Skip to main content
NCBI home page
Bioinformatics logoLink to Bioinformatics
. 2008 Aug 21;24(20):2384–2390. doi: 10.1093/bioinformatics/btn451

Novel pathway compendium analysis elucidates mechanism of pro-angiogenic synthetic small molecule

Kristen A Wieghaus 1,†, Erwin P Gianchandani 1,†, Mikell A Paige 2, Milton L Brown 2, Edward A Botchwey 1,3,*, Jason A Papin 1,4,*
PMCID: PMC2562016  PMID: 18718940

Abstract

Motivation: Computational techniques have been applied to experimental datasets to identify drug mode-of-action. A shortcoming of existing approaches is the requirement of large reference databases of compound expression profiles. Here, we developed a new pathway-based compendium analysis that couples multi-timepoint, controlled microarray data for a single compound with systems-based network analysis to elucidate drug mechanism more efficiently.

Results: We applied this approach to a transcriptional regulatory footprint of phthalimide neovascular factor 1 (PNF1)—a novel synthetic small molecule that exhibits significant in vitro endothelial potency—spanning 1–48 h post-supplementation in human micro-vascular endothelial cells (HMVEC) to comprehensively interrogate PNF1 effects. We concluded that PNF1 first induces tumor necrosis factor-alpha (TNF-α) signaling pathway function which in turn affects transforming growth factor-beta (TGF-β) signaling. These results are consistent with our previous observations of PNF1-directed TGF-β signaling at 24 h, including differential regulation of TGF-β-induced matrix metalloproteinase 14 (MMP14/MT1-MMP) which is implicated in angiogenesis. Ultimately, we illustrate how our pathway-based compendium analysis more efficiently generates hypotheses for compound mechanism than existing techniques.

Availability: The microarray data generated as part of this study are available in the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/).

Contact: botchwey@virginia.edu; papin@virginia.edu

Supplementary information: Supplementary data are available at Bioinformatics online.

1 INTRODUCTION

In recent years, high-throughput experimental technologies have facilitated the functional characterization of small molecules and other compounds hypothesized for therapeutic intervention. For example, microarray technology has been used to perform transcriptional profiling for discovery of unknown drug mechanism, as well as for prediction of overall drug efficacy (Lamb et al., 2006). However, extraction of biologically relevant signaling pathways and unambiguous mechanistic insights from high-throughput data remains a significant challenge (Bader et al., 2004; Butcher and Schreiber, 2005). ‘Compendium analyses’ comparing the genetic expression profiles of unknown drugs with a reference collection of profiles (i.e. a database) of known drugs have been proposed as a form of mechanistic interrogation, as drugs with similar targets are likely to have correlative transcriptional footprints (Fig. 1A) (Butcher and Schreiber, 2005). For instance, gene set enrichment analysis (GSEA) was recently developed to associate molecular function between unknown and known compounds by matching similar gene expression profiles (Subramanian et al., 2005). However, parallel transcription data from potentially similar drugs at appropriate concentrations in relevant cells often do not exist, or alternative therapeutics/proteins must be chosen arbitrarily for comparative microarrays. Likewise, docking studies and other modalities for evaluating drug mechanism with an unknown target require exhaustive screening of all possible binding partners (Taylor et al., 2002).

Fig. 1.

Fig. 1.

Open in a new tab

The novelty of a pathway-based compendium analysis. High-throughput experimentation, including gene expression profiling, has facilitated discovery of small-molecule mechanism and target specificity. (A) The existing approaches for compendium analysis, wherein transcriptional profiles of a compound with unknown mechanism are compared to a reference database of profiles of various compound treatments with known mechanistic functionality (Butcher and Schreiber, 2005). With this approach, matching profiles are suggestive of similar mode-of-action. (B) Our novel pathway-based compendium analysis. Our approach involves overlaying a putative therapeutic compound's transcriptional profile on a pathway knowledge database, and identifying signaling networks that are most directly correlated with the expression data.

