Identifying Biomarkers in RA by Laser Scanning Cytometry
Identifying Biomarkers in RA by Laser Scanning Cytometry
The study protocol has been approved by the local Ethics Committee of the University of Leipzig. All patients gave their informed written consent.
Synovial tissue was obtained by joint surgery (hand, foot, or shoulder) from patients with RA (n = 17) who met the American College of Rheumatology (ACR) revised criteria for the classification of RA. The control group consisted of 14 patients with trauma-induced acute arthritis (knee). Information about patients is summarized in Table 1. Histologic evaluation of RA tissue sections by a pathologist specified the local current-activity status in RA synovitis, according to the histopathologic scoring system by Stiehl. This scoring system of RA synovitis allows the characterization of synovial tissue into type I or type II synovitis. Moreover, based on a qualitative and quantitative characterization of cell infiltrations, disease activity is differentiated. The current-activity status provides information on exudative inflammatory processes on the synovial surface (in particular, fibrin exudation and granulocyte emigration) and is graded on severity from grade 1 to 3. RA tissue specimens were subclassified in five patients having highly current activity (grade 2 or higher; RA(++)) or 12 patients having mild current activity (less than grade 2 (RA(+)). This subclassification does not correlate with other patient data (for example, medication).
Molecules on cell types likely to be involved in the pathogenesis of RA (that is, adhesion molecules, activation markers, and receptors) were selected, according to the literature (Additional file 1, Table S1) and investigated with fluorescence labeling on cryosections of the RA synovium. Samples were scored as follows: absent (0), mild (1), moderate (2), or marked (3) expression. Antibodies against CD11b, CD38, CD29, CD90, HLA-DR, CD64, CD304, CD4, and CD271 provided a score > 2. These fluorochrome-labeled mAbs were used either alone or in combinations to stain serial sections of synovium for subsequent quantitative LSC analysis. For further information about antibody selection and staining, see Supplementary Information 1.
Quantitative analysis of synovial tissue was performed by using a laser scanning cytometer with the corresponding WinCyte software (CompuCyte; Westwood, MA, USA). All sections from each patient were analyzed within 24 hours after staining.
The region for scanning was set around the entire tissue section to get overall information of marker expression of the investigated material. Phantom contouring (non-cell-based analysis algorithm) was applied for analysis, as it resembles the in vivo situation better than a single-cell-based analysis (for details, see Supplementary Information 1).
Because phantom contouring is a non-cell-based analysis algorithm, it is essential to distinguish between tissue and background as a first step. This was based on the scatter signal of the tissue. Only phantoms including information about underlying tissue were used for further analysis (Additional file 1, Figure S3).
Each phantom contains information on integral fluorescence intensity and MaxPixel (the brightest pixel within the phantom). The median fluorescence intensity (MFI) of all phantoms was taken as a value of the averaged marker expression in the tissue. Based on negative control measurements, unspecific background intensity was defined. MFI values were corrected by subtraction of the respective MFI-negative control value. To determine the area of tissue positive for the marker, the parameter MaxPixel was used. Based on the negative control, the threshold (< 5% positive events) for autofluorescence was set for each fluorescent dye and patient, respectively. For each analysis, the percentage of phantoms with a MaxPixel value above this threshold (affected area) was determined (see Additional file 1, Figure S4).
For statistical evaluation of results, TANAGRA (version 1.4), an open-source data-mining software, was used. MFI values and affected area of RA patients and the control group were compared with the Mann-Whitney U test. A Kruskal-Wallis one-way ANOVA was applied for comparison of control and the RA subgroups RA(+) and RA(++). Values were considered significantly different if P ≤ 0.05 with P values ≤0.001 as highly significant. In case of significance, a Mann-Whitney U test was used for detailed group comparison. Significance level was adjusted by Bonferroni correction in case of multivariate analyses to P ≤ 0.017.
The discriminatory capability of the potential biomarkers identified by single antibodies or mAb combinations was tested with linear discriminant analysis (LDA). As case numbers are limited and to abate possible overfitting of data in LDA, cross-validation (10-fold with 10 repetitions) was performed after LDA. Values for sensitivity and specificity were taken from cross-validation. Likelihood ratio was calculated as LR: sensitivity/(1 - specificity).
Receiver operating characteristic (ROC) analysis was used to obtain information about the discriminatory capability and the variation of sensitivity and specificity in changing the decision threshold. AUC values > 0.85 were considered to be highly discriminative, and AUC < 0.55, to be nondiscriminative. ROC analysis was performed by using the web-based ROC analysis tool from Johns Hopkins University.
Graphic visualization of results was performed with SigmaPlot software (Systat, Erkrath, Germany).
