Abstract

   Aging represents one of the most complicated and highly integrated
   somatic processes. Healthy aging is suggested to rely upon the coherent
   regulation of hormonal and neuronal communication between the central
   nervous system and peripheral tissues. The hypothalamus is one of the
   main structures in the body responsible for sustaining an efficient
   interaction between energy balance and neurological activity and
   therefore likely coordinates multiple systems in the aging process. We
   previously identified, in hypothalamic and peripheral tissues, the G
   protein-coupled receptor kinase interacting protein 2 (GIT2) as a
   stress response and aging regulator. As metabolic status profoundly
   affects aging trajectories, we investigated the role of GIT2 in
   regulating metabolic activity. We found that genomic deletion of GIT2
   alters hypothalamic transcriptomic signatures related to diabetes and
   metabolic pathways. Deletion of GIT2 reduced whole animal respiratory
   exchange ratios away from those related to primary glucose usage for
   energy homeostasis. GIT2 knockout (GIT2KO) mice demonstrated lower
   insulin secretion levels, disruption of pancreatic islet beta cell
   mass, elevated plasma glucose, and insulin resistance.
   High-dimensionality transcriptomic signatures from islets isolated from
   GIT2KO mice indicated a disruption of beta cell development.
   Additionally, GIT2 expression was prematurely elevated in pancreatic
   and hypothalamic tissues from diabetic-state mice (db/db), compared to
   age-matched wild type (WT) controls, further supporting the role of
   GIT2 in metabolic regulation and aging. We also found that the physical
   interaction of pancreatic GIT2 with the insulin receptor and insulin
   receptor substrate 2 was diminished in db/db mice compared to WT mice.
   Therefore, GIT2 appears to exert a multidimensional “keystone” role in
   regulating the aging process by coordinating somatic responses to
   energy deficits.

   Keywords: GIT2, metabolism, keystone, aging, dysglycemia, trajectory

Introduction

   The natural aging process is associated with an accumulation of
   molecular perturbations affecting almost all cells, tissues and organs
   of the body. These alterations have long been known to affect multiple
   processes related to cell survival, genomic instability, altered gene
   expression patterns, aberrant cellular replication, oxidative damage by
   reactive oxygen species (ROS), and fluctuations in protein expression
   and coherent protein post-translational modification ([49]1). Life
   expectancy in most westernized and developing countries is
   demonstrating an inexorable increase ([50]2, [51]3), therefore either
   natural or pathological aging could be considered the primary threat to
   global health in the future. With increasing age, there is a
   system-wide reduction in the ability of the body to cope with stress
   in-part due to a degradation of the efficiency of energy-generating
   [e.g., ATP (adenosine triphosphate)] metabolic systems ([52]4–[53]6).
   Disruption of the primary energy-synthesizing system, i.e., glucose
   catabolism by oxidative phosphorylation, leads to both ATP depletion
   (thus affecting electrical cellular excitability, proteolytic
   activities, transmembrane transport processes, and kinase activity) as
   well as an increase in the deleterious effects of unregulated
   hyperglycemia, e.g., systemic inflammation, arterial stenosis, impaired
   tissue healing, neuronal damage, and renal failure. Hence, one of the
   most characteristic factors in the aging process are the generation of
   insulin resistance, disruptions to oxidative phosphorylation of
   glucose, and changes in body fat composition ([54]6). Therapeutic
   interventions, such as caloric restriction or visceral fat depletion,
   designed to address features of metabolic/glycemic aging, have
   succeeded in improving insulinotropic activity and life span in rodents
   ([55]7, [56]8). The importance of the efficiency in glucose metabolism
   in the aging process is underscored by the demonstration that multiple
   glucose-regulatory factors in nematode worms were also major
   age-regulating genes, e.g., daf-2 [insulin receptor analog ([57]9)] and
   age-1 [phosphoinositide 3-kinase analog – a primary downstream factor
   of insulin receptor functions ([58]10)]. Translation of these
   fundamental insulin/glucose-mediated life span alterations from
   nematodes to higher organisms has proven problematical: a situation
   likely due to the increased functional diversity of
   insulin/insulin-related ligands in more complex, higher organisms.
   Despite this however, adipose tissue-specific insulin receptor deletion
   is associated with life span alterations in mice ([59]11). Linked to
   this observation, experimental paradigms in rhesus macaques that
   attenuate age-related glycemic disruption and increase insulin
   sensitivity (i.e., caloric restriction) can significantly reduce the
   incidence of age-dependent pathologies ([60]12).

   Due to the gradual loss of metabolic integrity with increasing age, the
   body becomes more prone to a variety of pathophysiologies linked to
   energy insufficiency such as neurodegeneration, metabolic syndrome
   (MetS), and chronic inflammation. These accumulated and progressive
   changes in complex physiological systems such as the endocrine or
   central nervous system (CNS) are highly likely to be mediated by gene
   and protein groups that are trophically controlled by “keystone” or
   “hub” factors that orchestrate the communication between
   lower-complexity functional signaling networks ([61]13). Considerable
   research demonstrates that both neurodegenerative diseases and
   pathophysiological aging involve a functional interplay between a
   series of diverse biological systems including neurological,
   endocrinological, sensory, and metabolic activities ([62]13–[63]23).
   Many of these systems are functionally integrated together in one
   crucial CNS organ, i.e., the hypothalamus ([64]24). The hypothalamus is
   responsible for the regulation of many metabolic pathways by
   synthesizing and secreting numerous neuronal hormones that stimulate or
   inhibit the secretion of trophic hormones from the anterior pituitary.
   The hypothalamus therefore can control global metabolism, body
   temperature, thirst, hunger, fatigue, and circadian rhythms ([65]25).
   Not only does the hypothalamus act as a trophic master-controller of
   the endocrine system but it also possesses neuronal projections to many
   autonomous and higher centers of the brain ([66]26), thereby providing
   a crucial link between aging and age-related disorders such as dementia
   ([67]27). We have previously shown, through the development of in
   cellula aging models ([68]19), that regulation of the G protein-coupled
   receptor kinase interacting protein 2 (GIT2) is highly sensitive to the
   chronological aging process and also to the presence of age-linked
   stressors such as ROS. Using a combinatorial informatics and proteomics
   approach, we demonstrated that GIT2 acts as a hypothalamic aging
   “keystone” factor ([69]13). Subsequently, GIT2 has also been implicated
   in regulating DNA integrity via its interaction with ataxia
   telangiectasia-mutated (ATM), a DNA damage repair (DDR) kinase
   ([70]18). In response to multiple forms of DNA damaging insults, GIT2
   is rapidly phosphorylated by ATM and then forms a complex with multiple
   DDR proteins including MDC1, MRE11, and phosphorylated H2AX ([71]18).
   Genomic deletion of GIT2 in mice (GIT2KO) resulted in a significant
   increase in the rate of DNA damage accumulation compared to wild type
   (WT) age-matched control mice ([72]18). DNA instability has been
   associated with an advanced rate of aging for several decades
   ([73]28–[74]30) and this has been reinforced with the appreciation that
   accelerated molecular aging occurs in patients possessing mutations in
   multiple DDR proteins including ATM ([75]31), WRN [Werner syndrome,
   RecQ helicase-like ([76]32, [77]33)], and BLM [Bloom syndrome, RecQ
   helicase-like ([78]34)]. Glucose-related diabetic pathologies often
   result in elevated cellular levels of ROS due to an attenuated glucose
   uptake capacity and support of oxidative phosphorylation causing cells
   to generate ATP via other less-efficient synthetic mechanisms.
   Prematurely elevated ROS levels, as with mutations in DDR proteins, are
   considered to be one of the major causes of advanced aging
   ([79]35–[80]37). Reinforcing the link between dysfunctional aging,
   diabetic pathologies, and DNA instability, DDR proteins such as ATM are
   repeatedly linked to impaired insulin secretion capacity and diabetic
   pathophysiologies ([81]38–[82]41). Therefore, given the demonstration
   of a crucial hypothalamic function of GIT2 both in the aging process
   and DNA integrity management, we investigated the potential role of
   this G protein-coupled receptor (GPCR)-interacting protein in
   controlling/coordinating glycemic or metabolic functions in the context
   of aging. We have analyzed the molecular metabolic signatures of
   pathological aging in mice at a relatively early phase of life, i.e.,
   2–8 months of life (approximately 10–30% of life span), in a similar
   vein to the human aging trajectory study of Belsky et al. ([83]42). In
   the study of Belsky et al., it was suggested that research into the
   molecular mechanisms of aging are significantly confounded, both with
   animal models and human patients, by the accumulation of extraneous
   co-pathologies during life span ([84]42). However, it is also likely
   that dysfunctional molecular aging paradigms are set in place during
   early life that effectively define either a pathological or
   “less-pathological (healthy)” aging trajectory ([85]43–[86]45). Here,
   we have found that high-dimensionality molecular transcriptomic
   signatures of pathological aging and pro-diabetic pathologies occur
   early in life and that GIT2 controls global somatic metabolism,
   circulating levels of multiple plasma energy-regulating hormones and
   pancreatic beta cell functionality.

Materials and Methods

Cell Culture

   Murine Beta-TC-6 cells (ATCC[87]^1) were maintained in a humidified 5%
   CO[2] environment at 37°C in Dulbecco’s modified Eagle’s medium (DMEM)
   media (Life Technologies, Carlsbad, CA, USA) supplemented with 15%
   heat-inactivated fetal bovine serum and 1% penicillin–streptomycin
   (Life Technologies, Carslbad, CA, USA). Murine GIT2 siRNA (Santa Cruz)
   was generated as a pool of three target-specific 19–25 nt siRNAs.
   Control siRNA-A consists of a scrambled sequence that does not induce
   specific degradation of any cellular mRNA. Cells were transfected with
   siRNA oligonucleotides using Lipofectamine RNAi MAX (Life Technologies,
   Carlsbad, CA, USA) according to the manufacturer’s protocol. To assess
   cell viability after any specific long-term treatment, cells were first
   washed in phosphate buffered saline (PBS: Sigma-Aldrich, St. Louis, MO,
   USA) and then collected from the growth plates using calcium-free,
   non-enzymatic cell dissociation solution (Sigma-Aldrich, St. Louis, MO,
   USA). Cells were gently pelleted by centrifugation at 1000 × g at room
   temperature before being resuspended in a Hanks’ Balanced Salt solution
   (cat.#H9269: Sigma Aldrich) and mixed with a 0.4% Trypan Blue solution
   and allowed to incubate for 15 min at room temperature. Cells were then
   transferred to a Bright-Line™ hemacytometer (Sigma-Aldrich) and a
   percentage of positive blue cells (representing cells possessing a
   severely compromised plasma membrane and thus at high risk of imminent
   death) were counted by an experienced researcher. As a positive control
   for loss of cell viability, cell cultures were pre-incubated with a
   2 μg/ml concentration of actinomycin D (Sigma-Aldrich) for 2 h before
   collection and Trypan Blue processing as described here previously.

Experimental Animal Models

   G protein-coupled receptor kinase interacting protein 2 gene-trap
   knockout (KO) mice (GIT2KO), were originally obtained from Dr. Richard
   T. Premont (Duke University Medical Center, Durham, NC, USA) and
   genotyped using genomic PCRs as previously described ([88]46). PCR
   amplification was performed by standard protocols. Primers used to
   screen the GIT2 knockout mice were as follows: forward primer
   5′-TCTCCTGGAACTCAGGGATT, reverse primers (WT) 5′-CATTTCAGAGTCTGCTGCCTTA
   and (KO) 5′-GGCTACCGGCTAAAACTTGA. Male WT (WT: C57BL/6) and GIT2KO mice
   were group-housed (genotypes were housed separately) in
   temperature-controlled (22°C) and humidity-controlled (45%) rooms with
   a 12 h light–dark cycle with food and water available ad libitum. Male
   leptin receptor-mutant (db/db) mice, bred on a C57BL/6 background, were
   purchased from The Jackson Laboratories.[89]^2 All experimental mouse
   models were fed using a standard control animal chow containing 19%
   protein, 64% carbohydrates, and 17% fat (diet #101845 from Dyets Inc.,
   Bethlehem, PA, USA). Experiments were conducted during the light phase
   of the light–dark cycle in accordance with NIH guidelines. Animal care
   and experimental procedures followed NIH guidelines and were approved
   by the National Institute on Aging Animal Care and Use Committee
   (433-LCI-2015).

