Observations on neurophysiological pattern and behavioural traits as death-feigning mechanism in Eucryptorrhynchus scrobiculatus (Coleoptera: Curculionidae)

As adaptations to surrounding environmental stimuli, insects have receptors capable of receiving a variety of information while regulated by effectors and endogenous hormones to enhance stress tolerance and defense capability (Cook et al., 2017; Wu et al., 2020). Many insects exhibit adaptive defence behaviour in response to inimical stimuli, such as death feigning (also known as tonic immobility or thanatosis) (Ruxton, 2006; Ruxton et al., 2018), increasing the survival probability (Ohno and Miyatake, 2007). When insects are disturbed by stimuli, they remain immobile, with a restrained posture, and behave as if dead; arousal occurs anywhere between a few seconds or hours later (Humphreys and Ruxton, 2018). The death-feigning duration is an important index for evaluating death-feigning intensity under different environmental conditions (Li and Wen, 2021). Previous papers have reported that death-feigning duration decreased with time since feeding and emergence (e.g. tick; Oyen et al., 2021), but specific sounds (alarm calls) or low temperatures increased the duration (e.g. Gallus gallus, Callosobruchus chinensis; Miyatake et al., 2008). This evidence suggests that variations of death-feigning duration have significant implications for the adaptation of defense capability. Nowadays, there is agreement that the shifts in death-feigning behaviour traits, mainly duration, are involved with physiology and neurobiology (Li and Wen, 2021, 2022; Matsumura and Miyatake, 2022; Nakayama et al., 2012), affecting death-feigning mechanisms. However, no empirical study has yet identified the death-feigning signal input of receptors, regulation of the brain, feedback of endogenous hormones or expression of effectors to systematically define the mechanism of death-feigning behaviour.

It was previously hypothesized that the neurophysiological pattern of the brain explains the death-feigning behaviour, as it is mediated by the neural circuits in the brain that induce a series of complex physiological changes (Nishino, 2004; Rogers and Simpson, 2014), evidence in insects has been controversial so far (Li, H. J. et al., 2021). One important reason is that the difficulties of neurophysiological research in tiny insects are often underestimated, so evidence of the death-feigning mechanism may not be precise. Preliminary studies have shown that the death-feigning behaviour is related to the distribution and density of c-Fos labelling, which reflects neural activity in specific regions of the pigeon brain (Melleu et al., 2017; Humphreys and Ruxton, 2018). Moreover, the head of Gryllus bimaculatus stimulated arousal from the death-feigning state when facing cold stress (Rillich et al., 2019). But more detailed studies, such as on the neural structure of the stimulus points and the signal input of different parts of the body, would help to verify the neurophysiological pattern of death feigning.

Evidence for the neurochemicals in the death-feigning brain has gained further attention. The biogenic amine in stimulated animals is involved in behaviour function and traits (e.g. death-feigning duration) (Rogers and Simpson, 2014; Wu et al., 2020), and it is often believed that this would allow the defence behaviour to be more adaptive (Claßen and Scholz, 2018). In Tribolium beetles, dopamine (DA) and octopamine are key amino acids that regulate death-feigning duration (Miyatake et al., 2008; Nakayama et al., 2012), and the expression of DA-related genes in the tyrosine metabolism pathway of Tribolium castaneum differs in populations selected for the long and short duration of death feigning (Uchiyama et al., 2019). Importantly, Rogers and Simpson (2014) demonstrated that death-feigning duration was affected by DA receptor antagonists and agonists in both birds and mammals, but did not fully explain the behavioural mechanism. Given the effects of biogenic amines on death feigning, we hypothesized that the amino acids in the brain would be important factors in the death-feigning mechanism. Systematic evidence is nonetheless lacking.

