Joint loss of PAX2 and PTEN expression in endometrial precancers and cancer

Abstract

Latent endometrial carcinoma precancers are normal appearing endometrial glands with sporadic loss of tumor suppressor gene function such as PTEN. Progression to carcinoma is inefficient and requires additional genetic damage that creates a histologic precursor lesion called endometrial intraepithelial neoplasia (EIN). In this study, we examined loss of PAX2 expression, a gene required for embryonic uterine development, during endometrial carcinogenesis. Normal proliferative, EIN, and malignant (endometrial adenocarcinoma) endometrial tissues were immunostained for PTEN and PAX2. Proliferative samples with loss of protein in at least one gland were scored as latent precancers. EIN and cancer lesions were scored by the majority pattern. Overall prevalence and topography of joint PAX2-PTEN expression loss was examined. The prevalence of PAX2 protein loss in the sequence of normal to precancer to cancer was 36%, 71%, and 77% respectively, and for PTEN 49%, 44%, and 68%. Normal endometrial prevalence of PAX2 or PTEN deficient latent precancers was unaffected by biopsy indication, but increased significantly with age. Coincident loss of PAX2 and PTEN expression in an individual normal endometrium was seen in 21% of patients, but usually involved different glands. Coincident loss was more common in precancers (31%) and carcinoma (55%), in which case both markers were protein null in an overlapping clonal distribution. PAX2 and PTEN protein loss occur independently and accumulate with increasing age in latent precancers of normal premenopausal endometrium. Loss of function of both genes in an overlapping distribution characterizes clinical emergence of a premalignant lesion which is carried forward to carcinoma.

INTRODUCTION

The most common endometrial malignancy, endometrioid type endometrial adenocarcinoma, is often preceded 3–4 years earlier by a monoclonal genetically mutated precursor called endometrial intraepithelial neoplasia (EIN)(1). These have a distinctive histopathologic appearance which allows them to be diagnosed by pathologists(2), but carcinogenesis begins long before any specific lesion such as EIN develops. Following cessation of known endometrial cancer risk increasing (unopposed estrogens) and risk reducing (such as use of hormonally inert intrauterine devices(3) or administration of progestin-containing oral contraceptives(4, 5)) exposures the endometrium usually reverts to normal histology within a few menstrual cycles, but the interval of altered risk can last for years(6). This has led to the prediction that there are stable long term alterations within the histologically unremarkable endometrium which occur in response to nongenetic risk modifiers. Biomarkers which are informative in disclosing the earliest stages of disease offer the possibility of directly observing these events in normal tissues long before clinical detection.

Inactivation of the tumor suppressor gene PTEN has been associated with development of endometrial carcinoma in mouse knockout (7), and human observational studies(8), a role entirely consistent with its anti-tumorigenic role as a mediator of cell division and enabler of apoptosis (9). More surprising was the discovery that very small burdens of PTEN protein deficient glands caused by somatic PTEN gene mutation and/or deletion are frequently found in the endometria of otherwise normal cycling premenopausal women(10). These may be retained for years, through many menstrual cycles, during which mutant cells are continuously exposed to changing systemic hormonal conditions(11). Affected glands are undetectable by routine exam. These occult lesions are appropriately designated “latent precancers”, to indicate their hidden nature and requirement for additional hits before they can be recognized clinically.

PTEN mutation alone, however, is insufficient to cause endometrial cancer. The 35% latent precancer rate defined by loss of PTEN function is counterbalanced by a lifetime endometrial cancer risk of only 2.5%(12). Emergence from latency to overt clinical disease (whether EIN or carcinoma) is so inefficient as to be an uncommon event, but one known to occur (11), with the accumulation of additional genetic damage. Risk increasing exposures such as estrogens unopposed by progestins have been proposed as positive selectors for conversion of pre-existing latent precancers to EIN or carcinoma, perhaps acting indirectly through an increase in glandular proliferative or mutation rates(13). Risk reduction below that of the general population appears to be effected in part through hormonal or non-hormonal exposures such as oral contraceptive or (non-hormonally impregnated) intrauterine device use that serve as negative selectors of pre-existing latent precancers(13, 14).

