Thrombospondin-1 Type 1 Repeats in a Model of Inflammatory Bowel Disease: Transcript Profile and Therapeutic Effects

Thrombospondin-1 (TSP-1) is a matricellular protein with regulatory functions in inflammation and cancer. The type 1 repeats (TSR) domains of TSP-1 have been shown to interact with a wide range of proteins that result in the anti-angiogenic and anti-tumor properties of TSP-1. To ascertain possible functions and evaluate potential therapeutic effects of TSRs in inflammatory bowel disease, we conducted clinical, histological and microarray analyses on a mouse model of induced colitis. We used dextran sulfate sodium (DSS) to induce colitis in wild-type (WT) mice for 7 days. Simultaneously, mice were injected with either saline or one form of TSP-1 derived recombinant proteins, containing either (1) the three type 1 repeats of the TSP-1 (3TSR), (2) the second type 1 repeat (TSR2), or (3) TSR2 with the RFK sequence (TSR2+RFK). Total RNA isolated from the mice colons were processed and hybridized to mouse arrays. Array data were validated by real-time qPCR and immunohistochemistry. Histological and disease indices reveal that the mice treated with the TSRs show different patterns of leukocytic infiltration and that 3TSR treatment was the most effective in decreasing inflammation in DSS-induced colitis. Transcriptional profiling revealed differentially expressed (DE) genes, with the 3TSR-treated mice showing the least deviation from the WT-water controls. In conclusion, this study shows that 3TSR treatment is effective in attenuating the inflammatory response to DSS injury. In addition, the transcriptomics work unveils novel genetic data that suggest beneficial application of the TSR domains in inflammatory bowel disease.


Introduction
Thrombospondins (TSPs) are glycoproteins that are secreted into the extracellular matrix. They have important functions in development, inflammation, angiogenesis and cancer. The five members of this family (TSP-1 through 5) are all multimodular extracellular proteins. TSP-1 comprises a 450 kDA protein composed of three 150 kDA disulfide-linked polypeptide chains [1,2]. TSP-1 and TSP-2 organize into trimeric structures. Each subunit of the trimer consists of multiple domains: an N-terminal globular domain, a procollagen domain, three types of repeated sequence motifs (type 1, type 2, and type 3 repeats) and a Cterminal globular domain. TSP-1 and TSP-2 both have the thrombospondin type 1 repeat, also called thrombospondin structural repeats (TSRs). TSP-3, TSP-4 and TSP-5, on the other hand, lack the TSRs and procollagen domain; they also differ from TSP-1 and TSP-2 in their pentameric structure. The thrombospondins have been characterized in a variety of organisms, including Drosophila, other arthropods and vertebrates [3].
The TSRs are about 60 amino acids in length and are evolutionarily conserved (e.g., [2], [3][4][5]). TSRs have roles in cell attachment and have been implicated in binding a variety of transmembrane and extracellular proteins. They have been shown also to have functions in the regulation of cell proliferation, migration and apoptosis in a variety of physiological and pathological events, such as wound healing, inflammation and inhibition of angiogenesis [6]. For example, by interacting with CD36 [7,8] and integrins [9], TSRs inhibit endothelial cell migration.
The various functions of the TSRs have been attributed to several recognition motifs. Characterization of these motifs has led to the use of recombinant proteins that contain these motifs; these recombinant proteins are deemed useful in cancer therapy [10,11]. The TSP-1 3TSR (that is, all three TSRs of the type 1 repeat domain) can activate transforming growth factor beta 1 (TGFb1) and inhibit endothelial cell migration, angiogenesis, and tumor growth [10,12]. In an efficacy study of 3TSR on human pancreatic cancer cells, 3TSR reduced tumor volume by 69% and induced extensive necrosis after 3 weeks of treatment. 3TSR treatment also reduced tumor microvessel density and increased apoptosis in the endothelia of tumors [12].
The TGFb1 activating sequence RFK is located between the first and second TSR. When a squamous cell carcinoma cell line transfected with TSR2+RFK was injected into nude mice, TSR2+RFK inhibited in vivo tumor angiogenesis and growth in the mice. The tumors were shown to have increased levels of active TGFb1. Treatment with TSR2 without RFK produced tumors that were slightly larger than the 3TSR and TSR2+RFK tumors [13]. However, mimetic peptides derived from TSR2, such as ABT-510 [14], decreased microvessel density and increased apoptosis in gliomas. ABT-510 was also effective in diminishing inflammation and angiogenesis in a murine model of inflammatory bowel disease (IBD) [15].
In this study, we evaluated the therapeutic effects of the subdomains of the TSP-1 TSRs using the DSS model of colitis and used a microarray approach to analyze the transcript profile of the mice treated with DSS and the TSRs. Our goals were to gain insight into the genes and pathways that are regulated upon TSR treatment and to determine the specific TSR treatment that would have the best potential for drug therapy. We were also interested in identifying novel genes not yet associated with and characterized in IBD, as well as putative intestinal stem cell genes. The results herein establish the efficacy of the 3TSR-containing recombinant protein in alleviating the inflammatory response to DSS injury as shown by improved clinical signs and lower perturbation of the transcriptome.

