Larval development, metabolism and diet are possible key factors explaining the decline of the threatened Dytiscus latissimus

In Western Europe, the diving beetle Dytiscus latissimus (Coleoptera: Dytiscidae) has become rare and went extinct in several countries during the last century. This study investigated whether larval development rate, metabolism and feeding ecology differ between D. latissimus and the congeneric D. lapponicus to explore factors explaining its decline. During instar I and II, D. latissimus larvae developed faster and gained more weight than D. lapponicus larvae. In accordance, D. latissimus larvae had higher oxygen consumption rates than D. lapponicus larvae, which signifies a greater energy expenditure. Food preference tests showed that D. latissimus larvae strongly prefer caddisfly larvae (Trichoptera: Limnephilidae) with early instars being obligatory dependent on caddisfly larvae for their development. Only instar III larvae readily fed on alternative prey items. In contrast, D. lapponicus larvae had a broader diet and even rejected caddisfly larvae. Based on field observations, availability of caddisfly larvae strongly declined before the end of the larval development of D. latissimus, suggesting that time constraints on food availability limit completion of larval development. Our results suggest that food limitation during (early) larval stages is a possible bottle‐neck for this species, potentially explaining its disappearance from former localities. Promoting caddisfly larvae in the vicinity of D. latissimus oviposition sites, may possibly safeguard the present distribution of D. latissimus and support the species recovery. Although more research is needed, promoting leaf litter in shores may be beneficial to the shredding caddisfly larvae and in turn for their predator D. latissimus.


Introduction
Worldwide scientists have expressed their concern for the current loss of biodiversity (Barnosky et al., 2011;Ceballos et al., 2015). Research has suggested that the extinction rates of insects exceed those of birds and vascular plants (Thomas et al., 2004). For successful recovery of threatened insect species, it is paramount to understand their life history and what habitat conditions they require (e.g. Anthes et al., 2003;Thomas et al., 2009).
One of those insects for which that valuable knowledge is required but largely missing is Dytiscus latissimus (Coleoptera: Dytiscidae), Europe's largest predacious diving beetle. Its geographical range includes Central-and North-Europe and the west of Siberia (Nilsson & Holmen, 1995). During the twentieth century, its population numbers in Central and Western Europe have declined drastically (Hendrich & Balke, 2000;Holmen, 1993). In seven countries, D. latissimus has not been observed for the past 50 years and here the species is considered to be extinct (e.g. Gerend, 2003;H ajek, 2004;Queney, 2004;Cuppen et al., 2006;Scheers, 2015). In the Netherlands, D. latissimus was also believed to be extinct (Huijbregts, 2003) until it was rediscovered in 2005 (Van Dijk, 2006).
In Europe, D. latissimus inhabits a wide range of water types, including fish ponds, non-polluted clear or dystrophic lakes, old river oxbows and water bodies formed by damming or peat and turf excavation (Holmen, 1993;Hendrich & Balke, 2000, 2005Cuppen et al., 2006;Vahru sevs & Kalnin ß s, 2013). These habitats show a large variation in depth, transparency and hydrochemical aspects. The presence of plants that are suitable for egg deposition is considered a key factor for its reproduction and includes marsh-marigold (Caltha palustris), acute sedge (Carex acuta), cyperus sedge (Carex pseudocyperus), bottle sedge (Carex rostrata) and bogbean (Menyanthes trifoliata; Holmen, 1993;Balke, 2000, 2005). However, these factors by themselves are inadequate to explain its current rarity and scattered distribution.
Thus, it seems likely that additional factors limit the population size and distribution of D. latissimus (Cuppen et al., 2006;Reemer et al., 2008). There is an urgent need to identify these factors as these will enable the development of conservation strategies for existing habitat and the development of new habitat, and hereby realize the international protection of this species [Council of Europe 1979, enforced by the European Habitats Directive of 1992 (annex II)].
Several characteristics of D. latissimus suggest that the species, especially its larva, is sensitive to food shortage. First of all, the larvae of D. latissimus take about 2 months to develop from an egg into an adult (Vahru sevs, 2009), which is about the same amount of time required by larvae of other Dytiscidae like Dytiscus lapponicus and Cybister lateralimarginalis. The adults of D. latissimus are however larger than those of any other species of diving beetles found in Europe with an average size of 3.9-4.4 cm, and adults weighing 2.8 and 2.2 times more than adults of respectively D. lapponicus and C. lateralimarginalis (Nilsson & Holmen, 1995;own data). Completing larval development in the mentioned time span may thus be more challenging for D. latissimus than for smaller dytiscid species and is expected to require a higher metabolic rate and food intake.
Previous studies furthermore indicate that D. latissimus larvae have a strong preference for caddisfly larvae (family: Limnephilidae) above other prey species like tadpoles, isopods (family: Asellidae) and mayfly nymphs (family: Siphlonuridae; Blunck, 1923;Johansson & Nilsson, 1992;Vahru sevs, 2009). Being dependent on one or a few food sources, oligophagous species require higher densities of suitable food sources which likely results in a more stringent demand on the quality of the habitat for their preferred prey. Moreover, oligophagous species are more sensitive to decreasing prey availability since they cannot switch to other food sources.
Finally, the type of ponds where D. latissimus is found in the Netherlands is rich in other (large) carnivorous diving beetles like C. lateralimarginalis and D. lapponicus (Cuppen et al., 2006;Reemer et al., 2008). If the larvae of these other species compete with the larvae of D. latissimus for the same prey items, this could further increase food limitation in D. latissimus.
A fast development, possibly resulting in a high metabolism and high food demand, in combination with a specialist diet of the larvae may play an important role in habitat selection and population dynamics of D. latissimus. In order to test this hypothesis, the growth and moult of D. latissimus larvae was followed in the laboratory throughout the larval stages when subjected to a diet with and without Limnephilidae larvae. Furthermore, the metabolic rate and thermal sensitivity of metabolism of larvae kept under natural feeding conditions was determined by measuring their oxygen consumption from both air and water at different temperatures. Finally, the diet of D. latissimus larvae was studied with food preference tests. All the experiments were repeated in larvae of D. lapponicus to be able to compare development, metabolism and diet between the two species. The reason for choosing D. lapponicus as a comparison is that this species is common and frequently co-occurs with D. latissimus in reasonable numbers in the Netherlands (Cuppen et al., 2006). A comparison of lifehistory traits between the successful D. lapponicus and less successful D. latissimus in the same environment may give an indication why the latter is rare. The laboratory data were extrapolated to the field situation with the help of the phenology and densities of potential prey species and the phenology of D. latissimus larvae monitored in the field.

