Toll signalling promotes blastema cell proliferation during cricket leg regeneration via insect macrophages

Tissue regeneration allows the restoration of lost tissues from cells. Various animals, including planarians, insects, fishes, newts and frogs, have regenerative abilities. The regenerative abilities of amniotes, including chicks, mice and humans, are limited (Agata and Inoue, 2012). One crucial difference between regenerative and non-regenerative animals is the ability to form a blastema, a population of stem cells or dedifferentiated cells that proliferate and differentiate into several types of cells to restore the lost tissue, although some species, such as hydra, do not require blastema formation for regeneration (Agata et al., 2007; Vogg et al., 2019). A key goal in the field of regenerative biology is to identify the factors that trigger blastema formation, which is an initial step in tissue regeneration.

Following tissue injury, a defence response occurs around the wound site. Neutrophils and macrophages expressing proinflammatory genes migrate to the wound region to eliminate infectious microbes and clear debris from injured cells (Anders and Schaefer, 2014; Westman et al., 2020). These phagocytic cells also express pattern recognition receptors, such as Toll-like receptors (TLRs), to detect infectious microbes, and Janus kinase/Signal transducer and activator of transcription (JAK/STAT) signalling components, including interleukin receptors (IL-Rs), to receive cytokine (Hu et al., 2007). In vertebrates, infectious microbes are directly detected by TLRs (Kawai and Akira, 2011; Pandey et al., 2014; Szatmary, 2012). In insects, infectious microbes, such as gram-positive bacteria, yeasts and fungi, are mostly detected by Toll via proteoglycan recognition proteins (PGRPs) and clip-domain serine proteinases (clip-SPs), and gram-negative bacteria are detected by PGRP-LC and immune deficiency (Imd) signalling (Fig. S1) (Anthoney et al., 2018; Leulier and Lemaitre, 2008; Myllymäki et al., 2014). In both vertebrates and invertebrates, cytokines are received by IL-Rs and activate JAK/STAT signalling (Arbouzova and Zeidler, 2006).

Recent studies have shown that macrophages promote tissue regeneration in axolotl and zebrafish (Godwin et al., 2013; Petrie et al., 2014). Phagocytic uptake of liposome-encapsulated clodronate (Clo-lipo) is a well-established method for depleting macrophages, as liposome is specifically incorporated into phagocytes. Clodronate induces apoptosis by antagonising ATP metabolism (Van Rooijen and Sanders, 1994). In one study, macrophage-depleted axolotls (Ambystoma mexicanum) treated with Clo-lipo did not regenerate the amputated portions of limbs caused by downregulation of blastema marker genes, whereas regeneration was successful in control axolotls treated with PBS-lipo (Godwin et al., 2013). Axolotl macrophages respond to pathogen-associated molecular patterns (PAMPs) and damage-associated molecular patterns (DAMPs) via TLRs (Debuque et al., 2021). In zebrafish, macrophages infiltrate the affected area to engulf cellular debris (Li et al., 2012), and maintain appropriate levels of inflammation to induce expression of regeneration-promoting genes (Hasegawa et al., 2017). Thus, macrophage-depleted transgenic zebrafish exhibit altered fin regeneration, likely mediated by a reduction in blastema cell proliferation (Petrie et al., 2014). In earthworms, which are regenerative invertebrates, depletion of phagocytic cells impairs tissue regeneration (Bodó et al., 2021). Macrophages promote cell proliferation, even in partially regenerative vertebrates such as Xenopus froglets or mice. When a portion of a limb is amputated in these species, a cartilaginous callus is formed by proliferation of chondrocytes. However, Clo-lipo treatment of either species inhibits callus formation (Miura et al., 2015). Hence, the efficient functioning of macrophages is not sufficient, but is required, for regeneration. The precise signalling pathways that function in phagocytic cells during regeneration remain unclear.

Insects have an open blood-vascular system and their body fluids contain several types of haemocytes, including prohaemocytes, phagocytes and non-phagocytic cells (Hillyer, 2016). Prohaemocytes are the stem cells of other haemocytes. Plasmatocytes and granulocytes of Lepidoptera and plasmatocytes of Drosophila are the primary phagocytic and encapsulating cells involved in defence responses, and are analogous to mammalian macrophages (Browne et al., 2013; Evans et al., 2003; Ribeiro and Brehélin, 2006). Depletion of phagocytic cells using clodronate has also been achieved in mosquitoes and fruit flies (Kumar et al., 2021; Kwon and Smith, 2019).

To investigate the molecular link between insect immunity (Hillyer, 2016) (Fig. S1) and blastema formation during tissue regeneration, we focused on the role of Toll signalling pathways and plasmatocytes during leg regeneration of a hemimetabolous insect, because its regenerative abilities are greater than those of holometabolous insects. The two-spotted cricket Gryllus bimaculatus can restore the lost part of an amputated leg in the nymphal stage (Bando et al., 2017; Mito and Noji, 2008). When we amputated a cricket leg at the tibia, the wound surface was covered by a scab and wound epidermis (Mito et al., 2002). Prior studies have demonstrated that the formation of a blastema (Mito et al., 2002; Nakamura et al., 2008a) occurs through cell proliferation processes regulated by the JAK/STAT and Hippo signalling pathways (Bando et al., 2009, 2013). The lost part of the leg is recognised depending on positional information along the proximodistal axis, mediated by Dachsous/Fat protocadherins (Bando et al., 2009, 2011a,b). This is followed by redifferentiation of blastema cells. During repatterning, leg-patterning genes are re-expressed in the blastema (Ishimaru et al., 2015; Nakamura et al., 2008b) via an epigenetic mechanism (Hamada et al., 2015). In G. bimaculatus, as in other insects, two out of the six types of haemocytes – plasmatocytes and granulocytes – have been reported to respond to infection (Cho and Cho, 2019; Sokolova et al., 2000). The plasmatocytes are insect macrophages; however, the molecular link between immunity and the plasmatocytes is still unclear, although it may involve JAK/STAT and Toll signalling pathways.

