**5. DNA microarray technology**

Practically, in the exploring stage, the expression of ~40,000 gene spots and replicates can be simultaneously analyzed on a couple of glass array in a single experiment by means of

To pursue the genomic impacts of any gene based medicine, it is necessary to exploit high throughput screening methodologies (e.g., DNA microarray) for evaluation of global gene changes induced by the gene medicine or any other chemicals/compounds. Such genome

The DNA microarray technology combines standard molecular techniques with highthroughput screening to monitor the expression of up to ~40000 genes, which may provide a means for toxicity prediction prior to classical toxicological endpoints such as histopathology or clinical chemistry (Goldsmith & Dhanasekaran, 2004). In gene silencing experiments, such approach may allow a genomic characterization of delivery systems leading to identification of possible incompatibilities with intended target genes or biological effects of the gene based medicine. This may allow screening of compatible or useful delivery systems early in drug development that could subsequently save time and money in pre-clinical and clinical studies (Fielden & Kolaja,

Cytotoxicity and genotoxicity potentials of CPs and CLs are going to be well acknowledged, and accordingly these cationic nanosystems should undergo a rigorous genocompatibility evaluation prior to *in vitro* and *in vivo* exploitation (Kabanov, 2006; Omidi et al., 2005a). These systems alone or in combination with biologically active molecules (e.g., siRNA, antisense, aptamer) are able to alter cell signaling and biological responses in cells and organisms, emerging a cluster of genomic and post genomic consequences. In general, toxic responses to these kinds of nanomaterials are deemed to be very profound, in which various signaling pathways such as oxidative stress, immune responses and apoptosis pathways may be involved in response to generation of reactive oxygen species in the membranes (Kabanov, 2006). Cationic liposomes, irrespective of complexation with DNA, can downregulate the synthesis of pro-inflammatory mediators such as nitric oxide (NO) and tumor necrosis factor-alpha (TNF-alpha) in lipopolysaccharide (LPS)/interferon-gamma (IFN-gamma)-activated macrophages (Filion & Phillips, 1997a; Filion & Phillips, 1997b). Under the oxidative stress, cells may undergo the Nrf-2 signaling or the proinflammatory signaling cascades such as mitogen-activated protein kinase (MAPK) and nuclear factor kB (NFkB) cascades and eventually a programmed cell death may occur (Kabanov, 2006). Certain proteins such as protein kinase C (PKC) may also be affected detrimentally by cationic amphiphiles (Aberle et al., 1998), which function as PKC inhibitors and may inevitably result in inadvertent toxicity. It seems that the cationic amphiphiles with steroid backbones can exert more potent inhibitors of PKC than their straight-chain analogues, resulting in greater toxic impacts (Bottega & Epand, 1992). Polycations such as PEI formulated with plasmid DNA and administered to mouse lungs was reported to activate the p38 pathway involved in endocytosis, phagocytosis and hydrogen peroxide production. The observed *in vitro* and *in vivo* toxicity of such PEI polyplex formulations appeared to link to a general stress reaction, inflammatory responses, cell cycle regulation and DNA damage repair (Regnstrom et al., 2006). To obtain a complete image, it is essential to recruit high throughput screening methods such

Practically, in the exploring stage, the expression of ~40,000 gene spots and replicates can be simultaneously analyzed on a couple of glass array in a single experiment by means of

**4. Genocompatibility and toxicogenomics of polycationic nanostructures** 

based impact could be termed as "genotoxicity" or "toxicogenomics".

2006; Lettieri, 2006).

as DNA microarray.

**5. DNA microarray technology** 

microarray technology. However, for accomplishment of a significant correlation between the gene expression profiles and their functionality expression, it is important to implement substantial complementary investigations to verify the results at the molecular level and as a result extend our understanding of gene expression patterns and molecular pathways.

Microarray technology can be exploited to attain a wealth of data that can be used to develop a more complete understanding of gene expression, which can be used for transcriptional regulation and interactions as well as functional genomics. Despite its successful *in vitro* cell-based implementation, application of this technology for *in vivo*  investigations is deemed to be more sophisticated because of complexity of cytotoxicity and genotoxicity studies, which can be confounded by a number of variables such as type of target organ, effect of pharmacokinetics and/or pharmacodynamics parameters (Lobenhofer et al., 2001). Since its advent and application in life sciences, microarray has been widely applied for molecular/biological studies. In fact, a large number of indexed articles in various data banks (e.g., MEDLINE/PubMed) highlight the importance of microarray technology in post-genomics era.