We therefore propose in this study a novel method of compendium analysis (Fig. 1B), and we use it to probe the mechanism of phthalimide neovascular factor 1 (PNF1) (Fig. 2), a new bioactive small molecule that we previously demonstrated in vitro as inducing endothelial functions consistent with angiogenic processes and, more recently, in vivo as stimulating microvascular remodeling (Wieghaus et al., 2006, 2008a). Although there are emerging approaches that align network structures to infer underlying biology (Kelley et al., 2003), 2004; Sharan et al., 2005), our pathway-based compendium analysis is unique in that it overlays the transcriptional profile of a small molecule on a literature-derived pathway knowledge database, thereby identifying signaling networks specifically affected by the drug through correlation with expression data. Furthermore, our interest in PNF1 stems from its applicability in the promotion of vascularization, which is critical for successful treatment of many age-related diseases arising from impaired or abnormal function of microvasculature (Brey et al., 2005; Madeddu, 2005). PNF1 specifically exploits the stability of organic molecules and is more easily synthesized, relative to polypeptide growth factors that have also been proposed for therapeutic induction of angiogenesis, and consequently accommodates a wide range of storage and processing methods for in vivo delivery. Therefore, combined with appropriate design of biomaterial carriers, this novel drug may ultimately result in the development of ‘designer’ drug-release systems that can be tailored for tissue-specific applications. Elucidating the mechanism of the angiogenic effects of PNF1 is critical for drug optimization and proper therapeutic utilization.

Fig. 2.

Fig. 2.

Open in a new tab

Workflow for a controlled, multi-timepoint study of HMVEC stimulated by PNF1 treatment. Oligonucleotide microarray analysis was used to identify gene transcripts in HMVEC that were significantly regulated 1, 2, 4, 8, 16, 24 and 48 h after stimulation with 30 μM PNF1. Here we illustrate how these data, coupled with network analysis, were used to probe PNF1 mechanism as part of a novel pathway-based compendium analysis.

We previously interrogated the transcriptional profile associated with PNF1 24 h post-stimulation in human microvascular endothelial cells (HMVEC) (Wieghaus et al., 2007). By applying network analysis to genome-scale microarray data, we identified pathways of global biological control during PNF1 stimulation. We concluded that transforming growth factor-beta (TGF-β) signaling is associated with the angiogenic mechanism of PNF1 at 24 h, and we validated these findings with PCR of TGF-β pathway components (Wieghaus et al., 2007). Additionally, separately, we generated microarray data at key signaling and regulatory timepoints spanning 1–48 h post-supplementation by PNF1 in HMVEC, and we identified a TGF-β-induced gene implicated in angiogenesis, matrix metallopeptidase 14 (MMP14 or MT1-MMP), with differential regulation by PNF1 conserved over a maximum number of multiple timepoints (Wieghaus et al., 2008b).

Here, we developed and applied a novel pathway-based compendium analysis by coupling our previous transcriptional profile of PNF1 in HMVEC (i.e. gene expression data spanning 1, 2, 4, 8, 16, 24 and 48 h poststimulation) with network analysis (including substantial libraries of information regarding known associations among network components). By examining genome-scale expression arrays instead of single gene transcripts, key signaling events that may not have been elucidated by studying individual genes were revealed. Ultimately, the association of each gene transcript with several well-characterized angiogenic signaling pathways was determined efficiently, and relative ‘fingerprints’ of each angiogenic pathway within the PNF1 transcriptional profile were evaluated. We identified that PNF1 first induces function of the tumor necrosis factor-alpha (TNF-α) signaling pathway which in turn affects TGF-β signaling. The results of this novel pathway-based compendium analysis were consistent with our earlier findings of PNF1-directed TGF-β signaling at 24 h, including MMP14/MT1-MMP differential regulation at multiple timepoints (Wieghaus et al., 2008b). It is important to note that this new method for compendium analysis can easily be extended to the mechanistic exploration of other drugs and cell/tissue treatments as we discuss below.

2 METHODS

2.1 Cell culture and RNA isolation

HMVEC (Cambrex, Walkersville, MD, USA) were cultured in endothelial growth medium 2-microvascular (bulletkit, BioWhittaker, Walkersville, MD, USA) supplemented as directed with 5% fetal bovine serum, as previously reported (Wieghaus et al., 2007), 2008b). The cells (passage 9) were plated at 2.5 × 104 cells/cm2 at 37○C in a humidified chamber with 5% carbon dioxide. They were grown to confluence. After confluence, medium was refreshed, and 30 μM PNF1 or 0.6% dimethyl sulfoxide (DMSO) vehicle control was added to the sample. Total RNA from the cultures was isolated 1, 2, 4, 8, 16, 24 and 48 h post-supplementation using an RNeasy kit (Qiagen, Inc., Valencia, CA, USA) according to the manufacturer's protocol.

