Long noncoding RNA PAHAL modulates locust behavioural plasticity through the feedback regulation of dopamine biosynthesis

Summary

Some long noncoding RNAs (lncRNAs) are specifically expressed in brain cells, implying their neural and behavioural functions. However, how lncRNAs contribute to neural regulatory networks governing the precise behaviour of animals is less explored. Here, we report the regulatory mechanism of the nuclear-enriched lncRNA PAHAL for dopamine biosynthesis and behavioural adjustment in migratory locusts (Locusta migratoria), a species with extreme behavioral plasticity. PAHAL is transcribed from the sense (coding) strand of the gene encoding phenylalanine hydroxylase (PAH), which is responsible for the synthesis of dopamine from phenylalanine. PAHAL positively regulates PAH expression resulting in dopamine production in the brain. In addition, PAHAL modulates locust behavioral aggregation in a population density-dependent manner. Mechanistically, PAHAL mediates PAH transcriptional activation by recruiting serine/arginine-rich splicing factor 2 (SRSF2), a transcription/splicing factor, to the PAH proximal promoter. The co-activation effect of PAHAL requires the interaction of the PAHAL/SRSF2 complex with the promoter-associated nascent RNA of PAH. Thus, the data support a model of feedback modulation of animal behavioural plasticity by an lncRNA. In this model, the lncRNA mediates neurotransmitter metabolism through orchestrating a local transcriptional loop.

Keywords:

Gene expression – Small interfering RNAs – DNA transcription – Long non-coding RNAs – Transcriptional control – Nymphs – Metabolic pathways – Locusts

Introduction

Long noncoding RNAs (lncRNAs) are transcripts comprising > 200 nucleotides and possessing minimal or non-existent protein-coding capacity [1] and are increasingly recognised as key players in numerous cellular processes [2,3]. LncRNAs are the main products of RNA polymerase II, often polyadenylated and processed through splicing [3]. Growing evidence shows that most lncRNAs may be functionally relevant. LncRNAs engage in various biological processes, such as X-chromosome inactivation [4–6], DNA damage response [7,8], differentiation and development [5,9,10], metabolism [11,12] and immunity and disease response [12–14].

The brain and neuronal specificity of lncRNA expression has prompted the exploration of the potential roles of lncRNAs in neuronal development and cognitive and behavioural regulation [10,15–21]. In Drosophila species, cytoplasmic lncRNA yar regulates sleep behaviour [22]. The lncRNA CRG exhibits spatiotemporal specific expression patterns within the central nervous system; in addition, CRG affects the locomotor behaviour of Drosophila by positively regulating a neighbouring gene that encodes a Ca²⁺/calmodulin-dependent protein kinase [23]. LncRNAs can regulate gene expression through distinct modes [19,24–26], such as their association with transcription factors and/or chromatin modification factors [3,5,27–29]. In mice, Gomafu mediates anxiety-like behaviour by maintaining the polycomb repressive complex 1 at the promoter of the schizophrenia-related gene beta crystallin [30]. Many more findings implicate the widespread involvement of lncRNAs in neuronal response and neurological diseases [10,17,18,29,31]. Despite the recognised roles of lncRNAs in the neuronal system and behaviour, how neuronal and behavioural responses to environmental stimuli are modulated at the cellular and organismal levels by lncRNAs remains incompletely understood.

The metabolic and signal transduction pathways of dopamine (DA) are important for the behavioural responses of animals. DA is responsible for motor control, learning and memory and is associated with several important neural diseases [32,33]. Phenylalanine hydroxylase (PAH, also known as Henna in Drosophila and Locusta) catalyses the synthesis of tyrosine from phenylalanine and affects the production of DA and other bioamines in brain [34–37] (Fig 1A). We previously elucidated the metabolic and signalling pathways of DA-regulated aggregation behaviour in the migratory locust Locusta migratoria, which exhibits density-dependent behavioural plasticity [38–40]. Gregarious (G) locusts display high mobility and conspecific attraction. By contrast, solitarious (S) locusts are tardy and repulsive to other conspecific individuals [41]. Locusts can reversibly and rapidly shift their behaviours between the S and G phases in response to population density changes. Specifically, G nymphs exhibit significant behavioural solitarization upon 1 h isolation and full solitarization upon 16 h isolation. S nymphs display significant behavioural gregarization upon 32 h crowding [42]. Therefore, locusts are considered ideal models for studying animal behavioural plasticity [38,41,43]. PAH expression affects the catecholamine metabolic pathway (e.g., DA synthesis) and controls the behavioural phase shift of locusts [39,40]. Specifically, miR-133 targeting PAH coding region repressed PAH expression, reduced DA production and drove locust soliterization while activated, and vice versa while inhibited [36]. However, the molecular mechanism underlying the fine-tuning of PAH expression for the mediation of brain DA dynamics in behavioural shifts remains ambiguous.

In this study, we explored a new regulatory mechanism, which involves an lncRNA in the DA metabolic pathway and underlies the elaborate control of locust behavioural plasticity. We discovered a novel PAH-activating lncRNA (PAHAL), which is sense to the ancestral PAH gene. PAHAL tightly controls the metabolic regulation of DA biosynthesis in the brain and mediates reversible behavioural changes in response to population density. Mechanistic analysis revealed that PAHAL can enhance PAH expression by recruiting the transcription activator serine/arginine-rich splicing factor 2 (SRSF2) to the PAH promoter-proximal region. These results demonstrate that the lncRNA orchestrates gene expression for DA biosynthesis by regulating the local feedback transcription of PAH. Thus, our findings provide new insights into the role of lncRNAs in the fine tuning of neuronal and behavioural responses cued by environmental stimuli.

Results

Sense lncRNA PAHAL is expressed from the intron/exon of the PAH gene locus

We identified PAHAL, a novel lncRNA that overlaps with the PAH gene, in the genome and transcriptomes of the locust. We validated the transcriptional origin of PAHAL relative to that of PAH through 5′ and 3′ RACE. The locust PAH gene encodes PAH by a 1,652 bp-long transcript comprising eight exons (Fig 1B). PAHAL is unspliced and is 2,612- nucleotide (nt) long (S1 Fig). The PAHAL sequence covers the sixth exon and part of the two introns flanking the exon in the PAH gene (Fig 1B and S1 Fig). Sense-specific reverse transcription polymerase chain reactions (RT-PCRs) confirmed that PAH and PAHAL are different transcripts (Fig 1C). The 5′ exonuclease digestion assay indicated that PAHAL is 5′ capped (Fig 1D), and the RT-PCR from the 3′-end tailed-RNA demonstrated that PAHAL is a polyadenylated transcript (Fig 1E). These results indicate that the transcription direction of PAHAL is the same as that of the canonical PAH, and PAHAL transcription is sense to PAH gene (Fig 1B).

