The antigen-presenting environment in normal and human papillomavirus (HPV)-related premalignant cervical epithelium

Abstract

The activation of HPV-specific T cells within the cervical microenvironment is likely to play an important part in the natural history of cervical intraepithelial neoplasia (CIN). The extent and the type of T cell activation will depend critically on the expression of MHC, costimulatory cell surface molecules and cytokines by keratinocytes and Langerhans cells within the cervical lesion. Expression of MHC class II (HLA-DR and -DQ), costimulatory/adhesion molecules (CD11a/18, CD50, CD54, CD58 and CD86) and cytokines (tumour necrosis factor-alpha (TNF-α) and IL-10) was therefore investigated by immunohistochemistry in normal squamous epithelium (n = 12), low-grade (n = 23) and high-grade (n = 18) squamous intraepithelial lesions of the cervix. CIN progression was associated with de novo expression of HLA-DR and CD54, and increased expression of CD58 by keratinocytes. However, significantly, there was no expression of any adhesion/costimulation molecule by epithelial Langerhans cells in any cervical biopsy studied. Furthermore, TNF-α, a potent activator of Langerhans cells, was expressed constitutively by basal keratinocytes in normal cervix (12+/12), but expression of this cytokine was absent in a number of CIN samples (20+/23 for low-grade, 12+/18 for high-grade CIN). Conversely, the suppressive cytokine IL-10 was absent in normal epithelium (0+/12), but was up-regulated in a number of CIN lesions (12+/23 for low-grade, 8+/18 for high-grade CIN). The restricted expression of costimulation/adhesion molecules and the nature of the cytokine microenvironment within the epithelium may act to limit effective immune responses in some CIN lesions.

INTRODUCTION

In the aetiology of cervical cancer, and of its precursor lesion cervical intraepithelial neoplasia (CIN), HPV plays a major role [1, 2]. Several lines of research have implicated the host immune response as a critical factor in the control of these conditions. One example of this is that HPV16, the most prevalent type in the disease, encodes a tumour-specific antigen which can trigger a cytotoxic T cell response [3]. Adoptive transfer of these T cells can lead to destruction of tumours in syngeneic hosts.

A more complex, but possibly equally important reaction is via HPV-related changes in local CD4⁺ T cell responses [4, 5]. This T cell response is itself regulated by MHC class II molecules (HLA-DR and -DQ) which are expressed on professional antigen-presenting cells (APC) such as Langerhans cells (identified by surface CD1a) within the cervical epithelium. However, APC within peripheral tissue, such as cervix, do not normally function in antigen presentation in situ, but mature into an APC phenotype only after migration to lymph nodes or spleen. Increased numbers of ‘activated’ Langerhans cells outside lymphoid tissue are only found during chronic or persistent immunological responses [6, 7].

Keratinocytes, constituting 95% of the cervical epithelium, can also express class II molecules during inflammation and cell transformation [8, 9], and these cells can provide some accessory costimulatory signals to T cells [10]. The costimulatory properties of epithelial cells have been linked previously to up-regulation of some of the molecules that are important in antigen presentation, such as CD54 (intercellular adhesion molecule-1 (ICAM-1)), but analysis of several other associated costimulatory molecules, such as CD50 (ICAM-3), CD58 (LFA-3) and CD86 (B7-2) has not been documented within the cervix.

The antigen-presenting cell function of both Langerhans cells and keratinocytes is under exogenous control by a variety of soluble mediators. In this context, tumour necrosis factor-alpha (TNF-α) and IL-10 are an important pair of antagonistic cytokines. TNF-α, with a broad spectrum of proinflammatory activities, is also a potent activator of Langerhans cell antigen presentation [11]. In contrast, IL-10 is immunosuppressive, and this property is mediated via an effect on the APC [12–14].

In this study, we focus on two fundamental and interrelated questions regarding the immune microenvironment of preinvasive cervical disease. The first is the array of antigen-presenting costimulatory cell surface molecules expressed during disease progression, and whether these are localized to the professional APC, or more broadly to the epithelial keratinocyte, which is the cell where viral replication occurs and is the target of neoplastic transformation. The second looks selectively at the two cytokines, TNF-α and IL-10, known to be expressed by activated epidermal (cutaneous) keratinocytes [15, 16] which are likely to play an important role in regulation of antigen-presenting phenotype and function within the epithelium. Our studies demonstrate a very restricted costimulatory phenotype of both Langerhans cells and keratinocytes within either normal cervix or CIN lesions. Furthermore, expression of the proinflammatory cytokine TNF-α seems to be decreased, while expression of the suppressive cytokine IL-10 increases in these lesions. The failure to develop a local microenvironment conducive to appropriate T cell activation may contribute to the persistence and eventual progression of CIN lesions.

