**NHERF1 Loss Upregulates Enzymes of the Pentose Phosphate Pathway in Kidney Cortex**

**Adrienne Bushau-Sprinkle <sup>1</sup> , Michelle T. Barati <sup>2</sup> , Kenneth B. Gagnon <sup>2</sup> , Syed Jalal Khundmiri <sup>3</sup> , Kathleen Kitterman <sup>2</sup> , Bradford G. Hill <sup>4</sup> , Amanda Sherwood <sup>2</sup> , Michael Merchant <sup>2</sup> , Shesh N. Rai 5,6 , Sudhir Srivastava 6,7, Barbara Clark <sup>8</sup> , Leah Siskind <sup>1</sup> , Michael Brier 1,2 , Jessica Hata 9,10 and Eleanor Lederer 2,11,\***


Received: 7 August 2020; Accepted: 3 September 2020; Published: 14 September 2020

**Abstract:** (1) Background: We previously showed Na/H exchange regulatory factor 1 (NHERF1) loss resulted in increased susceptibility to cisplatin nephrotoxicity. NHERF1-deficient cultured proximal tubule cells and proximal tubules from NHERF1 knockout (KO) mice exhibit altered mitochondrial protein expression and poor survival. We hypothesized that NHERF1 loss results in changes in metabolic pathways and/or mitochondrial dysfunction, leading to increased sensitivity to cisplatin nephrotoxicity. (2) Methods: Two to 4-month-old male wildtype (WT) and KO mice were treated with vehicle or cisplatin (20 mg/kg dose IP). After 72 h, kidney cortex homogenates were utilized for metabolic enzyme activities. Non-treated kidneys were used to isolate mitochondria for mitochondrial respiration via the Seahorse XF24 analyzer. Non-treated kidneys were also used for LC-MS analysis to evaluate kidney ATP abundance, and electron microscopy (EM) was utilized to evaluate mitochondrial morphology and number. (3) Results: KO mouse kidneys exhibit significant increases in malic enzyme and glucose-6 phosphate dehydrogenase activity under baseline conditions but in no other gluconeogenic or glycolytic enzymes. NHERF1 loss does not decrease kidney ATP content. Mitochondrial morphology, number, and area appeared normal. Isolated mitochondria function was similar between WT and KO. Conclusions: KO kidneys experience a shift in metabolism to the pentose phosphate pathway, which may sensitize them to the oxidative stress imposed by cisplatin.

**Keywords:** cellular redox state; oxidative stress; mitochondrial function; cisplatin nephrotoxicity

#### **1. Introduction**

Although cisplatin is a widely used chemotherapeutic that treats a variety of solid malignant tumors (e.g., ovarian testicular, head and neck, and lung cancer), its nephrotoxicity limits its use [1]. Twenty percent to 30% of patients on cisplatin will develop cisplatin-induced acute kidney injury (AKI) with a single dose [1]. The mechanisms underlying cisplatin's nephrotoxicity remain perplexing. Furthermore, with no methods for the prevention or treatment of cisplatin nephrotoxicity, the identification of host factors (biomarkers) that confer susceptibility to cisplatin and new therapeutic targets for the prevention and/or treatment of cisplatin nephrotoxicity are promising areas of research.

The accumulation and bioactivation of cisplatin to a more nephrotoxic metabolite underlies the kidney's susceptibility to cisplatin-induced AKI. Cisplatin nephrotoxicity has been found to alter renal cell metabolism and induce cell death via both apoptosis and necrosis. However, the mechanism by which cisplatin induces these pathways remains unclear. Several studies have found that cisplatin treatment leads to renal tubular cell depletion of amino acids [2–5], influences lipid metabolism through the reduction of fatty acid oxidation resulting in the accumulation of fatty acids in kidney tissue [2,5,6], and decreases glycolytic enzymes and intermediates of the pentose phosphate pathway and the citric acid cycle [2,7]. Reactive oxygen species (ROS) and mitochondrial function may also contribute to cisplatin's mechanism of injury [8]. Mitochondria continuously produce ROS, such as superoxide, and scavenge these ROS using antioxidant enzymes (superoxide dismutase, glutathione peroxidase, catalase, and glutathione S-transferase) [9]. Cisplatin has been found to also accumulate in the mitochondria of renal epithelial cells [10,11], resulting in mitochondrial damage including decreased mitochondrial mass, disruption of cristae, and even mitochondrial swelling [12–14]. These morphological changes are associated with a significant reduction in mitochondrial activity and ATP production [12,13]. This structural damage to mitochondria leads to electron leakage and produces massive amounts of free radicals in the form of ROS, leading to cell injury and death.