Fig. 3 shows a schematic illustration of step-wise processes of the DNA microarray technology.

Technically, DNA microarray can be generated in two different types including printing pre-synthesized cDNAs (500–2000 bp) or synthesizing short oligonucleotides (20–50 bases) onto glass microscope slides, in which gene spots include either fully sequenced genes of known function or collections of partially sequenced cDNA derived from expressed sequence tags (ESTs) corresponding to the messenger RNAs of unknown genes. For example in practice, one may compare two different cells/tissues from untreated (UT) versus treated (T). For gene expression profiling, normally total RNA is extracted from the untreated and treated samples. Using an indirect labeling methodology, they are converted to labeled cDNA (e.g., with aminoallyle-dUTP). The aminoallyle-dUTP-cDNA is then labeled with cyanine dye (e.g., Cy3 or Cy5). The Cy3 and Cy5 labeled aminoallyle-dUTPcDNA from UT and T samples are hybridized on a single glass array, which is subjected to several washing steps, scanning with an appropriate scanner (e.g., using RS Reloaded™, TECAN, Switzerland) and data mining (e.g., using GeneMath™ software; Applied Maths, Sint-Martens-Lathem, Belgium); for detailed information reader is directed to see (Hegde et al., 2000; Omidi et al., 2005b; Omidi et al., 2008).

For microarray analysis, significantly upregulated and/or downregulated genes can be identified using traditional method (gene expression changes with a fixed cutoff threshold usually in 2 fold) to infer significance differences (i.e., the so called "fold change method"). The resultant data are normally presented as scatter plots of treated (T) versus untreated (UT) control. To reach this stage, data need to undergo a number of processes called as "transformation" and "normalization" to minimize the experimental erroneousness (i.e., the so called "data mining"). Since a scatter plot of T versus UT genes would cluster along a straight line, normalization of this type of data is equivalent to calculating the best-fit slope using regression techniques and adjusting the intensities so that the calculated slope is one. In many experiments, the intensities are nonlinear, and local regression techniques are more suitable, such as Locally WEighted Scatterplot Smoothing (LOWESS) regression (Berger et al., 2004; Chen et al., 2003).

In our studies, we have successfully exploited both approaches to study the impacts of the nonviral vectors (CPs and CLs based formulation) on global gene expression experiments. To get the significant alterations in gene expression, we rejected the arrays showing non-

Toxicogenomics of Nonviral Cationic Gene Delivery Nanosystems 557

equal intensity or variable intensity of control gene spots in replicates on the same slide or between slides in dye-flipping experiments (Hollins et al., 2007; Omidi et al., 2003; Omidi et al., 2005b; Omidi et al., 2008). Data for each gene were typically reported as an "expression ratio" or as the base 2 logarithm (log2) of the expression ratio of T to UT control. Genes were assumed to be up regulated or downregulated if they revealed an expression ratio of >2 and

Based on our findings, the starburst PAMAM dendrimer alone or as complexed with DNA can elicit inadvertent gene expression changes. We also found that the linear and branched PEI (25 kDa) are able to induce gene expression changes in A431 cells, as shown in Fig. 4

Fig. 4. Scatter plots of gene expression changes induced by cationic linear (A) and branched (B) PEI (25 kDa) in A431 cells. Data represent Log2 transformed gene expression values for large arrays housing 20000 genes. Above 2-fold change in expression of treated to untreated is indicated by bold circles and unchanged genes by unfilled circles. Panel C represents gene expression changes ratio between untreated A431 cells from different experiment. BPEI: branched polyethylenimine; LPEI: linear polyethylenimine (our unpublished data produced

In the case of arrays with thousands of spots, one needs to employ the "feature reduction" or "dimension reduction" to find the minimum number of the features (i.e., genes or maybe even the conditions) that can best describe the data and the classification using statistical methods such as principal component analysis (PCA), correspondence analysis (CA), multidimensional scaling (MDS), and cluster analysis, reader is directed to see the following citation (Hegde et al., 2000; Quackenbush, 2001; Quackenbush, 2002). Of the dimension reduction methods, PCA is the most widely used method as a tool in exploratory data analysis, which involves a mathematical procedure that transforms a number of possibly correlated variables into a smaller number of uncorrelated variables called principal components. PCA ignores the dimensions in which data do not vary significantly and it is

<0.5 (or >1 and <-1 for log2 transformed data), respectively.

