Preference of black soldier fly larvae for feed substrate previously colonised by conspecific larvae

The black soldier fly Hermetia illucens Linnaeus (Diptera: Stratiomyidae; BSF) is gaining interest as an alternative protein ingredient for livestock feed. Larval aggregation behaviour occurs commonly in larvae of various dipteran species. However, the cues initiating aggregation behaviour and its occurrence in subsequent larval instars are still unknown in BSF larvae. Here, we focus on understanding the attraction of larvae to cues left behind in the substrate by conspecific larvae. We developed a dual‐choice test in the dark and examined the preference behaviour of different instars with the use of video recordings. Larval choices were identified based on first substrate contact, cumulative number of substrate contacts, entry of the substrate and number of larvae present in a substrate over time. Our results demonstrate that BSF larvae discriminated between similar substrates with or without cues released by conspecifics, with a strong preference for previously colonised substrates. Our results also showed a difference in behaviour between head‐capsule classes. Substrate contacts occurred more frequently in larvae from a lower head‐capsule class and the number of larvae present in a substrate over time differed between the head‐capsule classes 0.71–0.80 mm and 0.91–1.00 mm. Demonstrating arrestment in response to chemical cues from conspecifics is the first step of understanding aggregation behaviour of BSF larvae and offers opportunities to identify the chemical cues involved.

An aggregation could also offer protection from unfavourable environmental conditions such as low temperature through increased temperature within an aggregation (Aubernon et al., 2016;Ruf & Fiedler, 2000). Increased temperature can reduce larval developmental time and decrease the risk of food shortage and predation (Podhorna et al., 2018), thus resulting in higher survival. The exact advantages of aggregation as well as the stimuli that initiate aggregation behaviour in BSF are still unknown. Initiation of aggregation behaviour can be due to chance, as BSF lay eggs in clutches (Rivers et al., 2011;Tomberlin et al., 2002). Clustering of eggs causes larvae to be in contact with conspecifics from the moment of hatching onward. However, for the initiation of aggregations, the interaction between conspecifics via visual, auditory, tactile or chemical cues may also be important (Wertheim et al., 2005).
The use of chemical cues is well-known and commonly used for communication in many insects such as caterpillars (Fitzgerald, 2003), ants, cockroaches and fruit flies (Buhl & Rogers, 2016). Chemical attraction and interaction between conspecifics can occur via cues left behind on surfaces (Fouche et al., 2018) or odour changes in the substrate previously colonised by conspecifics as is the case in larvae of Drosophila melanogaster Meigen (Diptera: Drosophilidae) (Durisko & Dukas, 2013) and Necrodes littoralis Linnaeus (Coleoptera: Silphidae) (Gruszka et al., 2020). Drosophila melanogaster larvae even showed an attraction to feed previously colonised by conspecifics when reared in isolation (Durisko & Dukas, 2013). The response to chemical cues left behind in the substrate may depend on developmental stage. For example, in N. littoralis larvae, the preference was instar specific (Gruszka et al., 2020). Third-instar N. littoralis larvae showed a preference for substrate previously colonised by conspecifics whereas first-instar larvae were less abundant and second-instar larvae even showed no preference for substrate previously colonised by conspecifics.
In BSF larvae there are a few differences between instars that may have an effect on preference behaviour. Duration of BSF 5th and 6th instars is longer and the growth rate is lower than that of earlier instars (Gligorescu et al., 2019). BSF larvae also show wandering behaviour in the 7th instar when close to pupation. In this wandering prepupal stage, larvae stop feeding and leave the substrate (Georgescu et al., 2020). The concentration and composition of chemical cues released by conspecifics may also vary between instars similar to the changes in chemical profile of larvae of Calliphora vicina Robineau-Desvoidy (Diptera: Calliphoridae) (Frederickx et al., 2012).
The present study investigates the response to cues from conspecifics left behind within feed substrates by BSF larvae. We aim to gain a better understanding of the causes of aggregation behaviour in different larval stages. The following hypotheses were experimentally tested employing a dual-choice test: (1) substrate previously colonised by conspecifics is preferred by BSF larvae; (2) the attraction to substrate previously colonised by conspecifics depends on larval instar.

