Human CLOCK enhances neocortical function

Abstract

The transcription factor CLOCK is ubiquitously expressed and important for circadian rhythms, while its human-specific expression in neocortex suggests additional functions. Here, we generated a mouse model (HU) that recapitulates human cortical expression of CLOCK. The HU mice show enhanced cognitive flexibility, which might be associated with alteration in spatiotemporal expression of CLOCK. Cell type-specific genomic profiling identified upregulated genes related to dendritic growth and spine formation in excitatory neurons of HU mice. We also found that excitatory neurons in HU mice have increased dendritic complexity and spine density, and a greater frequency of excitatory postsynaptic currents, suggesting a greater abundance of neural connectivity. In contrast, CLOCK knockout in human induced pluripotent stem cell-induced neurons showed reduced complexity of dendrites and lower density of presynaptic puncta. Together, our data demonstrate that CLOCK might have evolved brain-relevant gains of function via altered spatiotemporal gene expression and that these functions may underlie human brain specializations.

Introduction

The unparalleled advancement of human cognition can be attributed to the evolution of the human brain. Compared with non-human primates, the human brain has undergone neocortical expansion^1,2, increased complexity of neuronal dendritic arborization and neural networks^3–5, and alterations in gene expression^6–11. However, the underlying molecular mechanisms of these specializations are still not clear.

Comparative genomics among humans and non-human primates has demonstrated that the regulation of gene expression, rather than modifications in protein coding sequences, is a major facet of molecular evolution driving human brain specializations^12–14. Thus, elucidating the key regulatory players driving these gene expression programs is critical to understand human brain evolution. Moreover, functional analyses on these candidate genes are required to understand their relative contribution to human brain evolution^14,15. Few studies have functionally tested the role of specific genes in evolved brain function. Studies of human FOXP2, SRGAP2, and CBLN2 have demonstrated changes in transcriptional programs or neuronal function that are fundamentally altered compared to mouse or non-human primate function^16–23.

The transcription factor CLOCK (or circadian locomotor output cycles kaput) has significantly increased human-specific expression in the cerebral cortex compared to non-human primates and mouse^7,24–30 (Extended Data Fig. 1). Of note, CLOCK/Clock is the only core circadian gene that itself lacks rhythmicity in the neocortex^31–36. Transcriptome analyses discovered that a substantial number of downstream targets of mouse CLOCK are also nonrhythmic^37,38. We further showed that transcriptional targets of human CLOCK in neuronal cultures are also not regulated in a cyclical manner³⁹. Moreover, unlike most core circadian genes that have elevated expression following the opening of the eyes to coordinate light entry, CLOCK/Clock maintains stable expression throughout the day across all developmental stages³². Together, these results suggest that CLOCK has additional functions beyond its circadian function in the neocortex.

CLOCK has also been implicated in human brain function via its association with human cognitive disorders. We showed that CLOCK is a hub gene in a human-specific neocortical co-expression network that has significant enrichment for genes involved in neuropsychiatric disorders such as seasonal affective disorder, depression, schizophrenia, and autism⁷. Single nucleotide polymorphism analyses have linked CLOCK with a broad spectrum of neuropsychiatric disorders (e.g., autism spectrum disorder, bipolar disorder, schizophrenia, attention-deficit/hyperactivity disorder, major depressive disorder, and addiction)^40,41. Patients with epilepsy have decreased expression of CLOCK in the temporal cortex³³. We showed that CLOCK knockdown in human neurons leads to altered expression of genes important for neurodevelopment³⁹. However, the mechanisms downstream of CLOCK that are important for human brain function are unknown.

Here, we generated a humanized mouse model that contains both the protein-coding and regulatory regions of human CLOCK in a mouse Clock knockout background. We then leveraged this mouse model to investigate how human CLOCK affects cognitive behaviors, cell type-specific gene expression, neuroanatomy, and electrophysiology. We discovered that human CLOCK alters gene expression in a spatiotemporal manner to promote connectivity of neurons in the neocortex for enhanced cognitive flexibility.

Results

Generation of CLOCK humanized mice

We generated humanized CLOCK mice (HU) by expressing human CLOCK together with its flanking non-coding genomic regions in mice lacking endogenous Clock (knockout, KO). A bacterial artificial chromosome (BAC) containing the full-length human CLOCK gene was edited through recombineering to include flanking non-coding regulatory regions (Fig. 1A) and delivered via pronuclear injection into embryos. We selected positive lines (CLOCK^hum;Clock^+/+) (Supplementary Fig. 1A,B), which contained 7–8 BAC copies on average (Fig. 1B), and crossed them with Clock knockout mice (KO:Clock^−/−)⁴². F1 heterozygotes (Clock^+/− and CLOCK^hum;Clock^+/−) were crossed to obtain littermates of humanized (HU:CLOCK^hum;Clock^−/−), KO, and wildtype (WT:Clock^+/+) mice (Supplementary Fig. 1C). A mouse Clock BAC transgenic line (Clock^mus;Clock^+/+) that we previously generated was similarly crossed to Clock knockout mice to obtain mouse Clock overexpression mice (MS:Clock^mus;Clock^−/−) as a further control of human CLOCK overexpression in HU mice⁴³.

Fig. 1: Human BAC transgenic mice express human CLOCK in a greater percentage of cells.

(A) Recombineering of a BAC to obtain human CLOCK with flanking regulatory regions. (B) qRT-PCR results to calculate the number of BAC copies inserted into the genomes of two of the mouse lines. (C) qRT-PCR with species-specific CLOCK/Clock primers. For panels B and C, each data point is an individual. (D) Representative images of IHC staining for CLOCK (green), oligodendrocytes (OLIG2; red), and neurons (NEUN; magenta) in frontal cortex. DAPI indicates nuclei. (E-G) Quantification of CLOCK+ fractions in (E) all cells, (F) neurons, and (G) oligodendrocytes. Each data point is an image of subarea in the frontal cortex. We sampled 3–4 adult mice per genotype (HU: n = 4; WT: n = 3; KO: n = 3) and 3 images per mouse. For panels E-G, we applied a general linear mixed model (GLMM) with genotype and sex as fixed factors, image as random factor nested with individual, and Tukey’s test for post-hoc analysis. The open circles represent female samples, while closed circles represents male samples. Details of statistical results can be found in Supplementary Table 1.

All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Using qRT-PCR, we found that HU and MS mice exclusively expressed human CLOCK or mouse Clock with about 2.5-fold upregulation in the frontal cortex respectively compared to the WT mice (Fig. 1C). This 2.5-fold increase in CLOCK is similar to the relative increase we previously reported when comparing human and chimpanzee frontal cortex tissue⁷. Optical Genome Mapping (OGM) was conducted to understand the internal structure and the inserted location of the BAC. We found that in human CLOCK line 7 (hc7) and hc3 BACs have been inserted into Chromosome 6 and Chromosome 3 respectively (Supplementary Fig. 1D,F). In hc7, six copies of the BAC have been inserted at about 50,000bp away from the 3’UTR of Grm8 gene and three copies are intact (Supplementary Fig. 1E). While in hc3, eight copies of the BAC have been inserted at about 10,000bp away from the 5’UTR of Ralyl gene and three copies are intact (Supplementary Fig. 1G). These results could reconcile the BAC copy number and expression level observed in Fig. 1B and Fig. 1C respectively. Using immunohistochemistry (IHC), we observed protein expression of CLOCK in the frontal cortex of both HU and WT mice but not in KO mice (Supplementary Fig. 1H). We then quantified the fraction of CLOCK positive cells in the frontal cortex and found a greater fraction in HU mice than in WT mice (Fig. 1D,E). We used anti-NeuN and anti-OLIG2 to label neurons and oligos (including both oligodendrocytes and oligodendrocyte precursor cells) respectively, and found that HU mice demonstrated expression of CLOCK in a greater fraction of both cell types than WT mice (Fig. 1D,F,G). The percentage of CLOCK+ neurons in the HU mice (89%) is similar to what has been observed in the human cortex (86%), while the percentage of CLOCK+ neurons in WT mice (62%) is similar to previous reports in mice (66%)^33. To rule out the possibility that the increased fraction of human CLOCK positive neurons in HU mice is due to BAC transgenic overexpression, we compared it to the results of IHC in MS mice which overexpress mouse CLOCK. We found that the fractions of mouse CLOCK+ cells were similar to WT mice and significantly less than the percentages in HU mice (Fig. 1D–G). Furthermore, functional tests demonstrated that human CLOCK rescued deficits of KO mice in circadian activity, body weight, aggression, and depression (Supplementary Fig. 2A–G). Therefore, human and mouse have different expression patterns of CLOCK/Clock in cortical neurons, and our humanized mouse model shows a similar expression pattern of CLOCK as human in neocortical neurons.

To determine whether HU mice express functional CLOCK protein, we examined whether these mice sufficiently rescue Clock loss of function behaviors. We tested mice in a classic wheel-running paradigm for circadian rhythms. We observed a shortened locomotor activity period in both constant light and darkness in KO mice compared to the WT mice, which replicated previous results^42,44. HU mice partially rescued the reduction of period in both constant light and darkness conditions (Supplementary Fig. 2B,C), which is consistent with the rescue of period in ClockΔ19 mice with mouse Clock overexpression⁴³. Like ClockΔ19 mice, adult KO mice are heavier than WT mice, whereas HU mice partially rescue the increase in body weight normalized to body length⁴⁵ (Supplementary Fig. 2D,E). Because ClockΔ19 mice had previously been shown to demonstrate mania-like behavioral phenotypes^46,47, we carried out a tail suspension test. KO mice showed decreased immobility, in line with previous results, and this was partially rescued in HU mice (Supplementary Fig. 2F). Finally, we ran a resident-intruder test to quantify aggressive behavior. We found that KO mice were more aggressive than WT mice, while HU mice rescued the aggressive behavior completely (Supplementary Fig. 2G). The performance of KO mice was consistent with the behavior of Nr1d1 knockout mice⁴⁸. NR1D1 is a downstream target of CLOCK^49,50 and our result suggests that human CLOCK can effectively regulate downstream targets in the mouse genome. In summary, we generated a humanized mouse model that expresses functional human CLOCK.

Human CLOCK enhances cognitive flexibility

Because CLOCK has been linked with cognitive diseases, we next examined whether HU mice have altered performance in cognitive tests. We first assessed mice in a series of sensory-motor tests to determine if manipulation of CLOCK/Clock alters sensory or motor functions. We carried out an open field test that measures both activity and anxiety, an olfactory discrimination test, an eye blink reflex, an ear twist reflex, and a whisker twist orientation. There were no differences across all genotypes in any of these tests (Supplementary Fig. 2H–J). In the novel object recognition test, recognition memory was marginally impaired in KO mice, while there was no difference between HU and WT mice (Fig. 2A). In the fear conditioning test, we did not detect differences in associative learning of auditory cues across genotypes. KO mice showed a deficit in associative learning of context, but HU and WT mice did not differ in any type of associative learning (Fig. 2B). We conducted a 5-trial social memory test and found that KO mice were impaired in social learning, while the difference between HU and WT mice was not significant (Fig. 2C). These results suggest that human and mouse CLOCK play important and similar roles in general learning and memory. To avoid a potential ceiling effect and further differentiate the performance between the HU and WT mice, we challenged them in a set-shifting assay, which requires advanced cognitive ability for rule-based learning to measure executive function, working memory, and cognitive flexibility. We found that the KO mice were significantly impaired in all tasks of the set-shifting assay, while HU mice required significantly fewer trials to reach criterion than WT mice during the reversal task (Fig. 2D; Effect size: Cohen’s D = 1.11). To determine whether the improved performance of HU mice was because of the increase in human CLOCK or random insertion effects of this particular mouse line (hc7), we repeated the discrimination and reversal tasks in a second humanized line (hc3). The results of hc3 replicated what we found in hc7 (Fig. 2E; Effect size: Cohen’s D = 0.87), and we did not find a significant difference between the two lines (Supplementary Fig. 2K). To address whether overexpression of CLOCK in general or whether human-specific CLOCK regulation and function are driving the behavioral results, we repeated the discrimination and reversal tasks in MS mice. MS mice showed similar performances as WT mice (Fig. 2F) and required more trials to reach criterion than HU mice in the reversal task (Supplementary Fig. 2L). In summary, our results indicate that loss of CLOCK in mice results in broad cognitive deficits, while overexpression of human CLOCK, but not mouse CLOCK, improves cognitive flexibility independent of motor activity, anxiety or sensorimotor ability.

Fig. 2: CLOCK affects performance in cognitive tests.

(A) KO mice show a moderate deficit in a novel object recognition test. (B) KO mice have impaired memory of context but not the auditory cue in a fear conditioning test. (C) KO mice show social memory deficits in a 5-trial social memory test. The same intruder was used for the first 4 trials, and then a novel intruder was used for the 5th trial. (D) Trials to criterion in 5 continuous tests of a set-shifting assay. IDS: intra-dimensional shift, EDS: extra-dimensional shift. (E-F) Performance of reversal learning in (E) an alternative humanized mouse line (HU3) and (F) MS mice. Each data point is an adult mouse for panels A, D, E, and F. We applied a general linear model (GLM) with genotype and sex as fixed factors and with Tukey’s test for post-hoc analysis for panel A, D, E, and F. We did a repeated measures ANOVA with genotype and sex as between-subject factors plus time and trial as within-subject factors for panel B and C respectively. The open circles represent female samples, while closed circles represents male samples. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Human CLOCK alters cortical cell density but not lamination

Since CLOCK is expressed during cortical development^51,52, we hypothesized that manipulation of CLOCK might result in neurodevelopmental alterations that underlie the observed changes in cognitive behavior. We measured brain mass with and without normalization to body mass, but we did not find any significant differences across genotypes at postnatal day 7 (P07) or P21 (Supplementary Fig. 3A–D). Although brain mass is unaltered, we asked whether CLOCK alters the organization or structure of the cortex. Our previous work showed that knockdown of CLOCK expression in human neural progenitor derived neurons leads to altered migration in an in vitro neurosphere assay³⁹. Thus, we investigated potential outcomes of migration defects by measuring cortical thickness and lamination of somatosensory cortex in our mouse lines. We used anti-CUX1, anti-FOXP1, and anti-FOXP2 antibodies to label layers 2–4, layers 3–5, and layer 6, respectively (Supplementary Fig. 3E). We found that the absolute cortical thickness was not significantly different across genotypes (Supplementary Fig. 3F), while normalized cortical thickness was decreased in KO mice but not different between HU and WT mice (Supplementary Fig. 3G). Layer thickness was normalized to cortical thickness to quantify lamination. The absolute and normalized thickness of layers 2–4 and layer 6 did not show differences across genotypes (Supplementary Fig. 3H–K). Thus, CLOCK does not alter the laminar organization of the cortex. These results are consistent with other findings in Emx-Cre Clock conditional KO mice³³.

Then, we examined the density of cortical neurons and oligodendrocytes in orbital prefrontal cortex, which is involved in cognitive flexibility of reversal learning^53,54. We found a greater density of both neurons (anti-NEUN+) and oligodendrocytes (anti-OLIG2+ which marks both mature oligodendrocytes and oligodendrocyte precursor cells) in HU compared to WT mice (Fig. 3). To further confirm that this difference is due to the overexpression of human CLOCK specifically, we carried out the same quantification in MS mice and found similar densities as WT mice and significantly lower densities than HU mice (Fig. 3). To assess the generalization of this phenotype throughout the brain, we quantified cell density in two additional areas: somatosensory cortex and nucleus accumbens. We found an increased cell density also in somatosensory cortex but not in nucleus accumbens (Extended Data Fig. 2), suggesting that the change in cell density is a neocortex-specific function of human CLOCK. Although the cell density did not differ between WT and KO mice, the decreased cortical thickness of KO mice resulted in fewer total cells. In summary, CLOCK increases the number of cells in mouse frontal cortex without altering lamination.