Whole Animal Glycemic Function Testing

   Standard insulin tolerance tests (ITT), applied according to Jackson
   Laboratory handling protocols,[90]^3 involved blood glucose sampling
   from a distal tail snip after the specified time (0–90 min) of a
   subcutaneous Lantus^® (0.5 U/kg) bolus. Standard oral glucose tolerance
   tests (OGTT) were made using standardized protocols ([91]47–[92]49).
   Briefly, mice were fasted for 4 h, and blood was drawn from the tail
   and gathered using a heparinized micro-hematocrit capillary tube
   (Fisher Scientific, Pittsburgh, PA, USA), and clotisol (Agri-Med, St.
   Louis, MO, USA) was used to stop the bleeding. Mice were then
   administered with glucose (2 g/kg) via gavage and blood samples drawn
   at the specified time in minutes after glucose administration. For all
   time points, whole-blood glucose concentration was quantified at the
   time of blood draw using a Freestyle Glucose meter (Abbott
   Laboratories, Abbott Park, IL, USA). Blood samples were then
   centrifuged at 12,500 rpm for 12 min at 4°C, and plasma supernatant was
   removed and stored at −80°C. Insulin was quantified using a
   commercially available ELISA kit (Crystal Chem Inc., Downers Grove, IL,
   USA).

Whole-Body Murine Metabolic and Behavioral Analyses

   Murine behavior and metabolic parameters were measured, over a 48-h
   period, using a Columbus Instruments Comprehensive Lab Animal
   Monitoring System (CLAMS). The CLAMS system was used to evaluate food
   intake (gram), water intake (milliliter), X, Z total activity (counts),
   O[2] consumption (milliliter/kilogram/hour), CO[2] production
   (milliliter/kilogram/hour), and respiratory exchange ratio
   (RER = VCO[2]/VO[2]), as described previously with modifications
   ([93]50–[94]52). Eight mice per group were single housed for 24 h in
   the system. The mean weight of WT-control mice (27.25 ± 0.69 g) was
   slightly less but not significantly different from the GIT2KO
   counterparts (29.06 ± 1.17 g). Statistical analysis was performed using
   a two-tailed Student’s t-test, and p ≤ 0.05 was considered
   statistically significant.

Measurement of Glucose, Lipids, and Hormone Levels

   Blood was collected after the mice were euthanized with isoflurane.
   Eight to ten animals were included in each group. Glucose levels were
   measured using an EasyGluco blood glucose monitoring system (US
   Diagnostics, Inc.). For fasting glucose measurements, mice were
   food-deprived 12 h prior to blood collection. Plasma insulin, leptin,
   peptide YY (PYY), gastric inhibitory peptide (GIP), amylin, and
   pancreatic peptide (PP) levels were measured using Linco-Millipore
   multiplex kit ([95]53, [96]54). Briefly, plasma samples were incubated
   together with antibody conjugated beads for 16 h at 4°C. After three
   times washing, biotinylated detection antibody was added: these were
   further incubated for 1 h at room temperature, and then
   streptavidin–horseradish peroxidase was added. Fluorescent signals were
   detected by Bio-Plex 200 systems. Each sample was assayed in duplicate
   on a 96-well plate. Hormone levels were derived by interpolation from a
   reference curve generated in the same assay with reference standards of
   known concentrations of the detected hormones.

Pancreatic Tissue and Islet Extraction

   As described previously ([97]53), mice were euthanized with isoflurane
   followed by decapitation. Whole pancreata were removed from each
   animal. The pancreata were then processed differentially concerning
   their eventual usage. For immunohistochemical analyses, the pancreata
   were fixed in 4% formalin for 48 h and stored in PBS until processing.
   The pancreatic tissue was processed and embedded in paraffin wax.
   Pancreatic sections were cut at 5 μm thickness using a microtome, and
   the sections were adhered to poly-l-lysine-coated microscope slides
   (Fisher, Springfield, NJ, USA). For analysis of insulin- (beta cells)
   and glucagon- (alpha cells) secreting islet cells, pancreatic sections
   were immunostained according to a previously described protocol
   ([98]53). Briefly, the tissue was incubated with the primary insulin
   antibody (guinea pig anti-swine insulin; 1:300; DakoCytomation,
   Carpinteria, CA, USA) or glucagon antibody (1:500; guinea pig
   anti-glucagon; Millipore, Billerica, MA, USA) for 2 h at room
   temperature and then incubated with secondary antibody [Alexa Fluor 488
   (AF488) goat anti-–guinea pig, 1:200; or Alexa Fluor 568 (AF568) goat
   anti–guinea pig, 1:1,000; Life Technologies, Carlsbad, CA, USA] for 1 h
   at room temperature. Sections were imaged with an Olympus Fluoview IX70
   microscope (Olympus America, Center Valley, PA, USA). Quantification of
   immunohistochemistry images was performed in MATLAB (Mathworks) using
   novel software in conjunction with the image processing toolbox.
   Intensity readings of each image ranged from 0 to 256, with 256 being
   the greatest pixel density and hence the highest staining intensity.
   The region of interest (ROI) was drawn around each islet after
   background subtraction. The pixels within the bounds of the ROI and
   above the set threshold of eight were selected, from which the actual
   islet area was calculated. The normalized variance of the ROI was used
   to calculate an artificial ellipse from which the major and minor axes
   were determined. The major axis is the longest diameter that can be
   drawn in the ellipse, and the minor axis is the shortest diameter, both
   giving an accurate approximation for the range of the actual islet
   diameter. Islet morphometry and sizing analyses were performed in an
   unbiased, random fashion ([99]55). Specific murine islets were isolated
   using collagenase digestion, using the same protocol as described
   previously ([100]56). Murine islets suspended in DMEM-containing 5 mM
   glucose and 1% bovine serum albumin (BSA) were pelleted and stored
   before extraction of total RNA for microarray analysis.

Bioinformatic Analyses

   Microarray-derived gene transcript lists were initially separated into
   specific expression polarity regulation groups using VennPlex venn
   diagram analysis ([101]57) before pathway analysis using multiple forms
   [KEGG pathway using WebGestalt ([102]58), Ingenuity Pathway Analysis
   (IPA) Canonical Signaling, Textrous!-based natural language processing
   ([103]59)] of functional annotational clustering. For KEGG and IPA
   Canonical Pathway signaling analysis, we employed a cutoff of at least
   two transcripts (from the original filtered/analysis of variance
   datasets) needing to be present to minimally populate a particular
   canonical IPA or KEGG pathway with a probability (p) of enrichment
   value of ≤0.05. Where stated, a hybrid score process was employed to
   generate a single index for a specifically enriched KEGG pathway.
   Hybrid scores were generated by the multiplication of the negative
   log[10] of enrichment probability (p) with the enrichment factor (R).
   Latent Semantic Indexing (LSI)-based analysis ([104]60) was performed
   as described previously using Textrous! ([105]59). Textrous! as well as
   other LSI-based informatics platforms (e.g., GeneIndexer), correlate
   the strength of association, using LSI ([106]13, [107]19), between
   specific transcripts in a dataset with user-defined interrogation
   terms. Textrous! is able to investigate connections between input
   transcripts/proteins (using input Gene Symbols) and biomedical words
   and noun-word phrases using complete extracted data (in the form of
   curated gene-word documents) from PubMed Central,[108]^4 OMIM (Online
   Mendelian Inheritance in Man[109]^5), and the Mammalian Phenotypes
   Database at the Jackson Laboratories Mouse Genomic Informatics
   portal.[110]^6 For Textrous!-based natural language processing, the
   possible LSI correlation scores (termed Cosine Similarity Scores) for a
   gene/protein to be associated with an input interrogation term range
   from 0 to 1, with the stronger correlation scores approaching 1. A
   correlation score of ≥0.1 indicates at least an implicit correlation,
   between the specific gene and the user-defined input interrogation term
   ([111]61). To generate a holistic and more specific high-dimensionality
   data annotation for Textrous!-based text outputs, significant data word
   clouds were created from the Textrous!-based semantic noun and
   noun-phrase outputs using the web-based application Wordle[112]^7
   ([113]62). The text size within the word clouds is directly
   proportional to the input word frequency. Analyses of word-phrase
   frequencies from Textrous! noun-phrase outputs were made using
   WriteWords.[114]^8 The use of freeform quantitative wordclouds to
   convey complex non-canonical signaling-activity relationships is
   becoming more and more recognized as a novel technique to investigate
   high-dimensionality data ([115]63–[116]65).

Immunoprecipitation Assays

   Whole murine pancreata, WT or db/db, were extracted through
   micro-dissection. For protein immunoprecipitation, cytoplasmic cell
   lysates were prepared using the Qproteome^® cell compartment kit
   according to the manufacturer’s instructions (Qiagen, Valencia, CA,
   USA) and as described here. All protein extracts were quantified using
   BCA reagent (ThermoScientific, Waltham, MA, USA) before normalization
   to a standard 1 mg/ml protein concentration before immunoprecipitation.
   Normalized supernatant protein lysates were pre-cleared by a 1 h
   incubation, tumbling at 4°C, with 30 μl of protein A/G pre-conjugated
   agarose beads (EMD Chemicals, Gibbstown, NJ, USA). For
   immunoprecipitation, the pre-cleared cell lysates were then incubated
   overnight for 16 h at 4°C with tumbling with 10 μg of either pre-immune
   IgG sera (Sigma-Aldrich, St. Louis, MO, USA) or 10 μg of anti-GIT2
   antibody (Bethyl Laboratories, Montgomery, TX, USA). To collect the
   immunecomplexes, 30 μl of protein A/G pre-conjugated agarose beads (EMD
   Chemicals, Gibbstown, NJ, USA) were added to the lysates and then
   tumbled at 4°C for an additional hour. The agarose-associated
   immunecomplexes were then collected by centrifugation at 2000 × g for
   5 min. Immunoprecipitated protein complexes were removed from the
   agarose-associated antibodies by the addition of 50 μl of a
   room-temperature CHAPS-Urea buffer [4% CHAPS ((3-((3-cholamidopropyl)
   dimethylammonio)-1-propanesulfonate) (Sigma-Aldrich, St. Louis, MO,
   USA), 8M Urea)] through gentle agitation before being stored at −80°C
   for further mass spectrometry (MS)-based processing.

Mass Spectrometry

   Proteins were extracted from GIT2 immunoprecipitates using 4% CHAPS-8M
   Urea and were analyzed using reverse-phase nano LC-MS/MS on a
   ThermoFinnigan LXQ linear ion MS (ThermoFinnigan, West Palm Beach, FL,
   USA) as described previously ([117]66). Analytical samples were
   separated on an in-house fabricated 8-cm reverse-phase capillary
   emitter column or a c18 PicoFrit column (New Objectives, Boston, MA,
   USA), using 90 min gradients. The collision energy for the LXQ MS was
   set at 30%. Spectra were acquired in a data-dependent manner with the
   dynamic exclusion option enabled. The four most intense ions in each
   full MS scan were fragmented and analyzed. MS/MS spectral data were
   processed using MASCOT Daemon v. 2.2.2 (Matrix Science) for protein
   identification. For the MS/MS ion search using MASCOT, proteins with >1
   peptides and a score of each peptide higher than 45 with an unambiguous
   identification were considered for further analysis without manual
   spectra inspection. Proteins with one peptide and a score lower than 45
   were considered ambiguous and discarded from analysis. We used the
   pre-defined enzyme specificity in the Mascot Daemon v. 2.2.2. Search
   parameters included static mass modification to cysteine (iodoacetamide
   alkylation) and differential mass modification to methionine
   (oxidation). The decoy database option provided by MASCOT was selected
   to determine false discovery rates (FDR), which was kept below 1% by
   applying various significance thresholds to each search result.