We have recently published the novel finding that kidneys lacking the scaffolding protein Na/H exchange regulatory factor 1 (NHERF1) show increased susceptibility to cisplatin-induced AKI [15]. NHERF1 is a known monomeric membrane-associated protein that belongs to the NHERF family of post-synaptic density protein 9595/Drosophila Discs Large/ZonulaOcculens -1(PSD-95/DIg/ZO-1)homology (PDZ)-scaffold proteins [16]. NHERF1 is found in all epithelial cells and acts as a scaffold for multi-protein signaling complexes. In proximal tubule cells of the kidney, NHERF1 is anchored to the cytoskeleton in the subapical plasma membrane, where it acts as a key scaffolding protein of transport proteins and has critical roles in defining the renal proximal tubule brush border membrane (BBM) composition and in regulating ion transport [16]. Recent studies have uncovered increasing evidence that NHERF1 has a much broader role than as a scaffolding protein. In fact, alterations in NHERF1 expression, phosphorylation status, and/or localization have been associated with tumorigenesis [17–19], changes in cell structure and trafficking [19–21], inflammatory responses [19,22], and tissue injury [15,19,23,24]. These studies highlight the fact that changes in NHERF1 expression can affect cell proliferation through alterations in the WNT/beta-catenin pathway and cyclin D kinase pathways and that the absence of NHERF1 diminishes the inflammatory response by impairing the assembly of specific components of the response such as the intracellular adhesion molecule 1 (ICAM1) [23]. Previous studies from our laboratory comparing wild-type (WT) and NHERF1-deficient opossum kidney (OK) cells, a model of mammalian renal proximal tubule, demonstrated that NHERF1-deficient cells grew more slowly and were more likely to die in culture; however, they exhibited no overt morphological differences. The NHERF1-deficient cells did show decreased BBM expression of the type IIa sodium phosphate cotransporter (Npt2a), the sodium-dependent glucose cotransporter (SGLT1), and the enzyme γ-glutamyl transferase (GGTase)

(unpublished data), which are defects that reversed when NHERF1 was re-introduced into the cells [25]. Likewise, NHERF1-deficient mice show normal kidney morphology but a marked decrease in the BBM expression of Npt2a, resulting in significant phosphate wasting, hypophosphatemia, hypercalciuria, and stone formation [26]. Formal studies to address metabolic changes in NHERF1-deficient kidneys have not been performed, but proteomic analysis from our laboratory comparing BBM from WT and NHERF1 KO kidneys revealed significant differences in the expression of mitochondrial proteins and enzymes from a variety of metabolic pathways. Notably, similar differences in BBM protein expression from the intestine of NHERF1 KO mice have been reported, but again, no specific studies to determine whether these changes in protein expression translate into changes in metabolism have been performed [27].

These observations suggest the hypothesis that alterations in renal cell metabolic pathways and/or mitochondrial dysfunction resulting from the loss of NHERF1 could prime these kidneys for a 'second hit' with cisplatin, therefore sensitizing these cells to cisplatin nephrotoxicity. The goals of this manuscript were twofold: (1) to determine if NHERF1 KO mice demonstrate changes in renal metabolic pathways, and (2) to determine if NHERF1 KO mice have altered mitochondrial function and/or structure that predisposes these animals to cisplatin nephrotoxicity.