In this paper we show an association between loss of expression of the paired-box containing gene, PAX2, and endometrial cancer. Embryonic Pax-2 expression is required for development of the kidneys and ureters, the uterus and oviducts in females, and the vas deferens and epididymis in males(15). Persistent endometrial gland expression of Pax-2 in the adult (16, 17) is unlike the embryonic-only expression seen in other tissues derived from the intermediate mesoderm, likely reflecting an important function in endometrial proliferation and self-renewal.

There are some data that loss of constitutive PAX2 expression correlates with endometrial and cervical(18) malignant transformation. Quantitative RNA expression studies of PAX2 in human endometrial tissues shows high levels of expression in benign proliferative endometrium with 2-fold reduction upon tamoxifen therapy, and 5-fold reduction in the cancers(19). This presents the unique possibility that in the proliferating and self-renewing endometrial epithelial cells, PAX2 acts as a tumor suppressor. In this study we further examine the idea that loss of PAX2 function occurs in endometrial carcinogenesis, and relate changes in PAX2 to those which occur in parallel with the endometrial tumor suppressor gene, PTEN.

MATERIALS ND METHODS

Case Selection

Pathology reports at Brigham and Women’s Hospital were screened, with IRB approval, for a diagnosis of proliferative endometrium, EIN, or endometrial adenocarcinoma, and associated endometrial biopsies or curettage specimens retrieved as paraffin tissue blocks from the diagnostic archive. All diagnoses were confirmed by unblinded slide re-review (by GLM), and cases rejected if the expected diagnosis was not confirmed. Additional case details are described below. Each patient contributed only a single sample to the study, which were included in only one diagnostic group.

Malignant endometrial tissues received between July 1 and December 31, 2008 were retrieved based upon a report diagnosis of “endometrial adenocarcinoma.” Of the 112 candidate cases 36 were excluded because of unavailability of slides or blocks, 3 because of fragmentation artifact, 6 because of prior therapy, 2 due to lack of carcinoma in available slides, and 3 were immunohistochemistry failures. This left 62 sequential untreated endometrial cancers for which we report both PAX2 and PTEN immunohistochemical results.

Premalignant endometrial tissues received from August 1, 2006 and July 31, 2008 were retrieved by a pathology reports diagnosis of “Endometrial Intraepithelial Neoplasia.” Of the 127 candidate cases, we excluded the following: 23 due to concurrent adenocarcinoma, 12 because they were repeat biopsies in patients already represented (the first was retained as eligible), 19 due to unavailable slides or blocks, 1 due to cautery artifact, 9 due to insufficient material (minimum tissue requirement was defined by tissue area of approximately 0.5cm² of tissue, mostly functionalis, in the slide), and 7 due to absence of lesion in the slides available, and 4 due to immunohistochemistry failures. This resulted in successful PAX2 and PTEN immunohistochemistry results from 52 EIN bearing biopsies.

Normal Proliferative endometria from premenopausal (age < 50 years) women received between July 1 and December 31, 2008 were retrieved based upon a report diagnosis of “proliferative endometrium.” Of the 389 candidate cases we excluded: 3 because of missing slides, 102 because of co-existing pathologic endometrial conditions (59 with anovulation or co-existing neoplasm, 24 polyps, 5 with active gestation, 13 endometritis, and 1 IUD-bearing), 21 because of fragmentation or menstrual breakdown, 10 due to insufficient tissue, 10 because the patients were represented by an eligible prior biopsy, 4 because of immunohistochemistry failures, and 32 random cases unneeded to achieve desired study power. Patients with any of the following were then excluded: current history of sex steroid hormone use (n=6); previous known endometrial disease (n=7); or prior history of tamoxifen use (n=3). This left 191 proliferative endometrial biopsies which yielded both PAX2 and PTEN immunohistochemistry results.

Clinical Indications for biopsy within the normal proliferative group

Indications for biopsy were retrieved from the pathology requisition and classified into one of four general classes as follows: 1)extrinsic, endometrial biopsy performed as part of workup of known non-endometrial disease (ex. uterine fibroids, known non-endometrial pathology such as cervical disease); 2)intrinsic: endometrial biopsy performed because of known endometrial diagnosis documented prior to biopsy (ex. prior endometritis), or symptoms (bleeding) directly referable to the endometrium itself; 3)screen: endometrial sampling performed incidental to endometrial unrelated procedure (ex. tubal ligation), in reflex to a nonspecific screening test (endometrial cells on pap or thick endometrium on ultrasound), or in response to nonspecific symptoms or signs (infertility, pelvic pain); or 4)unknown: no indication for biopsy provided by the clinician.

