Abstract
The current study aimed to identify and validate an applicable immunohistochemistry panel including Ki-67, c-MYC, estrogen receptor-α (ER-α), and progesterone receptor isoforms A/B (PR-A/B) in correlation with clinicopathological parameters in patients affected by deep infiltrating endometriosis. Tissue microarrays were prepared from a cohort of 113 patients. Phenotypic profile of the panel molecules was evaluated in glands and stroma in parallel with microvessels and stroma density measurements. Principal component analysis was performed on 8 immunohistochemical variables, 2 histological variables, and 8 subgroups of clinical parameters. The immunohistochemical profiling showed consistent Ki-67 immunostaining in 17.9% of the samples and c-MYC in 83.1%, while intense ER-α immunoreactivity was detected in 84% of the samples and PR-A/B isoforms in 24.1% of them. The combination of clinical parameters and tissue phenotype allowed a stratification of endometriosis-affected patients. Such novel phenotypical and clinical correlation could be helpful in the future studies for a better stratification of the disease aiming at a personalized patient care.
Keywords
Introduction
Deep infiltrating endometriosis (DIE) is recognized as the most aggressive form of endometriosis, observed mostly as rectovaginal lesions that invade the uterosacral ligaments and the adjacent structures including bowel, ureters, and bladder. 1 Deep infiltrating endometriosis is heterogeneous in presentation and variable in its clinical evolution, suggesting subcategories of patients yet to be stratified. Endometriotic lesions with ability to infiltrate the bowel most frequently develop in the rectovaginal septum, 2 while such foci are rarely observed as a primary event in the sigmoid. 3 The prevalence of DIE has been reported to range between 5% and 12% of women affected by the endometriotic disease. 4 Currently, no molecular biomarkers are used in routine clinical practice to determine prognosis or support in the clinical decision-making. Among the main characteristics of endometriotic tissue lesions are an accelerated cell growth and an augmented resistance toward apoptosis. 5 –7 Therefore, number of cells expressing molecules involved in cell proliferation processes such as Ki-67 and c-MYC could delineate a degree of aggressiveness and invasiveness of these lesions. 6,8,9 In this respect, the cell proliferation-related c-MYC gene is known to be expressed in normal endometrium, while the other uterine structures are weakly immunoreactive. 10 Two histopathological parameters associated with the aggressiveness of the disease include the microvascular and stromal density. 11,12 At present, it is believed that estrogens regulate endometrial cell proliferation through binding to an atypical cis-element in the c-MYC promoter. 1,10 In endometriotic lesions, the levels of estrogens are 5 to 8 times higher than their levels in the serum, which may specifically regulate the expression of the c-MYC oncoprotein in these tissues. 13
Concerning the estrogen receptors expression itself, we have previously confirmed that the ovarian endometriotic lesions exhibit overall a decreased level of estrogen receptor-α (ER-α; decreased ∼10-fold at the messenger RNA [mRNA] level) and an increased level of estrogen receptor-β1 (ER-β1; increased ∼3-fold at the RNA level 14 ). Moreover, in spite of the elevated progesterone hormone levels in endometriotic tissues, progesterone receptors were found to be reduced at both mRNA (reduced ∼10-fold) and protein levels compared to normal endometrium. 14,15
In the present study, the expression of Ki-67, c-MYC oncoprotein, ER-α, and PR-A/B receptor isoforms in parallel to the microvessels and stromal density was evaluated by immunohistochemical (IHC) and histological analyses on tissue microarrays (TMAs) constructed from lesions of patients with DIE. Here, we show, through a multidimensional principal component analysis (PCA), that the level of expression of these markers correlates to certain specific clinical parameters. This procedure outlined subcategories of patients in whom the deregulated IHC profiles are linked to disease severity and progression, development of ovarian lesions (endometriomas), the intensity of endometriosis-associated inflammation, genetic susceptibility, tobacco consumption, and infertility. Such molecular tissue characterization could provide a scientific basis for a more personalized clinical management of endometriosis.
Material and Methods
Patients
The study cohort consisted of 113 patients affected with DIE in stage III and IV according to the revised American Fertility Society. 16 All the cases of DIE were surgically and pathological diagnosed. The patient’s tissue lesions were mainly located in the rectal wall and the rectovaginal septum. The medical/surgical procedures followed were approved by the institutional review board in accordance with the ethical standards established by our institution. According to the clinical records, all patients included in the cohort were treated with GnRH agonists at least 6 months before surgery. The age of the patients included in the cohort ranged between 21 and 47 years.