Laboratory study
Collection of the animals.
To minimise the impact on the Dutch population, larvae of D. latissimus used in the experiments were acquired from eggs in stems of C. acuta that were collected from fish ponds in Ruģel ßi, Latvia (55°52 0 33.1″N, 26°35 0 21.2″E), where the species is much more common. Here, its eggs are easier to find and collecting them does not negatively affect the population (Vahru sevs & Kalnin ß s, 2013 Potential prey animals used to assess food preference were collected from Booy's Veentje (52°49 0 00.7″N, 6°15 0 49.6″E) and Zandveen (52°49 0 39.1″N, 6°26 0 27.0″E) in the Netherlands.

Housing conditions.
Larvae were individually kept in plastic containers of 70 9 70 9 95 mm with 200 ml of surface water from the collection site (D. lapponicus) or from Booy's Veentje (D. latissimus). As the surface water temperature of moorland pools in May lies within the range of 14-17°C (unpublished data Bargerveen Foundation), the larvae were kept at a constant temperature of 15°C in a temperature controlled room. A piece of plastic coated wire mesh in the plastic containers was provided as structure for the larvae.
We distinguished two different groups of D. latissimus larvae. The first group was exclusively fed caddisfly larvae (Trichoptera: Limnephilidae sp.), with the exception of Instar III larvae who in addition were offered dragonfly larvae (Odonata: Libellulidae sp.) and tadpoles. The prey species were offered ad libitum, except in the case of instar III larvae, when food supplies were often no longer sufficient to feed the larvae ad libitum. The second group of D. latissimus larvae were denied Limnephilidae larvae and were instead fed ad libitum with mayfly larvae (Ephemeroptera: Cloeon), damselfly larvae (Odonata: Coenagrionidae sp.), water fleas (Diplostraca) and tadpoles.
Larvae of D. lapponicus were maintained on an ad libitum diet consisting of a large variety of prey, including larvae of Ephemeroptera sp., Coenagrionidae sp., and Libellulidae sp., and Diplostraca sp. and tadpoles.

Growth and development.
After hatching, larvae of Dytiscus spp. go through three larval stages. Their larvae crawl on land at the end of instar III to pupate in a shallow burrow near the water. Two to three weeks later it emerges as an adult. The development of D. latissimus and D. lapponicus larvae was closely followed in the laboratory. All eggs, larvae, pupae and adults were monitored daily and the date of hatching, moulting, pupation and death was recorded and used to calculate developmental time of the individuals. We compared the duration of the developmental period between D. latissimus and D. lapponicus for each larval instar using Mann-Whitney Utests. Too few larvae went through a successful pupation to include the pupal stage. These and all subsequent analyses were performed in RStudio (Version 0.99.473, August 12th 2015).
To track growth of all larvae, the length of the larvae measuring from the frontal clypeus to the urogomphi was taken regularly. To measure its length, the larva was taken out of the water and put on a narrow v-shaped surface coated with a thin film of water. Most of the time this immobilised the larva long enough to measure its length. In other cases a pair of tweezers was used to carefully keep the body of the larva straight. To obtain a better measure of growth, we converted length to weight using the relation between length and weight measurements of larvae of both species that were recorded after respiration trials (see below) or after a larva had died (Fig. S1). The weight of instar I, II and III larvae at different ages was analysed with a linear regression model including the factors age and species, and the interaction between these two factors. As most individuals were measured multiple times within an instar, the model was also run with a linear mixed-effect regression including 'individual' as a random factor to incorporate the non-independence of these data points. As this did not alter the results, the simpler linear regression model was used.
Finally, the growth and development of D. latissimus larvae fed on a limnephilid diet was compared with that of D. latissimus larvae that were fed a non-limnephilid diet using a Mann-Whitney U test. The latter group was not used in any of the other experiments described below.

