Gimme Shelter: differential utilisation and propagule creation of invasive macrophytes by native caddisfly larvae

In aquatic systems, invasive submerged macrophytes considerably alter the structure and functioning of communities, thus potentially compromising ecosystem services. The prolific spread of invasive macrophytes is often aided by vegetative fragment propagation, yet the contributions of various commonly occurring invertebrates to such fragmentation are often unquantified. In the present study, we examine fragmentary spread of invasive macrophytes by a group of shredder-herbivores, larval caddisflies. Through novel application of the comparative functional response (FR; resource use as a function of density) approach to the native case-building species Limnephilus lunatus, we compared utilisation of non-native waterweeds Elodea canadensis and E. nuttallii in mono- and polycultures. Furthermore, we quantified de-cased and cased caddisfly-induced fragment production and length changes among non-native E. canadensis, E. nuttallii, Crassula helmsii and Lagarosiphon major under two different plant orientations: horizontal (floating) versus vertical (upright) growth forms. Larval caddisflies exhibited Type II (hyperbolic) FRs towards both Elodea species, and utilised each plant at similar rates when plants were provided separately. When plant species were presented in combination horizontally, E. canadensis was significantly less utilised compared to E. nuttallii, corroborating observations in the field. De-cased larvae produced new plant fragments for all four aquatic macrophytes, whereas cased larvae fragmented plants significantly less. Elodea nuttalii and C. helmsii were fragmented the most overall. Crassula helmsii was utilised to the greatest extent when plants were horizontally orientated, and Elodea species when vertically orientated. This study identifies and quantifies a mechanism from a novel species group that may contribute to the spread of invasive macrophytes in aquatic systems. Whilst exploititative interactions are thought to impede invasion success, here we demonstrate how resource utilisation by a resident species may exacerbate propagule pressure from an invasive species.


Introduction
Freshwater ecosystems are increasingly invaded by alien species introduced accidentally or deliberately (Dudgeon et al. 2006;Ricciardi 2006;Seebens et al. 2017). Once established, alien species can impact biodiversity and alter key ecosystem functions such as productivity, nutrient cycling and hydrology (Dudgeon et al. 2006;Oreska and Aldridge 2011;Piria et al. 2017;Crane et al. 2020). For example, invasive macrophytes can form dense monotypic stands that alter physical habitat and biotic (vegetation, macroinvertebrates and fish) communities, as well as the interactions within and between these communities (Dibble et al. 1996). Invasive aquatic macrophytes are highly adaptable and competitive, and can, inter alia, grow rapidly, displace native plants and damage ecosystem services (Barrat-Segretain and Elger 2004;Redekop et al. 2016;Hussner et al. 2017;Szabó et al. 2018).
Many aquatic macrophytes reproduce asexually, propagating predominantly by vegetative rhizomes, turions or fragments (Barrat-Segretain 1996; Barrat-Segretain et al. 1998). Further, in response to abiotic factors such as wind, waves and water currents, human activities such as boating and fishing, as well as biotic factors such as herbivory, aquatic macrophytes are frequently broken into fragments. The fragments can disperse as propagules which later become new viable plants (Hussner 2009;Redekop et al. 2016;Coughlan et al. 2018). Production of macrophyte fragments by shredder-herbivores may be significant (Newman 1991;Pieczynska 2003;Maezo et al. 2010), and therefore when such consumers are present -particularly in high abundances -they could have serious implications for invasive macrophyte spread in terms of increasing propagule pressure (i.e. the number and frequency of individuals introduced to form an invader population; Lockwood et al. 2005). However, there is a distinct lack of quantitative data for utilisation or fragmentation rates of native/invasive aquatic macrophytes by shredder-herbivores (but see e.g. Carriera et al. 2014; Thouvenot et al. 2017).
