Research articlesImpacts of artificial light at night on the early life history of two ecosystem engineers
Abstract
Sessile marine invertebrates play a vital role as ecosystem engineers and in benthic–pelagic coupling. Most benthic fauna develop through larval stages and the importance of natural light cycles for larval biology and ecology is long-established. Natural light–dark cycles regulate two of the largest ocean-scale processes that are fundamental to larvae's life cycle: the timing of broadcast spawning for successful fertilization and diel vertical migration for foraging and predator avoidance. Given the reliance on light and the ecological role of larvae, surprisingly little is known about the impacts of artificial light at night (ALAN) on the early life history of habitat-forming species. We quantified ALAN impacts on larval performance (survival, growth, development) of two cosmopolitan ecosystem engineers in temperate marine ecosystems, the mussel Mytilus edulis and the barnacle Austrominius modestus. Higher ALAN irradiance reduced survival in both species (57% and 13%, respectively). ALAN effects on development and growth were small overall, and different between species, time-points and parentage. Our results show that ALAN adversely affects larval survival and reiterates the importance of paternal influence on offspring performance. ALAN impacts on the early life stages of ecosystem engineering species have implications not only for population viability but also the ecological communities that these species support.
This article is part of the theme issue ‘Light pollution in complex ecological systems’.
1. Introduction
Sessile marine invertebrates like corals, bivalves and polychaetes play a vital role as ecosystem engineers and in benthic–pelagic coupling. Ecosystem engineering species modify, maintain and create habitats and thereby drive the availability of resources to other species [1–4]. Sessile marine invertebrates can facilitate recruitment of conspecifics and other species and provide refugia by creating three-dimensional structures that enhance habitat complexity [2–4]. As suspension feeders, they form an important link between primary producers (mainly phytoplankton and bacteria) and upper-level consumers that prey on them, a function that contributes to benthic–pelagic coupling [5]. Most sessile marine invertebrates develop through pelagic and dispersive larval stages [6,7].
Marine larvae possess remarkable sensitivity to low-intensity light and the importance of solar and lunar light cycles for their biology and ecology is long-established [6,8–11]. Light–dark cycles drive large-scale ocean processes including broadcast spawning and diel vertical migration (DVM). Broadcast spawning is a dominant reproductive strategy in marine ecosystems, often synchronized around specific nights within the (annual) lunar cycle [12,13]. Alternatively, internally fertilizing parents release larvae synchronously, often in the night to avoid predation [7]. DVM plays a crucial role in the trade-off between feeding and predator avoidance for planktonic larvae. Commonly described as the largest daily movement of biomass, discrete changes in solar and lunar light cycles drive DVM down to water depths of 300 m, even during the Arctic Winter [11,14]. Larval dispersal and subsequent settlement drive recruitment in ecological networks and communities and shape marine biodiversity [7,15]. Consequently, larval performance (survival, development, growth, settlement) not only regulates the health and distribution of ecosystem engineering populations but also of biogenic reefs and the ecological communities they facilitate.
Surprisingly little is known about the impacts of artificial light at night (ALAN) on larval stages despite their light sensitivity and role in marine ecosystems. The first global atlas of ALAN underwater shows that 1.9 million km2 of the world's coasts are exposed to ALAN [16], many of which host global megacities where ALAN exceeds natural moonlight nearly all year round [17]. Extensive offshore development such as oil and gas platforms, windfarms and island development contribute to ALAN exposure of pelagic ecosystems [16]. Due to predicted expansion of human societies, ALAN is expected to increase exponentially in coastal regions [18]. The global transition towards energy efficient light-emitting diodes (LEDs)—a technology rich in short wavelengths that penetrate deep in the water column—means that ever more marine species and ecosystems are exposed to ALAN [19].
Evidence of ALAN impacts on larvae and early life stages is fragmentary. For example, exposure to artificially lit nights decreases survival in zebrafish Danio rerio embryos even before neural light-detecting structures are present [20]. Larvae in aquaculture settings often grow larger and faster at the expense of increased mortality under ALAN [21–24]. By contrast, pond snail Lymnaea stagnalis hatched under ALAN grew larger but mortality was not altered [25]. Tadpoles of the American toad Anaxyrus americanus showed reduced body mass and growth under ALAN [26] as did fry of the Atlantic salmon Salmo salar [27]. ALAN impacts also differ between developmental stages. While pre-settlement in the barnacle Semibalanus balanoides was not affected by ALAN [28], it reduced settlement of the barnacles Notochthamalus scabrosus and Jehlius cirratus and hence population recruitment [28,29]. The presence of such diverse responses of organisms to ALAN during their early life stages makes prediction beyond the focal species difficult.