(our unpublished data).

by Omidi et al.).

closely related to factor analysis.

Fig. 3. Schematic illustration of step-wise process of DNA microarray methodology.

Fig. 3. Schematic illustration of step-wise process of DNA microarray methodology.

equal intensity or variable intensity of control gene spots in replicates on the same slide or between slides in dye-flipping experiments (Hollins et al., 2007; Omidi et al., 2003; Omidi et al., 2005b; Omidi et al., 2008). Data for each gene were typically reported as an "expression ratio" or as the base 2 logarithm (log2) of the expression ratio of T to UT control. Genes were assumed to be up regulated or downregulated if they revealed an expression ratio of >2 and <0.5 (or >1 and <-1 for log2 transformed data), respectively.

Based on our findings, the starburst PAMAM dendrimer alone or as complexed with DNA can elicit inadvertent gene expression changes. We also found that the linear and branched PEI (25 kDa) are able to induce gene expression changes in A431 cells, as shown in Fig. 4 (our unpublished data).

Fig. 4. Scatter plots of gene expression changes induced by cationic linear (A) and branched (B) PEI (25 kDa) in A431 cells. Data represent Log2 transformed gene expression values for large arrays housing 20000 genes. Above 2-fold change in expression of treated to untreated is indicated by bold circles and unchanged genes by unfilled circles. Panel C represents gene expression changes ratio between untreated A431 cells from different experiment. BPEI: branched polyethylenimine; LPEI: linear polyethylenimine (our unpublished data produced by Omidi et al.).

In the case of arrays with thousands of spots, one needs to employ the "feature reduction" or "dimension reduction" to find the minimum number of the features (i.e., genes or maybe even the conditions) that can best describe the data and the classification using statistical methods such as principal component analysis (PCA), correspondence analysis (CA), multidimensional scaling (MDS), and cluster analysis, reader is directed to see the following citation (Hegde et al., 2000; Quackenbush, 2001; Quackenbush, 2002). Of the dimension reduction methods, PCA is the most widely used method as a tool in exploratory data analysis, which involves a mathematical procedure that transforms a number of possibly correlated variables into a smaller number of uncorrelated variables called principal components. PCA ignores the dimensions in which data do not vary significantly and it is closely related to factor analysis.

Toxicogenomics of Nonviral Cationic Gene Delivery Nanosystems 559

Hierarchical clustering plot**.** The algorithm used subjects the expression intensity ratio of treated versus untreated samples to single-linkage Hierarchical clustering (by means of Euclidean distance metric) analyses in order to arrange each gene with its related group members exhibiting a similar ratio of change in expression. We have shown that some overexpressed - or underexpressed genes display not only a similar pattern of expression but also a related cellular functionality and themes (e.g. apoptotic related genes) (Omidi et al., 2003; Omidi et al., 2005a). Such Hierarchical clustering maybe considered as a "genomic

Taken all these facts together, surprisingly, still little information is available upon specific genomic effects elicited by chemicals within various cells/tissues despite implementing the "omics" technology for discovery of intrinsic genomic signature of chemicals/compounds in various targets. As a result, extensive investigations are yet to be performed to get sufficient information on genetic-signature of chemical and pharmaceuticals in target cells/tissues. Accordingly, many individuals and some organizations have attempted to accomplish such aim. For example, the Comparative Toxicogenomics Database (CTD) is a useful platform providing insights into complex chemical–gene and protein interaction networks (http://ctd.mdibl.org/about) that can be used for successfully advancement of