Resource-use patterns of consumers can be quantified by measuring their functional response (FR), which describes the relationship between resource utilisation rate and resource availability (Solomon 1949;Holling 1959). Functional responses and associated resource preferences/switching have been identified as powerful tools to quantify invasive species impacts and invasion success (Dick et al. 2014;Cuthbert et al. 2018b;Skein et al. 2018), whereby Type II curves are thought to be resource destabilising due to high resource utilisation rates at low densities, whilst the converse is true for Type III FRs. However, whilst FRs are commonly applied to address consumer-resource interactions such as predation and herbivory (Dick et al. 2014;Xu et al. 2016a, b;Mu et al. 2018;Cuthbert et al. 2018a), there has hitherto been a lack of consideration for such per capita effects in shredder-herbivores, especially with regards to atypical resource utilisation behaviours, such as caddisfly case-building.
In consumer-resource (e.g. predator-prey) systems, invasion success is theoretically likely, and thus predictable, if invaders encounter lower biotic resistance compared to trophically analogous natives (see also the 'enemy release' hypothesis; Levine et al. 2004;Cuthbert et al. 2018b). Conversely, in the context of interactions involving invasive/native macrophytes, differential utilisation in favour of invaders may promote higher invasion success through greater propagation as a direct consequence of utilisation. Whilst stronger interactions towards invasive species over natives by grazers may contribute to biotic resistance (Oliveira et al. 2018), herbivore presence might enhance the fragmentation rate of invasive macrophytes that spread by vegetative propagules through direct and/or indirect effects (Thouvenot et al. 2019). For example, positive associations between invasive crayfish and macrophytes can result in reciprocal facilitations that heighten invasion dynamics (Thouvenot et al. 2017). Herbivore-plant interactions may also be mediated by plant traits, with characteristics such as nutritional properties, physical structure and secondary metabolites altering plant palatability (Hay 1996;Cronin et al. 2002;Elger and Lemoine 2005). In the case of resident herbivores, taxa that shred aquatic macrophytes (e.g. caddisfly larvae) may be important facilitative drivers of invasion via enhancing fragmentation and thus vegetative propagule pressure of aquatic plant species. Field observations of caddisfly larvae residing within swards of invasive Elodea species year-round have prompted us to assess the consumer-resource relationship and facilitated propagule creation.
Caddisflies (Trichoptera) are a group of insects that are widespread and abundant, and whose aquatic larvae comprise diverse functional feeding groups (Pescador et al. 1995;O'Connor 2015). Herbivorous case-building caddisfly larvae, such as Limnephilus spp., feed predominantly by shredding (Hanna 1959;Wiggins 2007), and are known to include submerged macrophytes in their diet (Jacobsen and Sand-Jensen 1994). The larvae of many caddisflies construct protective shelters (cases) using secreted silk to bind together particles that may include shells, mineral particles and plant parts (leaves and stems) (Gower 1967;Mouro et al. 2016). A new case is constructed at each larval instar stage, of which there are typically five during a life-cycle that spans up to 2 years, with case size increasing with each instar (Hanna 1959). Accordingly, dietary consumption coupled with case creation may drive high utilisation rates towards aquatic plants, year-round over an extensive larval life history. Given that the larvae recurrently create and augment cases across their larval life history, and feed on plant material, we hypothesise that they can play a significant role in fragmenting, and thus facilitating the spread of, invasive macrophytes that can disperse and establish vegetatively. The focal species, Limnephilus lunatus Curtis, is distributed widely throughout Europe and North America in both lenthic and lotic habitats, associated with plants, and forms an important component of aquatic food webs (Higler 1980).