Fig. 3: Human CLOCK increases cell density in adult orbital prefrontal cortex.

(A) Representative image of IHC staining for all cells (DAPI, blue), neurons (NEUN, green), and oligodendrocytes (OLIG2, red) in orbital prefrontal cortex. (B-D) Quantification of cell number in (B) all cells, (C) neurons, and (D) oligodendrocytes. Each data point is a section containing orbital prefrontal cortex. For all data panels, we sampled 4–6 adult mice per genotype (HU: n = 6; WT: n = 5; KO: n = 4; MS: n = 4) and 3 sections per mouse. We did GLMM with genotype and sex as fixed factors, section as random factor nested with individual, and Tukey’s test for post-hoc analysis. The open circles represent female samples, while closed circles represents male samples. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Spatiotemporal expression of human and mouse CLOCK/Clock

CLOCK is a transcription factor that is expressed in most cortical cell types^25,32,33. To understand the potential cell type-specific gene regulatory mechanisms underlying the observed alterations of cognitive flexibility, we performed snRNA-seq on the frontal cortex of P07 and P56 mice (n = 2 male and 2 female/genotype; 24 samples in total). Due to high levels of ambient RNA contamination⁵⁵, one sample of each genotype was removed at P07. All samples from both ages had similar quality control metrics (Supplementary Fig. 4A–H). The genotype composition of each cluster is similar (Supplementary Fig. 4I,J), suggesting that CLOCK does not alter the relative proportion of cell types, although we observed a greater overall number of cells in the frontal cortex (Fig. 3). The number of UMIs and expressed genes are consistent with previous work⁵⁶ (Supplementary Fig. 4K–N). We integrated nuclei from all genotypes of the same age and identified major cell types at P07 and P56, respectively (Fig. 4A,B; Supplementary Fig. 5; Supplementary Table 2).

Fig. 4: snRNA-Seq reveals that human CLOCK alters cell type-specific expression patterns in frontal cortex.

(A-B) Uniform manifold approximation and projection (UMAP) plots show the major cell types of frontal cortex at (A) P07 and (B) P56. ASTRO: astrocyte; ENDO: endothelium; EX2/3_IT, EX4/5_IT, EX5_IT, and EX6_IT: intratelencephalic projection neurons in layers 2–3, 4, 5, and 6, respectively; EX5_ET: extratelencephalic projection neurons in layer 5; EX5/6_NP: near projection neurons in layers 5 and 6; EX6_CT: corticothalamic projection neurons in layer 6; EX6b: excitatory neurons in layer 6b; IN_Pvalb, IN_Sncg, IN_Sst, and IN_Vip: inhibitory neurons subtypes exclusively expressing Pvalb, Sncg, Sst, and Vip, respectively; MICRO: microglia; MOL: mature oligodendrocyte; OPC: oligodendrocyte precursor cell; PVM: perivascular macrophage; VLMC: vascular leptomeningeal cell. (C-D) Comparison of CLOCK/Clock+ cells percentage in each major cell type among HU, WT, and previous work on human neocortical development⁵⁷ based on snRNA-seq data at (C) P07 mice and 12 years human and (D) P56 mice and 28 years human. (E-H) The number (bar plot) and Log₂ fold change (violin plot) of differentially expressed genes (DEGs) in each major cell type, and hypergeometric tests between DEGs from each pair of cell types (heatmap) at (E) P07 HU vs. WT comparison, (F) P07 KO vs. WT comparison, (G) P56 HU vs. WT comparison, and (H) P56 KO vs. WT comparison. The solid line in the violin plots indicates the median, while the two dashed lines mark the 25 and 75 percentiles. The color of each cell represents the −Log₁₀ of FDR corrected p value (q value). The number in each cell is the number of overlapped DEGs between two cell types. (I) Heatmap of gene ontology (GO) analysis to show enrichment of GO terms of the DEGs in EX2/3_IT. The color of each cell represents −Log₁₀ q value of hypergeometric tests. GO terms are grouped based on their functions (left strip) and GO categories (right strip). HW_UP, human CLOCK upregulated genes; HW_DOWN, human CLOCK downregulated genes; KO_UP, Clock KO upregulated genes; KO_DOWN, Clock KO downregulated genes.

The fraction of cells that express human CLOCK (CLOCK+ cells) in each major cell type are greater than the fraction of cells that express mouse Clock (i.e. in WT) (Fig. 4C,D). The results at P56 are consistent with our findings in IHC experiments (Fig. 1D–G). The fractions of CLOCK+ cells in HU mice are comparable with previous single-cell RNA-seq data from human neocortical development. At both P07 and P56, the percentage of CLOCK+ cells in human is higher than in WT mice, while HU mice show an increased percentage of CLOCK+ cells among all major cell types as well as cell subtypes except for microglia⁵⁷ (Fig. 4C,D; Extended Data Fig. 3A,B). Therefore, the overall expression pattern of CLOCK in HU mice is more similar to the human brain than WT mice. We also found that normalized expression of human CLOCK is greater than mouse Clock in all cell types at both P07 and P56 (Extended Data Fig. 3C–F), similar to the comparison between adult human and mouse across major cell types in cortex²⁵. These results further confirm that HU mice mimic the expression percentage and levels of CLOCK observed in human cortex. From P07 to P56, the fraction of CLOCK+ cells across the major cell types increased in both HU and WT mice (Fig. 4C,D). This result suggests that CLOCK might play different roles at different developmental stages. At both P07 and P56, the percentage of CLOCK+ cells varied across cell types, and neurons, especially excitatory neurons, showed the greatest percentage of both human and mouse CLOCK+ cells (Fig. 4C,D). Thus, CLOCK might have a major impact on neuronal function.

CLOCK regulates cell type-specific gene expression

To identify the genes regulated by CLOCK, we did pairwise (i.e., HU vs. WT and WT vs. KO) comparisons within each major cell type at P07 and P56 separately. To minimize the inflated type I error in the snRNA-Seq analysis, we used MAST with a mixed linear model to determine the differentially expressed genes (DEGs)⁵⁸. The HU vs. WT comparison detected genes that are specifically regulated by human CLOCK, while the WT vs. KO comparison determines genes that are regulated by mouse CLOCK. We overlapped the DEGs of each cell type at each age, and found that none of the cell types, except for excitatory and inhibitory neurons in HU vs. WT at P56 showed significantly overlapped DEGs (Fig. 4E–H; Supplementary Table 3–6). These results suggest that CLOCK regulates gene expression in a cell type-specific manner. To determine the cell type that is most affected by CLOCK, we considered both number and Log₂ fold change of DEGs (Fig. 4E–H; Extended Data Fig. 4A–D). At P07, CLOCK primarily regulates genes in excitatory neurons due to the dominant number of DEGs (Fig. 4E,F). At P56, CLOCK has broader effects on gene expression across cell types, although excitatory neuron is still one of the most affected cell types (Fig. 4G,H). These results, together with our finding that excitatory neurons are the cells with the greatest percentage of CLOCK+ cells (Fig. 4C,D), suggest that CLOCK has a significant impact on gene expression in excitatory neurons.

Excitatory neurons can be divided into subtypes based on their projection patterns. To further narrow down to the subtype of excitatory neuron that is most affected by CLOCK expression, we identified DEGs for subtypes. Projection neurons subtypes have distinct DEGs, while intratelencephalically projecting excitatory neurons (IT) from different cortical layers have significantly overlapping DEGs (Extended Data Fig. 4A–D; Supplementary Table 7–10). At both P07 and P56, the Log₂ fold changes of DEGs are similar across the subtypes of excitatory neurons, while the greatest number of DEGs is in IT neurons, especially in layer 2–3 IT neurons (EX2/3_IT) (Extended Data Fig. 4A–D). Therefore, we focused on EX2/3_IT in the following analyses.

Age-specific function of human and mouse CLOCK/Clock

To investigate whether human and mouse CLOCK regulate different genes in EX2/3_IT, we distinguished up- and down-regulated DEGs, and then overlapped human and mouse CLOCK-regulated genes within age groups (Extended Data Fig. 4E,F). Since HU mice are on the genetic background of mouse CLOCK KO, the non-significant overlaps between HU-WT and WT-KO comparisons suggest that the expression of genes regulated by mouse CLOCK are rescued by human CLOCK, consistent with our observation of rescued behaviors (Supplementary Fig. 2A–G). In addition, this lack of overlap suggests that overexpression of human CLOCK not only regulates and can rescue targets of mouse CLOCK but also can regulate human CLOCK-specific genes. We also overlapped DEGs from the two time points, P07 and P56 (Extended Data Fig. 4G,H). We found few overlapping genes, suggesting that both human and mouse CLOCK regulate genes in an age-specific manner.

To understand the function of the DEGs, we carried out Gene Ontology (GO) enrichment analysis on genes regulated by CLOCK in EX2/3_IT. Human CLOCK upregulated genes at both P07 and P56 are enriched for neurodevelopmental functions such as dendritic growth (e.g., neuron projection development) and spine formation (e.g., synapse organization) (Fig. 4I; Supplementary Table 11), suggesting these are enhanced in HU mice. Consistent with the greater number of DEGs for HU at P07, the number of enriched GO terms for HU DEGs at P07 are also greater than those enriched at P56 (Fig. 4I), suggesting a more important role for CLOCK in early postnatal development. In contrast, mouse KO led to enrichment of downregulated genes for only a few neurodevelopmental functions at P56 (Fig. 4I). In addition, DEGs of mouse KO at P56 but not P07 are enriched for circadian and sleep signal pathways (Fig. 4I). These results are consistent with a previous finding that showed that mouse visual cortex lacks rhythmic expression of circadian genes before eyes open at P12³².

CLOCK regulates genes with human-specific open chromatin

To investigate whether human CLOCK regulates human-specific expressed genes, we utilized a dataset that compared chromatin state from single-nucleus ATAC-seq in adult postmortem human cortex compared to chimpanzee and rhesus macaque²⁹. We subset the dataset to include those genes with human-specific open patterns of chromatin state that also contained a CLOCK:BMAL1 binding motif and overlapped them with DEGs from HU mice in EX2/3_IT. We found a significant overlap (Extended Data Fig. 4I), and 13 out of the 17 overlapping genes showed chromatin states consistent with their regulatory directions in HU mice (Supplementary Table 12).

Some of these human-specific expressed genes are specialized for neurodevelopmental functions. For example, Tenm2, which is upregulated with the greatest fold change in most major cell types of P07 HU mice, enables cell adhesion molecule binding and facilitates axon growth^59,60. Sorcs2, which is upregulated in P56 HU mice, responds to the neurotrophic factor BDNF to maintain neurite growth and spine formation^61,62. Finally, Srgap2, which is downregulated in P07 HU mice, inhibits dendrite branching and synaptogenesis¹⁶. Human-specific duplication of SRGAP2C suppresses activity of SRGAP2, resulting in complex dendrites and a greater density of spines in human^22,63.

We then validated the expression of TENM2 and SORCS2 proteins in the frontal cortex of P07 and P56 mice, respectively. We quantified the fluorescent intensity of antibody staining in IHC, and we found that HU mice showed significantly greater expression of both proteins compared to WT mice (Extended Data Fig. 4J–N). We examined the regulatory regions of both genes and detected CLOCK:BMAL1 binding motifs in the enhancer regions of both human and mouse genomes⁶⁴ (Supplementary Table 13). Together, these results confirmed the snRNA-seq data and suggest that TENM2 and SORCS2 are direct downstream targets of CLOCK.

CLOCK regulates dendrite complexity and spine density

We next assessed potential neuronal morphological changes that are suggested by the snRNA-seq DEGs related to dendrites and spines. We sparsely transduced cortical neurons with a recombinant adeno-associated virus (rAAV) expressing mCherry through unilateral intracerebroventricular injection in P0-P1 pups to visualize cell morphology including the soma, neurite arborization, and synaptic spines. At P18 and P56, we performed IHC to amplify the mCherry signal and co-stain with cortical layer and cell type markers (Extended Data Fig. 5A,B). We analyzed the morphology of the excitatory neurons in layers 2–4 of the frontal cortex. We found no difference in soma size across genotypes at both ages (Extended Data Fig. 5C,G). We then quantified the complexity of dendritic trees in P56 mice. Compared with WT mice, HU mice have a greater number of branches and longer total dendritic length, while KO mice did not show a significant difference in either parameter (Fig. 5A–C). Sholl analysis further confirmed that the concentric rings encountered more intersections with the neuron branches in HU mice and fewer intersections in KO mice compared to WT mice (Fig. 5D). This is consistent with previous studies where neurons in brain tissue of epilepsy patients that expressed less CLOCK showed less complexity of dendritic arborization³³. To summarize, our results suggest that CLOCK facilitates dendritic complexity of excitatory neurons in layers 2–4 of frontal cortex.

Fig. 5: CLOCK increases dendrite complexity and spine density in adult mice.

(A) Representative layer 2–4 excitatory neurons of frontal cortex in each genotype. (B-C) Quantification of (B) number and (C) total length of branch in each genotype. For panels B and C, each data point is a neuron. We sampled 3 adult mice per genotype and 5–9 neurons per mouse (total number of neurons: HU: n = 19; WT: n = 26; KO: n = 15). We applied GLMM with genotype and sex as fixed factors, neuron as random factor nested with individual, and Tukey’s test for post-hoc analysis. (D) Sholl analysis to quantify encountered intersections between neuron branches and concentric rings from soma as measures of dendrite complexity. The data are from the same samples in panels B and C. We did repeated measures ANOVA with genotype and sex as between-subject factors, distance to soma as within-subject factor, and Tukey’s test for post-hoc analysis. (E) Zoomed in segments to quantify density and morphology of spines. (F) Cumulative probability distribution of spine density. Each data point is a segment, we sampled 2–4 segments from 3–8 neurons per mouse. We applied GLMM with genotype and sex as fixed factors, segment as random factor nested with neuron which nested with individual, and Tukey’s test for post-hoc analysis. The open circles represent female samples, while closed circles represents male samples. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

We next quantified the density and morphology of spines in P56 mice. We measured the spine length in all spine types and the area of the spine head and width of the neck in thin and mushroom spines. Compared with WT mice, HU mice showed an increase in the density of spines, whereas KO mice instead demonstrated a decreased spine density (Fig. 5E,F). The results from KO mice are consistent with previous findings in Emx-Cre Clock KO mice and Clock knockdown in primary neuronal cultures³³. A comparison of spine morphology detected significantly enlarged head areas and elongated spines but not neck widths in HU mice (Extended Data Fig. 5D–F). Thus, these results indicate that human CLOCK affects spine density as well as spine morphology.

To determine whether the more complex dendritic trees and greater spine density results from an increased generation or decreased pruning of dendrites and spines, we also measured the spine properties at P18, which is before the peak of synaptogenesis at P21^65,66. We found that results of dendritic arborization and spine density are similar to what we observed at P56 (Extended Data Fig. 5H–K), while spine morphology did not show difference across genotypes at this age (Extended Data Fig. 5L–N). Thus, CLOCK is more likely to drive neuronal projection growth and spine formation rather than neurite elimination and synaptic pruning. In summary, these results are consistent with our snRNA-seq DEGs and demonstrate that CLOCK facilitates neurite growth and spine formation of cortical excitatory neurons.

Humanized mice show greater frequency of spontaneous EPSCs

To determine whether the increased complexity of dendrites and spine density result in an elevated number of neural connections in HU mice, we characterized the electrophysiological properties of layer 2–4 excitatory neurons in the somatosensory cortex of 3–4-week-old mice. We found that the firing versus current injection curves were similar across genotypes (Fig. 6A), suggesting that the excitatory neurons do not have differences in intrinsic excitability. The subthreshold membrane properties (i.e., resting membrane potential (RMV), input resistance, and normalized conductance) also did not show any differences (Fig. 6B–D). These results demonstrate that CLOCK does not affect the intrinsic properties of excitatory neurons. Similar results were found in inhibitory parvalbumin neurons of the visual cortex that had Clock deleted³².