Western Blotting

   Murine hypothalami or pancreata were extracted through
   micro-dissection, and cell lysates were also prepared from murine TC6
   cells. For protein extraction, differential cellular compartment
   (cytoplasm, plasma membrane, nuclear/large organelle, and cytoskeleton)
   lysates were prepared using the Qproteome^® cell compartment kit
   according to the manufacturer’s instructions (Qiagen, Valencia, CA,
   USA). All protein extracts were quantified using BCA reagent
   (ThermoScientific, Rockford, IL, USA) and then normalized for each
   specific experiment before resolution with SDS-PAGE and semi-dry
   electrotransfer (Bio-Rad, Hercules, CA, USA) to PVDF membranes
   (PerkinElmer Life Sciences; Waltham, MA, USA). Membranes were blocked
   using a 4% BSA for Western blot as described previously ([118]67), and
   primary antibody immune-reactive complexes were identified using
   alkaline phosphatase-conjugated secondary antisera (Sigma-Aldrich, St.
   Louis, MO, USA) with enzyme-linked chemifluorescence (GE Healthcare)
   and visualized with a Typhoon 9410 phosphorimager (GE Healthcare).
   Proteins were identified using primary antisera at 1:1,000 to 1:10,000
   dilutions, followed by species-specific alkaline phosphatase-conjugated
   secondary antibodies (Sigma-Aldrich, St. Louis, MO, USA) at 1:7,000
   dilution. The primary antibodies employed in this study are described
   in the next section.

Experimental Antibodies

   Primary antibodies specific for diablo homolog (Drosophila) (Diablo:
   ab8114), glyoxalase 1 (Glo1: 6F10: ab81461), sestrin 1 (Sesn1:
   ab103121), insulin receptor (Insr: ab5500), insulin receptor substrate
   2 (Irs2: ab134101), p53 and DNA damage regulated 1 (Pdrg1: ab74500),
   plasma membrane Ca^2+ ATPase (Pmca: ab2825), and tubulin alpha-1A
   (Tuba1a: ab76449) were obtained from Abcam (Cambridge, MA, USA).
   Primary antibodies specific for phosphatidylethanolamine binding
   protein 1 (Pebp1: clone EPR2875Y: C105420) and high mobility group
   nucleosomal binding domain 2 (Hmgn2: C151267) were obtained from LSBio
   (Seattle, WA, USA). Primary antibodies specific for receptor accessory
   protein 5 (Reep5: H-76), extracellular signal-regulated kinase 2 (Erk2:
   C-8), glucose transporter 2/solute carrier family 2 (facilitated
   glucose transporter), member 2 (Glut2/Slc2a2: H-67),
   glyceraldehyde-3-phosphate dehydrogenase (Gapdh: I-19), mitochondrial
   ribosomal protein L12 (Mrpl12: 397.1), F-box protein 24 (Fbxo24: D-14)
   microtubule-associated protein 1 A (Map1a: N-18), CASK-interacting
   protein 2 (Caskin2: P-13), and lamin A (Lmna: C-20) were obtained from
   Santa Cruz Biotechnology (Santa Cruz, CA, USA). Primary antibodies
   specific for doublecortin (Dcx: #4604), ataxia telangiectasia mutated
   (Atm: #2873), beta-PIX (Arhgef7: #4515), and p21-activated kinase 1
   (Pak1: #2602) were obtained from Cell Signaling Technology (Danvers,
   MA, USA). The primary antibody specific for regenerating islet-derived
   3 beta (Reg3b: Clone 518630) was obtained from R&D Systems
   (Minneapolis, MN, USA). Primary antibodies specific for NADH
   dehydrogenase (ubiquinone) 1 beta subcomplex and 10 (Ndufb10:
   15589-1-AP) and ribosomal protein L17 (Rpl17: 14121-1-AP) were obtained
   from ProteinTech (Chicago, IL, USA). The primary antibody specific for
   Finkel–Biskis–Reilly murine sarcoma virus (FBR-MuSV) ubiquitously
   expressed (fox derived) (Fau: NBP1-55090) was obtained from Novus
   Biologicals (Littleton, CO, USA). Primary antibodies specific for G
   protein-coupled receptor kinase-interactor 2 (Git2: A302-102A) and
   structural maintenance of chromosomes protein 5 (Smc5: A300-236A) were
   obtained from Bethyl Laboratories Inc. (Montgomery, TX, USA).

Differential Detergent Cellular Fractionation

   For cellular differential detergent fractionation, cell lysates were
   prepared using the Qproteome^® cell compartment extraction kit (cat#
   37502) according to the manufacturer’s instructions (Qiagen, Valencia,
   CA, USA) with minor modifications. In brief, cultured cells were first
   washed in ice-cold phosphate buffered saline (PBS: Sigma-Aldrich, St.
   Louis, MO, USA) and then collected from the plate using ice-cold
   calcium-free cell dissociation solution (Sigma-Aldrich, St. Louis, MO,
   USA). Intact cell pellets were created by centrifugation at 1000 × g at
   4°C. Cell pellets were then solubilized in the pre-cooled Qproteome
   proprietary cytoplasmic protein extraction buffer (CE1: 500 μl), which
   was supplemented with 1× of the kit Protease Inhibitor Solution. The
   lysates were then tumbled in the cytoplasmic CE1 buffer for 10 min at
   4°C. The CE1 lysate was then centrifuged at 1000 × g for 10 min at 4°C.
   The supernatant was then removed to a fresh Eppendorf tube to generate
   the cytoplasmic tissue extract. The extant cell pellet was then
   solubilized in 500 μl of ice-cold CE2 plasma membrane extraction buffer
   supplemented with 1× of the kit Protease Inhibitor Solution and then
   tumbled at 4°C for 30 min. The lysate was then clarified by
   centrifugation at 6000 × g for 10 min at 4°C. The supernatant,
   representing the plasma membrane enriched compartment was removed to a
   fresh Eppendorf tube, and the extant pellet was then incubated with
   7 μl of Qproteome-supplied Benzonase^® Nuclease and 13 μl of distilled
   water, mixed through gentle agitation and then incubated at standard
   room temperature (21–24°C) for 15 min. The pellet was then solubilized,
   with additional pipette trituration, in 250 μl of ice-cold Qproteome
   CE3 nuclear/large organelle compartment extraction buffer supplemented
   with 1× of the kit Protease Inhibitor Solution before being tumbled for
   10 min at 4°C. The lysate was next clarified by centrifugation at
   6800 × g for 10 min at 4°C. The supernatant, representing the
   nuclear/large organelle cellular fraction, was removed to a fresh
   Eppendorf tube. The extant cell pellet was then solubilized in 250 μl
   of room temperature Qproteome CE4 cytoskeletal fraction extraction
   buffer.

Glucose Uptake Assays

   For specific glucose uptake assessment, cultured cell monolayers were
   glucose- and serum-deprived in Krebs Ringer buffer (25 mM NaCl, 5 mM
   KCl, 1.25 mM NaH[2]PO[4], 2 mM CaCl[2], 1 mM MgCl[2], 25 mM NaHCO[3])
   at 37°C and 5% CO[2] for 40 min on the day of experimentation. To
   determine glucose uptake rates, the specified glucose bolus dose was
   added to the Krebs Ringer buffer, and 30 μL aliquots were taken at 0,
   1, 5, 10, 15 and 30 min. Glucose concentration from each aliquot was
   determined using a fluorescent glucose assay kit (# K606-100:
   BioVision, Milpitas, CA, USA) as per manufacturer’s instructions. Cells
   were lysed at the end of the experiment, and protein concentration was
   determined through a BCA assay (ThermoFisher Scientific, Waltham, MA,
   USA). Glucose uptake curves were then normalized for each experiment
   against the respective total protein concentration measured.

RNA Extraction and Oligonucleotide Microarray Hybridization

   RNA isolation from three animals in each experimental group was carried
   out using the Qiagen RNeasy Mini Kit (Qiagen, Inc., Valencia, CA, USA),
   as described previously ([119]68). RNA conversion to cDNA and
   subsequent hybridization with Sentrix MouseRef-8 Expression BeadChips
   (Illumina, San Diego, CA, USA) was performed as described previously
   ([120]19, [121]69). Microarray data were analyzed using DIANE
   6.0[122]^9, a spreadsheet-based microarray analysis program based on
   the SAS JMP7.0 system[123]^10. Raw microarray data were subjected to
   filtering and z-normalization and tested for significant changes as
   described previously ([124]19). Initial filtering identified genes with
   a z-ratio of ≥ ±1.50, with the z-ratio being derived from the
   difference between the averages of the observed gene z-scores divided
   by the SD of all of the differences for that particular comparison.
   Genes were then refined by calculating the FDR, which controls for the
   expected proportion of falsely rejected hypotheses, and including only
   those genes with FDR <0.05. These data were further analyzed using
   analysis of variance with significance set at p < 0.05. This allowed us
   to identify transcripts that differed in their intensity across all of
   the animal replicates and the various experimental conditions of the
   mice employed in this study.

Statistical Analyses

   In each histogram or figure, data represent the means ± SEM.
   Statistical analyses (Student’s t-test) were performed using GraphPad
   Prism (GraphPad Software, San Diego, CA, USA). p ≤ 0.05 was considered
   statistically significant. Significance is indicated in each figure as
   *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

Results

Loss of GIT2 Protein Affects Whole-Body Metabolic Activity Independent of
Physical Motor Activity

   As metabolic activity has been strongly linked with aging and multiple
   disease processes, we first investigated how genetic deletion of the
   aging-associated factor, GIT2, affected somatic metabolism and activity
   in post-pubertal 4-month-old (m.o.) male GIT2KO mice compared to age-
   and gender-matched control WT mice (Figure [125]1). Compared to WT
   mice, over the 48-h CLAMS-assessment period, we found that male GIT2KO
   mice demonstrated significantly lower levels of VO[2], VCO[2], and
   consequently RER (Respiratory Exchange Ratio) (Figures [126]1A–C). In
   addition, we also found that GIT2KO mice demonstrated a significantly
   lower thermal output temperature, compared to WT controls (Figure
   [127]1D). In contrast to the significant alterations in the
   aforementioned metabolic parameters, no significant differences in
   physical motor activity (in x total or z axis) across the test period
   between WT and GIT2KO mice (Figures [128]1E,F) were observed,
   indicating that the metabolic distinctions were not due to alterations
   in physical motor activity. In addition to the altered metabolic
   parameters present in the GIT2KO mice, a significant reduction in both
   food and water intake was observed compared to WT controls (Figures
   [129]1G,H). GIT2KO mice also demonstrated an increased propensity for
   longer sleep/inactivity epochs compared to WT mice (Figure [130]1I).

Figure 1.

