Introduction

Much research efforts have been made in uncovering the wiring and signaling pathways of neural circuits underlying specific behaviors. Multiple circuit dissection strategies have been developed, including genetic screening, genetic labelling, circuit tracing, live imaging, genetic sensors and central nervous system reconstruction via electron microscopy (EM) imaging. Recently, we have developed chemoconnectomics (CCTomics), focusing on building a comprehensive set of knockout and knockin tool lines of chemoconnectome (CCT) genes, to dissect neural circuitry based on chemical transmission (Deng et al., 2019).

Each strategy has advantages and disadvantages, for example: genetic screening is less biased but inefficient; circuit tracing with viruses provides information of connection, but is often prone to leaky expression and inaccurate labelling; and EM reconstruction is anatomically elaborate but does not allow for manipulation of corresponding neurons. CCTomics overcomes some limitations of previous strategies by allowing for behavioral screening of CCT genes and accurate labelling or manipulation of corresponding neurons. However, it is still limited in that knockout of some CCT genes can be lethal during development and that CCT genes may function differently in different neurons, which require a cell type-specific manipulation. Thus, we decided to invent a conditional CCTomics (cCCTomics) in which gene deletion was conditional.

There are three major strategies for somatic gene mutagenesis at the DNA/RNA level: RNA interference, DNA site-specific recombination enzymes, and CRISPR/Cas system. RNA interference targets RNAs conveniently and efficiently (Martin and Caplen, 2007; Oberdoerffer et al., 2005). Libraries of transgenic RNAi flies covering almost the entire fly genome have been established (Ni et al., 2011; Perkins et al., 2015). DNA site specific recombination enzymes such as Flp, B3 and Cre mediate specific and efficient gene editing (Gaj et al., 2014; Grindley et al., 2006). These strategies require flies with reverse repetitive sequences knocked into the corresponding genes, which is time-consuming with relatively complex recombination for genetic assays. CRISPR/Cas systems, particularly CRISPR/Cas9, which targets DNA with a sgRNA/Cas protein complex, have been broadly applied in gene manipulation over the last decade. The widespread use of CRIPSR/Cas9 in Drosophila somatic gene manipulation began in 2014 (Xue et al., 2014). Later, tRNA-flanking sgRNAs was designed and applied, which enabled multiple sgRNAs to mature in a single transcript (Xie et al., 2015), accelerating the application of this strategy in conditional gene manipulation in flies with impressive efficiency (Delventhal et al., 2019; Port and Bullock, 2016; Port et al., 2020; Schlichting et al., 2019; Schlichting et al., 2022). Additionally, libraries of UAS-sgRNA targeting kinases (Port et al., 2020) and GPCRs (Schlichting et al., 2022) have been established, but no sgRNA libraries covering all the CCT genes exists yet. The efficiency of CRISPR/Cas9 has not been validated systematically in the Drosophila nervous system.

The circadian rhythm can be used for proof of principle testing of cCCTomics. Organisms evolve periodic behaviors and physiological traits in response to cyclical environmental changes. The rhythmic locomotor behavior of Drosophila, for instance, shows enhanced activity before the light is turned on and off in a light-dark (LD) cycle, referred to as morning and evening anticipations, respectively (Collins et al., 2005; Helfrich-Förster, 2001). Under dark-dark (DD) conditions, the activities peaks regularly about every 24 hours (h) (Konopka and Benzer, 1971). Approximately 150 clock neurons, circadian output neurons and extra-clock electrical oscillators (xCEOs) coordinate Drosophila circadian behaviors (Dubowy and Sehgal, 2017; Tang et al., 2022). The regulation of morning and evening anticipations, the most prominent features in the LD condition, is primarily mediated by four pairs of sLNvs expressing pigment dispersing factor (PDF), six pairs of LNds and the 5^th s-LNv (Grima et al., 2004; Rieger et al., 2006; Stoleru et al., 2004). At the molecular level, Pdf and Pdf receptor (PDFR) are well-known, with their mutants showing an advanced evening activity peak and no morning anticipation (Hyun et al., 2005; Lear et al., 2005; Renn et al., 1999). Other neuropeptides and their receptors, including AstC/AstC-R2 and neuropeptide F (NPF) and its receptor (NPFR), have also been reported to modulate evening activities (Díaz et al., 2019; He et al., 2013; Hermann et al., 2012), while CCHa1/CCHa1-R and Dh31 regulate morning activities (Fujiwara et al., 2018; Goda et al., 2019). To date, no advanced morning activity phenotype has been reported in flies.

To develop a more efficient approach for somatic gene manipulation, we have now generated two systems for conditional manipulation of CCT genes. One is GFPi/Flp-out-based conditional knockout system of CCT genes (cCCTomics) and another is CRISPR/Cas9-based (C-cCCTomics). Both systems have achieved high efficiency of gene mutagenesis in the Drosophila nervous system. C-cCCTomics, utilizing chromatin-peptide fused Cas9 and scaffold optimized sgRNA, makes efficient conditional gene knockout as simple as RNAi. Further application of C-cCCTomics in clock neurons revealed novel roles of CCT genes in circadian behavior: CNMa-CNMaR modulates morning anticipation as an antagonistic signal of PDF-PDFR.