Fig. 6: CLOCK alters EPSCs without affecting the intrinsic properties of layer 2–4 excitatory neurons.

(A) The number of action potential (AP) versus current amplitude curves indicates that CLOCK does not affect intrinsic excitability. (B-D) CLOCK does not alter (B) resting membrane voltage (RMV), (C) input resistance, and (D) conductance, indicating unchanged membrane properties. (E) Representative raw traces of spontaneous EPSCs were recorded from the primary somatosensory cortex of layer 2–4 excitatory neurons. (F-H) Comparison of (F) frequency, (G) amplitude, and (H) frequency*amplitude of spontaneous EPSCs for each genotype. For panels A-H, we sampled 31 neurons from 5 HU mice, 28 neurons from 4 WT mice, and 30 neurons from 4 KO mice. All mice are 3–4 weeks old. For panels A, C, and D, we performed GLMM with genotype, sex, and x-axis as fixed factors, neuron as a random factor nested with individual, and Tukey’s test for post-hoc analysis. For panels B-D and F-H, each data point is a neuron, we performed GLMM with genotype and sex as fixed factors, neuron as a random factor nested with individual, and Tukey’s test for post-hoc analysis. (I) The amplitude of evoked EPSCs in response to the stimulus with a fold change of threshold intensity. We sampled 22 neurons from 4 HU mice, 13 neurons from 3 WT mice, and 21 neurons from 4 KO mice. We applied GLMM with genotype, sex, and stimulus intensity as fixed factors, neuron as a random factor nested with individual, and Tukey’s test for post-hoc analysis. The open circles represent female samples, while closed circles represents male samples. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

We next measured spontaneous and evoked excitatory postsynaptic currents (EPSCs) to determine if CLOCK changes the overall neuronal network. Neurons in HU mice showed an increased frequency but decreased amplitude of spontaneous EPSCs (Fig. 6E–G). The decreased amplitude could be attributed to a compensatory mechanism to the increased frequency⁶⁷. Thus, we calculated the total charge transfer (frequency × amplitude) to represent the overall effect and found that it is greater in neurons from HU mice compared to WT mice (Fig. 6H). This result might be linked to Nrxn2, a gene that is associated with spontaneous EPSC frequency^68,69, and one of the DEGs that is upregulated in HU mice. We then measured evoked EPSCs from upper-layer excitatory cortical neurons by stimulating an extracellular electrode placed in the same cortical layer. We found that the evoked EPSC amplitude decreased in KO mice, while there was no difference between HU and WT mice (Fig. 6I). This finding is consistent with the downregulation of Snap25 in KO mice, which is one of the plasma membrane proteins for synaptic vesicle fusion in evoked EPSCs⁷⁰. In summary, our results indicate that the increased frequency of spontaneous EPSCs in HU mice was neither because of altered intrinsic neuronal excitability nor likely due to greater vesicle release as evidenced by no significant change in evoked EPSCs. Therefore, the most plausible explanation is enhanced abundance of neural connectivity in HU mice, which is consistent with our results of enhanced dendrite complexity and increased spine density from snRNA-seq and neuroanatomical measures.

TENM2 rescues human iPSC-derived neurons with CLOCK knockout

We next investigated whether human CLOCK exerts similar functions in human neurons. We applied CRISPR-Cas9 to delete 1–17 bp of nucleotides at Exon 7 of CLOCK to induce a frameshift in iPSC clones (Supplementary Fig. 6A). We selected clone #2 and #7 (KO#2 and KO#7), which represented two types of altered reading frames with the longest deletions. The results of qRT-PCR and western blot confirmed that the core-circadian downstream targets were altered in expression and CLOCK was not expressed in the KO clones (Supplementary Fig. 6B,C), which suggested a successful knockout. We then followed an NGN2-overexpression protocol to rapidly differentiate the iPSCs into monolayers of excitatory cortical neurons which is consistent with previous work^71,72 (Supplementary Fig. 6D). We did immunocytochemistry (ICC) on Day 28 of differentiation for TUJ1 and MAP2, which indicated the successful differentiation process of iPSCs to mature neurons (Fig. 7A and Supplementary Fig. 6E–G), while CAMKII staining indicated that these were excitatory neurons (Supplementary Fig. 6E,F). Positive staining of CUX1 but not FOXP2, suggests that the differentiated neurons are more similar to the properties of upper-layer neurons (layers 2–4) rather than deep layer (layer 6) cortical neurons (Supplementary Fig. 6G). We quantified the length and segment of cell projections expressing TUJ1 and the density of puncta expressing VGLUT2 to analyze the dendritic complexity and spine density of the iPSCs-induced neurons respectively. We discovered that CLOCK KO neurons have decreased dendritic lengths (Fig. 7A,B and Extended Data Fig. 6A), number of dendritic segments (Fig. 7A,C and Extended Data Fig. 6B), and density of puncta (Fig. 7D,E and Extended Data Fig. 6C) in different KO lines at several different differentiation time points. These results are consistent with our findings in mouse excitatory neurons (Fig. 5) and suggest that CLOCK is functionally involved in the development of dendrites and spines in the human excitatory neurons.

Fig. 7: CLOCK KO reduces mature neuronal features in iPSC-derived neurons.

(A) ICC images of iPSC-derived neurons at Day 28 with nuclei marker DAPI (blue), neuronal dendrite mark TUJ1 (green), and presynaptic marker VGLUT2 (red). (B) Total length of dendrites, which was quantified from TUJ1 staining, normalized by the number of nuclei overlapping with TUJ1 signals. (C) Number of TUJ1 segments normalized by the number of nuclei overlapped with TUJ1 signals as a quantification of dendrite complexity. (D) Zoomed in segments to quantify number of VGLUT2 puncta. (E) Number of VGLUT2 puncta overlapped with TUJ1 signal and normalized by length of dendrites as a quantification of presynaptic spine density. For panels B, C, and E, each data point is an independent culture of neurons on a circle glass cover slip sitting on the bottom of a well in a 24-well plate. We did GLM with genotype/treatment as fixed factors with Tukey’s test for post-hoc analysis. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

We then tested whether the phenotypes of CLOCK KO in the human iPSCs-induced neurons could be rescued by overexpressing either of the two validated downstream targets of CLOCK (TENM2 and SORCS2). We leveraged CRISPR/Cas9 to target the promoter region of TENM2 and SORCS2 for transcriptional activation⁷³(Supplementary Fig. 6H,I). qRT-PCR showed that CLOCK KO neurons decreased expression of both genes in the induced neurons, while CRISPR-mediated activation resulted in similar expression levels of both genes as WT (Supplementary Fig. 6J). We then quantified ICC results on Day 21 for dendritic complexity and puncta density. The results of CLOCK KO and WT neurons with empty plasmid (Extended Data Fig. 7A–C) replicated the results of untreated neurons of Day 28 for dendritic arborization (Fig. 7A–C). The KO neurons with TENM2 sgRNA were partially rescued, showing a more complex arborization than KO neurons (Extended Data Fig. 7A–C). However, TENM2 overexpression failed to rescue the decreased puncta in KO neurons (Extended Data Fig. 7A,D), while SORCS2 upregulation failed to rescue either phenotype (Extended Data Fig. 7). The monolayer iPSCs-derived neurons mainly capture the early developmental stage⁷⁴. This could explain why TENM2, which is regulated by CLOCK at P7 in mice, could partially rescue dendritic arborization but SORCS2, a gene that is regulated by CLOCK in adult mice, could not rescue dendritic arborization.

Discussion

In this study, we generated a humanized mouse model to mimic the neocortical expression of human CLOCK. We demonstrate a robust function for human CLOCK in directing brain development and function. We find that human CLOCK regulates genes with human-specific expression to enhance dendritic growth and spine formation in EX2/3_IT. Moreover, the increased dendritic branches and synaptic spines in EX2/3_IT neurons likely facilitate the functional connections that we observed as a greater frequency of spontaneous EPSCs. Together, these genomic and physiological alterations downstream of human CLOCK present as the likely mechanism for the improved cognitive flexibility of HU mice.

It was unclear from previous studies, whether the observed increase in overall expression of CLOCK in the human cortex^7,24,27 could be attributed to higher cellular expression of CLOCK and/or a greater fraction of CLOCK+ cells. Using HU mice, we found that human CLOCK showed an increase in both measures of expression compared to mouse Clock. However, MS mice, which have increased cellular expression of mouse CLOCK but with similar percentages of CLOCK+ cell as WT mice, showed similar performance in reversal learning and cortical cell density as WT mice. These comparisons suggest that the possible cell-type specific expression of CLOCK rather than relative level of expression might determine the function of CLOCK in cortical excitatory neurons. There are more than twice as many CLOCK+ excitatory neurons in HU compared to WT at P07 (Fig. 4C), which might be a mechanism driving gained cortical functions.

CLOCK also exhibited a strong temporal effect on neuronal functions. We found that mouse CLOCK regulated genes associated with circadian rhythms and some neurodevelopmental functions at P56, but we observed very few DEGs downstream of mouse CLOCK alteration at P07 (Fig. 4F,H,I). In contrast, human CLOCK regulated 192 genes in excitatory neurons at P07 (Fig. 4E). Moreover, the results of GO analysis of DEGs from EX2/3_IT highlighted dendrite growth and spine formation (Fig. 4I), which is consistent with the developmental events around P07^66,75. Caution is needed to interpret the downstream DEGs as regulated by CLOCK through either direct binding to their regulatory regions or indirect pathways. Further experiments, such as ChIP-seq or CUT&RUN at single cell resolution could determine the direct targets of CLOCK in each cell type.

Although the functions of human CLOCK discovered in the current study could be attributed to its change in spatiotemporal expression, our study could not exclude the possibility that human CLOCK evolved novel molecular functions. Compared to human, CLOCK proteins are 98.7% and 96% identical in chimpanzee and mouse respectively^76–78. Since the main functional binding domains (i.e. bHLH and PAS) in the N-terminal of CLOCK are highly conserved, a major change in function is not expected. Species coding differences mainly occur in the C-terminal PolyQ region of CLOCK, and the function of this region is still controversial^79,80. A lack of polymorphisms in the human CLOCK PolyQ region suggests purifying selection⁸¹, but its function with respect to human brain evolution still needs elucidating.

The hallmarks of human brain evolution are enhanced cognitive flexibility, greater complexity of dendrites, and elevated density of spines compared to other non-human primates^{3–5,82–84}. Increased complexity of neural connectivity has also been linked to the altered computational properties of the human brain and might result in enhanced cognitive abilities^22,82,85, while alteration of neural connectivity could result in loss of cognitive flexibility and neuropsychiatric disorders such as ASD, ADHD, and schizophrenia⁸⁶. Our snRNA-seq data, dendritic morphology measures, and electrophysiology data together suggest that human CLOCK enhances abundance of neural connectivity in the frontal cortex of HU mice. Previous studies demonstrated that the complexity of dendrites and density of spines are positively correlated with rodents’ performance in cognitive tests^87,88. Therefore, the observed enhancement of cognitive flexibility in HU mice might result from the increased number of neural connections.

In addition to the increased abundance of neural connectivity of excitatory neurons, we also observed an increased neuron density in HU mice (Fig. 3C). The more complex neural connections of the human neocortex are also predicated on an increased abundance of neurons^89,90. These data suggest that human CLOCK might regulate the proliferation of neural progenitors during embryonic development. Dividing cells have rhythmic expression of circadian proteins⁹¹. Some cell cycle genes (e.g. c-Myc and Wee1) are under the control of circadian genes such as Cry1, Cry2, Per1, and Clock^37,92,93. The greatest CLOCK expression levels are detected in progenitor-like cells in early embryonic development⁵². Thus, future studies could investigate whether human CLOCK affects the proliferation of neural progenitors at embryonic stages.

Pleiotropy, which is defined as a single gene that controls multiple phenotypes, plays an important role during evolutionary processes⁹⁴. In this study, differential gene expression in circadian-related genes was only detected in KO mice at P56, not HU mice. These results suggest that human CLOCK might regulate neural functions independently of circadian related pathways. Extra-circadian functions have also been suggested for CLOCK and other core circadian genes such as BMAL1^95,96. Clock and Bmal1 mutant mice but not Cry1 or Cry2 mutant mice have deficits in place learning^97,98. Similarly, astrogliosis was found in the cortex of aged Npas2 and Clock double KO and Bmal1 KO mice, while Per1 and Per2 double KO mice did not show this phenotype⁹⁹. Per1, Per2, Cry1 and Cry2 are downstream targets of CLOCK and BMAL1 in the circadian signal pathway⁵⁰. These results suggest that CLOCK regulates place learning and astrogliosis through a non-circadian pathway. Additionally, the extra-circadian function of CLOCK has also been suggested in tissue outside of brain such as liver and skeleton muscle^37,38. However, it is important to point out that the arrhythmic expression and the possible extra-circadian function of CLOCK does not discount the role of CLOCK in circadian function, since it has been demonstrated that circadian expression of CLOCK is not required for circadian rhythmicity^100,101. In short, a gain of function through altered gene expression pleiotropy might be the evolutionary mechanism underlying CLOCK-directed human brain specialization.

There are a number of limitations to studying human CLOCK in a humanized mouse BAC model. Human-specific changes, which are those traits that evolved after divergence from the common ancestor of chimpanzee and humans, are of great interest for understanding evolution^14,102. Given the phylogenetic distance between mouse and human, however, we should be cautious that a humanized mouse model would inevitably capture primate-specific, ape-specific, and human-specific alterations. This could be addressed by generating a non-human primate model system or studying some of the cellular and molecular mechanisms by using human and chimpanzee iPSCs systems. The human BAC insertion location in hc7 is about 50,000bp away from the 3’UTR of the Grm8 gene (Supplementary Fig. 1E). Although Grm8 was downregulated in the deep-layer excitatory neurons of HU mice (Supplementary Table 7 and 9), previous work on Grm8/GRM8 knockout mice and human neurons did not detect apparent deficits in cognitive behaviors, axonal growth, and spine formation^103–105. Thus, the minor disturbance on the expression of Grm8 gene should not undermine our main conclusions. Additionally, hc3 mice with its BAC inserted about 10,000bp away from the 5’UTR of the Ralyl gene (Supplementary Fig. 1G) showed similar enhancement of cognitive flexibility in the set-shifting test (Fig. 2E; Supplementary Fig. 2K). This result further confirms that the neural function of human CLOCK is independent of the BAC inserted locations. The other limitation of this study is that the human CLOCK gene was expressed in the background of the mouse genome, thus the mechanism we uncovered in humanized CLOCK mice might not be the same as in vivo human specializations due to the mouse background. However, the humanized mouse model is currently the most effective way to directly study function of human genes on cognition and with fruitful results^18,22,106. We include the important control of the mouse CLOCK BAC to demonstrate that it is the pattern of overexpression driven by the human regulatory elements rather than overexpression alone that leads to the observed phenotypes in the humanized mice. To confirm the main results of our mouse model, we replicated the findings of dendritic arborization and spine density in human iPSC-derived neurons (Fig. 7). Additionally, the genes that are regulated by human CLOCK in mouse neurons significantly overlapped with human-specific differentially accessible regions²⁹ (Extended Data Fig. 4I). This suggests that human CLOCK could regulate human-specific genes in the mouse neurons. Therefore, our findings of gene expression, neuroanatomy, and cognition of HU mice successfully captured some human-specific alterations.

In conclusion, we found that the spatiotemporal alteration of human CLOCK expression in a mammalian model system might result in human-specific expression of genes that enhance dendritic growth and spine formation of excitatory neurons. The observed increased density of neurons and complexity of dendrites and spines form more abundance of functional connections in the frontal cortex. These changes may underlie the advanced cognitive flexibility of the humanized mice. Our study suggests that gain of cortical functions of CLOCK via altered spatiotemporal gene expression might be an important molecular mechanism for human brain evolution.