Mice
Wild-type (WT) mice (strain C57BL/6; n = 48) were purchased from Charles River Laboratories International (Wilmington, MA). Mice were bred at room temperature at the Wilkes University vivarium. All animal procedures were performed with the approval of the Wilkes University Institutional Animal Care and Use Committee and the U.S. National Institutes of Health (NIH) guidelines (Wilkes IACUC protocol # 189).
Dextran sulfate sodium (DSS) and TSR Treatments DSS (MW: 36,000,50,000) (MP Biomedical, LLC, Aurora, OH) was dissolved in the drinking water (distilled) of the mice at a dilution of 2.5% (wt/vol) and administered to 6-7 week old mice for 7 days to induce acute colitis. Simultaneously, mice were subcutaneously injected daily with the following recombinant proteins: 3TSR (n = 10), TSR2 (n = 9) TSR2+RFK (n = 15). TSR doses were administered according to published results [10]. As a control for the TSR treatment, mice (n = 14) were injected with saline. Following 7 days of DSS administration, mice were sacrificed by CO 2 asphyxiation.

Clinical parameters and Disease Activity Indices
The clinical severity of the colitis was determined daily by disease activity indices according to a published protocol [15]. Stool consistency was evaluated as follows: formed feces = 1, soft consistency = 2 and liquid feces = 3. Presence of fecal occult blood was tested using the stool guaiac test; scoring was done as follows: 1 for negative by guaiac paper, 2 for positive only with guaiac paper, 3 for bloody feces.

Histology and inflammation grading in colonic tissues
Intestines were removed, opened longitudinally, and rinsed with ice-cold PBS. For morphologic studies, tissues were fixed in Histochoice MB fixative (Electron Microscopic Sciences, Hatfield, PA) overnight, processed and cut in serial 5-mm sections. Sections were stained with hematoxylin and eosin (H&E) for histopathological analysis. Inflammation was graded as follows: 0, no inflammation; 1, modest numbers of infiltrating leukocytes in the lamina propria; 2, infiltration of leukocytes leading to separation of crypts and mild mucosal hyperplasia; 3, massive infiltration of inflammatory cells accompanied by disrupted mucosal architecture and complete loss of goblet cells. Slides were double-coded before pictures were taken and frames blindly analyzed in a monitor.

RNA isolation
Colon segments (,1 cm long) were collected from treated and untreated mice and snap-frozen in liquid nitrogen. Total RNA was extracted from these tissues using the RNAqueousH-4PCR kit (Applied Biosystems), following the manufacturer's instructions. RNA was DNase I-treated and quantified using the Nanodrop ND1000 spectrometer (Thermo Scientific, Wilmington, DE). RNA integrity was checked on a non-denaturing agarose gel. The microarray experiments are described in the following sections following the MIAME guidelines.

Microarray Processing
Total RNA from colons of at least three mice from each treatment group were submitted to the Center for Functional Genomics, University at Albany, Rensselaer, New York, for microarray processing. Treatment groups included the mice treated with DSS and the TSR-containing recombinant proteins 3TSR, TSR2 and TSR2+RFK and mice treated with DSS+ saline injections as controls. Untreated WT were used as additional control. RNA integrity of the samples used for microarray hybridization was verified using Agilent's 2100 Bioanalyzer (Santa Clara, CA). The Affymetrix (Santa Clara, CA) Mouse Gene 1.0 ST Array was used for the transcript profiling. 20 ng total RNA per sample was processed using WT Ovation Pico RNA Amplification System (NuGEN Technologies, Inc., San Carlos, CA). Sense target cDNAs were generated using the standard NuGEN WT protocol and hybridized to the arrays. Arrays were washed, stained on a FS 450 station and scanned on a GeneChip 3000 7G scanner using GeneChipH Command ConsoleH Software (AGCC).

