Altered learning, memory, and social behavior in type 1 taste receptor subunit 3 knock-out mice are associated with neuronal dysfunction

The type 1 taste receptor member 3 (T1R3) is a G protein-coupled receptor involved in sweet-taste perception. Besides the tongue, the T1R3 receptor is highly expressed in brain areas implicated in cognition, including the hippocampus and cortex. As cognitive decline is often preceded by significant metabolic or endocrinological dysfunctions regulated by the sweet-taste perception system, we hypothesized that a disruption of the sweet-taste perception in the brain could have a key role in the development of cognitive dysfunction. To assess the importance of the sweet-taste receptors in the brain, we conducted transcriptomic and proteomic analyses of cortical and hippocampal tissues isolated from T1R3 knock-out (T1R3KO) mice. The effect of an impaired sweet-taste perception system on cognition functions were examined by analyzing synaptic integrity and performing animal behavior on T1R3KO mice. Although T1R3KO mice did not present a metabolically disrupted phenotype, bioinformatic interpretation of the high-dimensionality data indicated a strong neurodegenerative signature associated with significant alterations in pathways involved in neuritogenesis, dendritic growth, and synaptogenesis. Furthermore, a significantly reduced dendritic spine density was observed in T1R3KO mice together with alterations in learning and memory functions as well as sociability deficits. Taken together our data suggest that the sweet-taste receptor system plays an important neurotrophic role in the extralingual central nervous tissue that underpins synaptic function, memory acquisition, and social behavior.

The type 1 taste receptor member 3 (T1R3) is a G proteincoupled receptor involved in sweet-taste perception. Besides the tongue, the T1R3 receptor is highly expressed in brain areas implicated in cognition, including the hippocampus and cortex. As cognitive decline is often preceded by significant metabolic or endocrinological dysfunctions regulated by the sweet-taste perception system, we hypothesized that a disruption of the sweet-taste perception in the brain could have a key role in the development of cognitive dysfunction. To assess the importance of the sweet-taste receptors in the brain, we conducted transcriptomic and proteomic analyses of cortical and hippocampal tissues isolated from T1R3 knock-out (T1R3KO) mice. The effect of an impaired sweet-taste perception system on cognition functions were examined by analyzing synaptic integrity and performing animal behavior on T1R3KO mice. Although T1R3KO mice did not present a metabolically disrupted phenotype, bioinformatic interpretation of the high-dimensionality data indicated a strong neurodegenerative signature associated with significant alterations in pathways involved in neuritogenesis, dendritic growth, and synaptogenesis. Furthermore, a significantly reduced dendritic spine density was observed in T1R3KO mice together with alterations in learning and memory functions as well as sociability deficits. Taken together our data suggest that the sweet-taste receptor system plays an important neurotrophic role in the extralingual central nervous tissue that underpins synaptic function, memory acquisition, and social behavior.
The gustatory sweet-taste receptor perception system, although crucial for nutrient sensing along the digestive system, appears more closely interconnected to complex endocrine and neurological functions than thought previously (1)(2)(3)(4)(5)(6). The sweet-taste perception process in the tongue, i.e. sugar/nonnutritive artificial sweetener gustation, begins with the activation of a heterodimeric G protein-coupled receptor complex, composed of the type 1 taste receptor member 2 (T1R2) 3 and type 1 taste receptor member 3 (T1R3). The T1R2/T1R3 heterodimer facilitates the sensation of a variety of ligands including natural and artificial sweeteners, amino acids, calcium, and sweet-taste proteins (7)(8)(9)(10)(11)(12)(13). The T1R3 subunit is widely expressed throughout the brain, with high abundance in the hypothalamus, hippocampus, and cortex (2,14,15). Interestingly, the T1R3 receptor possesses a strong structural similarity to the metabotropic glutamate receptors that are crucial for learning, memory, and behavior (16 -18). In addition, T1R3 knock-out (T1R3KO) mice demonstrate an inability to respond to monosodium glutamate (8) further supporting the potential importance of this receptor system in regulating neuronal/behavioral activity linked to glutamatergic transmission. Reinforcing a potential synergy between sweet-taste perception and cognitive systems, glucose availability has been strongly associated with effective learning capacity (19 -23). Neurometabolic dysfunction is a potent risk factor for multiple neurodegenerative brain diseases (24 -29), and alterations in chemosensory (taste/olfaction) perception are common in neurodegenerative diseases and aging (3, 30 -36). Specifically, neurodegenerative disorders with strong behavioral phenotypes, such as frontotemporal dementia (FTD) are characterized by alterations in gustation, eating behavior (37)(38)(39)(40), glucose metabolism (41,42), and energetic glucose utilization (43). In our present study, we found that T1R3KO mice demonstrated significant changes in neurosynaptic signaling proteins and structural architecture, altered learning and memory function, and disrupted sociability behavior. These findings demonstrate that an intact T1R2/ T1R3 receptor system in the brain is required for effective hippocampal neurotransmissive activity and neuronal architecture.

T1R3 genomic deletion affects taste perception but minimally affects somatic physiology
As part of the core unit of sweet-taste perception (i.e. T1R2/ T1R3 heterodimer) genomic deletion of the T1R3 receptor is likely to affect this modality. T1R3KO mice demonstrated a significantly blunted ability to respond in an appetitive manner to nutritive (sucrose) or non-nutritive (sucralose) sweet-tasting compounds (Fig. 1A). Despite alterations in taste perception we found no significant changes in general feeding behavior or body-weight gain between T1R3KO mice and WT controls. As the expression of T1R2/T1R3 heterodimers is more widespread than thought previously, we investigated the effect of genetic T1R3 disruption upon whole somatic physiology in mice (Fig. 1, B-E). Genomic deletion of the T1R3 receptor did not significantly affect oxygen consumption (VO 2 ), carbon dioxide production (VCO 2 ), respiratory exchange rate (RER), or heat generation compared with wild-type (WT) controls (Fig. 1, A-D). Furthermore, we failed to see any significant differences in ambulatory/vertical activity in the T1R3KO mice (supplemental Fig. S1, A-C). We did however find a consistent, but nonsignificant increase in food (Fig. S1D) and water (Fig. S1E) intake in T1R3KO mice compared with WT controls.