In this study, the expression of Toll signalling genes was altered during leg regeneration in G. bimaculatus. RNA interference (RNAi) of Toll family genes resulted in defective regeneration and impaired regeneration phenotypes, indicating that Toll signalling promotes leg regeneration. RNAi of Toll2-2 resulted in downregulated expression of JAK/STAT signalling components and decreased cell proliferation in the blastemas of regenerating legs. We also analysed the role of plasmatocytes, because Toll2-2 is expressed in plasmatocytes. Depletion of plasmatocytes in crickets resulted in decreased cell proliferation in the blastema and failure of leg regeneration. These findings suggest that Toll2-2-expressing plasmatocytes activate blastema cell proliferation, promoting leg regeneration via Toll signalling.

To identify signalling pathways activated in early regeneration processes, RNA-seq analysis was performed and gene expression between regenerating legs at an early phase [3 hours post-amputation (hpa)] and in non-regenerating legs (0 hpa) was compared. All reads obtained from early regenerating legs (RLs) and non-regenerating legs (NLs) were assembled into contigs and each read from RLs and NLs was mapped to the contigs to calculate the reads per kilobase of exons per million reads (RPKM) value. Comparison of RPKM values of each contig (Table S1) indicated that 908 contigs were only expressed in the RLs, and the expression of 2565 contigs was upregulated more than twofold in the RLs (Table 1). To exclude genes expressed at a low level, we omitted contigs for which read counts were <10 and RPKM values were <5, resulting in 59 contigs only expressed in the RLs and 957 contigs upregulated in the RLs being selected for further analyses. To obtain an overview of these contigs, 32 out of 59 and 549 out of 957 contigs, annotated using the BLASTX program (Table S2), were analysed using Blast2GO software. Gene ontology (GO) terms relating to biological processes, molecular functions and cellular components with which more than six contigs were annotated are summarised in Fig. 1 and Fig. S2A. Notably, transcripts related to ‘antimicrobial humoral response’ and ‘haemocyte migration’ in the biological process category were upregulated in the RL group compared with the NL group. These GO terms were not identified in our previous RNA-seq analyses at 24 hpa (Bando et al., 2013). In contrast, GO terms that were upregulated in the 24 hpa RL (Bando et al., 2013), such as ‘cell differentiation’ and ‘cell death’ in biological processes, were not upregulated in the 3 hpa RL (Fig. 1).

We identified transcripts encoding signalling pathway components that were upregulated in the RL (Fig. S2B). Transcripts encoding components related to vascular endothelial growth factor, insulin-like growth factor, fibroblast growth factor, transforming growth factor-beta (TGFβ), Wnt and Toll signalling pathways were upregulated in the RL group compared with the NL group. Previous studies have demonstrated the involvement of JAK/STAT (Bando et al., 2013), TGFβ and Wnt (Ishimaru et al., 2018; Mito et al., 2002; Nakamura et al., 2007), and Notch (Bando et al., 2011b) signalling pathways during cricket leg regeneration. We focused on the Toll signalling pathway in this study, because the function of Toll signalling during regeneration is unclear.

Eleven Toll genes from the Gryllus genome have been identified (Ylla et al., 2021) and designated Toll, Toll2-1-5, Toll6-1-2, Toll7, Toll8 and Toll9, based on amino acid homology with other insect orthologues (Hillyer, 2016) (Fig. 2A, Fig. S3). Five paralogues, Toll2-1 to Toll2-5, are phylogenetically close to termite Zootermopsis nevadensis Toll-like receptor 2, but not close to Drosophila Toll-2; Drosophila Toll-2 was close to Toll-7 (Fig. 2A). We cloned cDNA fragments of all Gryllus Toll genes from cDNA obtained from RLs. In the RNA-seq analysis, expression of Toll2-5, Toll8 and Toll2-2 were upregulated in RLs (Fig. S2B). Quantitative PCR (qPCR) examination of temporal expression changes in the Gryllus Toll genes revealed more than twofold upregulated expression of Toll2-1, Toll2-2 and Toll2-5 in RLs at 3, 24 and 48 hpa compared with RLs at 0 hpa (Fig. 2B). In particular, the expression of Toll2-1 and Toll2-2 was upregulated more than 3-fold and 5-fold, respectively, at 3 hpa. Their elevated expression was maintained at 24 hpa and slightly decreased at 48 hpa. Expression of Toll2-5 was gradually upregulated from 3-fold to 10-fold from 3 to 48 hpa. Conversely, expression levels of Toll2-4, Toll6-1, Toll6-2, Toll7, Toll8 and Toll9 were reduced to less than 50% at certain time points during regeneration (Fig. 2B). Expression changes of Toll and Toll2-3 were significant, but were upregulated to less than twofold (Toll) or downregulated more than 50% (Toll2-3) at 3, 24 and 48 hpa (Fig. 2B).

The observed changes in expression prompted us to examine further the involvement of the Gryllus Toll genes in leg regeneration. To identify regeneration-associated Toll genes, RNAi targeting each of the Toll genes was performed at the third instar. The morphologies of the RLs were compared with the morphologies of control legs at the fifth instar. Double-stranded RNA (dsRNA) targeting DsRed was used as a nonspecific control. We performed qPCR to estimate the reduction in mRNA levels of target genes in regenerating tibiae of RNAi crickets compared with those of DsRed^RNAi crickets at 48 hpa. RNAi against each Toll gene significantly reduced the transcript level to <50% of each control level (Fig. S4A), confirming the efficiency of RNAi.

Next, the phenotypes of RLs following RNAi targeting genes encoding Toll signalling molecules (Fig. 3B) at the fifth instar were determined and were categorised as class 1, class 2 and class 3 (Fig. 3A,C, Fig. 4). Class 3 was the normally regenerated leg found in the DsRed^RNAi crickets. In this class, the lost portion of the tibia was regrown, and tibial spurs were reconstructed at the distal end of the regrown tibia. The lost tarsus was reconstructed and segmented into tarsal segments 1 and 3 and the claws (Fig. 4). Class 2 was defined as an impaired regeneration phenotype. In this phenotype, the lost part of the tibia was not regrown, and tibial spurs were not reconstructed. The lost tarsus was reconstructed but was small and displayed an abnormal morphology. Class 1 was defined as a defective regeneration phenotype, in which no regeneration occurred, with no reconstruction of the lost tarsus (Fig. 3A, Fig. 4). In the RNAi experiments (Fig. 3C), >40% of crickets displayed class 1 phenotypes when RNAi targeted Toll2-1 or Toll2-2 (Fig. 3C). In addition, 18.8% and 32% of Toll2-1^RNAi and Toll2-2^RNAi crickets exhibited the class 2 phenotype, respectively; therefore, >70% of these RNAi crickets showed abnormalities during leg regeneration. These data, together with the observed expression changes during regeneration (Fig. 2B), indicate that Toll2-1 and Toll2-2 are important in leg regeneration.