To date, cationic lipids have been the most widely used delivery system for delivery of nucleic acids both *in vitro* and *in vivo*. For example, Lipofectin™ is the 1:1 mixture of DOTMA and DOPE. It is the first cationic lipid formulation that was received widespread attention. We found that cationic liposomes such as LF and OF, at concentrations routinely used to obtain efficient delivery of gene based medicines, were able to induce gene expression changes in human epithelial A431 cells (Table 1). Such alterations in gene expressions appeared to be largely dependent upon the physicochemical characteristics of the lipid, wherein OF elicited greater gene expression than LF, i.e., up to 16% of the genes studied (Omidi et al., 2003). We speculate that the surface charge may play a key role in terms of such genotoxicity. In these cells, we witnessed that the affected genes were functionally involved in various cellular processes such as cell proliferation, differentiation and apoptosis. The upregulated or downregulated genes include some important genes such as bcl-2-related protein a1 (BCL2A1), caspase 8 isoform c (CASP8), heat shock protein 70 (HSP70) and 60 (HSP60), annexin a2 (ANXA2), and tubulin beta 5 (TUBB5) (Omidi et al., 2003). Up regulation of caspase-8 clearly impart activation of procaspases and caspases that may provoke activity of a series of apoptotic signaling cascades such as electron carrier protein cytochrome C, adaptor protein Apaf-1, Bcl-2 family, p53 and various transcription factors (Kanduc et al., 2002). Given that the heat shock protein 70 acts as an inhibitor of apoptosis (Li et al., 2000), it's upregulation by OF in A431 cells is deemed to be a cellular compensatory or defense response. We assume that cells recognize the xenobiotics upon their biological properties. To examine such concept, we compared OF genotoxicities within

In A549 cells, the genomic impacts were intriguingly dissimilar compared to that of A431 cells (Table 1). Further, we observed some commonalities in gene expression modulation between two different cell lines (Omidi et al., 2008). Upon EASE analyses, the changes in gene expression fell into a number of various functional genomic ontologies. For example,

signature" of any chemical.

novel pharmaceuticals.

**7. Genomic impacts of cationic lipids** 

two epithelial cell lines (i.e., A431 and A549 cells).

#### **6. Pathway analysis for functional genomics and gene ontology**

To understand the functions of the genomic changes, one needs to implement appropriate methods on knowledge extraction from DNA microarray data. Such aim can be performed by means of "pathway analysis" (PA), which should be towards functional enrichment for establishing networks between genes. In fact, understanding the expression dynamics of gene networks helps us infer innate complexities and phenomenological networks among genes. Likewise, studying the regulation patterns of genes in groups, using clustering and classification methods may help us understand different pathways in the cell, their functions, regulations and the way one component in the system affects the other one. For pathway analysis, one of the most widely used methods is comparing the gene list to a pathway which gives a *p* value as a result. Basically, such scoring enrichment methods compare a list of the genes to that of a pathway and count the hits, so that the greater the number of the hits, the greater the score and the enrichment (Curtis et al., 2005). GenMAPP is an open source package that allows users to visualize microarray and proteomics data in the context of biological pathways (freely available at http://www.genmapp.org/). It represents biological pathways in a special file format called 'MAPPs' which are independent of the gene expression data. It is used to group genes by any organizing principle (e.g., apoptosis pathways). In addition, the gene set enrichment analysis (GSEA) is a novel method that uses all the data on the microarray in the order of expression, determining whether a priori defined set of genes shows statistically significant, concordant differences between two biological states such as phenotypes (Subramanian et al., 2005). In 2003, Hosack et al. developed a powerful software named, "the Expression Analysis Systematic Explorer" (EASE), which is customizable software for rapid biological interpretation of gene lists resulted by "omics" technology such as toxicogenomics, proteomics, or other high-throughput genomic data, in particular DNA microarray gene expression profiles. In fact, the biological themes returned by EASE recapitulate manually determined themes in previously published gene lists and are robust to varying methods of normalization, intensity calculation and statistical selection of genes (Hosack et al., 2003). We have largely exploited EASE to rapidly searching the Genbank in order to find the functional 'themes' in our microarray experiments. We have found various functional themes for the upregulated or downregulated genes induced by CLs in human epithelial cells, mainly: signal transducer activity, catalytic activity, response to external stimulus, cell growth and/or maintenance, cell cycle, response to biotic stimulus, regulation of programmed cell death, humoral immune response, cellular defense response, positive regulation of biosynthesis, negative regulation of cell proliferation, regulation of interferongamma biosynthesis, transcription factor binding, DNA repair, regulation of nucleocytoplasmic transport, apoptosis, apoptosis inhibitor activity, positive regulation of apoptosis, nuclease activity, transcriptional elongation regulator activity, regulation of caspase activation, response to oxidative stress, DNA damage response, and cell-mediated immune response (Omidi et al., 2005a).