   [131]Figure 1

   [132]Figure 1
   [133]Open in a new tab

   Whole-body metabolic analysis of WT and GIT2KO mice. Male WT or GIT2KO
   mice were housed for 48 h in a Columbus Instruments International
   Comprehensive Lab Animal Monitoring System (CLAMS). VO[2] (A), VCO[2]
   (B), RER (respiratory exchange ratio) (C), HEAT (ambient heat
   generation) measurements for WT and GIT2KO mice (D). Locomotor physical
   movement × total activity (E) and z total activity (F) measurements
   were made for WT and GIT2KO mice. Accumulated drinking (G) and
   accumulated food intake (H) and sleep epochs, total time sleeping, and
   % of time asleep (I) measurements were also recorded. For (A–H), WT
   datapoints are represented by black circles; GIT2KO datapoints are
   represented by open triangles. Gray areas in each panel (A–H) indicate
   the dark period of animal incubation. For (I), WT data is indicated by
   black bars and GIT2KO data by white bars. Statistical significance is
   indicated in each figure as *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

Transcriptomic Signatures of Metabolic Disruption in GIT2KO Hypothalami

   Our whole-body metabolic analyses of the GIT2KO mice demonstrated a
   strong alteration in metabolic activity. We have previously shown that
   aberrant hypothalamic signaling activity ([134]13) as well as
   transcriptomic profiles ([135]69) are strongly linked to both
   pathological aging as well as the presentation of neurodegenerative
   disease. We therefore assessed, using microarray, the effects of GIT2
   genomic deletion on hypothalamic transcriptomic profiles in male mice
   at 2, 4, and 8 months of age as these timepoints encompass our initial
   CLAMS 4-m.o. timepoint data. This timeframe was chosen to represent
   post-pubertal early life periods in which overt age-related damage may
   not be present but signatures of pro-aging behavior may be observed
   ([136]42). Gene transcripts differentially and significantly (p < 0.05)
   regulated in GIT2KO hypothalami compared to age-matched WT mice across
   the three time points were investigated using VennPlex ([137]57)
   (Figure [138]2A). Surprisingly, we found that transcriptomic profiles
   [Tables S1 in Supplementary Material (2 m.o.), S2 (4 m.o.) and S3
   (8 m.o.)] were strongly age-specific, with a considerable percentage of
   transcripts being uniquely regulated at a single time-point (2 m.o. –
   88.9% unique; 4 m.o. – 88.6% unique; 8 m.o. – 82.03% unique). Across
   the three experimental ages, we found 46 transcripts common to at least
   two timepoints (Figure [139]2B) – we randomly selected four of these
   transcripts (Pebp1, Reep5, Hmgn2, Diablo) to validate, using Western
   blotting (Figure [140]2C). Our Western blot analysis of hypothalamic
   tissues replicated our microarray-derived transcriptomic regulation
   patterns in the GIT2KO mice. The functional effects of the group of 46
   GIT2KO-modulatated transcripts identified across all timepoints were
   then investigated using unbiased Textrous!-mediated transcript-word
   association ([141]59, [142]62, [143]70). Using the collective
   processing module of Textrous! to generate a hierarchical word cloud, a
   strong focus upon neurotransmission and metabolism was evident (Figure
   S1 in Supplementary Material). With extraction of the primary word list
   associated with the 46 input transcripts (Table S4 in Supplementary
   Material), multiple words associated with glucose regulation
   (“insulin,” “proinsulin,” “euglycemic,” “glucose,” “intolerance,”
   “incretin,” “insulin-responsive,” “c-peptide”) were strongly evident. A
   typical healthy animal using glucose as its primary source of energy
   demonstrates an RER of approximately 1 (observed in the WT mice: Figure
   [144]1C), the GIT2KO mice however possessed an RER of approximately 0.8
   across the testing period (Figure [145]1C), which is suggestive of loss
   of glucose catabolism in favor of lipid metabolism. In line with this a
   strong representation of terms linked with lipid metabolism were also
   found in the Textrous! output form the 46 core GIT2KO transcripts
   (Figure S1 in Supplementary Material; Table S4 in Supplementary
   Material), e.g., “adipocytes,” “lipolysis,” “fatty,” and
   “triacylglycerol.”

Figure 2.

   [146]Figure 2
   [147]Open in a new tab

   Hypothalamic transcriptomic signatures in WT and GIT2KO mice across
   early life span timepoints. (A) Differential VennPlex analysis of
   hypothalamic transcriptomic data from 2-, 4-, and 8-month-old (m.o.)
   GIT2KO mice compared to age-matched WT controls. Contra-regulation
   denotes transcripts common to at least two Venn sectors that possess a
   different expression polarity. (B) Transcript z-ratios of significantly
   regulated transcripts in GIT2KO mice (compared to WT controls) common
   to at least two age timepoints. (C) Western blot validation of selected
   differentially regulated gene transcripts. WT data is indicated by
   black histogram bars: white histogram bars denote GIT2KO data. (D)
   Latent semantic indexing-based age-dependent ranking of biological
   functions related to GIT2KO physical and molecular phenotypes.
   Statistical significance is indicated in each figure as *p ≤ 0.05;
   **p ≤ 0.01.

Early Aging Pathological Trajectories in GIT2KO Mice

   With the combination of our whole-animal metabolic/behavioral analyses
   and our transcriptomic profiling of the GIT2KO hypothalamic, a strong
   metabolism-related pathological signature was evident. Using an
   analogous informatics process to Textrous!, i.e., GeneIndexer
   (Computable Genomix Inc.), that applies LSI to extract transcript-word
   associations in the reverse direction ([148]19, [149]62), i.e., from
   input text term to correlated transcript, we investigated how
   pathological metabolic trajectories may be evident in the GIT2KO mice
   even at these early life 2, 4, and 8 month timepoints. Therefore, we
   created lists of implicitly associated transcripts, across the three
   timepoints, linked to functional and behavioral aspects we observed in
   the GIT2KO mice, denoted by our input GeneIndexer interrogator terms
   (Table S5 in Supplementary Material; Table S6 in Supplementary Material
   – 2 m.o.; Table S7 in Supplementary Material – 4 m.o.; Table S8 in
   Supplementary Material – 8 m.o.). To assess the relative importance,
   using the hypothalamic transcriptomic datasets (2, 4, and 8 m.o.), of
   these terms at the specific experimental time points, the mean cosine
   similarity scores for the implicitly associated transcripts for each
   interrogator term were calculated. Using these mean cosine similarity
   score magnitudes, we ranked (from term 1 to term 18) the terms at each
   age timepoint – for presentation purposes we reciprocally transformed
   these ranks (Table S9 in Supplementary Material). Using this novel
   informatics approach, it is clear that multiple metabolic aspects
   (“mitochondria,” “diabetes,” “insulin,” “glucose,” “electron transfer
   chain,” “fat,” “energy,” “aging”) demonstrate, at these early life
   timepoints, a coherent age-dependent increase (Figure [150]2D).
   Interestingly, many of the behavioral factors we observed (Figure
   [151]1) that were disrupted (“thirst,” “thermogenesis,” “heat,”
   “sleep”) demonstrated an opposite age-dependent ranking mechanism
   (Figure [152]2D).

Multidimensional Signaling Network Alteration of Metabolic and Glycemic
Activity in GIT2KO Mice

   To complement our novel transcriptomic data investigation using
   LSI-based techniques, we also performed unbiased KEGG signaling pathway
   enrichment analysis on the 2-, 4-, and 8-m.o. transcriptomic datasets
   (Table S10 in Supplementary Material – 2 m.o.; S11 – 4 m.o.; S12 –
   8 m.o.). Similar to our basic transcriptomic analyses, we found that
   the predicted KEGG signaling pathway activity was age-specific, but to
   a much lesser extent than the primary transcripts, i.e., 2 m.o. 60% of
   pathways were unique; 4 m.o. 52% of pathways were unique and for 8 m.o.
   Sixty-five percent of pathways were unique (Figure [153]3A). In line
   with our CLAMS whole-body analysis and LSI-mediated transcriptomic
   investigation, the KEGG pathways demonstrating a link across the three
   timepoints (i.e., present in at least two of the age timepoints: 17
   pathways) possessed a strong focus upon energy metabolism (“Metabolic
   Pathways,” “Purine metabolism,” “Synthesis and degradation of ketone
   bodies,” “Glutathione metabolism”) as well as upon signaling functions
   associated with GIT2, such as cellular trafficking (“Endocytosis”) and
   DNA damage/repair (“Nucleotide excision repair”) (Figure [154]3B).
   Three pathways were common to all three animal ages: “Metabolic
   Pathways,” “Purine metabolism,” and “Huntington’s disease.” We have
   previously demonstrated that Huntington’s disease is often associated,
   in patients and especially in transgenic murine models of the disease,
   with considerable glycemic and metabolic dysfunction linked to
   pancreatic morphology alteration and diabetic pathological phenotypes
   ([155]53, [156]54, [157]71, [158]72). Therefore, this signaling pathway
   data reinforces the potential for GIT2 deletion to induce a systemic
   glycemic/metabolic disruption linked to central and peripheral
   degenerative actions. With respect to this pathological link, we chose
   to further investigate the high-dimensionality molecular metabolic
   phenotype of the GIT2KO mice using our KEGG data from the 8 m.o.
   timepoint as this age demonstrated the greatest linkages to
   glycemic/metabolic alterations (Figure [159]2D). Using our previously
   demonstrated KEGG signaling pathway-based keystone discovery workflow
   ([160]13), we created an LSI-based pathway-transcript matrix
   (0.67 × 10^6 total cell size: Figure [161]3C; Table S13 in
   Supplementary Material). From this matrix, we identified the 99% class
   percentile most cross-KEGG pathway conserved gene transcripts (92 in
   total: indicated in italic – Table S13 in Supplementary Material,
   Figure [162]3D-expanded view of top of heatmap in Figure [163]3C,
   Figure [164]3E-color-coded 99% percentile transcript identification).
   For illustration purposes, the top 30 cross-KEGG pathway correlated
   transcripts are indicated (Figures [165]3D,E). Many of the 92
   transcripts identified in the 99% percentile of cross-KEGG pathway
   group are critically involved in metabolic activities linked to
   diabetes-related pathophysiologies, e.g., Cox2 [cytochrome c oxidase
   subunit II ([166]73)], Coasy [CoA synthase ([167]74)], and Akt2
   ([168]75). Textrous!-based interpretation of this 92 “keystone”
   transcript dataset, using the individual processing mode [Figure
   [169]4A ([170]59)], revealed a strong link to glycemic regulation
   (“glucose-dependent,” “insulin,” “glycolysis,” “amp-activated”) and
   insulin functionality (“incretin,” “hyperglycemia,” “insulinotropic”).
   Using the ability of Textrous! to extract scientifically relevant
   noun-phrases linked to the input transcripts ([171]62), we created a
   “higher-order” word cloud using Wordle[172]^11 from these semantically
   associated nouns and noun-phrases (3773 input words in total: Figure
   [173]4B). This word cloud again underscored the importance of
   insulin-related glycemic regulation in GIT2KO mice compared to WT
   controls. Using the web-based phrase frequency counter suite of
   WriteWords[174]^12, we were able to extract from the input nouns and
   noun-phrases, the most common phrases of various word lengths (Figures
   [175]4C–H). Using this unbiased dataset description technique, we found
   sentences describing “DNA damage repair,” “fatty acid oxidation,”
   “mitochondrial activity,” and “ROS” were prominent – thus again
   confirming a functional intersection between metabolic actions and
   GIT2-related functionality [e.g., DNA repair and ROS responsivity
   ([176]18, [177]19)]. As we have consistently found a strong correlation
   between GIT2 genomic deletion and global metabolism disruption (Figure
   [178]1), predicted insulin/glucose system disruption (Figures
   [179]2–[180]4) and potential pancreatic pathology (Figure [181]2), we
   next assessed whether GIT2 genomic deletion affected the functional
   morphology of the murine pancreas.

Figure 3.