Open reading frame (ORF) analysis showed that 19 short ORFs are present in PAHAL. The longest ORF (201 nt) is located at most 3′ end and annotated as the long terminal repeat region of a Gypsy transposable element. No match to any known genomic sequence of insect species, except the contained PAH exon six (262 nt), was identified. Therefore, the lncRNA sequence is not conserved among insect species. The coding capacity assessment by using the Coding Potential Assessment Tool (CPAT) [44] showed that PAHAL lacks protein-coding capacity (Fig 1F). An in vitro translation assay demonstrated that no protein was produced by PAHAL expression (Fig 1G). Therefore, PAHAL is a sense noncoding RNA generated from the PAH locus.

PAHAL controls dopamine biosynthesis by regulating the PAH expression

We knocked down PAHAL through dsRNA (double-stranded RNA) and siRNA (small interfering RNA) interferences to test the regulatory relationship between PAHAL and PAH in the G locust brain. The dsRNA and siRNAs of PAHAL were targeted to the right intron flanking the sixth exon in the PAH gene and ought to be specific for PAHAL not influence the pre-mRNA of PAH. Primer sequences of these dsRNAs and siRNAs can be found in S3 Table. PAHAL dsRNA injection in the brain significantly repressed the PAHAL expression (t test: P = 0.002, N = 9). The PAHAL knockdown reduced the PAH mRNA level by 91% (t test: P = 0.006, N = 9) and PAH protein level by 96% (t test: P < 0.001, N = 4; Fig 2A). The PAHAL-specific knockdown through siRNA interference provided similar results (Fig 2B). However, PAH knockdown had no effect on PAHAL level (Fig 2C). These findings imply that PAHAL is involved in the positive regulation of PAH in locust brains.

Thereafter, we applied brain transcriptome sequencing to investigate the downstream pathways affected by the PAHAL expression. The RNA-seq data showed that 52 genes were significantly altered by PAHAL knockdown (Fisher’s exact test: P < 0.05, fold change > 2, N = 3; Fig 2D and S1 Table, Supplemental File 1). PAH was the most differentially expressed gene (Fisher’s exact test: P = 2.5 × 10⁻³⁷, N = 3). The functional clustering of the 52 genes revealed that the phenylalanine metabolism pathway to include the PAH gene was the most significantly enriched pathway (Fisher’s exact test: P = 4.9 × 10⁻⁴, N = 3; Fig 2E).

PAH encodes a dual function enzyme, that hydroxylates both phenylalanine for tyrosine synthesis and tryptophan for the synthesis of 5-hydrotryptophan (5-HTP), the precursor of serotonin in insects [34,45,46]. Therefore, we then quantified the levels of tyrosine, a direct product of PAH catalysis, and the six downstream metabolites in the catecholamine metabolic pathways, namely, l-dopa, DA, 5-hydroxytryptophan, serotonin, tyramine and octopamine by liquid chromatography–mass spectrometry (LC–MS) (Figs 1A and 2F). The PAHAL knockdown significantly decreased the concentrations of tyrosine (t test: P < 0.001, N = 10), and metabolites (i.e. l-dopa, t test: P = 0.001, N = 8) and DA (t test: P < 0.001, N = 7), in one downstream branch pathway but not those (i.e. tyramine and octopamine) in another one in the brain. Such knockdown also reduced the content of 5-hydroxytryptophan (t test: P < 0.001, N = 10) but not that of the downstream product, such as serotonin (Fig 2F). These results indicate that PAHAL affects the DA synthesis in the catecholamine metabolic pathway.

PAHAL is a positive regulator of PAH expression in locust behaviour

We tested the functional relationship between PAHAL and PAH by monitoring their expression levels in five tissues from G nymphs (Fig 3A). PAHAL and PAH were highly expressed in the locust brain. G locusts display high mobility and sociable. By contrast, S locusts are sedentary and live at very low densities [41]. The G and S individuals possess distinct, population density-dependent behavioural features [38]. Thus, we compared the expression levels of the two transcripts in the brains of G and S locusts (Fig 3B). The expression of PAHAL and PAH in the G locusts were 17.7- (t test: P = 0.004, N = 5) and 2.2-fold higher (t test: P < 0.001, N = 7), respectively, than those in the S locusts (Fig 3B). Accordingly, PAHAL and PAH were upregulated in the G locust brain. The brains of G locusts were subjected to fluorescence in situ hybridisation (FISH). The result showed that PAH and PAHAL transcripts are both localised to neuronal cell bodies, but as expected PAH is cytoplasmic whereas PAHAL is nuclear (Fig 3C and S2 Fig). Furthermore, PAH and PAHAL have broad expression patterns rather than being localised only in a subset of locust neurons.

We tested the time-course expression dynamics of PAHAL and PAH in the brain to determine their dynamic correlations during locust isolation and crowding treatments (Fig 3D). PAHAL was significantly upregulated at 4 h upon aggregation (t test: P = 0.002, N = 8) compared with that at 0 h. The upregulation of the PAHAL expression was sustained at 8 h (t test: P = 0.040, N = 8). By contrast, the PAHAL expression was significantly downregulated at 4 h after isolation (t test: P = 0.013, N = 8). Such expression continued to decrease at 8 h after isolation (t test: P = 0.001, N = 8). PAH exhibited the same time-course expression pattern. Thus, the expression levels of PAHAL and PAH are positively correlated during locust aggregation and isolation.

PAHAL controls phase-related behavioural transition

We investigated the role of PAHAL in behavioural regulation in the locusts using dsRNA and siRNA interference of PAHAL, respectively. Locust aggregation in an arena was measured by using an established automatic behavioural assay [40,42] (Fig 4A). Aggregation behaviour was quantified by using P_greg, a summary statistic provided by this behavioural assay and a reliable indicator of locust aggregation propensity [40]. The shift in the median P_greg value of G locusts from 0.89 to 0.23 when PAHAL was knocked down through dsRNA interference indicated a significant behavioural shift from the G to the S states (Mann–Whitney U test: P < 0.001, N_dsGFP = 30, N_dsPAHAL = 57, Fig 4B). The results from siRNA interference lead to a change of P_greg from 0.90 to 0.20 (Mann–Whitney U test: P < 0.001, N_siGFP = 33, N_siPAHAL = 35) (Fig 4C). This outcome from siRNA demonstrates an effect of PAHAL on locust aggregation similar to that from dsRNA.

Furthermore, we examined the specific parameters of the behavioural traits of locusts after the gene expression interference. In comparison with the dsGFP control treatment, the PAHAL knockdown reduced the total distance of movement by 71% (t test: P < 0.001, N_dsGFP = 30, N_dsPAHAL = 57), decreased the total duration of movement by 95% (t test: P < 0.001, N_dsGFP = 30, N_dsPAHAL = 57) and induced the reversion from attraction to repulsion, as shown by the attraction index (Mann–Whitney U test: P = 0.008, N_dsGFP = 30, N_dsPAHAL = 57; Fig 4D). We also examined the specific behavioural effect from siRNA knock down. The PAHAL interference by siRNA has behavioural effects similar to that by dsRNA in terms of mobility and conspecific attraction (Fig 4E). Overall, PAHAL effectively modulates the behavioural aggregation of the locusts.