To analyse further the role of Toll signalling, RNAi targeting ligands and intracellular component genes of the signalling (Fig. 3B) was performed. Whereas Drosophila has six spz paralogues (Viljakainen, 2015), we found two spz paralogues, which encode Toll ligands, in the Gryllus genome. Spz and Spz2 are paralogous to Drosophila Spz and Spz5, respectively. RNAi against Gryllus spz or spz2 resulted in class 1 phenotypes occurring at rates of 21.1% and 16.7%, respectively (Fig. 3C,D), and those against MyD88, tube, pelle or TRAF6, which encode intracellular signalling molecules of Toll signalling (Fig. 3B), were 57.1%, 10.0%, 11.1% and 16.7%, respectively (Fig. 3C,D). The percentage of class 1 and class 2 phenotypes after MyD88^RNAi was 78.5%, which is comparable to that of Toll2-1^RNAi (75.1%) or Toll2-2^RNAi (72.0%) (Fig. 3C), suggesting that Toll2-1 and Toll2-2 may play crucial roles in leg regeneration via MyD88.

Next, we performed RNAi against dorsal (dl) and Dorsal-related immunity factor (Dif), both of which encode nuclear factor-kappa-B (NF-κB) transcription factors that function in the Toll signalling pathway (Fig. 3B) (Lindsay and Wasserman, 2014). Single RNAi against dl or Dif showed weak defects in leg regeneration. For example, 18.2% and 50.0% of Dif^RNAi cricket legs showed class 1 and 2 phenotypes, respectively (Fig. 3C,D). The proportion of the class 1 phenotype caused by Dif^RNAi was slightly increased in crickets with dual RNAi against dl and Dif (Fig. 3C). Given that Dl and Dif function downstream of both Toll and TNF signalling pathways (Igaki and Miura, 2014), we performed RNAi against eiger (egr) and wengen (wgn), which encode TNFα and TNF receptors, respectively. Leg regeneration occurred normally in egr^RNAi and wgn^RNAi crickets (Fig. S7A), despite the reduction of egr or wgn mRNA to approximately 10% compared with DsRed^RNAi controls (Fig. S7B). These RNAi results indicate that Spz and Spz2 could promote leg regeneration through Toll receptors, MyD88 and Dl/Dif dimers.

Because Toll2-1^RNAi and Toll2-2^RNAi crickets showed the most severe defects in leg regeneration (Fig. 3C), we quantified the mRNA levels of all Gryllus Toll genes in RLs of Toll2-1^RNAi and Toll2-2^RNAi crickets at 48 hpa (Fig. S5A,B). In the Toll2-1^RNAi RLs, the relative amount of Toll2-2 mRNA was significantly decreased to 54%, whereas Toll and Toll2-4 mRNAs were significantly increased to 160% and 152%, respectively (Fig. S5A). In the Toll2-2^RNAi RLs, the relative amounts of Toll and Toll6-2 mRNA were slightly reduced to 75% and 71%, respectively (Fig. S5B), but expression levels of other cricket Toll genes, including Toll2-1, were not significantly changed. These results suggest that the RNAi phenotypes observed in Toll2-1^RNAi crickets might be caused by an additional reduction in the Toll2-2 mRNA level. Nucleotide alignment indicated that an off-target effect was unlikely (Fig. S5C), implying a possible epistatic regulation of Toll2-2 by Toll2-1.

We next focused on Toll2-2 function during cricket leg regeneration by carrying out single-gene functional analysis by RNAi. Notably, 28% of Toll2-2^RNAi crickets exhibited a class 3 phenotype (Fig. 3C) and 36% of endogenous Toll2-2 mRNA remained after Toll2-2^RNAi (Fig. S4A), implying that RNAi against Toll2-2 could show dose dependency. We injected a larger volume of Toll2-2 dsRNA solution (483 nl) and compared the resulting phenotypes with those obtained from standard Toll2-2^RNAi (207 nl). RNA reduction and phenotype ratio were not significantly different in either experimental condition (Fig. S6); therefore, for subsequent analyses, we performed RNAi using standard conditions.

To examine whether the class 1 or 2 phenotype observed in Toll2-2^RNAi crickets (Fig. 3A,C, Fig. 4) was caused by abnormal regulation of cell proliferation in the blastema, we analysed cell proliferation in Toll2-2^RNAi crickets at 48 hpa. The distribution of S-phase proliferating cell nuclei and total nuclei, revealed by 5-ethynyl-2-deoxyuridine (EdU) incorporation assay and Hoechst 33342 staining, respectively, was determined in Toll2-2^RNAi and DsRed^RNAi RLs (Fig. 5A). The proportions of S-phase nuclei out of all nuclei in the Toll2-2^RNAi and DsRed^RNAi RLs were 37.41±5.80% (n=4) and 3.63±2.26% (n=3), respectively (Fig. 5B). M-phase proliferating cell nuclei were also revealed by anti-phosphorylated histone H3S10 antibody (PH3) staining (Fig. 5A). The proportions of the number of M-phase nuclei out of total nuclei in the control and Toll2-2^RNAi RLs were 4.05±1.53% (n=4) and 1.15±0.26% (n=3), respectively (Fig. 5B). The proportions of S- and M-phase cells were significantly decreased by Toll2-2^RNAi in the 48 hpa blastema. These histochemical observations were validated by qPCR of Gryllus cyclin genes. The relative amounts of cycE and cycB mRNA were significantly decreased to 49% and 47%, respectively, in Toll2-2^RNAi RLs compared with those of DsRed^RNAi (Fig. 5C).

Given that transcript levels of Gryllus cyclin genes are decreased in the regenerating leg after RNAi against JAK/STAT signalling component genes (Bando et al., 2013) (Fig. S1), we quantified the expression levels of the Gryllus homologues of JAK/STAT signalling component genes – unpaired (upd), domeless (dome), hopscotch (hop) and Stat in Toll2-2^RNAi RLs. Expression of upd, dome and hop was downregulated by Toll2-2^RNAi in RLs compared with those of DsRed^RNAi (Fig. 5D). To determine whether Toll2-2 signalling might regulate transcription of Toll ligand genes, we examined the expression of spz and spz2 and found that it was decreased in Toll2-2^RNAi RLs (Fig. 5E). These results suggest that Toll2-2 signalling upregulates the expression of its own ligands, spz/spz2, as well as JAK/STAT signalling component genes and cyclins, or that Toll2-2 affects localisation of the ligand-expressing cells during Gryllus leg regeneration.