Here, we investigated the importance of the neurophysiological pattern of death feigning through a study of insect neurological structure, neurochemical expression and behavioural traits (the variation of death-feigning duration) in Eucryptorrhynchus scrobiculatus, the tree-of-heaven root weevil, which causes destruction to the widespread temperate deciduous tree Ailanthus altissima (commonly called the tree of heaven) (Herrick et al., 2012). Eucryptorrhynchus scrobiculatus exhibits stronger death-feigning behaviour as a forestry pest to enhance defensive adaptability in the wild (Li and Wen, 2021). The mechanical stimulus (e.g. physical contact) was the most effective external signal to induce death feigning, and the stimulus-sensing region was the metaventrite (the metathorax between the middle and hind legs) in E. scrobiculatus adults (Li and Wen, 2021, 2022). Although the importance of DA in death-feigning regulation has been highlighted (Matsumura et al., 2016; Uchiyama et al., 2019), its contribution to the systematic neurophysiological pattern of stimulus receptors, neurophysiology regulation and behaviour expression has never been evaluated. Incomplete taxonomic knowledge of death feigning, particularly for insects, is the main barrier to comprehending the mechanism of death feigning. To date, beetles are often used as study species because they exhibit large populations and a strong ability to feign death, thus allowing investigators to fill in the knowledge gaps.

Our primary goal in the present study was to test the expression of neurochemicals before and after death feigning in E. scrobiculatus to evaluate the potential mechanism associated with behavioural modifications (the variations of death-feigning duration and frequency). Thus, we studied the key regulatory body parts of E. scrobiculatus and analyzed the effects of neurotransmitters and metabolic pathways. Neuroanatomy and immunostaining techniques were used to analyze the distribution of neurons related to death feigning to clarify the regulation of the nerve centre. We utilized RNA-seq, which provided an opportunity to assess the physiological response related to the functional differences of receptor genes by the 42-target neurotransmitters on UHPLC−MS/MS and qRT-PCR. Our study will enhance the mechanistic understanding of death feigning, and allow us to predict the behavioural impacts of neurotransmitters such as biogenic amines across more species.

All experimental Eucryptorrhynchus scrobiculatus (Motschulsky) (Coleoptera: Curculionidae) adults were collected from the sample plot around the village of Haizi, Ningxia, China (106°31.301′E, 38°51.241′N), and reared at the Lingwu Forest Quarantine Station. Weevils were kept separately in stock culture buckets (15×10 cm, diameter×height) with fresh branches from the tree of heaven. Laboratory conditions were maintained at 25°C and 60% relative humidity.

The metaventrite (the metathorax between the middle and hind legs) has been shown to be the most effective stimulus-sensing region to induce death feigning (Li et al., 2019; Li and Wen, 2022). Thus, the E. scrobiculatus metaventrite was processed for scanning electron microscopy. In brief, the metaventrite was removed and washed with ultrasonic cleaner (PS-60AL, 10 min). The samples were dehydrated through a graded series of ethanol. Subsequently, the samples were sputter-coated with gold by an E-1010 sputter ion instrument (Hitachi, Tokyo, Japan) before examination with a Hitachi S-3400N scanning electron microscope (Yang et al., 2017).

We refer to the methods of Chen et al. (2018). In brief, the experimental adults were stored in cell tissue fixative for more than 48 h and dissected under a stereomicroscope (M165C, Leica). We used scissors to separate the pleurotergite of E. scrobiculatus adults, and carefully removed the excess muscle tissue, thus exposing the unwound central nervous system (CNS), which was then gently picked out with tweezers and placed in fixative for preservation and photography (M205FA, Leica).

Eucryptorrhynchus scrobiculatus individuals were randomly assigned to one of two treatment groups. In the death-feigning group, E. scrobiculatus was in the death-feigning state (immobile, restrained state: antennae and appendages are strongly folded and pressed close to the abdomen of weevils) for more than 30 s (3 replicates, 15 weevils per replicate); to induce death-feigning behaviour, a light stimulus was applied to the metaventrite using forceps, and the duration of death feigning was recorded. The duration was the time from when movement ceased until the first visible movement (Li et al., 2019). In the control group, E. scrobiculatus was in an arousal state (contrary to the death-feigning state, the antennae and appendages of aroused weevils show locomotor activity and walking) (3 replicates, 15 weevils per replicate); and the weevils in the same treatment were snap-frozen in liquid nitrogen and stored at −80°C until dissection. For the determination of neurotransmitter expression, we used the targeted metabolomics method of 42 kinds of neurotransmitters (Wong et al., 2016; Shanghai Biotree Biomedical Technology Co.).