PAHAL facilitates the transcription activation of PAH by interacting with the PAH promoter-proximal region

We determined the mechanism by which PAHAL mediates PAH transcription. The sequence analysis revealed that one nuclear localisation signal (NLS) [47] is present at the 3′ end of PAHAL but not in PAH (S3 Fig). The nuclear fractionation experiment showed that 94% of PAHAL mRNAs (possibly together with trace amount of PAH pre-mRNA) localises in the nucleus relative to nuclear RNA U6 (positive control) and cytoskeleton actin (negative control). By contrast, 97% of PAH mRNAs localised in the cytoplasm (Fig 5A). FISH results also showed that PAHAL and PAH localised in the nucleus and cytoplasm, respectively (Figs 3C and 5B). This finding indicates that these two transcripts are distinct in the subcellular location.

Whether PAHAL regulates PAH expression through a cis interaction with a PAH promoter was determined. The sequence from −1,169 nt to +89 nt relative to the PAH initiator was selected as the regulatory segment of the PAH promoter (labelled as P+5′-UTR). The luciferase assay showed that PAHAL substantially enhanced the luciferase expression driven by the PAH promoter in S2 cells (t test: P < 0.001, N = 4). By contrast, PAHAL expressed in the reverse orientation (i.e., reverse PAHAL) had no effect on PAH promoter activity (Fig 5C). Thus, PAHAL enhances the promoter activity of PAH not by acting as an enhancer element.

Moreover, the specific regulatory region in the PAH promoter targeted by PAHAL was determined by testing the transcription activities of truncated promoters coexpressing PAHAL. PAHAL significantly enhanced the activity of these promoter regions, even that of the shortest promoter (from −48 nt to +89 nt; t test: P < 0.001, N = 4; Fig 5D). Thus, the active PAHAL interacting region is close to the transcription start site (TSS) of PAH. The removal of the 5′-UTR abrogated the PAHAL-mediated transcriptional activation (Fig 5E). Therefore, PAHAL promotes the PAH expression by actively interacting with the promoter-proximal region, including the 5′-UTR, of the PAH gene.

PAHAL regulates PAH expression by binding with SRSF2

To determine proteins that interact with PAHAL, we performed RNA pulldown with brain tissue extracts (Fig 6A) and identified 11 protein candidates by mass spectrometry (S2 Table). Among these proteins, SRSF2 bound with the sense, but not antisense, strand of PAHAL (Fig 6B) and was associated with behavioural phase changes (Fig 6C and S4 Fig). Specifically, SRSF2 expression in the brain of a G locust significantly differed from that of an S locusts (t test: P < 0.001, N = 8). SRSF2 was continuously upregulated during locust crowding treatments and downregulated during isolation treatments (Fig 6C). Thus, the SRSF2 expression is positively related to the PAHAL and PAH expression during locust behavioural transition.

To determine the physical interaction between PAHAL and SRSF2 that binds to the promoter-associated nascent RNA during transcription regulation, we conducted several in vitro and in vivo RNA immunoprecipitation (RIP). In vitro RIP using an antibody against V5 tag in SRSF2-V5 protein-overexpressed cell lysis revealed significant PAHAL enrichments with SRSF2 (S5A Fig and Fig 6D, t test: P = 0.010, N = 6). In vivo RIP with brain tissues showed that endogenous PAHAL mRNA and SRSF2 were significantly enriched by the SRSF2 antibody (t test: P = 0.004, N = 5 for enriched PAHAL, Fig 6D; t test: P = 0.029, N = 5 for SRSF2, S5B Fig). Thus, the SRSF2 protein physically binds with PAHAL RNA.

Thereafter, we examined the regulatory effects of SRSF2 on PAHAL-mediated transcriptional activation through the SRSF2 knockout and overexpression. We first used SRSF2 protein-depleted cells generated from mouse embryonic fibroblasts (SRSF2-MEFs); an HA-tagged SRSF2 gene replaced the endogenous gene and expressed from a tetracycline (tet)-off promoter, thereby enabling the elimination of the protein by adding the tet analogue DOX [48,49]. The PAH promoter activity was significantly inhibited by 67% by SRSF2 knockout (One-Way ANOVA: P < 0.001, N = 6). The enhanced activity from co-transfected PAHAL was abrogated (One-Way ANOVA: P < 0.001, N = 6) by SRSF2 knockout (Fig 6E). By contrast, the SRSF2 overexpression in S2 cells significantly increased the PAH promoter activity by 56% (One-Way ANOVA: P< 0.001, N = 5). The elevated promoter activity was further enhanced by 91% (One-Way ANOVA: P < 0.001, N = 5) with PAHAL co-transfection. The effects of the SRSF2 and PAHAL expression exhibited a significant interaction in the PAH promoter activation (Mann–Whitney U test: P < 0.001, N = 5, Fig 6F). Therefore, SRSF2 substantially promoted the PAHAL-mediated cis-activation effects.

PAHAL–SRSF2 complex binds with sequence specificity for transcriptional regulation

We determined the specific sites for SRSF2 interaction in the PAHAL sequence. The domain mapping of sequence fragments containing the 3′ end of PAHAL revealed that all the fragments interacted with SRSF2 (Fig 7A, left panel). The mapping analysis of sequence fragments covering the 5′ end of PAHAL revealed that only the fragment that simultaneously covered the 3′ end (i.e., the last 245 nt) of PAHAL bound with SRSF2 (Fig 7A, middle panel). Two overlapping fragments at the 3′ terminal of PAHAL spanning 1,968–2,613 exhibited SRSF2 binding (Fig 7A, right panel). Thus, the sequence at the most 3′ end of PAHAL is necessary for PAHAL–SRSF2 binding.

Furthermore, we analysed the mechanism by which PAHAL and SRSF2 interaction facilitated the transcription activation of the PAH gene. SRSF2 facilitates transcription by binding to a high-affinity binding site (exonic-splicing enhancer, i.e., ESE) of nascent RNA [49]. We found three ESE motifs in the 5′-UTR of PAH (S5C Fig). The PAHAL-mediated activation of the PAH promoter activity was disrupted by the mutation of the three ESEs; by contrast, a single ESE mutation did not eliminate the activation effect of PAHAL (Fig 7B). Therefore, the regulatory function of PAHAL involves interaction with the ESEs of the PAH 5′-UTR. The luciferase assay further demonstrated that the three factors, namely, PAHAL, SRSF2 and tandem ESEs, had significant interactions in the effects of transcriptional regulation (One-Way ANOVA: P < 0.001, N = 6). The tandem ESEs exhibited the greatest effects among the three factors (Fig 7C). These results indicate that PAHAL directs SRSF2 to the ESEs of PAH 5′-UTR to promote PAH transcription.

Finally, we validated the role of SRSF2 in the transcriptional activation of PAH through RNAi. SRSF2 knockdown significantly reduced the mRNA level of PAH (t test: P = 0.006, N = 6) and protein level of PAH by 89% (t test: P = 0.037, N = 3, Fig 7D). Thus, SRSF2 regulates PAH transcription in vivo.