   [182]Figure 3
   [183]Open in a new tab

   KEGG signaling pathway-based keystone analysis of GIT2KO hypothalamic
   signatures. (A) Differential VennPlex analysis of hypothalamic KEGG
   signaling pathway data generated with the primary microarray
   transcriptomic data from 2-, 4-, and 8-month-old (m.o.) GIT2KO mice
   compared to age-matched WT controls. (B) Probability ranking scores for
   KEGG signaling pathways common to at least two different experimental
   age timepoints. ^#denotes KEGG pathways significantly populated by
   transcriptomic data from all three experimental timepoints. (C) KEGG
   pathway-generated cosine similarity score matrix generated using latent
   semantic output from GeneIndexer. Thirty two separate columns
   representing each significantly regulated KEGG pathway from the 8-m.o.
   hypothalamic microarray data were used to create the matrix of
   implicitly linked transcripts. Only transcripts demonstrating a
   correlation to at least two KEGG pathways were considered for analysis.
   Each level of transcript association (from 2 to 32 KEGG pathways) is
   indicated by a selective heatmap color (matrix size = 0.67 × 10^6
   cells: top and bottom of matrix are labeled). (D) Expanded inset of top
   30 most highly correlating matrix transcripts across the 32 KEGG
   pathways. (E) Class-based identification of the 92 transcripts within
   the 99% percentile of cross-pathway correlation probability.

Figure 4.

   [184]Figure 4
   [185]Open in a new tab

   Natural language processing-based analysis of multidimensional GIT2KO
   molecular pathway signatures. (A) Textrous!-based individual processing
   heatmap processing of the core 92 transcripts represented in Figure
   [186]3 comprising the multidimensional GIT2KO pathway signature. The
   strength of noun-transcript correlation is indicated by the intensity
   of teal color of the block. (B) Higher-order word cloud representation
   of expanded noun and noun-phrase output from the individual processing
   analysis in (A). The size of the resultant noun is indicative of the
   frequency in the output. (C–H) Textual phrase analysis (WriteWords*) of
   the higher-order word cloud data using length-dependent [(C) – two
   words, (D) – three words, (E) – four words, (F) – five words, (G) – six
   words, (H) – seven words] sentence analysis. Within each panel (C–H)
   the relative phrase frequencies is indicated relatively by the phrase
   text size.

   *[187]http://www.writewords.org.uk/

Genomic Deletion of GIT2 Disrupts Pancreatic Morphology

   Histochemical analysis of isolated pancreatic islets from control WT
   mice demonstrated a strong reactivity with GIT2 antibodies in
   pancreatic islet-like structures (Figure [188]5A, inset B). With
   differential histochemical staining for GIT2, insulin (for beta cells),
   or glucagon (for alpha cells), it was evident that GIT2 was present in
   both alpha and beta cells of the pancreatic islets (GIT2: Figures
   [189]5C–E, glucagon: Figures [190]5F–H). Using our previously described
   islet analytical workflow ([191]55), we found that 8-m.o. GIT2KO mice
   demonstrated a significantly reduced islet area (Figure [192]5I) and
   beta cell percentage (Figure [193]5J) with a concomitant significant
   increase in alpha cell percentage (Figure [194]5K). Linked to these
   data, we also found a significant increase in the total alpha cell area
   (Figure [195]5L), again with an expected reduction in total beta cell
   area (Figure [196]5M). Using our islet size differentiation parameters,
   we also found a general increase in the numbers of smaller islets in
   GIT2KO compared to WT controls (Figure [197]5N). Performing similar
   analysis on 2- and 4-m.o. GIT2KO mice, we found that while differences
   in pancreatic morphology compared to WT mice were evident, none however
   were significant (Figure S2 in Supplementary Material). Using visual
   morphological inspection of the 8-m.o. GIT2KO islets, it was clear that
   along with the decrease in beta cells and increase in alpha cells there
   was also an evident alpha cell involution into the beta cell mass – a
   facet indicative of diabetic pathologies in murine paradigms [WT mice:
   Figures [198]5O–Q, GIT2KO mice: Figures [199]5R–T ([200]76, [201]77)].
   This involution was generally absent in both 2-and 4-m.o. GIT2KO mice
   (Figure S3).

Figure 5.

   [202]Figure 5

   [203]Figure 5
   [204]Open in a new tab

   Genomic GIT2 deletion affects pancreatic islet functional structure.
   (A) DAB- (3,3′-diaminobenzidine) based immunohistochemical
   identification of GIT2-associated reactivity in discrete islet-like
   structures in WT murine pancreas. (B) High-magnification inset of
   DAB-stained, GIT2-immunoreactive pancreatic islet-like structure. (C)
   GIT2 AF488 immunostaining of WT pancreatic islet. (D) Glucagon AF568
   immunostaining of WT pancreatic islet. (E) Merge of panels (C,D)
   indicating AF488–AF568 colocalization. (F) Insulin AF488 immunostaining
   of WT pancreatic islet. (G) GIT2 AF568 immunostaining of WT pancreatic
   islet. (H) Merge of panels (F,G) indicating AF488–AF568 colocalization.
   MATLAB-based pancreatic islet mathematical automated assessment of WT
   (black bars) or GIT2KO (white bars) pancreas islet area (I), β-cell
   percentage (J), α-cell percentage (K), α-cell area (L), β-cell area (M)
   and islet number/size distribution (N). (O) Insulin AF488, glucagon (P)
   immunostaining of WT islet [8 month old (m.o.)]. (Q) Merge of panels
   (O,P). (R) Insulin AF488, glucagon (S) immunostaining of GIT2KO islet
   (8 m.o.). (T) Merge of panels (R,S). Statistical significance is
   indicated in a specific panel as *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

Genomic Deletion of GIT2 Disrupts Plasma Membrane Metabolic Hormone Levels
and Pancreatic Functionality

   To further investigate the systemic metabolic effects of genomic GIT2
   deletion, we measured the levels of multiple metabolic hormones
   [insulin, amylin, leptin, GIP (gastric inhibitory polypeptide), PP
   (pancreatic polypeptide), PYY (peptide YY)] in 8-m.o. GIT2KO mice
   compared to WT controls (Figures [205]6A–F). Deletion of GIT2 resulted,
   at the 8-m.o. timepoint, in significant reductions in plasma insulin,
   amylin, and PYY along with a significant increase in the PP levels.
   Significant age-dependent reductions of insulin were observed at the 2-
   and 4-m.o. timepoints (Figure S4): trends for non-significant
   reductions in both amylin and leptin where observed (Figure S4) as were
   non-significant increases in plasma GIP (Figure S4). Consistent with
   the multiple molecular glycemic pathologies observed in the GIT2KO
   mice, we found that as early as 4 months of age (and continuing onto
   8 m.o.), a significantly higher fasting plasma glucose was evident
   compared to age-matched WT controls (Figure [206]6G). We investigated
   whether this elevated plasma glucose was associated with a reduction in
   the insulinotropic functionality of GIT2KO mice. Using a standard ITT
   approach, we found that the 8-m.o. GIT2KO demonstrated a significant
   degree of insulin resistance to a bolus dose of Lantus^® (Figure
   [207]6H). Therefore, along with the disruption of beta cell mass, alpha
   cell involution, and reduced plasma insulin levels the GIT2KO mice also
   demonstrate significant insulin resistance. Performing standardized
   OGTT analysis, we also found that GIT2KO mice (at 8 m.o.) demonstrated
   a significantly lower ability to uptake a glucose bolus (Figure
   [208]6I). Coordinating with this reduced glucose uptake ability, we
   found that GIT2KO mice possessed a significantly diminished capacity to
   functionally secrete insulin in response to the glucose bolus (Figure
   [209]6J).

Figure 6.

   [210]Figure 6
   [211]Open in a new tab

   GIT2 genomic deletion affects multiple circulating energy-regulatory
   factors and response to external glycemic challenges. Differential
   circulating plasma levels of insulin (A), amylin (B), leptin (C), GIP
   [gastric inhibitory peptide: (D)], PP [pancreatic polypeptide: (E)],
   and PYY [peptide YY: (F)] between 8-m.o. (month old) WT (black bars)
   and GIT2KO (white bars) mice. (G) Fasting blood glucose measurements
   for 2-, 4-, 8-m.o. WT (black bars) or GIT2KO (white bars) mice. For
   (H–J) WT or GIT2KO mouse data is represented by black circles or open
   triangles respectively. (H) Standard bolus insulin tolerance test (ITT:
   0.5U/kg insulin) performed on 8-m.o. WT or GIT2KO mice. (I)
   Time-dependent elevation of plasma glucose in 8-m.o. WT or GIT2KO mice
   after intragastric introduction of a glucose bolus (2g/kg). (J)
   Time-dependent elevation of plasma insulin in 8-m.o. WT or GIT2KO mice
   after intragastric introduction of a glucose bolus (2 g/kg). For each
   experimental panel, statistical significance is indicated in each
   figure as *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

High-Dimensionality Transcriptomic Profiling of GIT2KO Pancreatic Islets

   From our metabolic, behavioral, informatic, and hormonal analyses, a
   clear disruption of pancreatic function in GIT2KO mice is evident. We
   chose to study this at a high-dimensionality level, using quantitative
   transcriptomics on mechanically isolated islets from 8-m.o. GIT2KO
   mice. Using purified islets from multiple GIT2KO or WT (n ≥ 3) mice, a
   differentially regulated transcriptomic profile was generated (Table
   S14 in Supplementary Material). At the basic transcriptomic level
   multiple transcripts closely linked to diabetic pathology and islet
   dysfunction were differentially regulated in GIT2KO compared to WT
   islets, e.g., Slc2a2 [glucose transporter 2 ([212]78)], Reg3a/b
   [regenerating islet-derived protein 3 alpha/beta ([213]79)], Ffa2 [free
   fatty acid receptor 2 ([214]80)], Trib3 [tribbles homolog 3 ([215]81)],
   PP [pancreatic polypeptide ([216]82)], Sesn1 [sestrin 1 ([217]83)],
   Glo1 [glyoxylase 1 ([218]84)], and Lars2 [leucyl-tRNA synthetase 2,
   mitochondrial ([219]85)]. Using “pancreas-specific” tissue database
   signaling investigation with IPA, we applied Disease/Bio-Function
   annotation to the GIT2KO islet transcriptome data (Table S15 in
   Supplementary Material). Corroborating our histochemical and functional
   pancreatic data, we found several prominent beta-cell focused functions
   (e.g., “development of pancreas,” “abnormal morphology of pancreas,”
   “abnormal morphology of beta islet cells”) within this unbiased
   analysis output (highlighted in Figure [220]7). To complement our
   Disease/Bio-Function annotation, we also assessed the most coherently
   associated functional networks within the GIT2KO transcriptomic data.
   The two highest scoring networks (based on numbers of enriched
   transcripts from the input data that form a coherent-curated network)
   focused on Scl2a2-Hnf4a-Hnf1a (Figure S5 in Supplementary Material) and
   Pdx1-Ins1-Glis3 (Figure S6 in Supplementary Material) were both
   officially (via IPA) denoted as involved in “Cell Death and Survival –
   Endocrine System Disorders – Metabolic Disease.” As several
   well-characterized murine pancreatic beta cell clonal lines exist,
   e.g., Beta-TC-6 (ATCC: abbreviated to “TC-6” hereafter), we decided to
   orthogonally investigate the link between GIT2 and several factors
   identified in our transcriptomic analysis using siRNA-mediated
   depletion of GIT2 followed by selective Western blot analysis (Figure
   [221]8), using the extracellular signal-regulated kinase 2 (Erk2) as a
   specific, well-characterized protein-gel loading control ([222]86,
   [223]87). Using in vitro siRNA-mediated GIT2 depletion (50–400 nM
   siRNA: Figure [224]8A), we were able to significantly attenuate the
   expression of GIT2 in the TC-6 cells. Employing the 400 nM
   concentration of level of GIT2 siRNA to reduce cellular GIT2 levels, we
   were able to recapitulate selected expression data generated from our
   microarray analysis, with specific relevance to beta cell function,
   metabolism, diabetes, and aging (Table S15 in Supplementary Material:
   Figures [225]8B–I). Hence acute siRNA; mediated reduction of cellular
   GIT2 caused increased expression of Dcx [doublecortin, Figure [226]8C
   ([227]88)], Glo1 [glyoxylase I, Figure [228]8D ([229]89)], Ndufb10
   [NADH dehydrogenase (ubiquinone) 1 beta subcomplex subunit 10, Figure
   [230]8E ([231]90)) as well as a decreased expression of Mrpl12
   [mitochondrial ribosomal protein L12, Figure [232]8F ([233]91)], Atm
   [ataxia telangiectasia mutated, Figure [234]8G ([235]18)], Reg3b
   [regenerating islet-derived 3 beta, Figure [236]8H ([237]92)], and
   Sesn1 [sestrin 1, Figure [238]8I ([239]93)].