Materials and Methods
Patients and Samples
The study protocol has been approved by the local Ethics Committee of the University of Leipzig. All patients gave their informed written consent.
Synovial tissue was obtained by joint surgery (hand, foot, or shoulder) from patients with RA (n = 17) who met the American College of Rheumatology (ACR) revised criteria for the classification of RA. The control group consisted of 14 patients with trauma-induced acute arthritis (knee). Information about patients is summarized in Table 1. Histologic evaluation of RA tissue sections by a pathologist specified the local current-activity status in RA synovitis, according to the histopathologic scoring system by Stiehl. This scoring system of RA synovitis allows the characterization of synovial tissue into type I or type II synovitis. Moreover, based on a qualitative and quantitative characterization of cell infiltrations, disease activity is differentiated. The current-activity status provides information on exudative inflammatory processes on the synovial surface (in particular, fibrin exudation and granulocyte emigration) and is graded on severity from grade 1 to 3. RA tissue specimens were subclassified in five patients having highly current activity (grade 2 or higher; RA(++)) or 12 patients having mild current activity (less than grade 2 (RA(+)). This subclassification does not correlate with other patient data (for example, medication).
Biomarker Candidates
Molecules on cell types likely to be involved in the pathogenesis of RA (that is, adhesion molecules, activation markers, and receptors) were selected, according to the literature (Additional file 1, Table S1) and investigated with fluorescence labeling on cryosections of the RA synovium. Samples were scored as follows: absent (0), mild (1), moderate (2), or marked (3) expression. Antibodies against CD11b, CD38, CD29, CD90, HLA-DR, CD64, CD304, CD4, and CD271 provided a score > 2. These fluorochrome-labeled mAbs were used either alone or in combinations to stain serial sections of synovium for subsequent quantitative LSC analysis. For further information about antibody selection and staining, see Supplementary Information 1.
Data Acquisition
Quantitative analysis of synovial tissue was performed by using a laser scanning cytometer with the corresponding WinCyte software (CompuCyte; Westwood, MA, USA). All sections from each patient were analyzed within 24 hours after staining.
The region for scanning was set around the entire tissue section to get overall information of marker expression of the investigated material. Phantom contouring (non-cell-based analysis algorithm) was applied for analysis, as it resembles the in vivo situation better than a single-cell-based analysis (for details, see Supplementary Information 1).
Quantitative Tissue Analysis
Because phantom contouring is a non-cell-based analysis algorithm, it is essential to distinguish between tissue and background as a first step. This was based on the scatter signal of the tissue. Only phantoms including information about underlying tissue were used for further analysis (Additional file 1, Figure S3).
Each phantom contains information on integral fluorescence intensity and MaxPixel (the brightest pixel within the phantom). The median fluorescence intensity (MFI) of all phantoms was taken as a value of the averaged marker expression in the tissue. Based on negative control measurements, unspecific background intensity was defined. MFI values were corrected by subtraction of the respective MFI-negative control value. To determine the area of tissue positive for the marker, the parameter MaxPixel was used. Based on the negative control, the threshold (< 5% positive events) for autofluorescence was set for each fluorescent dye and patient, respectively. For each analysis, the percentage of phantoms with a MaxPixel value above this threshold (affected area) was determined (see Additional file 1, Figure S4).
Statistical Analysis
For statistical evaluation of results, TANAGRA (version 1.4), an open-source data-mining software, was used. MFI values and affected area of RA patients and the control group were compared with the Mann-Whitney U test. A Kruskal-Wallis one-way ANOVA was applied for comparison of control and the RA subgroups RA(+) and RA(++). Values were considered significantly different if P ≤ 0.05 with P values ≤0.001 as highly significant. In case of significance, a Mann-Whitney U test was used for detailed group comparison. Significance level was adjusted by Bonferroni correction in case of multivariate analyses to P ≤ 0.017.
The discriminatory capability of the potential biomarkers identified by single antibodies or mAb combinations was tested with linear discriminant analysis (LDA). As case numbers are limited and to abate possible overfitting of data in LDA, cross-validation (10-fold with 10 repetitions) was performed after LDA. Values for sensitivity and specificity were taken from cross-validation. Likelihood ratio was calculated as LR: sensitivity/(1 - specificity).
Receiver operating characteristic (ROC) analysis was used to obtain information about the discriminatory capability and the variation of sensitivity and specificity in changing the decision threshold. AUC values > 0.85 were considered to be highly discriminative, and AUC < 0.55, to be nondiscriminative. ROC analysis was performed by using the web-based ROC analysis tool from Johns Hopkins University.
Graphic visualization of results was performed with SigmaPlot software (Systat, Erkrath, Germany).
Source...