A lipocalin protein, Neural Lazarillo, is key to social interactions that promote termite soldier differentiation

Abstract

Social communication among castes is a crucial component of insect societies. However, the genes involved in soldier determination through the regulation of inter-individual interactions are largely unknown. In an incipient colony of the damp-wood termite Zootermopsis nevadensis, the first larva to develop into a third instar always differentiates into a soldier via frequent trophallactic feeding from the reproductives. Here, by performing RNA-seq analysis of third instar larvae, a homologue of Neural Lazarillo (named ZnNLaz1) was found to be the most differentially expressed gene in these soldier-destined larvae, compared with worker-destined larvae. This gene encodes a lipocalin protein related to the transport of small hydrophobic molecules. RNAi-induced knockdown of ZnNLaz1 significantly inhibited trophallactic interactions with the queen and decreased the soldier differentiation rates. This protein is localized in the gut, particularly in the internal wall, of soldier-destined larvae, suggesting that it is involved in the integration of social signals from the queen through frequent trophallactic behaviours. Based on molecular phylogenetic analysis, we suggest that a novel function of termite NLaz1 has contributed to social evolution from the cockroach ancestors of termites. These results indicated that a high larval NLaz1 expression is crucial for soldier determination through social communication in termites.

1. Introduction

The transition from solitary to social living is one of the most striking events in evolution [1] and has occurred independently multiple times in diverse animal lineages. Because social living entails strong interactions among individuals in the group, to gain a comprehensive understanding of this major transition in evolution requires elucidation of the mechanisms that underlie social interactions.

Termites are one of the major social insect groups and acquired a division of labour among castes (reproductives, soldiers and workers) during the evolutionary transition to eusociality. They and their sister group, the subsocial cockroach Cryptocercus spp., diverged from a common ancestor, with termites subsequently evolving sociality independently of the hymenopteran social insects (bees, ants, and wasps). In termite society, soldiers are the first-evolved sterile caste (soldier-first eusociality) [2], which arose just once in a subsocial cockroach ancestor [3]. Soldiers are differentiated from workers via a presoldier stage, and the development of these individuals is regulated by intracolonial interactions [4]. Social interaction therefore plays a central role in termite evolution.

Despite its importance, the critical factors responsible for soldier development are still a mystery, primarily because soldier-destined individuals cannot be identified under natural conditions, thus hindering our ability to examine intrinsic mechanisms of soldier differentiation. However, in the damp-wood termite Zootermopsis nevadensis (figure 1a), in each incipient colony founded by alate pairs collected at the end of April, the first larva that becomes third instar (No. 1 larva) always differentiates into the first soldier of the colony while the next larva (No. 2 larva) moults into a normal fourth instar and function as a worker [5] (figure 1b). The No. 1 larva receives frequent proctodeal trophallaxis (i.e. anal feeding) from the reproductives (the queen and king) during the third instar stage. These findings provide a unique opportunity to monitor developmental processes during soldier differentiation under natural conditions.

Figure 1.

Differentially expressed genes during soldier differentiation in the incipient colony of Zootermopsis nevadensis. (a) In a termite nest, there are male and female reproductives (dark brown individuals), workers (white individuals) and soldiers (brown individuals with enlarged head and exaggerated mandibles). The first soldier is differentiated from the third instar (called No. 1 larva) via a presoldier stage, and maintained for at least several months. The other soldiers are differentiated from the older instar larvae (usually seventh instars in mature colony). (b) Moulting periods of the soldier-destined third instar (No. 1 larva) are shorter than those of the worker-destined third instar (No. 2 larva). Individuals were collected at day 0 (D0) and day 3 (D3) after their appearance for the RNA-seq analysis. (c,d) The scatter plots were described by using the count data of RNA-seq analysis between the No. 1 and No. 2 larvae at day 3 after the appearance in heads (c) and thoraxes + abdomens (d). Differentially expressed genes (q-value less than 0.05 for Cufflinks) are shown as the red circles. Blue circle indicates Znev_05665 (ZnNLaz1). (Online version in colour.)