#### **2. Materials and Methods**

#### *2.1. Animals and Treatments*

Two to 4-month-old male NHERF1(−/−) KO mice [28] and their WT littermates on C57BL/6J background were maintained on a 12:12 h light–dark cycle and were provided water and food ad libitum. At the time of sacrifice, animals were anesthetized with ketamine/xylazine (100/15 mg/kg, intraperitoneally (IP)). For enzyme kinetic assays, mice were given a single IP injection of 20 mg/kg cisplatin or vehicle (saline). Vehicle-treated and cisplatin-treated mice were euthanized after 72 h. All cisplatin studies were performed at the same time each day. Kidneys were removed and decapsulated; then, the cortex was separated from the medulla for homogenization. The kidney cortex was homogenized in 0.1 M Tris-HCl, pH 7.4 on ice. Then, tissue homogenates were sonicated on ice for 10 s/sample. Then, the samples were centrifuged for 15 min at 2500× *g* at 4 ◦C. For the mitochondrial studies, mice were not given a cisplatin or saline injection, and kidneys were removed and decapsulated for mitochondrial isolation or snap frozen in liquid nitrogen and stored at −80 ◦C for ATP analysis. All animal studies were approved by the Institutional Animal Care Use Committee (IACUC) at the University of Louisville and followed the guidelines of the American Veterinary Medical Association.

#### *2.2. Fructose-1,6-Bisphosphatase (FBPase) Activity Assay*

FBPase activity was measured as described previously [29]. Briefly, 20 µg protein (10 µL) was added to 170 µL assay buffer containing 0.1 M Tris-HCl, pH 8.6; 0.1 MgCl2; 0.1 M cysteine HCl in a 96-well plate and incubated at 37 ◦C for 5 min. Then, 20 µL of freshly prepared 50 mM fructose-1,6-bisphosphate (final concentration 5 mM) was added. Samples along with blanks (protein and substrate) and freshly prepared standards (0, 10, 25, 50, 100, 150, 200, 250 nM KH2PO4) were incubated at 37 ◦C for 60 min. The reaction was stopped by the addition of 50 µL 10% trichloroacetic acid (TCA), and the samples were centrifuged at 10,000× *g* for 5 min. Then, 200 µL of supernatant from samples, standards, and blanks was carefully pipetted into a fresh 96-well plate. Then, 50 µL freshly prepared 5% FeSO<sup>4</sup> (0.5 g of FeSO4 dissolved in 1 mL of 10% ammonium molybdate in 0.2 N H2SO<sup>4</sup> and the total volume was brought to 10 mL using distilled water) was added. The samples, standards, and blanks were read after 10 min at 820 nm wavelength. All samples, blanks, and standards were read in triplicate. The amount of inorganic phosphate released by the reaction of FBPase was determined by subtracting the substrate blank values from the sample values using the standards curve. All samples, blanks, and standards were read in triplicate. The average activity from the triplicate was considered

as *n* = 1 and shown as activity per mg protein (WT vehicle *n* = 3), (NHERF1 KO vehicle *n* = 4), (WT cisplatin *n* = 3), (NHERF1 KO cisplatin *n* = 5).

#### *2.3. Glucose-6-Phosphatase (G6Pase) Activity Assay*

G6Pase activity was measured as described [29]. Briefly, 20 µg protein (10 µL) was added to 170 µL assay buffer containing 0.1 M Tris-HCl, pH 7.4, and 0.1 M MgCl<sup>2</sup> in a 96-well plate and incubated for 5 min at 37 ◦C. Then, 20 µL freshly prepared substrate (50 mM glucose-6-phosphate, final concentration 5 mM) was added to each well. Samples along with blanks (protein and substrate) and freshly prepared standards (0, 10, 25, 50, 100, 150, 200, 250 nM KH2PO4) were incubated at 37 ◦C for 60 min. The reaction was stopped by the addition of 50 µL 10% TCA, and the samples were centrifuged at 10,000× *g* for 5 min. Then, 200 µL of supernatant from samples, standards, and blanks was carefully pipetted into a fresh 96-well plate. Then, 50 µL freshly prepared 5% FeSO<sup>4</sup> (0.5 g of FeSO<sup>4</sup> dissolved in 1 mL of 10% ammonium molybdate in 0.2 N H2SO<sup>4</sup> and the total volume was brought to 10 mL using distilled water) was added. The samples, standards, and blanks were read after 10 min at 820 nm wavelength. All samples, blanks, and standards were read in triplicate. The amount of inorganic phosphate released by the reaction of G6Pase was determined by subtracting the substrate blank values from the samples values using the standard curve. The average activity from triplicates was considered as *n* = 1 and shown as activity per mg protein (WT vehicle *n* = 3), (NHERF1 KO vehicle *n* = 4), (WT cisplatin *n* = 3), (NHERF1 KO cisplatin *n* = 5).

