Multiple spectral channels in branchiopods. II. Role in light-dependent behavior and natural light environments

In deep-sea aquatic habitats, attenuated ambient light is known to shape the spectral information received by the visual systems of its inhabitants (Frank et al., 2012; Lythgoe, 1984; Lythgoe and Partridge, 1989). Far less is known in comparison about the extent of light attenuation in shallow freshwater environments, or how the behaviors of temporary freshwater pool inhabitants might be configured for stereotyped behavioral responses when presented with light of differing wavelengths or intensities (Cohen and Forward, 2009). Branchiopod crustaceans are frequent inhabitants of shallow freshwater habitats ranging from highly light-transmissive lakes to heavily occluded inland waters (Dumont and Negrea, 2002; Williams, 2005). Fossils from the Devonian suggest that ancestral marine branchiopods colonized temporary freshwater habitats, and that this transition is responsible for many of the shared derived characteristics of extant branchiopods (Dumont and Negrea, 2002; Gueriau et al., 2016; Olesen, 2007). Ephemeral pools are types of temporary freshwater habitats that have periods in which they are fully desiccated (Williams, 2005). Many species of branchiopods are found in such habitats, as there are typically no fish to predate them (Brendonck et al., 2008; Dumont and Negrea, 2002; Gueriau et al., 2016). As a result, branchiopod crustaceans are passive dispersers, and many species have resting eggs that can withstand complete desiccation (Brendonck, 1996).

Branchiopods have evolved characteristics such as resting eggs that are clearly adaptations for temporary freshwater habitats. But they also have optic lobe neuroanatomy that is simpler in comparison to other pancrustaceans, a group of arthropods that includes hexapods and crustaceans (Ma et al., 2012). The structure of the optic lobe of branchiopods could be adaptive because of the costs of maintaining unnecessary neuroanatomical structures. In other pancrustaceans, three nested visual neuropils function as a series of information processing centers (Wolff and Strausfeld, 2016). Color information in other pancrustaceans is thought to be processed in two of their neuropils, the medulla and the lobula (Dyer et al., 2011; Kleinlogel et al., 2003; Melnattur et al., 2014; Morante and Desplan, 2008; Paulk et al., 2008, 2009; Strausfeld, 2012). The structural properties of the visual neuropils that have been identified in branchiopods to date, which do not include a medulla or a lobula, therefore suggest a lesser ability to process color information. In a companion paper, we have clarified how the structure of the lamina, the first optic neuropil, is configured (Lessios et al., 2018). Here, we hypothesize that light attenuation and the spectral content available in ephemeral pools act as a selective pressure for light-dependent behavior that is determined by a simplified set of optic neuropils. We describe light-dependent behavior and attenuation in natural habitats of the anostracan Streptocephalus mackini Moore 1966 and the notostracan Triops longicaudatus (LeConte 1846).

The biotic and abiotic factors that determine light attenuation in aquatic environments can be complex. Therefore, in situ measurements of irradiance are often the most descriptive measurements of the overall light level and environmental variation experienced by an aquatic animal (Johnsen, 2012). Studies of spectral attenuation in freshwater habitats first measured downwelling irradiance in highly transmissive waters (Jerlov, 1968, 1976). Subsequent study of inland turbid bodies of water found that they strongly attenuate light (Kirk, 1981). More recently, heavy attenuation in turbid bodies of water has been determined by remote sensing of large bodies of water and also from direct measurements of highly attenuating estuarine habitats (Lawrenz et al., 2010; Simon and Shanmugam, 2016). Downwelling irradiance in desert ephemeral pools has not been assessed to date (Scholnick, 1994; Williams, 2005). The temporary waters inhabited by many branchiopod crustaceans are characterized by high levels of particulate matter and by suspended soil in the water column (Brendonck et al., 2008; Williams, 2005). The size of a given soil particle determines how long it remains in the water column and how it attenuates light (Partridge and Cummings, 1999). Therefore, light attenuation in temporary waters also varies regionally because of soil particle size distribution (percent sand/silt/clay) affecting the number of particles of a given size that are suspended in the water column. How long particles remain in the water column depends on the presence of wind or biological activity of wading birds or other swimming animals. Given the potential for heavy attenuation in temporary waters, we predicted that ephemeral pools inhabited by branchiopods would be dimmer and more spectrally variable than other bodies of freshwater that have been categorized, or than terrestrial habitats.