Results

Complete Disruption of Target Genes by GFPi and Flp-out Based cCCTomics

For the purpose of conditional chemoconnectomics (cCCTomics), we initially leveraged the benefits of our previously generated CCTomics attP lines (Deng et al., 2019), which enabled us to fuse enhanced GFP coding sequence at the 3’ end of each gene’s coding region and flank most or entire gene span with FRT sequence through site-specific recombination (Fig 1A, Table S1). We designed this system so that it could be used to target genes tagged with GFP by RNAi (Neumüller et al., 2012) and at the same time to enable flippase(Flp) mediated DNA fragment excision between two FRT sequences when FRT sequences are in the same orientation (Vetter et al., 1983)(Fig1A).

Fig 1.

To validate the efficiency of cCCTomics, we performed pan-neuronal expression of either shRNA^GFP or flipase in cCCT flies. Immunofluorescent imaging showed that constitutive expression of shRNA^GFP (Fig 1B-1D, Fig S1A-S1I) or flipase (Fig 1E-1G, Fig S1J-S1U) almost completely eliminated GFP signals of target genes, indicating high efficiency. Knockout at the adult stage using either hsFLP driven Flp-out (Golic and Lindquist, 1989) (Fig 1H-1J) or neurally (elav-Switch) driven shRNA^GFP (Nicholson et al., 2008; Osterwalder et al., 2001) (Fig S2A-S2I) also showed high efficiency of suppression. Notably, control group of CCT^EGFP.FRT; elav-Switch/UAS-shRNAGFP flies fed with solvent (ethanol) showed obvious decreased GFP (Fig S2B, S2E, and S2H) comparing with UAS-shRNA^GFP/CCT^EGFP.FRT flies fed with RU486 (Fig S2A, S2D, and S2G), indicating leaky expression of elav-Switch.

We then applied cCCTomics in pan-neuronal knockout of nAChRβ2 which is required for Drosophila sleep (Dai et al., 2021). Ablation of nAChRβ2 in the nervous system dramatically decreased both day and night sleep of flies, which was the same as shown in the previous paper (Dai et al., 2021) (Fig 1K, 1L). Therefore, cCCTomics is an effective toolkit for manipulation of CCT genes and suitable for functional investigations of genes. Expression of in-frame fused eGFP-labelled CCT genes highly co-localized with signals revealed by immunocytochemistry (Fig S3A-S3I), allowing direct examination of gene expression without amplification, which is different from the GAL4/UAS binary system.

We then checked the viability of cCCT lines and found that cCCT lines including Capa^EGFP.FRT, ChAT^EGFP.FRT and Eh^EGFP.FRT were viable whereas their CCT mutants were lethal. Gad1^EGFP.FRT, GluRIID^EGFP.FRT and CapaR^EGFP.FRT were still lethal as their CCT mutants were (Table S2). This indicates that some of the cCCT knockin flies may functionally affect corresponding genes, which are not suitable for conditional gene manipulation. Combination of cCCT transgenic flies with UAS-flp, UAS-shRNA^GFP or specific drivers is complicated and unfriendly for screen work, despite the almost 100% efficiency of gene suppression. Because of the limitations of this method, we further created a CRISPR/Cas9 based conditional knockout system of chemoconnectomics (C-cCCTomics).

CRISPR/Cas9-Based Conditional Knockout System for CCTomics

To simplify effective manipulation of CCT genes, we designed a vector based on pACU2 (Han et al., 2011) with tRNA flanking sgRNAs (Port and Bullock, 2016; Xie et al., 2015) targeting CCT genes (Fig 2A). We also adopted an optimized sgRNA scaffold “E+F” (E, stem extension; F, A-U flip) (Chen et al., 2013), which facilitates Cas9-sgRNA complex formation and gene knockout efficiency (Dang et al., 2015; Poe et al., 2019; Zhao et al., 2016), to all sgRNAs to improve gene knockout efficiency. To balance efficiency and off-target effect, we selected three sgRNAs for each CCT gene with highest predicted efficiency and no predicted off-target effect based on previously reported models (Chu et al., 2016; Doench et al., 2014; Graf et al., 2019; Gratz et al., 2014; Heigwer et al., 2014; Stemmer et al., 2015; Xu et al., 2015) (see Table S3 and details at Methods).

Fig 2.

We generated UAS-sgRNA^3x transgenic lines for all 209 defined CCT genes (Abruzzi et al., 2017; Dai et al., 2019; Deng et al., 2019) and UAS-Cas9.HC (Cas9.HC) (Mali et al., 2013). We first verified that C-cCCTomics mediated precise target DNA breaking by ubiquitous expression of Cas9.P2 (Port et al., 2014) and sgRNA by targeting Pdf or Dh31. Sanger sequencing showed that indels were present exactly at the Cas9 cleavage sites (Fig S4A-S4H). To determine the efficiency of C-cCCTomics, we employed pan-neuronal expression of Cas9.HC with sgRNA^Dh31 or sgRNA^pdf. Targeting by Cas9.HC/sgRNA^Dh31 eliminated most but not all of the GFP signal in Dh31^EGFP.FRT (Fig 2B-2D), whereas all anti-PDF signals were eliminated by Cas9.HC/sgRNA^pdf (Fig 2E-2G). Furthermore, we used the C-cCCT strategy to conditionally knockout genes for pdfr, nAChRβ2 and nAChRα2, which were previously reported as essential for circadian rhythm or sleep (Dai et al., 2021; Renn et al., 1999). Pan-neuronal knockout of Pdfr phenocopied arrhythmicity of its null mutants (Fig 2H-2J) while there was no significant sleep decrease in these conditional knockout (cKO) flies (Fig 2K-2L) when we applied C-cCCTomics to manipulate nAChRβ2 or nAChRa2. Taken together, C-cCCTomics (with Cas9.HC) achieved a relatively high gene knockout efficiency, but it was not effective enough for all genes.