#### *2.4. Lactate Dehydrogenase (LDH) Activity Assay*

LDH activity was measured as described [29]. Briefly, samples (20 µg protein, 10 µL) were added to 170 µL assay buffer containing 0.1 M Tris-HCl, pH 7.4; 0.1 M MgCl<sup>2</sup> and 0.5 mM sodium pyruvate (substrate) and incubated at 37 ◦C for 5 min. Reaction was started by adding 20 µL 5 mM nicotinamide adenine dinucleotide and hydrogen (NADH) freshly prepared in 1 mM sodium bicarbonate, pH 9.0 solution (final concentration 0.5 mM). Samples and blanks (solution, protein, and NADH) were immediately read at 340 nm for 10 min. The optical density values were recorded every minute. The optical density measured before the addition of NADH was considered as time zero. Absorbance in the linear range was used for calculating activity. The activity was determined as a decrease in absorbance due to the oxidation of NADH per mg protein. The values of NADH blank were subtracted from samples to calculate the final activity. All samples were read in triplicate and the average value was considered *n* = 1 (WT vehicle *n* = 3), (NHERF1 KO vehicle *n* = 4), (WT cisplatin *n* = 3), (NHERF1 KO cisplatin *n* = 5).

#### *2.5. Malate Dehydrogenase (MDH) Activity Assay*

MDH activity is determined as described [29]. Briefly, a 10 µL sample (20 µg protein was added to assay buffer containing 0.1 M Tris-HCl, pH 7.4; 0.5 mM oxaloacetate; 0.1 M MgCl2. Reaction was started by adding 20 µL 5 mM NADH freshly prepared in 1 mM sodium bicarbonate, pH 9.0 solution (final concentration 0.5 mM). Samples and blanks (solution, protein, and NADH) were immediately read at 340 nm for 10 min. The optical density values were recorded every minute. The optical density measured before the addition of NADH was considered as time zero. Absorbance in the linear range was used for calculating activity. The activity was determined as a decrease in absorbance due to the oxidation of NADH per mg of protein. The values of the NADH blank were subtracted from samples to calculate the final activity. All samples were read in triplicate, and the average value was considered *n* = 1 (WT vehicle *n* = 3), (NHERF1 KO vehicle *n* = 4), (WT cisplatin *n* = 3), (NHERF1 KO cisplatin *n* = 5).

#### *2.6. Malic Enzyme (ME) Activity Assay*

ME activity was measured as described [29]. Briefly, a 10 µL sample (20 µg protein) was added to 170 µL assay buffer containing 0.1 M Tris-HCl, pH 7.4; 0.5 mM L-malic acid; 0.1 M MgCl2. Reaction was started by the addition of 20 µL freshly prepared 2 mg/mL of NADP<sup>+</sup> in 100 mM imidazole. Samples and blanks (solution, sample, and NADP<sup>+</sup> blank) were immediately read at 340 nm for 10 min. The optical density values were recorded every minute. The optical density measured before the addition of NADP<sup>+</sup> was considered as time zero. Absorbance in the linear range was used for calculating activity. The activity was determined as an increase in absorbance due to the reduction of NADP<sup>+</sup> to NADPH per mg protein. The values of NADP<sup>+</sup> blank were subtracted from samples to calculate the final activity. All samples were read in triplicate, and the average value was considered *n* = 1 (WT vehicle *n* = 3), (NHERF1 KO vehicle *n* = 4), (WT cisplatin *n* = 3), (NHERF1 KO cisplatin *n* = 5).

#### *2.7. Glucose-6-Phosphate Dehydrogenase (G6PD) Activity Assay*

G6PD activity was measured as described [29]. Briefly, a 10 µL sample (20 µg protein) was added to 170 µL assay buffer containing 0.1 M Tris-HCl, pH 7.4; 0.5 mM glucose-6-phosphate; 0.1 M MgCl2. Reaction was started by the addition of 20 µL freshly prepared 2 mg/mL of NADP<sup>+</sup> in 100 mM imidazole. Samples and blanks (solution, sample, and NADP<sup>+</sup> blank) were immediately read at 340 nm for 10 min. The optical density values due to the conversion of NADP<sup>+</sup> to NADPH were recorded every minute. The optical density measured before the addition of NADP<sup>+</sup> was considered as time zero. Absorbance in the linear range was used for calculating activity. The activity was determined as increase in absorbance due to the reduction of NADP<sup>+</sup> to NADPH per mg protein. The values of NADP<sup>+</sup> blank were subtracted from samples to calculate the final activity. All samples were read in triplicate, and the average was considered *n* = 1 (WT vehicle *n* = 3), (NHERF1 KO vehicle *n* = 4), (WT cisplatin *n* = 3), (NHERF1 KO cisplatin *n* = 5).