Despite having only two neuropils in their optic lobes, all branchiopods studied to date express four or more rhabdomeric opsins in their compound eyes (Kashiyama et al., 2009). In a companion paper (Lessios et al., 2018), we report that these opsins are most likely configured in four spectral photoreceptor classes in their compound eyes, which agrees with electroretinography of branchiopods studied to date, and the most recent rhabdomeric opsin phylogeny of pancrustacean compound eyes (Henze and Oakley, 2015; Smith and Macagno, 1990). In the first optic neuropil, the lamina, we identified lateral processes as neuroanatomical candidates for spatial summation in dim light conditions. Some pancrustaceans are thought to pool responses from multiple spectral photoreceptor classes in the lamina for visual behavior in dim light conditions (Menzel, 1974; Menzel and Greggers, 1985). However, it remains to be determined whether spatial summation takes place from both multiple ommatidia and spectral photoreceptor classes (Nilsson and Kelber, 2007). Given that there is evidence of broadband neurons in the lamina of other pancrustaceans (Kelber, 2006; Menzel, 1974; Menzel and Greggers, 1985; Yang and Osorio, 1996), broadband summation may take place in the lamina of branchiopods. Many pancrustaceans also possess long visual fibers that extend from the retina to the second optic neuropil. The long visual fibers of pancrustaceans can terminate in these deeper optic neuropils, and in some cases, lateral processes of the long visual fibers have been found in the lamina (Shimohigashi and Tominaga, 1999). Given the structure of the first optic neuropil we have reported for these two species, we predict that light-dependent behavior could involve summation from multiple spectral photoreceptor classes and would therefore have broad spectral bandwidth. Because anostracan branchiopods have been reported to have long visual fibers (Nässel et al., 1978), we predict wavelength-specific behavior of narrow bandwidth.

In the present study, we asked whether two species of branchiopods, T. longicaudatus and S. mackini, use multiple spectral channels for light-dependent behavior that uses luminance, rather than color vision, and if so, how this relates to their natural environment. We took in situ measurements of downwelling irradiance in temporary freshwater pools inhabited by T. longicaudatus and S. mackini. To better understand some of the factors affecting regional attenuation, we performed soil texture analysis on dry soil samples from two regions in Arizona to identify particle size distributions, and we analyzed total suspended solids from water samples taken in the field. We observed behavioral responses of populations from separate regions to calibrated light stimuli of narrow spectral bandwidth to ask whether visual performance is adapted for the spectral regime of each region. To ask whether branchiopods use light to maintain position in the water column by actively swimming, we observed behavioral responses under broad-spectrum lighting conditions, as well as in the dark. It should be noted that although comparing horizontal and vertical behavior provides insight regarding stereotyped responses to light, it still does not approximate responses to a natural light field in ephemeral pools, which as of yet remains unknown. An overarching hypothesis supported by the present study is that the visual system must serve the behavioral requirements of the organism in question, and that sensory structures at the periphery of the nervous system, such as eyes, are more easily modified, and potentially more developmentally plastic, than centralized neural structures (Land and Nilsson, 2012). Our aim was to study behavioral responses of branchiopods with simple optic lobe neuropils that explain the presence of multiple spectral channels in branchiopod compound eyes.