In this study, we examine caddisfly usage of four non-native macrophytes that can reproduce vegetatively via fragments and, in many instances, have a detrimental impact on the receiving environment (see Table 1): Elodea canadensis Michx., Elodea nuttallii (Planch.) H. St. John, Crassula helmsii (Kirk) Cockayne and Lagarosiphon major (Ridl.) Moss. Elodea canadensis, Canadian waterweed, is a non-native submerged macrophyte that spread rapidly following its introduction to the United Kingdom (UK) in the early 1800s and is now naturalised in the UK and Ireland with relatively benign impacts (Newman and Duenas 2010). Elodea nuttallii. Nuttall's pondweed, is an invasive submerged macrophyte that can form dense monocultures that can displace E. canadensis and produce adverse ecological impacts (Simpson 1990;Barrat-Segretain 2001;Larson 2007). Crassula helmsii, New Zealand pigmyweed, is an invasive submerged, emergent or semi-terrestrial macrophyte, depending on the conditions into which it is introduced. It can form dense, monotypic stands, is extremely difficult to control (OEPP/EPPO 2007), and is considered a major invasive threat to the UK (Dawson and Warman 1987;Dawson 1994;Dawson and Leach 1999;Huckle 2002). Lagarosiphon major, African elodea, is a submerged macrophyte that grows in dense mats, is exceptionally difficult to control, and considered highly invasive (Caffrey et al. 2010(Caffrey et al. , 2011. These plants were selected owing to their widespread establishment in the UK and Ireland, local availability and potential for coexistence in inland waters. We used laboratory-based experiments to test the predictions that larval caddisflies of L. lunatus: (1) exhibit a preferential utilisation between the two Elodea species that may contribute to invasion success and replacement of E. canadensis by E. nuttallii; and (2) can cause increased fragmentation to a range of invasive macrophytes: E. canadensis, E. nuttallii, C. helmsii and L. major, and hence positively influence propagule pressure and invasion success. Here, for the first time, the FR concept is thus applied to a native herbivorous invertebrate-the caddisfly, L. lunatususing as a resource two alien macrophytes, E. canadensis and E. nuttallii. Furthermore, we quantify the capacity of this species to promote fragmentation in these and other macrophytes.

Organism collection and plant preparation
Elodea canadensis, E. nuttallii, C. helmsii and L. major were collected throughout Northern Ireland from a variety of lakes and ponds (Table 1). Pilot studies indicated that caddisfly larvae would actively utilise each species when provided singularly. Stems of each species were cut at the base, placed in coolers, and transported in source water to Queen's University Marine Laboratory, Portaferry, Northern Ireland. Prior to use, plants from each species were maintained in separate outdoor aquaria for approximately three months. Plants were then maintained within the laboratory at a constant temperature of 12 ± 1°C with aerated source water in 2 L arenas (L 9 W 9 H: 34 9 34 9 14 cm). Larvae of L. lunatus were collected from Lough Erne (54°12 0 02.9 00 N 7°29 0 35.8 00 W) by hand and identified as per Wallace et al. (2003). In Lough Erne, this species is associated with swards of both E. canadensis and E. nuttallii (K. Crane personal observation). Individuals were maintained in the same laboratory as the plants within aerated aquaria, filled with locally sourced lake water (Lough Cowey: 54°24 0 41.79 00 N 5°32 0 25.96 00 W) and Elodea spp. ad libitum, and starved for 48 h prior to experimentation. Case removal was carried out by widening the posterior opening using dissecting forceps, then gently pushing the caddis out using closed, rubber-tipped forceps. No caddisflies were found to be damaged through this process, and thus their ability to fragment plants was not impeded. Photon Flux Density was supplied by four 52 W Arcadia 1200 mm Marine Stretch LED lamps so that plants received 270 lmol m -2 s -1 under a 16 h light and 8 h darkness regime. The conditions aligned with those at the collection sites, and were relevant to the time of year when organisms were sampled. All plants were acclimated to the laboratory conditions for a 48 h period prior to experimentation in Lough Cowey water, during which all species exhibited excellent health and very little necrosis. All waste invasive plant material was destroyed by autoclaving.
Apical plant fragments were harvested 16 h prior to the start of each experiment and washed in dechlorinated tap water to remove any debris. Where possible, fragments were cut from unbranched sections of stem. However, if present, axillary side shoots were carefully removed. Fragments were briefly maintained (\ 30 min) in de-chlorinated tap-water immediately prior to being measured or weighed for experimental use (see next).