Evaluation of Cas9 with Different Chromatin-Modulating Peptides

Since the establishment of the CRISPR/Cas9 system a decade ago, many groups have attempted to improve its efficiency in gene manipulation. Most attempts have been focused on the two main components of this system, the Cas9 protein (Ding et al., 2019; Ling et al., 2020; Liu et al., 2019; Zhao et al., 2016; Zheng et al., 2020) and the single guide RNA (Chen et al., 2013; Chu et al., 2016; Dang et al., 2015; Doench et al., 2014; Filippova et al., 2019; Graf et al., 2019; Labuhn et al., 2018; Mu et al., 2019; Nahar et al., 2018; Scott et al., 2019; Xu et al., 2015). At the beginning of C-cCCTomics design, we adopted an optimized sgRNA scaffold and selected sgRNAs with predicted high efficiency. We tried to further improve the efficacy by modifying Cas9 protein. We fused a chromatin-modulating peptide (Ding et al., 2019), HMGN1 (High mobility group nucleosome binding domain 1), at the N-terminus of Cas9 and HMGB1 (High mobility group protein B1) at its C-terminus with GGSGP linker, termed Cas9.M9 (Fig 3A, Methods). We also obtained a modified Cas9.M6 with HMGN1 at the N-terminus and an undefined peptide (UDP) at the C-terminus (Fig 3A). We replaced the STARD linker between Cas9 and NLS in Cas9.HC with GGSGP the linker (Zhao et al., 2016), termed Cas9.M0 (Fig 3A). None of these modifications have been validated previously in flies.

Figure 3.

To determine whether the modified Cas9 variants were more efficient, we first pan-neuronally expressed each Cas9 variant and sgRNA^ple, and assessed their efficiency by immunofluorescence imaging. By counting anti-TH positive neurons in the brain (anterior view) after targeting by Cas9/sgRNA^ple, we found that unmodified Cas9.HC/sgRNA^ple only achieved 69.58±3.04% (n=5) knockout efficiency (Fig 3G, 3K, 3L), while Cas9.M6/sgRNA^ple and Cas9.M9/sgRNA^ple significantly improved efficiency to 87.53±3.06% (n=7) and 97.19±2.15% (n=8), respectively (Fig 3I-3L). Fourteen additional CCT genes were subjected to pan-neuronal knockout, and the mRNA levels of the target genes were evaluated using real-time quantitative PCR with at least one primer overlapping the sgRNA targeting site (Fig S5). Cas9.M6 and Cas9.M9 demonstrated significantly higher gene disruption efficiency compared to the unmodified Cas9.HC, achieving average efficiencies of 87.51%±2.24% and 89.59%±2.39% for Cas9.M6 and Cas9.M9, respectively, in contrast to 70.72%±3.82% for Cas9.HC. (Fig 3M, Fig S5). To rule out the possibility of the observed variations in gene disruption efficiency being attributed to differential Cas9 expression levels, we quantified the Cas9 expression levels and noted that both Cas9.M6 and Cas9.M9 exhibited lower mRNA levels than Cas9.HC under the experiment condition (Fig 3N). Subsequently, genomic DNA of Drosophila head was extracted, and libraries encompassing target sites were constructed for high-throughput sequencing to verify disparities in genetic editing efficiency among these three Cas9 variants (Fig 3O). In almost all nineteen sites tested, the mutation ratio consistently showed a trend towards Cas9.M6 and Cas9.M9 having a higher gene disruption efficiency than Cas9.HC (Fig 3P, Fig S6). The single-site mutation rates varied from 5.81% to 43.47% for Cas9.HC, 22.40% to 53.54% for Cas9.M6, and 19.90% to 63.57% for Cas9.M9 (Fig 3P, Fig S6). It should be noted that genomic DNA extracted from fly heads contained glial cells, which did not express Cas9/sgRNA, leading to a larger denominator and consequently reducing the observed mutation rates. Unmodified Cas9 displayed mutation rates comparable to those previously reported by Schlichting et al., 2022. The findings indicated that both Cas9.M6 and Cas9.M9 displayed elevated efficiency compared to Cas9.HC, with Cas9.M9 exhibiting the highest mutagenesis proficiency. These results suggest that the implementation of modified C-cCCTomics using Cas9.M6 and Cas9.M9 achieved an elevated level of efficiency. While unmodified C-cCCTomics was not efficient enough to knockout nAChRβ2 and nAChRα2 to phenocopy their mutants, we employed Cas9.M9 in conditional knockout of these genes to verify its efficiency. Pan-neuronal knocking out of nAChRβ2 or nAChRa2 by Cas9.M9/sgRNA showed significant sleep decrease which was similar to their mutants (Fig 3Q-3R) (Dai et al., 2021).

Taken together, our results support that we have created a high efficiency toolkit for CCT gene manipulation in the nervous system, as well as more efficient Cas9 variants, Cas9.M6 and Cas9.M9, which can also be applied to genes other than those in the CCT.