Figure 7.

   [240]Figure 7
   [241]Open in a new tab

   Disease Bio-Function analysis of pancreatic islet transcriptomic data.
   Significant transcriptomic data from isolated GIT2KO pancreatic islets
   was functionally annotated using the Disease/Bio-Function suite from
   IPA Pathways Analysis. The regulation of three processes vital for
   pancreatic function is indicated: “development of pancreas, ‘abnormal
   morphology of pancreas,’ abnormal morphology of beta islet cells.” The
   nature of the specific node components and their type of functional
   interaction (noted as arrows between factors) are described in the
   associated key. Red factors were upregulated in the input dataset while
   green factors were downregulated (GIT2KO versus WT).

Figure 8.

   [242]Figure 8
   [243]Open in a new tab

   Depletion of GIT2 in murine pancreatic TC-6 cells modulates the
   expression of multiple proteins involved in stress management and
   energy regulation. (A) Dose-dependent GIT2 depletion curve for GIT2
   siRNA. A random sequence “scramble” siRNA was employed as a negative
   control. Extracellular signal-regulated kinase 2 (Erk2) was employed as
   a loading control throughout this figure. GIT2 siRNA-mediated (using
   400 nM siRNA concentration) modulation of TC-6 cells expression of GIT2
   (B), doublecortin [Dcx: (C)], glyoxylase 1 [Glo1: (D)], NADH
   dehydrogenase (ubiquinone) 1 beta subcomplex, 10 [Ndufb10: (E)],
   mitochondrial ribosomal protein L12 [Mrpl12: (F)], ataxia
   telangiectasia mutated [Atm: (G)], regenerating islet-derived 3 beta
   [Reg3b: (H)] and sestrin 1 [Sesn1: (I)]. Statistical significance is
   indicated in each figure as *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

GIT2 Functionality in Murine Pancreatic Beta Cells

   From our transcriptomic and functional analyses, it is evident that
   GIT2 integrity is important in regulating insulin sensitivity and beta
   cell function. We next investigated the molecular nature of GIT2
   activity, related to glycemic control, in beta cells using the TC-6
   murine model cell line. As the GIT2KO mice demonstrate insulin
   resistance, we mimicked this pathophysiological state in the TC-6 cells
   by chronic (24 h) treatment with palmitic acid ([244]94). In response
   to palmitic acid exposure, we found both increases in Glut2 as well as
   GIT2 (Figure [245]9A), here again we employed the extracellular
   signal-regulated kinase 2 (Erk2) as a specific, well-characterized
   protein gel loading control ([246]86, [247]87). Extension of the
   chronic exposure to 4 days (10 μM palmitic acid – replenished every
   6 h) demonstrated that the GIT2 expression was subsequently attenuated
   while the Glut2 potentiation persisted, mimicking our mouse model
   paradigm of GIT2KO status with an elevation of beta cell Glut2/Slc2a2
   (Table S15 in Supplementary Material; Figure [248]9B). We assessed the
   viability of the TC-6 cells in response to extended palmitate
   treatments using Trypan Blue exclusion cell counting and failed to
   demonstrate any significant loss of viability compared to actinomycin
   D-treated (2 μg/ml, 2-h exposure) positive controls (Figure S7 in
   Supplementary Material). In this in vitro experimental system, we then
   assessed the ability of the palmitic acid-treated cells to uptake
   glucose. Performing an in vitro glucose uptake assay using 5 mM glucose
   ([249]19), we found that cells possessing a diminished GIT2 in the
   presence of elevated Glut2 have an attenuation of glucose uptake
   capacity compared to non-treated cells (Figure [250]9C). Using
   Q-proteome^®-based subcellular fractionation of these palmitic
   acid-treated cells (4 days treatment), we found that in response to the
   introduction of the glucose bolus, the pre-treatment of cell with
   palmitic acid significantly affected the glucose-induced subcellular
   redistribution of GIT2 (Figure [251]9D). Appropriate subcellular
   fractionation of the cellular samples with the Q-proteome^® system was
   assessed using specific western blots for marker proteins characterized
   by cytoplasmic (Gapdh), plasma membrane (Pmca), nuclear/large organelle
   (Lmna), or cytoskeletal (Tuba1a) expression (Figure S8 in Supplementary
   Material). In control cells, GIT2 immunoreactive signals increased in
   the cytoplasmic and plasma membrane fractions in response to the
   glucose dose. This increase, in both cytoplasmic and plasma membrane
   compartments, was significantly attenuated in the palmitate-treated
   cells. No significant dynamic changes in GIT2 immunoreactive signals
   were observed in the nuclear fractions, however an elevation of basal
   nuclear GIT2 was observed in the palmitate-treated cells (Figure
   [252]9D). This effect is likely caused by the metabolic stress of
   chronic palmitate treatment causing ROS damage as we have previously
   demonstrated that multiple cellular stressors promote GIT2 nuclear
   translocation to regulate the DNA damage response ([253]18). These data
   therefore suggest that molecular attenuation of GIT2 affects glucose
   sensitivity of beta cells and cellular GIT2 redistribution is important
   for cellular functions such as glucose uptake. In line with these
   findings, we also noted that siRNA-mediated partial GIT2 depletion
   engendered a consistent but moderate (non-significant) attenuation of
   glucose uptake in TC-6 cells (Figure S9 in Supplementary Material). As
   GIT2 was first identified as a GPCR-associated scaffolding protein
   ([254]95), its functions are typically mediated via assembling and
   regulating the composition of multiprotein complexes. Therefore, using
   selective immunoprecipitation in control, non-diabetic, and
   pathophysiological conditions from whole pancreas, we next investigated
   the nature of its physical molecular interactions in these paradigms.

Figure 9.

   [255]Figure 9
   [256]Open in a new tab

   Pathological manipulation mimicking insulin resistance in murine TC-6
   cells affects GIT2 functionality and subcellular disposition. Treatment
   (1 day) of TC-6 cells with palmitic acid (10 μM) and mimicking an
   insulin-resistant cellular status potentiates GIT2 and Glut2 expression
   (A). Long-term palmitic acid treatment (2–4 days) differentially and
   biphasically regulates GIT2/Glut 2 expression (B). Extracellular
   signal-regulated kinase 2 (Erk2) was employed as a loading control in
   (A,B). Attenuation of bolus (5 mM) glucose uptake in TC-6 cells by
   long-term (4 days) palmitic acid treatment (C). Palmitic acid treatment
   (4 days) of TC-6 cells affects glucose-driven subcellular
   redistribution of GIT2 across the cytoplasmic, plasma membrane, and
   nuclear microdomains (D). Statistical significance is indicated in each
   figure as *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

Pathological Metabolic Alteration of GIT2 Interaction Partners

   As GIT2 genomic deletion induces a strong metabolic phenotype to
   appreciate the effects of metabolic stress upon GIT2 functionality, we
   next investigated the potential functional interaction partners of GIT2
   in pancreatic tissue from experimental animals experiencing significant
   metabolic stress, i.e., diabetic db/db mice ([257]55, [258]96). Db/db
   mice possessing a mutation in the leptin receptor demonstrate an
   increased propensity for diet-induced obesity and diabetes via
   congenital hyperphagia ([259]97). Upon inspection of the GIT2
   expression levels in 8-m.o. WT control and db/db mice, in both the
   hypothalamus and pancreas GIT2 expression was significantly elevated on
   db/db mice compared to the controls (Figure [260]10A). Therefore, in a
   pro-diabetic state, GIT2 expression is elevated, suggesting that GIT2
   expression is not only age- and stress-sensitive ([261]13, [262]18,
   [263]19), but also sensitive to the alteration of “metabolic” age in
   the db/db model ([264]42, [265]98–[266]100). To identify the effects of
   diabetes-related pathology on GIT2 protein–protein complex
   interactions, we performed selective GIT2 co-immunoprecipitation
   (co-IP) (with pre-immune sera as a negative control) in WT and db/db
   pancreatic cell lysates (Figure [267]10B). Due to the elevation of GIT2
   expression in db/db samples (Figures [268]10B,C), we normalized for
   GIT2 content of the co-IP samples used for subsequent Western blotting
   or MS analysis. Investigation of extracted and digested GIT2 co-IP
   samples with our linear ion trap LC-MS/MS identified 170 specific
   co-precipitating proteins from WT pancreatic lysates (Table S16 in
   Supplementary Material) and 153 specific co-precipitating proteins from
   db/db pancreatic lysates (Table S17 in Supplementary Material). The
   majority of these proteins were conserved (65%) between WT and db/db
   co-IP extractions (Figure [269]10D). However, it was interesting to
   note that within the proteins found only in the WT venn sector, i.e.,
   lost with metabolic dysfunction, multiple glycemic-related factors were
   evident (Table S18 in Supplementary Material), e.g., the insulin
   receptor (Insr), insulin receptor substrate 2 (Irs2), fatty acid
   binding protein 2 [Fabp2 ([270]101)], signal sequence receptor delta
   [Ssr4 ([271]102)] and G protein-coupled receptor 101 [Gpr101
   ([272]103)]. In addition to these WT-specific metabolic factors, it was
   also evident that additional proteins within the WT-only venn sector
   were associated with other activities strongly implicated in
   pathological aging including molecular aging, DNA damage and stress,
   e.g., structural maintenance of chromosomes 5 [Smc5 ([273]104)],
   eukaryotic translation initiation factor 2B [Eif2b2 ([274]105)],
   negative regulator of ubiquitin-like proteins 1 [Nub1 ([275]106)], and
   DEAD (Asp–Glu–Ala–Asp) box helicase 3, X-linked [Ddx3x ([276]107)]. We
   validated our MS data using selective Western blotting for several of
   the GIT2 co-precipitating factors unique to WT or db/db tissues (Figure
   [277]10E). Hence, we found a reduced association between GIT2 and Insr,
   Irs2, and Smc5 in the db/db samples. In contrast, we found an increased
   association between GIT2 and Rpl17 (ribosomal protein L17), Fau/MNSFβ
   (Monoclonal non-specific suppressor factor β), and Fbxo24 (F-Box Only
   Protein 24/IKK1) in the pathological db/db state. Interestingly, both
   Fau/MNSFβ and Fbxo24 are associated with apoptotic activity linked to
   inflammation and DNA damage ([278]108, [279]109). Proteins commonly
   co-precipitating with GIT2 in both WT and db/db conditions (Map1a,
   Pdrg1, Caskin2) were also validated and were found to equally associate
   with GIT2 in both conditions (Figure [280]10D). In addition to the
   investigation of individual GIT2 protein–protein complex interactions,
   we also investigated the potential cell signaling ramifications of
   these multiple associations. Using both KEGG (WT mice: Table S19 in
   Supplementary Material, Table S20 in Supplementary Material: db/db
   mice) and IPA Canonical Signaling Pathway analyses (WT mice: Table S21
   in Supplementary Material, Table S22 in Supplementary Material: db/db
   mice) of the respective co-IP datasets we again demonstrated a strong
   distinction between the functional links of GIT2-associated proteins in
   control WT conditions compared to db/db (Figures [281]10F,G).
   Inspection of the selective WT-unique signaling pathways (Figures
   [282]10E,F) revealed a strong link of GIT2-associated proteins with
   energy regulation (“Type II diabetes mellitus signaling,” “Fat
   digestion and absorption,” “PPAR signaling pathway,” “Maturity Onset
   Diabetes of Young (MODY) Signaling,” “AMPK Signaling”), cellular
   development (“Notch signaling”) and degenerative diseases linked to
   metabolic dysfunctions [“Huntington’s disease,” “Amyotrophic lateral
   sclerosis signaling” ([283]53, [284]110)]. As expected from the
   scaffolding nature of GIT2 and its well-characterized role in
   cytoskeletal regulation ([285]111), a strong representation of such
   factors was evident (Figures [286]10E,F: “Regulation of actin
   cytoskeleton,” “Focal adhesion,” “Actin Cytoskeleton Signaling”).