To explore the molecular mechanisms involved in social interactions that affect soldier caste differentiation, RNA sequencing (RNA-seq) analysis was conducted to compare differentially expressed genes (DEG) between the No. 1 and No. 2 larvae. The silencing of a homologue of lipocalin Neural Lazarillo (NLaz), the gene most significantly differentially expressed in the No. 1 larvae, resulted in the suppression of the proctodeal trophallactic interactions with the queen and reduced presoldier differentiation rates. Based on the results of molecular phylogeny and protein localization analyses, we propose that the lipocalin NLaz is a crucial factor in the social interactions leading to termite soldier caste differentiation.

2. Material and methods

(a). Termite collection and incipient colony foundation

Mature colonies of Z. nevadensis were collected in Hyogo prefecture, Japan, in April and June 2011–2015. Sections of decayed logs containing numerous nymphs (alate-destined individuals) were harvested, brought to the laboratory, placed into plastic boxes and maintained in constant darkness at room temperature until alates emerged. Alates were collected and separated by sex based on the configuration of the genital plate [6]. As in previous studies [5,7–9], male and female alates from different colonies were paired, placed into 60 mm plastic dishes, and supplied with crushed pieces of nest wood. These dishes were then kept in constant darkness at 25°C for several months. From each incipient colony, a single No. 1 larva at the third instar larval stage was observed at the beginning of August.

(b). Total RNA extraction

The oldest third instar (No. 1 larva) and the second larva that moulted into a third instar (No. 2 larva) were marked with different coloured waterproof inks and the head widths of larval instars measured in accordance with previous studies [5,10]. The No. 1 and No. 2 larvae were collected on days 0 and 3 after their appearance in the young colony (n = 5). All individuals were dissected using a stereomicroscope (SZX10, Olympus, Tokyo, Japan) and the heads and remaining body parts (thorax and abdomen) immediately immersed in liquid nitrogen, then stored at −80°C until RNA extraction. Total RNA was extracted from each body part and genomic DNA was removed using SV Total RNA isolation kit (Promega Madison, WI, USA).

(c). Library preparation for next-generation sequencing

The amounts of RNA and DNA in each sample were quantified using a Qubit fluorometer (Life Technology, Eugene, OR, USA), and the quality of RNA was validated using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). Total RNA (500 ng) was used for cDNA synthesis and purification was based on a low-throughput protocol using the TrueSeq sample preparation kit (Illumina, San Diego, CA, USA). The half-scale version of the standard protocol was applied for library preparation. One library was prepared in each category and eight total libraries (four head and four thorax/abdomen libraries) were sequenced. RNA-sequence analysis was performed by paired-end sequence using a next-generation sequencer Hiseq 2000 (Illumina, San Diego, CA, USA). All of the reads have been deposited in the DDBJ Sequence Read Archive (DRA) database under accession number DRA006998. Prior to the assembly and mapping (described below), low-quality reads and adaptor sequences used in RNA-seq were removed from all libraries.

(d). De novo assembly and mapping reads

Reads of four libraries derived from heads were used for the de novo transcriptome assembly, performed by Trinity software r2012-05-18 [11] with default settings. Prior to transcript assembly, the sequence reads were filtered to remove Illumina adaptor sequences and trim low-quality end sequences with cutadapt [12]. RNA-seq reads were mapped to the de novo assembled sequences, and mapped reads were counted using alignRead and RSEM equipped with Trinity. Counted reads were normalized by a trimmed mean of M values (TMM) method, and normalized data were used for the identification of DEGs by edgeR [13].

(e). Mapping reads to genome sequences

RNA-seq reads were mapped to the Z. nevadensis genome sequence [14] using TopHat v. 2.0.8 software [15] with aligner Bowtie2 v. 2.1.0.0 [16] with default settings. The mapping rate of each library against a genome was calculated using Samtools v. 0.1.18 [17]. Obtained mapping data were analysed for DEGs identified using Cufflinks v. 2.0.2 [18] with default settings. The annotated Z. nevadensis gene models OGS2.2 were used for cufflinks.

