Comparative Evaluation of P16, Ki-67, and Shape-correlated Genes as Biomarkers for Cervical High-grade Squamous Intraepithelial Lesions


 Background:Colposcopy was referred in cases with severe abnormalities in co-testing. However, approximately 60%-80% patients did not receive proper benefits. Although p16/Ki67 dual staining reduced the referral rate, its sensitivity and specificity need to be enhanced. Methods:The expression of p16, Ki-67, SMAD3, YAP1, RELA were evaluated inthe colposcopy referral population. Results:The sensitivity and specificity of p16+ combined with Ki-67+ for predicting CIN2+ were 62.1% and 89.5%, respectively. p16+ combined with YAP1+ and/or RELA+ provided a sensitivity and specificity of 70.9% and 89.5%, respectively, while 72.8% and 86.4% were achieved by p16+ combined with YAP1+ and/or SMAD3+ and/or RELA+. In HPV16/18+ and LSIL subgroups, the sensitivity and specificity of p16+ combined with Ki-67+ for predicting CIN2+ were 67.7% and 87.6%, respectively, for the former group and 58.6%, 88.8%, respectively, for the latter group, whereas those of RELA alone were 76.9% and 90.5%, respectively, and 82.8% and 92.1%, respectively. p16+, YAP1+/RELA+ showed a better performance for predicting CIN2+ with a better sensitivity (86.7% vs. 52.2%, p=0.028) and considerable specificity (86.7% vs. 90.0%, p=0.486) in the other HPV+ combined with ASCUS groupthan were achieved by p16+ combined with Ki-67+. RELA+ and the combination of p16 and RELA/YAP1 also provided the Max AUC area. Conclusions:Our study shows that RELA and the combination of p16 and RELA/YAP1 achieved better sensitivity and specificity for detecting morphologically CIN2+ lesions.

intraepithelial neoplasia [CIN1]) or high-grade squamous intraepithelial lesions (HSIL, cervical intraepithelial neoplasia [CIN2+]). However, many studies found that although co-testing effectively improved the effectiveness of cervical cancer screening, 60%-80% of patients who were referred to invasive colposcopy with HSIL did not receive proper bene ts. (6,7) Therefore, better triage strategies for colposcopy should be investigated to improve the accuracy of diagnoses and reduce referral rate of colposcopy.
In 2012, the Lower Anogenital Squamous Terminology (LAST) Standardization Project recommended using immunohistochemical staining (IHC) for p16 to improve consistency in the diagnosis of HSIL (CIN2+). (8) Recently, dual-stain cytology for p16 and Ki-67 has been proposed as a biomarker for colposcopy triage and to have a sensitivity ranging from 68-94.2% and a speci city ranging from 30.6-95.2% in different subgroups. (9)(10)(11)(12) A study based on the VUSA Screen compared genotyping results with the results of p16/Ki-67 dual-staining and reported that the sensitivity of p16/Ki-67 dual-stained cytology for the detection of CIN2 + tended to be higher than that of HPV16/18 genotyping (68.8% vs. 43.8%). However, the speci city for detecting CIN2 + was signi cantly lower for p16/Ki-67 dual-staining than for HPV16/18 genotyping (72.8% vs. 79.4%). (13) Another multi-center study analyzed 1357 patients and showed that sensitivity (92.7% vs. 94.5%) and speci city (52.7% vs. 53.5%) were higher for p16/Ki67 double staining than for cytology in the HPV-positive group. Moreover, although p16/Ki-67 showed better speci city than the HR-HPV test (66.4% vs. 55.8%), the sensitivity of p16/Ki-67 was worse than that of the HR-HPV test (87.5% vs. 91.7%) in the ASCUS/LSIL group. (14) Therefore, more disease-speci c molecular markers of cervical high-grade squamous intraepithelial lesions should be investigated to provide an immunocytochemical staining strategy with high sensitivity and high speci city for detecting cervical precancer. According to previous studies, the SMAD3, YAP1 and RELA were showed an overexpression in cervical cancer. (15)(16)(17) Here, we evaluated the correlations among the expression of p16, Ki-67, the shape-correlated genes SMAD3, YAP1, RELA and CIN2 + lesions using IHC in the colposcopy referral population, so that to explore more sensitive and speci c markers for further cytological staining.

Study population
A search was performed in the Peking Union Medical College Hospital and Beijing Chao-Yang Hospital database for data entered from January 1, 2015 to January 1, 2017 related to screening women who were 30-60 years and diagnosed with HPV16/18-positive or other HR-HPV-positive with ASCUS or LSIL or AGC (atypical glandular cell) in HPV genotyping and liquid-based cytology tests. Subsequent colposcopies, endocervical curettages and cervical biopsies were also collected. Cases were excluded if there were no biopsies, if the interval between cervical screening test and biopsies was more than six months, or if insu cient tissue was available as a formalin-xed para n-embedded block.