Figure 10.

   [287]Figure 10
   [288]Open in a new tab

   Pathological metabolic status directly affects GIT2 expression. (A)
   Advanced metabolic stress, independent of chronological age,
   potentiates GIT2 expression in hypothalamic and pancreatic tissues. (B)
   Selective immunoprecipitation of GIT2 (using pre-immune IgG as a
   negative control) from WT or db/db pancreatic whole-cell lysates.
   Coomassie staining of lysates is employed to indicate equal protein
   input for immunoprecipitation experiments. (C) Normalization of
   immunoprecipitated GIT2 from pancreatic lysates to adjust for
   db/db-induced elevation of GIT2. Normalized GIT2 co-immunoprecipitation
   extracts were then used for further semi-quantitative Western blot or
   mass spectrometry (MS) analysis. (D) Proportional Venn diagram analysis
   of the protein compositions of GIT2 co-immunoprecipitates from WT
   (solid line) of db/db (dashed line) pancreatic lysates. (E) Western
   blot validation of MS-identified GIT2-associating partners in WT and
   db/db pancreatic tissue lysates. WT data is indicated by black bars
   while db/db data is represented by white bars. (F) Proportional Venn
   diagram analysis of KEGG pathways significantly populated by GIT2
   co-immunoprecipitation data from either WT (solid line) or db/db
   (dashed line) pancreatic lysates. The associated pie chart indicates
   the relative KEGG pathway hybrid scores [(−log[10] (enrichment
   probability)) × enrichment ratio] for the 16 KEGG pathways unique to
   the WT dataset. Pathways related to metabolism, cytoskeletal signaling
   and disease are highlighted in bold. (G) Proportional Venn diagram
   analysis of IPA Canonical Signaling pathways significantly populated by
   GIT2 co-immunoprecipitation data from either WT (solid line) or db/db
   (dashed line) pancreatic lysates. The associated pie chart indicates
   the relative Canonical pathway score enrichment probabilities (−log[10]
   transformed) for the 15 Canonical Signaling pathways unique to the WT
   dataset. Pathways related to metabolism, cytoskeletal signaling and
   disease are highlighted in bold. Statistical significance is indicated
   in each figure as *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001.

Discussion

   Given the inevitable worldwide rise in the incidence of age-related
   disorders, it is imperative that a more nuanced understanding of the
   systemic control of the multiple processes that underpin aging is
   developed. Intensively complex, multidimensional events such as whole
   somatic aging are likely to be coordinated by a series of differential
   protein hierarchies, in which lower-dimensionality protein functions,
   e.g., simple mitogen-activated protein kinase (MAPK) cascades are
   orchestrated and connected to other linear pathways by
   master-controlling “keystone” or “hub” proteins ([289]13). Experimental
   research into the potential molecular interactions that regulate aging
   across an organisms life span quickly engenders signaling complexity
   issues comparable to the mathematical difficulties of optimally solving
   NP-hard (Non-deterministic Polynomial-time hard) or NP-complete
   calculations ([290]112, [291]113). Therefore, a more pragmatic approach
   to the investigation of hyper complex signaling architectures, both for
   mechanistic knowledge and for eventual therapeutic manipulation of
   highly complex systems, is to identify the central coordinators of
   these complex systems. We have previously illustrated a mechanism for
   uncovering such “keystone” regulatory factors [also known as “hub”
   proteins ([292]114) using our combinatorial bioinformatics approach
   ([293]13)]. Performing such an analysis with primary protein-based
   data, extracted from different aged samples of an endocrine organ
   thought to house the regulatory machinery of aging, i.e., the
   hypothalamus ([294]24), we identified the GPCR-interacting protein GIT2
   as a “keystone” of the aging process. Subsequently we demonstrated that
   this protein forms a link between age-related cellular damage (ROS,
   ionizing radiation) and the maintenance of DNA repair/stability – a
   process crucial to successful and healthy aging ([295]18). Genomic
   deletion of GIT2 resulted, with respect to the presence of accumulated
   DNA damage, in accelerated aging of murine CNS tissue. As the
   age-dependent development of insulin resistance and the deterioration
   of glucose-focused oxidative phosphorylation efficiency contribute
   significantly to accelerated aging mechanisms, we sought to investigate
   the potential role of GIT2 in systemic metabolic regulation. Using
   whole-somatic metabolic chamber analysis, we found that GIT2KO mice
   demonstrate a significantly reduced RER at a relatively young
   (4 months) age indicating a switch away from glucose use toward adipose
   and then, potentially, protein to generate usable energy (Figure
   [296]1). Upon investigation of the hypothalamic transcriptomic profiles
   in GIT2KO mice at younger and older ages (2 and 8 months old
   respectively), we uncovered, using multiple bioinformatics techniques,
   a core transcriptomic “signature” linked to metabolic/glycemic
   regulatory activity (Figure [297]2D, Figure S1). We further reinforced
   this observation using classical pathway annotation pipelines (KEGG
   pathway analysis: Figure [298]3) as well as novel natural language
   processing-based investigations using the hypothalamic transcriptomic
   data (Figures [299]3 and [300]4). In these highly nuanced data outputs,
   a strong intersection between classical GIT2 functions (endocytosis and
   cytoskeletal organization; Figures [301]3 and [302]4) ([303]111), novel
   functions (DDR; Figure [304]4) ([305]18), and metabolic activity
   (glucose, mitochondrial, insulin, and fatty acid oxidation: Figure
   [306]4) was evident. Given the multiple indications of an important
   role of GIT2 in glycemic regulation, we investigated a potential direct
   impact of GIT2 in pancreatic tissues that secrete insulin and related
   metabolic factors. We found that GIT2 was expressed in pancreatic alpha
   and beta islet cells and that genomic deletion in GIT2KO mice resulted
   in a disruption of both beta and alpha cell distribution in the islets.
   Reinforcing the potential for GIT2 to act as a neurometabolic
   controller, we also found moderately diminished levels of the
   GIT2-associaciated signaling factors Pak1 (p21 activated kinase 1) and
   beta-PIX (Arhgef7, Rho guanine nucleotide exchange factor 7) in both
   GIT2KO hypothalamus and the pancreas compared to WT controls (Figure
   S10 in Supplementary Material). Arhgef7 has been demonstrated to play a
   role in insulin secretion via regulation of Cdc42 (cell division
   control protein 42 homolog) activity ([307]115). In addition the
   downstream GIT effector Pak1, through physical binding and eventual
   phosphorylation, can control the activity of phosphoglucomutase 1, an
   important regulatory enzyme in cellular glucose utilization and energy
   homeostasis ([308]116). In pancreatic β-cells, Pak1 is also involved in
   insulin granule localization and vesicle release ([309]117), as well as
   playing a role in pancreatic cell development and incretin generation
   ([310]118).

   At the gross pancreatic histological level the GIT2KO mice effectively
   demonstrated (primarily at 8 m.o.) a classical alpha-cell involution of
   the insulin-secreting beta cell mass (Figure [311]5) ([312]77). This
   was accompanied by reductions in circulating levels of insulin, amylin,
   leptin, and peptide YY in 8-m.o. GIT2KO mice compared to WT controls
   (Figure [313]6). These multiple pancreatic alterations were associated
   with elevated fasting blood glucose, insulin (Figure [314]6), and oral
   glucose tolerance and a diminished insulin secretion capacity in
   response to glucose stimulation (Figure [315]6). We further reinforced
   the validity of these findings via a high-dimensionality transcriptomic
   investigation of isolated GIT2KO pancreatic islets. Using IPA
   Disease/Bio-Function analysis of the islet data, a strong unbiased
   demonstration of beta cell developmental disruption was revealed
   (Figure [316]7). Using murine beta cell culture models, we also found
   that the functional activity of GIT2 in these cells was disrupted by
   cellular stress that mimics insulin resistant pathophysiology (Figure
   [317]9), i.e., palmitate treatment. While representing a commonly
   employed experimental process ([318]119–[319]122), it is highly likely
   that palmitate-induced insulin resistance is a complex multifactorial
   process that can involve multiple collateral signaling systems effects
   including altered ceramide levels ([320]123), endoplasmic reticulum
   stress ([321]124), mTOR/S6K activation ([322]125), activation of
   protein phosphatase 2A ([323]126), modulation of TNF-α signaling
   ([324]127), and alterations in transmembrane free fatty acid receptor
   activity ([325]128–[326]130). It is interesting to note however, with
   specific respect to the functional intersection between insulin
   resistance and the aging process, Nakamura et al. ([327]131) suggested
   that palmitate-induced insulin resistance in their model was strongly
   associated with mitochondrially derived ROS, a pathological process
   that is considered one of the hallmarks of aging ([328]132). Linked to
   this it is therefore not surprising that in addition to its elevation
   in aging ([329]13), GIT2 expression is enhanced by both direct oxygen
   radical exposure ([330]19) and here in our current research with
   palmitate treatment. In line with the interconnection between aging and
   metabolic paradigms we subsequently demonstrated that in the db/db
   pathological paradigm GIT2 was prematurely upregulated compared to WT
   control mice, both in the hypothalamus and pancreas (Figure [331]10).
   In addition, we found in pancreatic tissues that the physical
   association of GIT2 with proteins involved in the glucose
   metabolic/insulin-regulatory system, e.g., insulin receptor and insulin
   receptor substrate 2, were disrupted in the db/db pathological state
   (Figure [332]10). With respect to the large number of potential
   GIT2-associating factors identified in our MS experiments, it is
   interesting to note that multiple interactome databases have already
   empirically identified well over 50 different partners, e.g., BioGRID
   reports 59 different GIT2 interactors[333]^13 and IntAct identifies at
   least 75 distinct GIT2 interactors[334]^14. The introduction of
   high-dimensionality proteomics-based protein complex analysis has
   greatly expanded the ability to discover both direct binary and
   co-complex binding partners ([335]133–[336]136). We believe that due to
   its widespread expression ([337]46) and modular structure, GIT2 is
   likely to possess a broad spectrum of potential interacting partners
   and therefore our finding of a large number of potentially interacting
   proteins may not unexpected. GIT2 possesses multiple protein–protein
   interaction domains including a GAP (GTPase-activating protein) domain,
   two paxilin-binding subdomains (PBS1, PBS2), three ankyrin repeats, a
   Spa2-homology domain (SHD), a focal adhesion-targeting homology (FAH)
   domain, a coiled-coil region, a leucine zipper motif, and a synaptic
   localization domain ([338]111). Hub or keystone proteins, which we
   contend GIT2 may be, are considered to play a potentially pivotal role
   in organizing complex molecular signaling networks via their ability to
   exert a more trophic level of signaling control ([339]137–[340]139).
   Using an unbiased discovery process of identifying proteins controlling
   multidimensional neurometabolic signaling networks in aging, we
   previously sought to discover target proteins that were associated with
   significantly more multiple signaling pathways compared to other
   proteins ([341]13). Therefore, it not surprising that GIT2 may indeed
   function as a hub protein as aging affects nearly every physiological
   system in the body. The seemingly large array of protein interactions
   suggested for such hub proteins may reflect the ability of multiple
   proteoforms of the same protein to interact physically with several
   distinct signaling protein complexes simultaneously ([342]140).
   Therefore, for any given protein, there are likely to be an ensemble of
   diverse interactomes that, unless differentially purified, will
   generate the appearance of a huge number of potential interacting
   proteins.