MATERIALS AND METHODS

Clinical specimens

Cervical samples were obtained from 41 patients attending the Colposcopy Out-patient Clinic at the Whittington Hospital, London, for diagnostic and therapeutic procedures, after recent evidence of abnormal cervical cytology. Clinical material ranged from colposcopically directed punch biopsies to resection specimens (loop excision). Tissues from patients undergoing total abdominal hysterectomy for benign conditions (n = 12), with normal cervical cytology and no evidence of previous cervical disease, were used as controls.

All biopsies were immediately snap-frozen on dry ice before being stored in liquid nitrogen until processed further. The samples were mounted in OCT compound (BDH, Poole, UK) and serial 6-μm cryostat sections were taken for immunohistochemistry and in situ hybridization. Intermediate frozen sections were stained with haematoxylin–eosin (H–E) and the biopsies were classified as normal or as representing low-grade (koilocytosis and/or CIN I) or high-grade (CIN II and/or CIN III) intraepithelial squamous lesions [17].

Human palatine tonsils, obtained from patients aged 1–20 years on whom routine tonsillectomy was performed at the Royal National Ear Nose and Throat Hospital, London, were used as positive controls for immunohistochemistry and in situ hybridization. The cryostat sections were prepared using the procedures described above.

Immunohistochemistry

Cryostat cervical sections mounted on vectabond-coated slides (Vector Labs, Peterborough, UK), air-dried for 30 min at room temperature followed by acetone fixation for 10 min at −20°C, were stained using an indirect immunoperoxidase technique. After rinsing in Tris-buffered saline (TBS) pH 7.6, sections were preincubated with normal rabbit (1:20) serum (Gibco BRL, Paisley, UK) for 15 min and incubated for 60 min at room temperature with specific antibodies (Table 1). Where more than one MoAb was available for a specific antigen, they consistently gave similar staining patterns. After washing, the sections were incubated for a further 60 min with normal human (1:25) serum and a secondary antibody, horseradish peroxidase (HRP)-conjugated rabbit anti-mouse (Dako, High Wycombe, UK) at a dilution of 1:50. Antibody binding was visualized using 3,3-diaminobenzidine (Sigma Chemical Co., Poole, UK) at 5 mg/ml and hydrogen peroxide (6 μl of 60% w/v H₂O₂) for 10 min. Slides were counterstained with Mayer's Haematoxylin (BDH), dehydrated and mounted in a resinous mountant (Eukitt; Kindler GmbH Co., Frisburg, Germany).

Table 1.

Primary MoAbs used in this study

Tonsillar sections used as positive controls were subjected to the same procedures. As negative control, a MoAb directed against type II collagen (IgG) and no primary antibody was also included. The expression of HLA-DR, HLA-DQ and CD54 (ICAM-1) by keratinocytes was evaluated using a semiquantitative immunohistologic grading system as described previously [9]. A total score was awarded which represented the sum of the intensity and the extent of staining for each biopsy. The two parameters were scored as follows. Intensity: 0, no staining; 1, weak staining; 2, moderate; and 3, intense staining. Distribution: 0, patchy basal positivity; 1, diffuse basal positivity; and 2, full-thickness positivity. The specimens were also classified according to the thickness of the cervical epithelium expressing IL-10 and TNF-α: no staining, basal, basal and parabasal staining, and full-thickness staining. HLA-DR, HLA-DQ, CD54 (ICAM-1), CD50 (ICAM-3), CD11a/18 (LFA-1) and CD58 (LFA-3) immunostained infiltrating cells were counted within the 200 μm of stroma immediately adjacent to the basement membrane of the epithelium. They were counted under light microscopy with an eyepiece graticule used in conjunction with a high-power (× 40) objective lens. The median number of these stroma subepithelial immune cells was determined for each biopsy and all cell counts were expressed as the number of cells/mm², as described previously [9]. Intraepithelial Langerhans cells were identified as cells possessing at least two dendrites attached to a cell body [18].

In situ hybridization

Frozen sections of cervical and tonsillar tissues mounted on vectabond-coated slides (Vector Labs), after air drying for 30 min, were fixed in 4% paraformaldehyde in PBS pH 7.4. Before hybridization, the sections were rewashed in PBS, dehydrated through an ascending series of ethanol and allowed to air dry.