Metabolism.
In order to assess whether the two Dytiscus species have a different metabolic rate or differed in thermal sensitivity of their metabolism, oxygen consumption rates were determined for instar I, II and III larvae of D. latissimus and for instar II and III larvae of D. lapponicus at 10, 15 and 20°C. Since larvae can perform aquatic and aerial gas exchange (Vahru sevs, 2009; Yee, 2014), we measured oxygen consumption from both air and water. Measurements took place in closed glass chambers of 19.4 ml (instar I and II) and 101.0 ml (instar III). Chambers were initially completely filled with Dutch Standard Water (DSW) (Nederlands Normalisatie Instituut, 1980). After insertion of the larvae in the chamber, an air bubble was injected comprising of 2.5 ml (instar I), 3 ml (instar II), 15 ml (instar III; D. latissimus) or 10 ml (instar III; D. lapponicus). The air injected was saturated with water vapour and temperature equilibrated in advance. After 10 min of habituation, the chamber was closed, and the larva was left undisturbed for another 60 min before the actual measurements commenced. A fine nylon mesh inside the chamber provided structure.
Oxygen pressure was measured optically every 15 s with a mini sensor spot connected to an OXY-10 mini oxygen meter (PreSens instruments, Germany), which was situated in the air compartment at the top of the chamber where the air bubble was held. At given intervals, the respiratory chamber was carefully inverted to bring the water in contact with the oxygen sensor allowing the oxygen pressure in the water to be measured. Oxygen concentration in the water was measured at least once every half hour in case of instar I larvae, and at the beginning and end of each trial in case of instar II and III larvae. During the measurements, the chambers were completely submerged in a temperature controlled water bath.
Trials at 10, 15 and 20°C always took place on the same day, in this order. The night before the trials the larvae were kept at 10°C with ad libitum access to food. After a trial, the chambers were opened and temperature was gradually increased with 5°C with an average rate of 0.23 AE 0.02°C per minute if another trial followed. Fifteen minutes after the water had reached the right temperature the chambers were closed and the larvae were left for another 15 min undisturbed before measurements of oxygen consumption commenced. Trials lasted 97 min on average, varying between 41 and 164 min. The length of a trial was determined by the rate of oxygen consumption which depended on species, larval stage and temperature (Fig. S2).
Per individual, a linear regression was fitted to the series of data points of a single trial to determine rate of oxygen consumption (expressed as lmol O 2 h À1 ). Oxygen consumption rates were corrected for oxygen diffusion from the water into the air bubble according to the difference in the partial pressure of oxygen between the two compartments, using empirically determined relationships for diffusive flux (see also Verberk & Bilton, 2015). In addition, oxygen consumption was corrected for background respiration, which was measured using blanks. Both the effect of background respiration and diffusive flux were small: corrected rates of oxygen consumption from water were 0.5% lower than the initial values, and highly correlated (R 2 = 0.995). Similarly, corrected values of oxygen consumption from air were 0.82% higher than the initial values and also highly correlated (R 2 = 0.999).
Linear regressions were used to investigate whether the oxygen consumption differed between D. latissimus and D. lapponicus in relation to their weight or temperature. We initially included both larval instar and weight, but since these were highly correlated we excluded larval instar from the model. The same model was run with a linear mixed-effect regression including 'individual' as a random factor to include the non-independency of measurements of the same individual that were taken at different temperatures. The linear mixed-effect regression produced qualitatively similar results as the linear regression model, therefore, the linear regression model was used. Furthermore, we used linear regression on the ratio between aquatic oxygen consumption and aerial oxygen consumption to compare the relative oxygen consumption from both media between D. latissimus and D. lapponicus larvae of different size and at different temperatures.