To address this major gap in understanding of ALAN impacts, we experimentally quantify larval performance (survival, growth and development) of two cosmopolitan temperate marine ecosystem engineers under ALAN. Both the blue mussel Mytilus edulis and the barnacle Austrominius modestus are known for their light sensitivity during their larvae phase [8,10] and are acknowledged habitat-forming species [2–4,30]. Mussels are also ecologically important as they accumulate significant amounts of calcium and carbon and are of global commercial value [2,31]. While mussels reproduce via broadcast spawning and thus external fertilization [32], barnacles fertilize eggs internally [33]. Austrominius modestus, native to subtropical and temperate Australasia and widespread in the Atlantic since the 1940s, is now considered a naturalized European species [34].
To enhance real-world application of our experiments, we included two of the principal factors affecting larval performance (survival, growth and development): 1) parental influences [33,35] and 2) rearing density as a proxy for food availability or competition [33,36,37]. Our experiments aim to quantify: 1) ALAN impacts on larval survival, growth, and development; 2) whether effects of ALAN are consistent among larvae from different parents, and 3) the interaction between ALAN and rearing density. Finally, our experimental design allows us to test for relationships between larval performance parameters.
2. Methods
(a) Lighting setup
We fitted incubators for rearing larvae with a unique, custom-built lighting system that simulates ALAN (in the form of LEDs), solar and lunar light cycles (for details see [38]). Briefly, daylight was simulated with an Aquaray Natural Daylight Tile set at 5000 lux diffused with 3 mm frosted Perspex to avoid bright spots [19] and powered by the BioLumen Control Unit (Tropical Marine Centre, UK) timed to match natural variation in sunrise and sunset at Plymouth, UK (50°22'34″ N, 4°8' 37″ W). The intensity and timing of natural night-time conditions were simulated using our novel moonlight system, which captures variability in night-time lighting as the moon transits the sky [38]. Moonlight was simulated using a bank of 2700–3500 K 1.2 cd LEDs housed within diffusing spheres. Lunar cycles were simulated using a pulse width modulated signal (scale 0–100%) applied to the 5 V output of Raspberry Pi 3 model B+, with maximum lunar brightness set to 0.5 lux. Lunar brightness was adjusted every minute from a look-up table of Zenith Sky Brightness values modelled for Plymouth, UK, accounting for lunar phase, altitude, opposition, parallax and atmospheric scattering [39]. ALAN was simulated between sunset and sunrise in Plymouth using Aquaray cool white Flexi-LED strips (Tropical Marine Centre, UK), with brightness controlled using voltage dimming. We kept larvae under one of six ALAN treatments: 0, 0.2, 0.5, 1, 10, 50 lux (measured in irradiance as W m2: 0.00, 0.18, 0.47, 2.94, 23.34, 124.80). ALAN brightness (lux) was measured using the SpectroSense2 Lux sensor, irradiance (W m2) using a broadband irradiance sensor (400–750 nm, Skye Instruments, UK).
(b) Experimental design
Larvae of both species were exposed to six light treatments in experimental laboratory systems (see §2a). The lighting setup was the same for both species, but the different reproductive strategies and life cycles of barnacles and mussels meant that we tailored the experimental design for each species (see specific sections below). Austrominius modestus is a hermaphroditic species, undergoes internal fertilization and releases multiple broods of larvae over the spring, summer and autumn period [33,40]. Larvae progress through multiple naupliar stages before reaching the final cyprid stage prior to settling [8]. We collected parents from the field and brought them to the lab to release larvae. Three adults were randomly selected, from which 10 larvae were randomly selected and cultured in 100 ml vials at a replication of four vials per each of the three parents for each of the six ALAN treatments (n = 72).
Mussels reproduce via broadcast spawning. Females release gametes into the water column where fertilization occurs from early spring to late summer [9,32]. Fertilized embryos develop into trochophores and after 72 h progress into veligers (various stages) before settling [10]. Due to the challenges in facilitating spawning in the laboratory [32], we used the sperm from a single male to fertilize eggs from two females. Fertilized eggs were reared in 100 ml vials at two density treatments: high-density (approximately 30 larvae ml−1) and low-density (10 larvae ml−1) [37]. Each of the two density treatments was replicated in four 100 ml vials for each of the two mothers across all six ALAN treatments (n = 96).
