tissue sections were retrieved from the department archive.The sample comprised OED sections (n = 60) of varying grades (mild, moderate, severe) with 5-year post-diagnosis data, in addition to non-dysplastic control samples (n = 8) which included cases of benign hyperplasia, scar tissue and inflammatory oral lichen planus.Verrucous and HPV-related OED lesions were excluded based on morphological features, as they are distinct entities with reportedly different behaviours.
Prior to the inclusion, cases were independently reviewed by a consultant oral and maxillofacial pathologist (SAK) to ensure there was sufficient epithelial tissue for analysis.Cases with insufficient tissue, gross artefact or tangentially cut sections were excluded.All cases were then blindly reevaluated by SAK, HM (clinician with extensive expertise and specialist interest in OED analysis) and PH (trainee oral and maxillofacial pathologist) to confirm the original diagnosis and where necessary assign an updated OED grade (using WHO and binary systems).Grading variability was measured by a Cohen's kappa score, which resulted in a value of 0.900, demonstrating good interobserver agreement.
New 5-μm-thick formalin-fixed paraffin-embedded sections of the selected cases were obtained for H&E and IHC staining.The sections were scanned at 40 × magnification using an Aperio-CS2 scanner (Leica Biosystems, Milton Keynes, UK) to obtain high-resolution whole slide images (WSI) producing 68 H&E slides and 67 IHC slides for analysis.The IHC sample had one less case due to technical scanning/imaging difficulties, resulting in its exclusion at the final stage.
Clinical data collection included patient age at diagnosis, sex, biopsy site, original histological grade (WHO, 2017), status of malignant transformation and recurrence (lesion that progressed to OSCC or recurred at the same clinical site following treatment within 5 years).

Immunohistochemical staining for PHH3
IHC staining was carried out for the mitosis marker PHH3 (Ser10) using a previously described protocol [23].A primary rabbit anti-human PHH3 polyclonal antibody (#9701; Cell Signalling Technology, 1:100 dilution) and a secondary goat anti-rabbit antibody was used.Following IHC, counterstaining with haematoxylin and mounting in DPX was done for further analysis.

Analysis of mitosis activity in OED
QuPath software (v.0.3.2) was used for identification of regions of interest (ROI) and subsequent mitotic feature analysis [24].For all slides, five rectangular-shaped ROIs of a consistent size (area≈165,000 mm 2 ) corresponding to representative dysplastic and non-dysplastic regions were selected at 20 × magnification and verified by two experienced clinicians (HM, SAK).
For the H&E sample (n = 68), two observers (HS, SAK), blinded to clinical outcomes, were asked to independently count and record (i) the total number of mitoses (TNOM), (ii) the number of 'normal' and 'abnormal' mitoses and (iii) the intra-epithelial mitosis location ('basal' or 'suprabasal') in each field.An agreement between the observers was made on how to qualify a 'normal' and 'abnormal' mitosis.An equational bipartition of the chromosomal material was used as standard for 'normal' mitosis [25], whereas the presence of abnormalities like binucleation, pyknotic nuclei, micronuclei and broken-egg appearances qualified the mitoses to be 'abnormal' [26].A kappa score of 0.646 was obtained between the two observers for independent mitosis counting.In cases of wide disagreement, a consensus score was agreed/used for the downstream analyses.The means and standard deviation for the mitosis variables (TNOM, type and location) from the five ROIs were recorded and an average obtained for each case.
For the PHH3-IHC sample (n = 67), QuPath's inbuilt 'positive cell detection' algorithm was applied for automated quantification of positively stained mitoses, and intra-epithelial mitosis location recorded through manual assessment (by HS, SAK).Due to the nature of the automated detection, the mitosis type could not be confirmed in the IHC sample.All data were exported onto a pre-structured spreadsheet in Microsoft Excel® (v.2206).