All the surgically obtained tissue samples were routinely fixed in buffered formalin, included in paraffin and serially sectioned for the diagnostic and the IHC analyses.
Tissue Microarrays
Five-micron sections stained with hematoxylin and eosin (H&E) were prepared initially from each tissue block to confirm the diagnosis and to identify representative areas containing endometriotic glands and stroma. Then, selected paraffin blocks from the cases were used for the TMAs construction. The pathologist identified the areas containing glands and stroma, and 2 tissue cores with a diameter of 2 mm were punched in duplicate from these areas. The core cylinders were then arrayed on a recipient paraffin block with a precision arraying instrument (Alphelys, 93-Plaisir, France) as described. 17,18 Five-micron-thick sections were then cut and placed on charged poly-lysine coated slides (Sigma, Fallavier, France).
Immunohistochemistry
The sections of TMAs were deparafinized, rehydrated in graded alcohols, and further processed for immunostaining. Sections prepared on slides were first submitted to antigen retrieval by incubation in citrate buffer (pH 6.0) for 25 minutes at 96°C, slides were subsequently incubated in 2% of normal goat serum for 30 minutes to block the nonspecific immunoreactivity. The primary antibody was layered over the section and incubated overnight at +4°C. Sections were then incubated with biotinylated antimouse immunoglobulins at appropriate dilution (LSAB2; Dako, Glostrup, Denmark) followed by incubation with avidin–biotin peroxidase complexes (Dako) for 30 minutes. The 3,3′-diaminobenzidine was used as chromogen and nuclear counterstaining was realized with Harris solution (RAL Diagnostics, Martillac, France).
Evaluation of the IHC Labeling
Two investigators analyzed the slides independently and estimated by consensus (less than 2 points of discrepancy on a scale 0-10), the distribution, and the percentage of endometriotic cells (glands and/or stroma) showing characteristic immunostaining. In addition, this immunostaining intensity in the lesions was evaluated as the mean of the 2 duplicates:
Strong intensity of staining: +++, or as an intensity arbitrary unit (au) 8-10, on a scale 0-10, with presence of 60% up to 90% of stained cells; medium intensity: ++, with an intensity au 5-7, with presence of 35%-50% of stained cells; weak staining intensity: +, with an intensity au 1-4, when 15% and 25% of stained cells were detected; undetectable immunostaining: −, or an intensity of 0.
Microvessel density quantification was performed as described previously 19 by counting the number of CD31 stained vessels per field in a given “endometriotic hot spot” area containing glands and stroma. The stromal density was quantified after immunostaining with CD10 antibody and counting the labeled cells within a microscopic area measuring 500 μm in diameter. 20
Antibodies
The IHC panel was composed of 4 relevant proteins for endometriosis, which were revealed with the following antibodies; rabbit antiestrogen receptor-α isoform (ER-α, clone 1D5; Dako), diluted at 1:50; monoclonal mouse antihuman progesterone receptor (clone PR 636), detecting PR-A and PR-B isoforms (Dako), diluted at 1:80; monoclonal mouse antihuman Ki-67 (clone MIB-1; Dako) diluted at 1:120; the monoclonal antihuman c-MYC antibody (9E10; Santa Cruz Biotechnology, Dallas, TX), diluted at 1:80; anti-CD31 antibody diluted at 1:100 (clone JC70A; Dako), and anti-CD10 diluted at 1:100 (clone 56C6; Dako). The dilutions for each antibody were made in phosphate buffered saline 1× without Ca++ and Mg++, supplemented with 3% of BSA (Sigma, Fallavier, France).
