Characterising the local immune responses in cervical intraepithelial neoplasia: a cross-sectional and longitudinal analysis

Authors


Dr YL Woo, Department of Gynaecological Oncology, Box 242, Addenbrooke’s Hospital, Cambridge CB2 2QQ, UK. Email ylw22@cam.ac.uk

Abstract

Introduction  Immunological competence influences the progression of cervical intraepithelial neoplasia (CIN) to invasive cancer. Information on the local immunological changes during the natural course of CIN is central for the development of new therapies.

Objective  This study defines the populations of tissue-infiltrating immune cells in a cross-sectional cohort of different grades of CIN and also in a longitudinal cohort of regressing, persistent and progressing low-grade (LG)-CIN.

Design  A cohort of 125 women with LG cytological atypia was recruited, of which 64/125 (51%) women with LG-CIN were followed prospectively for 1 year. Paraffin-embedded entry and exit cervical biopsies were used for immunohistochemistry analysis (CD4, CD8, CD56, FOXP3, CD1a and granzyme B).

Results  At recruitment, 74/125 (59%), 39/125 (31%) and 12/125 (10%) women referred with LG smears had histologically proven LG-CIN, high-grade (HG) and normal biopsies, respectively. Seventeen of 64 (24.6%) women with LG-CIN progressed to HG-CIN within 1 year. In both LG-CIN and HG-CIN, the predominant intraepithelial cell population were cytotoxic T cells, while CD4+ and FOXP3+ T cells predominated the stromal compartment. Women with LG-CIN who later on regressed displayed a significantly higher number of cytotoxic (granzyme B+) cells in their entry samples. In addition, the ratio between CD8+ cells and granzyme B+ cells was close to 1, suggesting that all infiltrating CD8+ T cells were highly active. In contrast, this ratio was three-fold lower in women, in whom the lesions persisted or progressed.

Conclusions  This study suggests that the early infiltration of lesions by highly cytotoxic effector cells protects against progression.

Introduction

To date, therapeutic vaccines targeted at cervical cancer or cervical intraepithelial neoplasia (CIN) have met with limited clinical success,1–3 despite their capacity to induce strong systemic immunological responses. This is because the characterisation of specific systemic immune responses does not always reflect the local immune responses in the lesion of interest. It is now generally accepted that persistent infection with high-risk human papillomaviruses (HPVs) initiates and, together with other factors, promotes the progression to overt cervical carcinoma. Epidemiological and natural history studies suggest that clinically relevant CIN2/3 (high-grade [HG]) lesions may be rapidly induced in a short period4–6 following infection but then requires another decade or so before invasive cervical cancer develops. The two different processes that require clear definition are (1) the productive viral infection associated with low-grade (LG)-CIN and (2) HG-CIN characterised by deregulation of HPV gene expression, prolonged cell cycle dysregulation, chromosomal instability and resistance to apoptosis, all of which hugely increase the risk of invasion.7

Another confounding factor in elucidating the natural history of CIN is the host immune response. Very little is known about the individual defence response to HPV infection. The host immune response to HPV is important in determining the outcome of infection. Thus, HPV infection is detected more frequently, the incidence of all grades of CIN is higher and the risk of CIN recurrence after treatment is higher in immunocompromised women.8–10 This is also supported by epidemiological studies showing that in both human and animal hosts with genetic, iatrogenic or acquired cell-mediated immune deficiencies, papillomavirus lesions show increased persistence and enhanced neoplastic progression,11 but systematic evaluation of the local immunological events in HPV-associated lesions in humans is limited.8,12,13 To modify and improve on current efforts, an understanding of the natural history of CIN is required, more specifically, the microscopic or even molecular evolution of disease.

In the present study, the local tissue-infiltrating immune cells are studied and defined in (1) a cross-sectional cohort of LG-CIN and HG-CIN lesions followed by (2) longitudinally among women with LG-CIN who regress, persist and progress over the period of 1 year.