Transcriptomic analysis of central-nervous-system perturbations in T1R3KO mice
From our somatic physiological analyses of the T1R3KO mice ( Fig. 1), it was evident that these mice do not present with any significant metabolic issues aside from the altered tasteperception modalities at this experimental age (4 -5-months old). Our previous research has demonstrated that systemswide taste-receptor functionality involves receptor activity not only in the tongue but also in peripheral tissues, e.g. pancreas (44), as well as central-nervous-system (CNS) tissues such as the hypothalamus (2). Considering the potential role of the sweet-taste receptor system in neurodegeneration, we next investigated, using differential (between WT and T1R3KO mice) quantitative transcriptomics, the impact of T1R3 deletion upon the integrity of diverse CNS tissues including the cortex (supplemental Table S1), the hippocampus (supplemental Table S2), and the hypothalamus (supplemental Table S3). We had demonstrated previously the applicability of high-dimensionality transcriptomic profiling to generate a holistic appreciation of complex CNS-related functions such as neurodegeneration (45), remedial drug efficacy (46), psychosocial disruption (47), and pathological aging (48). Investigation of the significant differential transcriptomic expression patterns between T1R3KO and WT mice ( Fig. 2A and supplemental Table S4) demonstrates the presence of multiple transcripts differentially regulated in multiple tissues from T1R3KO mice compared with WT controls, of which the cortex and hippocampus demonstrated the greatest degree of cross-over of T1R3KO-specific altered factors. To validate our transcriptomic data (supplemental Tables S1-S3), we chose five tran- The ability to perceive the appetitive stimuli, sucrose, and the non-nutritive sweetener, sucralose, was assessed in WT and T1R3KO mice (A). T1R3KO mice demonstrated a largely refractory response to the appetitive tastants. Whole-body metabolic status was assessed in T1R3KO mice using a CLAMS (Comprehensive Lab Animal Monitoring System) platform across a variety of outputs (WT, black triangles; T1R3KO, white squares) including oxygen consumption (VO 2 ) (B), carbon dioxide production (VCO 2 ) (C), Respiratory Exchange Ratio (RER-an assessment of the metabolic exchange of oxygen for carbon dioxide) (D), and heat production (E). Parameters (B-E) were measured over a 48-h period. The dark cycle time is indicated by the gray section of the pictogram bar in each CLAMS activity. Data are shown as means Ϯ S.E., and p Յ 0.05 was considered statistically significant (n ϭ 8).
scripts from our microarray data, as well as the T1R3 itself, confirming their differential expression in T1R3KO mice compared with WT controls by using quantitative real-time PCR (Fig. 2B). We confirmed our array data (supplemental Tables S1-S3) indicating a T1R3KO-dependent decrease in Actl6b (actin-like 6B), Ccl21b (chemokine (C-C motif) ligand 21B (leucine)), and Taf6 (TAF6 RNA polymerase II, TATA-boxbinding protein (TBP)-associated factor), along with a T1R3KOdependent potentiation of H2-Bl (histocompatibility 2, blastocyst) and Tmem87a (transmembrane protein 87A) transcript expression (Fig. 2B). Comparing transcriptomic differences between T1R3KO and WT mice, it is clear that considerable significant differences are present, e.g. 691 cortical transcripts, 584 hippocampal transcripts, and 441 hypothalamic transcripts. We employed signaling pathway analysis to elucidate the signaling activities that were potentially perturbed in T1R3KO mice. Using ingenuity pathway analysis (IPA), we investigated how the distinct transcriptomic states of the T1R3KO mouse cortex, hippocampus, or hypothalamus translated into potential canonical signaling or biofunction/disease pathways (supplemental Tables S5 (cortex), S6 (hippocampus), and S7 (hypothalamus) and Fig. 2C). With respect to the IPA canonical signaling pathway analysis, it was evident that the hypothalamus was effectively dominated by "up-regulated" pathways (comprising a predominance of up-regulated related transcripts). The cortex and the hippocampus demonstrated only a small number of up-regulated pathways. Conversely, both the cortex and the hippocampal tissues were dominated by a large number of "down-regulated" pathways (comprising a predominance of down-regulated related transcripts), whereas the hypothalamus possessed the inverse, i.e. only a small number of down-regulated pathways. As with our canonical signaling pathway analysis, we found a stronger commonality of predicted T1R3KO-related IPA biofunctions between cortical and hippocampal tissues compared with hypothalamic tissues (Fig.  2D). To further investigate the divergence between the hippocampal/cortical tissues and the hypothalamus, we generated a heat map of both the canonical signaling pathways and the biofunction pathways correlating to neurophysiological parameters linked to cognitive function and behavior (Fig. 2E and  supplemental Tables S8 (cortex), S9 (hippocampus), and S10 (hypothalamus)). For many of the important neurophysiological functionalities indicated in Fig. 2E, the associated Z-scores again demonstrated a tighter functional grouping between the cortex and hippocampus compared with the hypothalamus. Pathway annotation of the cortical and hippocampal data demonstrated multiple pro-neurodegenerative activities, e.g. reductions in cellular homeostasis and neurotrophin receptor signaling. The hippocampal data set demonstrated some of the strongest neurodegenerative activity (IPA signaling pathway and biofuntions) predictions, in terms both of up-and downregulation, among the three tissues. Hence, hippocampal tissues demonstrated the lowest Z-scores for cellular homeostasis, organismal death, organization of cytoskeleton, FGF signaling, HMGB1 signaling, and neurotrophin/TRK signaling and the highest Z-score for proliferation of neuronal cells. In light of these predictive functional signaling divergences and differential transcriptomic data, we next chose to investigate the actual proteomic differences between T1R3KO and the WT controls in the most closely, functionally clustered tissues that A, Venn diagram analysis of significantly regulated transcripts in the cortex, hippocampus, and hypothalamus of T1R3KO mice compared to WT controls. B, PCR validation of randomly selected transcriptomic targets (T1R3, Actl6b, Ccl21b, Taf6, B2-Bl, and Tmem87a: WT, black bars; T1R3KO, white bars). C, Venn diagram analysis of IPA canonical signaling pathway annotation generated using the significantly-regulated cortical, hippocampal, and hypothalamic transcriptomic data as the input. Numerical values for each pathway were generated through calculation (for a specific pathway) of the percent of transcripts up-regulated in the specific pathway minus the percent of the transcripts downregulated in the specific pathway. D, Venn diagram analysis of IPA biofunction/disease annotation data generated using the significantly regulated transcript data from each tissue as the input. Biofunction/disease activating or inhibiting Z-score estimates generated by IPA were employed for this numerical annotation. E, heat map illustration of clustered neurodegeneration-related biofunction/disease (upper panel) and canonical signaling pathways (lower panel: IPA-derived canonical signaling pathway Z-score estimates depicted) from hippocampal (HIP), cortical (CTX) and hypothalamic (HYP) tissues.
are strongly linked to neurodegenerative processes, i.e. the hippocampus and cortex.

Proteomic expression profile of T1R3KO hippocampal and cortical tissues
Using iTRAQ-based proteomics we assessed the high-dimensionality protein expression profiles of T1R3KO mice compared with WT controls. The top 10 significantly upand down-regulated proteins in the hippocampus and the cortex are organized in Tables 1 and 2 respectively. The following general trend is apparent, i.e. neuroprotection systems are down-regulated, whereas the cell growth and cell division systems are up-regulated. We identified 21 proteins significantly up-regulated and 28 proteins significantly down-regulated in the T1R3KO hippocampus compared with the WT controls ( Fig. 3A and supplemental Table S11, p Ͻ 0.05).
To validate our iTRAQ data, we selected three proteins and verified their predicted expression profile using Western blotting (Fig. 3B). For all selected proteins (Cplx1, complexin 1; Hint1, histidine-triad nucleotide-binding protein 1; and Slc8a1, solute carrier family 8 (sodium/calcium exchanger), member 1) our Western blot analysis confirmed the mass spectrometry For each significantly identified protein the official Gene Symbol, protein description, Log 2 iTRAQ expression ratio (T1R3KO vs. WT control), and molecular functionality are given. Proteins denoted in red were significantly elevated in T1R3KO vs. WT and those in green significantly downregulated in T1R3KO vs. WT controls.

Table 2 Top 10 most upregulated or downregulated proteins in T1R3KO murine cortex versus WT control cortex
For each significantly identified protein the official Gene Symbol, protein description, Log2 iTRAQ expression ratio (T1R3KO vs. WT control), and molecular functionality are given. Proteins denoted in green were significantly elevated in T1R3KO vs. WT and those in green significantly downregulated in T1R3KO vs. WT controls.