PAX2 and PTEN Immunohistochemistry

One representative paraffin tissue block was obtained from each pathology specimen, and stained for PTEN (murine monoclonal antibody 6h2.1 from Dako, Carpinteria, California, Catalog number M362729-2, used at 1:100 dilution overnight primary antibody incubation at 4’C) and PAX2 (rabbit polyclonal antibody Z-RX2 from Invitrogen, Carlsbad, California, catalog number 71-6000, used at 1:300 dilution overnight primary antibody incubation at 4’C). In brief, paraffin sections were rehydrated and underwent microwave antigen retrieval before adding primary antibody overnight at 4^oC. Slides were washed, incubated with appropriate secondary biotinylated immunoglobulin (Vectastain ABC kit, Vector Laboratories, Inc., Burlingame, CA) and signal detected by sequential addition of avidin peroxidase and 3,3′-diaminobenzidine. Endometrial glandular epithelial staining of independent replicate experiments was scored on two separate occasions by reviewers (GLM or KW) blinded to the patient group. Typically, PTEN defective glands are sharply offset at high contrast from endometrial stroma(10), and PAX2 defective glands offset by residual or overrun background endometrial glands which serves as an internal positive control. Discordant interpretations were resolved by consensus review at a multiheaded microscope. All endometrial tissue fragments were examined, and scored as PTEN null when signal was absent in the nuclear and cytoplasmic compartments of all cells in at least one gland, and PAX2 null when signal was absent in the nuclear compartment of all cells in at least one gland.

PAX2 immunohistochemistry validation by Multiple PAX2 antibodies

The PAX2 polyclonal antibody used throughout this study, Z-RX2, had been produced by immunizing a rabbit with the c-terminal domain (aa188–385) of the murine Pax-2 protein, and employed by other laboratories to detect PAX2 protein in mice and humans(17, 20, 21). In order to further validate performance of this reagent within our study material, we compared PAX2 protein detection by Z-RX2 with results obtained in the same tissues using alternative PAX2 antibodies. Four representative cases each from the proliferative, EIN, and carcinoma groups were selected based upon having localized foci of PAX2 protein null glands. Serial sections of tissue were immunostained with: 1) crude Z-RX2 used as above, 2)column affinity purified Z-RX2 (designated PAX-AP, supplied by Greg Dressler, used at 1:1000 dilution); and, 3)antibody 928, another rabbit polyclonal antibody directed against the same peptide used to create Z-RX2 (Dressler Laboratory, used at 1:500 dilution). All selected cases of normal proliferative, EIN, and adenocarcinoma with PAX2 null glands detected by antibody ZRX2, showed the same pattern of staining, and identification of individual null and expressing glands when using different lots of crude polyclonal antibodies (ZRX2 vs 928), or purified IgG fraction (ZRX2 vs PAXAP) (Data not shown).

Affected endometrial gland quantitation in proliferative endometria

An estimate of the burden of PAX2 and PTEN protein null glands in normal proliferative endometria was obtained by comparing the number of defective to the total number of glands present. The total number of endometrial glands seen in an average specimen from this study was estimated in twenty one randomly selected proliferative endometria stained for pankeratin (anti-human pankeratin cocktail of AE1 and AE3 murine monoclonal antibody, Dako, Carpinteria, California, catalog number M3515, used at 1:100 primary antibody incubation for 2 hours at room temperature) to clearly accent gland contours. Keratin stained slides were scanned in an iScan virtual microscopy device (Bioimagene, Cupertino, CA) and the total number of endometrial glands in each slide counted using the counter function in Image Viewer (v1.6.0, Bioimagene, Cupertino, CA). Incidental endocervical and lower uterine segment glands were excluded by reference to a serial hematoxylin and eosin stained section.

The number of PAX2 and PTEN defective glands in affected proliferative endometria was counted in replicate by two observers from immunostained sections using a hand counter. Replicate counts were plotted as a linear regression and 90% confidence interval outliers identified. The average of the two observations was rounded to the nearest whole number for non-outlier cases. Outlier specimens underwent a third gland count, and the two closest measurements were then averaged.