Discussion

In our study, we have identified a novel sense lncRNA named PAHAL in the PAH locus in the catecholamine metabolic pathways, and have revealed its specific role in regulating DA biosynthesis and locust behavioral aggregation. Moreover, we have also characterized the mode of transcriptional activation orchestrated by this lncRNA in the biosynthetic regulation. PAHAL exhibits a distinct mode of cis-regulation transcription, which is verified in locust, fruit fly and mice in vivo and in vitro. DA has been confirmed as a crucial modulator for locust behavioural phase transition [36,40]. Thus, our present study established the novel linkage between genetic and epigenetic modulation of behavioural plasticity of locust regulated by DA.

PAHAL knockdown attenuated the PAH expression and caused a behavioural shift toward S traits. Furthermore, PAH catalyses the reaction from phenylalanine to tyrosine and affects the downstream metabolic pathways of several bioamine neurotransmitters. Immunocytochemical studies show that aminergic neurons present wide distribution in the locust brain [50–52]. Specifically, most dopaminergic neurons are in the upper division of the central body [51]. Serotoninrgic neurons distribute 6 groups that innervate the central complex (groups l-5) [50]. Our study indicates that PAH and PAHAL are expressed in almost all neurons in the locust brain (Figs 3C and 5B). However, the transcriptome sequencing and LC–MS measurements illustrate that PAHAL predominantly affects the DA synthetic pathway in the locust brain (Fig 2), although the results do not exclude non-PAH players in DA metabolism and PAH’s role in other processes. Thus, this study establishes the regulatory role of PAHAL in the orchestration of the gene expression and DA metabolism to modulate animal behavioural transition in response to environmental stimuli.

This study also reveals the regulatory relevance of lncRNAs in strictly controlled genetic programs. Animal behaviours are generally deliberately tuned in response to a specific environmental cue. This response requires triggering the precise spatial–temporal regulation of the expression of specific genes or genetic pathways and the production of specific neurotransmitters [33,40]. The mode of cis-regulation of PAHAL lncRNA exhibits common transcription mechanism, which is verified in locust, fruit fly and mice. Numerous lncRNA loci act as local regulators and influence the expression of nearby genes through cis-regulation [24,25]. These lncRNAs with cis-regulatory roles may arise from diverse ways, such as from promoters or enhancers, antisense transcripts or from within introns or near TSSs of other host genes [25]. For example, about half fraction of antisense transcripts are noncoding RNAs [25]. In another study, genetic manipulation in mouse cell lines revealed that, among 12 genomic loci that produce lncRNAs, five loci influence the expression of a neighboring gene in cis [24]. These studies have revealed the prevalence of lncRNA-mediated cis regulation [53,54]. Cis-interference or silencing by intragenic lncRNAs is also a common phenomenon in numerous organisms. For example, the antisense Tsix RNA that originates from the Xist loci inhibits the expression of the maternal Xist allele [6]. By contrast, COLDAIR, which is a sense intronic lncRNA, antagonises FLC in Arabidopsis [55]. Nevertheless, PAHAL is mainly derived from the intronic region of PAH but also shares its partial sequence with the sixth exon of the coding PAH transcript. Such transcription mode does not solely occur to PAHAL. For example, the sense lncRNA GClnc1 is transcribed in large part from the intronic region of the SOD2 gene, but it also overlaps with the last exon of two SOD2 transcripts [56]. PAHAL and PAH are positively regulated in the behavioural change of the locusts. In vivo and in vitro analyses demonstrated that nuclear-enriched PAHAL promoted the PAH expression. PAHAL is engaged in a regulatory feedback loop that is stimulated by population density changes, by interacting with PAH (Fig 8). Thus, PAHAL acts as a sense transcriptional activator and represents an lncRNA subclass that participates in local cis-regulation.

In line with the behavioural phase changes, PAHAL is generated at transcription levels highly contingent upon locust population density. Thus, PAHAL transcription is precisely controlled commensurate with the crowding state and phase-related behaviour of locusts. The PAHAL transcription then triggers the fine tuning of the PAH expression and in turn the downstream biosynthesis of specific neurotransmitters such as the DA that controls the aggregation behaviour of locusts (Fig 8). Therefore, PAHAL is dedicated to orchestrating the PAH expression and downstream DA production during the population density-dependent behavioural aggregation of locusts. The density-contingent transcription of PAHAL could be also crucial for adjusting the DA dosage. A large proportion of lncRNAs in animal models present high spatial–temporal specificity of expression in various biological processes, such as development and behavioural responses [20,57–59]. Overall, these findings highlight the important role of lncRNAs as a regulator of tightly controlled genetic networks.

Interestingly, two distinct regulatory mechanisms for PAH expression rendered by two noncoding RNAs have been employed in the locust. On one hand, lncRNA PAHAL is induced by locust crowding, positively regulates PAH expression and DA production in the brain at the transcriptional level. On the other hand, miR-133 is upregulated by locust isolation, evokes degradation of PAH mRNA and inhibits DA production, thus representing a mechanism for post-transcriptional regulation of PAH [36]. Both PAHAL and miR-133 are involved in the modulation of behavioral aggregation but in contrasting directions. Therefore, the two mechanisms at different regulatory levels may be orchestrated in response to population density to ensure the fine tuning of PAH expression and DA biosynthesis and thus guarantee the deliberate behavioral response of locusts.

PAHAL mediates regulatory activation by recruiting the transcription activator SRSF2 to the promoter-proximal region of PAH. In this regulatory network (Fig 8), PAHAL involves a distinct regulatory element in the lncRNAs–SRSF2 interaction complex that is destined to the transcriptional machinery of the focal gene. Previous studies showed that some lncRNAs participate in regulatory complexes through different mechanisms, such as interaction with transcription factors and chromatin modifiers [44,60]. SRSF2 was first characterised as a splicing factor [49,61–63]. The nuclear-retained lncRNA MALAT1 is also involved in the regulation of alternative splicing by interacting with SRSF2 [64]. In contrast with MALAT1, PAHAL functions in transcriptional activation by binding with SRSF2. Notably, SRSF2 also plays an active role in transcription elongation and activation [49,62]. In this study, PAHAL RNA recruits SRSF2 to the promoter-proximal region of the target gene to promote local transcription (Fig 8). Therefore, SRSF2 possibly exerts different functions by interacting with various lncRNAs. SRSF2 bound to the ESE sites of the promoter-associated nascent RNA. Three tandem ESE motifs in the 5′-UTR of PAH significantly increases the PAHAL-mediated activation of the PAH-promoter activity in the presence of SRSF2. The ESE sites near the 5′ end of nascent RNA may act as critical signals for transcription activation by triggering the progressive release of SRSF2 and P-TEFb from the transcription pause complex [49]. Thus, PAHAL may enhance the transcription activation of the PAH gene by facilitating SRSF2 binding to the nascent PAH RNA. However, the activity assay with reverse PAHAL demonstrated that the enhancer effect rendered by PAHAL is not due to a genomic enhancer element residing in its sequence (Fig 5C).