Experiment 1: Limnephilus lunatus case-building functional responses
Elodea canadensis and E. nuttallii fragments were randomly selected from holding aquaria (see before) and excess liquid was gently removed by manually spinning individual fragments, ten times in both directions, within a handheld centrifugal spinner. Individual apical fragments were cut to an exact fresh mass of either 100 mg, 200 mg, 300 mg or 400 mg (i.e. four different 'supply' treatment levels) (Mettler Toledo AB104). To quantify FRs of caddisfly larvae, individuals were presented with increasing plant biomass (similar to increased prey numbers, see Xu et al. 2016a, b). Two plant fragments of either single-(E. canadensis or E. nuttallii) or mixed-species (one fragment of the same mass from each species) were placed in each container (800 ml plastic containers, L 9 W 9 H: 170 9 110 9 60 mm, with 400 ml water from Lough Cowey) within the laboratory (conditions as before). Individual de-cased or cased fifth-instar L. lunatus larvae were weighed (mean ± SE: de-cased, 0.14 ± 0.03 g; cased, 0.59 ± 0.10 g) and one added to each treatment to quantify usage by plant species and caddisfly larvae types (i.e. de-cased, cased). Preference for one plant species over the other was then recorded using comparative functional response analyses (see later). Controls consisted of three replicates of each plant combination and mass without L. lunatus. All experimental groups were replicated three times. The experiment lasted 48 h, after which wet masses of the original two plant fronds, provided at the start of the experiment, were quantified as before (i.e. accounting for utilisation via both case creation and consumption). We thus considered ''utlisation'' broadly in functional response analyses, whereby biomass changes resulting from direct consumption and case creation were pooled. The before-after differences in plant masses were used in analysis as caddisfly larvae-free control plants did not exhibit mass changes. Where two strands of individual species were used, the mean mass of both was used as a data point, whilst the single mass of each species was used for mixed species treatments. We then determined the proportion of plant mass utilised relative to the initial fresh mass supplied. Caddisfly larvae were weighed (wet mass) before and after the experimental period.

Experiment 2: Limnephilus lunatus plant fragmentation
Plants were randomly selected from holding aquaria (see before) and four apical fragments were cut to a standard length of 60 mm (i.e. total length per replicate 240 mm). Four fragments of a single species (wet mass ± SE: E. canadensis, 791 ± 35 mg; E. nuttallii, 574 ± 25 mg; C. helmsii, 380 ± 19 mg; L. major, 2268 ± 151 mg) were placed in 800 ml plastic containers with 400 ml lake water as before. Plants were either placed horizontally in containers or vertically as a bunch, weighted together at the base.
Orientation was varied to examine whether L. lunatus would more likely shred the plant when floating or sunk in mats, or upright whilst growing and rooted in the substrate. Where vertically presented, the base of each individual fragment of plant was protected using a small piece of cotton wool before being wrapped with a 60 9 5 mm lead plant weight, to keep the base of the fragment at the bottom of the container and the apical section positioned vertically. Individual decased or cased fourth-instar L. lunatus larvae were weighed (mean ± SE: de-cased, 0.11 ± 0.04 g; cased, 0.32 ± 0.1 g) and one added to each treatment. All experimental groups were replicated three times. Controls contained one replicate of each plant species in the absence of caddisfly larvae under both orientations. The experiment was run over 168 h (one week) to allow L. lunatus sufficient time to shred plants, construct new cases or supplement existing ones. After the experiment, new plant fragments were recorded and the lengths of the initial four fragments were measured in combination within each replicate (i.e. accounting for both fragmentation and consumption). Final differences in lengths were then considered against final lengths of caddisfly larvae-free controls within each replicate (i.e. for each species and orientation). Only stem fragmentation was recorded; the removal of leaves did not constitute a viable fragment, as there is no evidence that the focal macrophytes can propagate from leaves alone. Caddisfly larvae were weighed before and after the experimental period as before.