In both species larval development varies from two to several weeks, depending on temperature and food supply [9,33,40]. Larvae were reared in incubators set at 15°C to maintain constant temperature. Vials were placed in the centre of the incubator to ensure an even light field across all vials.
(c) Barnacle Austrominius modestus
(i) Adult collection and larval husbandry
Adult barnacles were collected from South Milton Sands, UK (50°15'40″ N, 3°51'33″ W), which is a light naive area [41], in March 2022. Intact barnacles were carefully removed from the substratum using a knife and inspected for egg masses. Adults were transported to the laboratory at the University of Plymouth, where they were placed in individual vials containing 100 ml UV-irradiated, 0.5 µm filtered seawater (FSW) to spawn [33]. After an hour, most parents had released larvae. We selected adults that had released a large proportion of their brood and whose larvae were actively swimming (established by a directional light source). Animals were kept in incubators set at 15°C and reared in 100 ml plastic vials filled to 80ml with FSW [33]. Every 2 days, we changed the water and fed larvae with live diatom Skeletonema marinoi (CCAP 1077/5) culture at a density of 4 × 105 cells ml−1 [33].
(ii) Data collection
Every 2 days, during the water changes, barnacle larvae were counted (alive, dead), checked for development stage (nauplii, cyprid) and pipetted into a clean vial containing fresh FSW [33]. We collected data for five responses: (i) mortality (vial-averaged number of dead individuals scaled up to 100 animals, n = 72), (ii) average developmental time (vial-averaged number of calendar days from freshly released nauplii to cyprid, n = 72) and size, measured as (iii) area (in µm2), (iv) length (in µm) and (v) width (in µm) using the opensource software imageJ (figure 1a–c). While we aimed to measure 10 cyprids for each vial, the naturally staggered development and mortality led to an unbalanced number of replicates for all size measurements and hence observations were averaged per vial (n = 63). The experiment was terminated after 32 days when only four nauplii had failed to develop into cyprids and were deemed unlikely to do so.
Figure 1. Example of larvae size measurements taken in ImageJ software. Carapace length (a), width (b), and area (c) measurements for Austrominius modestus cyprid. Maximum shell length measured parallel to the hinge line on day 60 for Mytilus edulis (d). The black dots inside the larvae are the eyes. Image credit: Fraser Brough.
(d) Blue mussel Mytilus edulis
(i) Adult collection, fertilization and larval husbandry
Adult mussels were supplied by Offshore Shellfish Ltd., a rope culturing mussel farm in Lyme Bay, UK (50°41′59″ N, 2°53′59″ W) with three sites 5 to 10 km offshore and away from the highly ALAN-lit region of Torbay (see ALAN underwater atlas for the North Atlantic by Smyth et al. [16]). Mussels were landed in Brixham Harbour (50°23′39″ N, 3°30′57″ W) in April 2022 and directly transported to the laboratories at the University of Plymouth, UK (50°22′34″ N, 4°8'37″ W). The brood stock of around 100 animals was maintained in a recirculating system supplying four aquaria of ca. 60 l each with FSW, at a salinity of 33/34 ppt, temperature of 10°C and 12 :12hr light : dark cycle (in accordance with seasonal daylight times for Plymouth). Animals were fed every other day with Isochrysis 1800 according to manufacturer's instructions (Reed Mariculture Inc., USA) until fertilization. Fertilization was initiated by thermal shock method to induce gamete release [32,37]. Mussels were exposed to air in a fridge at 3 ± 1°C for up to 48 h before being immersed in FSW in individual glass vials, which were moved between 10°C and 25°C water baths every 45–60 min. Only the most recently released spermatozoa and eggs were used for fertilization. Numbers of released eggs were estimated from five small aliquots counted under a microscope on a Sedgewick Rafter Cell chamber, while sperm concentration was assessed by fixing with Lugol's solution before counting individual cells on a haemocytometer. Sperm from a single male was used to fertilize eggs from two females at a ratio of 200 : 1 in 500 ml cylinders. After 45 min, over half of the eggs were observed to have a polar body, indicating a good fertilization rate [32]. The developing embryos were pipetted into 100 ml culture vials filled to 80 ml with UV-irradiated, 0.5 µm FSW at a density of 200 fertilized eggs cm−2 surface area of the vial and kept under one of the six ALAN treatments in incubators set at 15°C [32]. Embryos were left undisturbed for 72 h to develop from trochophore to the early veliger stage [32]. Once the veliger stage was reached, density was reduced to high-density (approximately 30 larvae ml−1) and low-density (10 larvae ml−1) treatments [37]. Water was changed three times a week by carefully sieving larvae onto a 30 µm mesh. Larvae were returned to a clean vial containing fresh FSW and fed with live Isochrysis galbana (CCAP 927/1) cultures at a density of 5 × 104 cells ml−1 [32].
