Cervical intraepithelial neoplastic (CIN) lesions are SOX2 positive, with increasing expression of this transcription factor upon progression from CIN1 to CIN3 8,10,11,19. In the underlying study we report on a discriminating. SOX2 expression pattern in CIN3 which can be clearly distinguished from the SOX2 expression patterns mainly seen in the lower grade preneoplastic lesions (for a schematic representation see Fig. 6). This particular pattern triggered us to more carefully study the molecular characteristics of the different epithelial layers in these high-grade premalignant lesions. On basis of the finding that the basal and parabasal compartments in a majority of CIN3 lesions show a normal genetic make-up, and the observation of direct morphological transitions between immature metaplasia and CIN3 with concomitant molecular switches, we propose three different models for the route of HPV infection, and the origin and progression of CIN lesions.
In the following paragraphs we will discuss different progression models for cervical carcinogenesis (Fig. 7A-D) that can explain the molecular make up of CIN3, taking into account the recent literature on the progression of CIN lesions 20–28.
Model 1: HPV infection of the (para)basal cell layers in normal squamous epithelium or mature metaplasia results in CIN1, which progresses to CIN2 and CIN3 (Fig. 7A, B).
The conventional model for the carcinogenic process in the uterine cervix, as proposed by amongst others Bosch et al20 and Woodman et al 21, suggests that HPV infection of the basal cells of the squamous epithelium or in mature squamous metaplasia initiates a cascade of events resulting in CIN1 (see also Herfs22) which in a minor fraction of patients progresses to invasive squamous cell carcinoma (SCC) via CIN2 and CIN323. Most of the CIN1 lesions have been described to originate from infected (para)basal cells in the ectocervix24. Our observation that the SOX2 positive basal cell compartments of normal and mature metaplastic epithelium (Fig. 7A) show a SOX2 expression pattern similar to that of CIN1 (Pattern 1), with subsequent increase of SOX2 positive cell layers with increasing severity of the lesion (Fig. 7B), indeed supports the suggestion that such a sequence of events can take place during progression of cervical preneoplasia. The observation that some of the CIN3 lesions show a SOX2 expression pattern similar to that of most CIN2 lesions suggests that at least this part of the high-grade lesions develops through progression from CIN2.
Model 2: HPV infection of the intermediate cell layers in immature metaplasia results in CIN3 (Fig. 7C).
A significant fraction (60%) of the CIN 3 lesions shows the typical SOX2 Pattern 3, often together with Pattern 2, typical for CIN2. Pattern 3 is characterized by a low or no expression of SOX2 in the (para)basal cell compartments and more extensive staining in the intermediate cell layers upwards. The fact that these (para)basal cell layers are SOX2 negative or weakly positive indicates that they may be part of immature metaplasia19. In the underlying study the basal compartments in a high frequency of CIN3 lesions were shown to exhibit a normal genetic make-up as defined by ploidy for chromosomes 1 and 3, and copy numbers for the SOX2 and SOX17 genes. This suggests that a HPV infection could have occurred in the intermediate cell layers in the normal or metaplastic epithelium in the transformation zone, resulting directly in CIN3, omitting the involvement of lower grade lesions as precursors.
This model is supported by reports in the literature that indicate that CIN3 lesions are rarely preceded by a CIN1 lesion 23, 25. The majority of metachronous CIN1 and CIN3 lesions, for example, were described to be caused by different HPV genotypes, indicating that a progressive biologic continuum from CIN1 via CIN3, leading finally to cervical cancer may be unlikely in at least part of the cases26. The SOX2 immunostaining results for Pattern 3 suggest a relatively rapid development of a high-grade lesion without a well-defined low-grade state. Indeed, Woodman et al21 have shown that detecting high-grade CIN was maximal 6 to 12 months after first detection of HPV16.
Model 3: HPV infection of the cuboidal squamocolumnar junction cells results in CIN3 (Fig. 7D).
HPV infection of a discrete population of cuboidal squamocolumnar junctional (SCJ) cells with a unique morphology and gene-expression profile, has been suggested by Herfs et al 22 to result in a premalignant squamous cervical lesion. The biomarker expression profile typical for these cells was also detected in a high percentage of high-grade CIN lesions.
Also, this model, suggesting a downward rather than an upward evolution from progenitor cell to the premalignancy, could explain the observation by several authors23, 25,26 that CIN3 can originate without previous precursor lesions. HPV infection of squamo-columnar junction (SCJ) cells was suggested to result in a trans-differentiation process with an outgrowth of subjacent squamous cells (so-called top-down differentiation) often leading to high grade lesions27. In the model as proposed in Fig. 7D the HPV infected SCJ cells overgrow the normal squamous epithelium and initiate a downward proliferation and differentiation to SOX2 negative (immature) squamous metaplasia, that further develops into CIN3, proliferating in an upward direction. This model cannot only explain the normal genetic make-up of the (para)basal cells, but also the aneuploidy and viral load detected in the intermediate cell layers.
Is the absence of SOX2 expression and a normal genetic make-up in the (para)basal cell layers an indication for regression of the CIN3 lesion?
As described by Doorbar et al29 lesion regression, when it does occur after HPV infection, is not associated with significant apoptosis or cell death. Animal model studies have shown that lesions are cleared by the replacement of actively infected cells with apparently normal basal cells that continue to divide29. These histologically normal cells can still contain viral genomes but without obvious viral gene expression. The characteristics of the (para)basal cells in Pattern 3 answer to a certain extent to these properties in that they are p16 positive, proliferating and show a low HPV load, but exhibit an apparently normal genomic composition.
If Pattern 3 is specific for a regressive process it could be expected that this SOX2 distribution pattern should be detected in the majority of CIN1 lesions. However, the cells targeted by HPV in these low-grade lesions are the basal cells shown to be SOX2 positive, while in the formation of higher grade lesions HPV infection occurs in SOX2 negative immature metaplasia.
We conclude that our data shed new light on the biological characteristics and dynamics in the development of premalignant cervical lesions, as well as on the initial viral infection of the target cells for HPV in the area of the squamo-columnar junction.