T1R3 KO mice have altered cognitive function
data (supplemental Table S11). Compared with the transcriptomics, which indicates functional intent, the verification of actual protein-expression profiles likely represents a more proximal mechanism by which to elucidate the differences in cell-signaling activity between T1R3KO and WT mice. Using our proteomics data along with a multidimensional informatics approach, we next investigated the distinct signaling patterns intrinsic to T1R3KO mice. Applying IPA canonical signaling pathway analysis (with an IPA hippocampal-specific background-reference data set) to the significant hippocampal proteomic data, we found that multiple pathways linked to cell cycle control (CDK5 signaling, telomerase signaling), develop- from hippocampal iTRAQ data are displayed; proteins with significantly differential expression (p Յ 0.05) in T1R3KO versus WT control are above or below the red lines (n ϭ 3 mice/genotype). B, Western blot validation of randomly selected proteins differentially expressed in the T1R3KO hippocampus. The iTRAQgenerated expression profiles of Cplx1, Hint1, and Slc8a1 were all confirmed by Western blotting of T1R3KO and WT mice. Additionally actin-normalized protein expression levels are shown. Error bars represent mean Ϯ S.E.; p Յ 0.05 was considered statistically significant. C-E, pie charts depicting the distribution of the canonical signaling pathway groups identified in the T1R3KO data set using IPA according to specific biological functions, i.e. cell cycle (C), developmental and protective signaling (D), and receptor signaling (E). F, pie chart identifying the multiple biofunction/disease pathway annotations of the significant proteomic data. G, using IPA-based protein signaling network generation, the most coherent interactions responsible for the associated T1R3KO biofunctions/ diseases (represented in F) are highlighted. mental/neuroprotective activity (Wnt/␤-catenin signaling, synaptic long-term depression), and receptor signaling (dopamine-receptor signaling and ␤-adrenergic signaling) were significantly enriched (Fig. 3, C-E, and supplemental Table S12). We also interrogated this proteomic data set for potential IPA biofunction/disease activity (Fig. 3, F and G, and supplemental Table S13). We found that there was a strong data-set enrichment for proteins associated with neuronal architecture and remodeling (neuritogenesis, dendritic growth/branching, quantity of neurons) (Fig. 3, F and G, and supplemental Table  S13). Using IPA-based network analysis between these significantly enriched IPA biofunction pathways, it was clear that these activities were likely related to one another and coordinated via multiple interacting proteins identified using iTRAQ (Ctnnd2, catenin ␦2; Nlgn3, neuroligin 3; Ppp2ca, serine/threonine-protein phosphatase 2A catalytic subunit ␣; Rap2a, Rasrelated protein Rap-2a (Fig. 3G)).
As with the T1R3KO hippocampus, we were able to identify multiple significantly regulated (p Ͻ 0.05) proteins (45 up-regulated and 10 down-regulated, Fig. 4A and supplemental Table S14) in the T1R3KO cortex compared with WT controls. We chose three exemplary proteins (Rps11, ribosomal protein s11; Hint1, histidine triad nucleotide-binding protein 1; and Itsn1, intersectin 1) that were significantly and differentially regulated between T1R3KO and WT mice to verify our proteomic data set (Fig. 4B). We also examined, at the signaling pathway level (Fig. 4, C-G, and supplemental Table S15) and the biofunction level (Fig. 4, H and I, and supplemental Table S16), the physiological nature of the cortex T1R3KO proteomic data. Using canonical signaling pathway analysis, a strong neurosynaptic functional signature was generated by the cortex T1R3KO data set, i.e. multiple signaling pathways clustered around functions such as cell-cycle control (EIF2 signaling and Cdk5 signaling), developmental signaling (mTOR signaling and ceramide signaling), receptor signaling (ephrin-receptor signaling and HMGB1 signaling), cytoskeletal regulation (RhoA signaling and RhoGDI signaling), and vesicular endocytosis (caveolar-mediated endocytosis signaling and macropinocytosis signaling).

Natural language-based informatics investigation of T1R3KOspecific proteomic disruption
To obtain a more data-specific appreciation of our proteomic data sets, we employed a biomedically focused natural language processing (NLP) informatics interpretation approach (49) using the Textrous! program (50). Using the individual pro-cessing mode of Textrous! to interrogate the significantly regulated hippocampal proteomic data-set, we extracted and correlated to the input data-set proteins multiple biomedically relevant words that clustered into several functional categories (Fig. 5, A (neuronal), B (synapse), C (vesicular trafficking), D (neurotransmission), E (Disease-related), and F (cytoskeleton)). Based on these functional Textrous! clusters, we generated a cumulative expression score for the proteins in these groups to indicate a general trend for the expression pattern (i.e. T1R3KO versus WT) of proteins associated with these semantically associated terms (Fig.  5G). Using this methodology we found that the "neuronal," "synapse," "neurotransmission," and "disease-related" clusters presented a profound trend for down-regulation (calculated using both the mean or sum of the log 2 expression ratio scores for the proteins linked to these functions (supplemental Table S17), whereas only "vesicular trafficking" and "cytoskeleton" were populated by generally up-regulated proteins ( Fig. 5G and supplemental Table S17).
Performing analogous Textrous!-based analysis of the cortical proteomic data set for T1R3KO mice, we found that the significantly regulated proteins clustered into two main groups, cytoskeleton and monomeric G proteins (Fig. 6, A and B). For both of these clusters all of the proteins linked to these Textrous!-extracted functions were largely overexpressed in T1R3KO mice compared with WT-control cortices ( Fig. 6C and supplemental Table S18). Compared with standard pathway annotation processes that use curated language-based interpretation (e.g. KEGG pathways) the Textrous! program is able to extract more experiment-specific data-descriptive noun phrases linked to the input proteins (50).
Therefore to reinforce our observed distinctions between NLP-based analyses of hippocampal and cortical data sets, we also used these semantically associated noun phrases linked to the hippocampal (supplemental Table S19) or cortical (supplemental Table S20) proteomic data sets as the input to generate complex and highly nuanced word clouds that represent a more gestalt-type interpretation of the proteomic data (Figs. 7A (hippocampus) and 8A (cortex)). The use of free-form quantitative word clouds to convey complex noncanonical signaling-activity relationships is becoming more and more recognized as a novel technique to investigate high-dimensionality data (51)(52)(53). In contrast to highlighting just a minimal subset of predicted output functions, biomedical word clouds create a more holistic representation of the full spectrum of signaling ramifications borne out of the data set(s). In accordance with our transcriptomic analyses (Fig. 2), we found that this unbiased NLP-based data-set interpretation was highly concordant, such that the terms synaptic, postsynaptic, endocytic, axonal, cytoskeleton, and PDZ ( Fig. 7A) were prominent. To analyze this highly nuanced output impartially, we used the textual analyzer Write-Words (http://writewords.co.uk/). 4 Upon textual frequency analysis we found several words that were disproportionally over-represented (Ͼ99% confidence limit of significance using individual word frequency scores (supplemental Table S21)) in this word- Figure 4. Differential proteomic expression patterns in T1R3KO cortex versus WT controls iTRAQ data. A, log 2 -transformed ratios (T1R3KO:WT) from cortical iTRAQ data are displayed; proteins with significantly differential expression (p Յ 0.05) in T1R3KO versus WT control are above or below the red lines. B, Western blot validation of randomly selected proteins differentially expressed in the T1R3KO hippocampus. The iTRAQ-generated expression profiles of Rps11, Hint1, and Itsn1 were all confirmed by Western blotting of T1R3KO and WT mice. Additionally actin-normalized protein expression levels are shown. Error bars represent mean Ϯ S.E.; p Յ 0,05 was considered statistically significant. C-G, pie charts depicting the distribution of the canonical signaling pathway groups identified in the T1R3KO data set using IPA according to specific biological functions, i.e. cell cycle (C), developmental and protective signaling (D), receptor signaling (E), cytoskeletal regulation (F), and vesicular endocytosis (G). H, pie chart identifying the multiple biofunction/disease pathway annotations of the significant proteomic data. I, using IPA-based protein-signaling network generation, the most coherent interactions responsible for the associated T1R3KO biofunctions/diseases (represented in H) are highlighted.
In contrast to the hippocampal data analysis, the word cloud interpretation of the cortical proteomic data suggested a less synaptically focused nature of the data, e.g. the most prominent . Textrous! natural language processing-based interpretation of hippocampal proteomic data. The potential functional relationships between significantly and differentially regulated proteins in T1R3KO hippocampi compared with WT controls were interpreted using the individual processing mode of Textrous! Proteins (up-or down-regulated in T1R3KO versus WT, indicated in red or green, respectively) and their correlated biomedical terms are represented as a heat-map matrix, with the strength of association indicated by the intensity of the teal blocks and "no correlation" indicated by a gray block. Specific functional clusters of protein activity were found using the hippocampal proteomic data set, i.e. neuronal (A), synapse (B), vesicular trafficking (C), neurotransmission (D), disease-related (E), and cytoskeleton (F) are depicted. G, mean log 2 iTRAQ expression scores (box and whiskers plot indicates 5-95% confidence limits; the mean score is indicated by either a red or green line depending on whether the mean possessed a positive or negative sign) for proteins subclassified into the specific functional Textrous! groups (A-F). Above each box the red or green number indicates the numerical sum of the log 2 -transformed iTRAQ expression scores for the proteins identified in that functional Textrous! group. Figure 6. Textrous! natural language processing-based interpretation of cortical proteomic data. The potential functional relationships between proteins significantly and differentially regulated in the T1R3KO cortex compared with WT controls were interpreted using the individual processing mode of Textrous!, proteins (up-or down-regulated in T1R3KO versus WT indicated as red or green, respectively), and their correlated biomedical terms are represented as a heat-map matrix, with the strength of association indicated by the intensity of the teal blocks and "no correlation" indicated by a gray block. Specific functional clusters of protein activity were found using the hippocampal proteomic data set, i.e. cytoskeleton (A) and monomeric G proteins (B). C, mean log 2 iTRAQ expression scores (box and whiskers plot indicates 5-95% confidence limits; the mean score is indicated by either a red or green line depending on whether the mean possessed a positive or negative sign) for proteins subclassified into the specific functional Textrous! groups (A and B). Above each box the red or green number indicates the numerical sum of the log 2 -transformed iTRAQ expression scores for the proteins identified in that functional Textrous! group.