(iii) Data collection
To record data, larvae were carefully sieved from the vials into a Petri dish for inspection and image-taking under a microscope and afterwards returned to vials with FSW. Four response variables were extracted using imageJ (figure 1d): (i) survival on day 60 (expressed as counts of live individuals per vial per 10 000; n = 96); (ii) length on day 9 (in µm for a subset of 10 individuals per vial; n = 960); (iii) length on day 60 (in µm due to high mortality averaged per vial; n = 77); and (iv) vial-averaged growth (in µm, n = 77) calculated as follows:
Survival of mussel larvae was expressed per 10 000 individuals because of the overall low number of survivors. A lower scale would lead to the loss of important data (many small values rounding to zero).
(e) Statistical analysis
The relationship between each response variable, ALAN and parent (∼iradiance * parent), and for mussels also density (∼irradiance * mother + irradiance * density), was modelled in an analysis of covariance. ALAN was analysed as a continuous response variable (irradiance) and log-transformed to reduce the leverage introduced by high intensity treatments. Parent was modelled as a factor (individual ID). Density was a factor with two treatment levels (low/high). The significance of model parameters was quantified using the Type III ANOVA approach of stepwise parameter removal, starting with the full model and removing interactions before main effects [42]. The significance of all responses was assessed at the 95% confidence level, unless explicitly reported otherwise. Where Gaussian errors were fitted to the response, model fit was visually assessed to comply with an assumption of residual normality. Significant differences between treatment levels of factors (parent, density) and confidence intervals were quantified by pairwise comparisons using the emmeans function (CRAN: emmeans [43]). The predicted relationships of irradiance and their 95% intervals for generalized linear models (GLMs) were modelled using the add_ci function (CRAN: ciTools [44]). The predicted relationship of irradiance and their 95% intervals for (generalized) linear mixed effects models (GLMMs) were modelled using the predictInterval function (CRAN: merTools [45]).
Mortality of barnacles was expressed as a count of the number dead per vial per 100 individuals and analysed using a negative binomial error distribution (dispersion variance/ mean greater than 1). The model was weighted by the number of individual larvae in each vial at the start of the experiment. The vial-averaged developmental time of barnacles was continuous but non-normally distributed, hence it was analysed using the Gamma error distribution. The vial-averaged size of barnacle larvae measured as area and width were normally distributed and modelled using a linear relationship. Length was non-normally distributed, hence analysed using the Gamma error distribution.
Survival of mussel larvae, counts of survivors scaled up to 10 000 animals per vial, was analysed using a negative binomial error distribution (dispersion variance / mean greater than 1). Lengths of mussel larvae after 9 days and 60 days were fitted with Gaussian error distributions. Length at day 9 was analysed using a linear mixed effects model (CRAN: lme4 [46]) with vial ID as a random factor to account for nestedness of 10 random observations per vial. Due to low survival at day 60, observations were averaged over vial and analysed using a linear model. Growth of mussel larvae between day 9 and day 60 was calculated as an average per vial due to low survival on day 60. Growth was analysed using a generalized linear model fitted using a Gaussian error distribution. All models were implemented in the R package, v. 4.2.2.
Note that although barnacle mortality was analysed to better compare responses between barnacle and mussel larvae, the data for both species will be displayed and discussed as survival. This was necessary to account for the different reproductive strategies and life cycles between barnacles and mussels described above and data distribution. The size difference of mussel and barnacle larvae at the beginning of the experiment (i.e. mussel larvae stemmed from eggs that were fertilized in the laboratory whereas barnacle larvae were too small to allow counting of dead individuals) meant that only survival data could be recorded for mussels.