   Highly connected hub proteins have been intensively studied in recent
   years as they potentially represent important therapeutic targets
   ([343]141). Hub proteins have been themselves categorized into two
   general types, i.e., “Party hubs” and “Date hubs” ([344]139, [345]142).
   The “Party” types are hubs whose expression is correlated with the
   expression levels of their interaction partners while “Date” types do
   not demonstrate this correlation. Linked to this distinction “Party”
   hubs are considered more likely to connect proteins within functional
   signaling modules, while “Date” hubs are more likely to connect
   different functional signaling modules. ”Party” hubs are generally
   considered to be multi-interface proteins whereas “Date” hubs are more
   frequently single-interaction interface proteins ([346]143). However,
   such strong distinctions of the nature of hub proteins though are
   unlikely to be indicative of the reality of multifunctional proteins
   ([347]144, [348]145). In the case of GIT2, and potentially many other
   “hub” proteins it is likely that it can, at various times depending on
   the specific cellular context, act as a “Party” hub, e.g., controlling
   skeletal dynamics, and act at times as a “Date” hub and cross-regulate
   multiple pathways involved in aging and metabolism, e.g., glycemic
   signaling and DNA repair. With regards to the pathophysiological
   functionality of GIT2, it is interesting to note that research over the
   last decade has generally concluded that human disease is rarely the
   consequence of an isolated abnormality in a particular gene or protein
   but is probably due to a series of complex perturbations in an
   underlying cellular network ([349]141), in this case the neurometabolic
   “axis.” The structure of these networks/axes is likely governed, at
   some level by “hub” or keystone proteins, whose alterations can exert
   changes in their global properties thus linking their function to
   multiple disease processes ([350]141, [351]146, [352]147). Further
   research into hub protein function is likely to uncover the biological
   significance of these network controllers in disease etiology
   ([353]147) as well as help in the delineation of trophic biomarkers and
   novel drug targets ([354]141).

   Our multifactorial analyses concerning the physiological status of the
   GIT2KO mice reveals a potentially important factor for the generation
   of aging trajectories, i.e., the metabolic shift away from primary
   glucose use to glucose/adipose use and then finally to adipose/protein
   use. Each of these transitions results in the reduction of RER away
   from a value of 1.0, i.e., glucose as primary energy source, toward
   0.85–0.8, i.e., glucose/adipose as primary energy source or toward
   <0.8, i.e., adipose/protein as a primary energy source. Our observed
   decline in RER of relatively young GIT2KO mice is in line with reports
   of age-dependent shifts in C57BL/6 mice ([355]148). Maintaining
   metabolic efficiency, through preservation of the utilization of
   glucose as the primary energy source, in early and later life, will
   potentially reduce hyperglycemia-induced chronic inflammation
   ([356]149, [357]150), attenuate generation of ROS ([358]151, [359]152),
   and also support the activity of energetic processes such as membrane
   electrical excitability ([360]153), activity of cation pumps
   ([361]154), unfolded protein management ([362]155, [363]156) and
   nucleic acid damage ([364]157, [365]158). Therefore, in-part, the
   transition from “healthy” aging to “pathological” aging may coincide
   with a significant downshift in RER as nearly all of the processes
   entrained by energy insufficiency result in feed-forward pathological
   loops. An enhanced understanding of the molecular events that could
   predict this event would therefore provide potential remedial
   mechanisms to postpone this transition. As GIT2 possesses important
   roles in stress sensitivity ([366]18, [367]19, [368]159), aging
   ([369]13), and somatic energy management, this molecule potentially
   provides a novel target for therapies designed to interdict
   pathological aging and age-related disease. With respect to potential
   therapeutic exploitation of this keystone factor, our transcriptomic
   and proteomic analyses have demonstrated that GIT2 is strongly linked
   with multiple therapeutically tractable protein targets, e.g.,
   glyoxylase 1 [Glo1: Figure [370]8 ([371]160–[372]163)] and sestrin 1
   [Sesn1: Figure [373]8; ([374]82, [375]92, [376]164, [377]165)].
   Considerable further experimentation will be clearly required to
   uncover the optimal therapeutic mechanisms of exploiting GIT2
   expression for the remediation of age-related pathologies, however, at
   the present time, it represents an interesting new master-regulator of
   the neurometabolic aging process.

   With the advent of such a nuanced hierarchical approach, specifically
   of functional keystone factors, to the understanding of physiological
   and disease protein networks our concept of disease causation needs to
   take this into account. If potential keystone factors such as GIT2 are
   more important than other proteins with respect to their organization
   of other signaling modalities, it would be presumed that genomic
   deletion/modification may result in the generation of significant
   health issues. In our hands, aside from the metabolic alterations
   studied here, we have found that with post-mortem tissue extraction
   from aged GIT2KO males (12–18 months of age), we have identified
   consistent structural deformation of their upper gut (primarily ileal
   distention) as well as renal/hepatic adipose deposition. It is
   interesting to note however that potential “hub” or keystone proteins,
   as we propose GIT2 to be an example of, while being crucial for the
   organization of complex molecular events are not often the sole causes
   of disease ([378]114, [379]166). Due to their trophic level of
   signaling pathway interactivity hub proteins are not often observed as
   genomic loci of disease ([380]142) but if they are perturbed they will
   likely generate a wide spectrum of related pathologies. For example,
   genetic alteration of GIT2 has been identified in a GWAS risk allele
   screen for MetS ([381]167). MetS represents a multifactorial disorder
   including an incredibly broad array of pathophysiologies including
   hypertriglyceridemia, insulin resistance, hypertension, vascular
   inflammation, atherosclerosis and renal, liver, and heart diseases
   ([382]168). Therefore, it is perhaps not surprising that the GIT2KO
   mice do not present such a singular dramatic pathological phenotype as
   the effects of GIT2 deletion are likely to be highly multidimensional:
   GIT2 genomic deletion has already been shown to affect central nervous
   system aging and DNA repair ([383]18), immune function ([384]169), bone
   mineral density, bone marrow adiposity ([385]170), and anxiety/stress
   phenotypic behavior ([386]159). Also, associated with issues of
   congenital knockout models, it is clear that there is a strong
   reflexive survival response of the GIT2KO mice to metabolic disruption.
   For example, we have noted that while presenting a distinct metabolic
   phenotype, male GIT2KO mice on a regular chow diet (not high in fat or
   glucose) maintain a relatively similar bodyweight across their
   moderately reduced total life span. This interesting phenomenon is
   currently the subject of our ongoing research. In addition, the GIT2KO
   mice appear to be attempting to systemically ameliorate the deleterious
   effects of elevated glucose levels by increasing the expression of
   Glo1. Protracted exposure to excessive blood glucose can lead to the
   formation and accumulation of advanced glycation endproducts (AGEs)
   that are related to diabetes and other age-related diseases.
   Methylglyoxal (MGO), a highly reactive dicarbonyl compound, is
   considered to be one of the major precursor in the formation of AGEs
   and therefore is strongly implicated in accelerated pathophysiological
   aging mechanisms linked to metabolic instability ([387]89). MGO is a
   cytotoxic, atherogenic ([388]171), and neurotoxic ([389]172) glycolytic
   intermediate that is elevated in diabetic patients ([390]173, [391]174)
   as well as those with cognitive decline ([392]175). Under physiological
   circumstances, MGO is detoxified by the glyoxalase system into
   D-lactate, with Glo1 as the key enzyme in the anti-glycation defense.
   Therefore, it is likely that congenital models of metabolic disruption,
   associated with a premature aging such as GIT2KO mice, may indeed help
   illuminate many of the endogenous protective mechanisms at the same
   time as demonstrating the pathophysiological effects of the genetic
   disruption.

   Our observations that GIT2 expression is modulated by stress ([393]18,
   [394]19), aging ([395]13) and metabolic pathologies (db/db) underscores
   the contextuality of what we consider to be “aging.” Therefore, rather
   than the process of aging being dictated merely by simple chronology,
   the effective physiological rate of aging (linked to damage
   accumulation) is considerably plastic and is contingent on the
   interaction and coordination of, multiple pathophysiological signaling
   pathways simultaneously across multiple tissues of the body. Acting as
   a protein with an ability to integrate numerous physiological processes
   intrinsic to the control of healthy aging, e.g., GPCR signaling systems
   ([396]176), insulin/IGF-1 signaling systems ([397]177), neurometabolic
   transport processes ([398]178, [399]179), DNA damage repair ([400]39),
   cytoskeletal organization ([401]180), and energy metabolism ([402]181)
   GIT2 may represent an important target for both further diagnostic and
   therapeutic research for age-related disorders. An important component
   of this future research however should be dedicated to the appreciation
   of how these differential cellular requirements may interact with each
   other via their co-dependence on this interesting keystone protein
   (Figure [403]11). Hence a more nuanced and contextually sensitive
   understanding of how multiple complex signaling systems are integrated
   and regulated will profoundly enhance our appreciation, and hopefully
   successful treatment, of age-related disorders.

Figure 11.

   [404]Figure 11
   [405]Open in a new tab

   Multifactorial functionality of GIT2 in the neurometabolic axis. From
   our data presented here and from previous research it is becoming clear
   that GIT2 likely acts as a keystone or hub protein for many
   physiological or cellular signaling processes involved in the
   maintenance of homeostatic metabolic function across life span. To
   illustrate the multidimensional functionality of GIT2 in such a complex
   system we have selected proteins/transcripts identified in this
   manuscript to highlight the network-level functional signature of GIT2
   activity. The ability of GIT2 to simultaneously control and integrate
   proteins involved in GPCR signaling/trafficking (Adora2b, adenosine A2b
   receptor; Kdelr2, Lys–Asp–Glu–Leu ER lumen protein retaining receptor
   2; Rgs5, regulator of G-protein signaling 5; Ramp1, receptor
   activity-modulating protein 1), insulin receptor (InsR) associated
   signaling (Irs2, insulin receptor substrate 2; Reg3b, regenerating
   islet-derived 3 beta; Insig 2, Insulin induced gene 2; Sesn1, sestrin
   1), metabolite/vesicular transport (Glut2, glucose transporter 2;
   Rims3, regulating synaptic membrane exocytosis 3; Slc2a13, solute
   carrier family 2 (facilitated glucose transporter), member 13; Gltp,
   glycolipid transport protein), DNA damage repair (DDR) (Atm,
   ataxia-telangiectasia mutated; Wrn, Werner syndrome ATP-dependent
   helicase; Gadd45a, growth arrest and DNA-damage-inducible protein
   GADD45 alpha; Parp2, Poly (ADP-ribose) polymerase 2), cytoskeletal
   remodeling (Vim, vimentin; Flna, filamin A; Spnb3, spectrin beta-3;
   Als2, Alsin) and energy metabolism (Pebp1, phosphatidylethanolamine
   binding protein 1; Glo1, glyoxylase 1; Mfn1, mitofusin 1; Gde1,
   glycerophosphodiester phosphodiesterase 1) clearly demonstrates
   keystone functionality of GIT2 in neurometabolic aging. Therefore the
   modulation of the expression of this protein, its post-translational
   and subcellular distribution are likely to strongly influence the
   orchestration of these vital functions in both health and
   pathophysiology.

Author Contributions

   BM and SM designed the work. WC, RS, SS, SSP, W-NC, CMD, BM, and SM
   performed experiments. BM, SS, SSP, W-NC, BM, SM, RP, JJ analyzed and
   interpreted the data. BM, WC, SM, and JJ wrote the manuscript.

Conflict of Interest Statement

   The authors declare that the research was conducted in the absence of
   any commercial or financial relationships that could be construed as a
   potential conflict of interest.

Acknowledgments