The local lncRNA-interwired transcription machinery could orchestrate locus-specific transcription regulation in response to an environmental stimulus. SRSF2 is a general transcription activator in the transcription elongation complex and targets a large set of genes [62]. The locus-specific sense lncRNA PAHAL is generated in response to the environmental signal and destined to the transcriptional machinery of the ancestral gene in collaboration with nascent RNA-bound SR proteins (e.g. SRSF2). The lncRNA–SRSF2 interaction complex may have also delivered conformation modifiers to its target genomic loci while still attached to the elongating RNAP II [28,60]. Hence, PAHAL may facilitate the regulatory role of SRSF2 in a gene locus-specific manner. The sequence at the 3′ end of PAHAL is crucial for the binding of SRSF2 with PAHAL, although the underlying mechanism needs further investigation. This region also harbours one NLS that determines the nuclear retention of PAHAL. SRSF2 is the only SR protein retained in the nucleus [65]. MALAT1 to bind with SRSF2 is required for the correct localisation of SRSF2 to nuclear speckles [64,66]. The sequence at the 3′ end of PAHAL may confer a structural component or binding site that promotes the nuclear retention of SRSF2. In summary, these data suggest that PAHAL could govern the spatial specification of the regulatory components of transcriptional machinery in cells and control the target specificity of interacting transcriptional factors in animal behavioural response. Such a precise regulatory mechanism derived from interactions between PAHAL and SRSF2 enlightens us to develop the novel approaches for behavioural manipulation of animals.

Materials and methods

Animals

Locusts were derived from one locust population (Hebei, China) and maintained under standard conditions at the Institute of Zoology, Chinese Academy of Sciences, Beijing, China [67]. The locust colonies were reared under a photoperiod regime of 14 h of light:10 h of darkness at 30°C ± 2°C and fed on fresh wheat seedlings and bran. The G locusts were reared in large cages (40 cm × 40 cm × 40 cm) at a density of approximately 500 insects per cage. The S locusts were obtained from the G colony and individually cultured in white metal boxes (10 cm × 10 cm × 25 cm) supplied with fresh air. The S locusts were reared in isolation for at least three generations prior to experimentation. One-day-old fourth-stadium nymphs, which are most prominent for the behavioural phase plasticity [40,67,68], were used in all experiments.

Cells

Drosophila S2 cells (Gibco, NY, USA, R69007) were maintained in SFX-insect medium (HyClone, Logan, UT, SH30278.02) at 28°C. Human HEK 293T cells (ATCC, Manassas, USA, CRL-3216) and SRSF2 protein-depleted SRSF2-MEFs were maintained in DMEM (GIBCO, NY, USA, C11965500BT) supplemented with 10% FBS (BI, Beit-Haemek, Israel, 04-001-1A) or tet-free FBS (Clontech, CA, USA, 631106) under humidified 5% CO2/95% air at 37°C. SRSF2-MEFs were treated with 10 μg/mL doxycycline (DOX, Sigma, MO, USA, D9891-1G) for 1 day to deplete SRSF2 transcription (DOX+) [48]. We used HEK 293T cells to express the locust SRSF2-V5 fusion protein because this cell line represents a rapid and high-yield protein production system and express more locust SRSF2 than Drosophila S2 cells. Lastly, we adopted the SRSF2-MEF cell line to verify the interaction of SRSF2 with PAHAL by using an endogenous SRSF2-KO system constructed in the cell line.

RNA isolation, 5′ and 3′ RACE and qPCR

Nymphal tissues were rapidly dissected and stored in liquid nitrogen for ≤ 2 months before RNA preparation. The total RNA was extracted by using TRIzol reagent (Invitrogen, CA, USA, 15596018) and incubated for 15 min at room temperature (RT) with 15 kunitzK units of RNase-free DNase I (Qiagen, Hilden, Germany, 79254) to remove genomic DNA. The RNA quality, integrity and quantity were detected using NanoDrop 2000c (Thermo, CA, USA) and 1% agarose gel electrophoresis. Analyses using 5′ and 3′ RACE were performed with a SMARTer RACE cDNA amplification kit in accordance with the manufacturer’s instructions (Clontech, CA, USA, 634858). The PCR products were purified and cloned into a pGEM-T vector (Promega, WI, USA, A1360) for sequencing. The cDNA was reverse-transcribed with a FastQuant RT Kit (Tiangen, Beijing, China, KR106-02). Briefly, 2 μg of DNase-treated total RNA was mixed with 1 μL of reverse transcriptase mix, 2 μL of FQ-RT primer mix and 2 μL of Fast RT buffer into RNase-free water to obtain a final volume of 20 μL. The mixture was incubated for 30 min at 42°C. The cDNA was attenuated to 100 μL and stored at −20°C. Each 10 μL qPCR reaction contained 2 μL of cDNA template, 5 μL of SYBR Green I Master and 0.5 μM of upstream and downstream primers. The qPCR was performed with LightCycler 480 SYBR Green I Master kit (Roche, Mannheim, Germany, 4887352001) on a LightCycler 480 instrument (Roche, Switzerland). Predenaturation was performed at 95°C for 10 min, followed by 45 cycles of PCR at 95°C, 58°C and 68°C for 20 s. The amplification specificity of the target genes was assessed on the basis of a melting curve. All PCR products were verified through sequencing before qPCR. The housekeeping gene ribosomal protein 49 (RP49) was used as the endogenous control in qPCR analysis [67]. The relative expression levels of the specific genes were quantified by using the 2^−ΔCt method, where ΔCt is the Cp value of RP49 subtracted from that of the gene of interest. Four to eight biological replicates and three technical replicates for every replicate were prepared for each treatment. The outliers were removed from the qPCR standard curves on the basis of the results of Grubbs’ test [69]. S3 Table illustrates all primers for RACE and qPCR.

Sense-specific RT-PCR

The sense-specific PCR was performed as previously described with slight modifications [56]. Briefly, 2 μg of DNase-treated total RNA was mixed with 2 pmol of a reverse primer (primer 13 in Fig 1C) for reverse transcription by using SuperScript IV Reverse Transcriptase (Thermo Fisher; Vilnius, Lithuania, 18090050). The PCR was performed with Taq polymerase (Takara, Tokyo, R001A). An initial denaturation of 94°C for 3 min was followed by 94°C for 30 s, 58°C for 30 s, 72°C for 40 s for 30 cycles. Next, primers 10 and 12 were used to specifically amplify PAH transcript, and primers 9 and 10 were used to specifically amplify the PAHAL transcript. The PCR with primers 11 and 12 is expected to yield no product if PAH and PAHAL are different transcripts. S3 Table present these primers.

5′ exonuclease digestion assay

5′-Phosphate-dependent exonuclease digestion was performed with a terminator 5′-phosphate-dependent exonuclease kit in accordance with the manufacturer’s instructions (Epicentre, IL, US, TER51020). Briefly, total RNA isolated from locust brains was treated with the exonuclease, which specifically digests RNA species with a 5′-monophosphate end, at 30°C for 60 min. The 5′-capped mRNA was isolated through phenol extraction and ethanol precipitation, and cDNA was generated using high-capacity RNA-to-cDNA kit (ABI, CA, USA, 4387406). The mRNA fraction of PAHAL, GAPDH and 18S rRNA was analysed through semi-quantitative PCR. GAPDH and 18S rRNA were used as positive and negative controls, respectively.