3. Results
(a) Barnacle Austrominius modestus
Increasing ALAN intensity reduced survival of barnacle larvae (figure 2a, χ2 = 5.97, df = 70, p = 0.015). The predicted relationship between survival and ALAN shows a survivorship of 67 individuals per 100 (lower 95% prediction interval 63; upper: 71; numbers extracted from prediction model described above) at no ALAN irradiance (control). However, at the highest irradiance (50 lux or 124.80 W m2), survivorship drops to 58 individuals (lower 95% prediction interval 52; upper 64); a reduction by 13%. Survival differed considerably between offspring from different parents (χ2 = 224.69, d.f. = 68, p < 0.001; see Supplement S1 for additional figures) without showing a significant interaction with irradiance (χ2 = 0.30; d.f. = 66, p = 0.861). The interactive effect of irradiance and parent on the average developmental time from nauplii to cyprid was significant at the 90% confidence level (figure 2b, table 1; F = 3.09, d.f. = 66, p = 0.052). The influence of the average developmental time on survival was modulated by parent (figure 2c, table 1; F = 20.57, d.f. = 66, p < 0.001). In offspring from parent 1, longer development time coincided with lower survival, but in offspring from parent 2 and parent 3 this pattern was reversed. Here larvae with the shorter developmental time showed the lowest survival (figure 2c, table 1). The size of barnacle cyprids was not affected by the interaction between irradiance and parent (area: F = 0.35, d.f. = 57, p = 0.706; length: F = 0.34, d.f. = 57, p = 0.712; width: F = 0.30, d.f. = 57, p = 0.743) or by irradiance (area: F = 0.01, d.f. = 61, p = 0.917; length: F = 0.03, d.f. = 61, p = 0.865; width: F = 0.06, d.f. = 61, p = 0.805). Larvae from different parents varied in size, i.e. in area (F = 3.41, d.f. = 59, p = 0.040), length (F = 3.46, d.f. = 59, p = 0.040) and width (F = 3.20, d.f. = 59, p = 0.048; table 1).
Figure 2. (a) Influence of irradiance on survival of the barnacle Austronimius modestus survival nauplii and cyprid per 100 animals. (b) Influence of irradiance and parent on the average developmental time of barnacle larvae from nauplii to cyprid. (c) Relationship between of the average developmental time and survival of barnacle from different parents. The figures show the raw data (dots are jittered, darker areas represent overlapping data points), predicted relationships (solid lines) and 95% prediction intervals (dotted lines). Note that barnacle data are displayed as survival to allow better comparison between species. Image credit: Harms [40].
| response | predictor | test statistic | residual d.f. | P |
|---|---|---|---|---|
| survivala | irradiance * parent | 0.30 | 66 | 0.861 |
| parent | 224.69 | 68 | <0.001 | |
| irradiance | 5.97 | 70 | 0.015 | |
| average developmental timeb | irradiance * parent | 3.09 | 66 | 0.052 |
| parent | 39.29 | 68 | <0.001 | |
| irradiance | 1.25 | 70 | 0.268 | |
| size as area (in µm2)c | irradiance * parent | 0.35 | 57 | 0.706 |
| parent | 3.41 | 59 | 0.040 | |
| irradiance | 0.01 | 61 | 0.917 | |
| size as length (in µm)b | irradiance * parent | 0.34 | 57 | 0.712 |
| parent | 3.46 | 59 | 0.040 | |
| irradiance | 0.03 | 61 | 0.865 | |
| size as width (in µm)c | irradiance * parent | 0.30 | 57 | 0.743 |
| parent | 3.20 | 59 | 0.048 | |
| irradiance | 0.06 | 61 | 0.805 | |
| survivala | average developmental time * parent | 20.57 | 66 | <0.001 |
| parent | 175.20 | 68 | <0.001 | |
| average developmental time | 45.72 | 70 | <0.001 |
(b) Blue mussel Mytilus edulis
Survival of mussel larvae declined with increasing ALAN irradiance (figure 3a; χ2 = 3.85, d.f. = 94, p < 0. 0.0498). The predicted relationship for mussel larvae survival and ALAN shows that at the highest irradiance (50 lux or 124.80 W m2) mussel survival falls below 50%, i.e. approximately 20 larvae per 10 000 (lower 95% prediction interval: 11; upper: 35) compared to approximately 46 larvae per 10 000 (lower 95% prediction interval: 32; upper: 67) without ALAN. This represents a drop in survival by 57%. Survival was neither affected by the interaction between irradiance and rearing density (χ2 = 0.02, d.f. = 91, p = 0.887), the interaction between irradiance and mother (χ2 = 0.36, d.f. = 90, p = 0.547) or either of the remaining main effects, namely rearing density (χ2 = 0.62, d.f. = 93, p = 0.432) or mother (χ2 = 0.13, d.f. = 92, p = 0.715; see electronic supplementary material, S1 for additional figures). After 9 days of exposure to ALAN, larvae under higher irradiance had grown longer than conspecifics under dimmer conditions (figure 3b; F = 5.28, d.f. = 958, p = 0.022). Length at Day 9 was not influenced by interactions between irradiance and density (F = 0.03, d.f. = 955, p = 0.957), between irradiance and mother (F = 1.11, df = 954, p = 0.292) or by density as a main factor (F = 0.23, d.f. = 957, p = 0.633). The length on Day 9 had no effect on survival (F = 0.07, df = 94, p = 0.800) and neither did mother (F = 1.06, d.f. = 93, p = 0.303) or an interaction between length Day 9 and mother (F = 1.99, d.f = 92, p = 0.158).