Methods

Experimental Model

Mouse

All mice in this study were kept on a C57BL/6J genetic background and housed in 12hr light/12hr dark cycle (LD) with room temperature around 22°C and humidity around 55%, lights were on at 7 am Central Standard Time. All mice were raised at the University of Texas Southwestern Medical Center (UTSW). We complied with guidelines of the UTSW Animal Care and Use Committees and followed procedures approved by the UTSW Institutional Animal Care and Use Committee (IACUCC# 2015-100918) for animal care and experiments. We weaned pups around 4 weeks of age and housed mice with same-sex littermates for experiments.

Human and non-human primates

We obtained human posterior cingulate cortex from UCLA VA Bank and UMiami Brain Endowment Bank. Tissues of the same brain region were collected from chimpanzees and macaques at Yerkes National Primate Research Center. All tissues are deidentified postmortem tissues that are exempt from IRB oversight at UTSW under STU 072016-025. These tissues are the same as our previous work²⁹. Tissues were embedded in Tissue-Tek CRYO-OCT Compound (Thermo Fisher Scientific, #14-373-65) and cryosectioned onto slides at 13 μm.

Cell culture

We conducted stem cell work under the oversight and approval of the UT Southwestern Stem Cell Research Oversight (SCRO) Committee (Registration #8). All cultured cells were kept at 37°C, 5% CO₂, and above 90% humidity.

HEK293T cells

HEK293T cells (ATCC, #CRL-3216) were cultured in DMEM (Cytiva HyClone^™, #SH3024301) with 1% Antibiotic-Antimycotic (Thermo Fisher Scientific, #15240062) and 10% Fetal Bovine Serum (FBS) (Thermo Fisher Scientific, #10437028).

Human iPSC culture

We obtained a human iPSC line (WTC-11, #GM25256) from the Gladstone Institute. We cultured the cells following the Technical Manual (Stemcell Technologies, Document #28315). Cells were maintained in six-well plates (Corning, #4936) with mTeSR^™1 medium (Stemcell Technologies, #85870). The plates were coated with growth factor-reduced Matrigel matrix hESC-qualified (BD Biosciences, #354277). We passaged cells when the confluency of culture reached about 85% by using Gentle Cell Dissociation Medium (Stemcell Technologies, #07174) and ROCK inhibitor (Fisher Scientific, #50-863-6). We changed mTeSR^™1 medium every day to maintain the cell line.

Rat primary astrocytes

Rat primary cortical astrocytes (Invitrogen, #N7745-100) were cultured according to Invitrogen protocols (MAN0001679 for thawing and establishing culture; MAN0001680 for expanding culture). We cultured astrocytes in Matrigel coated 6-well plate with DMEM and 15% FBS medium. Medium was half changed every 5 days. We passaged astrocyte cultures when they reached 100% confluency using 1:1 1X PBS (Cytiva HyClone^™, #SH3025601) diluted Accutase (Gibco^™, #A1110501).

SW102 cells

SW102 cell line was first generated by Copeland Lab at National Cancer Institute¹⁰⁷. This cell line was obtained from National Cancer Institute recombineering resource for BAC preparation.

Generation of humanized CLOCK mice

Recombineering human CLOCK BAC

We obtained a BAC (RP11-1029A14) that contained the full-length human CLOCK gene from BACPAC Resources (Children’s Hospital Oakland Research Institute). This BAC contains the full-length TMEM165 and part of PDCL2 genes flanking on the 3’ and 5’ end of CLOCK gene respectively (Fig. 1A). We first removed TMEM165 and PDCL2, and then we added a 3XFLAG-3XHA N-terminal tag through recombineering. The following primers were used to direct the recombination events (the underlined sequences are what binds the cassette, while the sequences before that are the 50bp homology arm overhangs):

5’ TMEM165 trim

Cassette Forward: 5’ CAAACAGGTATGA GGCATAATTCTGCTACTTACTAGC TGTGTAATTTGAGCCTGTTGACAATTAATCATCGGCA 3’

Cassette Reverse: 5’ ATTGGCCTATCAA TACTGATAATTTTAT ATGACACACAAGGAAAAACAAGTCAGCACTGTCCTG CTCCTT 3’

Oligo Forward: 5’ CAAACAGGTATGAGGCATAATTCTGCTACTTA CTAGCTGTGTAATTTGAGCTTGTTTTTCCTTGTGTGTCATATAAAATTATCAGTATTGATAGGCCAAT 3’

Oligo Reverse: 5’ ATTGGCCTATCAATACTGAT AATTTTATATGACACACAAGGAAAAACAAGCTCAAATTACACAGCTAGTAAGTAGCAGAATTATGCCTCATACCTGTTTG 3’

3’ PDCL2 trim

Cassette Forward: 5’ TTTCTGATCCCTAATATAAAAGCAGTCATAAGTAAAACA TTCAGTACACACCTGTTGACAATTAATCATCGGCA 3’

Cassette Reverse: 5’ ATTCTGCTACATGCTACAACATAGTCGAACCTTGAAAA CATGCTAAGCGATCAGCACTGTCCTGCTCCTT 3’

Oligo Forward: 5’ TTTCTGATCCCTAATATAAAAGCAGTCATAAGTAAAACATT CAGTACACATCGCTTAGCATGTTTTCAAGGTTCGACTATGTTGTAGCATGTAGCAGAAT 3’

Oligo Reverse: 5’ ATTCTGCTACATGCTACAACATAGTCGAACCTTGAAAACAT GCTAAGCGATGTGTACTGAATGTTTTACTTATGACTGCTTTTATATTAGGGATCAGAAA 3’

N-terminal 3XFLAG-3XHA tag

Cassette Forward: 5’ TAAGGAGAAGTACAAATGTCTACTACAAGACGAAAACG TAGTATGTTATGCCTGTTGACAATTAATCATCGGCA 3’

Cassette Reverse: 5’ ACATACCTGTCAACAATCGAGCTCATTTTACTACAGCTT ACGGTAAACAATCAGCACTGTCCTGCTCCTT 3’

Oligo Forward: 5’ TAAGGAGAAGTACAAATGTCTACTACAAGACGAAAACGTA GTATGTTATGGACTACAAAGACCATGACGG 3’

Oligo Reverse: 5’ ACATACCTGTCAACAATCGAGCTCAT 3’

Electroporation

We inoculated SW102 cells into 5ml LB Broth (Sigma, #L3022) with 25ug/ml chloramphenicol (Sigma, #C0378) and grew cells at 32°C overnight. We took 4mL of the saturated SW102 culture to inoculate into 120ml LB Broth with 25ug/ml chloramphenicol and grew the culture on a 32°C shaker to an OD600 between 0.2–0.3 for 2 hours. We placed the SW102 culture in a 42°C water bath for 15 minutes and then transferred to 50mL conical tubes and chilled on ice. The cultures were centrifuged at 4°C at 2000 g for 10 minutes to pellet cells. The cell pellet was resuspended in 25mL ice-cold 10% glycerol (VWR, #EM-4750). These centrifuge-resuspension steps were repeated for two more times with ice-cold 10% glycerol. After decanting the supernatant of the last wash, the recombineered BAC DNA was electroporated into these cells using a 1mm cuvette at 1750 V, 25 uF, 200 Ohms. 1mL of LB Broth was added and the cells then recovered for 2 hours at 32°C with shaking. 1mL cultures were spun down and washed with minimal M9 salt wash. Finally, the cells were plated on minimal plates with 25ug/ml chloramphenicol and incubated at 32°C.

Cesium Chloride BAC preparation

We picked a single colony from the minimal plate and grew it in a 2ml LB Broth culture overnight at 32°C with chloramphenicol. We inoculated 2L LB Broth cultures and incubated ~24 hours at 32°C with chloramphenicol. We harvested the bacteria by centrifugation at 4,000 g for 20 minutes. The pellets were resuspended in 100ml Buffer P1 without RNase (45.5 ml 20% Glucose (Sigma, #G6152), 25ml 1M Tris pH8.0 (Fisher, #BP152-5), 20ml 0.5M EDTA (Sigma, #E-5134), ddH₂O to 1L). We added 50ml Buffer P2 (20ml 10N NaOH, 50ml 20% SDS, ddH₂O to 1L) and mixed by inversion, and then incubated at room temperature (RT) for 5 minutes. 37mL cold Buffer P3 to the lysate (600ml 5M Potassium Acetate (Sigma, #P1190), 115ml Glacial Acetic Acid (Sigma, #695092), ddH₂O to 1L) was then added, mixed immediately 4~6 times, and placed on ice for 10 minutes. Samples were centrifuged at 15,000 g for 20 minutes, then the supernatant was transferred to new tubes and 100ml isopropanol was added. We then centrifuged the sample at 15,000 g for 30 minutes, followed by resuspending the pellet in 7ml TE. We then added 1.1g CsCl per ml of TE and 300ul Ethidium Bromide (EtBr; Fisher, #BP1302). The sample was separated into phases in a VTI 80 ultracentrifuge overnight at 300,000 g at 22°C. We then pulled the EtBr-marked DNA band with a 3ml syringe. The remainder was filled with TE mixed with 1.1g CsCl per ml of TE, and spun again for 4 hours at 300,000 g at 22°C. The BAC DNA band was pulled again and EtBr was removed with repeated 3ml washes of n-Butanol saturated with ddH2O. The resulting mix was diluted with 3X volume of TE, and 1/10 volume of 3M Sodium Acetate (Sigma, #71183) was added. 2X volume of 100% ethanol was added and samples were chilled on ice for 20+ minutes. DNA was pelleted at 2,500 g for 30 minutes at 4°C. The final ultrapure BAC DNA was diluted in BAC injection buffer (10mM Tris pH 7.4, 0.25mM EDTA).

Selection of mouse line

The CsCl purified human CLOCK BAC was integrated into C57BL/6N mice by direct pronuclear injection. Since we intended to generate a mouse model to mimic overexpression of human CLOCK, founders were screened by qRT-PCR, which will be described later, to generically amplify both human CLOCK and mouse Clock. Four founders showed increased CLOCK expression in all three brain regions (i.e. frontal cortex, hippocampus, and cerebellum, Supplementary Fig. 1A). We then did an immunoprecipitation (IP) using cortex from the mice as previously described¹⁰⁸. We used primary CLOCK antibody (MilliporeSigma, #ab2203) in 1:100 dilution in each sample. We then did Western Blotting on the IP’d sample as previously described¹⁰⁹ using both primary FLAG antibody (Sigma, #F1804) in 1:400 dilution and a different primary CLOCK antibody (Abcam, #ab3517) in 1:200 dilution. The result showed that only three founders demonstrated bands representing FLAG-tagged human CLOCK at the expected size (Supplementary Fig. 1B). The three founder lines (HUf2, HUf3, and HUf7) were kept and backcrossed onto a C57BL/6J genetic background for 10 generations. In our breeding, we found that the human CLOCK BACs of HUf2 were likely inserted into the X chromosome. Thus, we used lines HUf3 and HUf7 for our experiments. Given the relative higher expression of CLOCK in frontal cortex of HUf7, we used HUf7 mice for all of our experiments and used HUf3 mice to confirm the results of critical experiments.

Genotyping of human CLOCK and mouse Clock KO

We extracted genomic DNA from an ear or tail snip. The PCR mixture contains 5 ul EmeraldAmp GT PCR Master Mix (Takara, #RR310B), 1 ul genomic DNA, 1ul 10uM primers mix and 3 ul molecular grade H₂O. We amplified a 298bp fragment to detected human CLOCK with following primers: F: 5’ GTCGTCACCCTCGTTTGAGT and R: 5’ TTGCTGCATCCTACAGTGCT; and PCR cycle settings: 5 min at 95°C; 35 cycles of 15 sec at 95C, 30 sec at 60°C, and 30 sec at 68°C; 5 min at 72°C; hold at 12°C. We detected the WT and KO of endogenous mouse Clock by using the following primer set: F_WT: 5’GGTCTATGCTTCCTGGTAACG, F_KO: 5’ATTCCCCATCCAAAGATATTTGC, and R: 5’CCAGGCTTACGCTGAGAGC; and PCR cycle settings: 3 min at 94°C; 35 cycles of 30 sec at 94°C, 30 sec at 60°C, and 90 sec at 72°C; 10 min at 72°C; hold at 4°C to amplify 280bp and 506bp fragments for WT and KO allele respectively⁴². We amplified a ~300bp fragment to detected mouse Clock BAC with following primers: F: 5’ ATACGACTCACTATAGGGCGAATTCG and R: 5’ TTAGCCTCTTCTGTGCTACATCG; and PCR cycle settings: 2 min at 96°C; 35 cycles of 15 sec at 95°C, 30 sec at 55°C, and 15 sec at 72°C; 3 min at 72°C; hold at 12°C. The PCR products were analyzed on 1% agarose gel. However, for genotyping of MS mice, the primers to detect endogenous mouse Clock could also amplify mouse Clock BAC, so MS (Clock^mus;Clock^−/−) and Clock^mus;Clock^+/− mice could not be distinguished by using above mentioned PCR tests. Therefore, we further distinguished the two genotypes by quantitatively determining the copy number of LoxP in Clock knocked out allele through quantitative real-time PCR (qRT-PCR).

Tissue preparation and qRT-PCR

Quantification of copy number of LoxP for genotyping of MS mice

In order to distinguish MS mice (Clock^mus;Clock^−/−) from Clock^mus;Clock^+/− mice, we designed the primers (F: 5’ AGTGGGTGGCACCGATAAG and R: 5’ CACCAAATGTTTGTTGAATGAAA) to amplify a 166bp fragment containing a LoxP sequence in the Clock KO allele. We extracted DNA from 0.7 cm tail snips and performed phenol-chloroform purification¹¹⁰, and then we diluted the DNA to 0.1 ng/ul. The DNA solution was further diluted 4X as the DNA template. We always ran these DNA samples with known KO, Clock^+/−, and WT samples as a reference to determine copy number of LoxP. The qRT-PCR mixture contains 5ul iTaq Universal SYBR Green Supermix (Bio-Rad Laboratories, #172-5121), 2ul DNA template from phenol-chloroform purification, 1ul 10uM primers mix and 2ul molecular grade H₂O in one well of a 384-well plate (MicroAmp^™, # 4343370), each sample was run in replicated reactions in 4 wells. We performed qRT-PCR using a Touch Real-Time PCR Detection System (Bio-Rad Laboratories, #CFX384) with the following PCR cycle settings: 2 min at 50°C; 10 min at 95°C; 39 cycles of 15 sec at 95°C, 45 sec at 65°C; 5 sec at 65°C; to 95°C and stop. The Cq value of the 4 replicates from the same sample were averaged, and we compared the Cq values of the samples with the genotype-known reference samples to determine their genotypes.

TaqMan qPCR quantification of BAC copy number

We did phenol-chloroform extraction of DNA from 0.7 cm tail snips of WT mice and then diluted them to 100 ng/ul as the genome background. We prepared the sample DNA template in the same way and concentration. Purified BAC solution was diluted to 2 ng/ul in BAC injection buffer. Then 1.684 ul of BAC solution was added to 40 ul of genome background solution. This would result in 16 BAC copies per diploid mouse genome. We further diluted the BAC copies to 8, 4, 2, 1 in the genome background solution for a standard curve. The qPCR mixture contains 5 ul TaqMan^™ Universal Master Mix II no UNG (Applied Biosystems, #4440040), 0.9 ul 10 uM primer mix (F: 5’ GCCAGCCCCTTCTTGTCA and R: 5’ ACTGCAGTCTGAGCGGTTAGTG), 0.25 ul 10 uM TaqMan probe (5’ TTAGATCCTAGCTCACGTGTC; Applied Biosystems, #5371391), 1 ul DNA template, and 2.85 ul molecular grade H₂O in one well of a 384-well plate, each sample was run in replicated reactions in 4 wells. We performed qPCR using the same platform with the following PCR cycle settings: 10 min at 95°C; 39 cycles of 15 sec at 95°C, 1 min at 60°C and stop. The Cq value of the 4 replicates from the same sample were averaged, and we generated a linear model between the Cq values and the known copy number per diploid mouse genome. Then we calculated the copy number of the samples by substituting the Cq value into the model.