Statistical Analysis
Statistical analysis of the CEL files was done using GeneSpring GX 11.5.1 (Agilent Technologies). Signals were quantile-normalized using PLIER16 algorithm and baseline-transformed to the median of all samples. The log2 normalized signal values were filtered to remove entities that show signal in the bottom 20th percentile across all samples. This list was further filtered to only include entities where at least 1 out of 6 conditions have CV,20.0% (that is, to remove probes that are highly variable across replicates in a condition). The list was then subjected to ANOVA (p,0.05) that compares all conditions to the WT-water or WT-DSS saline condition. A Benjamini-Hochberg False Discovery Rate correction (p,0.05) was also included in the analysis. A 1.5-fold filter was applied to identify genes that are differentially expressed between any two specific conditions. Microarray data have been deposited in the Gene Expression Omnibus (GEO) database under accession number GSE32697.

Real-Time Quantitative PCR Validation of Array Data
Real-time quantitative PCR (RT-qPCR) was performed on colonic total RNA isolated from another set of three mice. To obtain cDNAs, RNAs (2 ug) were reverse-transcribed using the High Capacity RNA-to-cDNA Kit (Applied Biosystems). This was followed by amplification of the undiluted cDNAs with the Fast SYBR Green master mix (Applied Biosystems) and real-time quantification in the StepOne Plus Real-time PCR system. Amplification was performed with 40 cycles of 95uC for 3 sec, and 60uC for 30 sec, using 150 nM of each primer.
The use of Gapdh as endogenous control gene in the RT-qPCR validation was problematic with samples that were treated with the recombinant proteins. Preliminary qPCR runs showed fluctuating levels of Gapdh among the TSR-treated samples. Ubc, Gapdh, Rps23 and Tpt1 have been evaluated for suitability as internal control in colon cancer studies [16]. Using the GeneMANIA [17] prediction server (www.genemania.org), we found that Gapdh interacts with Igfbp5 ( Figure S2). Our array data shows that Igfbp5 is up-regulated in WT-DSS mice. The Gapdh-Igfbp5 interaction suggests coexpression and could explain the high variability of Gapdh expression in our samples. On the other hand, query of Tpt1 in GeneMANIA showed that it is unrelated to any of our genes of interest.

Evaluation of microvascular density (MVD)
MVD was determined in colon sections of the treated mice. Sections stained with CD31/MECA antibodies were screened for colitic lesions. Serial pictures were taken at high power (6400) using a digital camera (Olympus Corporation, Tokyo, Japan), covering both mucosa and submucosa. Images were blindly evaluated; MVD was assessed by counting the vessels in lesions in which endothelial cells were positive for MECA/CD31.

Evaluation of leukocytic infiltration
Slides stained with antibodies CD68 for macrophages and Ly-6G/Ly-6C for granulocytes were scanned at low magnification and examined for areas with the highest number of leukocytic infiltration. The number of positive cells stained with Mac/CD68 or Ly-6G/Ly-6C was assessed at 4006 magnification per field.