First, the brain was removed with dissecting scissors in 0.1 mol l⁻¹ ice phosphate buffer (pH balanced). The brain samples of 15 weevils were transferred to a 1.5 ml EP tube and homogenized with 200 μl acetonitrile in 0.1% formic acid (pre-cooled at −20°C) and sonicated for 5 min at 4°C. This step was repeated three times. The samples were centrifuged at 79.92 g at 4°C for 10 min. Then, 80 μl supernatants were collected and mixed with 40 μl of 100 mmol l⁻¹ Na₂CO₃ solution and 40 μl of 2% benzoyl chloride acetonitrile solution and incubated at room temperature for 30 min. After the spike of 1.6 μl of stable-isotope labelled standards, the samples were centrifuged again at 79.92 g for 5 min at 4°C. Finally, we collected 20 μl supernatants and mixed them with 10 μl of acetonitrile in 0.1% formic acid to measure the neurotransmitter concentrations on a UHPLC (SCIEX ExionLC).

The 42-target neurotransmitter method was carried out using an ExionLC System, equipped with a Waters ACQUITY UPLC HSS T3 (100×2.1 mm, 1.8 μm). The standard product information of 42 neurotransmitters for UHPLC–MS/MS analysis is shown in Table S1, and the ion chromatogram is shown in Figs S1 and S2.

Eucryptorrhynchus scrobiculatus individuals were randomly assigned to one of two treatment groups (death-feigning and arousal) as detailed above. The brains of 12 weevils from the two groups (death-feigning and arousal, 3 replicates, 2 weevils per replicate) were dissected on ice, and the RNA was extracted immediately using the RNApure Total RNA Kit (Aidlab, Beijing, China). For the determination of the transcriptional levels of three targeted genes (dar2, oar1 and oab2) involved in the DA signalling pathway, the brains of weevils were preserved in RNALater at −80°C. The cDNA was synthesized using the TRUEscript 1st Strand cDNA synthesis kit (Aidlab, Beijing, China). Primer3Plus online software (https://primer3.ut.ee/) was employed to design the gene-specific primers. Primer sequences are shown in Table S2. The reference gene UBC was used for the normalization of the data of the polymerase chain reaction (PCR) analysis, and relative expression was quantified using the 2^−ΔΔCt method (information on the primers and details on quantitative PCR are in Tables S2 and S3) (Gao et al., 2020). The RT-qPCR reactions were performed on a CFX96 Real-Time PCR Detection System with TB Green Premix Ex Taq II (Takara, Beijing, China). Cycling parameters were 95°C for 30 s, followed by 40 cycles of 95°C for 5 s and 60°C for 30 s.

Each E. scrobiculatus weevil was randomly selected from culture buckets and transferred to a culture container (2×3.5 cm, diameter×height) in which the behaviour tests were carried out. Each weevil was tested only once and then placed in a different culture box.

In the present study, various body parts of E. scrobiculatus were excised to verify their response to physical contact in death-feigning weevils. To prevent excessive nerve damage and death, each weevil was anaesthetized on ice, and the different body parts were quickly excised with scissors. This experiment had seven treatments, detailed below. There are six points of excision in the treatment groups (10 weevils per treatment). As shown in Fig. 1F, the points were located between the head and the protergum (H), the base of the elytra (E), the coxa of foreleg (FL), the coxa of midleg (ML), the coxa of hindleg (HL), and the joint between the metaventrite and abdomen (A). Ten weevils in the control group (CK) were not excised. Each experimental E. scrobiculatus was placed in a stock culture container (2×3.5 cm, diameter×height) with no feeding to observe the death-feigning behaviour as in the above method (see ‘Sample collection’, above). In brief, death feigning was induced by a light stimulus that was applied to the metaventrite using forceps, and the duration and frequency were recorded to comprehensively evaluate the intensity of death feigning (Li et al., 2019; Li and Wen, 2022).

To compare the impacts of biogenic amines on death feigning, we also established five treatments (10 weevils per treatment, three replicates, 150 total): four drug treatments and physiological saline, which served as vehicle control. We selected the DA antagonists chlorprothixene hydrochloride (Chl) and chlorpromazine (CH) (Wu, 2013), and the agonists pramipexole dihydrochloride (Pra) and clonidine (CL) (Huang, 2019) (Sigma-Aldrich, Deisenhofen, Germany) to evaluate the roles of neurochemicals in the physiological regulation of death-feigning behaviour. In the experiment, in each case, 0.01 g of the drug was dissolved in 10 ml of physiological saline and configured into a solution of concentration of 1 μg ml⁻¹, respectively. Weevils were anaesthetized on ice for 5 min, and 1 μl of the prepared drug solution was taken up with a microsyringe and injected through the front side of the first segment of the abdomen. The control group received vehicle only. After 60 min, durations and frequencies of death-feigning weevils were recorded as the above method.