RNA isolation and quantification of gene expression

Mice were rapidly decapitated and brains quickly removed. The fresh brain was quickly transferred to ice-cold coronal Acrylic Mouse Matrice - 1mm Coronal (Braintree Scientific, # BS-A-5000C) and washed with ice-cold 1X PBS. The boundary between the olfactory bulb and frontal cortex was aligned to the first dent where the first razor blade (Fisher Scientific, #12-640) was inserted to remove the olfactory bulbs. The second razor blade was inserted into the third dent. We then removed the subcortical region from this section and separated the left and right hemisphere cortical samples into different Eppendorf tubes. The sampled frontal cortex corresponds to the cortical regions from #23 to #37 sections in the mouse P56 Coronal Atlas from the Allen Brain Atlas released in 2011. We also dissected out hippocampus and cerebellum and put them in separate tubes. The tubes were flash frozen in liquid nitrogen.

We extracted the total RNA of frontal cortex, hippocampus, cerebellum, 293T cells, and iPSC-induced neurons by using the miRNeasy Mini kit (Qiagen, #217004) according to the manufacturer’s protocol. The total RNAs were depleted of genomic DNA by applying DNaseI Amplification grade (Invitrogen, #18068015), and then reverse transcribed used SuperScript III First-Strand Synthesis SuperMix (Invitrogen, #18080400) to amplify cDNA following the manufacturer’s protocol. The qRT-PCR mixture contains 5ul iTaq Universal SYBR Green Supermix, 3ul cDNA template, 0.3ul 5uM primers mix and 1.7ul molecular grade H₂O in one well of a 384-well plate, each sample was run in replicated reactions in 4 wells. We performed qRT-PCR using the same setup as LoxP copy quantification. Cq values of the 4 replicates from the same sample were averaged, and data of qRT-PCR were calculated ΔΔCq using a Touch Real-Time PCR Detection System and CFX Manager software (CFX384, Bio-Rad Laboratories, Hercules, CA). We used 18S and β-actin as the reference for normalization of expression. We designed the primers (F: 5’ AGCATGGTCCAGATTCCATC and R: 5’ CCCACAAGCTACAGGAGCAG) to amplify 146bp fragments for both mouse Clock and human CLOCK to check generic CLOCK/Clock expression in different brain regions of human CLOCK founder mice. To compare the expression of CLOCK/Clock in different brain regions of HU, WT, KO, and MS mice, we designed the primers (F: 5’ CACAAAACAGCACCCAGAGT and R: 5’ ATGACTGCCCCACAAGCTAC) to amplify 128bp fragments of human CLOCK exclusively, the primers (F: 5’ AGGAGCTGGGGTCTATGCTT and R: 5’ AGCATCTGACTGTGCAGTGG) to amplify 116bp fragments of mouse Clock exclusively, and the primers (F: 5’ GAGAAGTACAAATGTCTAC and R: 5’ CCATCAAAAATACTACTGTC) to amplify 114bp fragments of generic CLOCK/Clock. We also designed the primers (F: 5’ TGTCTCCTAGGTGCCTGCTT and R: 5’ CTCTCACGGGCTTCTCAGAC) to amplify 115bp fragments of mouse 18S, and the primers (F: 5’ CCAACCGTGAAAAGATGACC and R: 5’ CCATCACAATGCCTGTGGTA) to amplify 120bp fragments mouse β-actin as reference genes in mouse brain tissues. We then designed the primers (F: 5’ TACTGCAGCTGGAAATGTGC and R: 5’ TTGTTGCCACATCATCGTTT) to amplify 184bp fragments of TENM2 (primer1), the primers (F: 5’ GGCTTACCAGGAAACGATGA and R: 5’ CACCAAAGAGAGCGTCCTTC) to amplify 217bp fragments of human TENM2 (primer2), the primers (F: 5’ GTCCTCGCCTACACAAAGGA and R: 5’ GCAGGTGACGTACCGAAAAT) to amplify 186bp fragments of human SORCS2 (primer1), the primers (F: 5’ CTCGGTGGAGATTTTCGGTA and R: 5’ AATGCATACTTCGGCAGCTT) to amplify 212bp fragments of human SORCS2 (primer2), and the primers (F: 5’ GAGGGAGCCTGAGAAACGG and R: 5’ GTCGGGAGTGGGTAATTTGC) to amplify 68bp fragments of human 18S in HEK293T cells and human iPSCs-derived neurons. At last, we designed the primers (F: 5’ TGCCTACGGTTGTCATCTTT and R: 5’ CTCAACTCGCAATATCGCCA) to amplify 303bp fragments of human CLOCK, the primers (F: 5’ AGTAGGTCAGGGACGGAGGT and R: 5’ CAGGATTCCACATTGGAAAAG) to amplify 170bp fragments of human ARNTL, the primers (F: 5’ CTCTGCCTTCCCCATCGTC and R: 5’ TGGAACATGCTGAACTGTGC) to amplify 116bp fragments of human NPAS2, the primers (F: 5’ AACCCCGTATGTGACCAAGA and R: 5’ CCGCGTAGTGAAAATCCTCT) to amplify 138bp fragments of human PER1, the primers (F: 5’ GTCCCAGGTGGAGAGTGGT and R: 5’ AAATTCCGCGTATCCATTCA) to amplify 159bp fragments of human PER2, the primers (F: 5’ ACTTCGAGAAGCAGCTCACC and R: 5’ GCACCGGTAGTAGCAGGAC) to amplify 161bp fragments of human PER3, the primers (F: 5’ CTACATCCTGGACCCCTGGT and R: 5’ CACATCTGCTGGTTGTCCAC) to amplify 150bp fragments of human CRY1, the primers (F: 5’ GAACTCCCGCCTGTTTGTAG and R: 5’ CATCTTCATGATGGCTGCAT) to amplify 144bp fragments of human CRY2, the primers (F: 5’ CCAGTTTGAATGACCGCTCT and R: 5’ AAGCTGCCATTGGAGTTGTC) to amplify 160bp fragments of human NR1D1, the primers (F: 5’ GAGCAGGGGATCTGCTAAACT and R: 5’ GCACGAATGAGAGTTTCCTGC) to amplify 172bp fragments of human NR1D2, and the primers (F: 5’ CCCAGCTGATCTTGCCCTAT and R: 5’ CTCGTTGTTCTTGTACCGCC) to amplify 182bp fragments of human DBP in human iPSCs. All primers were designed by using Primer3¹¹¹.

Optical Genome Mapping

25 mg fresh frozen brain tissues, including neocortex, striatum, hippocampus, and cerebellum, were shipped to Bionano Genomics (San Diego, CA) for Optical Genome Mapping. DNA was isolated and labeled by using Bionano Prep SP Tissue DNA Isolation Kit and Direct Label and Stain-G2 Kit respectively. The assay was conducted with 100X of human genome (400Gbp). Raw data was run with Bionano’s de novo assembly pipeline. The assembled data was processed and visualized with Solve 3.7/Access 1.7 in reference to a custom GRCm38/mm10 mouse genome appended with BAC sequence as chromosome 22. Finally, structure variants (SV) were detected by comparing the genome of our samples with Bionano’s mouse control SV database.

Immunohistochemistry (IHC)

We anesthetized P18 and adult mice by injection of 80–100 mg/kg Euthasol. All mice were transcardially perfused with 4% ice-cold paraformaldehyde (PFA) (Sigma-Aldrich, #P6148), and we kept brains in 4% PFA in 4°C overnight for postfixing. Brains of P0 pups were kept in 4% PFA in 4°C overnight and then transferred to 4°C 30% sucrose (Sigma-Aldrich, S5016) with 0.01% sodium azide (Fisher, #S227I) until sectioning. For sagittal sections, we embedded adult brains in Tissue-Tek CRYO-OCT Compound and cryosectioned onto slides at 20 μm. For coronal sections, we embedded brains in 2% Low Melting Temperature Agarose (UltraPure^™, #15517-022) and sectioned with a Compresstome (Precisionary Instruments, #VF-300). The thickness of section depends on the experiments which are specified in corresponding parts. We stored free-floating sections in 1X PBS with 0.01% sodium azide. We used 1X TBS, which is diluted from 10X TBS (30.5g Tris base, 45g NaCl (Sigma-Aldrich, S3014), ddH₂O to 500ml final volume, pH 7.6), for all washes. We diluted primary antibodies in 3% normal donkey serum (Jackson ImmunoResearch, #017-000-121) and 1% bovine serum albumin (BSA) (Sigma-Aldrich, #A9647) in 0.4% Triton X-100 (Thermo Fisher Scientific, #BP151-100) in TBS (TBS-T). Secondary antibodies were diluted in 1% BSA in TBS-T. Instead of TBS, we used 0.1 M sodium phosphate buffer (PB) which was generated from mixture of 0.1 M sodium phosphate dibasic solution (1.42 g sodium phosphate dibasic (Sigma-Aldrich, #S7907) in 100 ml ddH₂O) and 0.1 M sodium phosphate monobasic solution (0.6 g sodium phosphate monobasic (Sigma-Aldrich, #S0751) in 50 ml ddH₂O) to pH 7.2 for IHC involved CLOCK antibody.

Lamination

We coronally sectioned adult mouse brains at 40 μm. Sections from frontal cortex, which are from #40 to #52 sections in the mouse P56 Coronal Atlas released as the Allen Brain Atlas in 2011, were used for lamination analysis. We put sections in citrate buffer (10 mM Sodium Citrate Tribasic Dihydrate (Sigma, #C8532), 0.05% Tween-20 (Fisher, #BP337), pH 6) for 10 min at 95°C for antigen retrieval. Aldehydes were quenched with 0.3 M Glycine (Sigma-Aldrich, #G8898) in TBS-T for 1h at RT. We then incubated sections in primary antibody solution overnight at 4°C. Secondary antibody incubations were performed for 1h at RT. We mounted sections onto slides (Fisherbrand^™, #12-550-15) and waited until dry. #1 Coverslips (Fisherbrand^™, #12-545-MP) were mounted using ProLong Gold Antifade Mountant with DAPI (Thermo Fisher Scientific, #P-36931). We used rabbit anti-CUX1 (1:200; Proteintech, #11733-1-AP) to label layers 2–4, mouse anti-FOXP1 (1:500; Abcam, #ab32010) to label layers 3–5, goat anti-FOXP2 (1:1000; Santa Cruz Biotechnology, #sc-21069) to label layer 6. We used species-specific secondary antibodies produced in donkey and conjugated to Alexa Fluor 488, Alexa Fluor 555, or Alexa Fluor 647 (1:2000). 20X images with fixed 13 z stacks and tile scans to capture the hemisphere of the coronal section were collected using a Zeiss confocal laser scanning microscope (LSM880) at the UT Southwestern Neuroscience Microscopy Facility. Image analysis was performed in FIJI¹¹² with self-written Macro codes. For each section, we did three measurements spanning the width of the primary somatosensory cortex for all layers, layers 2–4, and layer 6 respectively). We also measured the area of the hemisphere in which we measured layers. We divided the length of all layers to area for relative cortical thickness, and we divided the length of particular layers to cortical thickness for relative thickness of layers. All three measurements were averaged to represent the section.

Cell density

We coronally sectioned adult mouse brains from frontal cortex, which correspond to images from #27 to #36 in the mouse P56 Coronal Atlas released as the Allen Brain Atlas in 2011, into 40 μm slices for cell density analysis. Human and non-human primate sections on slides follow the same IHC protocol here. We put sections in 3% H₂O₂ (dilute in ddH₂O from 30% Hydrogen Peroxide Solution; Sigma-Aldrich, #H1009) for 1h at RT to block intrinsic peroxidase activity. We then incubated sections in blocking solution (2% normal donkey serum and 0.3% Triton X-100 in 0.1 M PB) for 1h at RT. Sections were transferred to primary antibody solution (antibody in blocking solution) for 48 hours at 4°C. Secondary antibody (antibody in 0.3% Triton X-100 in 0.1 M PB) incubations were performed overnight at 4°C. We then incubated sections in 300 nM DAPI solution in 1X PBS for 5 min at RT. Sections were mounted onto slides, and #1 coverslips were mounted using ProLong Gold Antifade Mountant (Thermo Fisher Scientific, #P-36970). We used 0.1% Triton X-100 in 0.1 M PB for all wash steps. We used rabbit anti-CLOCK (1:200; Santa Cruz Biotechnology, #sc-25361) to label CLOCK+ cells, mouse anti-NEUN (1:1000; Millipore, #MAB377) to stain neurons, and goat anti-OLIG2 (1:500; R&D Systems, #AF2418) to stain oligodendrocytes (including both mature oligodendrocyte and OPC). We used species-matched secondary antibodies produced in donkey and conjugated to Alexa Fluor 488, Alexa Fluor 555, or Alexa Fluor 647 (1:2000). A 20X objective lens with fixed 13 z stacks and tile scans was used to capture the hemisphere of the coronal section using a Zeiss LSM 880 confocal microscope at the UT Southwestern Neuroscience Microscopy Facility. We performed cell count analysis in a 1513 × 1411 pixels (623 × 587 um) field of orbitofrontal cortex (Fig. 3A) by using FIJI¹¹² with self-written Macro codes. We counted DAPI_ NEUN double positive cells as neurons and DAPI_OLIG2 double positive cells as oligodendrocytes. At last, all cells, neurons, and oligodendrocytes were colocalized with CLOCK + cells for CLOCK+ neurons and CLOCK+ oligodendrocytes respectively to calculate CLOCK + fractions.

Fluorescent intensity

P07 and P56 coronal sections of brain from similar regions as cell density analysis were used for TENM2 and SORCS2 imaging respectively. The IHC and imaging steps are similar to those in lamination. We used sheep anti-TENM2 (1:100, Invitrogen, #PA5-47638) and sheep anti-SORCS2 (1:500; R&D Systems, #AF4238) to stain the two proteins, and mouse anti-CAMKII (1:100; Abcam, #ab22609) for excitatory neurons in P07 sections. Image capture used the same settings as described above. The fluorescent intensity was quantified from excitatory neurons for TENM2. The fluorescent intensity of SORCS2 and CLOCK were quantified from 511 × 441 pixels (212 × 183 μm) and 313 × 313 pixels (100 × 100 μm) fields respectively. The readout A.U. values were normalized to the number of cells.

In vivo lateral ventricle injection, dendrite and spine morphology

We performed in vivo lateral ventricle injection with recombinant adeno-associated virus (rAAV/hSyn-mcherry, Virus Vector Core, University of North Carolina at Chapel Hill) that express mCherry to visualize dendrites and spines of neurons as previously described¹¹³. We prepared the needles for virus injection by pulling glass pipettes (World Precision Instruments, # 1B100-6) on a puller (Narishige International, #PC-10). We connected the needle to the micro syringe of Standard Infusion Only Pump 11 Elite Syringe Pumps (Harvard Apparatus, #70-4500) and fixed the needle on a stereotaxic (DAVID KOPF INSTRUMENTS, Model 940). P0-P1 mice were wrapped with a disposable rubber glove and buried in ice for 5–10 min for cryo-anesthetization. We placed the mouse on an ice-cold platform and fixed the head. We adjusted the needle to the intersection dot of transverse and sagittal suture lines and set this dot as the initial position (i.e. x=0, y=0, z=0). We then injected the rAAV in the left lateral ventricle (x=0.8mm left, y=1.5mm rostral, z=1.5mm deep) with infusion rate 3.33nl/second for a total of 50nl. The injected mouse was rubbed with its home cage bedding material gently and put back to its home cage which was placed on a heating pad. Mice were transcardially perfused as described above and the brains collected on P18 and P56.