T1R3 KO mice have altered cognitive function
words were linked primarily with cytoskeleton-modulating behavior (cytoskeletal, dynamics, actin, motility, and polymerization) (Fig. 8A). Using WriteWords-based quantification of the cortical word cloud data, we were able to demonstrate that the significant textual output from these data was linked predominantly to GTPase-associated cytoskeletal regulatory behavior (cytoskeletal, GTPase, exchange, and actin) as opposed to the focused neurosynaptic activity apparent from hippocampal data-set interpretation ( Fig. 8B and supplemental Table S22).

Altered architectural structure of the T1R3KO hippocampus
As our high-dimensionality data investigation of T1R3KO mice suggested a specific hippocampus-focused neurosynaptic phenotype, we next investigated the effects of genomic T1R3 deletion upon hippocampal neuronal architecture using Golgi staining. Upon inspection of Golgi-stained hippocampal CA1 regions from WT and T1R3KO mice, we noted a considerable increase in hippocampal neuronal density in the T1R3KO mice compared with the WT controls (Fig. 9, A and B). This distinction in T1R3KO mice appeared to be generated by an increased presence of large neuritic densities (Fig. 9, C-E (WT) and D and F (T1R3KO)). Concomitant with this increased neuritic density we also noted that the T1R3KO dendrites possessed significantly fewer spines across their length (Fig. 9, G and H). Indicative of the increased neuritic density in the T1R3KO hippocampus, we also found that the average dendrite length in the T1R3KO mice was greater (non-significantly) than in the WT controls (Fig. 9I). The evident neuronal somatic "expansion" (Fig. 9, D and F) is potentially associated with the observed dendritic spine density and alteration of primary dendrite morphology (i.e. apparent dendritic collapse). Taken together with Figure 7. Quantitative high-dimensionality hippocampal proteomic data word cloud. A, correlated full noun phrases were extracted from the Textrous!based semantically associated word lists created using the input significantly altered T1R3KO hippocampal proteomes. The font size in the word cloud is directly related to the word frequency present in the extracted noun-phrase lists. B, applying word frequency analysis to this semantically associated list, nine specific words significantly (p Յ 0.01) overrepresented in the list (box and whiskers plot indicates 1-99% confidence limits) were identified (left panel). The specific word frequency count for these words is represented to indicate their ranking.

T1R3 KO mice have altered cognitive function
the increased neuritic density, reduced spine density, and increased dendrite length, these results suggest a potential neurological dysfunction/synaptic failure in the T1R3KO mice. Reinforcing this conclusion, and perhaps in reactive response to neurosynaptic failure, increases in hippocampal NR1 (NMDA receptor subunit 1 (Fig. 9, J and K)) and NR2B (NMDA receptor subunit 2B (Fig. 9, L and M)) expression were observed in T1R3KO mice compared with WT controls. These findings were further supported via Western blotting, which demonstrated significant increases in both the NR1 and NR2B subunits in T1R3KO versus WT controls (Fig. 9N).
Based on our identification of altered neurosynaptic structure (Golgi staining, Fig. 9, A-I) and molecular signaling activity (NR1 and NR2B expression, Fig. 9, J-N) in T1R3KO mice, we next investigated whether individual neuronal excitability in the T1R3KO mice was also affected. Primary neurons from both WT and T1R3KO neonatal mice were isolated and cultured for 7 days to allow mature dendrite and axon growth. The fluorescence intensity of cell membrane-permeable Ca 2ϩbinding fluorophore Fluo-8 AM was used as an indicator of intracellular calcium concentration. Using simultaneous microscopic cellular analysis, in response to the addition of 100 M glutamic acid (Fig. 9O), Fluo-8 AM intensity rose quickly after excitatory ligand addition in both WT and T1R3KO neurons with 2-3.5-fold amplitude elevation in fluorophore levels. With respect to intracellular Fluo-8 AM intensity, WT neurons recovered gradually, and the cytoplasmic calcium [Ca 2ϩ ] c returned to basal levels after 2-3 min (Fig. 9P), whereas the isolated primary T1R3KO neurons remained excited with sustained and elevated [Ca 2ϩ ] c for more than 280 s (Fig. 9Q). Measuring the mean Fluo-8 AM intensity across the time period of the glutamate stimulation, we found a significant difference in the kinetic profile of stimulated calcium release/management between the T1R3KO and WT neurons (Fig. 9R).

Altered neurosynaptic protein expression levels in T1R3KO hippocampus
We next investigated the expression level of proteins specifically involved in neurosynaptic activity in T1R3KO mice. In line with our findings of potentially disrupted neurotransmission in T1R3KO mice (compared with WT controls), we found a significant reduction in phosphorylated cAMP responseelement-binding protein (p-CREB), phosphorylated synapsin (p-synapsin or p-Syn1), phosphorylated extracellular-signalregulated kinase 1/2 (p-Erk1/2), spinophilin (Ppp1r9b), PSD-95 (post-synaptic density protein, 95 kDa), and phosphorylated Figure 8. Quantitative high-dimensionality cortical proteomic data word cloud. A, correlated full noun phrases were extracted from the Textrous!-based semantically associated word lists created using the input significantly altered T1R3KO cortical proteomes. The font size in word cloud is directly related to the word frequency present in the extracted noun-phrase lists. B, applying word frequency analysis to this semantically associated list, six specific words significantly (p Յ 0.01) overrepresented in the list (box and whiskers plot indicates 1-99% confidence limits) were identified (left panel). The specific word frequency count for these words is represented to indicate their ranking.

T1R3 KO mice have altered cognitive function
TrkB (p-TrkB) (Fig. 10). The reductions in the levels of these factors all pointed to an effective neurotransmission deficit in the T1R3KO mice compared with WT controls. We also further investigated the same protein expression profile in cortical tissues in both T1R3KO and WT mice (Fig. 10). Reinforcing the specific hippocampal nature of these deficits in T1R3KO mice, although similar cortical expression trends were apparent (e.g. p-synapsin 1 and spinophilin), no significant alterations in this same panel were found in T1R3KO cortex samples.