The results described above indicated that Toll2-2 promotes cell proliferation during regeneration. In vertebrates, TLRs localise to the cell membrane of macrophages to detect infectious microbes (Satoh and Akira, 2016). In insects, including G. bimaculatus, plasmatocytes are professional phagocytic cells and are regarded as insect macrophages (Cho and Cho, 2019). To visualise the cricket presumptive plasmatocytes, India ink and boron-dipyrromethene-cholesterol within liposomes (BODIPY-lipo) were injected into the haemolymph of third instar nymph abdomens. India ink and BODIPY-lipo are characteristically incorporated into plasmatocytes. Thus, Gryllus plasmatocytes were identified by black droplets of India ink in Giemsa-stained smears and by green fluorescence of BODIPY-lipo (Fig. 6A). Phalloidin staining revealed the actin cytoskeleton of haemocytes and part of the haemocytes incorporated with BODIPY-lipo (Fig. 6B). During leg regeneration, plasmatocytes accumulated near the blastema region of the regenerating tibia, and some plasmatocytes remained in the femur (Fig. 6C). Accumulation of plasmatocytes into the blastema region was observed at 48 hpa compared with the pattern at 0 hpa (Fig. 6D), suggesting that plasmatocytes accumulate at the wound region following amputation.

To clarify the role of plasmatocytes in leg regeneration, plasmatocytes were depleted using 100 nm diameter Clo-lipo. The mean lifespan was estimated as the day when the survival rate was 50% after injection of Clo-lipo and PBS-lipo (liposome-encapsulated PBS buffer for control experiments). The mean lifespan of Clo-lipo-injected crickets was 15 days (Fig. S8A). To confirm the depletion of plasmatocytes, BODIPY-lipo was injected 24 h after the injection of Clo-lipo or PBS-lipo. The number of BODIPY-labelled plasmatocytes was decreased in the haemolymph of Clo-lipo-injected crickets compared with PBS-lipo-injected crickets (Fig. S8B). We also confirmed the depletion of plasmatocytes in the RLs at 48 hpa: Phalloidin-positive haemocytes were present beneath the epidermal cells in PBS-lipo- or Clo-lipo-injected RLs (Fig. 6E). However, BODIPY-incorporated plasmatocytes were not present in the blastema region of Clo-lipo-injected RLs (Fig. 6E).

To clarify the relationship between plasmatocyte function and leg regeneration, we observed the leg regeneration processes of Clo-lipo- and PBS-lipo-injected crickets. In PBS-lipo crickets, the wound surface was covered by a new cuticle at the fourth instar and the lost parts of the legs were regenerated by the sixth instar in a regeneration process that was indistinguishable from that of untreated crickets (Fig. 7A). In contrast, in Clo-lipo-injected crickets, the wound surface was covered by a new cuticle at the fourth instar, but the lost parts of the legs were not regenerated (class 1 phenotype) in 30/39 (76.9%) crickets or were regenerated in a small-malformed structure (class 2 phenotype) in 9/39 (23.1%) crickets (Fig. 7A). Reduction of plasmatocytes in the RLs at 48 hpa was quantified by qPCR for Gryllus homologues of croquemort (crq) and glial cells missing (Gcm), which are expressed in Drosophila plasmatocytes (Wang et al., 2014). A significant reduction in both gene expression was observed. Expression of crq was downregulated in Clo-lipo RLs compared with PBS-lipo RLs (P<0.001 by Student's t-test; Fig. 7B), indicating that Clo-lipo injection led to a decrease in the number of plasmatocytes in RLs .

The distribution of proliferating cells in the S and M phases was examined using the EdU-incorporation assay and anti-phosphorylated histone H3S10 antibody staining, respectively (Fig. 7D). In cricket RLs, the blastema appears as a whitish tissue localised in the distal region (Fig. 7C, brackets). In Clo-lipo RLs, however, scabs adhered to the wound surface. Thus, the wound epidermis and tibial epidermis were fragile when we removed the scab and cuticle from the RLs (Fig. 7C). The proportions of S- and M-phase cell nuclei out of total cell nuclei in the PBS-lipo- and Clo-lipo-injected RLs at 48 hpa were 27.57±7.83% (n=3) and 9.73±5.07% (n=3) for S phase, and 5.72±1.173% (n=3) and 2.35±0.14% (n=3) for M phase, respectively (Fig. 7E), a significant reduction in Clo-lipo RLs compared with PBS-lipo RLs (Fig. 7D,E). The relative amount of cycE and cycB transcripts was reduced to 45% and 27%, respectively, in Clo-lipo RLs compared with PBS-lipo RLs, as revealed by qPCR (Fig. 7F). The reduction of S- and M-phase cell nuclei and reduced expression of cycE and cycB observed in Clo-lipo-injected RLs (Fig. 7) were similar to those of the Toll2-2^RNAi RLs (Figs 3, 4 and 5).

To clarify the link between plasmatocytes and Toll signalling, we sought to determine the Toll genes expressed in the plasmatocytes during leg regeneration. Haemolymph was collected from PBS-lipo-injected and Clo-lipo-injected regenerating cricket nymphs at 48 hpa. After RNA extraction and reverse transcription, the expression levels of Toll genes in the haemolymph samples were compared by qPCR. Because the number of plasmatocytes was lower in Clo-lipo-injected haemolymphs (Fig. S8B), transcripts expressed in other haemocytes might be relatively increased in mRNA from Clo-lipo-injected crickets compared with that of PBS-lipo-injected crickets. mRNA expression levels of Toll2-1, Toll2-2 and Toll2-5 were significantly decreased in Clo-lipo-injected haemolymph (Fig. 8A), suggesting that Toll2-1, Toll2-2 and Toll2-5 are dominantly expressed in plasmatocytes. In contrast, expression levels of Toll2-4 and Toll6-1 were significantly increased and that of Toll2-3 was slightly increased in Clo-lipo-injected haemolymph (Fig. 8A), suggesting that Toll2-4 and Toll6-1 genes might be expressed in other types of haemocytes in addition to phagocytic plasmatocytes.

