Blimp-1 is a prognostic indicator for progression of cervical intraepithelial neoplasia grade 2

This study was part of a larger multicentre trial, the PRINCess study, which was designed to evaluate observational management of CIN2 in women < 25 years (Sykes et al. 2016). The PRINCess study clinical trial (ANZCTR number ACTRN12611000547943), was approved on April 14, 2010, by the Multi-region Ethics Committee (Ethics reference: MEC/09/07/079) and had site-specific local authorization.

Subjects were participants in the PRINCess study, which ran over the course of 24 months. Inclusion criteria for the subset of women in this study were biopsy-confirmed and p16-positive CIN2 and a high-risk HPV-positive DNA test on a cytological sample at baseline. Women were followed up with repeat cytology and biopsies at 6-monthly intervals and, based on clinical outcome during the 2-year follow-up term, were divided into ‘progressors’ to histologically confirmed CIN3 + (34 cases)’ or ‘persisters/regressors’ (histologically confirmed persistent CIN2 or regression to a lower grade or normal; 35 cases). Confirmed regression was defined as colposcopy, histology (where available), and cytology findings of CIN1 or less on two consecutive colposcopy visits at least 6 months apart. Regression was not confirmed if a low-grade colposcopic lesion had been visualized but not biopsied. Women were observationally managed but with a clinical endpoint of progression to CIN3 + , at which point they were withdrawn from the observational management and treated.

The CIN2 tissues stained and analysed in this study were from the baseline biopsy sample. Formalin-fixed paraffin-embedded (FFPE) serial sections were used. The first and last of the serial sections of each series was confirmed by a second Pathologist, in addition to the original pathological assessment, to contain CIN2.

Four-micrometer‐thick sections were incubated overnight at 37 °C. Slides were deparaffinized in xylene and rehydrated in graded ethanol. After rehydration in distilled water, antigen retrieval was performed by microwaving either in 50 mM tris buffer (pH 9.5) containing 5% urea or in 10 mM sodium citrate buffer (pH 6.0) twice for 10 min at 900 W. Slides were cooled at RT for 20 min before being washed in tris-buffered saline, pH 7.4 (TBS). Non-specific antigen binding sites were blocked with 100 μl of 5% skim milk in TBS for 20 min. Sections were incubated with primary antibodies (Table S1) in a humidified chamber at RT for 90 min. After the three washes in TBS, sections were incubated with the corresponding cross-absorbed secondary antibody (Table S1) at RT for 60 min. For the final 30 min of secondary antibody incubation, 40 μl of a 1:50 dilution of DAPI (Invitrogen) was added to stain cell nuclei. Sections were washed as described above, then incubated in the dark with 0.1% Sudan Black B in 70% ethanol for 5 min, to block autofluorescence. Sections were washed in a stream of TBS, immersed in TBS for 10 min, then mounted with 5 µl of SlowFade™ Diamond (Cat. No. S36972; Invitrogen) and stored at 4 °C prior to imaging. When staining with more than one primary antibody for same host species (Table S2; Panels I, IV, VI and VII), sections were incubated overnight at 4 °C with an excess of unconjugated donkey anti-mouse IgG Fab fragment (Table S1; Jackson ImmunoResearch, West Grove, USA) after the first primary and secondary antibody incubations, prior to the application of the second primary antibody. Sections were washed in TBS as above then incubated with second primary antibody, followed by the steps outlined above. A known positive control slide was included in each batch of staining. p16 and Ki67 staining was carried out by Southern Community Laboratories, Dunedin, New Zealand.

Images of the entire tissue for each slide were taken with the DP80 camera (OLYMPUS) with the 20 × objective on a BX53 microscope (Olympus). The CIN2 lesion was defined by a pathologist, and the area of lesion was measured using the tool in the Image J. The dermal area within 500 µm of the epithelial basement membrane was separately measured. The positive cells were counted on each tissue using the cell counter plugin in the Image J software (https://​imagej.​nih.​gov/​ij/​), and the cells per mm² were calculated. IDO-1, PD-L1 and E1^E4 staining was classified as either negative (absent) or positive (including focal staining) in the CIN2 region only. Ki67 was staining was analysed as has been reported previously, with positive staining in the lower one third, lower two thirds or three thirds of the epidermis being described (Reuschenbach et al. 2012).