Statistical analyses
Statistical analyses were conducted in GraphPad Prism (v9) and IBM SPSS Statistics (v29.0.1.0).Data was tested for normality following which appropriate statistical tests were selected.Unpaired Student's t-tests and one-way ANOVA were performed to compare differences in the TNOM, mitosis type and intra-epithelial location between OED grades and relative to control.Where relevant, an appropriate post hoc analysis (Tukey's/Dunnett's) was performed for pairwise comparisons.For the H&E analysis, the mean mitosis number and ratio of normal-to-abnormal mitoses were measured and compared between grades.Paired sample t-tests were conducted to compare the number of normal and abnormal mitoses across OED grades.
Multivariable logistic regression models were explored separately for H&E and PHH3-IHC samples, to assess statistical relationships between individual and combined mitotic variables (TNOM, mitosis type, intra-epithelial location) with clinical outcomes (malignant transformation and OED recurrence).The effect of adding clinical variables (age, sex, intraoral site) and histological grade (WHO, binary) on model performance was assessed.The area under the receiver operator characteristic (ROC) curve was used to assess model accuracy and visualise performance.A p value of < 0.05 was considered statistically significant.Figure 2 depicts the workflow methodology for this study.

Normal mitotic figures
There was a significant difference in the average number of 'normal' mitoses between WHO grades (p = 0.0016) and binary grades (p = 0.0040) (Table 1).Significant differences were also seen between the following groups: control vs severe OED (p = 0.0004), control vs high-grade OED (p = 0.0023), mild OED vs severe OED (p = 0.0026) and moderate OED vs severe OED (p = 0.0143) (Table 1).

Normal-to-abnormal mitosis ratio
The ratio of normal-to-abnormal mitoses was higher in OED (1.61) compared to control (1.25).This ratio was found to reduce with increasing grade severity.The ratios for mild, moderate and severe grades were 3.26, 1.49 and 1.43, and for low and high grades, 2.75 and 1.44, respectively.Statistically significant differences were observed when comparing the ratio of normal/abnormal mitoses across different grades (p = 0.0001 mild OED, p = 0.0289 moderate OED, p = 0.0470 severe OED, p < 0.0001 low-grade OED, p = 0.0137 high-grade OED).

Multivariable model development exploration
The association between mitosis variables, clinical characteristics, histological grades and clinical outcomes was assessed (for H&E and PHH3-IHC analysis) using multiple logistic regression.For comparative purposes, the prognostic strength of conventional grading systems (WHO and binary) was also evaluated (Tables 2 and 3).