We collected dry soil from the upper 2.0 cm of dry ephemeral pools located in the Sonoran Desert in central Arizona and near Willcox Playa in Southeastern Arizona (33.670°N, 111.464°W, and 32.221°N, 109.827°W). Resting eggs were contained in the soil samples. We collected soil throughout the catchment area of each dry pool (approximately 2000 m²), then mixed soil thoroughly within each sub-sample to ensure that eggs produced by different individuals were contained in all samples. Streptocephalus mackini lack a head-shield, so catching them in the field and transporting them without damage was not possible. We therefore only reared S. mackini from soil samples. We caught adult T. longicaudatus in a single pool in Central Arizona, as well as from a single pool in Southeastern Arizona (32.221°N, 109.827°W) using a dip-net that caused no damage. We reared shrimp to test their behavioral responses to light. Shrimp were reared under conditions with total suspended solid particulate matter similar to that measured during seasonal monsoonal wet periods in 2012 and 2013 (0.40±0.23 g l⁻¹ mean±s.d.).

We measured downwelling irradiance in the field at 1.0 cm depth intervals, with the shallowest measurement taken with the collector just below the water surface. We took measurements every 3 days from the time of an initial filling event in six ephemeral pool communities (100 m² or greater), between 10:00 and 14:00 h. These measurements were taken for 16 days total. We used a USB 2000+ spectrometer (Ocean Optics, Dunedin, FL, USA), 10 m cable, and a CR2-OR 0.635 cm diameter waterproof cosine corrector (StellarNet, Tampa, FL, USA) with a spectroradiometric calibration from Ocean Optics. At the time of initial filling of each pool, a GPS point was taken above the deepest point of each pool. Depth at these coordinates was measured on the day of each irradiance measurement. The mean (±s.d.) depth of these pools was 122±44 cm. Latitude and longitude coordinates, determined by GPS, are presented for each ephemeral pool in Table S1. Local attenuation coefficients were calculated using paired measurements from just below the surface and at 1 cm depth. These were averaged from three replicate measurements taken at each site, from a different spot in the pool, on each sampling day. Because of heavy attenuation at depths greater than 1 cm at some field sites leading to loss of signal at short wavelengths, measurements were compared at this depth from 350 to 700 nm.

To determine whether ephemeral pool light environments are affected by attenuation that occurs because of soil suspended in the water column, we performed soil texture analysis (percent sand/silt/clay) in the laboratory on dry soil samples using a hydrometer procedure according to methods used by McKenna and Sala (2016). We took hydrometer readings 40 s after we added processed soil (after all sand particles had settled) and after 7 h (after all silt particles settled). Clay particles can remain suspended in the water column for a week or more (Elliott et al., 1999). We asked whether suspended soil in the water column differed between two regions because of soil particle size distribution. To estimate total suspended solids in filled ephemeral pools, we performed vacuum filtration analysis on 10 ml water samples collected from the upper 10 cm of each pool every 3 days from the time of filling. These measurements were taken for 16 days total. We filtered total suspended solids using GF/F filters, and then dried and weighed them on a balance (Mettler Toledo XS105, Columbus, OH, USA) according to methods used by Deng et al. (2015). We used soil particle size distributions to estimate the suspended silt and clay found in water samples. We used a log transformation of total suspended solid estimates to establish normality (Shapiro–Wilk test) for statistical comparisons.

We reared shrimp in wading pools in a greenhouse facility under natural sunlight (Fig. S1). Wading pools were each filled with 80 liters of deionized water and then we mixed 1.5 liters of collected soil into the water. After 24 h, we added an additional 20 liters of deionized water and maintained a constant volume of 100 liters by adding deionized water as needed. We maintained a 25°C:15°C, 12 h:12 h light:dark diel cycle using a programmable thermostat and light timer. Under these conditions, shrimp developed brood pouches within 5–7 days after hydration of resting eggs, comparable to development rates observed in the field (N.L., personal observation). We did not add food supplement until we observed shrimp with brood pouches because most larval food is contained in the soil (Centeno et al., 1993). After observing adult shrimp, we mixed 10 ml of 0.025 g ml⁻¹ Brewer's yeast solution (Twinlab, Inc., Grand Rapids, MI, USA) into each wading pool once a week. In order to verify that the shrimp we were testing were the species we had identified using their morphology, we extracted DNA from a subset of individuals and amplified a cytochrome oxidase I (COI) mitochondrial segment (Folmer et al., 1994) from 20 individuals used in behavioral trials, verifying the species as T. longicaudatus and S. mackini. The tested populations of T. longicaudatus consist only of self-fertilizing hermaphrodites (Maeda-Martínez et al., 1997), verified using the presence of brood pouches in every individual. Sexes are separate in S. mackini. Individuals of each sex in the latter were identified using the morphology of second antennal appendages (Belk, 1975).