Principal Component Analysis
The PCA was performed using the Statist’XL Addin–of the Excel Spreadsheet (http://www.statistixl.com/). The initial clinical database contained more than 70 items, collected either in face-to-face interviews conducted by the surgeon during the month preceding the surgery or during surgery. 21 Collected data include demographic data, medical and surgical history, type and duration of symptoms, ethnicity, and lifestyle habits such as smoking, pain scores, and associated symptoms—dysmenorrhea or gastrointestinal and lower urinary tract symptoms assessed using a visual analogue scale. 22 From these 70 items, we excluded general information, such as personal address or education levels, contraception- and pregnancy-related details (eg, date of first contraception, gravidity, or parity, etc) as well as items with missing values for >15% of the patients (eg, blood type, assisted reproduction details such as number of stimulations, etc); thus a total of 41 items was included in the analysis as presented in Table 1. We ordered the 41 items into 8 subgroups according to their pathophysiological significance: inflammation, disease severity, age, infertility, presence of endometrioma, body mass index, genetic predisposition, and tobacco consumption. We then analyzed 15 sensitive clinical parameters by PCA within these subgroups, belonging to the series of 41 items, to avoid giving equal importance to the less relevant clinical parameters (eg, distinction between digestive or urinary lesions, etc). For 37 items out of 41, less than 1% of the data are missing. For the following items, lymphocytes count and C-reactive protein level variables, 5% of data are missing while for Ca 125, and ferritin item levels, 15% are missing. When missing, the values were replaced by the mean value of the other data in order to minimize any influence on the trends of the data, on any given axis. Off note, there is no missing value for the following categorical variables: age > 35, infertility, and endometrioma. Principal component analysis allowed the conversion of the 10 tissue markers and the 15 clinical parameters to 25 eigenvectors, classified according to decreasing eigenvalues. The first 5 eigenvectors, with eigenvalues greater than 1.5, which explain 39.4% of the total variance of the initial data, were further examined in order to recognize the biological and medical signification of a given axis (Table 2). This allowed determination of the variable’s potential to classify the patients according to their disease status in relation to the IHC as well as to the targeted histological profile (microvessel and stromal density).
Extraction of the 8 Subgroups of Parameters From the 41 Relevant Clinical Items Available in the Patients’ Database.
Abbreviations: BMI, body mass index; CRP, C-reactive protein; rAFS, rate according to the revised American Fertility Society Classification; VAS, visual analogue scale; PCA, principal component analysis.
a0: No; 1: Yes.
b0: Fertile; 1: Primary infertility; 2: Secondary infertility.
Patients’ Clinicopathological Characteristics From the Cohort, Distributed Along the First Five Axes Identified by Principal Component Analysis.
Abbreviation: ER-α, estrogen-α; rAFS, rate according to the revised American Fertility Society Classification.
Results
Histopathological Data
The histological parameters of the microarrayed DIE lesions were scored at low microscopic magnification. The tissue lesions consisted mainly of nodular structures of different sizes containing a number of cystically changed endometrial-like glands surrounded by variable amounts of vessels and stromal components. At higher magnification, almost all of the lesions appeared embedded in a fibromuscular cellular component seldom infiltrated with numerous microvessels (Figure 1A). The stromal component surrounding the glands was of different thickness and appeared as either proliferative (Figure 1B) or inactive stroma without cellular atypia (not shown). There was an inverse direct correlation between the abundance of the stromal component and the size and the number of endometriotic glands. As a rule, the small glands were mostly embedded in abundant stroma, while the larger glands were surrounded by a thin stromal layer. The stromal component abundance was assessed by immunolabeling with anti-CD10 antibody. Thirty-one percent of the samples exhibited an intense staining (Figure 1D), whereas 33% and 36% of them showed medium or weak stromal density, respectively (not shown). The microvessel density quantification by CD31 immunostaining showed highly developed vascularization in 42% of the samples (Figure 1C), medium vessels density was present in 32%, while small vessel density was found in 26% of the samples.
Histological and immunohistochemical features of lesions on tissue microarrays (TMAs). Hematoxylin and eosin (H&E) stained TMAs showing representative endometriotic lesion with abundant microvessels network (A) and a large and dense stromal component surrounding the endometriotic glands (B). Immunostaining of the microvessels with anti-CD31 antibody recognizing the endothelial cells (C). Immunostaining of the stromal component with anti-CD10 antibody (D). Bar scale, 100 µm.
Immunohistochemistry Analysis
Ki-67
As assessed, the immunolabeling of Ki-67 was distributed in both nuclei and cytoplasm in endometriotic glandular epithelial cells and stroma. Noticeably, the pattern of Ki-67 staining showed different intensities and correlated with the thickness of the stromal layer surrounding the glands. Indeed, while the proliferating glands of smaller size surrounded by an abundant stroma were intensely immunolabeled, the larger glands surrounded by a thin rim of stroma were often weakly immunoreactive (Figure 2A and B). A strong nuclear and Ki-67 immunostaining of the endometriotic glands was observed in 17.9% of the cases (n = 28), moderate immunoreactivity was found in 27% (n = 43), and weak staining was present in 21.79% (n = 34) of the cases. The stromal cells showed nuclear and intense pattern of immunoreactivity in 51 of the cases, while a weak Ki-67 immune labeling was detected in the 105 analyzed cases.
