Mature R. speratus colonies were collected in Toyama Prefecture, Japan, in November 2018 and 2019. Pieces of logs housing the termites were brought to the laboratory and kept in plastic cases in constant darkness. Three colonies collected in 2018 (colonies A–C) were used to validate the injection volume. A colony collected in 2019 (colony D) was used for the RNAi analysis.

Total RNA for double-stranded RNA (dsRNA) synthesis was extracted from whole bodies of the 5th-6th stage workers (old-age workers) (five individuals in each sample) using ISOGEN II (Nippon Gene, Tokyo, Japan). Old-age workers were discriminated from other developmental stages based on the body size and antennal segments (26, 27). The extracted total RNA was purified using DNase treatment to remove genomic DNA. RNA purity and quantity were measured using a NanoVue spectrophotometer (GE Healthcare BioSciences, Tokyo, Japan). cDNA was synthesized from the purified RNA using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster, CA, USA).

The ecdysone receptor homolog of R. speratus (RsEcR) was obtained from genome sequence data [RsEcR (gene ID: RS006194; 8)]. Using gene-specific primers ( Supplementary Table 1 ; Supplementary Figure 1 ), the dsRNA of RsEcR was amplified. RsEcR-specific primers were adapted to the T7 promoter sequences. Gene-specific primers were designed using the Primer3 Plus software (28). To obtain template dsRNA, amplified RsEcR dsRNA was purified using a QIAquick gel extraction kit (Qiagen, Tokyo, Japan). In accordance with previous studies (14, 16, 24, 29), the GFP sequence for the control experiment was amplified using the GFP vector pQBI-poll I (Wako, Osaka, Japan). GFP-specific primers with T7 promoter sequences were newly designed using the Primer3 Plus ( Supplementary Table 1 ; Supplementary Figure 2 ), because we intended to synthesize a shorter dsRNA (about 300 bp) (see Results). RsEcR and GFP dsRNA were synthesized using the MEGA script T7 Transcription Kit (Invitrogen, Carlsbad, CA, USA).

Nuclease-free water (NFW: solvent of dsRNA; 50.6, 101.2, 151.8, 202.4, 253.0, 303.6 354.2, and 404.8 nL, respectively) was injected into the lateral thorax of old-age workers of R. speratus (n = 60 per treatment) using a Nanoliter 2000 microinjector (World Precision Instruments, Sarasota, FL, USA) attached to a glass capillary. These individuals were kept in 65 mm petri dishes with 55 mm filter paper treated with 80 μg juvenile hormone III (JH III; Santa Cruz Biotechnology, Dallas, TX, USA) or 40 µg 20-hydroxyecdysone (20E; Sigma Aldrich, St. Louis, MO, USA) dissolved in 200 μL of acetone (20 individuals in each dish), in accordance with previous studies (19, 20, 30). All dishes were maintained in constant darkness at 25°C for 2 weeks and checked every 24 h to monitor the individuals. We calculated the rates of presoldier molt, worker molt and mortality. The rates of presoldier and worker molting individuals were percentages of those in survived individuals. The mortality were percentages of those in all the treated individuals. Statistical analysis was performed using two-way ANOVA followed by Tukey’s test with Mac statistical analysis ver. 3.0 (Esumi, Tokyo, Japan). Prior to the use of the ANOVA, we performed the Levene’s test on the variance equality using Mac statistical analysis ver. 3.0 (Esumi).

RNAi was performed according to methods described in the previous study (24). RsEcR and GFP dsRNA (2 μg/151.8 nL) were injected into the side of the thorax of old-age workers of R. speratus (n = 50 per treatment) following the method described above. RNAi-treated individuals were kept in 65-mm petri dishes with 55-mm filter paper treated with 80 μg JH III (Santa Cruz Biotechnology) or 20 µg 20E (Sigma Aldrich) dissolved in 200 μL acetone (10 individuals in each dish). Before the RNAi experiment, workers collected from colony D were treated with different 20E concentrations (40, 20, 10, and 5 µg), because high mortality were observed when workers were treated with 40 µg 20E (see Results). These petri dishes were kept in an incubator at 25°C for 2 weeks to monitor the rates of molted presoldiers and workers. Induced and dead individuals were immediately removed from dishes and preserved in FAA solution (ethanol:formalin:acetic acid = 16:6:1) for 24 h and stored in 70% ethanol.