All data are presented as means±s.e.m. To test for the effects of excision and drugs on the duration of death feigning, the statistical significance was analysed using one-way ANOVA and Tukey's multiple comparison tests for differences among treatments (SPSS Statistics V 17.0).

The relative expression levels of dopamine (dar2) and octopamine (oar1, oab2) were calculated using the 2^−ΔΔCt method and analyzed using GraphPad Prism 5.0 (GraphPad Software, La Jolla, CA, USA) with a one-way ANOVA.

The metabolic analysis of neurotransmitters is based on the public information database of metabolites and the Biotree DB database built by Shanghai Biotree Biomedical Technology Co. Metabolites were quantified, and the multiple reaction monitoring (MRM) mode of triple quadrupole high-resolution mass spectrometry was used to quantify metabolites. SIMCA software (v14.1, Switzerland) was used for principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA). Owing to the high similarity between the death-feigning and arousal treatments, the metabolites with fold change ≥1.5 or ≤0.7 and variable importance in projection (VIP) ≥1 were selected as differential metabolites.

The surface of the metaventrite is randomly distributed with many scales in the shape of sensilla brushes (178.16±1.16 μm; Fig. 1A), sensilla scopula (63.11±7.08 μm; Fig. 1A), cuticular pores (12.50±2.43 μm; Fig. 1B) and a small number of blunt protuberances (19.50±1.43 μm; Fig. 1C). But there were no obvious sensilla (Fig. 1D,E).

The ventral nerve cord clung to the E. scrobiculatus metaventrite. The prothoracic ganglion (ProTG) is located in the metathoracic venter between the foreleg and middle leg. The posterior ganglion (PG) and terminal abdominal ganglion (TAG) are located in the metathoracic venter between the middle leg and hind leg, connected by connective tissue (Con) (Fig. 2).

The PCA analysis results (R²=0.918) showed that the samples were separated, indicating that there was a significant difference between the two groups of samples, and the PCA model has a good fit. The PCA scores were all within the 95% confidence interval (Hotelling's T² ellipse). The values (R²X, R²Y, Q²) of the OPLS-DA model of death-feigning versus arousal were 0.602, 0.896 and 0.204, indicating that the OPLS-DA model was reliable.

The analysis of 42 neurotransmitters showed that the quantity and relative content of 16 neurotransmitters were upregulated, and that of 26 were downregulated (Table S1). Nine chemicals inhibited or promoted by death-feigning behaviour among the 42 neurotransmitters and metabolites in the brain determined in this study were grouped into four major signalling pathways: dopamine (DA), serotonin (5-HT), l-asparagine (Asn) and l-phenylalanine (Phe) (Table 1, Fig. 3).

For the DA signalling pathway, the concentrations of octopamine, vanillylmandelic acid, 3-hydroxytyramine hydrochloride, 3,4-dihydroxyphenylalanine and l-tyrosine were decreased in the death-feigning treatment (Fig. 3A); for the 5-HT signalling pathway, the concentrations of tryptamine and 5-HIAA were increased in the death-feigning treatment (Fig. 3B). Meanwhile, decreased levels of Asn and Phe were also found in the head in the death-feigning treatment (Fig. 3C,D).

Differential metabolites were annotated, and metabolic pathway enrichment analysis was performed in the KEGG metabolite database. The phenylalanine, tyrosine and tryptophan biosynthesis (tca00400: C0007, downregulation; C00078, upregulation; C00082, downregulation) (P=0.004, impact=1) showed significant enrichment (Table 1).

The quantitative real-time PCR (RT-qPCR) analyses showed that the relative expression of three receptor genes, dar2 (P=0.042, 0.09-fold), oar1 (P=0.013, 0.51-fold) and oab2 (P=0.005, 0.21-fold), was lower in the death-feigning treatment (Fig. 4A).