Immunohistochemical localization of Ki-67, c-MYC, estrogen receptor-α (ER-α), and PR-A/B in tissue microarray (TMA) cores. Aspect of strong Ki-67 expression in a deep infiltrating endometriosis (DIE) sample containing medium-size proliferating glands surrounded by an abundant stromal layer (A) and weak Ki-67 expression in a sample with large endometriotic glands lined with flattened atrophic epithelial cells surrounded by a thin rim of stroma (B). Consistent c-MYC expression level in a sample showing various size proliferating glands (C), and sample with weak c-MYC expression in gland surrounded by rich stroma (D). Consistent ER-α expression in the small-size proliferating glands and in surrounding stromal component (E), and tissue area from a sample showing weak ER-α expression (F). Strong level of PR-A/B expression in glands and stromal layer (G). Tissue area from a sample with low level of expression of PR-A/B in glands and stroma (H). Bar scale, 100 µm.
c-MYC oncoprotein
Overall, the immunolabeling of endometriotic glands by the c-MYC antibody was mostly cytoplasmic, appearing as a perinuclear granular staining, while the stromal cells exhibited both cytoplasmic and a faint nuclear staining in low percentage of cells (Figure 2C and D). In some areas, the epithelial cells in a low number of proliferative glands exhibited both intense cytoplasmic and weak nuclear c-MYC staining. A strong cytoplasmic immunostaining of the endometriotic glands was observed in 83.2% of cases (n = 123), moderate labeling was found in 11.5% of cases (n = 17), 3.3% of analyzed cases (n = 5) showed weak, and 2% of the cases (n = 3) showed negative staining (not shown). The stromal cell cytoplasmic staining was strong in 49.3% of the cases (n = 73), moderate in 9.4% (n = 14), and negative in 1 case (n = 1).
Estrogen receptor-α
Informative ER-α immunoreactive core spots generally showed nuclear ER-α immunostaining of the endometriotic glands. Similarly to c-MYC and Ki-67, a correlation was observed between the gland size, the thickness of the stromal layer, and the staining intensity (Figure 2E and F). In 52.1% of the cases (n = 76), the glandular staining was strong (Figure 2E), it was moderate in 21.2% of cases (n = 31), whereas in 13.0% of cases (n = 19), the immunostaining was weak and absent in 13.7% (n = 20) of the cases. The stromal component was strongly immunoreactive for ER-α in 53.8% of the cases (n = 78), moderate in 24.1% (n = 35), weak in 6.2% (n = 9), and the protein was absent in 15.9% of the cases (n = 23).
Progesterone receptor isoforms-A/B
The analyzed core spots showed mainly nuclear immunostaining with variable intensity in both glandular and stromal cells (Figure 2G and H). Here again, the staining intensity of the PR isoforms was correlated with the size of the glands and with the thickness of the stromal layer. The PR-A/B expression was not detected in endometriotic glands in 69.6% of cases (n = 101) by the quantification procedure. Consistent nuclear staining was observed in 8.2% of cases (n = 12; Figure 2G), 11.7% of cases were moderately stained (n = 17), and 10.3% (n = 15) of the cases showed weak immunostaining. Distinct areas of the stromal compartment showed intense nuclear staining in 24.1% of cases (n = 35), moderate intensity in 48.2% (n = 70) of the cases, whereas 13.7% of cases (n = 20) were stained weakly or were not stained.
Association Between the IHC/Histological Panels and the Patient’s Clinical Characteristics Assessed by PCA
The analysis included a total of 18 variables encompassing 8 IHC variables, corresponding to separate specific labeling quantifications in glands and in stromal components for c-MYC, ER-α, PR-A/B, and Ki-67, 2 histological variables (microvessels and stromal component density), and 8 parameter categories related to the clinical data from the 113 patients of the cohort (Table 1). A total of 25 axes (eigenvectors) were extracted by PCA, which were sorted according to the percentage of explained variance within the individual axis, from the “heaviest” to the “lightest.” The analysis, in particular, was focused on the first 5 eigenvectors with eigenvalues >1.5 (Figure 3) and proportion of variance for each component with more than 5%, which represented 39.4% of the total variation, that is, 11.3%, 9.2%, 6.6%, 6.5%, and 5.8% for the axes 1 to 5, respectively (Table 2). Globally, the analysis strongly associates the ER-α and PR-A/B expression, both involved in the endocrine regulation, with clinical variables related to disease severity and presence of endometrioma as well as histological parameters including microvessels and stroma density in the first 2 axes. No IHC or histological variables are associated with the third axis, which mainly relates to inflammatory reaction intensity and environmental contribution. The levels of c-MYC and Ki-67 protein expression, respectively in stroma and glands, differentiate the fourth axis, while ER-α level in glands and infertility status are characteristics of the fifth axis.