Next, the distribution of plasmatocytes in the Toll2-2^RNAi RLs was visualised and compared with those in DsRed^RNAi, PBS-lipo-injected, and Clo-lipo-injected RLs. BODIPY-lipo-incorporating plasmatocytes accumulated in the blastemas in the DsRed^RNAi and PBS-lipo-injected RLs (Fig. 8B, left). In contrast, few plasmatocytes accumulated in the Clo-lipo-injected blastema (Fig. 8B, bottom right), similar to what was seen in the Clo-lipo-injected haemolymphs (Fig. S8B). The number of BODIPY-lipo-incorporated plasmatocytes was decreased in Toll2-2^RNAi blastemas compared with DsRed^RNAi blastema (Fig. 8B, top). The BODIPY-lipo-incorporated plasmatocytes in the blastema were counted and the number of cells in the same volume of blastemas was calculated for comparison. In DsRed^RNAi and PBS-lipo-incorporated RLs, 16.6±3.5 (n=4) and 18.6±3.7 (n=7) plasmatocytes, respectively, accumulated in the blastema in a volume of 2.5×10⁵ µm³ . In contrast, accumulated plasmatocytes were significantly decreased to 10.7±2.1 and 3.4±2.3 in the Toll2-2^RNAi (n=4) and Clo-lipo-incorporated (n=6) RLs, respectively, in the same volume (Fig. 8C), indicating that Toll2-2^RNAi reduced the accumulation of plasmatocytes into the blastemas.

In vertebrates, activated macrophages release cytokines to stimulate other immune cells. In Drosophila, haemocytes release insect cytokines, including Upd and Spz, to induce an immune response (Agaisse et al., 2003; Shaukat et al., 2015). In Clo-lipo RLs, the relative expression levels of upd, spz and spz2 were significantly decreased to 43%, 49% and 48%, respectively, compared with PBS-lipo RLs (Fig. 8D,E). Taken together with the downregulation of upd, spz and spz2 in Toll2-2^RNAi RLs (Fig. 5D,E) and defective regeneration phenotypes caused by RNAi of each of these genes (Fig. 3C,D), the collective data suggest that Upd and Spz/Spz2 activate the JAK/STAT and Toll signalling pathways, respectively, during Gryllus leg regeneration. Thus, plasmatocytes that accumulate in RLs release cytokines, which activate the JAK/STAT and Toll signalling pathways, leading to blastemal cell proliferation during cricket leg regeneration.

In mammals, PAMPs, including lipopolysaccharides, lipoproteins and flagellin, directly bind to TLRs (Pandey et al., 2014). In insects, PGRPs detect PAMPs, leading to the activation of Toll or Imd signalling pathways to produce antimicrobial peptides (AMPs) (Lemaitre and Hoffmann, 2007; Lemaitre et al., 1996; Lindsay and Wasserman, 2014) (Fig. S1). To identify whether pathogen infection is a trigger for regeneration, we cloned partial cDNA fragments of Gryllus homologous genes of PGRP-SA and PGRP-SD, which recognise gram-positive bacteria, and PGRP-LC and imd, which recognise gram-negative bacteria, and performed RNAi to observe phenotypes of regenerating legs. The Gryllus genome encodes a single gene for each of the four genes. Determination of leg regeneration phenotypes after PGRP-SA^RNAi, PGRP-SD^RNAi, PGRP-LC^RNAi and imd^RNAi revealed that all crickets regenerated legs normally, in a similar manner to DsRed^RNAi crickets (Fig. S9A). This finding indicates that infectious microbes do not play a major role in leg regeneration. To further substantiate this, we examined the gene expression change of the Gryllus homologue of defensin, which encodes an evolutionarily conserved AMP (Fig. S10). qPCR analysis revealed that the expression level of Gryllus defensin was decreased at 3, 24 and 48 hpa in RLs (Fig. S10C), suggesting that PAMP-mediated activation of Toll signalling is suppressed during regeneration.

Toll molecules recognise PAMPs and DAMPs, mediated by clip-SP Persephone in Drosophila (Ming et al., 2014). We could not find a persephone (psh) homologue in the Gryllus genome (Fig. S11). DAMPs are released from cells in response to injury or cell death, including apoptosis and necrosis. We focused on the function of the scavenger receptor CD36 homologue Crq, which recognises the apoptotic body and is required for phagocytosis of plasmatocytes in Drosophila. The Gryllus crq homologue was expressed in plasmatocytes (Fig. 7B). During leg regeneration, 16.7% and 45.8% of crq^RNAi crickets had class 1 and class 2 phenotype, respectively (Fig. S9B-D). These results suggest that recognition and engulfment of apoptotic bodies released from injured cells by Crq likely triggers leg regeneration in crickets.

Toll/TLR signalling is evolutionarily conserved from insects to humans for the recognition of pathogens including viruses, gram-positive bacteria and fungi (Lindsay and Wasserman, 2014; Pandey et al., 2014). Toll/TLRs induce gene expression of cytokines and antibacterial components via MyD88 and NF-κB (Lemaitre and Hoffmann, 2007), although Drosophila Toll-2 does not require MyD88 for cell proliferation (Li et al., 2020). In insect immunity, activation of Toll signalling is different from that of vertebrates; infectious microbes are detected by PGRPs that activate clip-SPs to catalyse pro-Spz to Spz, which binds to a Toll receptor to activate signalling (Fig. S1) (Krautz et al., 2014; Lindsay and Wasserman, 2014). We previously reported that expression of Spz, Toll receptors and NF-κB (Relish, dl) genes is upregulated in RLs at 24 hpa (Bando et al., 2013). In this study, RNA-seq data showed that the expression of three Gryllus Toll genes (Toll2-2, Toll2-5, Toll8) and Rel were upregulated as early as at 3 hpa in RLs. Upregulation of Toll2-1, Toll2-2 and Toll2-5 genes was confirmed by qPCR during regeneration, suggesting that Toll signalling is involved in the cricket regeneration process from the early phases.