Deficits of learning and memory function in T1R3KO mice
Our multidimensional interpretation of both transcriptomic and proteomic data from the T1R3KO mice generated a focused assessment of potential physical and functional neurosynaptic dysfunction in these mice. We next investigated whether these phenotypic molecular "signatures" would also be predictive of actual disrupted cognitive/behavioral activity in these mice. We employed both novel object preference (NOP) and Morris water maze (MWM) tests to assess learning and memory functions in T1R3KO mice.
Using the NOP test we found that, unlike with WT mice, T1R3KO mice did not demonstrate a significant preference for novel objects (Fig. 11A), resulting in a significantly lower discrimination index in the T1R3KO mice for novel objects compared with WT control mice (Fig. 11B). Using the MWM test to assess spatial learning and memory performance we found that, compared with WT control mice, T1R3KO mice demonstrated a significantly longer escape latency time at days 3-4 of MWM training (Fig. 11C). Additionally, we found that T1R3KO mice demonstrated a significantly increased swim speed (Fig. 11D) and swimming path length (Fig. 11E) in the MWM test on days 3-4 compared with WT controls. With the transition from a hidden platform to probe trials in the MWM test, we found that T1R3KO mice spent signifi- cantly less time in the MWM quadrant that housed the platform previously (Fig. 11F).

Altered social and behavioral activity in T1R3KO mice
With the evident association with appetitive taste perception and behavioral modalities (36), we next assessed whether genomic T1R3 deletion affected social and behavioral activity. Using the classical three-chambered sociability test we found that T1R3KO mice, unlike the WT controls, failed to demonstrate a significant preference for interaction with an animate target mouse compared with an inanimate target (Fig. 11G). This resulted in the generation of a significantly lower discrimination index for the T1R3KO mice compared with WT controls for social interaction (Fig. 11H).
Further behavioral distinctions were observed between T1R3KO and WT mice using the open field test. T1R3KO mice demonstrated significantly lower ambulatory distance (Fig. 11I) and significantly greater physical inactivity (Fig. 11J), as well as a significantly greater propensity to spend time in the periphery of the field and less time in the field center (Fig. 11K). Related to this final finding, we also observed a significantly reduced propensity for T1R3KO mice, compared with WT controls, to remain in the light zone of the open field, coupled to a concomitant increase in their time spent in the open field dark zone (Fig. 11L). Assessing stress/anxiety using the elevated plus maze demonstrated only a subtle (non-significant) propensity for T1R3KO mice to spend more time in the closed arms rather than the open arms, compared with WT controls (Fig. 11M). Although they demonstrated multiple neuropsychological and behavioral deficits, the T1R3KO mice, compared with WT controls, displayed no motor coordination deficits using the Rota-Rod test (Fig. 11N).

Discussion
It is becoming clear that in addition to our deeper appreciation of the neuroendocrine complexity of taste perception, the sweet-taste receptor system is expressed in many organs other than the tongue (3, 24, 36, 54 -57). Here we have demonstrated that loss of the T1R3 subunit causes learning and memory impairments, CNS synaptic failure, attenuated neurotrophic activity, and proteomic molecular signatures and behavioral activities reminiscent of neurodegenerative phenotypes.
Despite research indicating a role of somatic metabolic hormones in taste perception, we found no significant metabolic disruption in T1R3KO mice (Fig. 1, B-E). These observations, however, do not necessarily preclude the capacity for these animals to generate a metabolic phenotype later in life. Our transcriptomic (Fig. 2), proteomic (Figs. 3-8), neurosynaptic (Figs. 9 and 10), cognitive, and psychosocial and behavioral investigations (Fig. 11) of T1R3KO mice however suggested a strong pro-neurodegenerative impact of T1R3 genomic deletion. CNS transcriptome profiling can generate an effective appreciation of complex functions such as neurodegeneration (45) and pathological aging (48). T1R3KO transcriptomic profiling demonstrated a greater coherency between cortical/hippocampal tissues compared with hypothalamic tissues (Fig. 2); this may underpin the relatively greater importance of CNS dysfunction (Figs. 2-4) versus somatic metabolic functions (Fig. 1) to the T1R3KO mouse phenotype. With specific reference to the strongest scoring hippocampal bio-function prediction, i.e. "proliferation of neuronal cells," the factors populating this group consisted of multiple transcripts linked to Alzheimer's disease (AD) (K-Ras (58), Wnt7b (59), Rasgrf1 (60), and Tardbp (61, 62)), calcium-dependent hippocampal neuronal excitability (K-Ras (63) (77)), neuroprotective activity (Cbr1 (78)), anxiety-related behavior (Cxcl12 (79) and Tsc2 (80)), and learning/memory functions (Xpa (81), RasGrf1 (82), and Ccnd1 (83)). Proteomic analyses of Figure 10. Genomic deletion of T1R3 results in significant alterations in multiple neurosynaptic proteins. A, selective Western blot analysis of standard marker proteins indicative of neurosynaptic functionality was performed in T1R3KO and WT mouse hippocampal extracts (n ϭ 4 mice/genotype). B, compared with WT tissue expression (black bars), T1R3KO hippocampal levels (white bars) of phosphorylated CREB (p-CREB), phosphorylated synapsin 1 (p-Syn1), spinophilin (Ppp1r9b), PSD-95, and p-TrkB were significantly reduced. No significant differences between WT and T1R3KO tissues in the expression levels of non-phosphorylated CREB, Syn1, TrkB, or Erk1/2 were noted. Levels of synaptophysin (Syp) and non-phosphorylated Erk1/2 were also unchanged between WT and T1R3KO tissues. In contrast to the effects of T1R3 deletion in the hippocampus, genomic deletion of T1R3 results in subtle non-significant alterations in multiple neurosynaptic proteins. C, selective Western blot analysis of standard marker proteins indicative of neurosynaptic functionality was performed in T1R3KO and WT mouse cortical extracts. D, compared with WT tissue expression (black bars), T1R3KO hippocampal levels (white bars) of phosphorylated synapsin 1 showed a strong (non-significant) trend toward reduced expression. For the other constituents of the neurosynaptic protein panel investigated previously (Fig. 9), no strong differences in expression were noted between WT or T1R3KO tissues. Error bars represent mean Ϯ S.E.; p Յ 0.05 was considered statistically significant. Figure 11. Impaired cognitive, social, and exploratory behavior in T1R3KO mice compared with WT controls (n ‫؍‬ 8 mice/genotype). A and B, recognition memory was assessed using NOP (WT, black bars; T1R3KO, white bars). A, time spent investigating familiar or novel objects. B, discrimination index in percentage of T1R3KO mice compared with WT controls. C-F, spatial learning was assessed using MWM tests to measure escape latency (C), swimming speed (D), path length (E), and percentage of time spent in (target) platform (F). T1R3KO mice required more time to find the escape platform (C), even though the swimming speed (D) and the path length (C) significantly increased during the last 2 days of training (E). In the probe task, WT mice spent 44.67% of the total time in the quadrant where the platform was placed compared with 25.64% of total time for T1R3KO mice (F). G, sociability was evaluated using the three-chamber social behavior apparatus. Compared with WT controls, T1R3KO mice had no preference in staying with another mouse and showed a lower hippocampal and cortical tissues further demonstrated significant alterations in multiple proteins associated with neurodegenerative conditions (Nrxn1 (84) and Pin1 (85), Prdx3 (86)), psychosocial disorders (Agk (47), Arhgap18 (87), and Itsn1 (88)), and pathological aging (Nlgn3 (89), Ipo5 (90), and Ogt (91)). Through classical pathway-based informatics, a profound alteration in neuronal architecture (Figs. [3][4][5][6][7][8] linked to dysfunctions in neuritogenesis, dendritic growth, and hippocampal neuronal morphology was evident in the T1R3KO mice, together with the concerted activity of multiple significantly regulated proteins. Textrous!-based natural language processing of our proteomic data demonstrated that reductions in neuronal, synaptic, and neurotransmissive functionalities were likely (Fig. 5G). Using Golgi staining we confirmed these informatic predictions, i.e. we found cortico-hippocampal neurodendritic clumping, poor spatial synaptic connectivity, and a significantly reduced dendritic spine density in T1R3KO mice (Fig. 9). Neurosynaptic architecture deterioration is strongly linked with cognitive dysfunction and memory loss in neurodegenerative diseases such as AD and FTD (92)(93)(94). Excitatory postsynaptic terminals are associated with dendritic spines (95)(96)(97). Calcium influx into these spine heads, through NMDA-type glutamate receptors, is well known to induce long-term potentiation, which underpins hippocampal learning and memory formation (98 -100). Furthermore, dendritic spines undergo activity-dependent remodeling to support synaptic plasticity and learning/memory function (101)(102)(103). Because alterations in synaptic marker expression and phosphorylation are indicative of abnormalities in synaptic vesicle dynamics, long-term potentiation, and impaired memory function (104,105), the synaptic markers that were found downregulated in the T1R3KO mice (i.e. synapsin 1, synaptophysin, spinophilin, and PSD-95) can be interpreted as critical indicators of cognition (106 -110). PSD-95 knock-out mice demonstrate spatial memory impairment due to the excessive excitation of synapses (111). Reduced PSD-95 expression levels are also correlated with disease severity for multiple neurodegenerative conditions such as AD (112), mild cognitive impairment (113), FTD (114), Huntington's disease (115), and pathological aging (116). Spine loss and subsequent spinophilin reductions are also evident in multiple transgenic murine models of neurodegeneration, e.g. AD (117)(118)(119) and advanced aging (120). Reductions in synaptic spinophilin are also strongly correlated with neurodegenerative disease pathology progression rates (121) as well as with the severity of learning and memory deficits (122).