(f). Sub-cloning, sequencing and molecular phylogenetic analysis

Total RNA was extracted from whole bodies of the No. 1 larvae (n = 5) using ISOGEN (NipponGene, Tokyo, Japan). The extracted RNA was purified by DNase treatment to remove genomic DNA. RNA purity and quantity were measured using a NanoVue spectrophotometer (GE Healthcare BioSciences, Tokyo, Japan). The purified RNA (2 µg) was reverse-transcribed for cDNA synthesis using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster, CA, USA). ZnNLaz1 gene-specific primers, to amplify the ORF and 5′ and 3′ UTR regions, were newly designed based on de novo assembled sequences obtained using Primer3Plus [19] (electronic supplementary material, table S1). PCR products were purified using a QIAquick Gel Extraction Kit (Qiagen, Tokyo, Japan), and subcloned into a pGEM easy T-vector (Promega, Madison, WI, USA). The inserted DNA sequence was amplified by PCR, and purified products were sequenced using a BigDye Terminator v. 3.1 Cycle Sequencing Kit and an automatic DNA Sequencer 3130 Genetic Analyzer (Applied Biosystems). Obtained sequences were subjected to BLAST database searches. The determined nucleotide and putative amino acid sequences are available at DDBJ/EMBL/GenBank databases (accession no. LC382016). NLaz homologues in other insects were obtained from the NCBI database, and molecular phylogenetic analyses were performed (see the electronic supplementary material).

(g). ZnNLaz1 expression analysis using quantitative real-time PCR

The No. 1 and No. 2 larvae were collected at day 0 and 3 after appearance in each colony. Nine different Νo. 1 and Νo. 2 larvae from 9 different incipient colonies were sampled at each time point. Individuals were immersed immediately in liquid nitrogen and stored at −80°C until use (three individuals in one tube). All individuals were separated into heads and other body parts, and to compare gene expression levels between Νo. 1 and Νo. 2 larvae, each sample was used for cDNA preparation. Total RNA extraction using ISOGEN, and DNase treatment and cDNA synthesis were performed in accordance with the procedure described above. For cDNA preparation, equal concentrations of total RNA (145 ng) were reverse-transcribed using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Real-time quantitative PCR was performed with an MX3005P qPCR system (Agilent Technologies). The suitability of six genes as internal controls used in previous studies [9,20] was evaluated using software GeNorm [21] and NormFinder [22]. qPCR analysis was performed in biological triplicates. Specific primers used in the qPCR analysis of ZnNLaz1 was designed from an obtained sequence by sub-cloning using Primer3Plus [19] (electronic supplementary material, table S1).

(h). Small-interference (si) RNA preparation

Double-stranded (ds) RNA was synthesized from plasmid DNA including the target region of the ZnNLaz1 gene sequence (DDBJ/EMBL/GenBank databases accession no. LC382016) by in vitro transcription. The nucleotide sequence for insertion into a plasmid was amplified from cDNA of the No. 1 larvae by PCR using gene-specific primers (electronic supplementary material, table S1). GFP vector pQBI-polII (Wako, Osaka, Japan) was used as the control, and a 706 bp portion was amplified as described previously [20,23]. The PCR products were purified using a QIAquick Gel Extraction Kit (Qiagen, Tokyo, Japan) and subcloned into a pGEM easy T-vector (Promega, Madison, WI, USA). All constructs were sequenced using a BigDye Terminator v. 3.1 Cycle Sequencing Kit and an automatic DNA Sequencer 3130 Genetic Analyzer (Applied Biosystems). PCR products including the target regions were amplified from these plasmids using T7 and SP6 primers and purified using a QIAquick Gel Extraction Kit (Qiagen, Tokyo, Japan). ZnNLaz1 and GFP dsRNA were synthesized from the purified products using SP6 (Roche, Grenzacherstrasse, Basel, Switzerland) and T7 RNA polymerase using a MEGA script T7 transcription kit (Ambion, Austin, TX, USA). ZnNLaz1 and GFP small-interference (si) RNA were produced from each dsRNA prepared using si-RNAse III (Takara, Siga, Japan).