For each continuous variable samples were separated into a high group (above the median) and a low group (below the median). Progression to CIN3 + over the 24-month follow-up was assessed for the low and the high groups and survival curves were generated with CIN3 + as the endpoint. A univariate analysis was carried out using the Log-rank (Mantel–Cox) test to calculate P values and the Mantel–Haenszel test was used to calculate univariate hazard ratios. Chi-square analysis and Fisher’s exact tests were used to test the relationship between categorical measures. The Mann–Whitney U (M–W U) test was used for pairwise comparisons. The Spearman’s rank-order correlation analysis was performed to calculate the correlation coefficient when comparing different markers. Multiple logistic regression and receiver operating characteristic (ROC) curve analyses were carried out to assess the specificity and sensitivity of selected markers for the detection of progression to CIN3. A Bonferroni-corrected P value of < 0.05 was considered statistically significant when carrying out multiple comparisons. All statistical analyses were performed with Prism software (Version 9.0, GraphPad Software, Inc).

The host immune system plays a key role during cervical carcinogenesis (Doorbar 2006). In this study, we have characterized antigen presenting cells, T and B cells and other specific populations of interest in CIN2 and the dermal region below the lesion in a subset of women under 25 who participated in an observational management trial (the PRINCess study). The regression rate (CIN1 or less) for the PRINCess study was 73% at the 12-month follow-up (Innes et al. 2018). For this study, we over-sampled women from the PRINCess study who had progressed to CIN3 + to generate similar group sizes for the persister/regressor and the CIN3 + progressor groups that were studied here.

A total of 69 cases with high-risk HPV-positive test results and histologically confirmed p16 and Ki67-positive CIN2 were recruited to this study. The participants were separated into the persister/regressor group consisted of (35 cases) or the ‘progressor’ group (34 cases). The characteristics of the study population at baseline are shown in Table 1. Patient age, smoking history and HPV-vaccination status was comparable between the persister/regressor and the CIN3 + progressor groups. The CIN2 baseline lesions in the progressor group were significantly larger (P = 0.0039; M–W U test) than the lesions that persister/regressor group. However, a univariate analysis of area with progression to CIN3 + as the outcome did not show any significant association (Table 2).

Expression of the viral E1^E4 protein occurs at the onset of viral genome amplification and accumulates late in the viral lifecycle (Doorbar 2013). We hypothesized that tissues that were actively infected and E1^E4 positive would not show full thickness staining with the proliferation marker Ki67. Tissues from both groups predominantly showed Ki67 staining in the upper two thirds of the lesion. Only 11 tissues in total stained positive for E1^E4, using the pan-E1^E4 antibody (Table 3). All the E1^E4 positive tissues were Ki67 positive in the lower third or two-thirds only, and not the full thickness of the lesion. There was no significant difference in Ki67 (P = 0.3927; Chi-square test) or E1^E4 positivity (P = 0.3221; Chi-square test) between the persister/regressor and the progressor groups.

A detailed analysis of immune-related markers in the CIN2 lesion and in the dermal region spanning the basement membrane up to 500 µm below the lesion was carried out. Examples of all staining are shown in Suppl. Figure 1. Several of the markers, granzyme B, fascin, GATA3, TSLP and CD138, stained some or all the keratinocytes in the CIN2 region. Except for TSLP, all positive cells in the lesion, including the keratinocytes, were counted for each of the markers. In the case of TSLP, the CIN2 lesions were uniformly stained; therefore, the individual cells were not counted.