Immunohistochemistry
All formalin-xed para n-embedded blocks were obtained from the Department of Pathology, Peking Union Medical College Hospital and Beijing Chao-Yang Hospital. Hematoxylin and eosin staining (H&E) had been previously performed on all cases at the time of original diagnosis. p16, Ki-67, SMAD3, YAP1 and RELA immunohistochemistry was performed in all included cases. Antibodies were purchased from Abcam (Cambridge, USA). The concentration of each primary antibody was set as follows: p16 (1:1000), Ki-67 (1:1000), SMAD3 (1:2000), YAP1 (1:1000) and RELA (1:2000). In addition, immunohistochemistry was conducted according to the manufacturer's protocol.
p16-positive staining was de ned as continuously strong, diffuse, bank-like nuclear or nuclear plus cytoplasmic staining involving at least one third of the squamous epithelium thickness. Negative staining was de ned as all other staining patterns, such as cytoplasmic only, wispy, blob-like, puddled, scattered, and single cells, according to the LAST guidelines.(8) Positive staining for Ki-67, SMAD3, YAP1 and RELA was de ned as follows: ve random microscope elds were selected to evaluate semi-quanti cation staining. The intensity of immunostaining was graded as 1+, weak; 2+, moderate; or 3+, strong. The area of positive squamous epithelium cells in each microscopic eld was categorized as 1 + for 0 to 33%; 2 + for 33 to 66%; or 3 + for 66 to 100% coverage. A sum between 5 and 45 was obtained by multiplying the 2 scores by 5. A sum from 0 to 17 was de ned as "low expression (-)", while a sum from 18 to 45 was de ned as "high expression (+)". Examples of observed staining characteristics of different expression are represented in Fig. 1. No speci c quantitative criteria were de ned for the number of positive cells. Two pathologists independently reviewed the H&E and IHC slides using their individual criterion, and the consensus interpretations were used as the nal interpretations. Discrepant interpretations were adjudicated by a third pathologist.

Follow-up
All cases which showed NILM (No intraepithelial lesion or malignant lesion) and CIN1 in cervical biopsies were informed to re-test by HPV genotyping and liquid-based cytology 12 months later, which lasting 3 years. The colposcopy and cervical biopsies were referred when the results of re-tests were in line with screening guidelines.

Statistical Analysis
The associations between p16, Ki-67, SMAD3, YAP1 and RELA positivity and subsequent CIN2 + were analyzed using Chi-squared Tests. The sensitivity, speci city, Youden's index and Area Under Curve (AUC) of each biomarker as a predictor for CIN2 + was determined, and the different assays were compared using the McNemar test. A P-value of < 0.5 was considered signi cant.

Page 5/22
1273 patients with complete results for HPV genotyping and liquid-based cytology tests met the initial inclusion criteria. Of these 1273 patients, 79 were excluded for no available cervical biopsy results. Of the remaining 1194 patients, 106 underwent endocervical curettage, and 52 consisted of vaginal HPV and cytology tests followed by vaginal biopsy. 1194 cases were available for evaluation, including 534 diagnosed with no intraepithelial lesion or malignant lesion (NILM), 289 diagnosed with CIN1, 173 diagnosed with CIN2, 160 diagnosed with CIN3 and 38 diagnosed with cancer. Patient ages ranged from 30 to 60 years old, and the median age was 40 years old.