We used a behavioral assay in a horizontal trough with light at one end to assess responsiveness to light under constant gravitational cues that could otherwise elicit a gravitaxis response, following testing procedures described in Cohen et al. (2010). We placed the testing trough 55 cm away from the light source, which consisted of a 500 W tungsten lamp housed in a slide projector (ViewLex, Inc., Holbrook, NY, USA) with a cooling fan. In control trials we only turned on the cooling fan. We limited infrared radiation from the light source using a hot mirror (Edmund Optics, Barrington, NJ, USA). We controlled wavelength using interference filters with a bandwidth at half maximum of 12.5 nm (Ditric Optics Inc., Marlborough, MA, USA), with center wavelengths ranging from 410 to 636 nm. We controlled irradiance using neutral density filters (Edmund Optics) and measured this using the spectrometer and equipment described earlier for field measurements. The minimum stimulus intensity centered at each wavelength in these experiments is shown in Fig. S1B.

where n_towards is the number of animals moving towards a calibrated light source, n_away is the number moving away (see below) and n_total is the total number of animals.

To assess behavior during active swimming of S. mackini and T. longicaudatus in the vertical plane in the presence of gravitational cues that could elicit a gravitaxis response, we marked a cylindrical glass testing column into five horizontal sections of 4 cm each (Fig. 2A, inset). We used this for testing in both a greenhouse setting under broad-spectrum light including UV (Fig. S2A), and for controls in a darkroom. Testing groups included 10 shrimp in each trial, with groups of S. mackini consisting of either 10 males, 10 females, or five males and five females. We placed the testing column in the center of a cardboard outer housing. The outer housing had walls that limited the view of the surroundings. The outer housing walls and the base of the column were painted flat black. A hole in the outer housing was used to position a video camera that maintained the entire column in its field. We placed a removable frame containing parchment paper above the apparatus during testing to serve as a broad-band diffuser of downwelling greenhouse light (Fig. S1).

We acclimated shrimp in deionized water for 3 h prior to testing active swimming. We tested shrimp in the greenhouse between 10:00 and 14:30 h. Six trials were carried out on each testing day. We alternated the testing order to avoid effects of time of testing. Between trials, we removed the water and cleaned the column. We carried out 12 replicates for each group composition of S. mackini and 10 replicates of T. longicaudatus. At the beginning of a trial, we pipetted a testing group gently into the center of the vertical column. During a day of testing, the solar orb passed symmetrically over the testing apparatus. We only tested on cloudless days. Downwelling light levels at the top of the testing column were measured at 14:00 h on a cloudless day.

We recorded video in the greenhouse light condition using a SDR-S7 digital video camera (Panasonic), through a 4 cm diameter opening in the outer cardboard housing. We recorded video in the dark using a C5900 H.264 wireless IP camera (Esky, Beijing, China) with a built-in infrared light source. We observed shrimp on the camera's viewscreen or a laptop screen during testing, recording the numbers of dead or non-responsive shrimp every 12 min. We later extracted individual frames using a VLC media player (VideoLAN, Paris, France) from the end of each 1-min interval of recording. We did not include the first minute interval in analyses as shrimp often displayed an initial flight reaction before resuming regular swimming. We assigned a random accession number to video files from each trial and scored each frame manually. The scorer was blind to the group composition of S. mackini as it was not possible to resolve sexual characteristics in the extracted frames. Based on preliminary results, we stopped experimental trials after 36 min of continuous video to ensure only active animals were being tested. Preliminary analysis indicated the 10th frame had the fewest individuals with a flight reaction, and the fewest number of unresponsive animals. Therefore, to avoid pseudoreplication from each trial, we compared the frame extracted at 10 min among replicates. A two-way chi-square test determined whether the frequencies of animals in each vertical section was dependent on group composition and whether greenhouse experiments differed from darkroom controls. To compare the numbers of animals in each section of the vertical column, we performed a sequential Bonferroni correction (Hochberg, 1988). We carried out a control experiment to compare sinking speeds of shrimp that were not swimming. We anesthetized shrimp in carbonated water for 3 s, after which we observed no gill-swimming. We then transferred them into the testing column and recorded the time to sink to the bottom to determine an average sinking rate.