Joint loss of PAX2 and PTEN expression

Specimens containing both PAX2 and PTEN protein defective glands were evaluated for overlap within a shared population of individual glands. For EIN and carcinoma specimens, histologic lesions occupying most or all of the specimen, side-by-side comparison of immunostain results of adjacent serial sections allowed scoring as overlapping or not. A modified approach was necessary with the normal proliferative endometria, both because the number of affected glands was very small, and localization required a fine degree of spatial resolution. Proliferative samples in which both PAX2 and PTEN deficient endometrial glands had been detected were re-examined and each tissue region containing protein null glands highlighted on the coverslip with an ink pen. The marked up glass slides for PTEN and PAX2 were then aligned to identify those specimens where both stain markups co-localized to the same tissue fragments. Slides with PAX2 and PTEN protein null glands within the same tissue fragments were then examined together and the number of individual glands deficient in both proteins enumerated with a hand counter.

Statistical Methods

Categorical comparisons were assessed using the Pearson chi-square test; categorical comparison across categories of an ordered variable were assessed using the Kruskal-Wallis test. Comparisons of quantitative variables were assessed using the Student t-test with a pooled estimate of variance when between two groups of cases, and with analysis of variance when among more than two groups. Pearson chi-squared tests were used to evaluate whether loss of PAX2 and PTEN proteins were significantly associated within each diagnostic group. The McNemar test was used to assess whether a case was more likely to have PTEN or PAX2 null glands when only one of the genes was protein deficient.

RESULTS

Subject ages varied significantly between the groups (ANOVA p<0.001) averaging 41.8 (SD 6.1) years for normal, 50.3 (SD 10.1) years for premalignant, and 60.5 (SD 12.3) years for malignant endometria.

Latent precancer prevalence in the normal proliferative endometria was not associated with the clinical indication for biopsy (Supplemental Table 2, Chi-Square p=0.146 for PAX2, and p=0.715 for PTEN), or sampling device employed (curet vs. biopsy) (Supplemental Table 2, Chi-Square p=0.282 for PAX2, and p=0.588 for PTEN). Clinical indications for endometrial sampling did not vary significantly (Supplemental Table 2, Chi Square p=0.373) between diagnostic groups of normal, premalignant, and malignant endometria. The sampling instrument (curet vs. biopsy device) employed was more frequently (55%) a curet in cancer patients, compared to 25% and 42% in normal and premalignant patients, respectively (Supplemental Table 2, Chi square p<0.001).

Joint loss of PAX2 and PTEN expression in EIN and cancer specimens occurred in a clonal distribution, involving most or all neoplastic glands (Figure 1). When protein products of both genes were lost, there was extensive or complete geographic overlap of PAX2 and PTEN loss across the lesional field in all (34/34) of the cancers and most (94%, 15/16) of the EIN lesions.

Figure 1. Overlapping PAX2 and PTEN protein loss in EIN and adenocarcinoma lesions.

Premalignant EIN (top row) and malignant carcinoma (bottom row) lesions are clonal neoplasms offset by background endometrial glands (separated by the dashed line). Routine stains (hematoxylin and eosin, “H&E,” left) of EIN and carcinoma show crowded areas of glands with altered cytology which on immunostaining of adjacent sections with PTEN (center) and PAX2 (right) are seen to be protein deficient. Examples of protein deficient glands are marked with “*”, and examples of protein expressing glands marked with “o”. In expressing tissues, stroma and glands stain for PTEN, whereas PAX2 expression is confined to glandular nuclei. Note overlap of loss of both markers within all glands of the lesion.

In proliferative endometrium, 21% of cases were both PAX2 and PTEN null, and an additional 43% were either PAX2 null or PTEN null (Table 2 and Supplemental Table 1). Proliferative cases which were null for only a single gene were more likely to be PTEN null, p=0.004. In EIN, 31% of cases were both PAX2 and PTEN null, and an additional 54% were either PAX2 null or PTEN null. EIN cases which were null for only a single gene were more likely to be PAX2 null, p=0.008. In the cancers, 55% were both PAX2 and PTEN null, and an additional 35% null for only one of the two genes. There were no significant differences in the prevalence of PAX2 and PTEN null cases among those with cancer, p=0.201.

Table 2.

Proportion of Endometrial Tissue Samples Showing Loss of PAX2 and PTEN Protein Expression, by Diagnosis.