Statistical analyses
All statistical analyses were performed using R v3.5.1 (R Core Development Team 2018). In Experiment 1, plant utilisation (response variable: proportion of initial plant mass used) under single-species exposures was analysed with respect to 'plant species' (2 levels: E. canadensis; E. nuttallii), 'caddisfly larvae type' (2 levels: de-cased, cased) and 'plant supply' (4 levels: 100 mg, 200 mg, 300 mg, 400 mg) using linear models (LMs). For mixed plant exposures, linear mixed models (LMMs) with a random effect structure to account for repeated measures of the two plant species within each experimental replicate at a single time points were used (Bates et al. 2015). Proportioned plant utilisation was arcsine square-root transformed prior to analyses to improve normality and variance homogeneity (tested using Shapiro-Wilks and Levene's tests, respectively). Similarly, differences in de-cased caddisfly masses (response variable: change in mass) were examined using LMs separately for both single and mixed exposures, according to plant species and supply, or supply alone, respectively. Cased caddisfly masses did not change after the experiment and thus were not considered.
In Experiment 1, owing to negligible utilisation by cased caddisfly larvae in specific groups, comparative FR modelling was only feasible for de-cased caddisfly treatments. The 'frair' package in R (Pritchard et al. 2017) was used to analyse de-cased larval caddisfly FRs using maximum likelihood estimation (Bolker 2010) and the Lambert W function (Bolker 2008). For each plant species separately, we deciphered FR types through logistic regression of the proportion of plant utilised (response variable: fragment mass change relative to original mass) as a function of the original mass supplied (continuous predictor). Here, Type II FRs were defined through a significantly negative linear coefficient (Juliano 2001).
We fit Rogers' random predator equation to account for non-replacement of plant material by the experimenter (Rogers 1972): where N e is the quantity of plant utilised, N 0 is the initial plant mass, a is the attack (i.e. shredding) rate, h is the handling time and T is the total time available. In a herbivory context, handling times may be considered as the time taken to utilise the plant resource. Regardless, the present study does not consider the attack rate and handling time parameters mechanistically, but instead considers them for comparative purposes in a factorial experimental design (Alexander et al. 2012). The attack rate and handling time are both central parameters of FR curves, with the attack rate corresponding to the original slope (i.e. search coefficient) and handling time corresponding to the asymptote (i.e. maximum feeding rate). Indicator variables were used to compare FR parameters between E. canadensis and E. nuttallii within single and mixed groupings (Juliano 2001). This approach compares FR parameters between groups via substitution of the a/h estimate from Eq. 1 plus a coded predictor for the focal variable (see Paton et al. 2019). We used a non-parametric bootstrapping procedure (n = 2000 iterations) to generate 95% confidence intervals around FR curves. In Experiment 2, raw counts of new fragments (response variable: number of new fragments) were analysed using negative binomial GLMs with log links. Fragment length changes relative to caddisfly larvae-free controls (response variable: fragment length change) of plants over the experimental period were analysed using LMs. For each of these models, 'plant species' (4 levels: E. canadensis, E. nuttallii, C. helmsii, L. major), 'plant orientation' (2 levels: horizontal, vertical) and 'caddisfly larvae type' (2 levels: de-cased, cased) were incorporated factorially. Yeo-Johnson transformations were used on plant length changes to homogenise variances and normalise residuals prior to analysis (Fox and Weisberg 2011). As before, de-cased caddisfly larvae masses were examined using LMs (response variable: change in mass) as a function of plant species and orientation, as cased caddisfly larve masses did not change over the experiment.
For all models, non-significant terms were removed stepwise to obtain the most parsimonious fit, and thus the final models included only significant terms. Tukey's comparisons were used for top model post hoc tests where terms were found to be significant at the 95% confidence interval via analysis of deviance. Effect sizes were derived through F-tests for LMs and LMMs and Chi square-tests for GLMs (Fox and Weisberg 2011). Owing to our sample size, the power to detect significance was low, and thus our results may be viewed as conservative in some instances.