We quantified a behavioral response index of notostracan and anostracan branchiopods to narrow-bandwidth light in the horizontal plane under constant gravitational cues that could otherwise elicit a gravitaxis response (Fig. 1). Triops longicaudatus moved away from the light source at all tested wavelengths except 550 nm and at all intensities greater than 10^3.5 photons cm⁻² s⁻¹ (mixed ANOVA: between-subjects factor F_8,18=3.67, P=0.010; within-subjects factor F_4,72=15.99, P<0.0001; interaction F_32,72=0.702, P=0.865; Fig. 1A,B). The lowest response intensity different from the control for T. longicaudatus was 10^5.5 photons cm⁻² s⁻¹ (P<0.01; Fig. 1A, asterisks). Male S. mackini moved away from light at intensities above 10^5.5 photons cm⁻² s⁻¹ (mixed ANOVA: between-subjects factor F_8,18=5.611, P=0.001; within-subjects factor F_4,72=76.29, P<0.0001; interaction F_32,72=4.465, P<0.0001; Fig. 1C, asterisks). Female S. mackini also moved away from light at intensities above 10^5.5 photons cm⁻² s⁻¹ (mixed ANOVA: between-subjects factor F_4,72=8.817, P<0.0001; within-subjects factor F_4,72=64.445, P<0.0001; interaction F_32,72=1.704, P=0.032; Fig. 1D, asterisks).

At intensities of 10^5.5 photons cm⁻² s⁻¹, S. mackini males moved towards a light source of peak emission of 532 nm (Fig. 1C, asterisks). Female S. mackini did not move towards light at any of the tested wavelengths and intensities (Fig. 1D). The minimum stimulus intensity that elicited a behavioral response for both species of branchiopods is 10^5.5 photons cm⁻² s⁻¹ (Fig. 1). This falls below the intensity of dim starlight in a terrestrial habitat, which is 10^6.0 photons cm⁻² s⁻¹ (Kelber et al., 2002).

To determine whether notostracan and anostracan branchiopods use light for depth selection in the presence of gravitational cues that could elicit a gravitaxis response, we assessed behavior under diffuse broad-spectrum light (Fig. S1). Groups of shrimp were tested in a vertical cylindrical column (inset Fig. 2A). Streptocephalus mackini tended to swim away from the surface in the greenhouse and were more likely to sink in the dark (Fig. 2B). We found the vertical distribution of S. mackini to be dependent on group composition (chi-square test of independence, χ²=19.13, P=0.014, N=257; Fig. 2A). Males were more likely to maintain a vertical position of 8.0–12.0 cm from the surface than females, independently of the presence of females (Fig. 2A, asterisks). We found that T. longicaudatus tended to maintain a vertical position between 16.0 and 20.0 cm away from the surface in light conditions, with more sporadic movements towards the surface under dark conditions (chi-square test of independence, χ²=22.96, P=0.004, N=187; Fig. 2B). To further assess whether shrimp actively maintained a vertical position, we measured the sinking rates of anesthetized shrimp (Fig. 2C). Sinking rates indicated that male S. mackini sink faster than females (one-way ANOVA followed by Tukey's post hoc test F_2,40=24.20, P<0.01), even though males tended to swim slightly above females in the greenhouse experiments (Fig. 2A). Together, these results suggest that both species of branchiopods use downwelling light to regulate their position in the water column when gravitational cues could also be used for gravitaxis.