2.2 Gene array and analysis

At each timepoint, RNA samples were prepared for hybridization using the GeneChip One-Cycle Target Labeling and Control Reagents kit (Affymetrix, Santa Clara, CA, USA), as previously reported (Wieghaus et al., 2007). Total RNA integrity was assessed by analysis with an Agilent Bioanalyzer using RNA 6000 Nano chips (Agilent Technologies, Santa Clara, CA, USA). Subsequently, biotin-labeled cRNA was generated from the total RNA samples and fragmented in preparation for array hybridization, according to standard Affymetrix GeneChip protocols. Fragmented cRNA (10 μg) was hybridized to Affymetrix Human Genome U133 Plus 2.0 probe arrays for 16 h. The arrays were washed and stained in an Affymetrix Automated Fluidics Station 400 and scanned with the Affymetrix GeneArray Scanner. Scanned images were examined for visible defects and checked for proper grid alignment, and acceptable image files were analyzed to generate raw data in the form of ‘CEL’ files. Quality control, including replicate analysis, measurement of noise and background levels, and assessment of the 3′/5′ ratio of glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and beta-actin (β-actin) was performed with Microarray Analysis Suite 5.0 (MAS 5.0, Affymetrix).

Expression profiles were generated and analyzed as previously described (Wieghaus et al., 2007). Briefly, using MAS 5.0, each gene transcript was detected as ‘present’, ‘absent’ or ‘marginal’, and the change between the experimental (PNF1-stimulated) and baseline (control) conditions at each timepoint was assigned to one of several standard categories, i.e. increase, marginal increase, no change, marginal decrease and decrease, depending on the P-value (calculated using the Wilcoxon signed rank test). The ranges for each call were 0.0000–0.0025, 0.0025–0.0030, 0.0030–0.9970, 0.9970–0.9975 and 0.9975 to 1.0000, respectively. These data were then inputted into dChip version 1.3 (Li and Wong, 2001), where they were normalized using the invariant set approach. The normalized expression data were used for the network analysis described below. The expression data are available in the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) (Wieghaus et al., 2008b).

2.3 Differentially expressed gene transcripts

The gene transcripts identified as differentially (up or down) regulated by microarray analysis at each timepoint were evaluated with further network and gene ontology analyses as previously reported (Wieghaus et al., 2007, 2008b). For these analyses, we employed the Ingenuity Pathway Analysis (IPA) software (Ingenuity® Systems, Redwood City, CA, www.ingenuity.com), including the Ingenuity Pathways Knowledge Base (IPKB), although any literature-derived pathway database [e.g. the publicly available Kyoto Encyclopedia of Genes and Genomes (KEGG) (http://www.genome.jp/kegg/)] may be utilized. IPA has been used recently to investigate genetic pathways and cell and tissue culture mechanisms (Brey et al., 2005). Graphical representations of the molecular relationships between gene transcripts were generated by IPA to further investigate the significantly regulated networks in the case of PNF1 stimulation of HMVEC (Wieghaus et al., 2007, 2008b). Basic network characteristics, including most highly connected and most differentially regulated nodes, were previously analyzed and reported (Wieghaus et al., 2007, 2008b). In addition, as described below, a pathway-based compendium analysis evaluating how known angiogenic pathways are affected by the drug over time was completed.

2.4 Pathway compendium analysis

In order to probe differential regulation of genes associated with key angiogenic functions, the IPKB was utilized to associate each significant (up or down) gene product with one or more of the following angiogenic pathways: angiopoietin 1 (Ang1), chemokine (C-C motif) ligand 2 (CCL2), basic fibroblast growth factor (bFGF), platelet-derived growth factor (PDGF), placental growth factor (PGF), TGF-β, TNF-α and vascular endothelial growth factor (VEGF) (Conway et al., 2001). Importantly, as PNF1 has been observed to induce angiogenic activity in vitro (Wieghaus et al., 2006) and more recently in vivo (Wieghaus et al., 2008a, 2008b), these eight pathways were selected on the basis of their known involvement in angiogenesis [see (Conway et al., 2001) for further details]. Future extension of our pathway-based compendium analysis approach to other drugs with unknown mechanisms would require selection of pathways relevant to the hypothesized functional effects of those drugs. For each angiogenic pathway, the percentage of differentially expressed genes associated with that pathway (out of the total number of genes contained within that pathway within the IPKB) was plotted, classified as genes that were either regulated by, or involved in the regulation of, that pathway (Fig. 3A).

Fig. 3.

Fig. 3.