Scree plot illustrating the 25 eigenvalues generated by the principal component analysis (PCA). The first 5 principal components, presenting an eigenvalue >1.5, were retained in the final analysis.
In addition, the patient’s clinical characteristics were projected on the corresponding axis according to their geometrical coordinates. The “10 upper” or high-coordinates group (HCG) of the axis and the “10 lower” or low-coordinate group patients (LCG) were used to further interpret the given axis. Overall, this analysis allows stratification of patients in subgroups by means of combinations of the panel of IHC/histological parameters and clinical variables as follows (Figure 4):
Correlation between the immunohistochemical/histological markers and the clinical parameters which delineate the 3 first axes by principal component analysis (PCA). Arrows corresponding to the PCA axes were disposed on the immunohistochemical/histological marker scales, ranging from 0 to 10 (A). The arrow is black in color when the tissue marker does not discriminate between the subgroups of patients in a given axis. c-MYC labeling on glands was always very strong and could not be used as a diagnostic tool. Colored arrows correspond to the high-coordinates group (HCG) or low-coordinates group (LCG) for each axis with its attributed color (B). Green arrows representing the HCG1 correlate a low labeling of c-MYC in the stromal component, of ER-α and PR-A/B in the glands and stroma, a high density of vessels and high quantity of stroma with endometrioma and disease severity but without genetic links. Red arrows representing LCG1 display the inverse profile (low severity, no endometrioma, but a strong genetic component). While yellow and blue arrows representing HCG2 and LCG2, respectively, do not show clear-cut difference in a given marker staining intensity, a low labeling of Ki-67 can be associated with tobacco consumption as shown by the violet arrow representing LCG3, the HCG3 being depicted by orange arrows. (The color version of this figure is available in the online version at http://rs.sagepub.com/.)
Axis 1
The HCG1 group exhibited a weak immunolabeling in glands and stroma for ER-α and PR A/B (Figures 2E-H and 4A and B, green arrows). But there is a high index for microvessel density and moderate stromal quantity, associated with several markers including disease severity, the presence of endometriomas, and weak genetic susceptibility. In contrast, LCG1 group is characterized by high ER-α and PR A/B expression levels and lower microvessel and stromal density. This axis also inversely linked the low stromal component density and proliferation rate, with the high intensity of IHC labeling for c-MYC, merely in stroma (Figure 4A and B, red arrows).
Axis 2
Likewise, this axis comprises parameters including those of disease severity, endometriomas, angiogenesis, proliferation kinetics, and endocrine regulation (Table 2). This axis dichotomizes the HCG2 subgroup, showing moderate PR-A/B glandular labeling (Figure 4A, yellow arrows), extensive inflammation, higher disease severity, and the presence of endometrioma, with the LCG2 subgroup exhibiting higher Ki-67 staining index, increased microvessels density, and genetic predisposition (Figure 4A and B, blue arrows).
Axis 3
This axis points out smoking status and systemic inflammation as the most significant parameters characterizing the HCG3 subgroup, associated with high c-MYC in the stroma (Figure 3A, orange arrows). Besides the LCG3 subgroup showed only moderate level of Ki-67 expression in the endometriotic stromal component (Figure 4A and B, violet arrows).
Axis 4
This axis delineates the HCG4 subgroup of patients with an increased stromal density associated with strong c-MYC staining in the stromal component and intense staining for Ki-67 in glands. This suggests greater tissue proliferation rates in this group in comparison to the lower proliferation rates in the LCG4 subgroup (Table 2).
Axis 5
Relates a HCG5 subgroup of patient with low glandular staining for ER-α and high infertility rates with an LCG5 subgroup showing a high level of Ki-67 expression in stroma and glands and greater disease severity (Table 2).