Patients and methods

Patients

A total of 125 women, aged 20–30 years, referred to Addenbrooke’s Hospital for colposcopy were recruited for a prospective study investigating the immunological responses in women with LG-CIN. All colposcopic assessment was performed by a single colposcopist (Y.L.W.) throughout the study. As part of the study, women had colposcopic assessment and cervical biopsy performed at entry and exit of the study. Those with histologically confirmed LG-CIN received follow-up visits. Others did not receive follow up as part of the study because of the following reasons: HG-CIN requiring excision, normal entry biopsy and default because of pregnancy or loss of follow up. This study was approved by the Local Research Ethics Committee, Addenbrooke’s Hospital, Cambridge.

Histology report

All histological slides were reviewed by a single consultant histopathologist (N.C.) for the purpose of this study. The pathologist was blinded to all other clinical information. Cases were classified as normal, LG-CIN and HG-CIN. Women who completed the study were classified histologically as progressors, persistors and regressors. The definition of a progressor was when the exit histology was worse than the entry, a persistor was when there was no change in histology and a regressor was when there was an improvement in the histological status.

Immunohistochemistry

Multiple 5-μm serial sections from formalin-fixed, paraffin-embedded blocks were cut onto aminopropyltriethoxysilane-coated slides (Department of Histopathology, Addenbrooke’s Hospital). The sections were dewaxed with xylene and rehydrated with graded ethanol. Antigen retrieval was achieved through pressure-cooking for 3 minutes in citrate buffer (pH 6.0). The following mouse monoclonal antibodies and dilutions were used: CD4 (1 in 40), CD8 (1 in 40), CD56 (1 in 50), granzyme B (1 in 60) all from Novo Castra, Newcastle-upon-Tyne, UK and FOXP3 (1 in 60) from AbCam, Cambridge, UK. Biotinylated goat anti-mouse was used as the secondary antibody. The colour change reaction was performed with horseradish peroxidase-labelled streptavidin–biotin complex using diaminobenzidine as substrate. The slides were then counterstained with Harris haematoxylin, dehydrated and mounted.

Quantification of cells

The numbers of immunophenotyped cells in the stained sections were counted by visual inspection. For each section, photographic images of each microscopic high-power field at 40× magnification were obtained. Multiple (2–6) photographs were taken at the epidermis/stroma interface of each section (Y.L.W.). Intraepithelial and stromal cell counts were made separately before an aggregate was calculated. To reduce intra- and inter-observer variability, duplicate counts of five slides were undertaken by the two individuals who performed all the counting (Y.L.W. and I.D.). For each woman, the average counts were calculated, and for each group, the results were expressed as median count.

Statistical analysis

The Mann–Whitney U test and Kruskal–Wallis test (nonparametric) were used to analyse the difference between the medians of the two and three histological groups, respectively, while chi-square test for trend was used to calculate significance of relationships observed. The values of P < 0.05 were regarded as significant. Statistical analyses were performed using GraphPad InStat software.

Results

Comparing local immune factors between LG-CIN and HG-CIN cohorts

From the cohort of 125 women, 115 LG-CIN and 59 HG-CIN cross-sectional slides (not taking into account the time course) were available for immunohistochemistry. The average CD8, CD4, FOXP3, CD56, granzyme B and CD1a counts within the intraepithelial and stromal compartments were enumerated (Table 1).

Table 1.  Immune cell counts in LG-CIN and HG-CIN
 CD4CD8FOXP3CD56GBCD8/CD4GB/CD8GB/CD56CD8 + CD56/CD4FOXP3/CD4CD1a
  1. GB, granzyme B.

  2. The cell counts represent the average value of multiple 40× high-power fields. P values were calculated using Mann–Whitney statistics.