45 CCT Genes Found in Clock Neurons by Genetic Intersection

We analyzed the expression profile of CCT genes in circadian neurons with CCTomics driver lines in all clock neurons expressing Clk856 (Gummadova et al., 2009). With the Flp-out or split-lexA intersection strategy (Fig 4A, 4B), we found 45 out of 148 analyzed CCT genes expressed in circadian neurons (Fig 4C, Fig S7-S8, Table S4, Table S5). In all eight subsets of clock neurons, 24 CCT genes were expressed in DN1s, 20 in DN2s, 21 in DN3s, 29 in LNds, 15 in l-LNvs, 11 in s-LNvs, 5 in 5^th s-LNv and 3 in LPNs, with a total of 128 gene-subsets.

Fig 4.

Prior to transcriptomic analysis (Ma et al., 2021), 23 CCT genes had been reported in clock neurons (Díaz et al., 2019; Duhart et al., 2020; Erion et al., 2016; Frenkel et al., 2017; Fujiwara et al., 2018; Goda et al., 2016; Hamasaka et al., 2007; He et al., 2013; Hermann-Luibl et al., 2014; Hermann et al., 2012; Hyun et al., 2005; Renn et al., 1999; Selcho et al., 2017). In our intersection dissection, expression profiles of 15 CCT genes were similar but not identical to previously reported 18 CCT genes in clock neurons (Fig 4C, Table S4). An additional 22 out of 30 CCT genes identified in our genetic intersection were also detected in the transcriptome analysis (Ma et al., 2021). Among the 45 CCT genes expressed in eight clock neuron subsets, 128 gene-subsets, ∼36% (46 out 128) were found in the transcriptome analysis (Ma et al., 2021) (Fig 4C, Table S4). This suggests that the expression profile obtained through genetic dissection overlapped with transcriptome results (see DISCUSSION).

Conditional Knockout of CCT Genes in Circadian Neurons

To investigate the function of CCT genes in circadian neurons with our conditional knockout system, we knocked out all 67 (45 genes identified above and additional 22 genes reported previously) CCT genes in Clk856-labeled clock neurons by C-cCCTomics.

In the pilot screen, we monitored fly activity by video recording (Dai et al., 2019) and analyzed rhythmic behavior under LD and DD conditions. We analyzed morning anticipation index (MAI) and evening anticipation index (EAI) under the LD condition (Harrisingh et al., 2007; Im and Taghert, 2010)(Fig 5A), power, period and arrhythmic rate (AR) under the DD condition. Fly activities tended to rise rapidly after ZT22.5 at dawn and ZT10 at dusk. Thus, we added two more parameters to describe the anticipatory activity patterns of LD condition. Morning anticipation pattern index (MAPI) was defined as the difference between Pi_a[arctan(ZT20.5∼ZT22.5 activity increasing slope)] and Pi_p[arctan(ZT22.5 ∼ZT24 activity increasing slope)], M(Pi_a-Pi_p). Evening anticipation pattern index (EAPI) was defined similar to MAPI (Fig 5A, see Methods). Pi_a and Pi_p were positive, while MAPI and EAPI were negative, for wild type (wt) flies as their activities gradually increases at dawn or dusk at increasing rates. Knocking out Pdf or Pdfr in clock neurons phenocopied their mutants with lower MAI, advanced evening activity, low power, high arrhythmic rate and shorter period (Hyun et al., 2005; Lear et al., 2005; Renn et al., 1999) (Table S6, Fig S9A-S9D). The MAI-decreasing phenotype of Dh31 knockout was also reproduced in this pilot screen (Goda et al., 2019) (Table S6). All the above results verified the effectiveness of C-cCCTomics. Unexpectedly, additional experimental replications with full controls using Cas9.M9 revealed that leaky expression of Cas9.M9 and sgRNA might have caused disruption of Dh31, Dh44, Pdf and Pdfr (Fig S9A-S9D) (see DISCUSSION), which was not suitable for neuronal specific mutagenesis of some genes. Therefore, in the following work we primarily focused on Cas9.M6 instead.

Fig 5.

Analysis of the newly defined parameters MAPI and EAPI showed that control flies (Clk856-GLA4>UAS-Cas9.M9) had negative EAPI but slightly positive MAPI. The positive MAPI of control flies in this screen might be caused by Cas9.M9 toxicity. Only the Pdf and Pdfr clock-neuron knockout flies showed positive EAPIs, indicating an advanced evening activity (Table S6, Fig S9A, S9C). nAChRα1, MsR1, mAChR-B, and CNMa cKO flies had highest MAPI values (Table S6). We further confirmed their phenotypes using Cas9.M6 which revealed that CNMa plays a role in regulating morning anticipatory activity (Fig S10A).

Regulation of Morning Anticipation by CNMa-positive DN1p Neurons

Conditionally knocking out CNMa in clock neurons advanced morning activity (Fig 5B, Fig S10B) and increased MAPI (Fig 5C, Fig S10A, S10C), leaving the power and period intact in male flies (Fig 5D, 5E). The same advanced morning activity phenotype were also observed in female flies (Fig S10D-S10G). To further validate this phenotype, we generated a CNMa knockout (CNMa^KO) line by replacing its whole coding region with an attP-splicing adaptor element (Deng et al., 2019) (Fig 5F). Both male and female CNMa^KO flies exhibited the same phenotypes as seen in the CNMa cKO (Fig 5G-5I).