Multiple-choice food preference test.
Prey preference of D. latissimus and D. lapponicus larvae was assessed by performing multiple-choice preference tests. Different instars were tested separately. Based on field observations at Booy's Veentje, the most abundant prey species during larval development of D. latissimus and D. lapponicus were selected. To instar I larvae we offered Diplostraca, and larvae of Ephemeroptera and large Limnephilidae sp.
(i.e. L. flavicornis, L. lunatus, L. marmoratus, L. stigma and L. subcentralis). We offered the same prey species to instar II and III larvae, and considering the larger size of these instars we offered them two additional, larger prey species, namely larvae of Coenagrionidae sp. and larvae of Libellulidae sp. Three additional preference tests were performed with instar I, II and III larvae of D. latissimus that included tadpoles in addition to the prey species listed above. Five individuals of small prey types, i.e. Diplostraca, Ephemeroptera and tadpoles, and one individual of the large prey types were offered during a trial to roughly equalise encounter rates between the Dytiscus larva and the different prey species.
The food preference test took place in a glass aquarium of 90 9 90 9 110 mm with a wire mesh on the bottom. Larvae were tested in 200 ml (instar I and II) or 250 ml (instar III) of filtered surface water, which was kept at 20-21°C. The night before the food preference test, instar I and II larvae were deprived of food in order to ensure that the larvae would eat during the preference test. Prior to each trial, randomly selected individuals of each prey type of approximately the same size were added to the aquarium and left for 15 min to habituate. Subsequently, a larva was introduced and its feeding behaviour was observed. Observation trials with instar I larvae were terminated after a larva ate a prey bigger than Diplostraca, as instar I larvae did not eat for several hours after that. These trials lasted 50-280 min (172 min on average). Instar II and III larvae were observed for a minimum of 150 min and until a larva had consumed at least one prey. Observation duration for trials with instar II and III larvae lasted for 150-321 min (180 min on average). Observation time did not influence the number of prey that a larva consumed during one trial (ANOVA: F 1,85 = 2.08, P = 0.153).
To determine whether larvae of D. latissimus and D. lapponicus discriminated between the prey types offered, successful attacks on all prey types were recorded. Each captured prey was immediately replaced to ensure that prey density and composition remained constant throughout the test. For each preference test on a given larva, the relative consumption of each prey type that was offered during a trial was calculated as: where R i is the number of individuals of prey type i consumed during one trial, and R tot the number of individuals consumed including all prey types during one trial. Furthermore, the relative availability of prey types was calculated, which accounts for differences in the number of individuals available of a given prey type: where A i is the number of individuals of prey type i offered during one trial, and A tot the number of individuals offered including all prey types during one trial. The relative consumption and relative availability were combined in the so-called electivity index (Ivlev, 1961): Consequently, per individual larva one electivity index was obtained for each prey type that was offered during the trial. Electivity indexes consisted of values between À1 and 1. Values below zero indicate rejection of a certain prey type in favour of others, values of or close to zero indicate indifference, and values above zero indicate a preference for a given prey. To determine prey preference or rejection, we tested whether the electivity indexes deviated significantly from zero (i.e indifference) by means of multiple binomial tests, one for each prey type. To prevent false positives due to the usage of multiple tests, P values were corrected according to the Benjamini and Hochberg procedure (Benjamini & Hochberg, 1995;Glickman et al., 2014).
In addition to successful attacks, attempted attacks were recorded. The ratio between the number of successful and attempted attacks was calculated to determine catching efficiency. Furthermore, handling time, i.e. the moment that a larva captured a prey to the moment it released the prey, was recorded to discriminate between complete and incomplete feeding of prey that were successfully attacked. Dytiscid larvae use their hollow mandibles to catch and inject digestive enzymes from the midgut into prey, after which they ingest the liquefied tissue (Yee, 2014). After complete consumption, merely the skin of the prey remains.

Field study
The field study was conducted at Booy's Veentje, which is one of the last known localities were the species still occurs in the Netherlands (Cuppen et al., 2006;Reemer et al., 2008;Koese et al., 2010). For the physical and chemical characteristics of Booy's Veentje see Supporting Information.

Phenology of prey species.
Aquatic macroinvertebrates in Booy's Veentje were sampled on April 5, April 28, May 23 and June 13, to determine their phenology. The samples were taken at a total of six sampling locations along two fixed transects perpendicular to the shore covering different vegetation zones. One transect was located at the south side (sampling locations 1A, 1B, 1C; 52°48 0 59.3″N, 6°15 0 47.1″E) and the second on the north side (sampling locations 2A, 2B, 2C; 52°49 0 02.6″N, 6°15 0 50.8″E). The characteristics of the different vegetation zones of both transects are summarised in Table S1.
The macroinvertebrate samples were taken alongside each transect with a D-frame net (width: 0.3 m; depth: 0.3 m, mesh size: 1 mm). Of each vegetation zone of both transects, 0.3 m 2 was sampled. In the laboratory, samples were washed over three sieves with 2, 1 and 0.5 mm mesh size and sorted. The animals were counted and identified to order level, with the exception of Trichoptera which were identified to species level based on Higler (2005) and Waringer and Graf (1997). Samples collected in zone A and C contained a lot of material. Therefore, sorting was stopped after 1 h and the numbers in the total sample material were calculated using the ratio between the volume of the remaining and the sorted sample material. Samples of zone B were analysed completely. For each collection date, densities (number of individuals m À2 ) were averaged across all samples. Due to a low water table, no samples were taken at sampling location 1A on May 23, and at 1A and 2A on June 13.