Experiment 1: Limnephilus lunatus case-building functional responses
In L. lunatus-free controls, 100% of the original mass of both Elodea species remained at the end of the experiment. Thus, experimental reductions in plant mass were deemed a result of utilisation by larval caddisflies, which was also evidenced by case creation/augmentation. In the final model, de-cased caddis utilised significantly more plant material than cased individuals, and utilisation was further mediated by plant supply. These terms also interacted, with utilisation rates decreasing with supply only in the case of de-cased caddis (Table 2a; Fig. 1a, b). Cased caddisly larvae always utilised significantly less than de-cased individuals (all p \ 0.001). We did not detect significance in differences between plant species or for other terms. Under mixed-species exposures (i.e. both plants together), however, significantly more E. nuttallii was utilised as compared to E. canadensis, whilst de-cased caddisfly larvae again utilised significantly more than cased equivalents (Table 2b; Fig. 1c, d). All other terms had no detected significance and were removed from the final model. De-cased caddisfly larvae masses did not change significantly between plants when presented singularly (F 1, 22 = 0.242, p = 0.628), irrespective of supply (F 3, 22 = 0.445, p = 0.724), with caddis gaining a mean of 52 mg (± 7.5 SE) in mass over the experiment. Similarly, in the mixed treatments, caddisfly larvae masses did not change significantly across supplies (F 3, 8 = 2.018, p = 0.190), with mean gains of 60 mg (± 12.7 SE).
Type II case-building FRs were exhibited by decased caddisfly larvae towards both plant species, either when presented individually or simultaneously, owing to significantly negative linear coefficients (Table 3). When both E. canadensis and E. nuttallii were presented separately, we did not detect attack Table 2 Model terms, degrees of freedom (df) and p values for (a) linear models (LMs) on proportioned plant utilisation under singlespecies exposures as a function of plant species (P, 2 levels), caddisfly larvae type (C, 2 levels) and plant supply (S, 4 levels); (b) linear mixed models (LMMs) on proportioned plant utilisation under mixed-species exposures as a function of plant species (P, 2 levels), caddisfly larvae type (C, 2 levels) and plant supply (S, 4 levels); (c) GLMs on fragmentation counts as a function of plant species (P, 4 levels), plant orientation (O, 2 levels) and caddisfly larvae type (C, 2 levels); (d) LMs on fragment length changes as a function of plant species (P, 4 levels), plant orientation (O, 2 levels) and caddisfly larvae type (C, 2 levels) Significant p values are emboldened rates and handling times (and hence maximum utilisation rates) to differ significantly towards the two plants (attack rate, z = 0.862, p = 0.389; handling time, z = 1.712, p = 0.087). Accordingly, confidence intervals overlapped across all plant supplies (Fig. 2a) (Fig. 3a, b). On the other hand, no new fragments were created in caddisfly larvae-free controls. Plant species alone significantly influenced the quantity of fragments produced (Table 2c). In total, 19 fragments were created in the case of C. helmsii, whilst 6, 5 and 3 were produced for E. nuttallii, E. canadensis and L. major, respectively. Overall, C. helmsii was significantly most susceptible to fragmentation (Fig. 3a, b). Cased caddis did not fragment E. canadensis or L. major, whilst de-cased caddis fragmented all plant species (Fig. 3a, b). Significance was not detected for all other terms and these were thus removed. For total fragment length changes of the original fragments, a significant three-way interaction among plant species, orientation and caddisfly larvae was exhibited (Table 2d). As such, length changes were differentially mediated by orientation among plant species, depending on whether caddis were de-cased or cased (Fig. 4a, b). Specifically, for de-cased caddisfly larvae, macrophyte length differences between species were not statisticaly clear irrespective of plant orientation (all p [ 0.05). For cased caddisfly larvae, however, vertical C. helmsii grew significantly compared to all other plants (all p \ 0.01), that in turn generally exhibited similar length reductions following treatment (all p [ 0.05). Overall, regardless of orientation, E. nuttallii and C. helmsii trended towards greatest length reductions where caddisfly larvae were de-cased, whilst L. major were reduced most by cased caddisfly larvae (Fig. 4). De-cased caddisfly larvae mass gains did not change significantly according to plant orientation (F 1, 20 = 0.947, p = 0.343), yet differed according to plant treatment (F 1, 20 = 13.377, p \ 0.001), with larvae always significantly heavier following treatment with L. major (all p \ 0.01) (mean mg ± SE: L. major, 145.0 ± 10.9; E. canadensis, 80.0 ± 11.6; E. nuttallii, 58.3 ± 3.1; C. helmsii, 80.0 ± 12.7).

