Patients undergoing laparoscopy for pelvic pain, investigation of infertility or tubal ligation were recruited at Monash Surgical Private Hospital and at Monash Medical Centre, Moorabbin and Cranbourne Campus, all in greater Melbourne, Australia. Ethical approval was obtained from appropriate institutional ethics committees and informed consent was obtained prior to surgery.

Control subjects were those with no laparoscopic evidence of endometriosis. Endometriosis patients were assessed according to the revised American Society of Reproductive Medicine classification system 29. Eutopic endometrium was obtained from both control and endometriosis patients by standard curettage. Ectopic endometriotic tissue was obtained via laparoscopic excision. Areas biopsied included peritoneum, uterosacral ligaments, and the ovary (endometriomata). Women on any hormonal preparations were excluded from the study. Patient details are provided in Table 1.

Eutopic endometrial tissue sections were designated to menstrual cycle stage by an experienced gynecological pathologist using the established Noyes endometrial dating criteria 30. For this study, selected tissue samples, derived only from women in the mid-secretory phase (POD 6–10) were assessed for chemokine expression. Endometrial curettings (eutopic tissue) and endometriotic biopsies were snap frozen immediately at -80°C in OCT embedding medium (Sakura Finetek USA Inc., Torrance CA, USA) and stored at -80°C until use.

Sectioning of frozen tissue samples and processing of sections were performed using methods described previously 31. Frozen tissue blocks were maintained at -20°C and sectioned at 10 μm, and sections adhered to sterile glass slides (uncharged and uncoated; Objektträger, HD Scientific, Melbourne, Australia) by gentle warming of the slide undersurface and then refreezing at -20°C. Sections were then fixed in cold (0°C) acetone for 1 min, stained with HistoGene (Arcturus Bioscience Inc., CA, USA) for 1 min and dehydrated in 95% ethanol (1 wash, 30 sec), 100% ethanol (three washes, 30 sec each) and xylene (two washes, 5 min each). Sections were then allowed to air dry at room temperature and used for laser capture microdissection.

Laser capture microdissection (LCM) was used for the isolation of epithelial glands from endometrial eutopic tissue samples. Frozen tissue sections from patients with endometriosis (n = 4, marked * in Table 1) and control subjects (n = 4) were used for LCM. The PALM^® MicroLaser Microdissection System (P.A.L.M. MicroLaser Technologies AG, Burnried, Germany) was utilized. Staining of frozen sections with HistoGene™ (Arcturus Bioscience Inc., CA, USA) enabled the precise selection of the glandular structures. The laser dissected glands were then ejected off the slide with a single defocused laser pulse and catapulted directly into the cap of a microfuge tube containing 10 μl droplet of TRIzol (Gibco BRL, Rockville, USA). For each patient, epithelial glands were laser-captured from approximately 20 tissue sections

Total RNA was isolated from each LCM sample using the TRIzol method according to manufacturer's instructions (Gibco BRL, Rockville, USA) following an initial incubation with glycogen (250 ng/μl) and TRIzol reagent for 30 min at room temperature to improve yields. RNA samples were stored at -80°C.

The RNA concentration for each sample was determined with the RiboGreen fluorescence RNA assay (Molecular Probes Europe BV, Leiden, The Netherlands) with an Escherichia coli ribosomal RNA preparation of known concentration as standard and RiboGreen reagent at a final dilution of 1:500. A standard curve of 0–80 ng/well in a 96-well flat-bottomed plate was employed. This assay gave an intra-assay coefficient of variance (CV), as defined by the repeated analysis of a single RNA sample, of 3.3% (n = 16) and an inter-assay variation of 10.3% (n = 14). To assess RNA quality and tissue specificity, RNA samples were reverse transcribed and primers specific for a mid-secretory phase endometrial gland-specific gene (HtrA3) used for PCR amplification as previously described 32.

Amplification of mRNA was performed to achieve the required amount of RNA for cDNA gene arrays. To obtain a representative molecular profile comparison of eutopic endometrium in each of control and endometriosis subject groups, 5 ng of LCM glandular epithelial cell-derived RNA from each of the 4 subjects in the group were pooled. T7 specific RNA amplification of mRNA was performed with the Arcturus HS RNA Amplification Kit (Arcturus) according to manufacturer's instructions. Final RNA yields following amplification, ranged from 0.5 μg to 1.5μg.