Poly(A) tail identification

Poly(A) tails were detected with the Poly(A) tail length assay kit (USB Corporation, Vilnius, Lithuania, 76455). Briefly, a limited number of guanosine and inosine residues were added to the 3′ ends of poly(A)-containing RNAs that were prepared from locust brains. The tailed-RNA was converted to cDNA through reverse transcription by using the newly added G/I tails as the priming sites. PAHAL U6 and GAPDH were amplified from the cDNA with gene-specific primers (S3 Table). The RT-PCR of PAHAL, U6 and GAPDH gave a fragment of 242, 71 and 180 bp, respectively.

In vitro translation assay

TNT Quick for the PCR DNA kit (Promega, WI, USA, L5540) was used for the in vitro translation assay to assess the coding capability of PAHAL. The assay was performed in accordance with the manufacturer’s instructions. Sense primers containing T7 promoters were used to amplify the Renilla luciferase gene (coding negative control) and the full-length PAHAL sequence. The PCR products (1 μg) were mixed with 0.75 μg of transcend biotin–lysyl–tRNA and 1 μL of methionine (1 mM) in 40 μL of TNT T7 Quick Master Mix to a final volume of 50 μL. The mixture was incubated for 1 h at 37°C. Thereafter, 15 μL of the reaction mix was added into 50 μL of the rehydration/sample lysis buffer (8 M urea, 2 M thiourea, 1% SDS and 0.02% [wt/vol] β-mercaptoethanol) in a tube and mixed by inverting the tube until the mixture became homogeneous. After incubation overnight at RT, 10 μL of the homogenised samples was mixed with 1 μL of 1% bromophenol blue buffer (1% bromophenol blue and 50 mM Tris base) and subjected to polyacrylamide gel (15%) electrophoresis. The samples were then transferred to polyvinylidene difluoride (PVDF) membranes (Millipore, CA, USA, ISEQ00010). The membranes were blocked with 5% (wt/vol) skimmed milk at RT for 1 h. The newly synthesised protein containing biotin was detected by using a chemiluminescent nucleic acid detection module (Pierce, CA, USA, 89880).

RNA interference and RNA-seq

dsPAHAL, dsPAH and dsSRSF2 were used to knock down the PAHAL, PAH and SRSF2 expression, respectively. This method has been widely used for specific and efficient knockdown in the locust [40,70–72]. The dsRNAs of GFP (dsGFP) were used as the negative control. The dsRNAs were synthesised by using the T7 RiboMAX express RNAi system (Promega, WI, USA, P1700). Three PAHAL-specific siRNAs (Fig 1B) were designed for siRNA interference. The siRNAs were synthesised by RiboBio (Guangzhou, China) and used as a siRNA pool to interfere with the PAHAL expression. The short interfering RNAs of GFP (siGFP) were used as the negative control. The brains of G locusts were injected with 69 nL of dsRNAs (2 μg/μL) or siRNAs (2 μg/μL) with a glass micropipette tip mounted on a nanoliter injector (World Precision Instruments, FL, USA) under an anatomical lens. The injected locusts were returned to normal rearing conditions and reared for 3 days before their brains were harvested for RNA or protein extraction. Three independent biological replicates were prepared for the RNA-seq analysis of dsGFP and dsPAHAL. The integrity of total RNA was quantified with an Agilent 2100 Bioanalyser (Agilent). The cDNA libraries were prepared and sequenced in accordance with the manufacturer’s protocol of Illumina Nova-seq 6000 (150 PE) at Berry Genomic Corporation, Ltd., Beijing, China. HISAT2 software was used to acquire clean reads from the raw data and map them with the locust genome sequence. The gene-expression level was calculated using the clean reads per kb million mapped reads by using the HTseq tool. Genes with corrected P < 0.05 were considered differentially expressed. Differentially expressed genes were defined as corrected P < 0.05 and log2|FoldChange|> 1. The expression levels were shown by a heat map signal that indicates log2 fold-change values relative to the median expression level within the group. The high and low expression levels relative to the median expression level within the group are represented by yellow and blue signals, respectively. KEGG enrichment was performed by using KOBAS software. Significance was analysed through Fisher’s exact test.

Liquid chromatography–mass spectrometry (LC–MS)

Brains were immediately dissected and stored in liquid nitrogen before being assayed. Ten brains were collected for each of the ten biological replicates. Each sample was homogenised with 100 μL of ice-cold PBS. Thereafter, 10 μL of 1 M perchloric acid was added to 90 μL of the homogenate for protein precipitation. The homogenate was centrifuged at 5200 ×g for 30 min at 4°C. The supernatants were filtered through 0.22 μm filters (Millipore, MA, USA). An equal volume of methanol (Fisher, NJ, USA) was added to the filtered supernatant of each sample for immediate LC–MS assay. The total protein of the remaining homogenate was applied as the loading control and quantified through the BCA assay (Fisher, NJ, USA, 23227).

LC–MS was performed with a rapid resolution liquid chromatography system (ACQUITY UPLC I-Class, Waters, USA). Hydrophilic interaction chromatography separation was performed with an ACQUITY UPLC^™HSS PFP column (100 mm × 2.1 mm, 1.8 μm). The autosampler was set at 10°C. The mobile phase A of the system was gradient elution with 0.1% formic acid acetonitrile (Pierce, CA, USA; Fisher, NJ, USA), whereas phase B was 0.1% formic acid water. The linear gradient for A was as follows: 10%, 0–3 min; 10%–15%, 3–5 min; 15%–100%, 5–7 min; 100%, 7–8 min; 100%–10%, 8–8.1 min; and 10%, 8.1–11 min. The flow rate was adjusted to 0.2 mL/min, and the injection volume was 10 μL. The total run time for each sample was 11 min.

LC/MS/MS analysis was performed on AB SCIEX Triple Quad 4500 (Applied Biosystems, CA, USA) with an electrospray ionisation source (Turbo Ion spray). Mass spectrometry detection was performed in positive electrospray ionisation mode. The [M + H] of the analyte was selected as the precursor ion. The quantification mode was the multiple-reaction monitoring (MRM) mode using mass transitions (precursor/product ions).

The ESI ion source temperature was 500°C. The other mass spectrometric parameters were as follows: curtain gas flow: 10 psi; collisionally activated dissociation gas setting: medium; ion spray voltage: 5500 V; and ion gas 1 and 2: 50 psi. Acquisition and processing were performed by using AB SCIEX Analyst 1.6 Software (Applied Biosystems, CA, USA). The collision energies for different MRM pairs were individually optimised with octopamine (m/z 154 > 136, CE 11V), DA (m/z 154 > 137, CE 14 V), serotonin (m/z 177 > 160, CE 17 V), tyramine (m/z 138 > 77, CE 38 V), l-dopa (m/z 198.1 > 152.1, CE 18 V), 5-hydroxytryptophan (m/z 221.1 > 204.1, CE 14 V) and tyrosine (m/z 182.1 > 136.1, CE 18 V).