Only 3TSR ameliorated clinical signs of colitis in the DSSinduced colitis
The presence of fecal blood was evaluated and rated in mice with DSS-induced colitis and treated with the TSRs ( Figure 1A). While no significant differences were observed during the first 5 days of treatment, by days 6 and 7, 3TSR mice showed lesser fecal bleeding when compared with mice treated with saline, TSR2 and TSR2+RFK (day 6, 3TSR and saline p = 0.020; day 7, 3TSR and saline p = 0.045). Fecal consistency was also evaluated daily ( Figure 1B). Again, by day 7, 3TSR mice displayed more solid feces compared to the liquid feces of saline-treated mice (p,0.0001) and TSR2+RFK (p,0.0001). TSR2 and TSR2+RFK mice showed more severe diarrhea than the controls (p,0.0001). Inflammation and epithelial injury were significantly decreased in colons of the 3TSR mice when compared with the saline-treated mice (p = 0.001) ( Figure 1C). Differences between 3TSR and the other two recombinant proteins were statistically significant (3TSR and TSR2+RFK: p = 0.034; 3TSR and TSR2: p = 0.025) ( Figure 1D).
Colonic tissues from TSR-treated mice showed differential patterns of leukocytic infiltration Sections stained with the antibody Ly-6G/Ly-6C were evaluated to identify granulocytes and monocytes ( Figure 2A). Colonic sections from 3TSR-and TSR2-treated mice showed counts considerably lower than saline-treated colons, (with p = 0.030 and 0.047, respectively) ( Figure 2B). Interestingly, counts obtained from the TSR2+RFK-treated colons were similar to the ones detected in saline-treated colons. The differences between TSR2 and TSR2+RFK were statistically significant (p = 0.021).
Macrophage infiltration was analyzed using Mac/CD68 antibody staining ( Figure 2C). Though no differences were detected when each group was compared with the saline control, TSR2+RFK mice showed a significantly lower influx of macrophages compared with the controls and the TSR2 and 3TSR groups (p = 0.0006) ( Figure 2D).
TSR treatment reduced the number of differentially expressed genes in the DSS induced colitis  Figure 3B and the gene list is presented in Table 1. Focusing on the 3TSR, only 43 genes were upregulated and two genes were down-regulated in the 3TSR vs WT-DSS saline comparison. Figure 3C presents a heatmap of the top 14 DE genes in this pair. The numbers of up-and down-regulated genes in the 3TSR treatment group diminished when compared to either TSR2 or TSR+RFK (Figures 3D and 3E). Table 2 presents the top 12 genes that are up-regulated in the 3TSR-treated group and the two down-regulated genes, using the WT-DSS-saline as reference.
Although some genes involved in the inflammatory response were up-regulated in 3TSR, their level of up-regulation was much lower than in the other TSR treatment groups. The complete list of differentially expressed genes (FC.1.5; p,0.1) is provided in the Table S2.

DSS and the TSR treatments up-regulated key genes involved in inflammation, cell adhesion and apoptosis
Genes up-regulated in the DSS and TSR-treatment groups belong to the following pathways: (1) matrix metalloproteinases, (2) inflammatory response, (3) peptide GPCRs, (4) eicosanoid/ prostaglandin synthesis regulation, (5) TGF-beta signaling, (6) Figure 1. Fecal blood and consistency rating in mice treated with DSS and TSRs over one week. While no significant differences were observed during the first days of treatment, by days 6 and 7, mice treated with the 3TSR (n = 10) showed lesser fecal bleeding when compared with mice treated with the saline (n = 14), TSR2 (n = 9) and TSR2+RFK (n = 15), (day 6: p = 0.02; day 7: p = 0.045, asterisk) (A). Fecal consistency was also evaluated daily (B). By day 7, 3TSR mice displayed more solid feces compared to the liquid feces of saline-treated mice (p,0.0001, asterisk) and TSR2+RFK (p,0.0001). Inflammation grading (C, D) was assessed and H&E stained sections. Decreased inflammation was observed in sections of colons from 3TSRtreated mice (n = 9) compared to colons from mice treated with TSR2+RFK (n = 12), TSR2 (n = 10) and saline (n = 7). 4006 original magnification. doi:10.1371/journal.pone.0034590.g001 smooth muscle contraction, (7) integrin-mediated cell adhesion and (8) apoptosis (Table 3). S100 calcium binding protein A9 (S100a9; also known as calgranulin B) had the highest transcript level in all treatment groups, with the WT-DSS group showing the highest FC (81), whereas the WT-DSS-3TSR has FC = 18 (Table  S2). An inflammatory response was indicated by the high transcript levels of Il Genes coding for proteins that are well known to interact with TSP-1 during inflammation were differentially expressed in DSS and TSR-treated mice, except in the 3TSR group. Cd47 was only slightly up-regulated in WT-DSS-TSR2 (1.8). It was not DE in TSR2+RFK and 3TSR. Tgfb1 was also slightly up-regulated in TSR2 (FC = 2.5) and TSR2+RFK (FC = 1.8), but not in 3TSR. None of the treatment samples showed differential expression of Cd36.

Real-time qPCR validation
Actual qPCR data revealed no fluctuation in expression of the internal control gene, Tpt1, which was consistent with the microarray data. We validated 15 DE genes to verify the array data. Figure 4 summarizes the validation for representative genes coding for acute phase proteins, growth factors and novel proteins. Validation for the metalloproteinases and genes coding for cytokines and a protein important for intestinal homeostasis are shown in Figure S3. Although Srgn and Wfdc12 have p-values greater than 0.05, (Srgn p = 0.056, Wfdc12 p = 0.065), the real-time qPCR validation shows that they are indeed differentially expressed.