Previous studies have found CNMa expression in DN1 neurons (Abruzzi et al., 2017; Jin et al., 2021; Ma et al., 2021). Our intersection showed four DN1p and one DN3 CNMa positive neurons in Clk856 labelled neurons (Fig S8 - 16, Fig 4C, Table S4). Analysis with an endogenous CNMa-KI-GAL4 knockin driver showed that six pairs of CNMa neurons located in the DN1p region and three pairs located in the subesophageal ganglion (SOG) had the brightest GFP signals (Fig 6A). The anatomical features of CNMa neurons were further confirmed using stingerRed and more neurons were found in regions, the anterior ventrolateral protocerebrum (AVLP), and the antennal mechanosensory and motor center (AMMC) (Fig S11A). Dendrites of CNMa neurons were concentrated in DN1p and SOG, with their axons distributed around DN1p region, lateral horn (LH), and prow region (PRW) (Fig 6B, 6C). Using the Trans-tango strategy (Talay et al., 2017), we also found that downstream of CNMa neurons were about fifteen pairs of neurons in the SOG, five pairs of LNd neurons, one pair of DN3 neurons and six pars intercerebralis (PI) neurons (Fig 6D, arrowhead).

Fig 6.

Because we had found that knocking out CNMa in Clk856-GAL4 labeled neurons produced advanced morning activity, and that CNMa intersected with Clk856-Gal4 labeled neurons in 4 pairs of DN1ps and one pair of DN3 neurons (Fig S8-16), we focused on these neurons and performed more intersections. Taking advantage of a series of clock neuron subset-labeled drivers (Sekiguchi et al., 2020), we intersected CNMa-p65AD with 4 DN1 labelling drivers: GMR51H05-GAL4, GMR91F02-GAL4, Pdfr-KI-GAL4 and GMR79A11-GAL4 (Fig 6E-6H). We found two arborization patterns: Type I with two neurons whose branches projecting to the anterior region, as in CNMa∩GMR51H05, CNMa∩Pdfr, and CNMa∩GMR79A11 (Fig 6E, 5G, 6H), and type II with four neurons branching on the posterior side with few projections to the anterior region, as in CNMa∩GMR91F02 (Fig 6F).

CNMa knockout in Type I or Type II neurons (GMR51H05-GAL4, GMR91F02-GAL4, and GMR79A11-GAL4) all reproduced the MAPI-increased phenotype of clk856 specific CNMa knockout (Fig 6I). However, Type II neurons-specific CNMa knockout (CNMa ∩ GMR91F02) showed no advanced morning activity peak (Fig 6K), while Type I neurons-specific CNMa knockout did (Fig 6J), indicating that these two type I CNMa neurons are the main functional subset regulating the morning anticipation activity of fruit fly.

Pdf or Pdfr mutants have weak or no morning anticipation, which is in reversely related to the phenotype of CNMa knockout flies. We also identified two Pdfr and CNMa double-positive DN1ps, which have a type I projection pattern (Fig 6G). Reintroduction of Pdfr in Pdfr knockout background revealed that GMR51H05 and GMR79A11Gal4 drivers, which covered the main functional CNMa-positive subset, could partially rescue the morning anticipation and power phenotype of Pdfr knockout flies to a considerably larger extent than the GMR91F02 driver (Fig 6L-6M). Moreover, knocking out Pdfr in type I CNMa neurons decreased morning anticipation of flies (Fig 6N). Because the main subset of functional CNMa is also PDFR-positive, it is possible that CNMa secretion is regulated by PDF/PDFR signal.

Role of Neuronal CNMaR in Morning Anticipation

There is only one CNMa receptor reported in the fly genome (Jung et al., 2014). We generated a CNMaR^KO-p65AD line by CRISPR/Cas9 (Fig 7A) and this knockout showed advanced morning activity (Fig 7B, 7D) and increased MAPI (Fig 7C, 7E) in both sexes. CNMaR^KI-Gal4/UAS-mCD8::GFP and CNMaR^KI-Gal4/UAS-stinger::Red showed expression of CNMaR across the whole brain (Fig 7F, Fig S11B), especially in DN1p, DN3, the PI and the SOG. The dendrite arborization and synaptic projections of CNMaR neurons also covered broad regions (Fig 7G, 7H), at the PI, the SOG, the posterior ventrolateral protocerebrum (PVLP) and the central complex (CC). Further conditional knockout of CNMaR in neurons by C-cCCTomics phenocopied CNMaR^KO-p65AD phenotype (Fig 7I-7L). These results indicate that CNMaR is similar to CNMa in regulating morning anticipation.

Fig 7.

Discussion

Conditional CCTomics Strategies and Toolkit

We have generated conditional gene manipulation systems based on Flp-out/GFPi or CRISPR/Cas9. cCCT based gene deletion after heat-shock or mifepristone (RU486) eliminated most GFP signals, and pan-neuronal constitutive expression of shRNA^GFP or flippase disrupted seven out of eight tested genes completely while targeting of SIFa^EGFP.FRT achieved 96±3% efficiency. Although the recombination of genetic elements is relatively cumbersome when using cCCTomics, it is worthwhile to apply this method to specific genes given its high level of efficiency. While two UAS-sgRNA libraries have been established, one primarily targeting kinases (Port et al., 2020) and the other targeting GPCRs (Schlichting et al., 2022), both libraries only cover a portion of CCT genes, and are thus insufficient for manipulating all CCT genes. The development of C-cCCTomics, however, makes CCT gene manipulation as simple as RNA interference. Furthermore, the use of modified Cas9.M6 or Cas9.M9 highly enhances the efficiency of gene disruption in the nervous system, allowing for efficient manipulation of all CCT genes in a cell-specific manner.