To compare behavioral responses of two natural populations of branchiopods from two separate regions in Arizona, we carried out behavioral assays on field-caught T. longicaudatus. We found that T. longicaudatus from Central Arizona had a broader response spectrum than the population from Southeastern Arizona (Fig. 3). The latter population responded significantly to light at >470 nm (one-way ANOVA followed by Dunnett's post hoc test relative to control F_8,64=5.419, P<0.001; Fig. 3).

We took measurements of downwelling irradiance during seasonal monsoonal wet periods to describe the spectral properties of ephemeral pool light environments. We found that the spectral quality (λ_P50, the wavelength at which 50% of photon flux distribution occurs) of ephemeral pools varied considerably among two regions (Fig. 4A). From the time of initial filling until 16 days later, we found a standard deviation of 70.4 nm among all habitats of two regions, a standard deviation of 26.4 nm within Central Arizona habitats and a standard deviation of 57.4 nm within Southeastern Arizona. The λ_P50 of the ephemeral pool habitat where eggs were collected after its desiccation for the behavioral response index assay was 534±18.0 nm (mean±s.d.; Fig. 4A, dark gray). In order to model intensity loss with depth, irradiance measurements were used to calculate diffuse attenuation coefficients. At less than 1 m of depth, ephemeral pool habitats tended to reach light intensities lower than those of terrestrial habitats under dim starlight (Fig. 4B). Diffuse spectral attenuation coefficients are presented in 25 nm bins in Fig. 4C for comparison to other aquatic environments (Jerlov, 1976). Note that the clearest waters transmit the most light at 480 nm, which is not the case for other highly attenuating waters such as coastal marine environments, or as we describe here, ephemeral pools (Fig. 4C). Light attenuation per centimeter in the water column of these pools is approximately two orders of magnitude higher than the most heavily attenuating coastal waters (Fig. 4C).

To determine whether soil particle size distribution could explain regional differences in light attenuation, we used hydrometers (Fig. 5A) to carry out soil texture analysis on dry soil samples from Central and Southeastern Arizona (Fig. 5B,C). Particle size distribution was dependent on the region (two-way ANOVA with region and particle size distribution as factors, interaction F_2,54=180.22, P<0.001; Fig. 5C). Soil texture analysis includes the largest particles (sand) that settle from a suspended solution in 40 s or less. It is unlikely that sand particles remain suspended in the water column throughout the life cycle of branchiopod populations in the field. Excluding sand, Central Arizona had a mean±s.d. particle size distribution of 66.9±1.9% silt and 33.1±1.9% clay. Southeastern Arizona had 41.0±2.0% silt and 59.0±2.3% clay.

To determine total suspended solids of water samples taken in the field, we performed vacuum filtration of water samples. Southeastern Arizona had higher total suspended solids (mean±s.d. 1.734±0.98 g l⁻¹) than Central Arizona (0.397±0.190 g l⁻¹). We then used soil particle size distribution measurements from each region to estimate suspended soil by particle size. Southeastern Arizona had significantly more suspended clay than silt particles, whereas Central Arizona had more suspended silt than clay (two-way ANOVA on log-transformed suspended soil, with region and particle size distribution as factors, interaction F_1,32=6.136, P=0.019; Fig. 5D). Clay particles have smaller diameters (<0.002 mm) than silt particle diameters (<0.05, >0.002 mm). Therefore, clay includes particles with radii that are roughly one-tenth the wavelength of UV and short wavelength light, which leads predominantly to scatter of shorter wavelength light than that caused by silt particles (Johnsen, 2012). Through scattering, clay contributes to higher attenuation of short wavelength light in Southeastern Arizona than in Central Arizona.