(i). RNA interference (RNAi)

The No. 1 larvae (one No. 1 larva per colony [5]) were separated from their colonies at day 1 after their appearance. The sampled larvae were anaesthetized on ice for 90 s, and an equal volume (32.2 nl) of DDW or siRNA solution (5 ng) coloured with Fast Green FCF (Tokyo Chemical Industry, Tokyo, Japan) was injected into the thorax using a Nanoliter 2000 (World Precision Instruments, Sarasota, FL, USA). These larvae were then replaced with their parents (the king and queen) in 46 mm Petri dishes under constant darkness at 25°C. To evaluate the reduction in ZnNLaz1 gene expression, the No. 1 larvae (13 individuals in each treatment) were collected 24 h after the injection of siRNA solution. Individuals were immersed immediately in liquid nitrogen and stored at −80°C until use. As described above, total RNA of the No. 1 larvae was extracted in each individual, and cDNAs were synthesized from the extracted RNA. Real-time quantitative PCR was performed using a MX3005P qPCR system (Agilent Technologies). The qPCR analysis was performed in biological replications (ZnNLaz1, n = 13; gfp, n = 13).

(j). Behavioural observation after RNAi treatment

Behavioural observations were performed in accordance with the previous study [5]. Each dish was placed in constant darkness for 30 min after the injection of DDW or siRNA solution at room temperature before video recording. The behaviours of the injected No. 1 larvae were recorded for 5 days (120 min per day) using a CX1 (Ricoh, Tokyo, Japan) digital camera under red light (electronic supplementary material, figure S1). The frequencies of proctodeal trophallactic behaviour (i.e. food transfer via anal feeding) from reproductives to the No. 1 larva were counted. We confirmed that these behaviours continued more than 3 s from the beginning according to the criteria of previous work [5]. After video recording, all dishes were kept in constant darkness at 25°C and checked every 24 h to determine whether the No. 1 larva had moulted into a presoldier or a fourth instar.

(k). Western blotting and immunohistochemistry

We requested that the Hokudo Company (Sapporo, Japan) produce a synthetic peptide antigen and a rabbit antibody against the synthetic peptide. Western blotting was performed using proteins extracted from whole bodies, guts, and additional bodies without guts of No. 1 and No. 2 larvae (see the electronic supplementary material). Next, in accordance with a previous study [23], the ZnNLaz1 protein was detected by immunohistochemistry. Whole bodies of the No. 1 larvae at day 3 after their appearance (n = 10) were fixed at 4% formaldehyde in PBS overnight at 4°C. All fixed individuals were dehydrated in increasing concentrations of ethanol (70%, 90%, 95% and 100%), and finally cleared in xylene. Fixed samples were embedded in paraffin. Paraffin blocks stored at −30°C were used for the histological sections. Sagittal sections (10 µm) were cut serially using an MRS80-074 microtome (Ikemoto, Tokyo, Japan). Serial sections on MAS-coated glass slides (Matsunami, Osaka, Japan) were deparaffinized in xylene two times for 5 min, rehydrated in decreasing alcohol (100%, 95% and 70%) and deionized water. Secondary antibody response was performed with a goat anti-rabbit secondary antibody Alexa Fluor 488 (Molecular Probes, Carlsbad, CA, USA). Sections were counterstained with 4, 6-diamidino-2-phenylindole (DAPI; 30 nM solution in PBT) (Lonza, Walkersville, MD, USA) in consistent darkness for 30 min at room temperature. Sections were washed in PBT for 10 min and mounted in VECTASHIELD (Vector Laboratories, Burlingame, CA, USA). Sections were observed using a BZ8100 microscope (Keyence, Osaka, Japan).

3. Results

(a). RNA-seq analysis between the soldier- and worker-destined individuals

In accordance with the previous study [5], proctodeal trophallactic behaviours from reproductives to the No. 1 larva were the most frequently observed at day 3 after their appearance. Consequently, to identify the genes that are differentially expressed during presoldier differentiation, we compared gene expression levels between No. 1 and No. 2 larvae at day 0 and 3 after they appeared. The numbers of DEGs counted by the TopHat-Cufflinks pipeline (using genome database [14]) were 52−253 genes in the heads and 53−330 genes in the thoraxes and abdomens. In the lists of DEGs, one gene (Gene ID: Znev_05665) was differentially expressed in the No. 1 larvae at day 3 after appearance, compared to those of day 0, both in the heads and thoraxes/abdomens (electronic supplementary material, tables S2 and S3). Moreover, Znev_05665 was differentially expressed in the No. 1 larvae at day 3 after appearance both in the heads and thoraxes/abdomens, compared to those of No. 2 larvae (electronic supplementary material, tables S4 and S5). On the other hand, Znev_05665 was not differentially expressed in the No. 2 larvae between day 0 and 3 (electronic supplementary material, tables S6 and S7), and significant differences in its expression levels were not observed between the No. 1 and No. 2 larvae at day 0 after their appearance (electronic supplementary material, tables S8 and S9). High expression levels of Znev_05665 in the No. 1 larvae at day 3 after appearance were confirmed by a Trinity-edgeR pipeline (using de novo assembly data newly constructed in this study) and real-time qPCR (electronic supplementary material, figure S2).