Phenology of D. latissimus larvae.
We sampled the shore vegetation (mainly floating rafts of M. trifoliata and C. rostrata) for larvae of D. latissimus on April 28, May 2, 9, 13, 23 and 26, and June 9 and 13, with a sieve (diameter 200 mm and mesh width 1 mm). D. latissimus larvae were identified in the field, its instar was noted and it was subsequently released. The phenology of the three instars of D. latissimus larvae in Booy's Veentje was reconstructed with the use of these observations and the developmental rates of the different instars observed in the laboratory.

Laboratory study
Growth and development.
There were clear differences in rate of development across the different instars resulting in difference in duration of development. Developmental time of instar I and instar II larvae of D. latissimus were significantly shorter than that of D. lapponicus larvae ( Fig. 1; Mann-Whitney U = 1.5, P < 0.001, n1 = 17, n2 = 25; U = 9, P = 0.002, n1 = 11, n2 = 8, respectively). On the contrary, development of instar III larvae took longer in D. latissimus larvae than in D. lapponicus larvae, but sample size was low and the difference was not significant (Mann-Whitney U = 4, P = 0.236, n1 = 4, n2 = 1).
None of the 15 instar I larvae of D. latissimus that were subjected to a diet without Limnephilidae larvae reached the second larval stage. Instead they died at an average age of 9.8 AE 4.74 days (n = 15) during which they reached an average length of 1.77 AE 0.25 cm (n = 7). This was probably the result of malnutrition, as these individuals were observed to start eating Ephemeroptera larvae, Coenagrionidae larvae and tadpoles only after 3 days since the treatment started and never in great quantities. In comparison, 22 individuals of the 68 instar I larvae of D. latissimus that were raised on Limnephilidae larvae developed into the second instar at an average age of 7.06 AE 1.25 days (n = 17). They also grew significantly larger compared to larvae raised on a diet without Limnephilidae, reaching an average length of 2.36 AE 0.30 cm during their first instar (Mann-Whitney U = 254.5, P < 0.001, n1 = 39, n2 = 7). Of the 46 individuals that did not develop into the second instar, mortality was caused through predation by Limnephilidae larvae in 23 cases (50%), which may have been caused by the small housing conditions as this results in an uncommonly high interaction rate between the D. latissimus larva and Limnephilidae larvae.

Metabolism.
Total oxygen consumption (i.e. both aquatic and aerial oxygen uptake) by D. latissimus and D. lapponicus larvae was positively related to weight and temperature. This increase in oxygen uptake (i.e. slopes) did not differ between the two Dytiscus species (Table 1). The overall oxygen consumption (i.e. intercept) however was higher in D. latissimus than in D. lapponicus larvae ( Fig. 3; Table 1).
The contribution of aquatic gas exchange to total oxygen consumption was always outstripped by the contribution of aerial gas exchange in all larval stages across both species (Fig. 4). With increasing weight however the contribution of aquatic gas exchange further decreased in D. latissimus larvae. In contrast, larvae of D. lapponicus, if anything, increased their reliance on aquatic gas exchange with increasing weight.

Food preference.
Prey preference was very different between larvae of both species, with D. latissimus showing strong preference for Limnephilidae larvae, a prey type that was rejected by D. lapponicus. Across all three instars, larvae of D. latissimus showed the greatest preference for Limnephilidae larvae, feeding exclusively on this prey as  Fig. 2. Weight of instar I (circles), II (triangles) and III (squares) larvae of D. latissimus (red symbols) and D. lapponicus (blue symbols) in relation to their age. Linear regressions are displayed by the dotted lines. The grey zone indicates the 95% confidence interval. The encircled data points were excluded from the linear regression analysis and represent the weights (measured at death) of individuals that had stopped growing, but were somehow unable to moult and thus died several days after the moment that all other larvae moulted. instar I (Fig. 5a, b). Instar II and III larvae supplemented their diet with tadpoles, and in the case of the last instar also with Libellulidae larvae. Nevertheless, electivity index values for larvae of Limnephilidae remained highest. Moreover, tadpoles did not appear to be fully consumed and despite the seemingly small difference in prey volume the average handling time of tadpoles was short compared to the average handling time of In contrast, the diet of D. lapponicus larvae was less specialised. Individuals showed no clear preference for one of the offered prey types (Fig. 5c). However, Diplostraca seem to be rejected by instar II and III larvae, as were Ephemeroptera larvae by instar III larvae (Fig. 5c). Furthermore, D. lapponicus larvae were never observed to eat Limnephilidae larvae in any stage. Nonetheless, instar II and III larvae of D. lapponicus did show interest in Limnephilidae larvae. They were observed to bite repeatedly in the case of the Limnephilidae larvae [number of attempts per individual: 3.11 AE 2.08 (n = 9), instar II; 5.33 AE 2.8 (n = 6), instar III; mean AE SD], but they never  succeeded to capture the larva within. In comparison, more than sixty percent of the attacks on Limnephilidae larvae attempted by D. latissimus larvae were successful (instar I: 69.61% AE 40.50, n = 17; instar II: 62.77% AE 34.92, n = 12; instar III: 75.09% AE 24.25, n = 14; mean AE SD).