Imd signalling, another pathway involved in insect immunity (Fig. S1), detects infections by gram-negative bacteria. PGRP-LC and Imd together induce the expression of antibacterial components through the NF-κB transcription factor Rel (Krautz et al., 2014; Lemaitre and Hoffmann, 2007). Although expression of Rel was upregulated in 3 hpa RLs (Fig. S2B), RNAi of PGRP-LC, imd and Rel did not show obvious defects in leg regeneration (Fig. 3C,D, Fig. S9A), suggesting that Imd signalling is not essential for leg regeneration in G. bimaculatus.

Our previous study showed that JAK/STAT signalling is involved in leg regeneration (Bando et al., 2013). Ligands for JAK/STAT signalling are interleukins in vertebrates (Morris et al., 2018). In Xenopus tadpole tail regeneration, interleukin-11 induces undifferentiation state (Tsujioka et al., 2017). In Drosophila melanogaster, ligands of JAK/STAT signalling are Upds, which act as cytokines (Arbouzova and Zeidler, 2006) and are required for the regulation of stem cell proliferation during intestinal regeneration (Kux and Pitsouli, 2014). In the present study, the expression of upd was decreased in plasmatocyte-depleted RLs. Lost parts of legs of upd^RNAi crickets were not fully regenerated, similar to observations after RNAi of JAK/STAT signalling genes (Bando et al., 2013). In a similar manner to the production of interleukins by macrophages to activate other immune cells in vertebrates (Schett et al., 2016), plasmatocytes release insect cytokines to induce the production of antibacterial components in Drosophila (Lemaitre and Hoffmann, 2007; Ramond et al., 2020). Thus, it is likely that Gryllus plasmatocytes produce Upd to activate JAK/STAT signalling in leg regeneration.

We cloned 11 Toll genes from Gryllus, not all of which were directly homologous to those of the nine Toll genes of Drosophila (Hillyer, 2016). The Gryllus genome has five Toll2 paralogues (Toll2-1–5) that are not present in Drosophila, and two Toll6 paralogues (Toll6-1–2). Gryllus has no genes homologous to Drosophila Toll3, Toll4 and Toll5. Drosophila Toll (McIlroy et al., 2013; McLaughlin et al., 2016; Ward et al., 2015) is important for defence against pathogens (Lemaitre and Hoffmann, 2007) and dorsoventral patterning during embryogenesis (Moussian and Roth, 2005). Toll-2, -6 and -8 are involved in the anteroposterior patterning of the fly (Paré et al., 2014). Drosophila Toll-2 regulates cell proliferation and planar cell polarity (Li et al., 2020; Tamada et al., 2021) and Toll-6 promotes neuronal cell shape, survival and interactions (McIlroy et al., 2013; McLaughlin et al., 2016; Ward et al., 2015). In Gryllus, Toll6-1, Toll6-2, Toll7 and Toll8 are expressed during embryogenesis (Benton et al., 2016). Their functions related to immunity and dorsoventral patterning are unknown. This study has clarified that diversified Gryllus Toll2 subfamily members play crucial roles in the early phase of leg regeneration.

Expression of Toll2-1, Toll2-2 and Toll2-5 was upregulated during leg regeneration, in which >70% of Toll2-2^RNAi crickets showed regeneration defects. The proportions of S- and M-phase proliferating cells were decreased in Toll2-2^RNAi RLs, indicating that Toll2-2 signalling is required for cell proliferation during leg regeneration. In Toll2-2^RNAi RLs, accumulation of plasmatocytes was reduced and expression of upd was downregulated. Thus, Toll2-2 in plasmatocytes may induce upd expression, which activates the JAK/STAT signalling pathway (Bando et al., 2013) and cyclins, leading to cell proliferation in regenerating legs.

During early leg regeneration processes, expression of Toll2-4, Toll6-1, Toll6-2, Toll7, Toll8 and Toll9 was decreased (Fig. 2B). However, Toll2-4^RNAi, Toll6-2^RNAi, Toll7^RNAi and Toll9^RNAi crickets showed defective or impaired regeneration phenotypes (Fig. 3C). Toll2-4^RNAi, Toll6-2^RNAi, Toll7^RNAi and Toll9^RNAi decreased target gene transcripts by >40%, verifying the RNAi efficiency (Fig. S4A). Our preliminary study showed that RNAi of cricket nymphs continued to suppress target gene expression for 2 weeks. Therefore, it is possible that the continuous suppression of gene expression caused defective or impaired regeneration phenotypes in Toll2-4^RNAi, Toll6-2^RNAi, Toll7^RNAi and Toll9^RNAi crickets.

Crosstalk between Toll signalling and Hippo signalling, and between Hippo signalling and JAK/STAT signalling may also regulate tissue growth (Liu et al., 2016; Jiang et al., 2016). Specifically, the Hippo signalling component Warts, together with protocadherins Fat and Dachsous, suppresses blastema cell proliferation, whereas Yorkie promotes proliferation during cricket leg regeneration (Bando et al., 2009). Therefore, Toll signalling may interfere with Hippo signalling to regulate blastema cell proliferation in crickets, although Drosophila Toll-2 cooperatively promotes cell proliferation with Yki (Li et al., 2020).

Insect haemocytes play a major role in host defence (Lavine and Strand, 2002; Ribeiro and Brehélin, 2006), as insects have an open blood-vascular system and lack oxygen-carrying erythrocytes. Insect haemocytes vary depending on the species. For example, the cricket G. bimaculatus has plasmatocytes, granulocytes, and three or four other haemocytes (Cho and Cho, 2019; Sokolova et al., 2000). Plasmatocytes in several insect species engulf pathogens and produce inflammatory cytokines and antipathogenic components (Evans et al., 2003; Lavine and Strand, 2002). In the present study, plasmatocyte-depleted crickets did not regenerate missing leg parts. Given that cytokines promote regeneration in axolotl and zebrafish (Godwin et al., 2013; Petrie et al., 2014), cricket plasmatocytes could promote tissue regeneration, possibly by these evolutionarily conserved molecular mechanisms. Although we had no direct evidence for localisation of the Toll2-1, Toll2-2 and Toll2-5 genes in plasmatocytes, their reduced expression levels in plasmatocyte-depleted haemocytes strongly support this idea. Likewise, the finding that cricket plasmatocytes express the cytokine genes upd, spz and spz2 is reminiscent of the secretion of interleukins from axolotl macrophages during regeneration (Godwin et al., 2013).