	Normal Proliferative N=191	Intraepithelial neoplasia (EIN) N=52	Cancer N=62	Kruskal-Wallis P value1
PAX2 null	35.6%	71.2%	77.4%	p<0.0001
PTEN null	49.2%	44.2%	67.7%	p=0.064
Joint PAX2 and PTEN Null	20.9%	30.8%	54.8%	p<0.0001
Joint PAX2 and PTEN Expression	36.1%	15.4%	9.7%	p<0.0001
McNemar chi-square test of agreement of PAX2 and PTEN loss2	p=0.004	p=0.008	p=0.201
Pearson’s chi-square 3	p=0.048	p=0.822	p=0.335

¹Non-parametric test of equality of proportions across diagnosis groups

²McNemar Chi Square test measuring whether PAX2 compared to PTEN protein loss were equally frequent among cases with loss of a single protein within each diagnostic group

³Pearson chi-squared tests measuring whether PAX2 and PTEN loss were significantly associated in individual patients within each diagnostic group.

The histologic presentation of PAX2, and PTEN, deficient glands in proliferative endometrium was distinctive(Figure 2) in two regards, and contrasted greatly with the extensive or complete overlap of joint protein losses seen in EIN and cancers (Figure 1). First, the number of protein deficient glands in normal tissues was very small, only involving a few glands of the hundreds present in the average specimen (Table 1). A random sample of 21 normal endometria averaged 784 (SD 525, median 691, range 91-1,902) total glands per specimen, identified by a keratin stain that highlighted all glandular epithelium irrespective of PAX2 or PTEN status. The total number of glands per specimen did not vary significantly (t-test p=0.668) between sample format, comparing endometrial curettage to endometrial biopsy. Protein deficient glands were a minor component of those present in the normal proliferative endometria, with only 0.46% of all glands lacking PAX2, and 1.34% null for PTEN. Second, when PAX2 and PTEN protein loss did occur in the same normal endometrial specimen, these usually involved mutually exclusive subsets of glands (Figure 2). Only 15 individual glands (0.01% of all examined) had loss of both PAX2 and PTEN proteins amongst all 191 proliferative specimens, and these were distributed amongst only 7 different samples (Table 1).

Figure 2. Non-Overlapping PTEN (center left circle and left inset) and PAX2 (center right triangle and right inset) protein loss in normal proliferative endometrium of a single patient.

PTEN protein null glands are independent of those deficient for PAX2 protein. Adjacent serial sections show the same tissue areas stained with PTEN and PAX2. This case overall would be scored (as in Table 2 and Supplemental Table 1) as containing both PTEN and PAX2 null glands, but the illustrated fragment has no individual glands lacking both PTEN and PAX2 proteins.

Table 1. PAX2 and PTEN proteins are rarely lost in the same normal proliferative endometrial glands.

Proportion of glands with loss of PAX2, PTEN, or both, in 191 samples of normal proliferative endometria.

Proportion of glands with loss of protein expression
PAX2 Only	PTEN Only	PAX2 and PTEN
0.46% (694/149,744*)	1.34% (2,003/149,744*)	0.01% (15/149,744*)

^*Number of affected glands in all samples divided by estimated total number of glands in all samples

The prevalence of PAX2 and PTEN protein loss varied significantly within normal proliferative and within premalignant EIN tissues, but not within cancers(Table 2, McNemar test within each diagnostic group). One of our hypotheses was that of accumulation of increased genetic damage in the sequence of normal→premalignant→malignant tissues. Using a Kruskal-Wallis test (Table 2), we found this to be significant with PAX2 (p<0.0001) and a nonsignificant trend (p=0.064) with PTEN. Interestingly, the greatest stepwise change in the prevalence of loss of protein expression occurred at different junctures for the two genes: for PAX2 at the normal-premalignant transition (increase from 36 to 71%), and for PTEN at the EIN-cancer transition (increase from 44 to 68%).

The proportion of endometria containing both PAX2 and PTEN deficient glands increased in the normal→premalignant→malignant sequence, from 21% to 31% to 55% (Table 2 and Supplemental Table 1), respectively. This was paralleled by a decline in the proportion of patients free of both PAX2 and PTEN null glands, from 36% in normal to 15% in premalignant and only 10% in malignant endometria. 90% of carcinomas lacked protein from one or both of these genes.