The temporal variation of light and suspended soil for each region is presented in Fig. 6. The pools in this study filled from seasonal rains that took place in 2012 and 2013. Large single rains filled each pool (time since filling=0). λ_P50, total irradiance and total suspended solids varied irregularly over time. A clear peak in measured suspended solids and a minimum in total irradiance occurred 6 days after initial filling (Fig. 6C,B). Clay particles can remain suspended for weeks, silt for up to 8 h and sand particles for a matter of seconds. Mixing from wind or other disturbances would result in an increase of suspended silt, and a temporary drop in total irradiance over this time frame. Our estimates of suspended clay versus silt would not detect a change over time within a region, as texture analysis was performed on dry soil samples to determine particle size distribution (Fig. 4B), and was used to estimate suspended soil (Figs 5D and 6D). The filled area of each pool used in this study was >100 m². In smaller pools, it has been reported that the presence of notostracans contributes to suspended solids in the water column, but these experiments have not been carried out for pools >100 m², and the density of notostracans could vary with total water volume. In mesocosm experiments, the presence of notostracans increases total suspended solids because of their swimming activity (Waterkeyn et al., 2011). We first observed adult branchiopods 5 to 7 days from initial filling in every pool, coinciding with the peak in suspended solids at 6 days from filling.

Light is one of the primary environmental factors used by zooplankton for depth selection behavior (Cohen and Forward, 2009; De Meester et al., 1999). We show that two species of branchiopod crustaceans found in ephemeral pools, T. longicaudatus (Order Notostraca) and S. mackini (Order Anostraca), use light to regulate their depth in the water column. Early studies suggested that a suspension feeding branchiopod, Daphnia magna (Order Cladocera), might use spectral preferences, termed ‘color dances’, to regulate its vertical and horizontal position toward alga-rich water (Lubbock, 1888; Smith and Baylor, 1953). Recent studies have shown considerable regional variation in shallow freshwater spectral environments (Sabbah et al., 2011). All branchiopods are passive dispersers and are often found in hydrologically isolated, environmentally variable freshwater habitats (Brendonck et al., 2008; Williams, 2005), so adaptation to local environments would be expected. Because behavioral responses of Daphnia depend on wavelength, branchiopod responses may depend on regional light environment, rather than spectral preferences associated only with the transmission properties of algae (Stearns, 1975). We found that regional light environments inhabited by notostracan and anostracan branchiopods differed in the attenuation of short wavelength light, and that smaller suspended particles (clay) led to UV attenuation with depth. Our results from field-caught T. longicaudatus from separate regions support the prediction that behavioral responses are adapted to the spectrum of downwelling irradiance of a given region at depth.

The levels of suspended soil we have described in ephemeral desert pool habitats lead to rapid light attenuation with depth. The branchiopod species studied here inhabit spectrally variable light environments that are often dimmer than terrestrial habitats under starlight. Such patterns of regional variation in light environments have consequences for the expression of multiple opsin proteins, which are the primary genetic components of visual pigments expressed in photoreceptors (Bowmaker, 1999). A branchiopod selective history in variable environments has led to one of the highest numbers of duplicated opsins found in animal genomes sequenced to date. There is also evidence that recently duplicated paralogs are maintained at a much higher rate than would be expected given homogenization by gene conversion (Colbourne et al., 2011). In a companion paper (Lessios et al., 2018), we suggest that branchiopods have maintained four spectral classes of photoreceptors in their compound eyes, which is supported by the number of opsins expressed in anostracan and notostracan compound eyes (Kashiyama et al., 2009). Based on the number of spectral photoreceptor classes alone, it might be assumed that these species possess tetrachromatic vision. A 518-million-year-old stem mandibulate fossil had optic lobe neuropils that are associated with color vision in extant pancrustaceans (Ma et al., 2012). Branchiopod fossils that are 365 million years old have been found with evidence of resting eggs, suggesting that they inhabited ephemeral and most likely turbid bodies of water (Gueriau et al., 2016). Given that branchiopods have simplified optic lobes, we suggest that the opsins which are configured for four spectral photoreceptor classes have diversified throughout the evolutionary history of notostracans and anostracans for broad-bandwidth behavioral responses in temporary habitats.

Based on the spatial variation and regional effects of light environment on behavior (Fig. 3), we suggest that each population is adapted to the spectral conditions of its specific region. There are many compound eye parameters that could change spectral sensitivities of the compound eyes. Some studies have modeled visual pigment concentration, k, the absorption coefficient per micrometer of the photoreceptors, and have been able to identify differences in relative contribution of one or more photoreceptor classes based on habitat or regional light environment (Johnson et al., 2002; Lessios, 2017). However, empirical evidence is currently lacking of routine changes in visual pigment concentration in rhabdomeric photoreceptors among individuals (Fuse et al., 2014), and other photoreceptor parameters that have an effect on spectral sensitivity could change in development.