Mammalian TLRs are pattern recognition receptors, but Drosophila Tolls are not, as they bind endogenous ligands instead of pathogens (Krautz et al., 2014; Lindsay and Wasserman, 2014; Pandey et al., 2014). When pathogens are recognised by PGRPs, the PGRPs catalyse the maturation of pro-Spz to Spz. Spz subsequently binds to Toll proteins to activate Toll signalling. In the present study, crickets subjected to RNAi of the pathogen recognition protein-coding genes PGRP-SA, PGRP-SD, PGRP-LC or imd showed normal regeneration (Fig. S9A). Expression of the antimicrobial peptide gene defensin was not upregulated during regeneration (Fig. S10C), although the upstream region of the defensin gene contains many NF-κB-binding sites (Fig. S10D). This suggests that pathogen infection is not required for leg regeneration. Thus, non-pathogenic molecules may activate Gryllus Toll signalling during leg regeneration. Candidate molecules include DAMPs (Krautz et al., 2014), which are endogenous proteins that are likely released from injured cells, such as necrotic or apoptotic cells, and are recognised by specific receptors, such as RAGE, TREM1 and TLRs, in vertebrates (Piccinini and Midwood, 2010; Wang et al., 2015). The evolutionarily conserved proteins HMGB1, S100, HSP, histone, actin, DNA and RNA are well-known DAMPs (Piccinini and Midwood, 2010). In the mammalian kidney, DAMPs released from dying cells cause inflammatory responses and acute injury. Regeneration is accelerated by the TLR-mediated release of interleukins from macrophages or dendritic cells (Kulkarni et al., 2014). In the liver, denatured DNA from dying hepatocytes stimulates TLRs to mediate hepatocyte stem cell differentiation (Seki et al., 2011). RAGE and TREM1 are not conserved among insects; hence, Toll signalling is able to respond to DAMPs in insects. In Drosophila, the clip-SP Psh is involved in DAMP recognition and catalyses Pro-Spz to Spz to activate Tolls (Ming et al., 2014). The psh homologue is not present in the Gryllus genome (Ylla et al., 2021), but a snake homologue and three grass paralogues, which encode other clip-SPs, are (Fig. S11). Some of these genes may replace a particular role of psh.

Crq, which is a scavenger receptor CD36 homologue, is expressed in the plasmatocytes and is required for engulfment of apoptotic cells in Drosophila (Franc et al., 1996, 1999; Guillou et al., 2016). In mice, CD36 mediates phagocytosis by cooperating with TLR (Erdman et al., 2009). In the present study, crq^RNAi crickets showed defective and impaired regeneration phenotypes, similar to those found in Toll2-2^RNAi crickets. As mentioned above, RNAi of PGRP genes and imd resulted in normal regeneration, implying that the defective and impaired regeneration phenotypes caused by crq^RNAi and Toll2-2^RNAi could occur as a result of defects in phagocytosis of cell debris after injury. We propose that Toll2-2 and Crq cooperatively recognise damage-associated molecules near the wound region (Fig. S12).

Our cricket leg amputation experiments were conducted without artificial pathogen inoculation. In the early phase of leg regeneration, when blastema cells form, activated plasmatocytes migrate to the injury site and facilitate the proliferation of blastema cells. These plasmatocytes are most likely polarised toward enhanced phagocytic function by regulating Toll-related and JAK/STAT signalling (Fig. S1). The extent of plasmatocyte polarisation would be influenced by surrounding tissue damage and pathogenic status. Recent single-cell studies of plasmatocytes (Cattenoz et al., 2020) suggest that crq-expressing cells and Toll-2-2-expressing cells would not be identical subpopulations in the plasmatocytes during early cricket leg regeneration, which will require further study.

In conclusion, this study provides new insights into the function of Toll-related signalling for leg regeneration via plasmatocytes, cooperatively with JAK/STAT signalling. Recognition of apoptotic cells via the scavenger receptor Crq on plasmatocytes also promotes leg regeneration.

All nymphal and adult two-spotted crickets (G. bimaculatus) were reared under standard conditions (Mito and Noji, 2008). Third instar nymphs were used for RNAi and liposome injection to analyse the regeneration processes (Nakamura et al., 2008a).

The distal one-third tibial regions of regenerating legs at 0 hpa (NLs) and 3 hpa (RLs) were separately collected from a few hundred crickets. Total RNA was extracted using ISOGEN II (311-07361, Nippon Gene). Poly(A)+ RNAs were purified using a MicroPoly(A)Purist kit (AM1919, Thermo Fisher Scientific). The cDNA libraries constructed from poly(A)+ RNAs were sequenced using a GS FLX Titanium next-generation sequencer (454 Life Sciences). To construct the assembled transcripts, the 595,425 reads of NL and 519,961 reads of RL samples (14,368,694 bp in total) that were sequenced were assembled into 20,317 contigs using GS De Novo Assembler v2.8 software. The average length of the assembled contigs was 2252 bp. To estimate the normalized expression levels of each transcript, each read obtained from the RL and NL samples was mapped to the contigs to calculate the RPKM value using GS Reference Mapper v2.8 software. Expression changes of assembled transcripts were calculated by dividing the RPKM values of RLs by corresponding RPKM value of NLs, which are listed in Table S1. Assembled transcripts that were upregulated more than twofold in RLs compared with NLs were annotated with the BLASTX program against the NCBI non-redundant protein sequence database, with an E-value cut-off of 0.001 (Table S2). Functional annotation of BLASTed transcripts was performed using Blast2GO software (https://www.blast2go.com/).