There was a significant difference in age (Figure 3, ANOVA p=0.002) amongst women with normal proliferative endometrium grouped by latent precancer status, with an older age of those who had any combination of PAX2 and/or PTEN deficient glands. Women with a PAX2 latent precancer averaged 43.3 years (SD 4.5) compared to 41.0 years (SD 6.7) without (t-test p=0.007). Women with a PTEN latent precancer averaged 43.1 years (SD 5.0) compared to 40.5 years (SD 6.8) without (t-test p=0.003).

Figure 3. Somatically acquired sporadic loss of PAX2 and PTEN protein function is age dependent in normal proliferative endometrial tissues (ANOVA p=0.002).

Notched box plot of patient age at time of endometrial biopsy with the specified PAX2 and PTEN immunostain findings. Each solid circle is one patient, boxes indicate central 50% of values, notches are confidence intervals, and whiskers represent 1.5 times the interquartile range. “wt” = wild type, protein expressed in all glands. “null” = latent precancer with isolated protein deficient glands.

DISCUSSION

The clonal pattern of loss of PTEN protein in 68%, and PAX2 protein in 77%, of endometrial adenocarcinomas is compelling evidence for a functional role of these genes during endometrial carcinogenesis. PTEN is already well established as tumor suppressor gene(9) commonly inactivated by mutation and/or deletion in endometrial carcinoma, whereas the mechanism and significance of loss of PAX2 protein is less well characterized. Loss of PAX2 function in the majority (77%) of endometrial adenocarcinomas, in contrast to high abundance in normal proliferative fields, has now been shown at both the protein(our data) and RNA levels(19). The endometrium changes dynamically throughout the menstrual cycle, being driven primarily by estrogens in the follicular proliferative phase, and by progesterone in the secretory luteal phase. Although we have seen high levels of PAX2 expression in endometrial glands throughout the menstrual cycle (data not shown), we have chosen to include a large series of proliferative endometrium as a normal reference. Endometrial adenocarcinomas have a global expression profile closest to that of estrogen stimulated proliferative endometrium(22), making it a reasonable comparison. In addition to separate comparison groups of normal proliferative vs. neoplastic endometrium, the same pattern of clonal loss of PAX2 protein is seen between normal and neoplastic regions of endometrium within a single patient. Many of the tissue sections of EIN and adenocarcinoma contained flanking normal endometrial tissues, which when present always displayed high levels of PAX2 nuclear signal (Figure 1). Thus, loss of PAX2 protein in carcinomas and EIN specimens relative to normal tissues is not an artifact of differences in the hormonal environment between compared tissues. The experimental methods employed have been further cross validated. PAX2 specificity of the reagent systems we employed for immunohistochemistry was confirmed previously by others(23), was consistent between multiple antibody batches prepared in different animals, and corresponds to previous RNA data.

Our data showing frequent clonal loss of PAX2 protein in endometrial neoplasia, independent of PTEN, suggests it acts as a tumor suppressor in this tissue and undergoes inactivation during carcinogenesis. The functional impact of loss of PAX2 protein is context dependent, as PAX2 is a transcription factor that acts by modulation of other genes. For example, PAX2 expression trans activates endogenous expression of the WT1 tumor suppressor gene in urogenital tissues(24), and in breast cancers down regulates a poor prognostic indicator, the growth factor ERB-2(25). PAX2 expression increases upon cell division, a pattern shown previously for some tumor suppressor genes that act to control the rate of proliferation(26). At this time we do not know the mechanism of PAX2 protein loss in endometrial carcinogenesis, but it may be due to a primary PAX2 gene inactivating event. Loss of PAX2 protein on the basis of PAX2 gene mutation(27) or deletion(28) has been described in the hereditary renal-coloboma syndrome, and epigenetic silencing (29) is seen in melanomas. The incidence of endometrial carcinoma has not been studied in patients with renal-coloboma syndrome, in which PAX2 gene inactivation causes malformations of the kidneys and eyes(30).

The pattern of PAX2 protein loss across the range of tissues examined is similar to that of PTEN, supporting its application as a biomarker for latent precancers. Latent endometrial precancers identified through use of PTEN are common in normal tissues(10), contain mutations carried forward years later to adenocarcinoma(11), and involute when exposed to cancer protective factors such as oral contraceptive or intrauterine device use(13). This has led to the theory that some exposures modulate endometrial cancer risk by their effect upon latent precancers, and these can be detected directly within the target tissue with markers such as PTEN. This report presents PAX2 as an additional marker for this effect.