Fluorescence in situ hybridization (FISH)

A double FISH experiment was performed as previously described with slight modifications [73]. The RNA probes for PAHAL (digoxigenin [DIG]–PAHAL) and PAH (biotin–PAH) were synthesised by using a T7/SP6 RNA polymerase kit (Promega, WI, USA, P2075, P1085) and DIG RNA labelling mixture/biotin RNA labelling mix (Roche, Mannheim, Germany, 11277073910, 11685597910). Nymphal brains were fixed in 4% (wt/vol) paraformaldehyde for 30 min at RT or overnight at 4°C. The brain tissue was embedded in 5% agarose. Hardened brains were trimmed and dissected into 40 μm slices with a Leica VT1200 S Vibrating Microtome (Leica, Bensheim, Germany). Brain slices were washed twice with PBST (0.5% Triton X-100 in 1× PBS) and then soaked in PBST for 10 min at RT to permeabilise the brain tissue. Brains were digested with proteinase K (160 μg/mL; Invitrogen, CA, USA, AM2548) at 37°C for 20 min and then washed thrice with PBST. Prehybridisation was performed by using a prehybridisation buffer (Wuhan Boster, Wuhan, China, AR0152) at 37°C for 30 min. Brain slices were hybridised with DIG–PAHAL and biotin–PAH probes (5 ng/μL) at 37°C overnight and then blocked at 4°C in 2% BSA (2% BSA in 0.2× SSC) for 20 min. Next, brain slices were incubated in anti-DIG alkaline phosphatase-conjugated antibody (1:500; Roche, Mannheim, Germany, 11093274910) and streptavidin–HRP (1:100) for 1 h at RT and then washed thrice with PBS. The fluorescent signal of DIG for PAHAL or that of biotin for PAH was detected with an HNPP Fluorescent Detection Set (Roche, Mannheim, Germany, 11758888001) or TSA Fluorescein System (Perkin-Elmer, MA, USA, NEL701A001KT). Images were captured under a LSM 710 confocal fluorescence microscope (Zeiss, Oberkochen, Germany) at 10× and 40× magnifications. S3 Table lists the primers used for the probe synthesis of PAHAL and PAH.

Isolation and crowding of locusts

The locusts were isolated by introducing typical G nymphs into metal boxes and individually rearing them under standard conditions. The locusts were crowded by introducing 10 labelled S nymphs and 20 G nymphs into an optic Perspex box (10 cm × 10 cm × 10 cm). Adequate fresh food was provided. After 0, 4 or 8 h of treatment, the locust brains were dissected and immediately frozen in liquid nitrogen for RNA preparation. All insects were sampled at the same time of the day to avoid the effects of circadian rhythm on the locust phenotypes. Equal numbers of male and female insects were sampled for each biological replicate.

Behavioural assays

Behavioural assays were performed as previously described [40,42]. A rectangular Perspex arena (40 cm × 30 cm × 10 cm) was used in the assay. One of the separated side chambers (7.5 cm × 30 cm × 10 cm) contained 30 G locusts as a stimulus group. Another chamber with the same dimensions was left empty. A locust was released into the centre of the arena and monitored for 300 s. Individual behavioural data were automatically recorded and analysed with the EthoVision video tracking system (v.3.1.16, Noldus Inc., Wageningen, Netherlands). The three behavioural variables were as follows: total distance moved (TDM) and total duration of movement (TDMV), which represent motor activity levels, and attraction index (AI, i.e., total duration in stimulus area minus total duration in opposite area), which represents the attraction or repulsion to the stimulus group. The behavioural phase state of each locust was assessed by applying a single probabilistic metric of gregariousness P_greg, P_greg = e^η/(1 + e^η), where η = −2.11 + 0.005 × AI + 0.012 × TDM + 0.015 × TDMV [36,42]. P_greg indicates the probability of a locust regarded as G. P_greg = 1 indicates fully G behaviour, whereas P_greg = 0 means fully S behaviour.

Nuclear fractionation

Nuclear and cytoplasmic fractionation was performed as previously described [73]. Twenty nymphal brains were harvested and homogenised in a cold lysis buffer [1× PBS containing 0.2% IGEPAL CA-630 (Sigma, MO, USA, I8896-50ml) and 1× proteinase inhibitor (Pierce MA, USA, 88266) and RNase inhibitor (Promega, WI, USA, N2111S)] for nuclear fractionation. The homogenate was then centrifuged at 30 ×g for 2 min at 4°C. The supernatants were transferred to a fresh tube and centrifuged at 425 ×g for 15 min at 4°C to obtain the nuclear pellet. The cytoplasmic fraction in the supernatant was centrifuged at 2000 ×g for 10 min at 4°C to remove residual nuclei. The nuclear pellet and cytoplasmic supernatant were maintained at −80°C prior to RNA extraction.

Reporter and expression plasmid construction, luciferase assay and antibodies

Luciferase assays were performed to verify whether PAHAL regulates PAH in cis. Plasmids containing different promoter regions of PAH and their deletions were constructed into pGL4.10 (Promega, WI, USA, E665A) by using the KpnI and XhoI restriction sites. Promoter regions were amplified from locust genomic DNA. A series of ESEs with mutations in the PAH 5′-UTR (containing −554/+89 fragment) were inserted into pGL4.10 obtained from PolePolar Biotechnology Co., Ltd., Beijing, China.

The full-length PAHAL sequence, the PAHAL sequence in the reverse orientation (reverse PAHAL), the PAHAL sequence with deletions, the SRSF2 ORF and a 3 kb lacz ORF (as the negative control) were cloned into the pcDNA3.1 (+) vector (Invitrogen, CA, USA, V79020) and/or pAc5.10/V5-His A vector (Invitrogen, CA, USA, V4110-20) by using KpnI and XhoI. These vectors were transfected with Drosophila S2 cells, HEK 293T cells and SRSF2-MEFs.

Cells were plated on 48-well plates and transfected by using Lipofectamine 3000 in accordance with the manufacturer’s instructions (Invitrogen, CA, USA, L3000015). Reporter plasmids (10 ng) with 200 ng of the expression plasmid or negative control vector were co-transfected with 5 ng of the internal control plasmid pRL-TK/pGL4.73 (Promega, WI, USA, E2241/ E691A). Luciferase activity was measured with Dual-Luciferase Reporter Assay System (Promega, WI, USA, E1960) at 30 h after incubation.

The polyclonal antibodies for PAH and monoclonal SRSF2 antibody were produced by immunising mice with the prokaryotic expression peptide (Beijing Protein Innovation, Beijing, China). The tubulin (rabbit) and GAPDH (rabbit) antibodies [74] were provided by Dr. Yun-Dan Wang.