T1R3 KO mice have altered cognitive function
In addition to dendritic spine disruption, our observed upregulation of post-synaptic NMDA glutamate receptors (NR1/ NR2B) is a potential indication of denervation supersensitization, i.e. reflexive potentiation of receptor expression (123), occurring due to a reduction of a coherently regulated stimulating input. Linked to this we also found altered responsivity to glutamate in T1R3KO mice, indicative of a disrupted calcium sequestration and storage functionality (Fig. 9), a process strongly linked with neurodegeneration (124,125) and pathological aging (126).
Synaptic plasticity requires both neurotrophic ligand support and the regulation of rapid protein translation. Hence an interplay exists between neurotrophins, such as brain-derived neurotrophic factor (127) and nerve growth factor and transcriptional/translational controllers such as CREB (128,129)), Erk1/2 (130,131), and upstream transcription factor 1 (Usf1 (132,133)). T1R3KO mice demonstrated significant reductions in hippocampal CREB phosphorylation as well as TrkB phosphotyrosine content (Fig. 10). A similar but non-significant trend for the reduction in these factors was also observed in T1R3KO cortical tissues (Fig. 10). CREB forms a convergence point of pathways and mechanisms activated during the processes of synaptic strengthening, plasticity, and memory formation, as CREB phosphorylation leads to the transcription of memory-associated genes (134). The reduced levels of CREB/ TrkB phosphorylation in the T1R3KO mice suggest an extant deficit in memory formation and/or brain-derived neurotrophic factor signaling in these animals.
When we investigated the potential behavioral effects of the multiple neurodegenerative indices identified in our study, we found that T1R3KO mice demonstrated impaired acquisition of novel objects as well as an impaired acquisition and learning of cognitive tasks (Fig. 11). T1R3KO mice also presented blunted sociability, reduced exploratory behavior, and moderately elevated anxiety responses (Fig. 11). Impaired sociability and restricted interests are hallmarks of autism spectrum disorder (ASD) and are strongly linked to disrupted neurosynaptic architecture. T1R3KO mice demonstrated behavioral (Fig. 11) and molecular transcriptomic (supplemental Tables S1-S3)/ proteomic (supplemental Tables S13 and S14), signatures associated with ASD-like phenotypes. T1R3KO mice demonstrated alterations in multiple factors linked to ASD etiology in the murine BTBR Tϩtf/J inbred mouse strain model, e.g. reductions in cortistatin along with elevations in Cadm3 and Wfs1 (47). Neuroligin 3 (Nlgn3), a post-synaptic cell-surface protein associated with ASD (135), was significantly reduced in the T1R3KO hippocampus. Neuroligins are trans-synaptic ligands for pre-synaptic neurexins and may be involved in the formation and remodeling of CNS synapses. Interestingly, we found a significant elevation of cortical Nrxn1 in the T1R3KO mice. Impaired regulation of neuroligin-neurexin connectivity at glutamatergic synapses appears to be crucial in ASD-related pathophysiology (136). In autistic patients, an increased neuron excitation/inhibition ratio has been found associated with the pathology of neural dysfunction (137), again reminiscent of our observed altered T1R3KO primary neuron responsivity to glutamate (Fig. 9).
Our NOP/MWM data suggested that neurosynaptic failure, a common feature of dementias such as AD (138) and FTD (139), likely contributes to the cognitive deficits observed in T1R3KO mice. Our T1R3KO hippocampal proteomic signatures identified a significant reduction in the vesicular trafficking protein ␥-taxilin (Txlng), which has recently been implicated in stress-induced AD-related tauopathy formation (140). Hotokezaka et al. (140) have demonstrated that Txlng is stressresponsive and vital for maintaining the integrity of the unfolded protein response system. Genetic ablation of Txlng expression in neuronal cells has thus been shown to induce AD-related Tau hyperphosphorylation. Tau pathologies in AD are also linked to amyloid deposition via Pin1 (peptidyl-prolyl T1R3 KO mice have altered cognitive function cis-trans isomerase NIMA-interacting) (141)(142)(143). Pin1 activity prevents neurons from exiting their post-mitotic state and pathologically re-entering the cell cycle. We observed a significant elevation of T1R3KO hippocampal Pin1, which may indicate a reflexive response to prevent excessive amyloid and tau pathologies. Pin1 overexpression is involved in the degradation of wild-type Tau as seen in AD (144) and FTD with parkinsonism linked to chromosome 17 (FTDP-17) (145). Unlike AD, FTD patients often present first with significant social/syntactic abnormalities prior to the later onset of cognitive decline/ memory loss. In this respect the combination in the T1R3KO model of (i) altered social interactions, (ii) cognitive decline, and (iii) neuronal dysfunction presents a potentially intriguing murine insight into this condition. FTD is also strongly associated with NMDA receptor dysfunction (146), seen in the T1R3KO mice; NMDA receptor antagonists are also considered potential therapeutic strategies for FTD (146). We also noted from our proteomic T1R3KO signatures a significant reduction in the putative L-DOPA (L-3,4-dihydroxyphenylalanine) receptor (ocular albinism type I (OA1)/Gpr143). Dysfunctions in the dopamine and L-DOPA signaling systems have been strongly associated with FTD etiology (139,(147)(148)(149). OA1 is a discrete L-DOPA receptor, and given that FTD is often presented with linguistic abnormalities it is interesting to note that L-DOPA has been demonstrated to engender a highly specific semantic nootropic effect (150,151). We also noted a significant reduction in hippocampal T1R3KO expression of Slc6a17 (sodium-dependent neutral amino acid transporter Slc6a17), in which homozygous mutations are linked with FTD-like neuropathologies including, diminished cognitive ability, linguistic degradation, and psychosocial deficits (152). Effective loss of this protein has also been correlated with disruptions in dendritic spine formation and perturbations of glutamatergic signaling patterns, both features evident in the T1R3KO mice (153).
Using ourWordCloud-based appreciation of the differential effects of T1R3 loss between hippocampal and cortical tissues, it was evident that a strong cerebellar component of the resultant pathophysiology was prominent (Fig. 7A). Recent research has suggested that in addition to its classical role in fine-motor control, the cerebellum may also possess a higher cognitive functional role (154 -157). The cerebellum has also been implicated as a functional locus in cognitive/psychosocial disorders such as ASD (158), a role controlled by physical communication tracts between the hippocampal/amygdalic regions and the cerebellum itself (159). Further research into this potential hippocampal-cerebellar axis may prove fruitful for identifying novel multifunctional neurotherapeutics.
In addition to the demonstration of dementia-like cognitive and social deficits, T1R3KO mice also demonstrated some degree of reluctance to explore an open field, potentially indicative of a spatial anxiety. As with the altered Nlgn3-Nrxn1 stoichiometry described previously, we also observed an intriguing differential expression pattern of Hint1 in T1R3KO mouse tissues, i.e. significant hippocampal reductions combined with significant cortical elevation. A similar cortico-hippocampal expression divergence for Hint1 is associated with perturbed social behavior in mice (160), and Hint1 knock-out mice demonstrate heightened levels of anxiety-related behavior (161).
Our data reported here demonstrate the complex multidimensional impact that loss of the T1R3 exerts upon neuronal functionality. Further research into this paradigm should now be focused perhaps on a more reductionist approach, to piece together the physical molecular connectivity between our phenomenological results and the actual sweet-taste heterodimer receptorsome structures themselves (162).
Considering our data, which ranges from alterations in molecular signatures in the CNS, to neuronal architecture, in vitro cellular signaling activity, and behavioral modifications, it is evident that the sweet-taste receptor system likely plays a multidimensional role in cognitive, psychosocial, and neuroprotective activities in the brain. This concept then necessitates further investigation into the role of this receptor system in complex processes such as neurometabolic aging and dementia, and its future therapeutic exploitation may demonstrate important efficacy profiles for multiple disorders linking the glucose-sensory/ metabolic system and cognitive/psychosocial health.