Open in a new tab

Identification of involvement of TNF-α and TGF-β pathways in PNF1 activity using a pathway-based compendium analysis. (A) The results of the pathway-based compendium analysis when pathway data are extracted from the IPKB. The percentages of gene transcripts in the top-scoring IPA networks (as generated by comparing PNF1-treated versus control expression profiles) that either regulate or are regulated by eight known angiogenic factors, namely Ang1, CCL2, bFGF, PDGF, PGF, TGF-β, TNF-α and VEGF, are displayed. The TNF-α and TGF-β pathways have the highest percentages of genes with regulatory interactions. (B) The results of the pathway-based compendium analysis when pathway data are extracted from the NCI PID. Among six known angiogenic factors contained within the NCI PID, namely Ang1, bFGF, PDGF, TGF-β, TNF-α and VEGF, the TNF-α and TGF-β pathways have the highest percentages of genes with regulatory interactions.

In addition, to illustrate the applicability of our method across multiple data sources, including publicly available data, we repeated our pathway-based compendium analysis on pathway data extracted from the National Cancer Institute (NCI) Pathway Interaction Database (PID). Here, we considered a subset of the angiogenic pathways, as only six of the eight angiogenic factors in the IPKB were also included in the PID: Ang1, bFGF, PDGF, TGF-β, TNF-α and VEGF. Importantly, for each pathway, we selected the core components and interactions unique to that pathway and implicated in angiogenesis based on supporting evidence in the literature (see Supplementary Material). Again, the percentages of differentially expressed (up or down) genes associated with each of these six angiogenic pathways within the PID were plotted, normalized to the total number of genes delineated in each pathway according to the PID (Fig. 3B). Notably, the pathway-based compendium analysis is strictly an overlaying of regulated (up or down) genes on top of known biological pathways and relationships. Although a pathway knowledge database is necessary, no sophisticated and computationally intensive (in time and space) software is required.

Subsequently, to further probe the regulation of genes associated with functions of TNF-α and TGF-β, the two pathways observed to be most regulated by drug treatment via our pathway-based compendium analysis, genes integral to these pathways were identified. Specifically, genes associated with the TGF-β and TNF-α signaling pathways were obtained from KEGG and the BioCarta Proteomic Pathway Project (http://www.biocarta.com/genes/index.asp), respectively (Figs 4A and 5A, respectively). For each gene, the fold change from the microarray analysis was plotted as a function of time (Figs 4B and 5B).

Fig. 4.

Fig. 4.

Open in a new tab

Regulation of the TNF-α pathway by PNF1. (A) The members of the TNF-α pathway that were significantly regulated by PNF1 over at least one timepoint (spanning 1, 2, 4, 8, 16, 24 and 48 h post-supplementation). These renderings are based on available literature on these pathways, as taken from the BioCarta Proteomic Pathway Project (http://www.biocarta.com/genes/index.asp) and supplemented by the IPKB. Shaded gene products represent PNF1-stimulated genes; gene products not shaded (white) were not significantly regulated by the drug but are required for signaling between PNF1-regulated genes. (B) Plots of the fold changes over time for the individual gene products in the TNF-α pathway that were significantly differentially expressed after PNF1 stimulation.

Fig. 5.

Fig. 5.

Open in a new tab

Regulation of the TGF-β pathway by PNF1. (A) The core TGF-β pathway members, as rendered from KEGG (http://www.genome.jp/kegg/), that were significantly regulated by PNF1 over at least one timepoint (1–48 h). Shaded gene products represent PNF1-stimulated genes; gene products not shaded (white) were not significantly regulated by the drug but are required for signaling between PNF1-regulated genes. (B) Plots of the fold changes over time for the individual gene products in the TGF-β pathway that were significantly differentially expressed after PNF1 stimulation.