(b). Identification of Neural Lazarillo homologue

The coding region of Znev_05665 (897 bp, 298 aa) was determined by sub-cloning and sequencing in order to confirm the Znev_05665 sequence obtained from the genome sequence. Similarity searches indicated that the obtained sequence was similar to Neural Lazarillo (NLaz) of Drosophila melanogaster (E-value = 3 × 10⁻¹⁸). The obtained sequence was confirmed to have a signal sequence at the N-terminus and a conserved lipocalin-like domain. The protein structure predicted from the deduced amino acid sequence was completely confined to lipocalin [24] with a ligand-binding pocket in other insect species. We confirmed that Znev_05665 was included in the same orthologue group with D. melanogaster NLaz (Orthodb7 ID: EOG7TBPX9). We found that Znev_08072 was also included in the same group. Molecular phylogenetic analysis using conserved domains revealed that both Znev_05665 and Znev_08072 were the most closely related to the respective homologues identified from other termites (Cryptotermes secundus, Macrotermes natalensis) with genome sequences available (XP_023718675 and XP_023713938, MN_004031 and MN_003354, respectively) (electronic supplementary material, figure S3). These clades were closely related to the homologues of cockroaches (Cryptocercus punctulatus and Blattella germanica), and we named them termite NLaz1 and NLaz2, respectively.

(c). Molecular evolution for NLaz homologue

We performed the phylogenetic analysis by maximum-likelihood (PAML) analysis to determine whether the termite NLaz1 evolved more rapidly than the other NLaz homologues by calculation of synonymous (dS) and non-synonymous (dN) substitution rates. Although two NLaz copies were found in the Z. nevadensis, C. secundus, M. natalensis and B. germanica genome sequence data, only one homologue was found in a member of the sister-group of termites (the woodroach C. punctulatus) based on transcriptome data (DRA004598) [25]. The results showed that the ω (=dN/dS) ratio in the branch leading to the clade with Znev_05665 (hereafter called ZnNLaz1), XP_023718675 and MN_004031 was not significantly higher than those in the other branches (p = 0.308) (clade (1) in electronic supplementary material, figure S3 and table S10). On the other hand, a significantly higher ω ratio was detected in the branch leading to the termite NLaz1 (ZnNLaz1, XP_023718675 and MN_004031) + Cryptocercus linage (p = 2.10×10⁻³) (clade (2) in electronic supplementary material, figure S3 and table S11) and an amino acid residue at position 75 [S (Blattella), K (termites + Cryptocercus)] was positively selected in this branch (posterior probability above 99% in Bayes Empirical Bayes analysis). A significant higher ratio was not detected in the branch leading to the termite NLaz2 (Znev_08072, XP_023713938 and MN_003354) (p = 1.000) (clade (4) in electronic supplementary material, figure S3 and table S12). Two branches leading to the clade with cockroaches and termites were not significantly higher than those in the other branches (p = 0.997 and 0.999, respectively) (clade (3) and (5) in electronic supplementary material, figure S3, tables S13 and S14).

(d). ZnNLaz1 knockdown in soldier-destined larvae using RNA interference (RNAi)