We coronally sectioned P18 and adult mice at 100 μm. We used sections from frontal cortex, which correspond to images from #37 to #47 in the mouse P56 Coronal Atlas released as the Allen Brain Atlas in 2011, for dendrite and spine analysis. The IHC procedure was the same as the above description for cell density. Instead of 1# coverslip, we used #1.5 coverslip (Globe Scientific, #1415-15) for use with a high magnitude oil object. We used mouse anti-NEUN (1:1000; Millipore, #MAB377) to label neurons, and goat anti-OLIG2 (1:500; R&D Systems, #AF2418) to stain oligodendrocytes, chicken anti-GFAP (1:500; Abcam, #ab4674) to label astrocytes, rabbit anti-CUX1 to label layers 2–4, mouse anti-CAMKII to stain excitatory neurons, and rabbit anti-DsRed (1:500; Clontech, # 632496) and goat anti-tdTomato (1:500; LifeSpan BioSciences, #LS-C340696) to enhance the mCherry signal. We used species-specific secondary antibodies produced in donkey and conjugated to Alexa Fluor 405, Alexa Fluor 488, Alexa Fluor 555, or Alexa Fluor 647 (1:100 for Alexa Fluor 405, 1:2000 for the others). A 63X objective lens with fixed 13 z stacks and tile scans were used to capture the entire arborization of a given neuron using the airyscan function of the Zeiss LSM 880 confocal microscope at the UT Southwestern Neuroscience Microscopy Facility. We used self-written Macro codes to apply Simple Neurite Tracer^114,115 and Sholl analysis^116,117 plugins in FIJI to analyze the complexity of dendritic arborization (i.e. length of dendrites, number of branches, number of terminal, and number of intersections). Spine analysis was conducted by using self-written Macro codes and SpineJ¹¹⁸ plugin in FIJI. We measured the density and morphological traits (i.e. spine length, head area, head width, and neck width) of spines.

Immunoprecipitation (IP)

We performed IPs as we did previously¹⁰⁸. We homogenized mouse cortex in 1X RIPA buffer (250 mM Tris-HCl pH 7.4, 750 mM NaCl, 5 mM EDTA, 5mM Na₃VO₄, 2.5% sodium deoxycholate, 0.5% SDS, 5% Igepal) with 10ul/ml of 100mM PMSF, 25ul/ml of 200mM Na3VO4, and 10ul/ml Protease Inhibitor Cocktail (Sigma-Aldrich, #P8340) by using TissueLyser LT (QIAGEN, #85600). We then sonicated the homogenized tissue using a Bioruptor (Diagenode, #UCD-200) in the following procedure: 10 seconds pulses of 200 W with an interval of 20 seconds for 5 min twice. Samples were centrifuged at 13,500 g for 15 min at 4°C. The supernatant of each sample was transferred to a new tube and incubated with 100ul Dynabeads Protein G (Invitrogen, #10004D) and rabbit anti-CLOCK (1:100; MilliporeSigma, #AB2203). At last, we used 40ul 1X SDS (1mM EDTA, 0.5% SDS, 20mM Tris-Cl) to elute proteins from Dynabeads for western blot.

Western blot

In addition to using eluates from IP, we also collected protein from cultured cells. We washed cells with 1X PBS, collected in lysis buffer (0.5% Igepal, 50mM Tris pH 8, 10% glycerol, 0.1mM EDTA, 250mM NaCl, 50mM NaF, 0.7mM PMSF, 100X diluted Protease Inhibitor, 10mM DTT, 100uM Na3VO4). and centrifuged at 13,500 g for 10 min at 4°C. We performed Bradford assays (Bio-Rad) on the supernatants to quantify the protein concentration. We ran 50ug protein of each sample on 10% SDS-Page gels, and then transferred proteins to PVDF membrane (Bio-Rad, 162-0177). PVDF membrane was incubated in blocking solution (TBS with 1% Skim milk and 0.1% Tween-20) for 30min at RT, and then we added rabbit anti-CLOCK (1:200; Abcam, #ab3517), mouse anti-FLAG (1:400; Sigma, #F1804), rabbit anti-CLOCK (1:200; in house antibody raised in rabbit with epitope 377–556 a.a), and mouse anti-GAPDH (1:5000; Millipore, #MAB374) overnight at 4°C. PVDF membrane was washed in TBS with 0.1% Tween-20, and then incubated with corresponding secondary antibodies similar to IHC in blocking solution for 1h at RT, washed in TBS with 0.1% Tween-20 and imaged on an Odyssey infrared imaging system (LI-COR Biosciences).

Wheel-running rhythm

Mice (N = 6–10/genotype) at 8–12 weeks of age were housed in wheel-running cages¹¹⁹ to monitor their locomotor activity rhythm. We initially exposed mice to a 12:12 hours of light and darkness cycle (LD) for two weeks followed by 4 weeks of constant darkness (DD), and then we returned to a LD cycle for two weeks followed by 4 weeks of constant light (LL). We recorded wheel-running activity and analyzed the free-running period of DD and LL by using χ² periodogram analysis in ClockLab (Actimetrics, Wilmette, IL).

Behavioral tests

All tested mice were adult and handled 5 min/day by the experimenter for two weeks prior to experiments to alleviate their handling stress in experiments¹²⁰. We tested young adult mice (8–12 weeks of age) in the behavioral test battery. Mice underwent tests in the following order: open field test, novel object recognition test, social memory and aggressiveness test, set-shifting test, olfactory test, neurological reflex test, and tail suspension test. The order of the behavioral tests was organized to go from least to most stressful. We also applied two weeks of inter-test intervals to mice between two adjacent tests to minimize the effects from the previous test¹²¹. Fear conditioning was conducted using a different cohort of animals.

Open field test

The open field was conducted in the same apparatus and with the same procedure as previously described¹²². We individually released mice into the center of the arena and allowed them to explore for 1 hour. The first 30 min was video-recorded and analyzed using Limelight4 (Actimetrics, Wilmette, IL). We quantified the total distance moved as a measure of activity. We divided the whole arena into 25 (5X5) small square areas. We defined the 16 areas on the edge as the outside areas, while the 9 areas in the middle were defined as the center areas. We quantified the duration in the outside areas and the center of areas as a measure of anxiety.

Novel object recognition test

We did the novel object recognition test in the same arena as the open field test. 24 hours after the open field test, we provided two identical glass vials in the arena as a training session. We individually released mice and allowed 30 min to investigate the objects. We video-recorded the first 15 min for analysis. The testing session happened 24 hours later; we replaced one of the glass vials with a nail polish bottle, and then individually released the mouse to investigate the objects for 15 min. Performance was recorded and analyzed using Limelight 4. A recognition index of the ratio of nose investigation duration between the novel and familiar object in the testing session was calculated¹²³.

Social memory and aggressiveness test

We used a five trials social memory test to mimic the type of short but frequent interactions that occur in a colony setting¹²⁴. Juvenile (4–6 week old) sex-matched mice were used as social stimuli. We placed the test mouse into a new cage for a 30 min habituation period. A stimulus mouse was introduced into the cage for 1min. After a 10min delay, this was repeated for a total of five 1min exposures, with the same stimulus mouse in the first 4 exposures, and a novel stimulus mouse in the fifth exposure. We video-recorded and manually quantified the frequency and duration of social interaction (i.e. body investigation and attack).

Set-shifting test

Mice were calorie restricted. Food was only provided for 4 hours in the first 24 hours. For the second 24 hours, food was provided for 2 hours. For the remainder of the testing period, food was only provided for 1 hour per day. The mice maintained 80–85% of their ad libitum feeding body weight during the test. In the acclimation session before the real test, mice were trained to dig bedding medium in a bowl to obtain a ~10mg piece of peanut butter chip. This session consisted of two days of training. On the first day, we trained mice to recognize the peanut butter chip as food, and successfully obtained 23 pairs of chips from two empty feeding bowls. On the second day, mice started with obtaining 13 pairs of chips from two empty feeding bowls, followed by obtaining 13 pairs of chips from bowls with kay-kob (Kaytee, #1851118) gradually added, and at last obtaining 23 pairs of chips that are fully covered by kay-kob. The set shifting test consisted of four consecutive tests in the following order: discrimination, reversal, intra-dimensional shift (IDS), and extra-dimensional shift tests (EDS). We did all and tested in home cages. Two sensory modalities (visual and olfactory) were used. For each test, we paired 2 odors with 2 mediums (4 possible combinations) to the buried chip. A food reward was only associated with a particular stimulus in one modality. The mouse had to not only discriminate reward and non-rewarded modalities, but also discriminate and associate the correct stimulus in the correct modality with food reward. In discrimination, reversal, and intra-dimensional shift tests, the reward modality was consistent, while the reward modality was switched in the extra-dimensional shift test. Except for the discrimination and reversal tests, all stimuli were changed from test to test. Discrimination and reversal tests shared the same stimuli, but their association to reward were opposite. Supplementary Table 14 shows an example of combinations of stimuli across tests, and we counterbalanced all combinations of reward modalities and stimuli. A mouse was trained for up to 50 trials in each test. The criterion to learn was considered as 8 consecutive correct trials. A mouse that reached the criterion or 50 trails was forwarded to the next test. We quantified their performance by using the number of trials to criterion¹²⁵. The set-shifting test assesses the rule-based learning of animals. Thus, animals can also learn the rules of the shift and improve their performance across tests^125,126. Therefore, the first shift is more difficult than the subsequent ones. Reversal learning has been demonstrated as the most difficult task in the set shifting test^126,127. To avoid a potential ceiling effect and maximize the ability to distinguish enhanced performance between HU and WT mice, we tested reversal learning as the first rule shift.

Olfactory test

Since the set shifting test relies on olfactory discrimination, we ran an olfactory test to see if genetic modification of Clock/CLOCK affects olfactory discrimination. Mice were individually kept in a new cage for 30 minutes with a dry cotton swab attached to the center of the cage ceiling. We then tested mice in nine consecutive trials in the following order: three water dipped cotton swab trials, three vanilla cotton swab trials, and three lemon cotton swab trials. Each trial lasted for 2 minutes with 1 minute inter-trial intervals. We quantified the duration of nose investigation on the cotton swab¹²⁸.

Neurological reflex test

We conducted eye blink reflex, ear twitch reflex, and whisker orienting reflex with a cotton swab to examine any sensory-motor deficits. We quantified these data in a binary manner as blink or not, twitch or not, and orienting to touched whisker or not¹²⁹.

Tail suspension test

We first attached a 5cm Tygon tubing to the base of the mouse tail to prevent climbing onto its own tail, and then we attached a 20cm piece of yellow lab tape to the tail tip of each mouse. Mice were then placed at a height of 30cm height from the surface allowing them to hang from their tails. We tested 4 mice per batch. Adjacent individuals were about 20cm apart from each other, and there was no barrier between them. We video-recorded their performance for 6 min, and then manually quantified duration of immobility as previously described¹³⁰.

Fear conditioning test

We carried out fear conditioning testing as previously described¹²². We trained young adult mice in the fear-conditioning chambers. Mice were allowed to freely explore the chamber for 3 min. Then we applied a tone (2.8 KHz, 85 dB) for 30 sec, and a 2 second 0.85 mA foot shock overlapped with the last 2 sec of the tone. We applied the paired tone and foot shock stimuli for 4 times with 1 min inter-trial intervals. After 24 hours, we re-exposed the mice to the context (the same chamber in training) for 5 min, followed by the sound cue memory test (in a different chamber with different context and odor cues in another room but with the same tone) for 5 min. We used Ammeter (Med Associates, # ENV-421) to measure the shock intensity from each pair of neighboring wire bars on the grids, and sound was calibrated with a sound meter. The metal grid floors (for training) and plastic floors (novel context) were washed with water, sprayed with 70% ethanol and dried before each trial to avoid odor cues. All sessions were video-recorded and freezing of mice was scored by FreezeFrame3 (Actimetrics, Wilmette, IL).

Body and brain mass

We measured the body mass of living mice at P7, P21, and young adult age (8 to 12 weeks). From euthanized mice, we removed the brains, which included medulla and olfactory bulb, from P7 and P21 mice, and weighed them immediately. All measurements were conducted by using a tabletop scale (Denver Instruments, #MXX-412).

Single-nuclei RNA-seq

Tissue collection, nuclei isolation, and library preparation

Our nuclei isolation procedure was modified from our previous work¹³¹. The frozen tissue was transferred to a Dounce homogenizer (DWK Life Sciences, # 885300-0002) with 2ml of ice-cold Nuclei EZ lysis buffer (Sigma, #NUC-101). We then inserted pestle A for 23 strokes followed by pestle B for 23 strokes on ice. The homogenized sample was transferred to a 15ml conical tube. We added 2ml of ice-cold Nuclei EZ lysis buffer and incubated on ice for 5min. Nuclei were collected by centrifuging at 500 g for 5min at 4°C. We discarded the supernatant and added 4ml of ice-cold Nuclei EZ lysis buffer to resuspend nuclei. We then repeated the incubation and centrifuge steps and resuspended the nuclei in 200ul of nuclei suspension buffer (NBS) (1x PBS, 1% ultrapure BSA (Thermo Fisher Scientific, #AM2618), and 0.2 U/ul RNAse inhibitor (Thermo Fisher Scientific, #AM2694)), and finally filtered the nuclei suspension twice through a Flowmi Cell Strainer (Bel-Art, H13680-0040).

We mixed 10ul of nuclei suspension with 10ul of 0.4% Trypan Blue (Gibco, #15250061) and loaded onto a hemocytometer (INCYTO, #DHC-N01) to determine the nuclei number concentration. We then diluted each sample to 1000 nuclei/ul in NBS. 10,000 nuclei/sample was used to generate snRNA-Seq libraries using 10X Genomics Single Cell 3’ Reagent Kits v3 (10x Genomics, #1000153) following the manufacturer’s protocol¹³². We processed 24 mice (2 males and 2 females per genotype per age) for snRNA-Seq. Libraries were sent to the McDermott Sequencing Core at UT Southwestern for sequencing using Illumina NovaSeq 6000.

Data analysis

Data processing

For mouse single-nucleus RNA-Seq, raw BCL data for sequenced libraries was downloaded from the McDermott Sequencing Core at UT Southwestern. Self-written codes were used for data processing. FASTQ files were extracted using CellRanger mkfastq (v3.0.2)¹³². FASTQ files were next processed through UMI-Tools (v0.5.4) whitelist with ‘--set-cell-number=50000’ to identify the top 50,000 cell barcodes per sample¹³³. Reads corresponding to identified cell barcodes were then extracted using extract from UMI-Tools suite. Extracted FASTQ reads were then aligned to the reference mouse genome (MM10 GRCm38p6) and reference mouse Gencode annotation (vM17) using STAR (v2.5.2b)¹³⁴. Samples corresponding to genotype HU were additionally aligned to a custom reference assembly containing only human CLOCK that was extracted from the reference human genome (HG38 GRCh38p12) and human Gencode annotation (v28) using STAR (v2.5.2b)¹³⁴. Uniquely aligned BAM reads were further assigned to genes using featureCounts from Subread package (v1.6.2) using reference Gencode annotations¹³⁵. BAM files with alignment and gene assignment information were then sorted and indexed using Samtools (v1.6)¹³⁶. Sorted and indexed BAM files were further used to generate a raw counts matrix (per gene per cell barocde) using count from UMI-Tools suite. Raw count matrices per sample were re-organized using R SeuratDisk package (https://github.com/mojaveazure/seurat-disk) in order to match the input requirements for CellBender. CellBender was then run on each sample to remove noise from ambient RNA¹³⁷. Potentials doublets were filtered out using DoubletFinder R package¹³⁸. Cleaned and filtered count matrices were then used to generate Seurat objects and QC plots per age group per genotype. Before generating Seurat objects for samples that belong to genotype HU, data were updated to retain gene expression information for both the endogenous copy and humanized copy of Clock/CLOCK.