| Insects
Insects from the BSF colony at the Laboratory of Entomology at Wageningen University were used. This colony is maintained in a climate chamber at a temperature of 27 ± 1°C, a relative air humidity of 70 ± 10% and an L:D cycle of 16 h:8 h. Larvae were reared on a standard chickenfeed diet; a mixture of two parts tap water and one part chickenfeed (Kuikenopfokmeel 1; Kasper Faunafood) and completed larval development in 12 days. Females that had eclosed from puparia laid eggs in corrugated cardboard strips (±3.5 × 10 cm) placed above an oviposition box filled with sawdust mixed with mouse faeces. After providing the female flies access to the oviposition boxes for 6 h, the cardboard strips with freshly deposited eggs were collected and stored in a Petri dish with moist filter paper until larval hatching occurred 48-60 h later. After hatching, 250 neonate larvae were transferred with a soft brush to a 250 mL experimental rearing container (108 × 82 × 45 mm, Polypropyleen-PP, Gédé Verpakkingen B.V.). A window (8.0 cm × 5.5 cm) was cut into the lids of the rearing boxes and covered with mesh to allow ventilation. The larvae were provided with 150 g of standard chickenfeed diet. Experimental rearing containers were incubated in the climate chamber under the conditions mentioned above.

| Video observation set-up
To record larval behaviour two set-ups were developed, each in a wooden cabinet (60 cm × 60 cm × 40 cm, 402.055.56 IKEA) ( Figure S1).

| Experimental substrate preparation
To investigate the attraction to conspecific cues, colonised and uncolonised substrates were created 24 h before the experiments ( Figure 1a). The substrates consisted of 75 g of standard chickenfeed diet incubated in a 250 mL rearing container. The colonised substrate was produced by adding 50 randomly selected larvae of known age from the rearing container to the container with fresh feed. The uncolonised substrate was exposed to the same environmental conditions but no larvae were added. The containers with colonised and uncolonised substrates were incubated in a climate chamber at 27 ± 1°C, 70 ± 10% RH for 24 h. Just before using the substrates in the experimental setup, the larvae were removed from the colonised substrate.

| Head-capsule measurements
After the dual-choice assay, the groups of 10 larvae were collected from the arena, weighed (Mettler Toledo NewClassic MF, model F I G U R E 1 Overview of the methods used to explore substrate preference of a group of 10 H. illucens larvae. (a) The two different substrates that were created for the choice test. The colonised substrate had 50 larvae present within the substrate during 24 h. The uncolonised substrate had no contact with larvae during the same 24 h period. (b) The dual-choice test arena where 10 simultaneously released H. illucens larvae were able to choose between the uncolonised or the colonised substrate from which the larvae had been removed.

| Statistical analysis
Statistical analysis was conducted in R (version 4.1.0; R Core Team, 2021).

| First substrate choice
The first choice of the larvae included the first contact with and the first entry into one of the two substrates (Table 1). These variables were tested for substrate choice and differences between headcapsule classes using a generalised linear (mixed-effects) model (GL(M)M) with a binomial distribution and logit link function (lme4 package) (Bates et al., 2015). For the first entry, the experimental data were used as a random intercept based on Akaike's information criterion (AIC) (Bertrand et al., 1988). For the first contact, the experimental data explained minimal variability and was not included as a random intercept. After model selection, the significance of substrate choice and head-capsule class was tested with the GL(M)M providing the best fit via a type-II Wald Chi-square test (car package) (Fox & Weisberg, 2019). In cases where the larvae did not enter a feed dish within the 35 min trial time or the first contact was uncertain, the replicate was excluded from the first-choice analysis.

| Substrate preference
Substrate preference was analysed considering the total number of contacts and the total number of entries (Table 1). These variables were also tested for the difference between head-capsule classes.
The total number of contacts and the total number of entries were both tested with the use of a generalised linear mixed-effects model (GLMM). Since a Poisson log link function resulted in overdispersion, a negative binomial distribution and log link function were used (lme4 package; Bates et al., 2015). The experimental data were used as a random intercept based on AIC (Bertrand et al., 1988

| Time course in larvae presence
The time course of the presence of larvae in the substrate (Table 1) was analysed for substrate type and differences between head-

Behaviour Description
First contact The first larva of the group of 10 larvae making contact with a feed dish with its mouthparts. This is considered the first contact choice.