Discussion
In this study for the first time to our knowledge, we performed a cross-sectional analysis of a cohort of patients affected by DIE and evaluated the relationship between the IHC profile of the lesions and a total of 41 clinicopathological parameters distributed in 8 subgroups. The presented data show that combining the clinical parameters with the defined IHC panel allows the segregation of patients in distinct meaningful prognostic groups. Moreover, the markers used in the proposed panel appear also suitable to identify patients who might benefit from a specific hormonal therapy.
The studied IHC and histological markers included 2 cell proliferation associated molecules (Ki-67, c-MYC), steroid hormones receptors (ER-α and PR-A/B), and 2 histological markers, that is, microvessel and stromal density. The 4 IHC markers panel was chosen since previous studies ascertained their prognostic and therapeutic values in endometriosis. 23
Overall, our results showed elevated levels of Ki-67 and c-MYC proteins, and in general, a tendency of an increase in microvessels and stroma density, suggesting augmented cellular proliferation. In addition, a significant amount of ER-α protein in the endometriotic glands and stroma was observed in the majority of the samples, whereas the levels of PR-A/B receptors were drastically decreased.
The strong glandular immunostaining for nuclear Ki-67 and for cytoplasmic c-MYC oncoprotein was evidenced in 17.9% and 83.1% of the cases, respectively. This is in accordance with previous studies that showed higher level of Ki-67 and c-MYC expression in both glands and stroma in different forms of endometriotic lesions. 8,24 An increased level of expression of Ki-67 was also described in eutopic endometrium from patients with endometriosis. 9 Regarding the c-MYC expression, our findings indicate intense and diffuse cytoplasmic staining detected in the endometriotic glands with either proliferative or secretory-like appearance. The c-MYC oncoprotein expression in both nuclei and cytoplasm of glands in ovarian endometriomas and eutopic endometria has been previously reported in a small cohort of patients with endometriosis, 24,25 which is in agreement with our observations. The increased expression of nuclear c-MYC was also observed in endometrial hyperplasia and endometrial carcinoma compared to normal endometrium, suggesting that the different staining patterns of c-MYC, that is, nuclear or cytoplasmic, might correlate with the degree of invasiveness of the endometrial tumors. 26
Concerning the steroid hormone receptors expression, consistent level of ER-α expression in parallel with a drastic decrease of the level of PR-A/B isoforms in DIE was observed. While isolated tissue areas were consistently immunostained for ER-α in 52.3% of the cases, the immunoreactive PR-A/B isoforms were not detected in 69.6% of the samples. Although extensively studied, the precise expression levels of ER-α and ER-β1 in the different subtypes of endometriotic lesions are still not well characterized. 1,27 While some authors reported an increased relative ratio of ER-α and estrogen receptor-β (ER-β) mRNA in red peritoneal lesions, 28 other studies described reduced ER-α expression in the different forms of endometriotic lesions. 29 –31 In a previous study, we have reported that the mRNAs for progesterone receptor were diminished for more than 10-fold in ovarian endometrioma in comparison with the eutopic endometria. 14 In the same report, the estrogen receptor profile was more complex and indicated that the ER-α (estrogen receptor 1) mRNA level was reduced by more than 5-fold, while the ER-β (estrogen receptor 2) mRNA level in the lesions was increased by more than 2.5-fold. 14 This suggests that the levels of estrogen receptor mRNAs are not in accordance with the tissues protein levels and indicate transcriptional and/or translational differences between the different forms of endometriosis.
Clinically, while some studies reported no significant relationship between c-MYC as an histopathological factor and the disease progression, 32 –34 others showed that an increased c-MYC expression was associated with higher grade, advanced stage, and poor differentiation of endometrial gynecological tumors. 35,36 In addition, it was reported that the intensity of either nuclear or cytoplasmic c-MYC staining is an important factor in predicting survival of patients with endometrial carcinomas. 37
The decreased levels of PR-A/B receptors in endometriotic tissues were confirmed by several reports in relation to the very low response to hormonal treatments of women affected by the disease. 1,38 It was proposed that this phenomenon is related to the higher amount of the inhibitory isoform PR-A and the absence of the stimulatory isoform PR-B that by itself is a strong activator of the progesterone target genes such as glycodelin, stromal insulin-like growth factor-binding protein-1, and prolactin. 38,39 Alternatively, it was suggested that the lower levels of steroid hormone receptors in endometriotic lesions might be in relation with a yet unknown ovarian and tubal dysfunctions phenomena that could contribute to the development of the disease. 40 –42 To strengthen the fact that the IHC scoring methods have great potential for therapeutic and prognostic purposes, a correlation between the level of the selected endometriotic tissue markers and the clinical parameters was systematically performed by PCA. In fact, PCA is a powerful multivariate analysis method enabling to describe large data sets and to represent correlations expressed by 2-dimensional graph. Remarkably, this analytical strategy unveils a significant correlation between the DIE tissue phenotype and the 8 distinct clinicopathological markers. Each of 5 main axes was able to separate 2 groups of patients with features outlining different clinical courses of the DIE disease. It was then possible to draw several assumptions. In fact, the endometriosis-affected patients could be subcategorized through 5 main key words: the disease severity, presence of endometriomas, genetic susceptibility, inflammatory reaction intensity, and tobacco consumption. Although only 1 lesion per patient was deposited in the TMA arrays and despite the fact that all the lesions for one given patient may not show the same phenotype, our PCA strategy authorizes a clear stratification of the patients.
