MOLECULAR CORRELATE OF MOUSE EXECUTIVE FUNCTION. TOP-DOWN AND BOTTOM-UP COMPLEMENTATIONS BY PRESYNAPTIC VERTEBRATE BRAIN-SPECIFIC Ntng GENE PARALOGS

Executive function (EF) is a regulatory construct of learning and general cognitive abilities. Genetic variations underlying the architecture of cognitive phenotypes are likely to affect EF and associated behaviors. Mice lacking one of Ntng gene paralogs, encoding the vertebrate brain-specific presynaptic Netrin-G proteins, exhibit prominent deficits in the EF control. Brain areas responsible for gating the bottom-up and top-down information flows differentially express Ntng1 and Ntng2, distinguishing neuronal circuits involved in perception and cognition. As a result, high and low cognitive demand tasks (HCD and LCD, respectively) modulate Ntng1 and Ntng2 associations either with attention and impulsivity (AI) or working memory (WM), in a complementary manner. During the LCD Ntng2supported neuronal gating of AI and WM dominates over the Ntng1-associated circuits. This is reversed during the HCD, when the EF requires a larger contribution of cognitive control, supported by Ntng1, over the Ntng2 pathways. Since human NTNG orthologs have been reported to affect human IQ (1), and an array of neurological disorders (2), we believe that mouse Ntng gene paralogs serve an analogous role but influencing brain executive functioning.


INTRODUCTION
Executive function (EF) is a heterogeneous construct that can be viewed as a set of processes executively supervising cognitive behaviors (3). EF is an umbrella term for working memory (WM), attention and impulsivity (AI), and response inhibition, and is thought to account for the variance in cognitive performance (4). WM, due to its storage and processing components, is viewed as a bimodal flexible system of a limited capacity. Since WM maintains current information and simultaneously supports its execution, as a latent factor underlying intelligence (5), it has been termed as "the central executive" (6) attention-controlling system dependent on consciousness (7).
However an awareness-independent model has been also proposed (8,9). General learning (Ln) ability depends on attention and WM interaction (10) as well as perception, the causal and informational ground for the higher cognitive functions (11). Perception guides our thinking about and acting upon the world and serves as an input to cognition, via a short-term memory mediated interactions (12). A possible mechanism linking perception and cognition would be attention (13 14), consequently elaborated into the brain predictive coding approach currently dominating cognitive neuroscience (15), and positing that attention is a property of brain computation network (16). However this has been challenged by the opposite opinion that "cognition does not affect perception" (17). Regardless whether or not such a cognitive-sensory dichotomy exists, herein we view perception and cognition as two main information streams the EF exerts its actions upon, possibly through active association.
We have previously described the function of two vertebrate-specific brainexpressed presynaptic gene paralogs, the human brain phenotypically fragile (2).
Despite the fact that EF abrogation is a major determinant of problem behavior and disability in neuropsychiatric disorders (19), the genetics underlying EF remains elusive with no causative vector agents (e.g. genes) have yet been reported.
Herein we show that NTNG paralogs affecting human IQ also affect mouse learning and brain executive functioning.   as (x,y) coordinates we have assessed the phenotypic proximity of the Ntng1 -/and      Figure 2. Working memory (WM) estimate and the effect of cognitive demand by the analysis of rank and its variance for Ntng1 -/and Ntng2 -/mice. A,E. Mouse ranks and rank PVE (proportion of variance explained) based on four parameter rank measures (SF2) as detailed in ST2-1 (for Ntng1 -/-) and ST2-2 (for Ntng2 -/-). The rank sorting was done in a genotype-independent manner treating all mice together. Ranking for each out of four parameters was done independently of other parameters with a final reranking of the ranks sum to generate the final rank (shown). In case of an equal sum of the ranks, the mice were given identical ranks. PVE was calculated as a square of within genotype rank variance divided on the sum of each genotype variances squares multiplied on 100%. B,F. Mice rank distribution across one-to-four parameters as top 4 and bottom 4 performers. C,G. Genotype-specific placing among the mice. D,H. Behavioral consistency of mice across the sessions (y axis, sum of r 2 correlations of a single session ranks vs. final ranks for each mouse across the sessions) and behavioral parameter crosscorrelations (x axis, the r 2 correlation of a parameter final ranking vs. final ranking for all 4 parameters). The gene ablation-specific phenotype severity can be assessed visually by matching each parametercorresponding vertexes of the obtained quadruples. p value represents a Wilcoxon rank sum test.!    ) and (8)(9)(10)(11)(12)(13)(14), middle panel, defined as low cognitive demand (LCD) and high cognitive demand (HCD) sessions, respectively. One and two-way ANOVA was used for the statistics. B. Ranks and PVE comparisons over the LCD and HCD. The rank sorting was done in a genotype-independent manner, similar to Fig.1 and Fig.2, but using only one parameter, success (Sc), see ST2-1 and ST2-2 (Ln). Rank statistics was by Wilcoxon rank sum test. C. Learning (Ln) vs. attention and impulsivity (AI) rank correlations (from Fig.1A-1,2).!      (2-7) and (8)(9)(10)(11)(12)(13) indicates the data split on the low cognitive demand (LCD) and high cognitive demand (HCD) sessions. (B-E, middle left and right) The same data as above but averaged per LCD and HCD sessions. The data for each parameter are presented as a mean±SEM (standard error of mean) and fully provided in ST1-1 (for Ntng1 -/-) and ST1-2 (for Ntng2 -/-). Two-way and one-way ANOVA were used for statistics.!    7) and (8)(9)(10)(11)(12)(13)(14) indicates the data split on the low cognitive demand (LCD) and high cognitive demand (HCD) sessions. (B-E, middle left and right) The same data as above but averaged per LCD and HCD sessions. The data for each parameter are presented as a mean±SEM (standard error of mean) and fully provided in ST2-1 (for Ntng1 -/-) and ST2-2 (for Ntng2 -/-). Two-way and one-way ANOVA were used for statistics.!   Table 6-2 (ST6-2). C-means fuzzy clustering (Euclidean C-means) rank probabilities for Ntng1 -/and Ntng2 -/mice performance on the radial arm maze (RAM) task.!