Animal care and use
All animal procedures were approved by the Animal Care and Use Committee (ACUC) of the NIA, National Institutes of Health. T1R3KO mice (8), kindly provided by Dr. Charles Zuker (Columbia University), were cross-bred with WT C57BL/6J mice (The Jackson Laboratory) for at least four generations. All animals were kept in a 12-h light/dark cycle and received food (standard Harlan-Teklad chow, LM-485) and water ad libitum. A total of 10 -20 mice (4 -5 months of age) in each group were employed for behavioral, histological, and protein expression analyses. These 4 -5-month-old animals were used in experiments and euthanized right after the experiments. For post mortem tissue extraction, animals were humanely euthanized with isoflurane, and specific organs/tissues were then carefully dissected, snap-frozen, and stored at Ϫ80°C until further analysis.

Murine behavior assessment
The assessment of general activity levels, exploratory behavior, and locomotor activity in T1R3KO mice compared with control WT mice was performed using an open-field test as described previously (163). Open-field activity-monitoring software (ANY-Maze, Stoelting Co., Wood Dale, IL) automatically recorded multiple parameters over the duration of the 10-min test.
Anxiety-related behavior was assessed using a light-dark transition assessment as well as performance in an elevated plus maze. The light-dark transition test was performed as described previously (164). The square chamber used in the open-field test was also used to complete this test, with the additional insertion of a dark chamber occupying up to 50% of the apparatus area. Time spent in either the light or dark compartment was automatically recorded by the ANY-Maze monitoring software (Stoelting). The elevated plus maze (San Diego Instruments, San Diego) consisted of two open arms and two closed arms that extended from a common central platform; the experiment was performed as described previously (163). Each mouse was placed in the center square facing an open arm and allowed to freely explore the apparatus for 10 min. ANY-Maze behavioral tracking software was used to record time spent in either the closed or open arms of the apparatus, and the relative time spent in the closed or open arms was calculated by dividing the time spent in either arm by the total exploration time (600 s for all animals).
Social behavior was assessed using a three-chambered sociability apparatus (Stoelting) with a protocol based upon previously established methods (165). Another mouse of the same strain, age, and gender and an inanimate object housed in identical wire-mesh containers were placed in opposing chambers of the apparatus. Experimental animals were then returned to the center chamber of the apparatus where they were free to interact with either the mouse or object. Interactions were recorded for 10 min and analyzed with ANY-Maze monitoring software. A discrimination index taking into account the relative exploration times between the target mouse and the inanimate object, was calculated.
Differences in learning and memory between T1R3KO and WT mice were investigated using both the NOP and MWM tests as described previously (163). For the NOP test, both the habituation session and the testing session were videotaped and the animal's behavior was tracked automatically by ANY-Maze monitoring software. The discrimination index between the two objects was calculated. The MWM was used as described previously (163) to further assess cognitive function. After 4 days of training, animals were given a probe trial on day 5 in which the platform was removed from the pool and the amount of time spent in each quadrant was recorded within 1 min. The probe trial indicates whether the animal can remember in which quadrant the escape platform was located previously. The escape latency, swimming path length, and swimming speed of each animal during each trial were also recorded and analyzed.
Motor coordination was evaluated using an accelerating Rota-Rod treadmill (Med Associates Inc., St. Albans, VT) as described previously (164). On the test day, the mice were placed on the Rota-Rod, which gradually accelerated from 4 to 40 rpm over the 5-min test time. The test was performed twice/ day, and the latency to fall was measured and averaged.

Gustatory activity assessment
Taste-testing experiments took place during daylight hours in a manner similar to that described previously (1,164). T1R3KO and WT mice (n ϭ 8/testing group) were habituated to the laboratory environment for 35 min each experimental day before the initiation of taste testing. Test stimuli consisted of various concentrations of sucrose (1, 3, 10, 30, and 100 mM, Fisher Scientific, Atlanta, GA) or sucralose (0.001, 0.01, 0.1, 1, and 5 mM, Toronto Research Chemicals, Toronto, Canada). Brief-access taste-testing took place in a Davis Rig gustometer (Davis MS-160, DiLog Instruments, Tallahassee, FL), as described previously (166 -169). Brief-access procedures minimize post-ingestive effects that may confound other assays such as intake tests (168). Before taste-testing was initiated, mice were trained to lick a stationary tube of water in the gustometer after being placed on a 24-h restricted water-access schedule. For appetitive stimuli (i.e. sucrose and sucralose) animals received 3 days of testing with the five stimulus concentrations and purified water. A water rinse presentation (1 s) was interposed between the test trials for normally avoided stimuli to help control for potential carryover effects of the agents.

Western blot analyses
Brain tissues were lysed in a modified radioimmune precipitation assay buffer (150 mM NaCl, 50 mM Tris, 0.5% sodium deoxycholate, 1% Nonidet P-40, and 1% SDS). Subcellular fractionation (into the cytosolic, plasma membrane, large organelle, and cytoskeleton compartments) of brain tissues was performed using the Qproteome cell compartment kit according to the manufacturer's instructions (Qiagen, Valencia, CA). All protein extracts were quantified using BCA reagents (Thermo Fisher Scientific) before resolution with SDS-PAGE and electrotransfer to PVDF membranes (PerkinElmer Life Sciences). Membranes were blocked with 5% nonfat milk and incubated overnight at 4°C using primary antibodies. After a 1-h incubation with the appropriate secondary antibody, membranes were developed and scanned by a Typhoon 9410 variable-mode phosphorimaging machine. Blots were probed with the following antibodies: rabbit anti-spinophilin, mouse anti-NR1, and rabbit anti-NR2B from Millipore (Billerica, MA); rabbit anti-phospho-CREB, rabbit anti-CREB, rabbit antiphospho-ERK1/2, rabbit anti-ERK1/2, rabbit anti-phospho-synapsin1, rabbit anti-synapsin1, rabbit anti-synaptophysin, and rabbit anti-PSD95 from Cell Signaling Technology, Inc. (Danvers, MA); rabbit anti-CPLX1 and rabbit anti-HINT1 from Abcam (Cambridge, MA); rabbit anti-SLC8A1 from Epitomics (Burlingame, CA); and mouse anti-␤-actin from Sigma-Aldrich. The experiments were repeated at least three times. Results were normalized to actin expression.