Because high expression levels of ZnNLaz1 were specifically observed only in the No. 1 larvae at day 3 after appearance (figure 1; electronic supplementary material, figure S2), it could be an important candidate as a determinant of presoldier differentiation. To test this possibility, we performed an RNAi experiment to inhibit ZnNLaz1 expression in No. 1 larvae at the third instar stage. ZnNLaz1 expression levels were significantly reduced by the ZnNLaz1 siRNA injection compared with gfp siRNA (electronic supplementary material, figure S4). RNAi of ZnNLaz1 had no lethal effect, and all treated larvae survived until they moulted into presoldiers or fourth instar larvae. DDW-injected (n = 7) and gfp siRNA-injected No. 1 larvae (n = 9) always differentiated into presoldiers (figure 2a,b), and it took 11.0 ± 1.3 and 11.1 ± 0.7 days (mean ± s.d.) from their appearance to presoldier differentiation, respectively. By contrast, for ZnNLaz1 siRNA-injected No. 1 larvae (n = 17), 8 and 9 larvae differentiated into presoldiers and moulted into fourth instar larvae, respectively (figure 2c), which took 11.4 ± 0.7 or 15.2 ± 0.7 days from their appearance to the moult, respectively. Presoldier differentiation rates after the ZnNLaz1 siRNA injection (47.1%) were significantly lower than those of control treatments (DDW, 100%; gfp, 100%) (Ryan's test for multiple comparisons of proportions, RD = 0.43952 and 0.46267, respectively, p < 0.05). Behavioural observations indicated that the frequencies of proctodeal trophallaxis from the female reproductive (queen) to the fourth instar larvae (i.e. worker)-destined ZnNLaz1 RNAi-treated individuals were significantly lower than those from a queen to the presoldier-destined individuals, either ZnNLaz1 RNAi-treated or gfp RNAi-treated (figure 2d; two-way ANOVA followed by Tukey's test, p < 0.05). On the other hand, proctodeal trophallaxis rates from the male reproductive (king) were not significantly different among the three treatments. These results indicated that high ZnNLaz1 expression is necessary for presoldier differentiation through the regulation of social interaction with the queen.

Figure 2.

Effects of RNAi-mediated ZnNLaz1 knockdown. (a–c) The external head morphologies of moulted individuals after siRNA injection. Scale bars, 1 mm. (a) Presoldier differentiated from a gfp-siRNA injected No. 1 larva. (b) Presoldier differentiated from a ZnNLaz1-siRNA injected No. 1 larva. (c) The fourth instar larva moulted from a ZnNLaz1-siRNA injected No. 1 larva. (d) Total occurrences (120 min×5 days of observation) of trophallaxis from reproductives to siRNA-injected No. 1 larvae. Individuals observed are ZnNLaz1-siRNA injected No. 1 larvae before the larval moult into fourth instar (left), ZnNLaz1-siRNA injected No. 1 larvae before presoldier differentiation (middle) and gfp-siRNA injected No. 1 larvae before presoldier differentiation (right). The boxes in orange and blue represent the frequencies from female reproductives (queens) and male reproductives (kings), respectively. Numbers of colonies examined are shown in parentheses. The boxes and whiskers indicate median, quartiles, and range. The data are consistent with the use of parametric statistics by Levene's test (p = 0.506) prior to the use of the ANOVAs. Two-way ANOVA was performed for observed frequencies of proctodeal trophallaxis from reproductives to the siRNA-injected No. 1 larvae (treatments, F = 4.91, p = 1.63×10⁻²; sexes, F = 20.1, p = 1.53×10⁻⁴, interactions, F = 5.70, p = 9.44×10⁻³). Different letters above the columns indicate significant differences among categories (Tukey's test, p < 0.05). (Online version in colour.)

(e). ZnNLaz1 protein localization in the No. 1 larvae

In order to assess ZnNLaz1 protein localization, we performed western blotting and immunohistology using a specific antibody. Based on the deduced amino acid sequences, the ZnNLaz1 protein size was calculated to be approximately 32 kDa. Using proteins extracted from whole bodies, one common band was observed both in the proteins from the No. 1 and No. 2 larvae, and one specific band was observed below the common band only in the former (electronic supplementary material, figure S5A). It is possible that NLaz could function by protein processing, such as the cleavage of signal sequences at the N-terminus or glycosylation at the C-terminus, as in grasshoppers and fruit flies [26,27]. This specific band was observed in proteins extracted from the guts of No. 1 larvae (electronic supplementary material, figure S5B). Based on immunohistology, ZnNLaz1 signals were observed in the mid- and hindguts in the No. 1 larvae at day 3 after their appearance (figure 3a,d,e). In particular, specific signals were localized in the intestinal walls of the guts, especially at extranuclear regions (figure 3g,h). In the negative control (without a ZnNLaz1 primary antibody), ZnNLaz1 signals were not observed (figure 3b,c).