A 738-bp EcoRI fragment of human TNF-α cDNA clone 142–4, corresponding to nucleotides 337–1070 of TNF-α mRNA, was subcloned into the EcoRI site of pBluescript SK + plasmid (Stratagene, San Diego, CA). An antisense strand RNA probe for TNF-α transcripts was synthesized from NotI-linearized plasmid, using T7 RNA polymerase. A control sense strand RNA probe was synthesized from XhoI-linearized plasmid, using T3 RNA polymerase.

A 410-bp XhoI fragment of human IL-10 cDNA clone H15C, corresponding to nucleotides 0–410 of IL-10 mRNA, was subcloned into the XhoI site of pBluescript II KS + plasmid (Stratagene). An antisense strand RNA probe was synthesized from XhoI-linearized plasmid, using T7 RNA polymerase. A control sense strand RNA probe was synthesized from HindIII-linearized plasmid, using T3 RNA polymerase. All probes were labelled with the non-isotopic hapten, digoxigenin (Boehringer Mannheim, Lewis, UK). Finally, probes were resuspended in 50 μl of diethylpyrocarbonate-treated water containing 1 μl of RNase inhibitor (500 μg/ml) and stored at −20°C. The method for in situ hybridization has been described previously [19].

Statistical analysis

Differences between the three histological groups of biopsies (normal, low- and high-grade) for HLA-DR, HLA-DQ and CD54 (ICAM-1) expression by keratinocytes were analysed with the non-parametric Wilcoxon rank sum test. To compare the number of stromal immune cells in relation to keratinocyte expression of HLA-DR and CD54, the same test was used. The expression of HLA-DR, CD54 and the cytokines (IL-10 and TNF-α) by keratinocytes in CIN lesions was evaluated with the Fisher's exact test. Finally, the correlation between the number of subepithelial immune cells expressing both MHC class II and costimulatory molecules was investigated with the non-parametric Spearman rank correlation. In all these statistical tests P values are two-sided.

RESULTS

CD1a⁺ cells in the cervical epithelium do not express costimulatory or adhesion molecules

Intraepithelial Langerhans cells were identified with a specific marker, CD1a, and they were present in all tissues studied (n = 53). This cell population was invariably positive for HLA-DR, occasionally for HLA-DQ. However, as shown for one example in (Fig. 1), in no sample did these cells express any of the adhesion/costimulatory molecules CD11a/18, CD50, CD54, CD58, or CD86. They also failed to express either TNF-α or IL-10 (not shown). CD1a⁺ cells were also occasionally observed in the stromal tissue underlying the cervical epithelium. In one sample with normal epithelium (n = 12) and in five out of 41 CIN lesions, in which CD1a⁺ stromal dendritic cells were detected, they expressed all the above mentioned adhesion molecules, as observed in serial cryostat sections (not shown).

Fig 1.

Langerhans cells do not express costimulatory molecules within the epithelium. Serial cryostat sections of a low-grade intraepithelial squamous lesion of the cervix were stained for CD1a (A) HLA-DQ (Ia3) (B), intercellular adhesion molecule-1 (ICAM-1) (B-H19) (C), LFA-1 (AZN-L20) (D), ICAM-3 (KS128) (E), and LFA-3 (AICD58.9) (F). Note that Langerhans cells stain for CD1a but are negative for all the other markers tested, while mucosal infiltrating immune cells express adhesion molecules. In (F) keratinocytes are positively stained for LFA-3. Samples were counterstained with Mayer's Haematoxylin. (Mag. × 40.)

Expression of MHC class II antigens and costimulatory molecules by cervical keratinocytes

The expression of both HLA-DR and HLA-DQ in normal cervix and CIN lesions has been documented in detail previously [20]. The expression of CD54 (ICAM-1) was similar to that of HLA-DR, in that it was absent in normal cervix, detected in 4/23 low-grade lesions and 9/18 high-grade lesions. The co-ordinate expression of HLA-DR and CD54 was apparent both in terms of which lesions expressed the molecules and the pattern of expression. Thus, the epithelial cells from 22 specimens out of a total of 41 were negative for both HLA-DR and CD54 expression, and up-regulation of both molecules was detected in 13 samples (Fisher's exact test, P < 0.0001). In each case, staining of serial sections for HLA-DR and CD54 showed co-localization of the expression of these molecules by keratinocytes.