To check for the knockdown of gene expression, RNAi-treated individuals of R. speratus (n = 6 per treatment) were collected 6 and 9 days after RNAi treatment. Total RNA was extracted from the whole body of each individual using ISOGEN II (Nippon Gene, Tokyo, Japan). As described above, extracted RNA was used for cDNA synthesis, which was prepared for gene expression analysis of RsEcR. Relative quantification of transcripts was performed using PowerUp SYBR Green Master Mix (Applied Biosystems, Foster, CA, USA) and the QuantStudio 3 Real-Time PCR System (Applied Biosystems, Foster, CA, USA). Gene-specific primers for qPCR were designed using Primer3 Plus ( Supplementary Table 1 ). According to the previous study (31), the suitability of six reference genes was evaluated using GeNorm (32) and NormFinder (33) software [EF1-alpha (accession No. AB602838; 34), NADH-dh (No. AB602837; 34), beta-actin (No. AB520714; 35), glutathione S-transferase 1 (GstD1, gene ID: RS001168; 8), ribosomal protein S18 (RPS18, ID: RS015150; 8), and eukaryotic initiation factor 1A (eIF-1A, ID: RS005199; 8)]. These were used as candidate reference genes in the respective termite species (24) and other insects, including Drosophila melanogaster and Apis mellifera (36, 37). Expression levels were calculated using biological replications (number of replications: n = 6). Statistical analysis was performed using the Mann-Whitney U test with Mac statistical analysis ver. 3.0 (Esumi).

To evaluate the effects of RsEcR RNAi, we calculated the rates of individuals with gut purging (elimination of gut contents before the molt), presoldier induction and worker molting. The rates of gut-purged individuals were percentages of those in all the treated individuals. The rates of presoldier induction and worker molting individuals were percentages of those in gut-purged individuals. Statistical analysis was performed using Fisher’s test with Mac statistical analysis ver. 3.0 (Esumi).

To determine the optimal injection volume, we injected various volumes of NFW into the hormone-treated individuals. Presoldiers were strongly induced from workers (non-injection) by JH III treatment [73.3% ± 10.4% (colony A, mean ± SD) and 80.0 ± 0% (colony B)], and no presoldiers were observed in the control (acetone) treatment of both colonies (0%) ( Figure 1A ; Supplementary Table 2 ). The rates of presoldier molt in NFW-injected individuals varied among treatments with different injection volumes [two-way ANOVA, colony: p = 0.96, injection volume: p = 1.31E-17, interaction (colony vs injection volume): p = 0.86] ( Figure 1A ; Supplementary Table 2 ). In the treatments with 50.6–253.0 nL NFW injection, rates of presoldier molt were essentially similar to those of non-injection in both colonies (about 60–80%). However, presoldier molting rates were drastically dropped in the treatments with 303.6–404.8 nL injection (below 30%). Mortality were significantly different among treatments with NFW injection volumes after feeding JH III [two-way ANOVA, colony: p = 0.8, injection volume: p = 2.43E-7, interaction (colony vs injection volume): p = 0.44] ( Figure 1B ; Supplementary Table 2 ). High mortality was observed in treatments with large injection volumes.

Worker molt was strongly induced in workers (non-injection) by 20E treatment [83.3 ± 10.4% (colony A) and 75.0 ± 13.2% (colony C)], but almost never occurred in the control treatment [5.0% ± 5.0 (colony A) and 3.3 ± 2.9% (colony C)] ( Figure 2A ; Supplementary Table 3 ). The rates of worker molt were also affected by the different NFW injection volumes [two-way ANOVA, colony: p = 7.87E-5, injection volume: p = 3.37E-19, interaction (colony vs injection volume): p = 4.2E-3]. In the treatments with 50.6–253.0 nL injection, rates of worker molt were similar to those of non-injection in both colonies (about 60–80%). However, molting rates were significantly lower than those of non-injected individuals in the treatments with 303.6–404.8 nL injection in both colonies (below 40%). Mortality was significantly different among treatments with NFW injection volumes after feeding 20E [colony: p = 2.64E-3, injection volume: p = 1.58E-11, interaction (colony vs injection volume): p = 0.43] ( Figure 2B ; Supplementary Table 3 ). High mortality were observed in the treatments with large injection volumes, similar to JH III application.