Figure 3.

ZnNLaz1 protein localization in the No. 1 larvae at day 3 after their appearance. (a) Sagittal section with hematoxylin and eosin staining. (b) DAPI staining without a ZnNLaz1 primary antibody. (c) ZnNLaz1 immunostaining without a ZnNLaz1 primary antibody (only with secondary antibody). (d,f) DAPI staining or (e,g) ZnNLaz1 immunostaining with ZnNLaz1 primary and secondary antibodies. Merged DAPI staining and ZnNLaz1 immunostaining images (h). foregut (fg), midgut (mg), hindgut (hg), fat body (fb). Scale bars indicate 500 µm (a), 300 µm (b−e), 50 µm (f−h). (Online version in colour.)

4. Discussion

(a). Molecular evolution of NLaz during termite social evolution

Based on the transcriptome and gene function analyses, we identified a candidate determinant for soldier differentiation, lipocalin family NLaz protein, in Z. nevadensis. Compared with other proteins lipocalins are generally divergent in animals due to their high rates of amino acid substitution [28]. In termites, a lipocalin gene, which is specifically expressed in the soldier caste of Hodotermopsis japonica (= sjostedti), is predicted to be involved in social communication between soldiers and other colony members [29]. In this study, the termite NLaz1 (Znev_05665, XP_023718675 and MN_004031) and C. punctulatus homologue (Comp41569_c0_seq1) evolved under positive selection after their divergence from the other NLaz homologues. Further sequencing efforts from termites and closely related insects may clarify how key amino acid substitutions, probably including the S to K change observed here, contributed to the social underpinnings of soldier differentiation.

(b). Molecular function of ZnNLaz1 and candidate ligands

Animal lipocalins have the capacity to bind multiple ligands including pheromonal substrates [30]. For example, NLaz of D. melanogaster has a role in transporting small hydrophobic ligands and in metabolic regulation in the fat body [27]; it binds some small molecular substances (e.g. palmitic acid and arachidic acid) as well as a cuticle hydrocarbon (7-tricosene) used as a sex pheromone [31]. Although the accurate expression sites of ZnNLaz1 are unclear, its product may bind with small hydrophobic molecules as shown in D. melanogaster. The cuticular hydrocarbon profiles of Z. nevadensis are known to differ among reproductives, soldiers and worker-like larvae [32]. Moreover, two hydrocarbons derived from the accessory mandibular gland serve as individual recognition pheromones in Z. angusticollis [33]. It is possible that chemical substances used for the synthesis of other hydrocarbons (e.g. methyl-branched hydrocarbons) are produced by gut microorganisms in Z. nevadensis [34]. Further biochemical analysis is needed to confirm the hypothesis that specific components (such as hydrocarbons and/or other small molecules discussed above) in the gut fluids of the queen are candidate pheromonal ligands that promote the trophallactic behaviour that induces presoldier differentiation.

(c). Role of ZnNLaz1 for soldier differentiation

Knockdown of ZnNLaz1 expression resulted in decreased frequency of trophallaxis from the queen and reduced presoldier differentiation rates (figure 2). Proctodeal trophallactic behaviours are broadly conserved among lower termite species, and these behaviours serve to integrate the nutritional, microbial, and social environments of the colony [35]. There is therefore a possibility that ZnNLaz1 plays a role in the integration and coordination of social signals in the termite gut through the regulation of proctodeal trophallactic behaviours.