Field study
Phenology of prey species.
During the sampling period, densities of potential prey species reached maximum densities in late May, and decreased from this point (Fig. S3). As the results in this study suggest a strong preference and dependence of D. latissimus larvae on Limnephilidae larvae, results in this section will focus on the phenology of this specific group of prey species.
Generally, Limnephilidae larvae initiated pupation around mid-May. On May 23, more than half of the Limnephilidae larvae were prepupating or pupating (0.67 AE 1.49, prepupa; 2.0 AE 4.47, pupa; mean number of individuals m À2 AE SD). Around the start of June, no larvae or pupae of Limnephilidae were found.

Phenology of D. latissimus larvae.
On April 28, thirteen instar I larvae of D. latissimus were found, and another nine on May 2. On May 9, two instar III larvae were found, and on each day of May 13, 23, 26 and June 9, one instar III larva was found. Assuming that May 9 was the first date that instar III larvae appeared in Booy's Veentje and June 9 was the last date that instar III larvae were present, the phenology of D. latissimus larvae in Booy's Veentje was estimated based on the average developmental time of instar I, II and III found in the laboratory (Fig. 7).
Field occurrences of instar I and II larvae of D. latissimus co-occurred with the presence of Limnephilidae larvae, although peak densities of the latter had already been reached earlier. After mid-May, when Limnephilidae larvae densities were declining, instar II larvae were still developing. The development of instar III larvae lasted beyond the availability of Limnephilidae larvae.

Discussion
In this study, we investigated the development, metabolism and diet of D. latissimus larvae as this could possibly explain the limited distribution and low abundance of D. latissimus in the Netherlands and elsewhere. Our study reveals a higher growth rate and higher mass specific metabolism in D. latissimus larvae than in its congener D. lapponicus. We also found that D. latissimus has a more specialised diet than D. lapponicus, consisting primarily of Limnephilidae larvae. Thus, the highly specialised diet combined with a high energy and hence food demand, and having to attain a larger body mass, all suggest that food limitation could be a major bottle-neck in the life-history of D. latissimus. Consequently, we believe that the availability of Limnephilidae larvae in the habitat of D. latissimus is a key factor in explaining the species' distribution.

High growth and metabolic rate
Compared to D. lapponicus, D. latissimus larvae gained weight faster and developed in a shorter time frame (at least for the first two instars). To fuel this physical performance D. latissimus larvae consumed more oxygen. Aquatic respiration can be challenging, since, compared to air, the bioavailability of oxygen in water is lower and water is much more dense and viscous raising the cost of breathing (e.g. Verberk & Atkinson, 2013). Many aquatic organisms depend at least partly on integument respiration, but the capacity of integument respiration to meet oxygen demand is limited (Graham, 1988). Bimodal breathers have been shown to increasingly rely on aerial gas exchange when warming increases metabolic demand in an aquatic hemipteran (Verberk & Bilton, 2015). Declining oxygen availability likewise favoured aerial gas exchange to maintain metabolic rates in pulmonate gastropods (Jones, 1961), demonstrating that aerial gas exchange is less challenging than aquatic gas exchange. To meet their increasing metabolic demands, growing larvae of D. latissimus likewise increased the proportion of oxygen taken up from air.
It is conceivable that the larval developmental time found in this study differs between the laboratory and the field, due to differences in environmental conditions, such as water temperature and prey availability. During the period that larvae were observed in the field, temperatures at the water surface ranged from 11.9 to 23.0°C.
Temperatures were kept constant in the lab so that in this study, instar I and II larvae in the field have been exposed to slightly lower temperatures than the larvae in the laboratory, whereas instar III larvae in the field were exposed to slightly higher temperatures.
As temperature requirements and thermal responses may differ between populations across the geographical range of a species (Kilian & Nielson, 1971), we cannot exclude the possibility that development rates differed inherently between the Dutch population studied in the field and the Latvian population used in the experiments.