For the re-analysis of the single-nucleus comparative genomic data, raw FASTQ files of human, chimpanzee and macaque were acquired from publicly available data repositories^28–30. CellRanger (v8.0.1) mkref command was run to generate reference genome index for each of the species included in the re-analysis^29,132. In order to make these results comparable with previous findings in bulk RNA-Seq, qRT-PCR, and IHC, CellRanger (v8.0.1) count command was run on each sample with parameter ‘--include-introns’ set to false to count the reads mapping to exons only instead of exons and introns (default for single-nucleus transcriptomic datasets). Using the raw count matrix generated by CellRanger, CellBender was run to identify and remove ambient RNA contamination¹³⁷. Similarly, DoubletFinder was run to identify potential doublets from each dataset¹³⁸. Cleaned count matrixs were then used to merge and integrate the datasets per species per data source and clustering was performed using Seurat v5¹³⁹. Further clean-up was performed to remove any mis-clustered and low-quality cells/clusters and clusters were labelled for cell-types using previously published meta data from source publications. Cluster labels were further confirmed using literature-based gene markers for every identified cell-type. Log₂ normalized CLOCK expression per species per source was further extracted from respective Seurat objects to generate heatmaps.

Clustering analysis

R package Seurat (v4.0.6)¹⁴⁰ and self-written scripts were used for downstream analyses. After filtering, the raw UMI counts were normalized with a scaled factor of 10,000 and regressed to covariates such as the number of UMIs per cell, percent mitochondrial content per cell, batch, and sex as described in the Seurat analysis pipeline¹⁴⁰. Expression of the top 2000 variable genes was used to calculate principal components (PCs). Based on the result of JackStraw analysis, 30 and 28 PCs, which include the majority variation for P07 and P56 samples respectively, were used to identify cell clusters by using the original Louvain algorithm as described in the Seurat analysis pipeline¹⁴⁰. Dimension reduction and clustering visualization was conducted by using Uniform Manifold Approximation and Projection (UMAP)^141,142. 5 and 4 clusters of P07 and P56 datasets were removed due to their low number of UMIs (<10).

Cell type annotation

Gene enrichment per cluster was determined by Log₂ fold change ≥ 0.3 and FDR adjusted p value ≤ 0.05 for a given gene/cell compared to the rest of the cells in a cluster. We then applied self-written code to do pairwise Fisher Exact Tests between cell types of this study and previously published brain single-cell datasets^56,143 (Supplementary Fig. 5). We assigned the cell types to clusters based on the statistical results and location of cluster on the UMAP. One cluster of P07 did not show enrichment with either published dataset, but it showed enrichment of striatal neurons using Cell-type Specific Expression Analysis tool (http://genetics.wustl.edu/jdlab/csea-tool-2/)¹⁴⁴. Hence, we removed this cluster.

Differentially expressed gene (DEG) analysis

DEG analyses were conducted pairwise (i.e. between HU and WT, and between KO and WT) within either P07 or P56. We subsetted cells of major cell types (i.e. astrocyte, endothelium, excitatory neuron, inhibitory neuron, mature oligodendrocyte, oligodendrocyte precursor cell, and vascular leptomeningeal cell) as well as the subtypes of excitatory neurons (i.e. EX2-3_IT, EX4-5_IT, EX5_IT, EX6_IT, EX5_ET, EX5-6_NP, EX6_CT, and EX6b). We used MAST (v1.8.2)⁵⁸ with linear mixed model (MAST::GLM) for a zero-inflated regression analysis¹⁴⁵ to calculate an FDR adjusted p value for each gene. Compared to previous statistics in analysis of DEGs of snRNA-Seq, MAST::GLM can significantly decrease type I errors and hence report fewer numbers of DEGs^146,147. The linear mixed model tested genotype as fixed factor, sample as random factor, and other parameters as covariates (such as batch, sex and so on). We also applied the Seurat analysis pipeline to calculate log₂ fold change and percentage of expressed cells¹⁴⁰. Genes expressed in at least 30% of cells with |log₂ fold change| ≥ 0.1 and FDR adjusted p value ≤ 0.05 were determined as DEGs.

Hypergeometric overlap test

A hypergeometric test was used to assess the significance of overlaps of DEGs across cell types, between different ages, and between human CLOCK and mouse Clock. It was also used to test overlap between DEGs of HU-WT comparison in EX2-3_IT and human-specific open/closed chromatin²⁹. FDR adjusted p values were calculated based on the number of overlaps that occurred in a set of pairwise overlaps (e.g. Fig. 4E–H and Extended Data Fig. 4A–I).

Gene ontology (GO) enrichment analysis

ToppGene Suite¹⁴⁸ was used for GO analysis to determine the functions that were enriched by the DEGs. We used the expressed genes in P07 and P56 as the background input to ToppGene Suite. We did not apply a p value cutoff for output. The outputs were downloaded and processed using self-written code for further analysis and plot generation. Significant enrichment was set using a Benjamini-Hochberg FDR adjusted p value < 0.05.

Motif analysis in enhancer regions

Enhancer regions of TENM2/Tenm2 and SORCS2/Sorcs2 were obtained from Search Candidate cis-Regulatory Elements by ENCODE (SCREEN)⁶⁴, and we further downloaded the sequences of these enhancer regions from UCSC Genome Browser¹⁴⁹. We then scanned the sequences of the enhancer regions for CLOCK:BMAL1 motif by using Mulan¹⁵⁰.

Electrophysiology

Brain slices and recordings

P21–27 mice of both sexes were used for the electrophysiological recordings. We anesthetized mice with a mixture of ketamine (125 mg/kg, UTSW Veterinary Drug Service) and xylazine (25 mg/kg, AnaSed^® INJECTION, #NADA139236), and then quickly decapitated to dissect out the whole brain. The brain was glued on the section panel of a Vibratome (Leica, #VT1200S) and immersed in ice-cold dissection buffer (110mM Choline-Cl, 2.5mM KCl, 1.25mM NaH₂PO₄, 25mM NaHCO₃, 25mM Dextrose, 3mM Ascorbic Acid, 3.1mM Sodium Pyruvate, 7mM MgCl₂, 0.5mM CaCl₂.2H₂O, and 0.2mM Kyneuric Acid, pH 7.4, 285 mOsm osmolality). The brain was sectioned into 300μm thickness slices (corresponding to the images from #39 to #47 in the mouse P56 Coronal Atlas released as the Allen Brain Atlas in 2011). The slices were immediately transferred to nominal Artificial Cerebrospinal Fluid (ACSF) (125mM NaCl, 3mM KCl, 1.25mM NaH₂PO₄, 25mM NaHCO₃, 10mM Dextrose, 2mM MgCl₂, and 2mM CaCl₂.2H₂O, pH 7.4, 295 mOsm osmolality, bubbled with 5% CO₂ and 95% O₂) at 35°C for 30 min, and then kept at RT for at least 30 min before recordings. We carried out this experiment in the somatosensory cortex due to its substantial overlap with the frontal cortex and its features that allow higher accuracy to consistently sample neurons from layers 2–4. Excitatory neurons were patched using glass pipettes of 4–6 MΩ resistance, filled with K-Gluconate containing internal pipette solution (130mM K-Gluconate, 6mM KCl, 3mM NaCl, 10mM HEPES, 0.2mM EGTA, 4mM ATPMg, 0.4mM GTP-Tris, and 14mM phosphocreatine-Tris, pH 7.25, 305 osmolality). Neurons with a stable resting membrane potential and series resistance less than 20 MΩ were used for analysis.

Current steps for the excitability measurements

To measure the intrinsic excitability of the recorded neurons, we injected incremental current steps of 50pA amplitude and 500 ms duration in the current clamp at resting membrane potential. The number of action potentials at each current step were counted and plotted in the firing versus injected current curve. Also, the resting membrane potential of each neuron was determined and plotted.

Input resistance and normalized conductance

Input resistance of the neuron was measured in voltage clamp by applying single 500 ms duration, −10 mV voltage step at −85, −65 and −55 holding potentials. The recorded data from the voltage step was used to calculate the conductance and capacitance of each neuron was also calculated. Finally normalized conductance was determined by dividing the conductance by the capacitance.

Spontaneous and evoked EPSCs

Spontaneous EPSCs were recorded from layer 2–4 excitatory neurons in voltage clamp at −70 mV for 60 seconds while circulating oxygenated normal ACSF in the bath solution. Amplitude and frequency of the spontaneous EPSCs were analyzed using MiniAnalysis (Synaptosoft, Decatur, CA). Events were first detected automatically by the software after setting up cutoffs for the area (30) and height (10) of an event.

AMPA-receptor mediated evoked EPSCs were recorded from the layer 2–4 excitatory neurons in response to the extracellular electrical stimulation of horizontal intercortical afferents. We placed a cluster bipolar stimulating electrode (FHC, Bowdoin, ME) parallel to the recorded layer 2–4 excitatory neurons and applied biphasic pulse of 100 us interval to evoke an EPSC. ET (threshold intensity) is the minimum current intensity that evoked an EPSC, whereas AT is the amplitude of the EPSC response at ET. We then applied stimulus greater than the ET (2-fold, 3-fold, 4-fold and 5-fold of the ET) and measured the amplitude of the evoked EPSC. Finally fold change increases in the amplitude of the EPSC (with respect to AT) were calculated and plotted with the fold change increase in the stimulus intensity.

CRISPR-Cas9 generation of CLOCK KO iPSC line

The CLOCK KO cell lines were generated by targeting the exon 7 of CLOCK with a pair of sgRNAs following published guidelines¹⁵¹. The following set of sgRNAs were used: 5’-CACCGCTAGTGAAATTCGACAGGAC-3’ and 5’-AAACGTCCTGTCGAATTTCACTAGC-3’. To maximize the efficiency and mitigate potential off-target effects of gene editing, sgRNA candidates were designed and analyzed using CHOPCHOP v3¹⁵² and CRISPOR¹⁵³ respectively. SgRNA sequences with the fewest off-target sites in the human genome were selected for use. Each sgRNA was inserted into pSpCas9(BB)-2A-GFP (PX458)¹⁵¹ (Addgene, #48138). Cells were incubated with ROCK inhibitor for 1–3 hours and then electroporated with the pair of sgRNAs. 48–72 hours post electroporation, cells were subjected to fluorescence-activated cell sorting (FACS) for GFP+ signals and re-plated as single clones for the subsequent 10–12 days. To verify deletion of CLOCK, positive clones were confirmed via Sanger sequencing, qPCR, and western blot (Supplementary Fig. 6A–C).

Generation of iPSC-induced neuron

We differentiated iPSCs into induced neurons following a protocol modified from the original Neurogenin-2 (NGN2) fast differentiation experiment⁷² (Supplementary Fig. 6D).

Generation of transduced iPSC line

We obtained plasmids of reverse tetracycline-controlled transactivator (rtTA) (FUdeltaGW-rtTA; Addgene, #19780) and NGN2 (pTet-O-Ngn2-puro; Addgene, #52047). We amplified and harvested these plasmids by using EndoFree Plasmid Maxi Kit (Qiagen, #12362) and sent them to the Neuroconnectivity Core at Baylor College of Medicine Intellectual and Developmental Disabilities Research Center for lentivirus packing. We dissociated iPSCs to single cells using Accutase and plated them into 12-well plates at 3.8 × 10⁵ per well in mTeSR^™1 medium and ROCK inhibitor. 1 day later, lentiviruses rtTA (titre: 3.108 × 10¹⁰) and NGN2 (titre: 1.578 × 10¹⁰) were added in medium with 1:2500 and 1:10000 dilution respectively to transduce iPSCs. The transduced iPSCs were expanded for differentiation and frozen down for future studies following the Technical Manual.

NGN2 differentiation

We dissociated transduced iPSCs in Accutase and plated 1.5 × 10⁴/well and 1 × 10⁵/well into 24-well and 6-well plates respectively. We placed an acid cleaned glass coverslip (Carolina Biological Supply, #633029) into each well of a 24-well plate. Cells in 24-well plates were used for morphological analysis, while cells in 6-well plates were collected for qRT-PCR. All wells were coated with Matrigel. The confluency was monitored every 12 hours. Once the confluency reached about 50%, we started Day 1 and replaced stem cell medium to differentiation medium KSR (415ml KnockOut DMEM (Gibco^™, #10829018), 75ml Knockout Replacement Serum (Invitrogen, #10828-028), 5ml MEM NEAA (Invitrogen, #11140-050), 5ml Glutamax (Gibco^™, #35050), and 0.5ml beta-mercaptoethanol (Invitrogen, #21985-023) with 4ug/ml Doxycycline (Sigma-Aldrich, #D9891), 10uM SB431542 (Cellagen Technology, #C7243-5), 100nM LDN-193189 (Selleck Chemicals, #S2618), and 2nM XAV939 (Tocris Bioscience, #3748)). The iPSCs started to differentiate while many cells continued to proliferate (Supplementary Fig. 6D). 24 hours later (Day 2), we changed the medium to a mixture of half KSR and half N2B (500ml DMEM/F-12 (Life Technologies, #11330057), 5ml Glutamax, and 7.5ml 20% Dextrose (Sigma, #G6152) with 1:100 N-2 Supplement (Thermo Fisher Scientific, #17502-048), 4ug/ml Doxycycline, and 0.7ug/ml Puromycin (Sigma-Aldrich, #P8833)); cells differentiated into neural progenitor-like cells and grew until about 90% confluency (Supplementary Fig. 6D). On Day 3, about 70% of cells died due to lack of successful transduction with rtTA and NGN2 (Extended Data Fig. 10D). Medium was changed to N2B. On Day 4, cells continued to die until about 10% of newborn neurons survived (Supplementary Fig. 6D). We changed to NBM medium (485ml NBM (Gibco, #21103), 5ml Glutamax, 2.5ml MEM NEAA, and 7.5ml 20% Dextrose with 4ug/ml Doxycycline, 1:50 B27 (Thermo Fisher Scientific, #17504-044), 10ng/ml GDNF (PeproTech, #450-10), 10ng/ml BDNF (PeproTech, #450-02), and 10ng/ml NT3 (Fisher Scientific, #267N3005CF)), and added 10uM Ara-C (Sigma-Aldrich, #C1768) to eliminate progenitors. On Day 5, medium was changed, and 20,000 astrocytes were added to each well in 24-well plate. After that, cell cultures were divided into three groups. In group 1, cells were kept in NBM until Day 28. The other cultures were CRISPR transfected on Day 6, cells in group 2 and 3 were kept in NBM until Day 21 and Day 35 respectively. After changing NBM medium on Day 6 for group 1 and Day 7 for group 2,3, NBM was half changed every other day, and 2.5% FBS was add to NBM once per week.

CRISPR-Cas9 overexpression of TENM2 and SORCS2

We performed CRISPR-Cas9 to overexpress TENM2 and SORCS2 similar to a previous study⁷³.