In social insects, juvenile hormone (JH) is the central endocrine regulatory factor for caste differentiation, and termite soldier differentiation requires an increased JH titre in workers prior to moult [4]. ZnNLaz1 in the No. 1 larvae was differentially expressed at day 3 after their appearance compared with other developmental stages, and concurrent with an increase in JH biosynthetic gene expression [8]. Previous studies showed that NLaz expression levels were involved in the regulation of fat storage and food intake through Insulin/IGF signalling (IIS) in adult flies and ant queens [27,36], suggesting that there are regulatory mechanisms for energy metabolism through ZnNLaz1 functions in the No. 1 larvae. Although it is unknown whether ZnNLaz1 regulates endogenous JH titre changes via energy metabolism through the regulation of IIS, the frequent interactions with the queen may determine the developmental fate of the soldier caste, potentially derived from food intake. In addition, brain dopamine level changes were also involved in the regulation of proctodeal trophallaxis between the No. 1 larvae and reproductives (both queen and king) [9]. Taken together, trophallactic behaviour promoted by high brain dopamine levels as well as the ZnNLaz1-mediated social interaction with the queen may contribute to JH titre changes during the ontogeny of presoldier differentiation. Whether the reduced proctodeal trophallaxis observed in ZnNLaz1-RNAi-treated individuals is an effect of the differentiation into workers rather than a mediator remains unknown. To clarify the relationship between ZnNLaz1 and trophallactic behaviour, crosstalk between ZnNlaz1 and intrinsic (i.e. hormonal and neuronal) cascades should be analysed in detail.

(d). Evolution of soldier caste through inter-individual communications

In the mature colonies of Z. nevadensis, soldiers are usually differentiated from seventh instar larvae [10,37]. However, it is difficult to identify the soldier-destined individuals in mature colonies, because the soldier ratio is usually quite low compared with workers (below 10% [38]). Soldier differentiation is generally promoted by interaction with reproductives [4,39], and proctodeal trophallaxis between colony members is frequently observed in mature colonies of Z. nevadensis [40]. Consequently, even in mature colonies, there is a possibility that ZnNLaz1-mediated social interaction is crucial for soldier differentiation.

In incipient colonies of termites, the first soldier plays an essential role in taking over the task of nest defence from reproductives, allowing the king and queen to devote themselves to reproduction [3]. The familial social structure of incipient colonies in termites resembles that of the subsocial cockroach Cryptocercus spp., the sister-group of termites, with the exception that termites have a soldier caste [41]. Trophallactic behaviours in Cryptocercus were observed from a donor adult, primarily the female, to all of the nymphs [42]. In Z. nevadensis, however, the frequencies of these behaviours are strongly biased toward the No. 1 larvae [5]. Although in termites proctodeal trophallactic behaviours are triggered by the recipients [43], reproductive castes (both queen and king) are the only trophallactic donors in the incipient colony. Consequently, the acquisition of a novel NLaz function that promotes trophallaxis with the queen may contribute to the evolution of a soldier caste in termites. From this point, further molecular and behavioural analysis of ZnNLaz1-mediated interactions among nest members should be performed in order to elucidate the mechanisms underlying social evolution from a familial lifestyle such as that in Cryptocercus, to one that includes the sterile soldier caste characteristic of termites.

5. Conclusion

RNA-seq analysis was performed during termite caste differentiation under natural conditions. A homologue of Neural Lazarillo (ZnNLaz1) was found to be the gene most highly expressed in soldier-destined larvae of Z. nevadensis. A high ω ratio was observed in the branch leading to the termite NLaz1 and its Cryptocercus homologue. RNAi-mediated knockdown analysis indicated that ZnNLaz1 expression level in soldier-destined larvae was involved in the regulation of trophallactic behaviour with the queen. Our results suggest that termite social evolution within the cockroach lineage could be promoted by the acquisition of a novel NLaz function that regulates the social interactions involved in producing the soldier caste.

Ethics

All procedures including Recombinant DNA Experiments were performed according to the University of Toyama ethics regulations and guidelines.

Data accessibility

RNA-seq data and DNA sequences obtained are available from the DDBJ Sequence Read Archive (DRA) database (accession no. DRA006998) and DDBJ/EMBL/GenBank databases (no. LC382016), respectively. All other relevant data are within the paper and its electronic supplementary material.

Authors' contributions

H.Y. and K.M. designed the study; H.Y., Y.M., K.T. and K.M. collected samples; H.Y., S.S. and Y.M. performed the experiments; H.Y., S.S., Y.H., S.M. and K.T. analysed the data; H.Y., S.S. and K.M. drafted the manuscript; all authors contributed to the final version of the manuscript.

Competing interests

The authors have declared that no competing interests exist.

Funding

This study was partly supported by JSPS KAKENHI grant no. JP25128705 and JP16K07511, and by Research Project Promotion grant (no. 16SP01302) and ‘the Spatiotemporal Genomics Project' of University of the Ryukyus.