The toxicity of CRISPR/Cas9 depends on the Cas9 protein (Port et al., 2014). When expressed pan-neuronally in nSyb-GAL4 (R57C10-GAL4, attP40), Cas9.M9 slightly reduced viability, while the expression of other Cas9 variants had no significant effect on viability (Fig S12). Although Cas9.M9 showed leaky-expression-efficiency, this was not a problem with Cas9.M6, which successfully disrupted Dh31, Dh44, Pdf, and Pdfr (Fig S9). A more restricted expression of Cas9.M9 with lower toxicity is necessary for better somatic gene manipulation in the future.

CCT of Clock Neurons

Intersecting Clk856-Gal4 or Clk856-p6AD with CCTomics, we identified 45 CCT genes in Clk856 labelled clock neurons. Clock neurons appear highly heterogeneous both in our intersection dissection and in a previous transcriptomic analysis (Abruzzi et al., 2017; Ma et al., 2021). Comparing these two CCT gene expression profiles in clock neurons, 45 out of 128 gene-subsets are identical. The accuracy of our genetic intersection is limited by two possibilities: 1) KI-LexA may not fully represent the expression pattern of the corresponding gene, and 2) the efficiency of STOP cassette removal in the Flp-out strategy is limited. Moreover, the leakage of LexAop-GFP may result in unreliable labelling in split-lexA strategy.

We have also observed that the expression profile of Pdfr in clock neurons is inconsistent across studies, with different clock subsets being identified(Hyun et al., 2005; Im and Taghert, 2010; Lear et al., 2005; Ma et al., 2021; Mertens et al., 2005; Shafer et al., 2008) . The variability in labelling clock neurons with Pdfr transgenic GAL4s, KI-myc tag, antibodies, and transcriptomic anatomy reflects the limitations of each approach. The expression of Pdfr in LNds and DN1s is considered reliable as they are labeled by all strategies. Our intersection dissection is most closely aligned to Pdfr antibody-labelled neurons (Hyun et al., 2005). Both genetic drivers and transcriptomic analysis contribute to our knowledge of the expression profile of neurons. The physiological significance of each gene in a particular neuron should be further investigated by genetic manipulation.

Regulation of Rhythmic Behavior by CCT genes

Multiple attractive genes have been identified in our functional screen of CCT genes in clock neurons: for example, knocking out of VGlut weakens both morning anticipation and rhythmic strength (Table S7). In further screening of brain regions, we have narrowed down the morning anticipation regulation role of VGlut in R18H11-GAL4 labeled neurons (Table S8). VGlut in these neurons has also been reported to regulate sleep in Drosophila (Guo et al., 2016). Its downstream neurons may be the PI neurons or LNvs (Barber et al., 2021; Guo et al., 2016).

Moreover, the deficiency of the neuropeptide CNMa results in advanced morning activity. We have validated that two Pdfr and CNMa double-positive DN1p neurons mainly regulate this process through intersectional manipulation of CNMa. Knockout and reintroduction of Pdfr in these neurons have verified that Pdfr partially functions in DN1p CNMa neurons, and PDF increases cAMP level in Pdfr positive neurons (Shafer et al., 2008), suggesting the regulation of CNMa signaling by PDF signaling. Furthermore, given that the morning anticipation vanishing phenotype of Pdf or Pdfr mutant is opposite to that of CNMa knockout flies, we consider the two signals to be antagonistic. However, knocking out CNMaR in Clk856 labelled clock neurons showed no significant phenotype (Table S7), whereas the mutant and pan-neuronal knockout flies had similar phenotypes to CNMa knockout flies, suggesting its role in the circadian output neurons. Previous studies have indicated that CNMa integrate thermosensory inputs to promote wakefulness, and CNMaR is thought to function in Dh44 positive PI neurons (Jin et al., 2021), a subset of circadian output neurons. To gain a deeper understanding of the downstream effects of DN1p CNMa postive neurons, further analysis focusing on specific brain regions is necessary.

We have also reproduced phenotypes of Pdf, Pdfr, and Dh31 mutant flies with C-cCCTomics as previous studies. Surprisingly, only five genes are functional among all sixty-seven CCT genes in this prior screen. This may be caused by limitations of the simple behavioral paradigm, single gene manipulation, and single GAL4 driver. For example, switching of light condition from L: D = 12h: 12h to L: D = 6h:18h, AstC/AstC-R2 would suppress flies’ evening activity intensity to adapt to the environmental change (Díaz et al., 2019), and only double knockout of AChRs and mGluRs in PI neurons can possibly result in alteration in behavioral rhythms (Barber et al., 2021). Further diversified functional analysis of CCT genes in clock neurons is required for clock circuit dissection.

Methods

Fly Lines and Rearing Conditions

Flies were reared on standard corn meal at 25 ℃, 60% humidity, 12 h light: 12 h dark (LD) cycle. For flies used in behavior assays, they were backcrossed into our isogenized Canton S background for 5 to 7 generations. For heat induced assays, flies were reared at 20℃. All CCT attP KO lines and CCT KI driver lines were previous generated at our lab (Deng et al., 2019).

Clk856-GAL4 and GMR57C10-GAL4 driver lines were gifts from Donggen Luo Lab (Peking University). 13XLexAop2 (FRT.stop) myr::GFP was gift from Rubin Lab.