First entry
The first larva of the group of 10 larvae entering a feed dish with the entire body. This is considered the first entry choice.

Total number contacts
The total number of contacts by the 10 larvae made with their mouthparts touching the feeding dish during the 35 minute experiment.

Total number entries
The total number of entries by the 10 larvae made with their entire body in a feed dish during the 35-min experiment.

| First choice
Black soldier fly larvae did not discriminate between substrate types during the first contact with the feed dish (χ 2 = 3.18, p = 0.075) ( Figure 3a, Table S1). Head-capsule class did not affect the total number of first contacts (χ 2 = 0, p = 1) ( Figure 3a, Table S1). There was no significant interaction between head-capsule class and first substrate contact (χ 2 = 5.77, p = 0.217) (Figure 3a, Table S1). The first entry into the substrate was influenced by the substrate type, with more larvae entering the colonised substrate first (χ 2 = 15.38, p < 0.001) (Figure 3b, Table S1). Head-capsule class did not affect the first entry (χ 2 = 6.67, p = 0.155) and there was no significant interaction with the first entry into the substrate (χ 2 = 4.25, p = 0.372) (Figure 3b, Table S1).

| Total number of contacts during the 35-min exposure
There was a strong effect of substrate on the total number of contacts, showing that the larvae prefer to contact colonised substrate (χ 2 = 111.80, p < 0.001) (Figure 4a, Table S1). Head-capsule class had a significant influence on the total number of contacts with either substrate type (χ 2 = 16.09, p = 0.003) (Figure 4a, Table S1). The total number of contacts larvae made differed significantly between headcapsule classes 0.61-0.70 and 0.91-1.00 (p = 0.010), 0.61-0.70 and 1.01-1.10 (p = 0.003) and 0.71-0.80 and 1.01-1.10 (p = 0.049) ( Figure 4a, Table S2). There was no significant interaction between substrate type and head-capsule class (χ 2 = 2.15, p = 0.71) (Figure 4a, Table S1). These results show a strong effect of substrate type and an influence of head-capsule class for the total number of contacts during the experiments.

| Total number of entries during the 35min exposure
There were more entries in the feed dishes which contained colonised substrate over uncolonised substrate (χ 2 = 83.78, p < 0.001) (Figure 4b, Table S2). However, in contrast to the total number of contacts with the feed dishes, the total number of entries was not significantly influenced by head-capsule class (χ 2 = 5.14, p = 0.273) (Figure 4b, Table S2).

| Time course in entry preference
The video recordings allowed us to monitor the time course in substrate entry and departure during the 35 min observation period.
There was a significant effect of substrate on the number of larvae present (χ 2 = 69, p < 0.001, Figure 5, Table S1). Head-capsule class also significantly affected the time course (χ 2 = 21, p < 0.001) showing a significant difference in responses of larvae with head-capsule classes 0.71-0.80 mm and 0.91-1.00 mm ( Figure 5, Table S3). There was no significant interaction between head-capsule class and substrate type (χ 2 = 7, p = 0.146) ( Figure 5, Table S1). There was a significant effect of time: the number of larvae present within both substrate types increased with time (χ 2 = 337, p < 0.001) ( Figure 5, Table S1).

| DISCUSS ION
This study shows that BSF larvae exhibit arrestment by substrate that was previously colonised by conspecifics during 24 h and demonstrates that the later substrate is also preferentially colonised.
This suggests that BSF can not only distinguish between substrates with widely different chemical and microbial contents (Parodi et al., 2020) but can also discriminate among batches of the same diet that have either been previously colonised by conspecifics or not. These observations indicate that aggregation of black soldier fly larvae could be mediated by chemical cues inherent to previously colonised feed substrate. The source of the cues could be the larvae themselves or micro-organisms associated with the larvae. Similar findings were reported for N. littoralis (Gruszka et al., 2020) and D.

