TABULAR RESULTS                     FIGURES
 

Methods: High density microarrays (Affymetrix Hu6800 chips) were used to study expression of approximately 6000 genes in (4) normal and (10 endometrioid adenocarcinomas) neoplastic endometrial tissues.  50 genes which discriminate normal from malignant tissues were selected by Permax test (actual data t-test interpreted against a background of permuted maximum t-tests for the dataset) and minimal 3-fold and 100 expression unit differences of means.  Genes which cycle between estrogenic (n=2) and secretory (n=2) phases of normal endometrium were stratified by t-test and those top 50 which meet minimal 3-fold and 100 expression unit differences of means selected.  

Table 1: Clinicopathologic Features of Tissues Studied.   Adobe pdf
All carcinomas were of the endometrioid type, with Grade shown on a scale of 1-3.  Percentage myometrial invasion (“Invasion”) is based upon pathologic evaluation of the hysterectomy specimen.  Cellular composition of tissue used for RNA isolation is shown as that percentage of tissue volume occupied by epithelial cells (“% epithelium”).  Since measurements excluded acellular areas (gland lumens, tissue fractures), the sum of  epithelial and stromal percentages is 100%.  Phenotype of “microsatellites” is shown as Stable or Unstable based upon comparison of matched tumor and normal myometrial genotype in each patient.  PTEN function was evaluated by immunohistochemistry of paraffin embedded tissues, and scored as PTEN-expressing (“positive”) or PTEN non-expressing (“Null”).  Lack of PTEN protein expression has been associated with genotypic deletions or mutations of the PTEN gene itself. NA = not applicable, ND= Not done.             

Tabular Results: Additional Information about Affymetrix probe sets is available online.

Raw Data Spreadsheet:   Excel File              Tab Delimited Text File
Data are provided for 2 normal proliferative ("P") endometria, 2 normal secretory ("S") endometria, and 10 endometrioid adenocarcinomas ("T").  Average Differences for each sample were rescaled to sum to 3,000,000 arbitrary units over all probe sets.  To avoid misinterpretation of very low expression levels, genes labeled as Absent or Marginal by the Affymetrix algorithms were reset to an Average Difference of 20.  Further, genes called Present but with an Average Difference expression level less than 20 were reset to a value of 20.  This resulted in a dataset truncated on the left tail (i.e., low expression level) at a value of 20, in which only genes determined to be “Present” by the Affymetrix call were included as positive expression values.  Each row is one gene, identified by Affymetrix Probe set (from the Hu6800 chip) and gene name.

Table 2.  50 Genes which discriminate normal from malignant endometrium.  Adobe pdf
Arranged by functional category, 50 genes selected by Permax test <0.5, fold>3, and difference>100 (Figure 1, 2) are listed along with their average values for four normal (“Normal”) and ten malignant (“Tumor”) endometria.  Lower Permax values have greater significance.  Gene identifiers include the Affymetrix probe set identifier (Affy ID), Locus Link ID number (“Locus Link”), and standard abbreviation (“Abrev”)as shown in Locus Link.   Those genes whose difference between groups exceeds four-fold  (fold difference of means, “FoldDiff”) cannot be simply ascribed to the differing abundance of epithelial tissues in normal compared to malignant tissues.  Gene abbreviations followed by “as” indicate an alternatively spliced RNA form. 

Supplemental Table I: 50 Genes Downregulated in Secretory Endometrium  Adobe pdf
50 hormonally regulated genes with lower expression in progestenic secretory (n=2) normal endometrium compared to estrogenic proliferative (n=2) normal endometria were identified using additive selection criteria including threshold filters (difference >100, ratio >3) for distance between group means and ranking of genes by t-test.  Genes are identified by the Affymetrix GeneChip probe set number (Affy Probe Set) , and brief descriptive name.  Mean expression is shown for proliferative (PE) and secretory (SE) endometria in normalized AD units without log transformation.  

Supplemental Table II: 50 Genes Upregulated in Secretory Endometrium.   Adobe pdf
50 hormonally regulated genes with higher expression in progestenic secretory (n=2) normal endometrium compared to estrogenic proliferative (n=2) normal endometria were identified using additive selection criteria including threshold filters (difference >100, ratio >3) for distance between group means and ranking of genes by t-test.  Key is identical to Supplemental Table I.
 

Figures:

Figure 1: Permax selection of genes which discriminate normal from malignant endometrium.
Panel A: Cumulative frequency distribution of the maximum and minimum t-statistics from 1001 dataset permutations of 4 normal and 10 carcinoma endometrial specimens.  Panel B: Scatterplot of  753 expressed genes (crosses) in normal (x axis, mean of 4 samples) and malignant (y axis, mean of 10 samples) endometrial tissues in which the means of normal and tumor tissues are separated by a minimum of 100 expression units and have a ratio difference ≥ 3.   t-statistics calculated from the non-permuted dataset were compared with the cumulative frequency distribution of the permuted t-statistics as shown in Panel A to assign a probability of error (Permax).  Solid circles indicate 50 selected genes with Permax values less than 0.50.    

Figure 2: Genes which distinguish normal from malignant endometrium.
Expression matrix of 50 genes which distinguish normal (2 proliferative, PE; 2 secretory, SE) from malignant (10 tumors, T) endometrium (Permax<0.50, 3-fold, 100 difference).  Columns show individual tissues, rows represent genes.  Color scale shows standard deviation from the mean expression value for each gene. Dendrograms on the margin show agglomerative hierarchical clustering (Wards linkage, Euclidean distances) of genes (right) and tissues (bottom).

 Figure 3:  Global expression patterns in malignant endometrium, organized by fluctuation in the normal menstrual cycle. 
Average expression for two phases of normal endometrium, proliferative (PE, n=2), and secretory (SE, n=2) is plotted from left to right in order of increasing SE-PE difference.  3003 genes that were moderately expressed (>100 normalized AD units) in at least one normal tissue are plotted as a least squares smoothed line, along with values for 10 carcinomas.  Genes which are induced in secretory endometrium are seen on the right as an upswing of the fine dotted line.  For this class of genes, the carcinomas demonstrate an expression level closer to that observed in PE. Vertical striped lines below show location of 100 genes which best discriminate proliferative from secretory endometrium.  These are divided into those genes upregulated in proliferative (50 genes, left group) or secretory (50 genes, right group) endometrium, as selected by meeting minimum filter requirements (normalized AD unit expression means, ratio>3, difference>100) and having lowest t-statistics. 

Figure 4: Expression matrix of 100 genes selected from Figure 3 which distinguish proliferative from secretory phases of the normal menstrual cycle.  
Color scale shows expression for each gene (row) normalized as number of standard deviations from the mean.  Included are the 2 secretory normal (S), and 2 proliferative normal (P) endometria compared to select this group of genes, as well as 10 endometrioid endometrial adenocarcinomas (T).  Dendrograms on the margin show agglomerative hierarchical clustering (Wards linkage, Euclidean distances) of genes (right) and tissues (bottom).