Discussion
Understanding mechanisms that allow introduced species to establish and spread is vital for management strategies and assessment of invasion risk (Flemming and Dibble 2015). For aquatic macrophytes, interactions with resident consumers may be a major determinant of invasion success. For example, Parker and Hay (2005) found that, in the receiving environment, native herbivores preferred non-native plants across taxonomic groupings that included crayfish  Redekop et al. 2018). However, the capacity of herbivore-shredders to enhance the invasiveness of alien macrophyte species through indirect effects has rarely been considered (Thouvenot et al. 2017(Thouvenot et al. , 2019. The results of the FR experiment showed that there was little difference in utilisation where the non-native E. canadensis and high-impact invader E. nuttallii were presented individually. However, when offered simultaneously, L. lunatus preferentially used E. nuttallii over E. canadensis. Palatability has been shown to differ among Elodea species, and change seasonally by being higher during faster growth phases (Barrat-Segretain et al. 2002). Given that E. nuttallii allocates more resources to rapid growth than E. canadensis (Eugelink 1998), it may allocate fewer resources to deter herbivory (Barrat-Segretain et al. 2002). In turn, this may enhance invasion success by exacerbating fragmentation rates by herbivores. Resource use by cased L. lunatus larvae was low and, under most masses supplied, E. nuttallii was consistently selected whilst E. canadensis was never (b) (a) used. Nevertheless, de-cased caddis exhibited a Type II FR towards both plant species, under both single and mixed species groups. Therefore, proportional resource use by de-cased caddisfly larvae was high at low plant masses supplied, suggesting that these plants will be utilised by de-cased larvae at a high rate when relatively rare in aquatic environments. Accordingly, even when plant abundances are low, feeding caddisfly larvae could affect dispersal by fragmentation via utilisation of the plants. Increased production of plant propagules associated with caddis shredding could thus generate a higher rate of range expansion than the natural diffusion from a single large population alone (Hengeveld 1989). Therefore, we suggest that herbivorous caddisflies are not an effective natural enemy of these plants. Rather, in cases where plant dispersal is otherwise strongly limited, the relationship with caddisflies could be described as a mutualistic one. In our study system, consistent preferential utilisation could promote better spread of the high impact invader E. nuttallii over the relatively benign non-native E. canadensis. However, further research which examines additional combinations of macrophytes, as well as fragment viability and competition, is required to elucidate consumptive dynamics and preferences in freshwaters.
Where macrophyte species were presented singly, attack rates and handling times (and hence maximum utilisation rates) by de-cased L. lunatus were similar towards the two Elodea species. However, in mixed macrophyte combinations, whilst attack rates remained similar, handling times were significantly shorter and thus maximum utilisation rates higher, towards the invasive E. nuttallii. We consider these parameters comparatively rather than mechanistically, as they were not validated using empirical measurements (Alexander et al. 2012). The greater selective utilisation of E. nuttallii may be driven by the presence of longer, more malleable leaves, which could be easier to manipulate than the shorter, wider E. canadensis. Mechanistically, the implications of these results are not only that L. lunatus preferentially uses E. nuttallii, but that such shredders are also likely fragmenting it and enhancing its propagation. Displacement of E. canadensis by E. nuttallii can occur over a relatively short period of one to two years (Simpson 1990). As such, shredders that selectively fragment E. nuttallii could help drive or accelerate this shift in macrophyte community composition, yet also in combination with other factors such as plant resource use and growth rates.