The broad spectral bandwidths we found for photonegative behavior at low intensity suggest that both T. longicaudatus and S. mackini generally employ multiple spectral photoreceptor classes for depth-selection behavior (Fig. 1). In T. longicaudatus, summation of outputs from multiple spectral classes could take place from broad receptive fields present in the lamina that we have identified in a companion paper (Lessios et al., 2018). The narrow bandwidth of photopositive responses we found for male S. mackini strongly supports the use, and the presence, of a single spectral photoreceptor class with an absorptance peak near 532 nm identified in a companion paper (Lessios et al., 2018). There are several possibilities regarding how a single receptor could provide input for narrow-bandwidth photopositive behavior. Long visual fibers have been reported in anostracans (Nässel et al., 1978). A long visual fiber that sums responses from several ommatidia in the proximal lamina could be configured to a single photoreceptor class to control wavelength-specific behavior. An alternative would be that short visual fibers from a single photoreceptor class make contact with other terminals in one of the main layers of the lamina and they are compared synaptically with other receptor classes. If there are synaptic comparisons of photoreceptor outputs, it could involve color vision for behavior. However, because male S. mackini photopositive responses to light centered at 532 nm were restricted to a narrow waveband, we refer to them as a wavelength-specific behavior (Goldsmith, 1990; Kelber and Osorio, 2010). Crustaceans and insects that employ wavelength-specific behavior tend to use a subset of their photoreceptors for behavior (Cohen et al., 2010; Kolb and Scherer, 1982), but it remains unknown whether color opponency is involved for such wavelength-specific behavior (Kelber and Osorio, 2010). Furthermore, as we have found males of S. mackini to swim slightly above females, we suggest they could be using a single photoreceptor class to actively maintain a position at shallower depth than females. This is supported by some accounts that report vision is involved in anostracan mate choice (Belk and Martin, 1991).

We have found that ephemeral pool habitats, at less than 1 m of depth, are often dimmer than terrestrial habitats under starlight. In a companion article (Lessios et al., 2018), we found that branchiopod apposition compound eyes are likely to generate reliable signals at such intensities. There is some neuroanatomical evidence that pancrustaceans might incorporate summation from compound eye ommatidia in the lamina of their optic lobes for reliable vision at low intensities (Glantz, 2014; O'Carroll and Warrant, 2017; Warrant, 2017). Support for summation from multiple photoreceptor spectral classes for achromatic visual behavior at low intensity has so far been limited to behavior described in honeybees (Menzel and Greggers, 1985). We suggest that T. longicaudatus and S. mackini are similarly employing summation for the behavior at low intensity described here. We infer that animals with simplified visual systems, such as branchiopods, are likely to have stereotyped light-dependent responses. If their habitats reach low intensities, their visual systems are likely to be configured to employ summation from multiple spectral classes in the peripheral nervous system. Research of light-dependent behavior that employs summation from multiple photoreceptor classes and wavelength-specific behavior is needed to fully understand the selective pressures that drive opsin diversification and maintenance. We suggest that future studies which make inferences about color vision should continue to use integrative approaches incorporating electrophysiology, neuroanatomy and behavior.

We thank two anonymous reviewers, and the editor, for their contributions to the manuscript. We thank Nicholas Strausfeld for useful comments on the manuscript. We thank undergraduate assistants Patrick Murphy and Molly Klein for their help in the greenhouse and with water sample filtration. We thank Susanne Neuer for providing her laboratory and equipment, and Wei Deng for help with vacuum filtration of water samples, as well as Owen McKenna and the Hall laboratory for their help with soil texture analysis of our dry soil samples. We thank Christopher Rogers, Jennifer Work, Kevin Smith and David Pearson for their help in locating ephemeral pools with branchiopod populations.