Gryllus homologues were cloned by PCR with Ex-taq or LA-taq with GC buffer (RR006A or RR02AG, TaKaRa Bio). Primers used are listed in Table S3. Template cDNAs were reverse transcribed using the SuperScript III reverse transcription kit with random primers (18080051, Thermo Fisher Scientific) from total RNA extracted from the regenerating legs of cricket nymphs at the third instar. The isolated Gryllus cDNA fragments were used as templates for the synthesis of dsRNAs.

dsRNAs were synthesised using the MEGAScript T7 Kit (AMB13345, Thermo Fisher Scientific) and adjusted to 20 μM. Third instar nymphs were anaesthetised on ice before RNAi. After injection of 207 nl of dsRNA into the abdomen of the third instar nymphs with an auto-nanoliter injector (Nanoject II, #3-000-204, Drummond Scientific Company), their tibiae were amputated between the second and third spines. Wound regions are usually covered with scab within a day in third instar. The third instar nymphs moult to fourth instar within 4 days. In the fourth instar, newly formed cuticles cover the wound region instead of the scab. The fourth instar nymphs moult to fifth instar within 5 days and the lost leg tissues are reconstructed in miniature in the fifth instar. After RNAi and leg amputation, we observed RNAi phenotypes of regenerating legs on the 10th day, which corresponds to the late fifth instar. As a negative control for RNAi experiments, we injected dsRNA for the DsRed2 exogenous gene. For dual RNAi experiments, a mixture of dsRNAs for the two target genes was used. In the mixture, the final concentration of each dsRNA was adjusted to 10 μM. Statistical differences in RNAi phenotypes were analysed by Fisher's exact test. For RNAi against Toll and Toll2-2, we generated two non-overlapping dsRNAs corresponding to either the 5′ or 3′ portion of the coding region (Fig. S3B, denoted by double-headed arrows). The RNAi experiments performed with these sequences were designated Toll(5′)^RNAi, Toll(3′)^RNAi, Toll2-2(5′)^RNAi and Toll2-2(3′)^RNAi, respectively.

Regenerating tibiae of control, RNAi-treated or liposome-injected nymphs were pooled into single tubes, and total RNA was extracted using the RNAqueous-Micro Kit (AM1931, Thermo Fisher Scientific). Each pooled RNA sample was divided into two samples and reverse transcribed to prepare cDNA. Each cDNA was used for qPCR performed with the FastStart Essential DNA Green Master kit (06402712001, Roche) and the LightCycler Nano System (Roche). The relative proportions of the transcripts were calculated using Gryllus β-actin as an internal control. Relative gene expression levels are shown as the mean±s.d. Statistical differences were analysed by an unpaired, two-tailed Student's t-test between two samples or Tukey's test for more than three samples and are shown by asterisks (*P<0.05, **P<0.01, ***P<0.001). The qPCR experiments were repeated twice for confirmation. The primers used are listed in Table S4.

Proliferating cells in the S phase were detected using the Click-iT EdU Alexa Fluor 488 Imaging Kit (C10337, Thermo Fisher Scientific). EdU was injected into the abdomen of third instar cricket nymphs at 44 hpa, and the RLs were fixed at 48 hpa (4 h after EdU injection) with 2% paraformaldehyde (PFA) in PBS containing 0.05% Tween-20 (PBT) overnight at 4°C. EdU-incorporated cells were detected according to the manufacturer's instructions. Hoechst 33342 (H3570, Thermo Fisher Scientific) was used for nuclear staining. Proliferating cells in the M phase were detected by immunostaining of phospho-histone H3S10. Regenerating legs at 48 hpa were fixed with 2% PFA and cuticles were removed, then regenerating tibiae were washed with PBT and blocked with 1% normal goat serum (NGS) in PBT for 1 h. Blocked samples were incubated with primary antibody [rabbit polyclonal anti-phospho-histone H3 (Ser10) antibody; 06-570, Millipore] at 1:500 in 1% NGS in PBT overnight at 4°C. The samples were washed with PBT and blocked with 1% NGS in PBT. The samples were incubated with secondary antibody (Alexa Fluor 488-conjugated anti-rabbit IgG antibody; A-11008, Molecular Probes) at 1:750 in 1% NGS in PBT for 3 h at 25°C. Finally, the samples were washed with PBT and incubated with a 1:1000 dilution of Hoechst 33342 in PBT for 15 min.

For plasmatocyte depletion in crickets, 207 nl of Clo-lipo 100 (6.4 mg/ml clodronate) and Clo-lipo 300 (9.8 mg/ml clodronate) (160-0432-1 and 160-0430-1, Katayama Chemical Industries) was injected into the abdomen of third instar nymphs. PBS-encapsulated liposomes were used as controls. For plasmatocyte detection in the haemolymph, 207 nl India ink or 207 nl BODIPY-lipo (130 μg/ml BODIPY-cholesterol, Katayama Chemical Industries) were injected into the abdomen of third instar nymphs. One day after the injection, an anticoagulant solution (20 mM EDTA in PBS) was injected into the abdomen of the nymphs and haemolymphs were sucked from the nymphs. Haemolymphs were fixed with 2% PFA in an anticoagulant solution for 30 min at 25°C. Fixed haemocytes were washed with anticoagulant solution and then stained with a 1:2000 dilution of Hoechst 33342 for 30 min at 25°C. The haemocytes were washed with an anticoagulant solution and stained with a 1:40 dilution of Rhodamine phalloidin (R415, Thermo Fisher Scientific) for 30 min at 25°C. For plasmatocyte detection in the RLs, RLs of BODIPY-lipo-injected crickets at 48 hpa were fixed with 2% PFA in PBT and the cuticles were removed. Stained haemocytes and cuticle-removed RLs were observed with a fluorescent microscope DM5000 B (Leica Microsystems), and with a confocal laser scanning microscope LSM780 (Carl Zeiss) at the Central Research Laboratory, Okayama University Medical School, Okayama, Japan.

Scanning electron microscopy images of RLs at the fifth instar were obtained. RLs of control and Toll2-2^RNAi were fixed in 4% PFA and 4% glutaraldehyde in PBS for 15 min at room temperature. Fixed legs were washed with PBS, then dehydrated through an alcohol series (20%, 40% and 60% ethanol in PBS, 80% ethanol in water, 45% ethanol and 45% tert-butyl alcohol in water) for 30 min and in 100% tert-butyl alcohol for 1 h. After dehydration, the fixed legs were substituted with 100% tert-butyl alcohol, frozen at 4°C, and freeze-dried. Dried legs were coated with osmium with osmium coater (HPC-1S; VACUUM DEVICE). Images were captured using a model S-4800 field emission scanning electron microscope (Hitachi) at the Central Research Laboratory, Okayama University Medical School, Okayama, Japan.

We are grateful to Dr Itsuro Sugimura (Hokkaido System Science Co., Ltd.) for assistance with data analysis; Dr Takayuki Otani (Katayama Chemical Industries Co., Ltd.) for preparing PBS-lipo, Clo-lipo and BODIPY-lipo; and Nobuaki Fujimori and Taiki Morino for their technical assistance.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199916