Discussion
This study highlights the potential importance of mitosis assessment and quantification in OED diagnosis and prognostication.Mitosis counting has been effectively implemented in the diagnosis of various malignancies [13,17,[27][28][29], but its diagnostic importance in oral precancers remains largely unexplored.Due to the limitations of manual mitotic figure counting, PHH3 was explored to evaluate its role as a diagnostic and prognostic adjunct to conventional H&E assessment.
The role of various oncogenes in OED progression to cancer still remains unvalidated [30].Ki-67 being a cell cycle marker, rather than a specific marker of mitosis, has shown conflicting results.In one study, the value of PHH3 and Ki-67 for measuring mitotic activity in OSCC demonstrated a significant association between expression of PHH3 (p = 0.016) and mitotic activity (p = 0.031) with survival time; however, no similar relationship was found with Ki-67 (p = 0.295) [31].In another study, the presence, location and pattern of Ki-67 positivity demonstrated variable results for differentiation between normal tissue, OED and OSCC [32].The unreliability of Ki-67 [32,33] and the successful use of PHH3 as an independent biomarker in various different malignancies [13,15,17,22,34] led us to explore this marker further.
The TNOM was shown to increase proportionally with grade severity on both H&E and PHH3-IHC analyses, supporting findings in the existing literature [35-38.This could be explained by the increased stem cell turnover and quantity of abnormal mutations [39].Overall, PHH3 mitotic count was greater than H&E, likely due to the inclusion of early prophase stage, which cannot be reliably distinguished on H&E-stained sections.In a previous study, a comparison in mitotic count between H&E and crystal violet-stained sections demonstrated significant differences between non-dysplastic oral mucosa, OED and OSCC [39].Whilst our findings revealed a greater difference between mild and severe OED, control and high-grade/severe OED, promising differences were also observed between the more 'demanding' groups (moderate vs severe OED) in terms of mitosis number, mitosis type and mitosis location.
H&E analysis of mitosis type demonstrated a higher ratio of normal-to-abnormal mitoses in OED than control, which decreased with grade severity.Mitosis location assessment on H&E and IHC analysis demonstrated significant differences in the number of 'basal' and 'suprabasal' mitoses between grades.'Suprabasal' mitoses were shown to be more predictive than 'basal' mitoses on PHH3-IHC.A study on meningioma demonstrated that PHH3 mitotic counts had a better interobserver correlation than H&E mitotic counts (R m = 0.83 vs 0.77, respectively) [40], with good discrimination between grades (AUROC 0.91).Our study suggested similar findings, with better generally performance for PHH3-IHC models than H&E models, particularly for TNOM and mitosis location (Table 2).This is likely to be related to greater objectivity of mitosis assessment with PHH3 staining.
Prognostic models combining TNOM, mitosis type, location and histological grading showed better prediction for transformation and recurrence.Generally, the addition of clinical variables had minimal impact on model performances, whereas histological grading boosted predictive potential.Such a trend was also observed in a study by Mahmood et al.where inclusion of grades improved prognostic strength of histological OED models [10].
The most predictive H&E models for malignant transformation ('abnormal mitoses' + 'suprabasal mitoses' + 'TNOM' + 'WHO grade' = AUROC 0.8113) and OED recurrence ('abnormal mitosis' + 'basal mitoses' + 'TNOM' + 'WHO grade' = AUROC 0.7895) (AUROC 0.65) incorporated multiple mitotic features and outperformed conventional WHO grading on its own.In the case of PHH3-IHC models, the most superior models utilised fewer mitotic features for prediction of transformation ('basal mitoses' + 'binary grading' = AUROC 0.7714) and recurrence ('TNOM' + 'WHO grading' = AUROC 0.7783).These findings indicate that PHH3-IHC may be important for prognostication of OED, complementing H&E analysis.The authors acknowledge a few limitations.First, the follow-up period comprised 5 years.Whilst transformation may occur later [41], a number of studies have shown transformation incidence to be highest during the first 5 years.[5,[41][42][43][44] A study by Hankinson et al. (2021) reported a median transformation time of 22 months (IQR 46.0) for a cohort of OED cases (n = 150) retrieved from the same centre as that used for this study [45].Second, cases were from a single-centre, and the sample size could be regarded as small [46,47].However, the unit in question is a national tertiary centre providing service to a large geographical region, thereby increasing the biological diversity of the sample.Furthermore, the sample has an equitable distribution of dysplasia grades with inclusion of transformed and non-transformed cases.For an early exploratory study that serves as a basis for future work, our sample is similar to many other studies [31,48,49] of this kind.The control cases were included for clinical interest and early comparative analysis, hence the small numbers.They did not contribute to the prognostic work, which was the important and novel aspect of this study.
In conclusion, we report increased mitotic activity with OED progression.Mitotic quantification using PHH3-IHC is potentially more reliable than H&E analysis, with typically greater predictive strength, even with inclusion of fewer variables.The addition of histological grading further improved performance of PHH3-IHC models, more so than the H&E models.To the best of our knowledge, this is one of the first studies to utilise mitosis quantification and compare H&E with PHH3-IHC for OED analysis and prognosis prediction.The promising results call for further exploration of H&E and IHC markers to contribute to a more objective grading of OED and reliable prognosis prediction.Further studies with larger multicentre cohorts are required for clinical validation.

Fig. 2
Fig. 2 Overall workflow methodology of the study.A Identification, retrieval and preparation of H&E sample (n = 68).B Preparation of PHH3-IHC sample (n = 67).Conversion of tissue sections to digital WSI and identification of ROI for H&E (C) and PHH3-IHC analysis (D).E Manual assessment of mitosis activity (number, type, location) on H&E.F Automated mitosis quantification for PHH3-IHC sample.G Statistical analysis to assess mitotic activity in OED with correlation to clinical outcomes

Table 2
Exploration of multivariate prognostic models based on the TNOM, mitosis location, clinical variables and histological grading systems (H&E n

Table 3
Exploration of multivariate prognostic models based on the type of mitoses, clinical variables and histological grading systems on H&E assessment (n = 68 − 5 ROI per WSI) The first two rows indicate the prognostic values for existing grading systems for comparative purposes.Highlighted rows indicate the top most predictive models overall.Asterisk indicates a statistically significant finding.AUROC area under receiver operating characteristic Text in bold indicate the most significant values/models