C-cCCTomics sgRNAs design

All sgRNAs target at or before functional coding regions (e.g. GPCR transmembrane domain, synthetase substrate binding domain) of each CCT genes. For each gene, about 20 sgRNAs with specific score ≥ 12 were firstly designed at CRISPRgold website (Chu et al., 2016; Graf et al., 2019), then their specificity and efficacy were further valued in Optimal CRISPR target finder(Gratz et al., 2014), E-CRISPR(Heigwer et al., 2014) and CCTop (Stemmer et al., 2015) system. The first three highest efficacy sgRNAs with no predicted off-target effect were selected. All selected sgRNAs are listed in Table S3.

Molecular Biology

All cCCTomics knockin (KI) lines and C-cCCTomics transgenic flies were generated through phiC31 mediated attB/attP recombination, and the miniwhite gene was used as selection marker.

For cCCTomics KI lines, backbone pBSK-attB-FRT-HpaI-T2A-EGFP-FRT was modified from pBSK-attB-loxP-myc-T2A-Gal4-GMR-miniwhite (Deng et al., 2019). Myc-T2A-GAL4 cassette was removed by PCR amplification while first FRT cassette was introduced. Second FRT cassette was inserted by T4 ligation between SpeI and BamHI. T2A-EGFP was cloned from pEC14 and was inserted into the backbone between two FRT cassettes. All gene spans, except for stop codon, deleted in CCT attP KO lines were cloned into pBSK-attB-FRT-HpaI-T2A-EGFP-FRT at HpaI site.

For C-cCCTomics UAS-sgRNA lines, backbone pMsgNull was modified from pACU2 (Han et al., 2011). Synthetic partial fly tRNA^Gly sequence was inserted between EcoRI and KpnI. An irrelevant 1749 bp cassette amplified from pAAV-Efla-DIO-mScarlet (addgeen#130999) was inserted between EagI and KpnI. All sgRNA spacers were synthesized at primers, and “E+F” sgRNA scaffold and rice tRNA^Gly was amplified from a synthetic backbone PM04. Finally, gRNA-tRNA^Gly cassettes were cloned into pMsgNull between EagI and KpnI by Gibson Assembly.

All UAS-Cas9 variants generated in this research were cloned into vector pACU2 and all Cas9 sequence were amplified from hCas9 (addgene#41815). Human codon optimized Cas9 was cloned into pACU2 to generate UAS-Cas9.HC. UAS-Cas9.M0 was modified from UAS-Cas9.HC by introduce a 3x HA tag after NSL and replace the SARD linker with the 3×GGSGP linker (Zhao et al., 2016). UAS-Cas9.M6 and UAS-Cas9.M9 were designed as HMGN1-Cas9-UPD and HMGN1-Cas9-HB1 respectively. All these chromatin-modulating peptides were linked with Cas9 by 3x GGSGP linker.

CNMaR^KO-p65AD is generated by replace the coding region of the first exon with T2A-p65AD by CRISPR/Cas9, the T2A-p65AD was linked in frame after first ten amino acids. Spacers of gRNAs used to break the targeted CNMaR region were 5’-GCAGATTTCAGTTCATCTTT-3’, 5’-GGCTTGGCAATGAC TATATA-3’.

Gene expression quantitation and high-throughput sequencing

Female flies were gathered six to eight days post-eclosion for gene expression quantification and high-throughput sequencing. Fly heads were isolated by chilling them on liquid nitrogen and subsequent shaking. mRNA extraction was performed using Trizol according to a previously established protocol (Green and Sambrook, 2020). Genomic DNA was removed, and cDNA was synthesized using a commercial kit (TIANGEN#DP419). For real-time quantitative PCR, at least one PCR primer was designed to overlap with the sgRNA target site.

Genomic DNA from fly heads was extracted using a standard alkali lysis protocol(Huang et al., 2009). Genomic regions approximately 130 to 230 bp in length, centered around the sgRNA target site, were amplified by PCR employing Q5 polymerase (NEB#M0494). Subsequently, libraries were prepared using the BTseq kit (Beijing Tsingke Biotech Co., Ltd.). These libraries were pooled and subjected to sequencing on the MiSeq platform (Illumina). Analysis of the libraries was conducted using Crispresso2(Clement et al., 2019).

Generation of KI and Transgenic lines

Generation of cCCTomics KI, CNMa^KI-p65AD, and CNMaR^KO-p65AD lines are the same as generation of CCTomics KI driver lines as previously described (Deng et al., 2019). To generate C-cCCTomics UAS-sgRNA or UAS-Cas9 variants lines, attB vectors were injected and integrated into the attP40, attP2 or VK00005 through phiC31 mediated gene integration.

All flies generated in this research were selected by mini-white and confirmed by PCR.

Behavioral assays

Unmated male or female flies of 4 to 5 days were used in circadian rhythm assays. Before measurement, flies were entrained under 12 h light: 12 h dark cycle at 25℃ for at least 3 days and then transferred to dark-dark condition for 7 days.

Virgin flies of 4 to 5 days were used in sleep assays. Flies were entrained to a 12 h light: 12 h dark cycle at 25℃ for 2 days to eliminate the effect of CO2 anesthesia before sleep record. Sleep was defined as 5 min or longer immobility (Hendricks et al., 2000; Shaw et al., 2000) and analyzed by in-house scripts as previously described ((Dai et al., 2021; Dai et al., 2019; Deng et al., 2019).