Intraepithelial compartment
LGCIN3.128.182.624.913.694.30.711.226.021.842.22
HGCIN3.327.372.647.354.773.620.741.497.211.282.78
P value (Mann–Whitney U test)0.250.990.730.110.080.70.130.570.430.790.07
Stromal compartment
LGCIN16.8512.567.364.572.431.120.241.431.390.53 
HGCIN11.7815.819.543.643.010.920.312.171.090.66 
P value (Mann–Whitney U test)0.150.030.190.20.20.870.220.120.610.66 
Total average counts
LGCIN19.9720.749.989.486.121.270.41.252.050.632.22
HGCIN15.123.1812.1810.997.781.210.422.261.610.692.78
P value (Mann–Whitney U test)0.160.140.310.370.030.860.360.530.640.600.07

In both LG-CIN and HG-CIN, the predominant intraepithelial cell population were cytotoxic T cells (CD8 and CD56+). When the two groups were compared, the only significant difference were a higher stromal CD8 population and total granzyme B positive cells in the HG-CIN group (P = 0.03). However, the higher cytotoxic activity among the HG-CIN appears to be within the epithelial compartment and not the result from the higher CD8 cells within the stromal compartment. It is possible that the higher cytotoxic activity within the epithelial compartment is contributed by natural killer (NK) cells. In comparison, CD4 and FOXP3 positive cells were significantly higher in the stroma than in the epithelium in both LG-CIN and HG-CIN groups. However, no significant differences were noted between the groups. In summary, the cytotoxic T cells were significantly more abundant within the intraepithelial compartment, while the CD4 and regulatory T cells predominated the stromal compartment.

Clinical outcomes of women referred with LG cytological atypia

At the end of the 12-month follow-up period, a total of 64/125 (51%) women completed the study. At the first visit, 39/125 (31%) women had histologically confirmed HG-CIN and therefore underwent loop excision of the transformation zone, 11/125 (9%) had normal histology, while 11/125 (9%) defaulted.

Using cytology as a starting reference and histology to provide the end-point diagnosis, 14.8% of women from this study regressed, while 60.7% persisted over the course of 1 year.

Histologically, of the 64 women who completed the 1-year follow up, 37/64 (58%) persisted, 17/64 (27%) progressed, while 10/64 (16%) regressed over that period. The immunohistochemistry results of these three groups taken at time of recruitment (entry biopsies) and at 1 year (exit biopsies) are shown in Table 2.

Table 2.  The average cell counts of entry and exit cervical biopsies
 Progressors (n = 17)Persistors (n = 37)Regressors (n = 10)
EntryExitEntryExitEntryExit
  1. The LG-CIN cohort was classified as progressors, persistors or regressors after 1-year observation.

Epithelial (average cell count per 40× high-power field)
CD43.502.634.662.432.821.73
CD87.246.739.626.216.285.89
FOXP32.291.543.281.752.292.64
CD563.254.294.624.278.743.92
Granzyme B2.744.932.643.635.062.96
CD1a1.992.711.892.311.901.56
Stroma (average cell count per 40× high-power field)
CD426.4619.9316.8214.0713.3518.34
CD814.4315.0912.739.8018.9510.18
FOXP38.134.687.685.419.654.37
CD569.063.023.503.065.4010.13
Granzyme B2.33.331.722.263.732.16

Comparing local immune cells in progressing, persistent and regressing LG-CIN

To gain more information on the relationship between the local immune response and the clinical outcomes, LG-CIN infiltrating CD4, CD8, CD56, FOXP3 or granzyme B positive immune cells present in the entry and exit biopsy samples of histological progressors, persistors and regressors were enumerated (Table 2).

No overt differences were found in the number of the different types of infiltrating immune cells between the three groups, either at entry or at exit. However, among the women with LG-CIN who later on regressed (regressors), a significantly higher numbers of cytotoxic (granzyme B+) cells (P = 0.019, Kruskal–Wallis test) were found in their entry samples.

When the T-cell subpopulations in progressors and regressors were studied more closely, a few observations could be made. First, the ratio between CD8+ cells and granzyme B+ cells is close to 1, suggesting that the CD8+ cells that already infiltrated the lesion at recruitment were highly active. Notably, this ratio was three-fold lower in the lesions that persisted or progressed. It was also observed that the CD8/CD4 ratio was 1.6-fold higher among regressors. Comparison of all the other medians revealed no other significant results.