We examined seven treatments and found that death-feigning duration was significantly affected by different body parts (F=59.44, P<0.0001) (Figs 1F and 5). The values of duration (Fig. 5A) and frequency (Fig. 5B) of death feigning were zero in the H group. However, in the E, FL, ML, HL and A groups, the duration decreased significantly compared with the CK group (E: P<0.0001; FL: P<0.0001; ML: P<0.0001; HL: P=0.002; A: P<0.0001). No differences were recorded in death-feigning frequency in the other groups.

We examined the agonists and antagonists of dopamine and found that the two antagonists significantly regulated the death-feigning duration. As shown in Fig. 4B, compared with CK, Chl and CH significantly increased the death-feigning duration (P<0.0001, P<0.0001). When considering the regulation of agonists Cl and Pra, no significant differences were detected. No differences were recorded in the frequency of death feigning (Fig. 4C).

In the present study, we first performed analyses of the neurophysiological pattern of death-feigning relating to the neural structure of stimulus points, the response of different parts of the body, and the effects of neurotransmitters and metabolic pathways; death-feigning duration was significantly related to the behaviour modification. Importantly, we detected the expression of neurotransmitters in the aroused and death-feigning brain of E. scrobiculatus. The analysis of four major signalling pathways demonstrated that nine neurochemicals were involved in mediating the death-feigning behaviour, i.e. DA (5), 5-HT (2), Asn (1) and Phe (1).

Previous studies have been conducted on the confused flour beetle (Tribolium confusum) and the red flour beetle (T. castaneum), in which brain DA levels were lower in artificially selected strains (L strains) exhibiting a longer duration of death feigning (Miyatake et al., 2008; Nakayama et al., 2012). However, in the first investigation of octopamine and 5-hydroxytryptamine concentrations in the brains of T. castaneum, concentrations were not significantly different between long-distance walking and short-distance walking strains (Matsumura et al., 2016). Contrary to these studies, our results analysed the signalling pathways of death-feigning behaviour, as well as the expression of receptor genes of DA and octopamine, and further compared the effects of antagonists and agonists in behaviour tests.

For the DA signalling pathway, tyrosine and 3,4-dihydroxyphenylalanine are the two precursors for the biosynthesis of DA, and octopamine, homovanillic acid are the two major metabolites of DA (de Vita et al., 2021; Kahsai, et al., 2012). The decreased concentrations of brain DA, its precursors and metabolites (Fig. 3A) in the death-feigning versus the arousal treatment suggest that the DA pathway was markedly involved with the variation in death-feigning duration. Our RT-qPCR results further demonstrated that the gene expression of DA receptors (OARs: oar1, oab2; DARs: dar2) was downregulated. Octopamine and DA associated with the DA-mediated mechanisms inhibit the motor activity to promote the behaviour maintaining death feigning in E. scrobiculatus head. Moreover, behaviour tests demonstrated that the selected antagonists of DA, Chl and CH significantly increased the duration of death feigning, suggesting that DA depression is involved in the death-feigning physiological response. However, there was no significant difference in death-feigning duration between the control and the agonist groups. The pathological and pharmacological properties of DA receptor genes are indeed unclear (Berke, 2018; Cousineau et al., 2020); thus, it is hard to describe the function of receptor genes in death feigning. In the future, we would like to discover more information about related genes and investigate their regulatory role in death feigning.

For the 5-HT signalling pathway, two neurotransmitters were identified: tryptophan is involved in the precursors for biosynthesis and 5-HIAA is the metabolite of 5-HT. Previous papers reported that the activation of 5-HT receptors is associated with an increase in anxiety, which can promote alertness (e.g. Rillich et al., 2019). We first considered that 5-HT was associated with death-feigning behaviour because of the increased concentrations of 5-HT precursors and metabolites, demonstrating the increased 5-HT levels inhibit the motor activity of the weevil, thereby maintaining immobility and a restrained state. Previous work found that 5-HT receptors also indirectly influenced neuronal function through their modulation of neurotransmitter release of DA (Murphy et al., 2021) and regulated behaviour, including the feeding behaviour of Rhodnius prolixus (Orchard, 2006) and the aggressiion and locomotion of Drosophila melanogaster (Silva et al., 2014). Although the 5-HT receptor has been well studied, we are not clear about the pharmacology of the receptor genes in weevils, thus no RT-qPCR and behavioural experiments of antagonists and agonists of 5-HT on E. scrobiculatus death-feigning have been verified.