Gene profiling was conducted using the GeArray Q Series Human Chemokine and Receptor Gene Array (SuperArray Bioscience Corp., Bethesda, USA) consisting of 77 chemokine and chemokine receptor cDNA probes, printed in quadruplicate on a nylon membrane. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH), cyclophilin and β-actin were included as positive controls and PUC18 plasmid DNA as a negative control. The gene array experiments were performed as previously described 27.

For each array experiment, 100 ng amplified RNA was reverse transcribed to cDNA using Moloney murine leukemia virus-reverse transcriptase (MLTV-RTase) (Promega, Annandale, NSW, Australia) in the presence of biotinylated uridine 5-triphosphate (udTP) (Roche), and incubated overnight in hybridization buffer (SuperArray) containing 100 μg/ml denatured sheared salmon sperm DNA (Invitrogen Australia, Pty. Ltd., Mount Waverley, Vic, Australia) with pre-hybridized nylon arrays at 68°C. Post-hybridization washes with saline/sodium citrate/sodium dodecyl sulfate (SSC/SDS) were also conducted at 68°C. Positive cDNA binding was detected by application of a streptavidin-alkaline phosphatase conjugate in blocking solution and chemiluminescent substrate, CDP-Star as supplied with the arrays (SuperArray).

Membranes were exposed to X-ray film for a range of times between 2 sec and 5 min, to ensure quantitation during the linear phase of the reaction. Arrays for the two subject groups were conducted simultaneously, and the entire experiment was then repeated. X-ray films were scanned at high resolution and densitometrically analyzed using GelDoc software (Bio-Rad Laboratories, Regents Park, NSW, Australia). Chemokine signals were normalized for background signal intensity, to correct for exposure times. Gene expression for each array was also normalised to cyclophilin expression levels. Fold changes in gene expression of 2-fold or higher were considered as significantly different.

To confirm array data and to determine the variability between subjects, real time RT-PCR was performed on unamplified RNA derived from individual LCM samples from endometriosis and control groups (n = 3 per group: insufficient starting material was available for the remaining 2 samples). Total unamplified RNA (5 ng) was reverse transcribed using avian myeloblastosis virus-reverse transcriptase (AMV-RTase) (Promega, NSW, Australia) and 1 ng random hexanucleotide primers (Amersham Biosciences, NJ, USA), and the cDNA generated was subsequently amplified by PCR for specific primer pairs encoding CCL16, CCL21 and 18S RNA genes (for primer sequence refer to 27).

The Roche Light Cycler Real-time PCR system was used (Roche Diagnostics Australia Pty Ltd, NSW, Australia) where 4 μl cDNA (diluted 1 in 10) was added to a master mix including SYBR Green I, deoxynucleotide triphosphates, Taq polymerase enzyme, optimized concentrations of MgCl₂, and specific primers (0.5 pmol/μl; Sigma Genosys, Australia Pty. Ltd., NSW, Australia). An initial denaturing step was performed for 10 min at 95°C, before 40 cycles of 95°C for 15 sec, 55–66°C for 5 sec (annealing temperature specific to primer pair 27) and 72°C for 10 sec. All samples to be compared were included within the same run and the entire PCR experiment was performed in duplicate.

Expression of mRNA was quantitated by comparison with a 6-point standard curve of serially diluted (10-fold) cDNA standards specific to the gene product. The standard doses ranged between 2.5 ng/μl and 0.5 fg/μl. Fluorescence from incorporation of SYBR green into double-stranded PCR products was monitored continuously during cycling at the end of each elongation phase, and quantitation of mRNA expression was performed when amplified products were in the log-linear phase and parallel to the standards. A quality control (a single endometrial cDNA sample) was included in every run. At the end of each program, melting curve analysis was carried out to ensure specificity of the reaction products. The sizes of the products were confirmed by gel electrophoresis and DNA sequenced to confirm identity for selected samples. Data was normalized for the expression of 18S RNA.