IHC validation of genes recognized in stem cell biology, cancer and inflammation
Array data show that Lgr5, an intestinal stem cell biomarker [18,19], was down-regulated in the TSR2-treated WT mice. Lgr5 was not differentially expressed in the 3TSR and TSR2+RFK mice. However, results obtained in the RT-qPCR validation show that Lgr5 was significantly down-regulated in the TSR2+RFK ( Figure 4B).
MMP3, IL-6 and Calgranulin (MRP8) are relevant in angiogenesis and inflammation. Validation by IHC of these genes ( Figure 5) showed the MMP3 protein localized in the cytoplasm of epithelial cells and endothelial cells. Strong positive staining for MMP3 was detected mainly in monocytes, polymorphonuclears, plasma and fibroblasts. Stromal staining was detected in the submucosa. Endothelial cells and the leukocytic infiltrate in inflamed areas were positive for IL-6. MRP8 staining was observed almost exclusively in polymorphonuclears.

Discussion
Our results indicate that 3TSR suppressed the inflammatory response induced by DSS injury and these data are supported by the array results. For pharmacological consideration, it is logical to assume that the lesser the perturbation of the transcriptome, the better the clinical outcome. Our results also confirm the finding that TSR2 is less effective than either 3TSR or TSR2+RFK [10]. 3TSR has been shown to induce endothelial cell apoptosis and inhibit tumor growth [11]. This function may be correlated to the modulation of VEGFR2 function by TSP-1 and 3TSR [20].
One of the unexpected findings when clinical parameters were evaluated was the severity of the bleeding and diarrhea in mice treated with TSR2 and TSR2+RFK. These results do not agree with previous studies using the ABT-510, a TSR2 mimetic peptide [15]. In the ABT-510 studies, osmotic mini-pumps were used for the continuous delivery of the peptide in contrast to the daily injections used in this study. Moreover, increased eNOS activity has been attributed to increased mucosal blood flow in induced colitis [21] and TSP-1 is known to regulate nitric oxide (NO) signaling [22]. Noteworthy is the lesser influx of granulocytes in the TSR2 mice compared to the TSR2+RFK and the 3TSR. This suggests that the TSR2 domain inhibits the influx of granulocytes independently of the RFK domain. TSR2 has been demonstrated to inhibit the secretion of NO during the acute phase of inflammation, and therefore decreasing the trafficking of neutrophils [23]. On the other hand, RFK presumably inhibits inflammation by activating TGFß1. However, recent studies suggest that when there is a high secretion of IL-1 and IL-6, TGFb1 will favor the production of Th17 cells. Th17 cells secrete pro-inflammatory cytokines and metalloproteinases that exacerbate the inflammatory response [23].
Our array data showed that metalloproteases and chemokines such as MMP10, MMP13, the CCs and CXCs are involved in the pathogenesis of DSS colitis. Expression of Cxcl2 and Cxcl5 ( Figure  S3B), the proteins of which are known to target neutrophils and have suspected roles in IBD, was significantly increased in the DSS and TSR-treated mice during the acute phase of colitis. The increase in Cxcl2 expression is the same observation reported for the leptin receptor-deficient mice that were treated with DSS [24]. An irregularity in chemokine signaling could jeopardize epithelial integrity, which could lead to the breakdown of defense against invading pathogens. Moreover, Cxcl5 has a great potential in eliciting a pro-angiogenic phenotype [25].
The concentration of acute phase proteins is known to increase or decrease during inflammation. Hp and Ptx3, both of which are enriched in the DSS and TSR-treated mice, code for acute phase proteins. In a study of Hungarian patients with Crohn's disease (CD), Hp polymorphism was determined to be associated with CD and inflammatory disease behavior [26]. Hp has been suggested to have a protective role in inflammatory colitis [27]. Ptx3 has been annotated to ''positive regulation of nitric oxide biosynthetic process'' and to ''zymosan binding''.
This study showed strong evidence for the alteration of gene expression in DSS-induced colitis that supports vulnerability to cancer. As an example, calgranulins, S100A8/A9 (calgranulins A and B, respectively), have been detected in various human cancers; hence they have been suggested to play a key role in inflammationassociated cancer [28]. They support epithelial barrier function and mediate responses to infection and inflammation [29] by amplifying pro-inflammatory responses [30]. In gastrointestinal epithelial cells, their expression is induced during inflammation and increased protein expression of S100A9 has been detected in colitis-related carcinogenesis in mice [31]. Key genes associated with colon tumor progression are reportedly activated by S100A8/A9 [32].