Increased expression with increased disease severity was also observed for the costimulatory molecule CD58 (LFA-3). CD58 was constitutively expressed by basal or basal and parabasal epithelial cells in normal epithelium, but this expression increased in extent in parallel with the severity of the cervical intraepithelial lesion (Fig. 1F).

Expression of both CD86 (B7-2) and CD50 (ICAM-3) by epithelial cells in normal (n = 12) or premalignant cervical squamous epithelium (n = 41) was never observed, although the antibodies to both of these molecules showed strong reactivity in tonsil (not shown).

Expression of TNF-α and IL-10 in cervical epithelium

In normal squamous epithelium (n = 12), TNF-α protein was always present in basal or basal and parabasal epithelial cells. In contrast, IL-10 expression was absent (Fig. 2 and Table 2). In 23 biopsies with low-grade lesions, IL-10 expression by keratinocytes was detected in 12 specimens (52%), and in contrast, TNF-α expression was lost in three samples (13%). Finally, in 18 biopsies with high-grade disease, whereas IL-10 continued to be up-regulated in eight cases (44%), TNF-α expression by epithelial cells was further down-regulated and the cytokine was undetectable in six out of 18 specimens (33%) (Table 2). No correlation between the expression of IL-10 and TNF-α was found, which suggests that they are independently regulated. Both molecules were found in 17 samples, both were absent in six, but in 14 TNF-α was present and IL-10 was not, and in two the opposite occurred (Fisher's exact test, P = 0.2). There was no statistically significant correlation between the expression of TNF-α and HLA-DR (P = 0.1) or CD54 (P = 1.0) by the epithelial cells. Likewise, no correlation was observed for the expression of IL-10 and HLA-DR (P = 1.0) or CD54 (P = 1.0) by cervical keratinocytes.

Fig 2.

The expression of TNF-α and IL-10 in normal epithelium and cervical intraepithelial neoplasia (CIN) lesions. Three consecutive frozen sections of cervical squamous epithelium from each sample were stained with antibodies to TNF-α (left panel), IL-10 (right panel) and an isotype matched control (not shown). (A,B) One example of normal ectocervix. Basal epithelial cells are positively stained for TNF-α, and no IL-10 protein is observed. (C–H) Three examples of low-grade lesions. TNF-α expression was quite variable, being sometimes undetectable (E), and sometimes full-thickness (G). IL-10 was up-regulated in some CIN lesions (F,H). In (G,H) the full thickness of the dysplastic epithelium shows strong cytoplasmic staining for both cytokines. Samples were counterstained with Mayer's Haematoxylin. (Mag. × 40.)

Table 2.

Immunohistochemical analysis of cytokine expression in normal squamous epithelium and premalignant cervical lesions according to the epithelial level involved

Figure 2 shows characteristic examples of TNF-α and IL-10 protein expression in normal and premalignant cervical squamous epithelium. Both cytokines were distributed diffusely throughout the cytoplasm of the epithelial cells and keratinocyte staining for TNF-α was always more intense than for IL-10. The distribution of mRNA for both cytokines, as documented by in situ hybridization (Fig. 3), paralleled that of protein, with high levels found in basal and parabasal cells.

Fig 3.

The distribution of TNF-α and IL-10 message in cervical intraepithelial neoplasia (CIN) lesions. Consecutive sections of cervical epithelium were stained with antisense (A,C) or sense (B,D) probes for TNF-α or IL-10. (A,B) Staining of a low-grade CIN lesion with IL-10 probes. (C,D) Staining of a high-grade CIN lesion with TNF-α probes. (Mag. × 40.)

Expression of MHC class II antigens and costimulatory molecules by immune cells in the subepithelial stroma

Table 3 summarizes the expression of HLA-DR and -DQ, CD54 (ICAM-1), CD50 (ICAM-3) and their counter-receptor CD11a/18 (LFA-1), and CD58 (LFA-3) by subepithelial immune cells in normal squamous epithelium and CIN lesions. It is apparent that there was a steady increase in the number of infiltrating cells expressing all the above mentioned molecules in relation to the severity of HPV-related cervical disease. Figure 1 depicts some representative examples of the expression of costimulatory molecules by mucosa-infiltrating immune cells in premalignant cervical epithelium. In the CIN group, using the Wilcoxon rank sum test, the number of stromal cells expressing MHC class II or adhesion molecules was not correlated to the expression of TNF-α or IL-10. This result suggests that production of both cytokines by cervical keratinocytes is not induced by local infiltrating immune cells.