For the further verification of gene array and real time RT-PCR data, frozen sections from 3 of the 4 same endometriosis patients and control subjects, along with tissue sections from an additional 8 normal and 7 endometriosis subjects, were used for the immunohistochemical comparison of chemokine (CCL16 and CCL21) protein expression (n = 12 and 10 respectively per group). Glands were not found in the frozen endometriotic lesion of one endometriosis patient who had been included in the array study and this subject was thus excluded from this part of the study. Affinity-purified polyclonal antibodies raised against human CCL21 and CCL16 peptides (Santa Cruz Biotechnology, Santa Cruz, CA, USA) were utilized for immunohistochemistry as previously described 27. Positive controls (normal eutopic endometrium) previously shown to have positive staining for CCL16 and CCL21, were included in every run. Negative controls were sections of each tissue included on the same slide but with non-immune IgG at the same concentration substituted for the primary antibodies. For immunostaining, 10 μm sections were post-fixed in 10% buffered formalin for 10 min, rehydrated and exposed to microwave antigen retrieval for 5 min before cooling. Primary antibodies were applied overnight (17 ± 1 h) at 4°C, diluted to 4 μg/ml in non-immune block containing 10% horse serum (Sigma-Aldrich, Sydney, Australia) and 2% human serum (in-house) in TBS with 0.1% Tween 20.

Detection of positive binding was performed by the sequential application of biotinylated horse anti-goat IgG (1:200 in nonimmune block; Vector Laboratories, Burlingame, CA, USA) and avidin-biotin-peroxidase conjugate (Dako, Glostrup, Denmark), followed by the substrate diaminobenzidine (Dako) for between 2 and 10 min. Sections were counterstained with Harris' hematoxylin (Sigma), dehydrated and mounted from Histosol with DPX mounting medium (BDH Laboratory Supplies, Poole, UK). Immunostaining was assessed semi-quantitatively by two independent assessors, blind to the identity of subject groups. Blinding was not possible in the assessment of ectopic tissue potentially biasing the results. Staining intensity and heterogeneity for each endometrial compartment (epithelium, stroma, leukocytes and vasculature) was assessed and allocated a score between 0 and 4 where 0 = no stain, 1 = faint staining, 2 = moderate staining, 3 = strong staining and 4 = maximally intense staining.

None of the data were normally distributed and therefore non-parametric tests were used. The direct comparison of two samples from different patients was conducted with the Mann-Whitney U-test. The direct comparison of two samples (eutopic and ectopic endometrium) from the same patient was conducted with the Wilcoxon Matched-Pairs Signed-Ranks Test. For the relationship between two independent interval variables (staining intensities for CCL16 and CCL21) the rank correlation coefficient was used. The significance level was set at a P value of < 0.05.

Laser capture technology enabled the isolation of glandular epithelial cells from mid-secretory phase endometrial tissue of control and endometriosis subjects (n = 4 per group) (Figure 1A and 1B). For each subject, 150 – 600 glands were microdissected from 10 to 20 individual frozen tissue sections of 10 μm in thickness. Extraction of RNA of the LCM cells produced RNA yields of varying quantities ranging from 0.18 ng to 1.15 ng per glandular structure isolated (Figure 1C). The positive amplification of the epithelial-specific gene product, Htra3, ascertained that glandular epithelial cells had been specifically isolated and RNA integrity retained (Figure 1C).

To provide a global chemokine expression profile in eutopic endometrium from normal women and endometriosis patients, and to overcome patient to patient variability, total RNA extracted from LCM glandular epithelial cells was pooled from 4 individuals in each group (Table 1). Of 77 genes profiled on the cDNA gene arrays, expression of 22 chemokine and receptors were reproducibly (two separate runs) upregulated in pooled endometrium from women with endometriosis compared with controls (Table 2). Downregulation of expression was observed in 11 genes; however only 2 of these demonstrated a 2-fold or higher difference when compared to expression levels in normal endometrium (Table 3). For many upregulated genes, fold increase was not quantifiable, due to the lack of detectable expression in the control pool compared to very high expression in the endometriosis pool. In contrast, the fold-change quantified in genes which were downregulated was relatively small, with the largest difference being a 2.5-fold decrease in CCL28 expression.