The 'novel' genes that we validated appear to have a role, one way or another, in inflammatory response and the promotion of inflammation-related cancer development. The function of these genes as they relate to diseases and how these genes interact with other genes have not been extensively studied. The description ascribed to them in the gene ontology database, however, indicates some involvement in the development of physiological disorders.
Genome-wide association studies have identified Capsl to have association with type 1 diabetes [33] and the autoimmune thyroid disease Graves' disease [34]. Cbr3, an enzyme that reduces carbonyl groups in xenobiotics and endogenous compounds such as steroids and eicosanoids, has been suggested to have a minor role in xenobiotic metabolism in humans [35]. Ereg is a growth factor that resides in the extracellular space and is a marker for ovarian cancer development [36]. Fpr3 is a member of a G-protein coupled receptor family that controls immune responses such as leukocyte chemotaxis and activation [37]. Serpina3n has been found to inhibit human Granzyme B activity in sertoli cells as a mechanism to suppress the immune response [38]. SRGN is a proteoglycan that has important roles in inflammatory reactions [39]. In human endothelial cells, it co-localizes and interacts with tissue-type plasminogen activator (PLAT) [40]. In macrophages, where it is the major secreted proteoglycan, it regulates the secretion of TNF [41]. The up-regulation of Srgn ( Figure S3B) in the DSS-and TSR-treated mice suggests that it is a component of the immune response in these mice.
The significance of Wfdc12 up-regulation in the DSS-and TSRtreated mice ( Figure S3C), lies in the fact that an altered microbiota composition in the intestinal tract characterizes the pathogenesis of IBD [42,43]. WFDC12 is an antibacterial protein that inhibits the growth of both E. coli and S. aureus at a IC90 of 10 mM [44]. With its anti-infective activity, it could serve as mucosal defense against pathogenic bacterial strains, in conjunction with chemokine signaling.
Evidence for involvement of Insulin-like growth factor (IGF) signaling in cancer progression has been increasing [45]. IGF signaling plays a critical role in promoting cell survival and proliferation of embryonic and adult stem cells [46]. IGFBP5 is associated with ''stemness'' and its expression has been linked to LGR5, which has been shown to be expressed in cycling columnar epithelial cells at the intestinal crypt base [18,19]. LGR5-deficient mice showed decreased Igfbp5 expression [47]. Erosive lesions in the intestinal epithelium due to DSS could account for the lower transcript levels of Lgr5 in the DSS-treated mice.
LGR5 deficiency was reported to deregulate the Wnt pathway and contribute to cancer [47].
In conclusion, this study showed that the 3TSR have antiinflammatory properties that seem to be independent of any angiogenic inhibition or ability to activate TGFß1. These data highlight the anti-inflammatory functions of the 3TSR of TSP-1 in the mouse intestine and its potential use as novel therapeutic tool in IBD. The transcript profiles of the treatment groups showed a pool of genes involved in critical pathways of inflammation. The identification of these genes has implications in improving our  Figure S1 Assessment of microvessel counts. Using combined MECA/CD31 antibodies (A), fewer blood vessels were observed in sections from TSR2 mice (n = 20) compared to sections from mice treated with TSR2+RFK (n = 35), 3TSR (n = 29) or saline (n = 16), (B). 4006 magnification. (TIF) Figure S2 A functional association network of selected mouse genes. GeneMANIA, a web interface for predicting gene function and determining usability of genes for functional assays (www.genemania.org) was used to determine suitability of endogenous control genes for validating array data by RT-qPCR. The following genes (grey circles) were inputted in GeneMANIA: Ubc; Tpt1; Rps23; Gapdh; Capsl; Igfbp5; Fpr3; Srgn; Kcne3; Mmp10; Mmp13; Wfdc12. The network generated shows interaction between Gapdh and Igfbp5; therefore, Gapdh expression is likely to be co-expressed with Igfbp5 and show varying expression levels similar to that of Igfbp5. Tpt1, on the other hand, has no predicted interaction with any of the input genes. (TIF) Figure S3 Additional RT-qPCR validation of array data.

(TIF)
Table S1 Genes used for the RT-qPCR validation and their primers. Tpt1 was used as the control gene after it was found to show no fluctuation compared to four other genes that were tested. (XLSX)