Limnephilus lunatus larvae induced fragmentation in E. canadensis, E. nuttallii, C. helmsii and L. major, creating new propagules from the original fragments. Unsurprisingly, de-cased caddisflies produced substantially more fragments of all species than cased individuals owing to active case construction. Indeed, cased caddisflies never fragmented E. canadensis or L. major. Further, excepting C. helmsii, abilities of caddisfly larvae to fragment plants were mostly unaffected by the orientation of plants in our study. The majority of aquatic plants are capable of producing highly viable fragments, yet the minimum fragment size required for successful regeneration differs among species (Hussner 2009). Notably, many aquatic plant fragments, including some of those examined here, can regenerate from stem fragments comprising of a single node alone (Hussner 2009) and from sizes as small as 1-5 cm (Hussner 2009;Coughlan et al., 2018). Whilst new fragment sizes were not recorded in our study, it is likely that any new fragment produced which included at least one node could be viable. Future studies should thus consider both fragment size, and also the amount of nodes present both within and between species, for assessing regeneration likelihoods. In waterbodies, fragmentation instigated at the base of macrophyte swards by benthic-dwelling caddisfly larvae could release larger propagules in the form of long plant stems, e.g. 60-100 cm. Moreover, larval L. lunatus have been recorded in abundances as high as 160 individuals m -2 in freshwaters (Liess and Schulz 2009), and given that a new case is created for each of five instar stages, this could result in substantial propagule pressure should fragmentation occur, in addition to fragments created by consumption more generally. However, whether net population-level effects (i.e., via different consumer numerical responses or abundances) of herbivory by caddisfly larvae are facilitative or antagonistic for invasive macrophytes should be further examined in light of minimum viable propagule sizes and the sections of macrophyte swards that are preferentially grazed. In turn, whether multiple conspecific caddisfly larvae effects combine additively, antagonistically or synergistically in terms of fragmentation rates requires testing.
Successful management of invasive aquatic plants requires sound knowledge of their biology and, in this case specifically, their ability to regenerate following fragmentation. Since all of the examined macrophytes are spread by fragments, shredding could potentially have an impact on the spread of many invasive plants. Of the focal species, C. helmsii has exhibited a particularly high regeneration capacity and has the ability to form new shoots from single nodes (Hussner 2009). Bearing in mind this species was the most fragmented of all the examined macrophytes, coupled with its propensity to invade and become abundant, it especially appears to be facilitated by de-cased caddisfly larvae. Palatability in aquatic plants is linked to a range of traits including nutritional qualities, physical structure and secondary metabolites (Hay 1996;Cronin et al. 2002;Elger and Lemoine 2005). Zhang et al. (2019) found plant palatability to be negatively related to dry matter content, carbon:nutrient ratios and total phenolics, yet positively related to phosphorus and nitrogen levels in individual species. Moreover, herbivory has been found to positively relate to calcium and lignin contents (Bonar et al. 1990), however the extent which chemical properties influence plant selection by caddisfly larvae requires further investigation. As such, factors such as lignin content within and between plants, as well as physical deterrent effects, should also be considered in the context of caddisfly larval case creation. Shredders in aquatic systems may also be affected by seasonality (Boiché et al. 2010), which could reflect different in palatabilities owing to temperature variation. Moreover, in natural systems, fragmentation by caddisfly larvae of larger fronds may be coupled with hydrodynamics and influenced by stream size (Heidbüchel et al. 2019). Further work should consider these biotic and abiotic factors comparatively in order to better understand the drivers of differences in utilisation and fragmentation among key aquatic plants.

Conclusion
We have identified a potential mechanism that promotes the differential propagation of invasive macrophytes, depending on the plant-specific fragmentation properties. Larvae of L. lunatus showed a preference for invasive E. nuttallii over naturalised E. canadensis, and de-cased caddis fly larvae generated fragments from all of the examined macrophytes, but to different degrees. On the other hand, cased L. lunatus fragment macrophytes less between casebuilding episodes. Accordingly, large caddisfly larvae populations in freshwater systems infested with these macrophyte species might facilitate their further spread owing to their shredding activities. Our study identifies and quantifies a mechanism of propagule pressure from a novel group of species, by which resident herbivores potentially facilitate rather than limit invasion. The viability of the resulting propagules in the field remains to be determined.
Author contributions KC and AR proposed the study. KC, RNC and NEC designed the experiments. KC conducted the experiments. RNC performed data analysis. All authors contributed considerably to writing the manuscript, which was led by RNC.
Funding Open Access funding provided by Projekt DEAL.
Data availability Underlying raw data will be made available in the online supporting information.

Compliance with ethical standards
Conflict of interest The authors declare no conflicts of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.