As a secondary goal of array experiments it necessitates to look for groups of genes that behave similarly across a series of treatments (i.e. clustering analysis). There are a number of methodologies for clustering that can be employed upon experimental and statistical objectives; for clustering methods see citations (Azuaje, 2003; Sturn et al., 2002; Yang et al., 2001). In our studies on toxicogenomics of gene delivery systems, we have used softwares such as GeneSight™ or GeneMath™gene expression to present data as a single linkage

To understand the functions of the genomic changes, one needs to implement appropriate methods on knowledge extraction from DNA microarray data. Such aim can be performed by means of "pathway analysis" (PA), which should be towards functional enrichment for establishing networks between genes. In fact, understanding the expression dynamics of gene networks helps us infer innate complexities and phenomenological networks among genes. Likewise, studying the regulation patterns of genes in groups, using clustering and classification methods may help us understand different pathways in the cell, their functions, regulations and the way one component in the system affects the other one. For pathway analysis, one of the most widely used methods is comparing the gene list to a pathway which gives a *p* value as a result. Basically, such scoring enrichment methods compare a list of the genes to that of a pathway and count the hits, so that the greater the number of the hits, the greater the score and the enrichment (Curtis et al., 2005). GenMAPP is an open source package that allows users to visualize microarray and proteomics data in the context of biological pathways (freely available at http://www.genmapp.org/). It represents biological pathways in a special file format called 'MAPPs' which are independent of the gene expression data. It is used to group genes by any organizing principle (e.g., apoptosis pathways). In addition, the gene set enrichment analysis (GSEA) is a novel method that uses all the data on the microarray in the order of expression, determining whether a priori defined set of genes shows statistically significant, concordant differences between two biological states such as phenotypes (Subramanian et al., 2005). In 2003, Hosack et al. developed a powerful software named, "the Expression Analysis Systematic Explorer" (EASE), which is customizable software for rapid biological interpretation of gene lists resulted by "omics" technology such as toxicogenomics, proteomics, or other high-throughput genomic data, in particular DNA microarray gene expression profiles. In fact, the biological themes returned by EASE recapitulate manually determined themes in previously published gene lists and are robust to varying methods of normalization, intensity calculation and statistical selection of genes (Hosack et al., 2003). We have largely exploited EASE to rapidly searching the Genbank in order to find the functional 'themes' in our microarray experiments. We have found various functional themes for the upregulated or downregulated genes induced by CLs in human epithelial cells, mainly: signal transducer activity, catalytic activity, response to external stimulus, cell growth and/or maintenance, cell cycle, response to biotic stimulus, regulation of programmed cell death, humoral immune response, cellular defense response, positive regulation of biosynthesis, negative regulation of cell proliferation, regulation of interferongamma biosynthesis, transcription factor binding, DNA repair, regulation of nucleocytoplasmic transport, apoptosis, apoptosis inhibitor activity, positive regulation of apoptosis, nuclease activity, transcriptional elongation regulator activity, regulation of caspase activation, response to oxidative stress, DNA damage response, and cell-mediated

As a secondary goal of array experiments it necessitates to look for groups of genes that behave similarly across a series of treatments (i.e. clustering analysis). There are a number of methodologies for clustering that can be employed upon experimental and statistical objectives; for clustering methods see citations (Azuaje, 2003; Sturn et al., 2002; Yang et al., 2001). In our studies on toxicogenomics of gene delivery systems, we have used softwares such as GeneSight™ or GeneMath™gene expression to present data as a single linkage

**6. Pathway analysis for functional genomics and gene ontology** 

immune response (Omidi et al., 2005a).

Hierarchical clustering plot**.** The algorithm used subjects the expression intensity ratio of treated versus untreated samples to single-linkage Hierarchical clustering (by means of Euclidean distance metric) analyses in order to arrange each gene with its related group members exhibiting a similar ratio of change in expression. We have shown that some overexpressed - or underexpressed genes display not only a similar pattern of expression but also a related cellular functionality and themes (e.g. apoptotic related genes) (Omidi et al., 2003; Omidi et al., 2005a). Such Hierarchical clustering maybe considered as a "genomic signature" of any chemical.

Taken all these facts together, surprisingly, still little information is available upon specific genomic effects elicited by chemicals within various cells/tissues despite implementing the "omics" technology for discovery of intrinsic genomic signature of chemicals/compounds in various targets. As a result, extensive investigations are yet to be performed to get sufficient information on genetic-signature of chemical and pharmaceuticals in target cells/tissues. Accordingly, many individuals and some organizations have attempted to accomplish such aim. For example, the Comparative Toxicogenomics Database (CTD) is a useful platform providing insights into complex chemical–gene and protein interaction networks (http://ctd.mdibl.org/about) that can be used for successfully advancement of novel pharmaceuticals.