#### *2.8. Liquid Chromatography-Mass Spectrometry of Kidney Cortex for ATP Quantification*

First, 10 mg of kidney cortex was snap-frozen and stored at −80 ◦C until the LC-MS procedure. The tissue was maintained in liquid nitrogen during the preparation for LC-MS. Tubes and beads were pre-cooled in liquid nitrogen and 2.5% kidney cortex homogenate was made using an extraction solution (70% acetonitrile (ACN) and 30% H2O). Tissue was homogenized using a bead beater at 5 m/s for 15 s. Then, the homogenized tissue was transferred to another tube for centrifugation at 16,000× *g* for 10 min at 4 ◦C. The supernatant was saved for subsequent LC-MS analysis. To quantitate the ATP present in mouse kidney cortex tissue, a standard curve was prepared using 1 mM ATP diluted serially with the extraction buffer containing 20 µM <sup>13</sup>C10ATP. Each concentration point was diluted 50× with 60% ACN and 40% 15 mM ammonium acetate. Then, 5 µL was injected onto a SeQuant ZIC-cHILIC 100 × 2.1 mm metal-free HPLC column. Separation was performed by pumping 60% ACN and 40% 15 mM ammonium acetate through the column with a flow rate of 0.5 mL/min with no change in composition using a Waters Acquity UPLC. ATP was eluted from the column and flowed into a Waters Quattro Premier XE mass spectrometer and subsequently quantitated using optimized MRMs. A calibration curve was constructed using ATP concentration on the x-axis and the response of ATP over <sup>13</sup>C10ATP on the y-axis. Mouse kidney extracts were also diluted with 50x 60% ACN and 40% 15 mM ammonium acetate, 5 µL was injected, and the concentration in kidney tissue was interpolated using the calibration curve previously made for (WT *n* = 5) and (NHERF1 KO *n* = 5).

#### *2.9. Perfusion Fixation of Total Kidney In Situ for Electron Microscopy*

The abdominal cavity of the mouse was opened along the *linea alba* and intestines were moved aside to expose kidneys. A suture was loosely put around the right kidney for removal without glutaraldehyde exposure. Then, the right kidney was used for kidney cortex homogenates. Next, the chest cavity was opened to expose the heart for perfusion through the left ventricle, and the vena cava was cut, while the right kidney was tied off and removed. The left mouse kidney was perfused with 3% glutaraldehyde solution at a rate of 6 mL/min for approximately 1–3 min. Perfusion was stopped once kidneys had a change in color and consistency. The kidney was removed and placed in a petri dish of

3% glutaraldehyde where three to four 0.2 cm × 0.4 cm slices were cut and then stored in 3–4 mL of 3% glutaraldehyde at 4 ◦C. At this point, the tissue was sent to Norton Children's Hospital Pathology Department for a blind EM analysis of kidney tubule mitochondria. Images were taken by a renal pathologist at the Pathology Department. Sections with the highest concentration of mitochondria were randomly picked in a 4x field to be used to calculate mitochondria number and mitochondria area by Image J (WT *n* = 6) and (NHERF1 KO *n* = 5).

