Open Access Research article

Systems analysis of primary Sjögren's syndrome pathogenesis in salivary glands identifies shared pathways in human and a mouse model

Steve Horvath1, Abu NM Nazmul-Hossain1, Rodney PE Pollard2, Frans GM Kroese2, Arjan Vissink2, Cees GM Kallenberg2, Fred KL Spijkervet2, Hendrika Bootsma2, Sara A Michie3, Sven U Gorr4, Ammon B Peck5, Chaochao Cai1, Hui Zhou1 and David TW Wong1*

  • * Corresponding author: David TW Wong

  • † Equal contributors

Author Affiliations

1 School of Dentistry, Dental Research Institute, University of California at Los Angeles, 10833 Le Conte Avenue, 73-017 CHS, Los Angeles, CA 90095-1668, USA

2 Department of Oral and Maxillofacial Surgery, University Medical Center Groningen, University of Groningen, P.O. Box 30.001, Groningen, the Netherlands

3 School of Medicine, Stanford University, 300 Pasteur Drive, R241 MC 5324, Stanford, CA 94305, USA

4 School of Dentistry, University of Minnesota, 7536 Moos HST, 515 Delaware Street SE, Minneapolis, MN 55455, USA

5 Department of Pathology, Immunology and Laboratory Medicine, School of Medicine, University of Florida College of Medicine, JHMHSC D6-33D, 1600 SW Archer Road, Gainesville, FL 32610-0275, USA

For all author emails, please log on.

Arthritis Research & Therapy 2012, 14:R238  doi:10.1186/ar4081

Published: 1 November 2012

Additional files

Additional file 1:

a table presenting MM membership for human data. For each gene, the table reports the MM measure, which is also known as eigengene-based connectivity. Each module gives rise to its own MM measure; for example, MM denotes the measure for the magenta module. Columns whose name starts MM report the Pearson correlation coefficient between the gene expression value and the respective module eigengene. For each MM measure (a correlation coefficient), one can also report a corresponding correlation test P value based on the Student t test (see columns whose name starts p.MM). For example, p.MM.magenta reports a two-sided correlation test P value based on the Student t distribution. Column B reports the original module assignment based on the hierarchical cluster tree but the module membership measures were used to select genes for the functional enrichment analysis.

Format: XLS Size: 14.8MB Download file

This file can be viewed with: Microsoft Excel Viewer

Open Data

Additional file 2:

a table presenting GO terms for genes in human pSS-associated co-expression modules. The table reports GO categories, uncorrected test P values, and corresponding P values that correct for multiple comparisons using the following methods: Bonferroni, Benjamini, and the false discovery rate. It also reports the number of population hits and related count data used to calculate the hypergeometric test P value. Genes column reports the Affymetrix probes that were hits for the corresponding GO term. The magenta module (highlighted in yellow) is of particular interest since it was preserved in the mouse model.

Format: XLS Size: 729KB Download file

This file can be viewed with: Microsoft Excel Viewer

Open Data

Additional file 3:

a table presenting standard differential expression analysis in human data. For each gene (transcript), the results of several two-group comparison tests (carried out with the WGCNA function standardScreeningBinaryTrait) are reported. The first group comparison test contrasts females versus males in control subjects only (columns B through L). The second group comparison test compares controls versus pSS (columns M through W). The third group comparison test compares controls versus sicca (columns × through AH). The fourth group comparison test compares pSS versus sicca (columns AI through AS). Further, columns AT through AZ report the results of a correlation test where each gene was correlated with an ordinal variable that encodes disease status (0 for controls, 1 for sicca, and 2 for pSS). Any column whose name is preceded by corPearson reports the Pearson correlation coefficients where the binary grouping variable (for example, female vs. males) was coded as a binary numeric variable (1 for the first group, 0 for the second group). The meaning of the first and second groups can be learned from column t.Student, which reports the Student t-test statistic. For example, t.Student.F.vs.MGenderInControls shows that the first group corresponds to females and the second group to males (among the control samples). Similarly, t.Student.Control.vs.pSS and t.Student.Control.vs.Sicca show that the first group is comprised of controls. t.Student.pSS.vs.Sicca shows that the first group is comprised of pSS patients. Column meanFirstGroup reports the mean expression value in the first group. FoldChange column reports a signed fold-change value defined by the ratio meanFirstGroup/meanSecondGroup if meanFirstGroup >meanSecondGroup. But if meanFirstGroup <meanSecondGroup, the fold change is defined as minus meanSecondGroup/meanFirstGroup. SE.FirstGroup reports the standard error in the first group. AreaUnderROC reports the area under the receiver operating characteristic curve. pvalueStudent reports the Student t test P value that corresponds to the Student t-test statistic. q.Student reports the corresponding q value (local false discovery rate) calculated with the qvalue R package. nPresentSamples reports the number of nonmissing observations that were available.

Format: CSV Size: 11.1MB Download file

Open Data

Additional file 4:

a table presenting correlations of mouse expression data with time. This comma-delimited file reports the results from a correlation test where each mouse gene (transcript) is correlated with time (measured in weeks). Column corTime reports the correlation coefficient, column ZCorTime reports the corresponding Student t-test statistic, and column pValueStudentCorTime reports the corresponding two-sided Student t test P value. Column AreaUnderROCCorTime reports the area under the receiver operating characteristic curve calculated with the function rcorr.cens in the Hmisc R package.

Format: CSV Size: 8.7MB Download file

Open Data

Additional file 5:

a figure showing Magenta module expression in the mouse data. For each week (x axis) the height of the bar shows the mean of the magenta module eigengene value (±1 standard error). P value calculated with the Kruskal-Wallis test, which is a nonparametric group comparison test. While the magenta module was defined based on the human data, this plot shows how the corresponding module eigengene relates to time course in the mouse data. To define the magenta module eigengene in the mouse data, human genes were mapped to orthologous mouse genes.

Format: PDF Size: 2KB Download file

This file can be viewed with: Adobe Acrobat Reader

Open Data

Additional file 6:

a figure showing the quantitative real-time PCR validation in the mouse data. The results of the quantitative RT-PCR validation analysis involving a select group of genes (CD1D1 ortholog of CD1D, CCR7, CXCL10, CXCL12, SLAMF9, LTA) are reported. Barplots in the upper and lower panels correspond to the expression values measured by microarrays and RT-PCR, respectively. The bars are colored according to time (weeks). To verify the selected gene expressions (n = 7), aliquots of salivary gland RNA originally used for the microarray data were pooled. Each cDNA preparation was quantified by spectrophotometry and PCR performed. Quantifications were determined by ImageJ. Relative gene expression values yielded by the PCR arrays are compared directly with data yielded by the Affymetrix 3' Expression Array GeneChip Mouse Genome 430 2.0 arrays. Pooling RNA from each time point prior to cDNA preparation is thought to be the underlying reason for higher transcript detection in a couple of RT-PCR reactions (for example, in Cxcl12 samples).

Format: DOCX Size: 204KB Download file

Open Data