Discussion
In the past 30 years, as a result of the extensive development of cervical cancer screening, the incidence of cervical cancer has dropped by more than 50% in the United States, and the rate of mortality has also decreased.(18) However, in developing countries, the incidence of cervical cancer is approximately 4-6 times higher than that of developed countries.(1) There are still many limitations related to cervical cancer screening technologies and strategies. (19) The independent use of cytology or HPV tests showed poor sensitivity and speci city for the detection of underlying CIN2+, resulting in unnecessary and invasive diagnoses. (20) In 2015, the updated cervical cancer screening guidelines of the American Society of Colposcopy & Cervical Pathology (ASCCP) indicated that the optimal screening methods are co-testing for women aged 30-65 as this simultaneously increases the sensitivity and speci city of cervical cancer screening. (21) However, studies have shown that although co-testing effectively improved the effectiveness of cervical cancer screening, 60%-80% of patients except for cytology HSIL who underwent colposcopy did not receive bene ts. (22)(23)(24) In the present study, the rate of CIN2 + in the population who were triaged to colposcopy ranged from 18.1-40% according to HPV genotyping and liquid-based cytology ( Table 1, except for HSIL). Despite the increasing coverage of HPV vaccines, screening will remain necessary for decades to control cervical cancer. Therefore, it is imperative to explore effective triage strategies to reduce misdiagnoses and unnecessary invasive examinations.  (14) Taken together, these data indicate that although p16/Ki67 dual staining reduces the referral rate for colposcopy resulting from current screening guidelines, its sensitivity and speci city need to be enhanced. Therefore, more sensitive and speci c biomarkers should be investigated to improve the accuracy of cervical cancer screening and reduce unnecessary invasive examinations.
SMAD family member 3 (SMAD3), a signal transducer and transcriptional modulator that mediates multiple signaling pathways, is activated by transforming growth factor-beta (TGF-β) and thought to play a role in the regulation of carcinogenesis. Recent studies have shown that Twist induces epithelialmesenchymal transition (EMT) by regulating the TGF-β/Smad3 signaling pathway in cervical carcinogenesis and that the activation of TGFA/Smad3 signaling can induce the migration and invasion of cervical cancer cell lines, suggesting that it plays a vital role in cervical cancer metastasis. (16,17) Yesassociated protein-1 (YAP1) encodes a downstream nuclear effector of the Hippo signaling pathway, which is involved in development, growth, repair, and homeostasis. This gene is known to play a role in the development and progression of multiple cancers by acting as a transcriptional regulator of this signaling pathway and may therefore function as a potential target for cancer treatment. Xiao et al. showed that nuclear levels of YAP are higher in high-grade cervical lesions and that YAP can function as a predictive marker for cervical cancer. reported that p16 immunostaining, performed using the strongest staining as the cutpoint, was 86.7% sensitive and 82.8% speci c for detecting CIN2+. (30) A review of p16 immunohistochemistry in cervical lesions also showed that the positive rates of p16 ranged from 0-37.5%, 0-100%, and 61.5-100% in the negative, LSIL and HSIL groups, respectively; while the positive rates of Ki-67 ranged from 0-81.1%, 12.5-100%, and 42.3-100% in the negative, LSIL and HSIL groups, respectively.(31) It has been suggested that the different positivity rates observed across studies may be due to multiple factors, such as different staining conditions, clones, patient populations, and lesion grading thresholds. (32) In the present study, the correlations between the expression levels of p16, Ki-67, and three shape-correlated genes (SMAD3, YAP1 and RELA) and CIN2 + lesions were evaluated by IHC. The sensitivity and speci city of p16 were 82.5% and 70.5%, respectively, for detecting CIN2+, consistent with the literature. RELA and a combination consisting of p16, YAP1 and RELA had better sensitivity and speci city than was achieved by p16 alone for detecting CIN2 + in different subgroups. Notably, 45 cases that showed < CIN2 at the initial biopsy were diagnosed as CIN2 + during the follow-up period, and 37 of these 45 cases were positive for p16+, RELA+/YAP1 + at the initial IHC staining, which also provided a Max AUC area for predicting CIN2+. Hence, we suggest that a combination consisting of p16, RELA and YAP1 may represent a biomarker for underlying CIN2 + in cervical screening. Our aim is to investigate more diseasespeci c molecular markers for detecting CIN2 + on cervical biopsy to provide an accurate immunocytochemical staining strategy for detecting high-risk cervical precancers. Our team has also conducted a project to evaluate the immunocytochemical staining performance of correlative markers in screening populations. We expected to present better results in the future. Cytology HSIL showed a high sensitivity and speci city for detecting CIN2+, therefore we did not focus on this subgroup.

Conclusion:
our study shows that the rate of CIN2 + ranged from 18.1-40% in a population of patients who were triaged to colposcopy according to HPV genotyping and liquid-based cytology (except for cytology HSIL).
SMAD3 and a combination consisting of p16 and RELA/YAP1 IHC achieved superior sensitivity and speci city for detecting morphologically CIN2 + lesions in the colposcopy referral population. However, further studies will be needed to evaluate the immunocytochemical staining performance of correlative markers.

Consent for publication
All authors have given their consent for the publication of this article.
Availability of data and material Supplemental Table (S1, S2, S3, S4, S5) are provided as supplemental material and are available online with the paper.

Competing interests
The author reports no con icts of interest in this work.    The AUC for p16, Ki-67, SMAD3, YAP1 and RELA in different groups. A, the colposcopy referral population (except for HSIL