| Chemical cues
Larvae themselves may be able to produce chemical cues causing attraction and/or stimulating substrate entry and arrestment.
Such cues left behind on surfaces by con-and heterospecific individuals were previously found to influence preference behaviour in larvae of Lucilia sericata Meigen (Diptera: Calliphoridae) (Boulay et al., 2013;Fouche et al., 2018) and Calliphora vomitoria Linnaeus (Diptera: Calliphoridae) (Fouche et al., 2018). These cues may serve as aggregation pheromones, which are present in a variety of non-eusocial insects frequently associated with microbes (Wertheim et al., 2005). In Drosophila simulans However, the preference of BSF larvae for the previously colonised substrate may also be explained by the presence of microbes in the substrate. Microbes can have an effect on massreared insects and are known to impact behaviour in several species (Jordan & Tomberlin, 2021). Rotten substrate can be nonpreferred by L. sericata larvae for at least the first 30 min (Fouche et al., 2021) as microbes can have a negative effect on development (Richards et al., 2013). However, preference for a microbeinoculated substrate can also benefit larval performance, as fly larvae likely utilise nutrients from microbes (Gold et al., 2018).
Therefore, BSF larvae may actively search for microbes, as has been demonstrated for D. melanogaster larvae and adults (Wong et al., 2017). Microbial changes traceable to the BSF gut bacteria can be found in certain substrates (Jiang et al., 2019). The bacterial composition of the substrate is mainly affected by BSF larval density and substrate type (Schreven et al., 2022), and attractiveness may vary due to the microbiome composition. It was demonstrated that D. melanogaster adults had a strong attraction to Lactobacillus-containing medium (Wong et al., 2017), a group of bacteria that was also found in the chickenfeed substrate containing BSF larvae (Schreven et al., 2022). Similarly, BSF larvae may respond to cues produced by Lactobacillus. Larvae of D. melanogaster showed a similar attraction to substrate colonised by larvae with an intact microbiome but not to substrates colonised by axenic larvae (Venu et al., 2014).

| Differences between instars
The instars used in this research were predominantly 4th and 6th

| Time course in behaviour
The number of larvae within any substrate type increased over time and significantly more larvae were present in the colonised sub-   within an aggregation found in L. sericata larvae (Boulay et al., 2013).
However, this increase was observed over a much longer time frame, from 30 min until 24 h (Boulay et al., 2013). In the woodlouse Porcellio scaber Latreille (Isopoda: Porcellionidae) the initiation of aggregation behaviour was much faster, as more than half of the test specimens were aggregating after 10 min (Devigne et al., 2011). In N. littoralis larvae the aggregations formed after 5 min (Gruszka et al., 2020).
Even though our results did not test aggregation behaviour directly, they hint toward a similar increase in aggregation tendency over time in BSF larvae.

| Significance of aggregation behaviour
This research shows the first evidence of chemical communication between BSF larvae. If the compounds involved can be isolated and identified they may be used as cues for initiation of aggregation behaviour, a relevant first step to discovering the underlying mechanisms of this frequently occurring behaviour in BSF larvae.
Aggregation behaviour can have advantages for BSF larvae as it may increase food intake via social digestion (Gregg et al., 1990), offer protection from predation or parasitism (Wertheim et al., 2005) and reduce environmental stress (Broly et al., 2013). Large aggregations can lead to a local increase in temperature (Charabidze et al., 2011) resulting not only in more optimal rearing conditions but may also be a behavioural response against pathogens. It is known that insects and other ectotherms will seek warmer environments when infected (Hunt et al., 2011;Kluger, 1979). This phenomenon is known as behavioural fever (Boorstein & Ewald, 1987;Bundey et al., 2003;Kluger, 1979).
Understanding what initiates and changes aggregation behaviour of BSF may help to develop conditions in mass-rearing facilities that allow natural behaviour and welfare of the insects.
Therefore, quantification of aggregation behaviour may aid discussions about insect welfare (van Huis, 2019). The strong preference for substrates previously colonised by conspecifics offers the first insight into the still poorly understood aggregation behaviour in BSF larvae.

ACK N OWLED G M ENTS
We thank Hans Smid for help with designing the experimental camera set-up and Thibault Costaz for the discussions about data analysis and Eleanor Gourevitch for linguistic advice. Our research has been supported by the Dutch Research Council (NWO; NWA programme, InsectFeed project, NWA.1160.18.144).

CO N FLI C T O F I NTE R E S T S TATE M E NT
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets generated during and analysed during this study are available in the 4TU repository (Kortsmit, 2022