RNA pulldown and Western blot analysis

Twenty brains in one biological duplicate were lysed to acquire total protein for RNA pulldown by using 200 μL of tissue lysis buffer [200 μL of T-PER tissue protein extraction reagent (Pierce, CA, USA, 78510) and 2 μL of Halt Protease Inhibitor Cocktail, EDTA-free (Pierce, CA, USA, 87785)]. Such an undertaking was carried out to identify the interactive proteins with PAHAL in vivo. RNA pulldown was performed with a magnetic RNA–protein pull-down kit in accordance with the manufacturer’s instructions (Pierce, CA, USA, 20164). The RNA probes for full-length PAHAL and their deletions (biotin–PAHAL) were synthesised with a T7 RNA polymerase kit. The RNA-associated proteins were separated on 15% SDS-PAGE gel and then subjected to protein silver staining. Single silver-stained bands that were present in PAHAL pulldown but absent from antisense PAHAL pulldown were excised and then bleached by using a destaining buffer (50% acetonitrile and 25 mM NaHCO3). Disulphide bonds of protein samples were disrupted by adding 10 mM DTT for 1 h at 56°C and 55 mM Iodoacetamide for 45 min at RT. Protein was hydrolysed by incubation in trypsin buffer (62.5 ng/μL trypsin and 25 mM NaHCO3) at 37°C overnight. The protein samples were subjected to mass spectrometry (Beijing Protein Innovation, Beijing, China). MASCOT software (Multiple Alignment System developed by iCOT) [75] was used to identify and quantity proteins from the mass spectrometric data. The protein functions were annotated by the protein scores > 85, which is a criterion of the significant high expression level of proteins (P < 0.05) in this single silver-stained band.

RNA pulldown and western blot analysis were performed on SRSF2-MEFs (2 × 10⁷) that co-transfected with pcDNA3.1/V5-His/SRSF2 ORF and a series of PAHAL-pcDNA3.1+ deletions. This task was carried out to identify the specific sites for SRSF2 interaction in the PAHAL sequence.

The total proteins for Western blot analysis were first extracted by using TRIzol reagent. The proteins were dissolved in rehydration/sample lysis buffer (in vitro translation assay) to a concentration of 10 μg/μL overnight at RT and then mixed with 1 μL of 1% bromophenol blue buffer. The proteins were then separated through SDS-PAGE on 10% NuPAGER Bis–Tris gel (Invitrogen, CA, USA, NP0315BOX) by using 1× NuPAGE MOPS SDS running buffer (Invitrogen, CA, USA, NP0050) in accordance with the manufacturer’s instructions. The separated proteins were transferred to PVDF membranes by using 1× transfer buffer and blocked with 5% (wt/vol) skimmed milk for 1 h at RT. The membranes were incubated in a blocking buffer at 4°C overnight with a primary antibody in the following concentration: anti-PAH (mouse), 1:500; anti-SRSF2 (mouse), 1:500; anti-SmD1 (rabbit), 1:2,000; anti-GAPDH (rabbit), 1:5,000; and anti-Tubulin (rabbit), 1:5,000. The secondary antibody (1:5,000; Easybio, Beijing, China, BE0101-100, BE0102-100) in the blocking buffer was incubated for 1 h at RT. The immunological blot was detected with SuperSignal West Femto Substrate Trial Kit (Pierce, CA, USA, 34094).

RNA immunoprecipitation (RIP) assay

The SRSF2 ORF was first cloned in frame with the V5 epitope of the pcDNA3.1/V5-His vector for the construction of pcDNA3.1/V5-His/SRSF2 ORF to test the binding of SRSF2 with PAHAL in vitro. HEK 293T cells were then co-transfected with this vector and pcDNA3.1 (+)/PAHAL for the following RIP assay. After 3 days, 2 × 10⁷ of HEK 293T cells were harvested for the RIP experiment.

The RIP assay was performed with Magna RIP Quad RNA-Binding Protein Immunoprecipitation Kit (Millipore, CA, USA, 17–704). The cell pellet was homogenised in ice-cold RIP lysis buffer containing 1× proteinase inhibitor and RNase inhibitor and stored at −80°C overnight. Magnetic beads were pre-incubated with 5 μg of V5 antibody (Invitrogen, CA, USA, R96025) or normal mouse IgG (Millipore, CA, USA, CS200621) for 30 min at RT with rotation. The supernatants of the lysate from the centrifugation were co-incubated with the bead–antibody complex overnight at 4°C with rotation. Thereafter, 10 μL of the supernatants was stored as the input. The RNA in the immunoprecipitates and input was extracted by using TRIzol reagent. Reverse transcription was performed with a high-capacity RNA-to-cDNA kit. The target gene expression was analysed through qPCR.

The binding of SRSF2 with PAHAL in vivo was tested by performing the RIP assay on brain tissues. A Dynabeads Protein G immunoprecipitation kit (Thermo Fisher, CA, USA, 10007D) was used. S4 Table shows the antibody epitopes of PAH and SRSF2. Fifty brains in one biological replicate were lysed with 700 μL of T-PER Tissue Protein Extraction Reagent containing RNasin Plus RNase Inhibitor (Promega, WI, USA, N2611) and EDTA-free 1× Halt Protease Inhibitor Cocktail. After the lysates were centrifuged at 10,000 ×g for 10 min, the supernatants were collected and subjected to RIP by following the protocol included with the kit. A total of 100 μL of Dynabeads was pre-incubated with 10 μg of antibody for 10 min at RT with rotation. The supernatants (10 μL) of the lysate were stored as the input. A total of 300 μL of supernatants were co-incubated with the bead-specific antibody complex with rotation for 10 min at RT. The RNA and proteins were extracted using a TRIzol reagent before quantitative measurement.

Bioinformatics and statistical analysis

Transcriptome data subjected to poly(A) lncRNA screening were obtained from a previous publication [76]. Locust lncRNA sequences were predicted and identified through an integrative method as previously described [44,57]. Sequence conservation of PAHAL was analysed by blasting in the genome assembly of all species in flybase (www.flybase.org). The sequence motifs were WNNNNSNNAGCCC (W = A/T, S = G/C) for the NLS [47] and WSSNGYY (W = A/T, S = G/C Y = C/T) for the SRSF2-responsive ESE [49,77]. Data from the tissue expression and truncated luciferase assay were analysed through ANOVA, then by post-hoc Tukey’s b-test for multiple comparisons. Independent sample Student’s t-tests were performed for comparing differences in gene expression and other values between treatments. The frequency data of behavioural features, namely, P_greg and AI, were analysed through Mann–Whitney U test. Two-sided P-values were provided. Data are presented as the mean ± SEM unless stated otherwise. All statistical data were analysed by using SPSS 21.0 (SPSS Inc., IL, USA). The locust genome data are available at the following website: http://www.locustmine.org. All sequences for PAHAL, PAH and SRSF2 have been deposited in GenBank under accession numbers KX962170, KX951493 and KX951494, respectively. The RNA-seq data have been uploaded to NCBI with the accession number PRJNA522953. Numerical data that underlies graphs or summary statistics and sample image data in support of all reported results have been uploaded to Harvard Dataverse Network with the websit https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/LXRBKJ and https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/KSVYWY

Supporting information

S1 Table [xlsx]
The FPKM of differently expressed gene.

S2 Table [docx]
The proteins from pulldown identified by mass spectrometry.

S3 Table [docx]
Sequences of all primers used in the study.

S4 Table [docx]
The antibody epitopes of PAH and SRSF2.<

S1 Fig [pdf]
Full-length sequence of the locus.

S2 Fig [tif]
Localisation of and in FISH.

S3 Fig [tif]
Nuclear localisation signal (NLS) identified in .

S4 Fig [a]
Gene expression of proteins from pulldown during locust isolation and crowding.

S5 Fig [a]
SRSF2 binding with .