To verify the results obtained from cDNA array experiments and to establish the variability between individuals, two candidate genes (CCL16 and CCL21, also known as HCC-4 and 6C-kine respectively) that had not previously been identified as associated with endometriosis, were selected for further analysis using quantitative real-time RT-PCR on individual unamplified RNA samples of glands from eutopic tissue (patients and controls: n = 3 subjects per cohort). All 3 samples used for each cohort, in this analysis, were included in cDNA array analysis. Selection of the candidate genes was based on the marked upregulation of mRNA levels on the cDNA arrays. These chemokines have been previously characterised in normal cycling human endometrium in our laboratory 27, the mRNA for CCL16 and CCL21 is maximal in entire endometrium during the mid-secretory phase of the menstrual cycle. In this study, mRNA expression for CCL16 was below the detection limit in the LCM glands from all 3 controls and 1 of the 3 endometriosis patient samples (Figure 2A). Where detectable, expression levels ranged from 0.01 – 0.05 fg/pg 18S RNA. Overall there was a significant elevation in eutopic endometrial glandular CCL16 expression in women with endometriosis compared to controls (n = 3; P = 0.049).

CCL21 mRNA was more abundant than that of CCL16 and was markedly elevated in eutopic glands from one endometriosis patient (mean, 0.61 fg/pg 18S RNA) (Figure 2B). The levels of CCL21 observed in the remaining endometriosis patients ranged from 0.02 – 0.19 fg/pg 18S RNA. In the control cohort, CCL21 mRNA expression ranged between 0.003 – 0.045 fg/pg 18S RNA. These combined data clearly show that the expression of CCL16 and CCL21 is disturbed at least in some patients with endometriosis. In women without endometriosis the expression was consistently low.

Since the findings from the cDNA array and real time RT-PCR experiments supported a role for CCL21 and CCL16 in at least some women with endometriosis, immunohistochemistry was performed to establish the presence and cellular location of the protein for these 2 chemokines in eutopic and ectopic endometrial samples. Immunoreactivity for both chemokines was evident in glandular epithelial cells but absent from stromal cells and was overall in accord with the real-time PCR quantitation. CCL16 protein was detected in the glandular epithelium of all but one of the 32 tissues examined and CCL21 protein was evident in glands in all tissues. However, staining was variable between individual tissues even within groups (Figure 3). No significant differences were detected for either CCL16 or CCL21 between mean glandular staining intensity in endometrium from normal women compared with eutopic endometrium from women with endometriosis (Figure 3). The staining pattern in some glands was predominantly apical (Figures 4A–C) suggesting secretion of the chemokine into the uterine lumen. Within each tissue, all glands were equally stained and staining was present in all cells in each gland (Figures 4A–D).

Comparisons were also made between the immunostaining in matched eutopic and ectopic endometrium in endometriosis patients. Positive staining was present in ectopic tissue for both CCL16 and CCL21 (Figures 3 and 4). In both cases, chemokines specifically localised in epithelial cells of glandular structures in ectopic lesions derived from ovarian, uterosacral ligaments and peritoneal sites. In some ectopic peritoneal lesions, large irregular cyst-like structures were present and strong staining for both CCL16 and CCL21 was apparent in glandular epithelial cells and in secretions within the cyst. In other peritoneal lesions, but more commonly, in ovarian lesions, glandular formation was very structured, closely resembling eutopic tissue morphology. Very strong staining was seen in glands in some ectopic lesions (one lesion for CCL16 (Figure 4E) and three for CCL21). Overall glandular staining intensity was higher in ectopic tissue than in eutopic tissue and this reached significance (P = 0.047) for CCL16. Importantly, there was a high degree of correlation between staining for CCL16 and CCL21 in each sample (Rank correlation coefficient R = 0.81, P < 0.001).

Leukocytes were evident in most tissues and stained strongly for CCL16 and CCL21: these were highly abundant in some lesions (Figure 4F insert) emphasizing the importance of laser capture microdissection in this study. Inclusion of leukocytes in the original cDNA array analysis would have skewed the data. As negative controls, endometrial sections with non-immune IgG substituted for the primary antibodies, demonstrated no localised staining (Figures 4G–H).