A highly specialised feeder
The food preference tests in this study confirm the strong preference of D. latissimus larvae for Limnephilidae larvae described by Blunck (1923), Johansson and Nilsson (1992) and Vahru sevs (2009). Moreover, the mortality of instar I larvae when they were denied Limnephilidae larvae as prey indicates that Limnephilidae larvae are essential for early larval stages, something which likely results in the species being extremely sensitive to low or fluctuating Limnephilidae densities. Our field study corroborates their importance as prey as instar I and II larvae of D. latissimus completed development before the Limnephilidae larvae are too scarce or inaccessible. This time constraint may also explain the fast development of instar I and II larvae of D. latissimus relative to those of D. lapponicus. Since D. latissimus larvae could no longer be fed ad libitum in the lab after they reached instar III, it is possible that larval development of instar III was delayed and that under normal food conditions they can develop faster. Field data showed however that at the time of instar III development, Limnephilidae density had dropped to such an extent, that larvae were probably forced to switch to alternative prey types to gain enough food. One alternative prey type may have been Limnephilus pupal cases, as instar III larvae were observed to be capable of piercing the case of the Limnephilidae larvae with their mandibles, and thus can reach the pupae within. It is unclear however, whether instar III larvae can actually locate an immobile pupa, considering that after chemical stimuli, movement seems to be the second important stimulus for Dytiscidae larvae to locate prey (Formanowicz, 1987). The results of our food preference tests suggest a greater need to switch to alternative prey in instar III larvae, as the larvae demonstrated an increase in diet breadth upon reaching instar III. This could be an adaptation to low numbers of Limnephilidae at the time of instar III development.
Although the experiments have been conducted on D. latissimus larvae of Latvian origin, we think that the results can be projected on the Dutch population and other European populations, as Blunck (1923) and Johansson and Nilsson (1992) already showed that the strong preference for Limnephilidae larvae is not unique for larvae of the Latvian population. Feeding behaviour  of the larvae seemed unaffected by the housing conditions as D. latissimus showed the same hunting behaviour as described by Johansson and Nilsson (1992).

Limnephilidae densities, an important factor?
A single larva of D. latissimus that was solely raised on Limnephilidae larvae consumed on average 198.54 Limnephilidae larvae during its entire larval development to an adult, which is on average six individual Limnephilidae larvae per day (see Supporting Information). Combining this information with the densities of Limnephilidae larvae that were found during larval development in Booy's Veentje, one larva of D. latissimus would need approximately 20 m 2 of riparian zone to develop into an adult. This is a considerable amount of space to maintain a sufficiently large viable population. Given that around 35% of the surface area of Booy's Veentje is riparian zone, Booy's Veentje could roughly support the development of 250 D. latissimus larvae each year.
This number drops when considering that this crude estimation assumes that the D. latissimus larvae consume all Limnephilidae larvae that are available and considering that parts of the riparian zone are unsuitable for Limnephilidae larvae. In addition, only a part of these individuals will survive to a breeding adult due to factors like predation, disease and food accessibility. Although it is assumed in this estimation that instar III larvae solely eat Limnephilidae larvae, it is unclear to what extent they rely on Limnephilidae larvae for their survival as they are able to feed on alternative prey items. Prey densities likely affect the feeding behaviour of the beetle larvae in terms of the time spent searching for prey, and the time for prey handling and consumption (Formanowicz, 1982;Kruse, 1983). Under high prey densities, larvae may switch to a new prey and only partially ingest their prey. Given that in our experiments most prey were eaten completely, we could have underestimated the prey densities required by D. latissimus larvae.
At sites where there is suitable habitat for Limnephilidae larvae, the accessibility of Limnephilidae larvae for D. latissimus larvae is of major importance for a sustainable population of D. latissimus. In order for the young instar I larvae of D. latissimus to reach the Limnephilidae larvae, the egg deposition sites for females of D. latissimus should be near or connected to suitable habitat for Limnephilidae larvae. In the study area of this study however, the zones where eggs are deposited are several metres out of the shore, whereas high Limnephilidae densities were found in the zones near banks that were lined with inclining trees and had high densities of leaf litter (data not shown). This distance between suitable egg deposition substrate and suitable Limnephilidae habitat can function as a bottle neck for the development of larvae of D. latissimus. Processes like acidification, eutrophication and desiccation will accelerate the succession to dominance of graminoids and Sphagnum mosses (particularly in weakly buffered waters) which could increase the distance between, or even disconnect, suitable egg deposition substrate for D. latissimus (consisting of floating rafts with M. trifoliata and larger Carex sp. species) and optimal Limnephilidae habitat along the shores with overhanging branches of scrubs (Kooijman, 1992;Lamers et al., 2015).
Preliminary feeding experiments showed that Limnephilid species (L. flavicornis and L. stigma) in Booy's Veentje prefer decaying leaves of Betula and Alnus over other food sources, such as algae (unpublished data Bargerveen Foundation). We therefore suggest that shoreline trees could be an important factor in the food chain that supports populations of D. latissimus. If so, Limnephilidae larvae may be threatened in weakly buffered systems in the Netherlands because shrubs and forests on lake shores are often removed to reduce the precipitation interception effect of forests and to reduce the nutrient and organic matter input of trees (Leuven et al., 1987;Wuyts et al., 2008). As a result, an important food supply in the form of leaf litter is removed from the Limnephilidae larvae. Furthermore research however needs to confirm this hypothesis. For instance, it is unclear to what extend the preference for leave litter is applicable to Limnephilidae larvae in other areas. Literature suggests that certain shredding caddisfly larvae may grow on a range of different food types. For example, Hickin (1968) observed that Limnephilus flavicornis, which was the most common species of Limnephilidae in our study, occurred in large numbers in a cattle pond where it apparently fed on algae. In addition, future research should take other factors like water quality into consideration, which could also play an important role in the occurrence of Limnephilidae larvae in D. latissimus habitat.