Discussion

The natural history of CIN is difficult to study systematically. Cross-sectional studies are just snapshots of a highly dynamic process, while longitudinal prospective studies are flawed with sampling errors. Furthermore, it is becoming clear that T cells that are phenotypically similar based on surface markers may in fact have different functions at different stages of viral infection.14 However, despite all these limitations, such studies complement one another to provide important information on the natural history of CIN.

Similar to other studies, we observed that up to 26.7% of borderline abnormal smears and 27.9% of mildly dyskaryotic smears had HG-CIN.15,16 More accurate predictions of the grade of CIN are acquired by cervical biopsies,17 although this still varies between 50 and 89.6%.18–22 Few studies use cervical biopsies to study natural history as there are concerns that a biopsy may affect the natural course of events. The effects of biopsy may theoretically heighten the inflammation process and lead to artificially higher rates of disease regression. In this prospective study, the data do not suggest an overestimate of regression rates.

Natural history studies using cross-sectional samples suggest that the modal time between HPV infection and CIN3 approximates 7–15 years.23 In contrast, in our prospective cohort, 27% of the women progressed from histologically confirmed LG-CIN to a histologically confirmed HG-CIN within a year. Other prospective studies, in particular those who follow young women from the time of first intercourse, reveal rapid conversions from LG-CIN to HG-CIN within 1–2 years,5,24–27 suggesting that clinically relevant lesions may rapidly progress in a short period and indicating that the current model on HPV-induced progression does not accurately reflect the natural course of disease.

Cell-mediated immunity is important for the regression of HPV-related pathology, which can be inferred from the animal models where infection and regression can be studied systematically.24,25 Studies that investigate the local immune environment in the cervix are in general cross-sectional, characterising immune infiltrates in normal, LG-CIN, HG-CIN and cervical carcinoma.8,26–29

Granzyme B is an enzyme that exists in cytoplasmic granules within cytotoxic T cells and is only expressed upon activation. Granzyme B serves to induce apoptosis within virus-infected cells, thereby destroying them.30 Cervical tumours have been found to be infiltrated with more granzyme-B-staining cytotoxic lymphocytes (CTLs) compared with CIN, where only a small proportion of CTL are active.31 Strikingly, we found that although there was no difference in the number of CD8+ T cells, a notable difference was observed in the number of tissue-infiltrating cytotoxic (granzyme B+) cells among the regressors at recruitment. This difference was not observed at the end of the study when the disease had already regressed, implying that cytotoxic activity was no longer needed. This is in contrast to some cancers where increased percentage of granzyme B+ tumour-infiltrating lymphocytes carry poor clinical prognosis.32,33 Similarly, the cross-sectional data from this study show a higher expression of granzyme B among HG-CIN, suggesting that indeed this may be a poor prognostic marker. However, the longitudinal data suggest that in early disease, it may in fact play a protective role, highlighting the differences between cross-sectional and longitudinal observations as well as the disease groups studied.

Furthermore, the regressors in the present study demonstrated a higher CD8/CD4 ratio compared with progressors, which is in keeping with studies that show that a high CD8/CD4 ratio was predictive of better clinical outcomes in cervical34 and other cancers.35–37 Recent studies demonstrated that regulatory T cells (FOXP3+ cells) are associated with progression and poor prognosis of several epithelial cancers,38–40 including cervical cancer,34 based on the fact that the prevalence of regulatory T cells was significantly higher in the women with carcinoma than in the normal tissue. We did not observe any difference in FOXP3+ cells among the three groups either at recruitment or at the end of the study, suggesting that regulatory T cells may not have a major role in premalignant neoplastic process.