Furthermore, l-asparagine (Asn) and l-phenylalanine (Phe) were also identified as antagonists. The decreased concentrations of Asn and Phe indicated that the locomotion of E. scrobiculatus was inhibited to maintain death-feigning behaviour. Statistics from KEGG (Biotree DB database) enrichment also confirmed that the two metabolites were key amino acids involved in brain energy metabolism and the inhibitory effects of excitatory behaviour (Das et al., 2020; Li, X. et al., 2021). As there was very little data on the effects of Asn and Phe signalling pathways on death feigning, the results are presented here with very limited discussion. To verify the influence of these metabolites, further investigation is needed regarding neurotransmitters and the regulation of death feigning.

In the excision test, we referred to previous methods (Meng and Liang, 2010; Nishino and Sakai, 1996) and found that the head significantly contributed to the regulation of death-feigning duration. The separation of the legs, elytra or abdomen decreased the death-feigning duration, which showed that the thoracic and abdominal ganglia were involved in the regulation of motor neurons and thus affected death feigning to maintain an unusual, restrained posture. Nishino (2004) found similar results, showing that the cricket head was the main regulatory body part, indicating that the behavioural mechanism of death feigning might be related to the brain nerve. However, in Laccotrephes robustus, death-feigning behaviour was not induced by mechanical stimulation when the head and thorax were removed (Meng and Liang, 2010), suggesting that the nerve centre, including the brain and thorax ganglia, were involved in the regulation mechanism of death feigning. The function of the ventral nerve cord in the death-feigning mechanism still requires more study.

Previous studies have shown that the abdomen of beetles including Callosobruchus maculatus, C. chinensis, T. castaneum and Leptinotarsa decemlineata are sensitive to contact stimulus, and the whole bodies of Timema cristinae and G. bimaculatus were sensitive (Li and Wen, 2021, 2022; Matsumura and Miyatake, 2022), indicating that there is a contact-sensitive area. We defined this area as an ‘effective stimulus point’ and predicted that insects exhibiting death-feigning behaviour may have adaptive effective stimulus points, which would help researchers study the death-feigning mechanism more simply and intuitively. The microstructure anatomy found that many cuticular pores and few blunt protuberances were distributed in the effective stimulation points, and no sensilla. Concerning the ventral nerve cord that clings to the E. scrobiculatus metaventrite, we supposed that the distribution of thoracic ganglia may be related to the effective stimulus point of death-feigning behaviour.

In conclusion, we verifed the key regulatory body parts involved in the regulation of the central nervous system in death feigning duration, meanwhile examining the variation of differential metabolite concentrations in death-feigning versus arousal treatments and analysing the metabolic regulatory network associated with death feigning, demonstrating the behavioural regulation mechanism of death feigning. We described the neurophysiological pattern of death feigning and illustrated the changing process from the arousal posture to the death-feigning posture (Fig. 6). In brief, when the effective stimulus point (metaventrite) of arousal-group E. scrobiculatus received mechanical stimulation, the thoracoabdominal ganglion transmitted signals into the brain through the ventral nerve cord. The brain then regulated the DA and 5-HT metabolic pathways, reducing the expression of DA (dar2) and octopamine (oar1, oab2) receptor genes, finally inducing death feigning. The death-feigning duration significantly increased with the expression of DA antagonists, suggesting that the death-feigning and arousal states of E. scrobiculatus show a trade-off between fleeing and protection. Additionally, it has been reported that death feigning has practical value (Humphreys and Ruxton, 2018). For example, the application of the live transportation of large death-feigning animals reduced unnecessary damage and production costs (Reebs, 2007). Moreover, the immobility response of death feigning in rats has been shown to reduce traumatic stress disorder (Zhou, 2018). Under such circumstances, our results might provide foundations of the physiological response for further research requirements in E. scrobiculatus, in which death feigning is a key behaviour modification that highlights a link between biological pest control and developing management practices.

The authors first thank the Forest Quarantine Station in Lingwu city for providing the laboratory and experimental device. The authors gratefully acknowledge undergraduates Qian Wang, Wenjuan Guo, Xuewen Sun, Weicheng Ding and Hongyu Li of the Beijing Forestry University College of Forestry for their assistance. Special thanks to Dr Almut Kelber and the reviewers.