#### *2.10. Seahorse XF24 Mitochondrial Respiration Analysis*

Mice for this study had food taken away 6 h before sacrifice. Mitochondrial oxidative capacity was measured in isolated kidney mitochondria using a Seahorse Bioscience XF24 extracellular flux analyzer (Billerica, MA, USA). For measurements in isolated mitochondria, tissue from the kidney cortex of both kidneys (approximately 50 mg) was isolated and homogenized in 1 mL of isolation buffer (220 mM mannitol, 70 mM sucrose, 5 mM 3-(N-morpholino) propanesulfonic acid (MOPS), 1 mM ethylene glycol tetraacetic acid (EGTA), 0.3% fatty acid-free bovine serum albumin (BSA), pH 7.2). The homogenate was centrifuged at 500× *g* for five minutes at 4 ◦C. The supernatant containing mitochondria was centrifuged at 10,000× *g* for five minutes. Following two wash centrifugation steps in BSA-free isolation buffer, the mitochondria were suspended in respiration buffer (120 mM KCl, 25 mM sucrose, 10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 1 mM MgCl2, 5 mM KH2PO4, pH to 7.2). Protein in the mitochondrial suspension was estimated by the bicinchoninic acid method (Sigma) using BSA as a standard. Then, 10 µg of mitochondrial protein was sedimented in XF culture plates. Succinate (10 mM), rotenone (1 µM), and ADP (1 mM) were injected to assess state 3 respiratory activity (phosphorylating respiration). The oxygen consumption rate (OCR) of mitochondria after exposure to oligomycin (1 µg/mL) was used to estimate state 4 activity (non-phosphorylating respiration). Finally, Antimycin A (20 µM) was utilized to stop all respiration. Data are expressed as pmol O2/min/µg protein (WT *n* = 6, NHERF1 KO *n* = 6).

#### *2.11. Brush Border Membrane Isolation*

Kidney cortex BBM from WT and NHERF1 KO mice were prepared at 4 ◦C using the MgCl<sup>2</sup> precipitation method as previously described [30]. Kidney cortex slices were minced and then briefly homogenized in 50 mM mannitol and 5 mM Tris-HEPES buffer, pH 7.0 (20 mL/g), in a glass teflon homogenizer with four complete strokes. Then, a polyron homogenizer was used to provide high-speed [20,500 revolutions/min (rpm)] homogenization for three strokes of 15 s each with an interval of 15 s between each stroke. MgCl<sup>2</sup> was added to the homogenate to a final concentration of 10 mM and slowly stirred for 20 min. The homogenate was spun at 2000× *g* in a Sorvall high-speed centrifuge using an SS-34 rotor. The supernatant was centrifuged at 35,000× *g* for 30 min using the SS-34 rotor. Then, the pellet was resuspended in 300 mM mannitol and 5 mM Tris-HEPES, pH 7.4, with four passes by a loose-fitting Dounce homogenizer (Wheaton, IL) and centrifuged at 35,000× *g* for 20 min in 15 mL Corex tubes using the SS-34 rotor. The outer final pellet was resuspended in a small volume of buffered 300 mM mannitol.

#### *2.12. Label-Free Quantitative Liquid Chromatography-Mass Spectrometry*

Protein samples were analyzed by 2D-LC-MS/MS using either a linear trap quadrupole (LTQ) ion trap mass spectrometer (Thermo Fisher Scientific, Waltham, MA, USA) and/or using a Thermo Scientific LTQ Orbitrap Elite hybrid FTMS system enabled with electron transfer dissociation (ETD) fragmentation. Routinely one-dimensional (1D) reversed phase (RP) or two-dimensional (2D) strong cation exchange (SCX)-(RP) HPLC experiments were conducted off line for the LC-MALDI approach or online for the LC-LTQ-XL-Orbitrap-ESI-MS/MS approach and protein assignments were made as previously described [31–34].

The acquired mass spectrometry data were searched against an appropriate REFSEQ protein database using the SEQUEST (version 27 revision 11) algorithm and Matrix Science Mascot v.1.4 using Proteome Discoverer v1.3 to prepare and process RAW files as well as aggregate results. Separately for comparative LCMS analyses, the MS/MS database analysis was performed with SequestSorcerer (Sage-N Research, San Jose, CA) and high-probability peptide and protein identifications were assigned from the SEQUEST results using the Protein and PeptideProphet (tools.proteomecenter.org/software.php) and SageN Sorcerer statistical platforms. Scaffold 3 proteomic analysis software (ProteomeSoftware, Inc, Portland, OR) was used for quantitative comparison using a label-free spectral counting method. The qualitative comparison of protein expression patterns was performed by Ingenuity Pathways Analysis software (http://ingenuity.com).

### *2.13. Interpretation of Protein Assignment Results*

Lists of identified proteins and expression patterns were submitted for pathways analysis using one of several publically available (David, http://david.abcc.ncifcrf.gov/; GO, http://www.geneontology.org/; KEGG Pathway database, http://genome.jp/kegg/pathway.html), GenMAPP (www.genmapp.org/) or commercially licensed tools (Ingenuity Pathways Analysis software, (Redwood City, California, USA) http://ingenuity.com).