A graphical representation of the cellular infiltrates into the lesion and the dermal region beneath the lesion is shown in Fig. 1. CD4 T cells were present in the CIN2 lesion but were more frequently detected in the dermal region beneath the lesion. Only around 3% of the CD4 cells were also positive for the Treg marker FoxP3, both in the lesion and the dermal region. Parallel staining for T-bet suggested that proportionally around a third of the CD4 cells might be Th1 cells. GATA3-positive staining identifies Th2 cells; however, the number of GATA3-positive cells in the dermal region exceeded the total number of CD4 T cells, due to GATA3 + keratinocytes also being present. In the case of IL-17, the number of cells with positive staining in the nuclei were similar to the total CD4 T cells, suggesting that cells other than CD4 T cells were also being stained. This may be a result of infiltration of other immune cells that express IL-17, including NK T cells, CD8 T cells and neutrophils.

CD8 T cells were detected in CIN2 lesions but were approximately half as frequent as CD4 T cells and were around fivefold less frequent in the lesion than in the dermal region beneath the lesion. Granzyme B expression by the CD8 T cells was measured as an indicator of cytotoxic activity by the cells, however only around 1% of the epidermal CD8 T cells expressed granzyme B, whereas a slightly higher proportion of dermal CD8 cells (7%) expressed granzyme B.

We explored markers that stained antigen-presenting cells (langerin, CD11c, fascin, Clec9A, DC-LAMP) in the CIN2 biopsies. Langerhans cells (LCs), detected using antibodies specific for langerin, were present in the epidermis in low numbers (< 20/mm²) and were close to undetectable in the dermis (< 2/mm²). Fascin expression in LCs is associated with maturation of LCs and formation of dendritic processes (Ross et al. 1998). The proportion of fascin-positive LCs was higher in the epidermis than in the dermis. We also looked at CD11c, which is a marker for the cDC2 subset in humans. As might be expected, CD11c-positive cells were around four times more frequent in the dermis than in the epidermis, but surprisingly were more commonly found in the epidermis than Langerhans cells. CD11c positive DCs were previously reported to be enriched in anogenital tissues compared to LCs (Bertram et al. 2019). We stained tissues with antibodies for Clec9A to detect cDC1 cells, and DC-LAMP for mature DCs. Both antibodies stained positive control cells in tonsil but no positive cells were detected in the CIN2 biopsies tested in this study.

Antibody secreting cells express CD138, regulating their survival (McCarron et al. 2017). CD138 is also expressed in some tumours and in normal squamous epithelium (Kind et al. 2019), consistent with the full thickness staining of the epidermis that we observed in this study. The staining pattern of CD138 + cells in the dermal region was for surface-stained infiltrating cells, and the numbers of positive cells was similar to the numbers of CD8 T cells that were found in this region and fewer than the CD4 T cells. We also determined the number of CD32B + cells in the epidermis and dermis, which is an inhibitory surface receptor that can down-regulate B cell function, including antibody production. CD32B + cells were detected in similar numbers in the lesion and the dermal region, around tenfold lower than the numbers of CD138 + cells that were detected in the dermal region.

We stained tissues for several positive and negative immune regulatory proteins. HMGB1 is a widely expressed protein with an alarmin function, initiating inflammation (Yang et al. 2020). Pang et al. (2014) found that elevated expression of HMGB1 was associated with a high recurrence of HPV infection in patients with cervical cancer. In this study, we found that HMGB1 was expressed predominantly in the nucleus of most nucleated cells in the dermis, which are primarily fibroblasts. Staining intensity varied from weak to strong, with around 50% of the cells showing strong positive staining. Staining was also observed in the nucleus of epidermal keratinocytes, but the number of positive cells was only about one fifth of the number observed for the equivalent area in the dermal region.

TSLP is a protein expressed T cells and by suprabasal keratinocytes. Its expression induces the release of T cell attracting chemokines, and DC maturation. We observed a small subset of infiltrating cells (< 5% of all TSLP-positive cells) that were strongly TSLP positive in the dermal region beneath the lesion, in addition to the uniform but weaker staining seen in the lesion itself. The numbers of TSLP positive cells were equivalent to around 40% of the total CD4 and CD8 T cells.