This study shows that the phenomenon of sporadic loss of gene function within normal endometrial tissues, including PTEN in 49% and PAX2 in 36% of normal cycling premenopausal endometria, is a process that may occur independently and in parallel with multiple genes. Combining these two markers in a single patient series, 64% of all women already bear small subpopulations of protein null endometrial glands before the menopause. Such a high frequency of occurrence challenges notions of normality, as in this study population the majority state is one in which normal tissues have already lost their genetic homogeneity through mutations and other stable gene inactivating events that are likely to occur in regenerative tissues. We do not believe this effect to be unique to our patient population, as it has been seen (using PTEN) in other institutions at comparable rates(14). Its consistent occurrence across different types of patients (Supplemental Table 2) suggests it is not a peculiarity of a specific diagnosis, symptom, or indication for physician visit. These normal tissues with sporadically acquired gene-specific defects are appropriately described as “latent precancers” to reflect both a very low efficiency of progression to carcinoma and a need for accumulation of additional genetic damage before they display any histopathologic alterations. These events are amongst the earliest ever demonstrated in a human tumorigenic pathway, and provide insights into preclinical stages of disease that previously have been inaccessible for lack of informative markers.

A striking feature of the current experiments is mutual exclusiveness of PTEN and PAX2 protein loss within individual endometrial glands in normal tissues. The burden of protein deficient glands is very low overall, averaging only 130 and 50 per ten thousand normal glands, respectively, for PTEN and PAX2. Almost all of these loss of expression events were independent, with only 15 glands amongst 191 normal premenopausal normal endometria (1 in 10,000 normal endometrial glands) displaying joint loss of both. A low frequency of joint PAX2 and PTEN protein loss caused by primary gene inactivation through mutation, deletion, or epigenetic silencing is expected, and that seen can be explained as a random convergence of independent events that occur individually at modest likelihood amongst the large number of glands in an endometrial field. Only 3.7% (7/191) of normal endometria contained PTEN and PAX2 double-null individual glands, a proportion close to the 2.5% lifetime risk of endometrial cancer(12). We propose that these marker-based observations in normal tissues are a histologic display of the multi-hit phases of epithelial carcinogenesis in which tumor risk increases with accumulation of defects in several genes within a continuous cell lineage.

In contrast to the normal proliferative endometrium, premalignant (EIN) and malignant (cancer) clones in which both PTEN and PAX2 are lost are predominantly overlapping, creating a double null state for all lesional cells.. This contrast allows some inferences regarding the biologic effects of individual, compared to combinatorial, loss of PAX2 and PTEN function. The most common phenotype associated with loss of only one gene product is an unremarkable normal appearing histology, and most common phenotype for loss of both gene products is EIN or carcinoma.

Separate loss of PAX2 and PTEN protein can occur at any time during carcinogenesis or tumor progression, as a first event within otherwise normal tissues or secondarily within an already neoplastic clone. Accumulation of PAX2 protein null glands is greatest in the normal-EIN transition, rising from 36% to 71% of all cases. Correspondingly, PTEN protein null endometria are equally frequent across normal and premalignant histologies, but undergo an increase from 44% to 68% across the EIN-cancer threshold. Loss of specific proteins at these differing stages may supply clues regarding their respective roles in the progressive events of carcinogenesis, from an initial phase of expansile monoclonal growth (normal-EIN transition) to acquisition of malignant behavior (EIN-cancer interface).

The notion of latent precancers, in which acquired sporadic loss of gene function in normal tissues represents early stages of carcinogenesis, began with data from a single gene, PTEN. Many of the model predictions, such as clonal continuity with cancer and reduced prevalence being associated with reduced cancer risk states have been experimentally confirmed with this single marker. This paper extends the model by showing that latent precancers may involve multiple genes, these frequently co-occur in individual tissues, and yield a phenotype that is highly dependent on the overlapping vs non-overlapping topography of affected glands. In the case of endometrium, PAX2 and PTEN seem to be inactivated independently, but when this does occur in the same glands promotes neoplastic transformation. This cross sectional study is not without its limitations, which include a sample size that does not enable adjustment for age differences, and lack of prospective followup of early stage lesions in individual patients over time. More work is needed to define a constructive role, if any, for latent precancer diagnosis in routine clinical practice, or use of PAX2 as a diagnostic tumor marker.