Synaptic scaffold evolution generated components of vertebrate cognitive complexity

Abstract

The origins and evolution of higher cognitive functions, including complex forms of learning, attention and executive functions, are unknown. A potential mechanism driving the evolution of vertebrate cognition early in the vertebrate lineage (550 million years ago) was genome duplication and subsequent diversification of postsynaptic genes. Here we report, to our knowledge, the first genetic analysis of a vertebrate gene family in cognitive functions measured using computerized touchscreens. Comparison of mice carrying mutations in each of the four Dlg paralogs showed that simple associative learning required Dlg4, whereas Dlg2 and Dlg3 diversified to have opposing functions in complex cognitive processes. Exploiting the translational utility of touchscreens in humans and mice, testing Dlg2 mutations in both species showed that Dlg2's role in complex learning, cognitive flexibility and attention has been highly conserved over 100 million years. Dlg-family mutations underlie psychiatric disorders, suggesting that genome evolution expanded the complexity of vertebrate cognition at the cost of susceptibility to mental illness.

Figure 1. Dissecting the role of Dlg paralogs in different components of cognition

a. Comparison of invertebrate Dlg and 4 vertebrate paralogs (Dlg1-4). Two pairs of vertebrate Dlg genes can be identified reflecting their evolutionary origins in the two rounds of whole genome duplication (1R, 2R) at the base of chordates ~550 million years ago (Mya). Yellow box highlights the 4 vertebrate Dlg proteins showing high conservation of domain architecture. b. A battery of rodent touchscreen tasks with increasing cognitive complexity was used to probe simple and complex forms of information processing. Seven tasks are grouped into 4 coloured boxes and representations of the stimuli displayed on the touchscreen are shown. Conditioning and simple forms of learning were measured using tests for operant and Pavlovian conditioning. More complex forms of learning (visual and visuo-spatial discrimination) and information processing (cognitive flexibility, inhibitory response control and attention) were measured using tests that involved more complex perceptual stimuli and/or required more complex response control.

Figure 2. Distinct roles of Dlg paralogs in simple forms of conditioning and associative learning

a. Mice were trained through several phases to nose-poke a stimulus displayed on the touchscreen to attain a reward (Operant conditioning). Animals were required to successfully complete and reach the set criterion at each phase before advancing to the next phase. Phase 1: animals acclimated for 20min on 2 days to the operant chamber and required to consume reward pellets freely available from the magazine. Phase 2: a single visual stimulus was displayed on the touchscreen after which, the disappearance of the stimuli coincided with delivery of a food reward, presentation of a tone and illumination of the pellet magazine. Phase 3: animals were required to nose poke a visual stimulus that appeared on the touchscreen in order to obtain a reward. Phase 4: animals were additionally required to initiate the commencement of a new trial with a head entry into the pellet magazine. Phase 5: In addition to that described for previous phases, responses at a blank part of the screen during stimulus presentation now produced a 5s time-out and were not rewarded. Dlg2^−/− and Dlg3^−/Y mice displayed normal acquisition rates relative to wt littermate controls. Dlg4^−/− mice were able to successfully complete Phases 1 and 2 but were unable to complete trials during Phase 3, even after 20 sessions of training (*p<0.05). b. Pavlovian conditioned approach. Number of approaches to CS+ and CS− (left graph). Wt (black) and Dlg4^−/− mice (white). Wt: mixed between-within subjects ANOVA, main effect of genotype p<0.001, stimulus (CS+/ CS−) x session interaction p<0.001, post hoc paired samples t-test *p<0.001. Approach latency to CS+ and CS− (right graph). Wt: independent samples t-test, *p<0.05. c. Visual discrimination learning. Total number of trials (left graph) (wt: 210.91±19.76, Dlg2^−/−: 222.7±26.18) (wt: 243.46±18.25, Dlg3^−/Y:173.38±10.06) and errors (right graph) (wt: 64.36±7.9, Dlg2^−/−: 68.0±10.13) (wt: 81.54±8.60, Dlg3^−/Y:57.46±5.27) to reach learning criterion on visual discrimination. Dlg3^−/Y: independent samples t-tests, trials *p<0.005, errors *p<0.05. d. Object-location paired-associates learning. L, left; C, centre; R, right. Percentage of correct responses across training sessions for Dlg2^−/− (left graph) and Dlg3^−/Y (right graph) mice. Dlg2^−/−: mixed between-within subjects ANOVA, main effect of genotype p<0.001, genotype x session interaction p<0.001, post hoc paired samples t-test *p<0.05. All values reported represent mean ± SEM.