Figure 3. Influence of irradiance on (a) the survival of blue mussel Mytilus edulis larvae per 10 000 animals (from fertilization until Day 60) and (b) the length of mussel larvae (measured in µm) on Day 9. The figure shows the raw data (dots are jittered, darker areas represent overlapping data points), predicted relationships (solid lines) and 95% prediction intervals (dotted lines). Image credit: Fraser Brough.
Any effect of ALAN irradiance on the length of mussel larvae was negated after 60 days (F = 0.153, d.f. = 958, p = 0.697) by which point differences in length were quantifiable only between individuals from different mothers (F = 30.85, d.f. = 956, p < 0.001). Similarly, growth was only influenced by mother (F = 16.51, d.f. = 956, p < 0.001). None of the interactions or main effects other than mother influenced the length of mussel larvae on Day 60 and their growth between Day 9 and 60 (table 2 for all test results).
| response | predictor | test statistic | residual Df | P |
|---|---|---|---|---|
| survivala | irradiance * density | 0.02 | 91 | 0.887 |
| irradiance * mother | 0.36 | 90 | 0.547 | |
| density | 0.62 | 93 | 0.432 | |
| mother | 0.13 | 92 | 0.715 | |
| irradiance | 3.85 | 94 | 0.0498 | |
| length at day 9 (in µm)b | irradiance * density | <0.01 | 955 | 0.957 |
| irradiance * mother | 1.11 | 954 | 0.292 | |
| density | 0.23 | 957 | 0.633 | |
| mother | 38.74 | 956 | <0.001 | |
| irradiance | 5.28 | 958 | 0.022 | |
| length at day 60 (in µm)c | irradiance * density | 0.28 | 955 | 0.600 |
| irradiance * mother | 0.78 | 954 | 0.380 | |
| density | <0.01 | 957 | 0.962 | |
| mother | 30.85 | 956 | <0.001 | |
| irradiance | 0.15 | 958 | 0.697 | |
| growth between days 9 & 60 (in µm)c | irradiance * density | 0.40 | 955 | 0.531 |
| irradiance * mother | 0.56 | 954 | 0.458 | |
| density | 0.01 | 957 | 0.938 | |
| Mother | 16.51 | 956 | <0.001 | |
| Irradiance | 0.81 | 958 | 0.365 | |
| survivala | length day 9 * mother | 1.99 | 92 | 0.158 |
| mother | 1.06 | 93 | 0.303 | |
| length day 9 | 0.07 | 94 | 0.800 |
4. Discussion
Demonstrated ALAN impacts on habitat-forming species like marine sessile invertebrates are rare, except for tropical corals [47–49] and three species of barnacles [28,29]. Here, we quantified ALAN impacts on two cosmopolitan, sessile invertebrates from temperate marine ecosystems, the barnacle Austrominius modestus and the mussel Mytilus edulis, by measuring larval performance (survival, growth and developmental time) under an ALAN gradient. Higher ALAN irradiance reduced survival in both species. While barnacle survival dropped by 13% and showed large variation in survival across ALAN intensities, i.e. larger spread of data (figure 2a), survival of mussel larvae dropped by 57% across replicates under the highest tested ALAN irradiance (figure 3a). Direct ALAN impacts on development and growth were largely absent and if present, small. These impacts varied between the two species and between time-points (9 versus 60 days in mussels). Within the first 9 days mussel larvae grew larger at higher ALAN irradiances. This pattern was absent on day 60 and ALAN had no influence on growth between days 9 and 60. There is some indication at the 90% confidence interval that ALAN impacts on developmental time in barnacles vary between offspring from different parents (some developed faster, others slower under increased ALAN intensity). Not surprisingly, we found strong parental influence on nearly all measured responses of larval performance in both species (note that due to challenges in facilitating spawning in the laboratory [32] maternal influences on mussel larvae are based on two mothers). Rearing density as a proxy for food availability did not affect any of the responses measured in mussels (survival, length at day 9, day 60, growth).