Immunohistochemistry
Animals were anesthetized with carbon dioxide and transcardially perfused with 4% paraformaldehyde. Brains were postfixed overnight in 4% paraformaldehyde and then cryoprotected in 30% sucrose and stored at 4°C until use. Whole brains were sectioned (40 m) using a Cryotome TM cryostat (Thermo Fisher Scientific). The sections were then processed for immunohistochemical localization of NR1 and NR2B (mouse anti-NR1 and rabbit anti-NR2B) using a Vectastain Elite ABC kit (Vector Laboratories, Burlingame, CA) and developed using a Dako liquid diaminobenzidine chromogen system (DakoCytomation California, Carpinteria, CA). Sections were imaged with an Olympus Fluoview IX70 microscope (Olympus America, Center Valley, PA).

Brain Golgi staining
Brains were processed for Golgi-Cox staining using an FD Rapid GolgiStain TM kit according to the manufacturer's instructions (FD NeuroTechnologies, Columbia, MD). 100-m coronal sections were sliced on a cryostat and mounted onto gelatin-coated glass slides. After development in NH 4 OH for 10 min, sections were dehydrated through a series of graded ethanol washes and cleared in 100% Histo-Clear (National Diagnostics, Atlanta, GA) three times. Slides were then coverslipped

T1R3 KO mice have altered cognitive function
with Permount TM mounting medium (Fisher Scientific) and dried in the dark for at least 1 week. Neurons within the hippocampal CA1 region, fully impregnated with the Golgi-Cox solution, were selected for further analysis using Neurolucida Version 8 software (MBF Bioscience, Williston, VT). These chosen neurons were not obscured by neighboring neurons and had no obviously truncated dendrites. For each animal (n ϭ 8), six representative neurons were traced throughout the sections using a Leica DMRB microscope ϫ100 (NA 1.30) lens. Dendritic spine density and dendrite length were analyzed using Neurolucida Explorer software (MBF Bioscience).

Primary neuron culture
Cortical neurons were prepared from WT and T1R3KO embryonic mice on day 16 as described previously (163). In brief, cells were dissociated by incubation with TrypLE TM Express (Life Technologies, Inc.) at 37°C for 25 min. Tissues were washed twice with ice-cold Hanks' balanced salt solution (Sigma-Aldrich) after the addition of minimum essential medium containing 10% FBS. A 70-m BD Falcon cell strainer was used to remove the undissociated tissue. Cells were plated in polyethyleneimine-coated 35-mm-diameter dishes at a density of ϳ100 cells/mm 2 culture surface. Cultures were maintained in Neurobasal medium containing B-27 supplements, 10 units/ml penicillin, 10 g/ml streptomycin (Life Technologies, Inc.), 2 mM L-glutamine, and 1 mM HEPES (Sigma-Aldrich). All experiments were subsequently performed on healthy primary neuronal cultures after 6 -8 days of in vitro culture.

Primary neuron calcium imaging
Primary neurons were loaded with 5 M Fluo-8 AM (TEFLabs, Austin, TX) for 20 min followed by washing three times. Confocal imaging was performed using a Zeiss (Oberkochen, Germany) 710 confocal laser-scanning microscope with a ϫ40/1.3 oil DIC M27 reverse objective. The 488-nm argon laser line was used to excite Fluo-8 fluorescence, which was measured by a band-pass filter from 493 to 630 nm. The illumination intensity was kept to a minimum (at 2.0% of laser output) to avoid photobleaching, and the pinhole was set to give an optical slice of 3 m. Glutamate was added to generate a final concentration of 100 M after an 80-s baseline recording period. A 5-min time series of fluorescence images was recorded. Images of Fluo-8 fluorescence intensities were collected every 2 s. The fluorescence intensity presented was normalized with respect to time zero, calculated by ⌬F/F ϭ (F Ϫ F base )/(F base Ϫ B). The experiment was repeated multiple times (n ϭ 4), and 20 representative cells from each group were chosen to perform the statistical analysis.

RNA extraction and microarray analysis
For microarray analysis of transcriptomic expression patterns in murine CNS tissues, specific, dissected brain regions were first homogenized using a Beadbeater (Biospec, Bartlesville, OK) followed by RNA purification using the RNEasy Mini Kit (Qiagen) according to the manufacturer's instructions. The RNA was examined for quantity and quality using an Agilent Bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA). Total RNA was used to generate biotin-labeled cRNA using an Illu-mina TotalPrep RNA amplification kit (Ambion, Austin, TX). A total of 0.75 g of biotin-labeled cRNA was hybridized at 58°C for 16 h to the Illumina Sentrix Mouse Ref-8 Expression BeadChip kit (Illumina, San Diego, CA). The microarrays were then washed and blocked, and the labeled cRNA was detected by staining with streptavidin-Cy3. The arrays were scanned using an Illumina BeadStation ϫ500 genetic analysis systems scanner, and the image data were extracted using Illumina BeadStudio software, version 3.0. We deposited the raw transcriptomic data at GEO/ArrayExpress under accession number GSE97677. All experimental details are MIAME (minimum information about a microarray experiment)-compliant.

iTRAQ quantitative proteomics
Isobaric tags for relative and absolute quantification (iTRAQ)-based proteomics was performed as described previously (170). iTRAQ labeling reagents were obtained from Applied Biosystems (Carlsbad, CA). WT and T1R3KO tissue samples (150 g of initial total protein from each animal) were treated in parallel throughout the iTRAQ labeling procedure. Post-labeling WT control and T1R3KO peptide samples (for a single specific tissue) were combined and dried down to a volume of 50 l to reduce the content of ethanol prior to strong cation exchange chromatography, which was performed as described previously (171); the labeled samples were introduced by nanospray mass spectrometry.

Mass spectrometry analysis
An Eksigent NanoLC Ultra 2D system (Dublin, CA) and a Thermo Fisher Scientific LTQ Orbitrap XL mass spectrometer were used for quantitative proteomic analysis. Tandem liquid chromatography-mass spectrometric analyses were performed as described previously (47). For iTRAQ quantification, the reporter ion intensities in MS2 spectra from raw data sets were used to calculate the expression ratios between the contrasting states. iTRAQ data were analyzed using DIANE 6.0, a spreadsheet-based microarray analysis program based on the SAS JMP7.0 system. Raw iTRAQ data, subjected to filtering and Z-normalization, were tested for significant changes as described previously (163). Briefly, sample quality was analyzed by scatter plot and protein sample Z-score-based hierarchical clustering to exclude possible outliers. The initial filtering identified proteins with a Z-ratio Ն 1.20, with the Z-ratio derived from the difference between the averages of the observed protein Z-scores divided by the standard deviation of all of the differences for that particular comparison. We were able to detect ϳ200 proteins after filtering. Proteins were then refined by calculating the false discovery rate (FDR), which controls for the expected proportion of falsely rejected hypotheses, and including only those proteins with FDR Ͻ 0.05. These data were further analyzed using a one-way analysis-of-variance design with significance set at p Ͻ 0.05. The analysis-of-variance design was compared between the T1R3KO and WT groups. This allowed us to identify proteins that differed in their expression for T1R3KO compared with the WT controls.

Bioinformatic analyses
After identifying significantly altered proteins using iTRAQ, the protein sets were analyzed using a multidimensional bioinformatic approach. To facilitate the specific separation of complex data sets, we employed the novel Venn diagram platform, VennPlex (172). IPA was employed for canonical signaling pathway enrichment as well as biofunction/disease annotation. For both of these pathway/functional enrichment techniques the same minimum pathway population criteria were applied, i.e. Ն2 experimental proteins present for each pathway at an enrichment probability of p Յ 0.05. In addition to these standard pathway enrichment approaches, we employed our inhouse-developed NLP latent semantic indexing-based informatics platform, Textrous! (50). Textrous! facilitates the creation of highly data-specific de novo signaling descriptions, thus generating a more nuanced appreciation of high-dimensionality pathological signaling paradigms (173,174). Word clouds were generated with Wordle (http://www.wordle.net/) 4 from Textrous!-based NLP analyses. To extract both word and phrase frequencies from our Textrous! output, we employed the Word Frequency Counter application from WriteWords (http://www.writewords.org.uk/word_count.asp). 4