Competition with other Dytiscid species
Competition with other large species of predaceous diving beetles during the larval stage of D. latissimus probably plays a minor role in the availability of Limnephilidae larvae in the study area of this study. Species that cooccur with D. latissimus in the Netherlands, i.e. Cybister lateralimarginalis, D. marginalis, and D. lapponicus, are generalists that eat various insect larvae and tadpoles (Wesenberg-Lund, 1943; this study). Moreover, larvae of D. latissimus are better able to exploit supplies of Limnephilidae larvae than these other species. D. latissimus larvae were extremely skilled in apprehending a Limnephilidae larva from its case. They attacked and handled cased Limnephilidae larvae in a specific way. After locating a Limnephilidae larva, they would sit on the case of the larva and wait at the case opening for the larva to come out and sometimes encourage it by tapping the rear end of the case with their hind legs. They captured the Limnephilidae larva in its thorax after it came out or in case of instar III larvae, through its case (for more details and illustrations see Johansson & Nilsson, 1992). More than half of the attempted attacks that were observed Ó 2018 The Authors. Insect Conservation and Diversity published by John Wiley & Sons Ltd on behalf of Royal Entomological Society, 11, 565-577 were successful, while D. lapponicus larvae never succeeded to apprehend a cased larva during this study. There are two other species of Dytiscidae that have developed a specialised hunting strategy to apprehend Limnephilidae larvae: D. harrisii and D. semisulcatus (Blunck, 1916;Blunck & Klynstra, 1929;Leclair et al., 1986). Like larvae of D. latissimus, larvae of D. harrisii starved to death when they were offered tadpoles instead of Limnephilidae larvae (Leclair et al., 1986).
Altogether, this study demonstrates that D. latissimus larvae need many Limnephilidae larvae to complete their development into an adult, suggesting that the availability of Limnephilidae larvae is essential for the successful population management of D. latissimus. This result may offer an explanation for the disappearance of populations of D. latissimus in the Netherlands and elsewhere in Europe. A more thorough research of Limnephilidae larvae under field conditions is needed to identify key factors and possible bottle necks in the ecology and lifecycle of Limnephilidae larvae. This will allow a better understanding of habitat requirements, conservation targets and effective measures to improve the conservation and recovery of populations of D. latissimus and their habitat in the Netherlands and Europe.
Identifying the main causes underpinning the decline of an endangered species is essential for its successful recovery. While most research targets adult stages, this study demonstrates that it is important to also consider other life stages as recovery requires insects to complete their entire life cycle. Research on some other threatened insects has reached the same conclusion (e.g. Anthes et al., 2003;Thomas et al., 2009), and together these may guide conservation and research efforts for other vulnerable beetles like Graphodorus bilineatus and Lucanus cervus for which knowledge on their larval ecology is still limited. sion to enter the area and carry out our research. We thank V. Vahru sevs for collecting egg material and providing valuable comments and information on how to handle D. latissimus under laboratory conditions. We thank J. Kuper for identifying the Limnephilidae larvae in our samples. Finally, we are grateful for the useful comments provided by two anonymous referees, which allowed us to significantly improve our manuscript.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Location (meters from shore) and type of coverage of vegetation zones of the south and north transect. Figure S1. The relation between length and weight of larvae of D. latissimus (red; y=0.108 x 2 -0.397 x + 0.413; F 2,39 =1143; p<0.001; R 2 =0.982) and D. lapponicus (blue; y=0.0754 x 2 -0.254 x + 0.245; F 2,28 =685; p<0.001; R 2 =0.979), shown by the dotted lines. Figure S2. Model residuals for oxygen consumption in D. latissimus and D. lapponicus at 10, 15 and 20°C show that oxygen consumption was positively related to temperature. Figure S3. Mean densities of taxonomic orders found in samples taken at several transects in Booy's Veentje on April 5 (n=6), 28 (n=6), May 23 (n=5) and June 13 (n=4).