We designed pairs of forward and reverse sgRNAs by using the above mentioned tools^152,153. F: 5’ ACCGTTTAATTGCCTAAGCGGTCT and R: 5’ AACAGACCGCTTAGGCAATTAAAC target the transcription start site (TSS) of TENM2, F: 5’ ACCGGTTTGCTTTATATGCTCATT and R: 5’ AACAATGAGCATATAAAGCAAACC target the site 33bp ahead of TSS of TENM2, F: 5’ ACCGAGATTGTGCAGGGCTGACCA and R: 5’ AACTGGTCAGCCCTGCACAATCTC target the site 137bp ahead of TSS of TENM2, F: 5’ ACCGAGTGACAAGTGCCAAAACTC and R: 5’ AACGAGTTTTGGCACTTGTCACTC target Exon 2 of TENM2, F: 5’ ACCGCTGGCGACCATGGCGCACCG and R: 5’ AACCGGTGCGCCATGGTCGCCAGC target TSS of SORCS2, F: 5’ ACCGCGGTGCGCGCTCGGCGTCGC and R: 5’ AACGCGACGCCGAGCGCGCACCGC target the site 60bp ahead of TSS of SORCS2, and F: 5’ ACCGGACCGGAGGAACCACCGAGC and R: 5’ AACGCTCGGTGGTTCCTCCGGTCC target the site 168bp ahead of TSS of SORCS2. Each pair of sgRNAs was cloned into pAAV-U6-sgRNA-CMV-GFP (Addgene, #85451) by using the SapI restriction enzyme (BioLabs, #R0569). To confirm the successful insertion of sgRNAs into the plasmids, we sequenced them in the Sanger Sequencing Core at McDermott Center of UTSW by using the forward primer: 5’ ACTATCATATGCTTACCGTAAC.

We transfected HEK293T cells (at about 70% confluency) and iPSC-induced neurons on Day 6 (Supplementary Fig. 6D) by following the Lipofectamine^® LTX DNA Transfection Reagents Protocol. We diluted 3ul of Lipofectamine^® LTX Reagent (Invitrogen, #15338030) in every 50ul Opti-MEM^™ I Reduced Serum Medium (Gibco, #31985088). Vectors of pMSCV-LTR-dCas9-VP64-BFP (Addgene, #46912) and pAAV-U6-sgRNA-CMV-GFP with/without sgRNAs insertion were diluted into 1ug/ul and mixed. 1ug of vector and 1ul of PLUS^™ reagents (Invitrogen, #15338030) were added to every 50ul Opti-MEM^™ I Reduced Serum Medium. The diluted lipofectamine reagent and vectors were mixed in a 1:1 ratio and incubated for 15 min at RT to generate DNA-lipid complexes. Then 50ul and 250ul of the DNA-lipid complex were added to each well of a 24-well and 6-well plate respectively. After 12 hours for HEK293T cells and 24 hours for iPSC-induced neurons, we stopped the transfection by changing to fresh medium.

Immunocytochemistry (ICC)

We washed wells containing cover slips with 1X TBS and then fixed them with 4% PFA and 2% sucrose in TBS for 20 min at RT. Aldehydes were quenched with 0.3 M glycine in TBS-T for 1 hour at RT. We then incubated cover slips in blocking solution (5% normal donkey serum and 3% BSA in TBS-T) for 1 hour at RT. We removed blocking solution and incubated cover slips in primary antibody solution (blocking solution with primary antibody) for 24 hours at 4°C. Secondary antibody incubations were performed overnight at 4°C followed by incubation in 300 nM DAPI solution for 5 min at RT. Coverslips were mounted in ProLong Gold Antifade Mountant. We used chicken anti-MAP2 (1:2000; Abcam, #ab5392) to label mature neurons, mouse anti-CAMKII (1:100) to label excitatory neurons, sheep anti-TENM2 (1:200) to validate expression of TENM2, sheep anti-SORCS2 (1:500) to validate expression of SORCS2, mouse anti-CUX1 (1:500; Santa Cruz Biotechnology, #sc-514008) to mark layer 2–4 like neurons, goat anti-FOXP2 (1:500) to mark layer 6 like neurons, chicken anti-GFP (1:2000; Aves Labs, #GFP-1010) to enhance GFP signal of transfected plasmid, mouse anti-TUJ1 (1:1000; Covance, #MMS-435P) to label dendrites of neurons, and rabbit anti-VGLUT2 (1:500; Synaptic Systems, #135402) to label the presynaptic puncta. We used species-matched secondary antibodies produced in donkey and conjugated to Alexa Fluor 488, Alexa Fluor 555, or Alexa Fluor 647 (1:2000). A 20X objective lens was used to capture images from cover slips on a Zeiss Axio Observer Z1 florescent microscope and a 10X objective lens was used to capture images from 6-well plates on a Leica Thunder Imager 3D Cell Culture with Tokai Stage Top Microincubator at the UT Southwestern Neuroscience Microscopy Facility. We analyzed the dendritic complexity in a 1388 × 1040 pixels (447 × 335 um) field which contains neurons (Fig. 7A and Extended Data Fig. 7A) by using FIJI¹¹² with self-written Macro codes which followed an available analysis pipeline¹⁵⁴. We thresholded the TUJ1 signal to binary, and then used the Watershed and Analyze Skeleton (2D/3D) plugins of Fiji to restore and quantify the number and length of segments in the image. A segment was defined as the line between each of two immediately connected intersections^154–156. These two data were normalized to the number of DAPI+ cells that colocalized with GFP and TUJ1 signal to determine the complexity of dendritic branching. VGLUT2 puncta that overlapped with GFP and TUJ1 were counted and normalized to the length of dendrites as a quantification of spine density.

Statistical analysis and reproducibility

No statistical methods were used to pre-determine sample sizes, but our sample sizes are similar to those reported in previous publications. Data distribution was assumed to be normal, but this was not formally tested. Instead, we plotted individual data points to show the data distribution whenever it is possible. For mouse experiments, individuals of all genotypes were from littermates. Human and non-human primate samples were not randomly selected due to the availability of the tissues. CLOCK KO iPSC-derived neurons were compared with the wildtype cell line which was used to generate the CLOCK KO iPSCs. Experimenters were blind to the genotypes and species at least during experiments and data quantification. The details of statistical analyses are described in each figure legend. We performed all statistical analyses in R (version 4.1.1) by using self-written code.

Extended Data

Extended Data Fig. 1. Human-specific upregulation of CLOCK in neocortex.

(A) CLOCK expression among human, chimpanzee, and macaque across major neocortical cell types through reanalyzing recently published comparative snRNA-seq genomics datasets28–30. We applied a general linear mixed model (GLMM) with species as fixed factors, cell type as random factor nested with individual, and Tukey’s test for post-hoc analysis. EX: excitatory neuron; IN: inhibitory neuron; ASTRO: astrocyte; Oligos: oligodendrocyte and OPC; ENDO: endothelium; _MICRO: microglia. (B) Representative images of immunofluorescent staining for neurons (NEUN, yellow), CLOCK+ cells (CLOCK, red), oligodendrocytes (OLIG2, green), and all cells (DAPI, blue) in posterior cingulate cortex. The colored dash polygons labelled CLOCK-neurons (yellow) and oligodendrocytes (green) in the CLOCK channel, while solid polygons indicated the corresponding cells in each channel. (C-E) Quantification of (C) fluorescent intensity of CLOCK staining, CLOCK+ fraction in (D) neurons and (E) oligodendrocytes. For panels C-E, the boxplots are defined by the first quartile, mean, and third quartile for the box, and whiskers extending to the most extreme data points within 1.5 times the interquartile range. Each data point is a subarea of a section in the posterior cingulate cortex. We sampled 3 subareas in 2 sections per individual and 3 individuals per species. Points with the same color are from the same individual. We applied a general linear mixed model (GLMM) with species as fixed factors, subarea as random factor nested with section which nested with individual, and Tukey’s test for post-hoc analysis. HS: human, PT: chimpanzee, and MM: macaque. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Extended Data Fig. 2. Human CLOCK increases cell density in adult somatosensory cortex but not nucleus accumbens.

Quantification of cell density in adult somatosensory cortex (A-D) and nucleus accumbens (E-H). (A,E) Representative image of IHC staining for all cells (DAPI, blue), neurons (NEUN, green), and oligodendrocytes (OLIG2, red). (B-D and F-H) Quantification of cell number in (B,F) neurons, (C,G) oligodendrocytes, and (D,H) all cells. For all data panels, each data point is a section containing somatosensory cortex. We sampled 3–5 adult mice per genotype (HU: n = 5; WT: n = 3; KO: n = 3) and 4 sections per mouse. We did GLMM with genotype and sex as fixed factors, section as random factor nested with individual, and Tukey’s test for post-hoc analysis. The open circles represent female samples, while closed circles represents male samples. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Extended Data Fig. 3. Spatiotemporal expression of CLOCK/Clock across cell types in the neocortex.

(A,B) Comparison of CLOCK/Clock+ cells percentage in cell subtypes among HU, WT, and previous work on human neocortical development50 based on snRNA-seq data at (A) P07 mice and 12 years human and (B) P56 mice and 28 years human. (C,D) Cellular expression of CLOCK in each cell type at (C) P07 or (D) P56. (E,F) Feature plot of CLOCK expression in UMAP at (E) P07 or (F) P56.

Extended Data Fig. 4. Comparison of downstream genes regulated by human and mouse CLOCK in subtypes of excitatory neurons across developmental stages.

Number (bar plot) and Log2 fold change (violin plot) of DEGs in each subtype of excitatory neurons, and hypergeometric tests between DEGs from each pair of cell types (heatmap) at (A) P07 HU vs. WT comparison, (B) P07 KO vs. WT comparison, (C) P56 HU vs. WT comparison, and (D) P56 KO vs. WT comparison. The solid line in violin plot indicates the median, while the two dash lines mark the 25 and 75 percentiles. In EX2–3_IT neurons, overlap of DEGs from human and mouse CLOCK at (E) P07 or (F) P56. (G) Overlap of DEGs result from human CLOCK between P07 and P56. (H) Overlap of DEGs from mouse Clock between P07 and P56. The color of each cell represents the −Log10 q value of hypergeometric test in heatmaps. The number in each cell is the number of overlapped DEGs. The number in brackets are the number of DEGs. (I) Venn diagram to show overlap between DEGs of EX2/3_IT and human-specific open chromatin in neurons. (J) Representative images to compare TENM2 fluorescent intensity (green) between HU and WT mice in excitatory neurons (CAMKII, red) of frontal cortex at P07. (K) Quantification of TENM2 fluorescent intensity. Each data point is an excitatory neuron, and we sampled 3 mice per genotype and 5 neurons per mouse. (L) Representative images to compare SORCS2 fluorescent intensity (green) between HU and WT mice in the frontal cortex at P56. (M-N) Quantification of (M) absolute and (N) normalized SORCS2 fluorescent intensity. Each data point is an image of a frontal cortex section, and we sampled 3 mice per genotype and 5–6 images per mouse. We did GLMM with genotype and sex as fixed factors, image as random factor nested with individual, and Tukey’s test for post-hoc analysis. The open circles represent female samples, while closed circles represents male samples. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Extended Data Fig. 5. Neuronal morphology of layer 2–4 excitatory neurons in frontal cortex.

(A) AAV transduced (tdTomato, red) excitatory neurons, but not astrocytes (GFAP, green) and oligodendrocytes (OLIG2, magenta). (B) AAV transduced (tdTomato, red) excitatory neurons across cortical layers (layers 2–4: CUX1, green; layer 6: FOXP2, magenta) of adult frontal cortex. (C-F) In adult mice, (C) comparison of soma area. Each data point is a neuron, and we sampled 39 neurons from 4 HU mice, 46 neurons from 4 WT mice, and 20 neurons from 3 KO mice. We did GLMM with genotype and sex as fixed factors, neuron as random factor nested with individual, and Tukey’s test for post-hoc analysis. (D-F) Comparison of cumulative probability distribution of (D) head area, (E) spine length, and (F) neck width of spines across genotypes. Each data point is a spine, and we sampled all spines from 2–3 segments from 3–7 neurons per mouse. We did GLMM with genotype and sex as fixed factors, spine as random factor nested with segment which again nested with neuron which further nested with individual, and Tukey’s test for post-hoc analysis. (G-N) In P18 mice, (G) CLOCK does not alter soma area. Each data point is a neuron, and we sampled 4 mice per genotype and 6–19 neurons per mouse (total number of neurons: HU: n = 59; WT: n = 71; KO: n = 37). (H-I) Quantification of (H) number and (I) total length of branch in each genotype. For panel G-I, each data point is a neuron, and we sampled 4 mice per genotype and 7–17 neurons per mouse (total number of neurons: HU: n = 48; WT: n = 63; KO: n = 35). For panel G-I, we did GLMM with genotype and sex as fixed factors, neuron as random factor nested with individual, and Tukey’s test for post-hoc analysis. (J) Sholl analysis to quantify encountered intersections between neuron branches and concentric rings from soma for dendrite complexity. The data are from the same samples of panel H and I. We did repeated measures ANOVA with genotype and sex as between-subject factors, distance to soma as within-subject factor, and Tukey’s test for post-hoc analysis. (K) Cumulative probability distribution of spines density across genotypes. Each data point is a segment, and we sampled 2–3 segments from 3–5 neurons per mouse. (L-N) cumulative probability distribution of (L) head area, (M) spine length, and (N) neck width of spines across genotypes. Each data point is a spine, and we sampled all spines from 2–3 segments from 3–5 neurons per mouse. We did GLMM with genotype and sex as fixed factors, spine as random factor nested with segment which again nested with neuron which further nested with individual, and Tukey’s test for post-hoc analysis. The open circles represent female samples, while closed circles represents male samples. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Extended Data Fig. 6. Additional data to confirm CLOCK KO reduces dendritic complexity and spine density in iPSC-derived neurons.

Date collected from different CLOCK KO cell lines (KO#2 and KO#7) at different differentiation stages (Day 21, Day 28, and Day 35). Data of WT and KO#2 at Day 28 are replotted from Fig. 7, and date of Day 21 are replotted from Extended Data Fig. 12. (A) Total length of dendrites normalized to number of neurons. (B) Number of segments of dendrites per neuron. (C) Number of puncta per μm of dendrites. For all panels, each data point is an independent culture of neurons on a circle glass cover slip sitting on the bottom of a well in a 24-well plate. We did GLM with genotype/treatment as fixed factors with Tukey’s test for post-hoc analysis. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Extended Data Fig. 7. TENM2 partially rescued CLOCK KO mature neuronal features in iPSC-derived neurons.

A) ICC images of iPSC-derived neurons on Day 21 with nuclei marker DAPI (blue), CRISPR-Cas9 transfection marker GFP (green), neuronal dendrite mark TUJ1 (red), and presynaptic marker VGLUT2 (magenta). (B) Total length of dendrites, which were quantified from TUJ1 staining, normalized by the number of nuclei overlapped with TUJ1 signals. (C) Number of TUJ1 segments normalized by the number of nuclei overlapped with TUJ1 signals as a quantification of dendrite complexity. (D) Number of VGLUT2 puncta overlapped with TUJ1 signal and normalized by length of dendrites as a quantification of presynaptic spine density. For all data panels, each data point is an independent culture of neurons on a circle glass cover slip sitting on the bottom of a well in a 24-well plate. We did GLM with genotype/treatment as fixed factors with Tukey’s test for post-hoc analysis. Details of statistical results can be found in Supplementary Table 1. All data are shown as means ± SEM. All statistics were two-sided tests. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.

Data Availability

Single-nucleus RNA-seq data are deposited into NCBI GEO (GSE224759). We used reference mouse genome (MM10 GRCm38p6) for annotation. Other datasets which have been used in this paper: single-nuclei RNA-Seq data from Allen Brain Institute (RRID:SCR_016152; https://assets.nemoarchive.org/dat-net1412), single-nuclei RNA-Seq data of Caglayan et al., (2023) from NCBI GEO (GSE192774), and single-nuclei RNA-Seq data of Khrameeva et al., (2020) from NCBI GEO (GSE127774). Due to the amount of data, all other source data will be made available upon request to the corresponding authors.

Code Availability

We have deposited all original code to GitHub (https://github.com/konopkalab/CLOCK-humanized-mice). DOI: 10.5281/zenodo.15319773¹⁵⁷