The transcriptional repressor Blimp-1 is a regulator of B and T cell development and differentiation (Fu et al. 2017). Loss of Blimp-1 indicates poor prognosis in B-cell-like diffuse large B-cell lymphoma (Xia et al. 2017), whereas Zhu et al. (2017) showed that Blimp-1 inhibited the T cell response in acute myeloid leukemia. Blimp-1-positive cells were identified in the lesion and more frequently in the dermal region beneath the lesion. We were able to characterize Blimp-1-positive cells for FoxP3 and CD4 staining. Around 10% of the Blimp-1-positive cells in the lesion stained for FoxP3 or CD4, and only 1% were triple positive cells. In the dermal region 14% of the Blimp-1-positive cells were CD4 positive, and only 3% of the cells were triple positive (Suppl. Figure 2).

The relationship between high or low numbers of the cells identified with each of the markers and progression to CIN3 + over the 24-month follow-up period was assessed (Fig. 2; Table 2). For the univariate analysis, high numbers of cells positive for Blimp-1 (Log-rank test; P = 0.002), CD4-FoxP3 + (P = 0.02) or HMGB-1 (P = 0.01) in the epidermal region (P = 0.002) were independently and strongly associated with progression to CIN3. Low numbers of CD4 + (Log-rank test; P = 0.002), CD8 + (P = 0.004) and CD11c + (P = 0.02) cells in the dermal region was also associated with progression to CIN3. Following a Bonferroni correction to adjust for multiple comparisons, high Blimp-1 in the epidermis and low CD4-positive cells in the dermis were confirmed to be significantly associated with progression to CIN3 + .

Expression of immune checkpoint molecules IDO-1 and PD-L1 was also assessed in the CIN2 lesions (Table S3). Overall, less than 20% of the lesions were positive for each of these markers. There were some double positive tissues (5/69), with most double positives (4/35 c.f. 1/34) in the progressor group. There was no relationship between expression of these markers and persistence/regression or progression of disease.

The relationships between the different cell subsets in the lesion and the dermal region beneath the lesion in the persister/regressor and progressor tissues were evaluated (Fig. 3A). In the epidermis, CD4 cells correlated more strongly with T-bet, IL-17, CD8 and CD11c cells in persister/regressors, compared with the progressor group. Interestingly, T-bet cells inversely correlated with CD11c cells in the persister/regressor group, not in the progressor group. Blimp-1 expression in the lesion did not strongly correlate with other cell populations, except for FoxP3 in the persister/regressor group. In contrast, HMGB1 correlated with FoxP3 positive cells in both groups. The dermal region of the persister/regressor tissues showed the strongest correlations between different cell populations. The related cells were consistent with the key components of an adaptive response. CD4 cells correlated with CD11c-positive cells, CD8 and CD8/granzyme B-positive cells and GATA3, T-bet and FoxP3-positive cells. Blimp-1 and TSLP expression was also positively correlated with CD4 T cells in persister/regressor tissues. This contrasted with the progressor lesions, where CD4 cells correlated only with FoxP3 and CD8 in the dermis.

A multivariate model (multiple logistic regression) was developed to adjust for the effects of age, smoking and vaccination status (Table 2). Blimp-1, HMGB1 and T-bet in the lesion retained statistical significance between the persister/regressor and progressor groups in the multivariate model; however, CD4 T cells in the dermal region were no longer significant following the adjustment, suggesting that one or more of these variables was a confounding factor in the analysis. Following a Bonferroni correction for multiple comparisons, Blimp-1 expression in the lesion retained the significant association with survival when adjusted for confounding factors.

An ROC analysis was carried out as part of the multivariate analysis and the ROC curves showing the sensitivity and specificity of Blimp-1 and CD4 expression in the lesion and the predicted/observed graphs are shown (Fig. 3B). Based on the median cut-off, the separation of samples on high or low expression of Blimp-1 correctly predicted 85% of CIN2 that would progress to CIN3 + and 79% of CIN2 cases that remained CIN2 or regressed to CIN1 or normal, giving Blimp-1 a negative predictive power of 83% and a positive predictive power of 81%.