3 RESULTS

3.1 Pathway compendium analysis

The results of our novel pathway-based compendium analysis are illustrated in Figure 3. As described in the Methods section, differentially expressed genes were associated with eight pathways commonly associated with angiogenesis (Conway et al., 2001) and contained within the IPKB (i.e. Ang1, CCL2, bFGF, PDGF, PGF, TGF-β, TNF-α and VEGF) (Fig. 3A), as well as six of these angiogenic pathways contained within the NCI PID (i.e. Ang1, bFGF, PDGF, TGF-β, TNF-α and VEGF) (Fig. 3B). In the case of our IPKB-based compendium analysis, genes associated with the angiogenic pathways were further classified according to whether they were regulated by, or were involved in the regulation of, the eight angiogenic pathways. As the resolution of the pathway data within the NCI PID is not as detailed, only the association of differentially expressed genes with the angiogenic pathways was tracked in the NCI PID-based compendium analysis. The numbers of genes associated with each pathway are represented as a percentage of all genes associated with the pathway within the knowledge database to normalize for different numbers of expressed gene products within the literature-derived databases (Fig. 3A and B). More genes were associated with the TNF-α and TGF-β pathways than any other angiogenic pathway at every timepoint across both pathway data sources. Additionally, while the TNF-α pathway dominates at earlier timepoints (particularly 1 and 2 h post-supplementation), the TGF-β pathway is implicated at later timepoints (particularly 24 h). These data indicate that the TNF-α and TGF-β pathways may be mechanisms through which PNF1 stimulates microvascular remodeling. Importantly, when a set of genes on the microarray was selected at random and the pathway-based compendium analysis repeated, the results (see Supplementary Material) were vastly different. Using the NCI PID pathway data, the VEGF pathway dominates in this random gene set, and no genes in the TNF-α pathway are present. This separate analysis further strengthens the results from our novel approach, as it illustrates how the prevalence of TNF-α and TGF-β pathways within the differentially expressed gene sets across the seven timepoints we considered is unique.

3.2 Identification of PNF1-regulated gene products in the TNF-α and TGF-β canonical pathways

The differential expression levels of all members of the TNF-α and TGF-β networks were subsequently examined (see complete listing in Supplementary Material). Transcripts from the TNF-α pathway with significant fold changes (>|2|) for at least one timepoint are diagrammed in the pathway map shown in Figure 4A, and their fold changes over control cultures are graphed in Figure 4B, classified as extracellular signals, receptors and intracellular signaling (subcategorized into caspases, mitogen-activated protein (MAP) kinases and others). The expression of TNF-α itself and TNF receptor 1A (TNFR1) were affected at 8 and 24 h, and 4 h, respectively. Additionally, various signal transduction molecules within the TNF-α pathway, including caspase 2 (CASP2), caspase 3 (CASP3) and MAP kinases, were affected by drug treatment at multiple timepoints.

Similarly, changes in the expression of TGF-β signaling pathway components by PNF1 over 1–48 h were also evident (Fig.5A and B). Extracellular signals associated with the pathway were differentially expressed, including thrombospondin (THBS1) after 48 h and multiple transcripts of decorin (DCN) at a total of five timepoints, i.e. 1, 4, 16, 24 and 48 h. Transforming growth factor-β1 (TGFB1) was significantly affected at 1, 4, 16 and 24 h, which verifies previous experiments implicating TGF-β in the PNF1 mechanism after 24 h (Wieghaus et al., 2007). We previously hypothesized that differential regulation of accessory TGF-β receptors, betaglycan (TGFBRIII) and endoglin (ENG), after 24 h may provide important insight into drug mechanism (Wieghaus et al., 2007) as their interplay is integral for balance of pro- and anti-angiogenic signaling (Wong et al., 2000). Here, investigation of these receptors at additional timepoints revealed differential expression of ENG at 16 and 48 h. Many TGF-β-related intracellular signaling molecules and transcription factors were also significantly affected. Hence, signaling molecules in the TNF-α and TGF-β pathways are differentially expressed by PNF1, as suggested by our novel pathway-based compendium analysis.

4 DISCUSSION

PNF1 is being developed as a novel stimulator of angiogenesis for regenerative medicine applications and treatment of ischemic diseases. Characterizing mode-of-action is a key challenge in commercializing PNF1 and drug development in general. Previously, we demonstrated that PNF1 promotes cellular processes relevant to angiogenesis in human endothelial cells in vitro (Wieghaus et al., 2006), that it stimulates robust microvascular remodeling in vivo (Wieghaus et al., 2008a), and that networks of transcripts (including many TGF-β-related) integral to the angiogenic process were upregulated 24 h after PNF1 application to endothelial cultures (Wieghaus et al., 2007). In addition, we recently expanded the transcriptional profile of PNF1 by performing comparative microarray analyses on PNF1-stimulated (versus control) HMVEC over 1–48 h (Wieghaus et al., 2008b). These data facilitated the identification of MT1-MMP modulation by PNF1. MT1-MMP activity is induced by TGF-β and linked to endothelial invasion in type I collagen (Cheng and Lovett, 2003). Here, we developed and applied a novel pathway-based compendium analysis to our 1–48 h transcriptional regulatory footprint of PNF1 activity in HMVEC to specifically identify the molecular mechanism by which the drug functions. We demonstrate through this compendium analysis that PNF1 stimulation in its most effective in vitro concentration (30 μM) (Wieghaus et al., 2006) produces a significant change in expression of HMVEC genetic networks and related cell signaling processes, particularly TNF-α and TGF-β pathways. These changes are highly consistent with pro-angiogenic activity.