Termite soldier differentiation requires high JH titers in the worker body (38, 39). Moreover, in insects, 20E titer levels generally increase before the larval molt (40). There is a possibility that the reduction in the molting rates is due to the dilution of hormone titer levels with the injected NFW (303.6–404.8 nL). Alternatively, the high mortality observed with large volumes of NFW injections may be due to some mechanical effects using pure water. To clarify this possibility, further injection analysis using a physiological saline solution should be performed. In the damp-wood termite Hodotermopsis sjostedti, RNAi-based knockdown was effectively induced by the injection of 1 µL dsRNA solution to the 7th instar treated with JH analog (15). We suggest that the proper (non-lethal) volumes of injection should be determined, especially considering the body size of the target individuals, because H. sjostedti 7th instars are much larger than the R. speratus workers used in this study. For RNAi-based knockdown with hormone treatment in R. speratus, injection volumes should be below 253 nL.

A previous study showed that RNAi-based knockdown was effectively caused by injection of 2 µg dsRNA (about 260 bp) in R. speratus nymphs without hormone treatment (24). In Drosophila S2 cells, the length of dsRNA for effective RNAi was shown to be >211 bp (41). In this study, we prepared 2 µg dsRNA (about 300 bp) dissolved in 151.8 nL. According to the results described above, injection volumes should be reduced as small as possible (50.6 nL in this study). However, when we used injection volumes of 50.6 nL, a glass capillary was immediately clogged probably due to high viscosity. We then selected injection volumes of 151.8 nL, and injected GFP or RsEcR dsRNA solution (2.0 µg/151.8 nL) into R. speratus workers. EF1-alpha was selected as the most appropriate reference gene for real-time qPCR analyses ( Supplementary Table 4 ). Gene expression levels of RsEcR were not affected by RsEcR RNAi 6 days after injection (day 6, Figure 3A ; Supplementary Table 5 ). However, the expression levels of RsEcR were significantly decreased by RsEcR RNAi on day 9 (more than 50%) compared to the GFP control ( Figure 3B ; Supplementary Table 5 ). We did not observe any gross morphological and phenotypic changes in the RNAi-treated individuals without hormone treatments.

In RsEcR RNAi with JH treatment, the rates of gut-purged individuals (26%) were significantly lower than those in the GFP control (64%) (Fisher’s test, p < 0.05; Figure 4A ; Supplementary Table 6 ). Furthermore, presoldiers emerged from most gut-purged individuals in the GFP control (87.5%; Figures 4B, C ), but never from those in RsEcR RNAi treatment (0%; Figure 4B ; Supplementary Table 6 ). All gut-purged individuals in the latter failed to molt ( Figure 4D ), as shown in Z. nevadensis (18).

Before the experiment with RsEcR RNAi with 20E treatment, mortality of workers with 40, 20, 10, and 5 µg 20E were compared, because high mortality was observed in workers (colony D) treated with 40 µg 20E dissolved in 200 µL acetone. Although there were no statistical differences among treatments, the 20 µg 20E treatment tended to be higher rates of worker molt and lower levels of mortality ( Supplementary Figure 3 ; Supplementary Table 7 ). Consequently, we decided to perform RNAi using 20 µg of 20E treatment. The rate of gut-purged individuals (60%) was significantly lower than that in the GFP control (76%) (Fisher’s test, p < 0.05; Figure 5A ; Supplementary Table 8 ). RsEcR RNAi decreased the rates of worker molt (33.3%) compared to the control (55.3%) ( Figure 5B ; Supplementary Table 8 ), and most RNAi-treated individuals failed to molt ( Figures 5C, D ). These results indicate that the RsEcR RNAi assay performed here (injection of 2 µg dsRNA dissolved in 151.8 nL NFW) was successful in R. speratus workers with hormone treatments. Although the proper amount of dsRNA should be recognized for each gene, we suggest that the concentration obtained here can be used as the starting point for RNAi method during caste differentiation using hormone treatment in this species. According to the present study and previous works (15–18, 24), proper amount of dsRNA may be around 0.5–2 µg for the RNAi experiments in termites.

The relatively weaker effects of RsEcR RNAi treated with 20E ( Figure 5 ), compared to those with JH III ( Figure 4 ), may be due to the timing of knockdown effects caused by RNAi. Knockdown effects of RNAi could not be observed 6 days, but observed 9 days after dsRNA injection ( Figure 3 ). Since the JH application induces the molting event to presoldiers, 20E-EcR action may be promoted after the increase of JH titer in workers treated with JH III. In contrast, 20E-EcR action may be immediately promoted by the 20E treatment in workers. Indeed, the initiation of molting event is occurred early in the 20E treatment, compared to the JH treatment (approximately 10-11 or 13-14 days after the treatment, respectively; 20). To clarify the possibility of the different timings of physiological action after each treatment, further expression analysis of hormone-related genes should be performed during caste differentiation.