In summary, this study clearly demonstrates that cross-sectional and longitudinal studies on CIN provide us with different but complementary information. Notably, the infiltration of cytotoxic effector cells (granzyme B+) into the local environment of LG-CIN is important for lesion regression. This is strikingly different from studies in cancers where it has been reported that active cytotoxic effectors, particularly CD8+ T cells are poor prognostic markers. The distinction supports the fact that LG-CIN is essentially a productive viral infection susceptible to classical cytotoxic attack, whereas HG and invasive lesions have acquired the ability to evade such immune attack. This supports the idea that identifying the specific antigens/viral targets for these local active cytotoxic effectors and subsequently boosting the numbers may be of therapeutic potential, particularly if early lesions are targeted.

Disclosure of interest

There is no conflict of interest.

Contribution to authorship

Y.L.W. conducted the research, undertook the experiments, analysed the data and wrote the manuscript. I.D. performed the experiments and analysed the data. N.C. collaborated in the immunohistochemistry analysis and interpreted all the histological slides. M.S., R.C. and S.H.v.d.B. conceived and collaborated in the conceptualisation and design of the project. J.S. supervised the experiments, undertook the interpretation and analysis of the data. All authors contributed to the revision of the manuscript.

Details of Ethics approval

This study was approved by Addenbrooke’s Hospital NHS Trust (LREC 03/061).

Funding

This work has been funded by Cancer Research UK. Y.L.W. received the Gordon Hamilton-Fairley CRUK Clinical Fellowship.

Editor’s Commentary

Cervical cancer and its precursor, cervical intraepithelial neoplasia (CIN), are caused by human papillomaviruses (HPVs). This is therefore an important disease model in the evaluation of cancer resulting from an infectious agent. The immune response has been shown to influence the control of HPV infections and the regression of CIN. Therefore, hypothetically, the modulation of the appropriate immune response can be used as therapy for CIN. The current treatment of high-grade CIN (HG-CIN) is at present based on ablative and excision measures. To develop immunotherapy capable of resolving the underlying persistent HPV infection, the local immune response in the different stages of CIN needs to be further defined. Low-grade CIN (LG-CIN), which represents productive viral infection, is different from HG-CIN and is likely to have different immune mechanisms in place. It is therefore helpful to compare the immune response in these two distinct entities.

A literature search highlights the limited number of studies investigating the local immune environment in CIN. Furthermore, most studies are cross-sectional in nature. The strength of this study is its prospective design and the discriminate use of immunohistochemistry analysis. It allowed the evaluation of critical immune cell markers to be undertaken longitudinally and also cross-sectionally such as cluster of differentiation CD4, CD8 and CD56, which are co-receptors that assist the T-cell receptor to activate its T cell following an interaction with an antigen-presenting cell; FOXP3, which is a transcriptional regulator and functions as the master regulator in the development and function of regulatory T cells and granzyme B, which is an exogenous serine protease that is released by cytoplasmic granules within cytotoxic T cells and natural killer cells. Their purpose is to induce apoptosis within virus-infected cells, thus destroying them. The findings of this study support the hypothesis that LG-CIN is indeed different from HG-CIN. More importantly, it has demonstrated that the local immune markers of LG-CIN that progress to HG-CIN are different from LG-CIN that regress.

What are the clinical implications? Many published studies on therapeutic vaccination for CIN target HG-CIN and are E6 and E7 based. Both E6 and E7 are oncoproteins that affect cell regulation in many ways, E6 primarily promotes carcinogenesis by inactivating P53 and E7 inhibits Rb proteins, which act as tumour suppressors. When tumour suppressor proteins are inactivated, tumour growth proceeds unchecked. These targets therapeutic vaccines against these oncoproteins have met with limited success. This study seems to support the notion that effective immune therapies should in fact be aimed at early lesions, that is LG-CIN. Since cytotoxically active cells have been shown to be important for the regression of LG-CIN, the next step should be aimed at defining the antigen specificity of these effectors. It is not inconceivable that the modulation of the different subsets of antigen-specific cytotoxic effectors can be used as an additional arm in the armamentarium to combat cervical cancer and potentially virally induced malignancies.

P Martin-Hirsch
Lancashire Teaching Hospitals NHS Trust, UK

Ancillary