Our unique method for angiogenic compendium analysis suggests that PNF1 predominantly affects transcription of genes associated with the functions of TNF-α and TGF-β over 1–48 h post-stimulation: TNF-α-related transcripts are important at earlier timepoints, while TGF-β-related signaling molecules are more influential at later timepoints. These experiments validate our previous observations implicating the TGF-β pathway, mechanistically, after 24 h (Wieghaus et al., 2007). Furthermore, our results agree with independent literature findings, in which increasing evidence implicates TNF-α induction of TGF-β through the ERK-specific MAPK pathway (Sullivan et al., 2005). Many individual gene products critical to their canonical pathways were differentially expressed by PNF1 over at least one timepoint between 1 and 48 h drug stimulation; these results served to validate the results of the novel compendium analysis and also allowed for isolation of a definitive number of TNF-α- and TGF-β-related transcripts for future probing of PNF1 mechanism.

As we are interested in identifying the mechanism of a drug that exhibits robust angiogenic and arteriogenic responses in vivo, it is significant that our analyses implicated two signaling pathways (TNF-α and TGF-β) with known activities in both forms of microvascular remodeling. While these molecules are both implicated as important growth factor signals in disease paradigms like inflammation and artherogenesis, it is important to note that the processes of arteriogenesis and inflammation share key events (monocyte recruitment, cell proliferation and migration) and are mediated by similar sets of growth factors, including TNF-α and TGF-β (Schaper and Buschmann, 1999; Van Royen et al., 2001). Indeed, inflammation may be the most critical stimulus for vessel growth, especially during tissue ischemia (Silvestre et al., 2008). TNF-α creates the inflammatory milieu necessary for arteriogenesis (Van Royen et al., 2001). Although there is potential for inflammatory side effects of arteriogenic therapies, it has been hypothesized that these effects may be largely diminished through local delivery of the agents (Van Royen et al., 2001), such as from PLAGA matrices. Indeed, initial experiments conducted to determine the extent of normal vascular pattern retention after PNF1 stimulation, measured via fractal dimension analysis, demonstrate that PNF1 stimulates marked microvascular remodeling and expansion with significantly increased regularity of vascular tissue patterning (Wieghaus et al., 2008a).

We now hypothesize a TGF-β-related mechanism of action for PNF1 (Fig. 6). Specifically, activation of endothelial cells by the novel compound results in activation of TNF-α, which recruits monocytes and promotes arteriogenesis and angiogenesis through local inflammatory responses (Lamb et al., 2006). TNF-α then induces TGF-β, which further stimulates angiogenesis and arteriogenesis in vivo through PDGF and bFGF (Van Royen et al., 2001) and mural cell differentiation (Chambers et al., 2003). TGF-β has been demonstrated to activate marked arteriogenesis in ischemic hindlimbs, and to induce MCP-1 and monocyte chemotaxis (Wahl et al., 1987; Wiseman et al., 1988). This potential mechanism parallels significant angiogenic and arteriogenic responses seen after PNF1 application in vivo and incorporates the substantial regulation of TNF-α and TGF-β pathways highlighted by our novel compendium analysis. This mechanism also parallels previous results implicating MT1-MMP as a downstream effector of PNF1 stimulation (Wieghaus et al., 2008b), as both TNF-α and TGF-β are well-known to modulate MT1-MMP signaling (Cheng and Lovett, 2003; Sternlicht and Werb, 2001). Future work exploring the precise (up or down) regulation of individual components of the TNF-α and TGF-β pathways, as illustrated in Figures 4A and 5A, respectively, might further characterize PNF1 mode-of-action. Indeed, extension of this work has led to identification of modulation of MT1-MMP by PNF1 [see Wieghaus et al. (2008b) for details]. As described above, MT1-MMP activity is induced by TGF-β and linked to endothelial invasion in type I collagen (Cheng and Lovett, 2003).

Fig. 6.

Fig. 6.