Figure 3. Dlgs play distinct roles in cognitive flexibility and response inhibition

a. Reversal learning. Percentage of correct responses across sessions for Dlg2^−/− (left graph) and Dlg3^−/− (right graph) mice. Dlg2^−/−: mixed between-within subjects ANOVA sessions 1-8, main effect of genotype *p<0.05. b. Dlg2^−/− mice on reversal learning using complex perceptual stimuli. Percentage of correct responses across sessions on reversal learning. Dlg2^−/−: mixed between-within subjects ANOVA, main effect of genotype *p<0.005. c. Extinction learning. Percentage of responses made across sessions by Dlg2^−/− (left graph) and Dlg3^−/Y (right graph) mice. Dlg2^−/−: mixed between-within subjects ANOVA, main effect of genotype p<0.05, genotype x session interaction p<0.005, post hoc paired samples t-test *p<0.05. Dlg3^−/Y: mixed between-within subjects ANOVA, main effect of genotype *p<0.05. All values reported represent mean ± SEM.

Figure 4. Dlgs are differentially involved in attentional processing and response control

a-c. Performance on the 5-Choice serial reaction time task (5-CSRTT). See methods for detailed description of task. Dlg2^−/− mice (graphs on left) and Dlg3^−/Y mice (graphs on right). a. Percentage accuracy (% correct responses). b. Percentage of premature responses. c. Number of perseverative responses. Dlg2^−/−: accuracy and premature responses, mixed between-within subjects ANOVA, main effects of genotype *p<0.01, genotype x session interaction p<0.05, post hoc paired samples t-test *p<0.05. Dlg3^−/Y: accuracy and premature responses, mixed between-within subjects ANOVA, main effects of genotype *p<0.05. All values reported represent mean ± SEM.

Figure 5. Dlg paralogs have diversified to play distinct roles in different cognitive functions

a. Summary of Dlg phenotypes in twelve measures within six cognitive tests. The cognitive repertoire is grouped into 4 boxes according to Figure 1b. Invertebrate Dlg mutants have lethal phenotypes as does mouse Dlg1^−/− however, presence of a single copy of the Dlg1 gene (Dlg1^+/−) was sufficient for these mice to perform normally across the different cognitive functions examined. Dlg4 was essential for simple forms of associative learning. Some cognitive functions were enhanced by a mutation in one Dlg gene (Dlg3) and attenuated or suppressed by a mutation in another Dlg gene (Dlg2), revealing that Dlg2 and Dlg3 play opposing regulatory roles in more complex cognitive processes. b. Clustering of gene-phenotype relationships shows 4 groups of cognitive functions (Cognitive Clusters 1-4).

Figure 6. Conservation of Dlg2 functions in mice and humans

Using the Cambridge Neuropsychological Test Automated Battery (CANTAB), 4 individuals with mutations in DLG2 were assessed on 3 tasks comparable to those within the rodent touchscreen battery. The Intra-extradimensional set-shifting task (IED) was used to assess discrimination acquisition and cognitive flexibility, the Paired Associates Learning task (PAL) to assess visuo-spatial learning and memory and the Rapid Visual Information Processing task (RVP) to assess sustained attention. Comparison of performance in touchscreen tasks for mice carrying mutations in Dlg1, Dlg2, Dlg3 or Dlg4 with humans carrying mutations in DLG2 (see methods). A standardised performance score compared to controls is shown, where a negative score indicates poorer than average performance. #Black bar denotes the lack of data for comparison due to the inability to test Dlg4 mutant mice on any of the three tasks represented.