ALAN impacts on survival, development and growth in early life stages have been shown across taxa. However, as in our experiments, the patterns are not uniform and can vary between species, within species between life stages and parentage. First, it should be noted that larval survival is generally low in the wild [7] and challenging in laboratory settings [32,37]. Second, extensive aquaculture research on the effects of ALAN across teleost fish commonly shows that larger and faster larval growth under ALAN comes at the expense of reduced survival, such as via malformations followed by mortality [21,22]. Exposure to artificially lit nights decreases survival as early as during embryogenesis, even before an organism possesses neural light-detecting structures [20]. Larvae of the tropical convict surgeonfish Acanthurus triostegus [50] grew larger and survival declined under ALAN. Similarly, crustacean larvae showed shorter developmental time between stages and lower survival [23,24]. Pond snail hatchlings Lymnaea stagnalis also grew larger under ALAN but without compromising survival [25]. Contrary to those ALAN impacts, tadpoles of the American toad Anaxyrus americanus [26], fry of Atlantic salmon Salmo salar [27] and juvenile orange-fin anemonefish Amphiprion chrysopterus monitored in the wild [51] all grew smaller; the latter also experienced higher mortality. ALAN reduced survival in both our species, but as described above, the pattern and percentage reduction differed between species. Under the highest ALAN levels, survival dropped by 57% in mussels and 13% in barnacles, suggesting that high ALAN irradiance could have a stronger effect on mussel compared to barnacle populations.
The two main differences between our species are their reproductive strategy (external versus internal fertilization) and the onset of ALAN exposure in the larvae's life history (embryonic versus larval stage). We fertilized mussel eggs in the laboratory and consequently, animals were under ALAN during embryogenesis. By contrast, we collected adult barnacles with developed larvae from the field where embryogenesis occurred under natural light conditions. ALAN during embryogenesis has been shown not only to lower survival as mentioned earlier [20] but also to diminish embryo quality (yolk area, egg length, eye diameter; [52]). The differences between the two species might also explain why we found different ALAN effects on development and growth between mussels and barnacles. Mussel larvae grew larger with increasing ALAN irradiance by day 9 (a transient effect since we found no effect on day 60). For barnacles, there is some indication that ALAN induced variability in developmental time between larvae from different parents (some developed faster, others slower under increased ALAN intensity). Given that our results are significant at the 90% confidence level, further experiments are needed to examine this pattern. However, the developmental time of barnacles correlated strongly with survival and again, the direction varied with parent. Altered growth rates away from optimal patterns (faster or slower) are known to incur both short- and long-term costs [53]. As growth is energetically costly, larvae that grew faster at the beginning of the experiment may not have survived the weeks after.
Our study validates the well-established pattern of strong parental influence on larval performance under environmental stress such as increased temperature and lower salinity [35,36,54]. Torres et al. [35] show that larval performance (here survival and development) differed between offspring of different egg-carrying mothers and the environmental conditions (temperature and salinity) that they experienced. Similarly, our results suggest that the influence of ALAN might be modulated by parental influence (offspring from some parents cope better than others), which might be driven either by genotypic variation, parental phenotype, its environment or a combination of these. Overall, our results provide two novel directions for further research on ALAN impacts. First, the response to ALAN may depend on the onset of ALAN exposure (embryonic versus larval stage). Second, evidence shows that the parental exposure to ALAN imposes transgenerational effects on human foetuses (reviewed in [55]) and rat offspring [56]. Future research on how parental influence changes ALAN impacts on non-human animals could disentangle the role of genotypic variation, parental phenotype and its environment and give insight into potential selective pressure from ALAN.