Open in a new tab

Known TNF-α and TGF-β signaling of angiogenesis, and putative pathways by which PNF1 functions. We hypothesize that PNF1 activates endothelial cells and stimulates their production of TNF-α, an angiogenic cytokine that also recruits monocytes and promotes arteriogenesis. Additionally, TNF-α stimulates production of TGF-β, which stimulates both angiogenesis and arteriogenesis in vivo (Wahl et al., 1987). Evaluation of PNF1 effects in CCL2 receptor mouse chimeras (CCR2−/−) recently suggested that monocyte recruitment is critical for PNF1 efficacy (Wieghaus et al., 2008a).

Ultimately, the microarray dataset coupled with network analysis represents a novel form of compendium analysis. Previous compendium analyses have probed mechanistic detail by comparing expression profiles of multiple stimulants (Butcher and Schreiber, 2005), i.e. in the case of PNF1, collecting microarray data for PNF1 and known pro-angiogenic factors, identifying the factor with an expression profile that most correlates with PNF1 via clustering methods, and associating the function of this factor with that of PNF1. By instead generating a comprehensive PNF1 transcriptional profile and coupling this dataset with network analysis, including a pathway knowledge database, significant mechanistic detail was elucidated much more efficiently. These data allowed for discovery of earlier, upstream signals (e.g. the TNF-α pathway) that further stimulate TGF-β-related gene products at 24 h and other later timepoints; these results would not have been revealed without the extension of our previous analysis to timepoints other than 24 h or our novel method for compendium analysis to extract patterns of gene expression information from multiple transcriptional profiles using pathway data. Importantly, our findings were consistent across two different sources of pathway knowledge, namely the IPKB as well as the publicly accessible NCI PID. Consequently, although our results and conclusions here are based on these databases, the pathway-based compendium analysis may be extendible to any literature-derived pathway knowledge database, including other publicly accessible databases such as KEGG. In addition, this pathway-based compendium analysis can be applied in the future to other novel compounds with unknown mechanism, simply by overlaying the expression profiles of those compounds on pathways that are hypothesized to be involved in their functional outcomes.

Funding

National Institute of Arthritis and Musculoskeletal Disease (K01AR052352-01A1 to E.A.B.); US Department of Education [Graduate Assistance in Areas of National Need (GAANN) to K.A.W.]; and National Institutes of Health (GM08715/Biotechnology Training Grant to E.P.G.).

Conflict of Interest: none declared.

Supplementary Material

[Supplementary Data]
btn451_index.html (891B, html)