In view of improving the clinical management of the patients, our data provide a basis for using combination of tissue markers as a multidimensional tool for diagnostic and/or prediction of disease severity.
The analysis of axis 1 demonstrates that the group of patients HCG1 exhibits a low expression of ER-α and PR-A/B in addition to high microvessel and moderate stromal density. In addition, lower expression of c-MYC only in stroma was associated with more frequent occurrence of ovarian endometriomas, higher disease severity, whereas a genetic susceptibility was not prominent. These tissue and clinical features suggest that the patients of this subgroup suffer of a longstanding and/or progressive disease and are unresponsive to treatment with steroid hormones. Thus, in those patients, steroid hormonal treatment would not be as efficient as it could be for the individuals of the opposite LCG1 subgroup.
The second axis defines inversely patients with high glandular PR-A/B expression and strong inflammatory reaction associated with the presence of endometriomas but also without genetic susceptibility. Moreover, the endometriotic lesions in these women neither exhibit a high tissue proliferation rate nor a high microvessel density in the lesions. Altogether, this indicates that patients from this subgroup experience a dissimilar disease pathogenesis, with a more recent beginning, slow progression, and higher sensitivity to progesterone treatment.
The third axis unveils a correlation between tobacco consumption, systemic inflammation, and weak expression of Ki-67 in the stromal component. This subgroup of women shows, in particular, additional symptoms related to a smoking status. In these cases, efficient counseling to stop smoking habit may help to lessen some unnecessary sufferings.
The forth axis delineates a group of patients with high stromal proliferation rate but without particular clinical parameters orienting toward a more severe disease. Over again, our data contribute to outline a subcategory of patients that differs strongly from the opposite subgroup. A deeper analysis of these patients’ clinical features may unravel more precisely the disease pathogenesis.
Finally, the fifth axis delineates a subgroup of individuals with high-glandular ER-α expression, but low Ki-67 expression in the glandular and stromal component is associated with high infertility rates. Once more, the presented data support the assumption that this subgroup of patients could benefit from particular counseling and/or selective hormonal treatment.
One limitation of the current analysis is that we used an already available clinical database, which was not primarily designed for the current study. The cross-sectional data give a snap-shot aspect of the patients, but specific inclusion and exclusion criteria were not taken into account at the design stage before the IHC processing. Furthermore, these data have been collected in only 1 center, which may not allow results to be extrapolated to all women with DIE. However, this is the first time that such strategy (association of IHC tissue and clinical markers) is used to portray patients with severe endometriosis. To further refine this kind of analyses, a clinical database must be designed to be used during the collection of the endometriotic tissues in a multicenter prospective study.
In conclusion, using TMAs constructed from deep endometriotic lesions, it was possible to determine particular Ki-67, c-MYC, ER-α, and PR-A/B histo/phenotypic profiles associated with several major clinicopathological parameters. Two subgroups of patients were projected on each of the 5 different axes obtained by the PCA allowing patient’s stratification based upon both IHC and clinical parameters that included endometriotic tissue burden, extensive inflammation, fertility, pain as well as the existence of familial genetic component. Thus, this strategy could be proposed to serve as a useful tool for a more refined classification of the different forms of endometriosis by an analytical quantification of 4 essential IHC and 2 pathognomonic tissue markers.
Footnotes
Acknowledgements
The authors are grateful to Dr Alexandre Goguin for his help in the preparation of the manuscript.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Inserm institutional funding.