Table 3.

Median number of cells (per mm²) expressing MHC class II antigens and costimulatory molecules in cervical subepithelial stroma

DISCUSSION

The results of this study paint an equivocal picture of immune activation in CIN lesions. On the one hand, the increased expression of HLA-DR and CD54 by keratinocytes, and the accumulation of leucocytes with an activated phenotype immediately below the epithelial lesions, all point to the existence of a CIN-related immune response. In contrast, the absence of mature APC, the absence of CD50 or CD86 on any cell type within the lesion, and the down-regulation in TNF-α argue against the development of a full-blown T cell-dependent cellular inflammatory response.

Langerhans cells expressing CD1a (a specific marker commonly used for their identification) and HLA-DR were found in all the cervical samples tested, although we [20] and others [21] have noted a progressive decrease in number with increasing severity of neoplasia. Furthermore, this study shows that in normal and premalignant cervical squamous epithelium Langerhans cells in situ do not express or express extremely low levels (undetectable by immunohistochemistry) of CD11a/18, CD50, CD54, CD58 and CD86. Viac et al. [22] have also not found expression of CD54 by Langerhans cells in genital warts. Human epidermal (cutaneous) Langerhans cells have been shown to express high levels of CD50 [23], but not the other adhesion molecules [24]. In contrast to those APC present within the epithelium, subepithelial CD1a⁺ cells, known as dermal dendritic cells, do express all the above mentioned adhesion molecules [25], as we have observed in five out of 41 CIN lesions. This supports the widely accepted view that Langerhans cells in situ (immature ‘tissue dendritic cells’) capture antigens and are very efficient in their processing but relatively poor presenters. Up-regulation of expression of costimulation molecules occurs after migration to the T cell area of lymphoid tissue [25, 26]. In contrast to the HPV-related pattern of migration and activation, persistent immune responses, e.g. associated with chronic infection, are often associated with increased numbers of Langerhans cells with an activated or mature phenotype [6, 7]. The phenotype of the Langerhans cells within CIN lesions is therefore suggestive of only limited immune activation within these lesions.

The co-ordinate expression of HLA-DR and CD54 (ICAM-1) on keratinocytes, and the parallel up-regulation of CD58, is consistent with a role for these cells, rather than the much rarer Langerhans cells, in antigen presentation to T cells within the cervix. However, what might be the functional outcome of the presentation event in vivo is difficult to predict. Although keratinocytes can activate antigen-specific effector T cells [27], they can also induce tolerance [28]. The absence of CD86, however, is probably more likely to lead to tolerance induction than activation of antigen-specific effector T cells.

Finally, we report for the first time the expression of both TNF-α and IL-10 protein and message in normal and premalignant cervical epithelium. Both of these cytokines are known to play a key role in regulating APC function. TNF-α is constitutively produced by the basal keratinocytes of normal cervical squamous epithelium. In contrast, IL-10 expression in normal cervical epithelium was never observed. Differences in expression of both cytokines by cervical keratinocytes in comparison with epidermal ones are apparent. In normal human epidermis (keratinized squamous epithelium, in contrast to non-keratinizing cervical squamous epithelium) TNF-α protein is localized to the upper level keratinocytes [15]. Immunohistochemically, IL-10 has also been identified throughout all levels of epidermis but with accentuation in upper level keratinocytes [29]. A comparison of cytokine staining between normal and CIN lesions suggests that progression of CIN is associated with some changes in the pattern of cytokine expression (though it should be noted that controls were not age-matched, and the independent effect of age variation can not therefore be assessed). In CIN lesions, the absence of correlation between the expression of TNF-α and IL-10 observed argues in favour of an independent regulation of these cytokines. Either down-regulation of TNF-α or IL-10 expression could result in failure to activate an effective T cell response within the microenvironment of the cervix.

The overall conclusion from this study is that premalignant cervical lesions are sites of only partial, and selective immune activation. The up-regulation in HLA-DR, CD54 and CD58 are consistent with the idea that a CIN-related immune response is initiated within the cervical microenvironment. However, the down-regulation of TNF-α and the increased production of IL-10 by the epithelial cells, and the complete absence of CD50 and CD86 expression, could contribute to persistence of infection, and hence, cellular transformation in some HPV-related cervical lesions. Strategies aimed at boosting the intrinsic immune response to HPV may therefore prove effective both for prophylaxis and therapy of HPV-related neoplasia.