REFERENCES

  1. Bader JS, et al. Gaining confidence in high-throughput protein interaction networks. Nat. Biotechnol. 2004;22:78–85. doi: 10.1038/nbt924. [DOI] [PubMed] [Google Scholar]
  2. Brey EM, et al. Therapeutic neovascularization: contributions from bioengineering. Tissue Eng. 2005;11:567–584. doi: 10.1089/ten.2005.11.567. [DOI] [PubMed] [Google Scholar]
  3. Butcher RA, Schreiber SL. Using genome-wide transcriptional profiling to elucidate small-molecule mechanism. Curr. Opin. Chem. Biol. 2005;9:25–30. doi: 10.1016/j.cbpa.2004.10.009. [DOI] [PubMed] [Google Scholar]
  4. Chambers RC, et al. Global expression profiling of fibroblast responses to transforming growth factor-beta1 reveals the induction of inhibitor of differentiation-1 and provides evidence of smooth muscle cell phenotypic switching. Am. J. Pathol. 2003;162:533–546. doi: 10.1016/s0002-9440(10)63847-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Cheng S, Lovett DH. Gelatinase A (MMP-2) is necessary and sufficient for renal tubular cell epithelial-mesenchymal transformation. Am. J. Pathol. 2003;162:1937–1949. doi: 10.1016/S0002-9440(10)64327-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Conway EM, et al. Molecular mechanisms of blood vessel growth. Cardiovasc. Res. 2001;49:507–521. doi: 10.1016/s0008-6363(00)00281-9. [DOI] [PubMed] [Google Scholar]
  7. Kelley BP, et al. Conserved pathways within bacteria and yeast as revealed by global protein network alignment. Proc. Natl Acad. Sci. USA. 2003;100:11394–11399. doi: 10.1073/pnas.1534710100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Kelley BP, et al. PathBLAST: a tool for alignment of protein interaction networks. Nucleic Acids Res. 2004;32:W83–W88. doi: 10.1093/nar/gkh411. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Lamb J, et al. The Connectivity map: using gene-expression signatures to connect small molecules, genes, and disease. Science. 2006;313:1929–1935. doi: 10.1126/science.1132939. [DOI] [PubMed] [Google Scholar]
  10. Li C, Wong WH. Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc. Natl Acad. Sci. USA. 2001;98:31–36. doi: 10.1073/pnas.011404098. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Madeddu P. Therapeutic angiogenesis and vasculogenesis for tissue regeneration. Exp. Physiol. 2005;90:315–326. doi: 10.1113/expphysiol.2004.028571. [DOI] [PubMed] [Google Scholar]
  12. Schaper W, Buschmann I. Arteriogenesis, the good and bad of it. Cardiovasc. Res. 1999;43:835–837. doi: 10.1016/s0008-6363(99)00191-1. [DOI] [PubMed] [Google Scholar]
  13. Sharan R, et al. Conserved patterns of protein interaction in multiple species. Proc. Natl Acad. Sci. USA. 2005;102:1974–1979. doi: 10.1073/pnas.0409522102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Silvestre JS, et al. Post-ischaemic neovascularization and inflammation. Cardiovasc Res. 2008;78:242–249. doi: 10.1093/cvr/cvn027. [DOI] [PubMed] [Google Scholar]
  15. Sternlicht MD, Werb Z. How matrix metalloproteinases regulate cell behavior. Annu. Rev. Cell. Dev. Biol. 2001;17:463–516. doi: 10.1146/annurev.cellbio.17.1.463. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Subramanian A, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci. USA. 2005;102:15545–15550. doi: 10.1073/pnas.0506580102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Sullivan DE, et al. Tumor necrosis factor-alpha induces transforming growth factor-beta1 expression in lung fibroblasts through the extracellular signal-regulated kinase pathway. Am. J. Respir. Cell. Mol. Biol. 2005;32:342–349. doi: 10.1165/rcmb.2004-0288OC. [DOI] [PubMed] [Google Scholar]
  18. Taylor RD, et al. A review of protein-small molecule docking methods. J. Comput. Aided Mol. Des. 2002;16:151–166. doi: 10.1023/a:1020155510718. [DOI] [PubMed] [Google Scholar]
  19. Van Royen N, et al. Arteriogenesis: mechanisms and modulation of collateral artery development. J. Nucl. Cardiol. 2001;8:687–693. doi: 10.1067/mnc.2001.118924. [DOI] [PubMed] [Google Scholar]
  20. Wahl SM, et al. Transforming growth factor type beta induces monocyte chemotaxis and growth factor production. Proc. Natl Acad. Sci. USA. 1987;84:5788–5792. doi: 10.1073/pnas.84.16.5788. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Wieghaus KA, et al. Small molecule inducers of angiogenesis for tissue engineering. Tissue Eng. 2006;12:1903–1913. doi: 10.1089/ten.2006.12.1903. [DOI] [PubMed] [Google Scholar]
  22. Wieghaus KA, et al. Mechanistic exploration of phthalimide neovascular factor 1 using network analysis tools. Tissue Eng. 2007;13:2561–2575. doi: 10.1089/ten.2007.0023. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Wieghaus KA, et al. Expansion of microvascular networksin vivoby novel therapeutic phthalimide neovascular factor 1 (PNF1) Biomaterials (in press) 2008a doi: 10.1016/j.biomaterials.2008.08.029. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Wieghaus KA, et al. Phthalimide neovascular factor 1 (PNF1) modulates MT1-MMP activity in human microvascular endothelial cells. Los Angeles, CA: Biomedical Engineering Society (BMES) Annual Fall Meeting; 2008b. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Wiseman DM, et al. Transforming growth factor-beta (TGF beta) is chemotactic for human monocytes and induces their expression of angiogenic activity. Biochem. Biophys. Res. Commun. 1988;157:793–800. doi: 10.1016/s0006-291x(88)80319-x. [DOI] [PubMed] [Google Scholar]
  26. Wong SH, et al. Endoglin expression on human microvascular endothelial cells association with betaglycan and formation of higher order complexes with TGF-beta signalling receptors. Eur. J. Biochem. 2000;267:5550–5560. doi: 10.1046/j.1432-1327.2000.01621.x. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

[Supplementary Data]
btn451_index.html (891B, html)
btn451_1.pdf (141KB, pdf)
btn451_2.pdf (147KB, pdf)
btn451_bioinf-2008-0837-File011.doc (95KB, doc)

Articles from Bioinformatics are provided here courtesy of Oxford University Press

RESOURCES