Other potential mechanisms of ALAN-induced mortality and changes in growth and development in offspring are manifold. Higher mortality after accelerated growth can be a result of directly incurred physiological costs, a trade-off that has been shown well beyond exposure to ALAN and various forms of environmental stress [53]. During accelerated growth, fewer resources are allocated to energy reserves. Instead, organisms experience a higher metabolic rate, which can be associated with faster production of reactive oxygen species (ROS) and thus oxidative stress reducing lifespan, potentially via higher vulnerability to cellular damage. While ALAN has been shown to increase metabolic demands [57,58], cause overproduction of ROS and increase in oxidative damage from corals [47,48] to vertebrates, including humans [59,60], there is plenty of evidence that ALAN does not necessarily affect organisms via this pathway [61,62]. Alternatives can be endocrine disruptions such as depressed levels of T3, an important hormone for metamorphosis [50]. Animals under ALAN also show reduced melatonin levels [63], which indicates a disruption of rest-like states on the molecular level. ALAN alters the expression of clock genes, which time protein synthesis, DNA repair, cell division and renewal [20,49,64] and impairs organisms' health by affecting the circadian immune system and microbiome [65]. While many studies have quantified ALAN impacts across biological scales within their study system, between-system comparison remains difficult, including our two ecosystem engineers. A recent meta-analysis shows that the direction of physiological and life history trait responses to ALAN can range from strongly negative to positive [66]. The diverse directions of responses to ALAN for survival, growth and development reported in the literature [66] together with our results suggest that further research is needed to consolidate and link patterns and mechanisms.
Here we show that ALAN clearly affects survival, and has differential impacts on development and body size in the larval stage of two marine ecosystem engineers. Impaired larval performance under ALAN is likely to affect not only population health and distribution but also their role as habitat-forming ecosystem engineers. Sessile fauna like our model organisms, the mussel Mytilus edulis and the barnacle Austrominius modestus, support other species by providing three-dimensional habitats [2–4,30]. As filter feeders with a complex life-cycle involving a larval phase, they couple benthic pelagic systems and contribute to energy flows through trophic chains [5] and the accumulation of calcium and carbon [2,31]. While ALAN impacts have been demonstrated at most scales of biological organization [19,66], few studies have been able to quantify ALAN effects on ecosystems and ecological functioning so far [47,49,64,67]. Targeting ecosystem engineering species can form an entry-point to eventually scale-up ALAN impacts to ecosystem level. Tools like biogeographical and predictive modelling can then identify hotspots of susceptible species and habitats at a global scale [15,68], which is yet to be realised for ALAN [19,68]. Given that larval dispersal shapes marine biodiversity and various ecosystem functions including the provision of nutrients for humans, ALAN impacts of the early life stages of ecosystem engineers have clear implications beyond the population level to the ecological communities they facilitate.
Ethics
This work did not require ethical approval from a human subject or animal welfare committee.
Data accessibility
The datasets supporting this article and code will be uploaded as part of the electronic supplementary material [69].
Declaration of AI use
We have not used AI-assisted technologies in creating this article.
Authors' contributions
S.T.: conceptualization, data curation, formal analysis, investigation, methodology, supervision, validation, visualization, writing—original draft; FT.B.: conceptualization, data curation, investigation, methodology, writing—review and editing; L.G.: conceptualization, funding acquisition, methodology, resources, writing—review and editing; S.R.J.: conceptualization, funding acquisition, methodology, writing—review and editing; T.W.D.: conceptualization, formal analysis, funding acquisition, project administration, resources, supervision, writing—review and editing.
All authors gave final approval for publication and agreed to be held accountable for the work performed therein.
Conflict of interest declaration
We declare we have no competing interests.
Funding
The work leading to the publication of this research was funded by the Natural Environment Research Council (grant no. NE/S003533/2 awarded to T.W.D. and NE/S003533/1 awarded to T.W.D., L.G. and S.R.J.).
Acknowledgements
We thank Thomas H. Galley for his advice on rearing Mytilus edulis and Kate Griffith on Austrominius modestus, both at the School of Ocean Sciences, Bangor University. We thank Andrew Grimmer and Cindy Pino, University of Plymouth, for assisting in animal husbandry and data collection.