Locomotion was obtained as previously described (Dai et al., 2021). Locomotion activity was measured and analyzed by Actogram J plugin (Dai et al., 2019). Morning anticipation index and evening anticipation index were defined as the ratio of last 3 hours activity before light-on or light-off accounts to last 6 hours activity before light-on or light-off (Index=sum(3hrs)/sum(6hrs)) (Harrisingh et al., 2007; Im and Taghert, 2010) and analyzed by an in-house python script.

Heat Shock and Drug Treatment

For hsFLP mediated conditional knockout, flies of 4 to 6 days were heat shock at 37℃ during ZT10 to ZT12 for 4 days. And they were reared at 20℃ for another 4-days and then dissected.

For mifepristone (RU486) induced conditional knockout, flies of 4 to 6 days were treated with 500 μM RU486 mixed in corn food and then dissected 4 days later.

Immunohistochemistry and Confocal Imaging

For all imaging without staining, adult flies were anesthetized on ice and dissected in cold phosphate-buffer saline (PBS). Brains or ventral nerve codes (VNC) were fixed in 2% paraformaldehyde (weight/volume) for 30 min, washed with washing buffer (PBS with 1% Triton X-100, v/v, 3% NaCl, g/mL) for 7 min three times and mounted in Focusclear (Cell Explorer Labs, FC-101).

For imaging with staining, brains and VNCs were fixed for 30 min, washed for 15 min three times. Then they were blocked in PBSTS, incubated with primary antibodies, washed with washing buffer, incubated with second antibodies and mounted as described previously(Dai et al., 2021; Dai et al., 2019).

All brains or VNCs were imaged on Zeiss LSM710 or Zeiss LSM880 confocal microscope and processed by Imaris.

The following primary antibodies were used: mouse anti-PDF (1:200, DSHB), rabbit anti-TH (1:1000, Novus Biologicals), rabbit anti-LK (1:1000, RaoLab, this paper). Rabbit anti-DSK (1:1000) was a gift from Dr. C. Zhou Lab (Institute of Zoology, Chinese Academy of Science) (Wu et al., 2020). The following secondary antibodies were used: AlexaFluor goat anti-mouse 488 (1:1000, Invitrogen), AlexaFlour goat anti-rabbit 488/633 (1:1000, Invitrogen).

For Fig 2, number of TH positive neurons were counted with Imaris Spots plugin.

Quantification and Statistics

MAI, MAPI, EAI, EAPI, power and period were calculated by python or R scripts. ZT0 was set as the time point when light was on and ZT12 was set as the time point for light-off. Activity bins started at ZT0 and each was calculated as a sum of the total activity within 30min. Flies were regarded as death and removed if their activity value within last 2 bins was 0. A representative 24hr activity pattern was the average between corresponding activity bins from 2 consecutive days. To minimize effects from singular values, each flies‣ activity was normalized using the following formula:

bi means the activity value for a certain bin. min(b0,…,b48) means the minimal bin value within 24hr and max(b0,…,b48) means the maximal bin value within 24hr. Nor_bi means the final normalized bin value for a certain bin from a given fly. Normalized activity was used for following analysis.

Morning activity arise (M_arise) was defined as the radian between the activity curve (ZT21-ZT22) and the time coordinate. Morning activity plateau (M_plateau) was defined as the radian between the activity curve (ZT22.5-ZT24) and the time coordinate. Evening activity arise (E_arise) was defined as the radian between the activty curve (ZT8-ZT11.5) and the time coordinate. Evening activity plateau (E_plateau) was defined as the radian between the activity curve (ZT10.5-ZT12) and the time coordinate. Morning anticipation pattern index (MAPI) was calculated by subtracting M_arise from M_plateau and Evening anticipation pattern index (EAPI) was calculated by subtracting E_arise from E_plateau.

The original activity data from 7 consecutive days in dark-dark condition was used for power and period calculation as described (Geissmann et al., 2019). Each flies‣ periodogram was calculated based on Chi-Square algorithm (Sokolove and Bushell, 1978) and flies with a null power value were regarded as arrhythmic.

All statistical analyses were carried out with Prism 8 (Graphpad software). The Kruskal-Wallis ANOVA test followed by Dunn’s posttest was used to compare multiple columns.

Author Contributions

R.M. and J.Y. performed majority of experiments and data analysis. B.D., J.Y., and R.M. found advanced morning activity phenotype and carried out experiments of CNMa/CNMaR. R.M., Y.D., and S.D. carried out experiments of cCCTomics fly generation. R.M., J.Y., X.D., and Y.R. wrote the manuscript.

Competing Interest Statement

The authors declare no competing interest.

Acknowledgements

We are grateful to Xiaofei Liu, Xueying Wang and Yile Jiao for sgRNA design and cloning, to Linyi Zhang, Wenli Xu, Yuqing Gong and Quiquan Chen for assistance with behavior assays, to Xiao Dong for cloning, to Xinwei Gao for assistance with imaging, to Pingping Yan, Lan Wang, Yonghui Zhang and Haixia Zeng for fly rearing, to Enxing Zhou and Wei Yang for Drosophila activity tracing scripts, to Donggen Luo, Chuan Zhou, Gerald M. Rubin, Vienna Drosophila RNAi Center and Bloomington Drosophila Stock Center for flies, to Yuh-Nung Jan for pACU2 construct.

Supplementary Information

Fig S1.

Fig S2.

Fig S3.

Fig S4.

Fig S5.

Fig S6.

Fig S7.

Fig S8

Fig S9.

Fig S10.

Fig S11.

Fig S12.

Table S1

Table S2

Table S3

Table S4

Table S5

Table S6

Table S7

Table S8