Collections were carried out in Lázaro Beach, located in Fortaleza Inlet (23°30’S 45°08’W) in Ubatuba, a municipality on the southeastern coast of Brazil (Fig. 1). The beach is bordered by rocky shores with dense cover of the brown alga Sargassum cymosum C. Agardh, 1820 and is moderately exposed to wave action (Széchy and Paula 2000). This area was selected for the present study because caprellids are very abundant, with densities higher than 20 ind./g of dry weight of S. cymosum (Jacobucci et al. 2002). Specimens were collected monthly from October 2010 to February 2012 in a slopping rocky shore area from 2.5 to 3.0 m depth from the surface. In each month, 25 fronds of S. cymosum were randomly collected through snorkelling. The fronds of S. cymosum were removed from the substrate with a spatula and individually covered with fabric bags (0.2 mm mesh size) (Jacobucci et al. 2002). This technique was previously used by Takeuchi et al. (1987) and proved to be efficient, and the mesh size was suitable to retain the caprellids. The bags with algae were stored in glass jars, fixed with 5% formaldehyde and transported to the laboratory.

In the laboratory, each frond was washed with freshwater to remove the associated epifauna. This process was carried out three times to increase collection efficiency. Sargassum cymosum fronds were dried at 60°C for 48 h and then weighed to determine the dry weight (biomass) of algae. The removed epifauna was filtered on a 0.2 mm sieve, placed into labelled jars and preserved in 70% ethanol for later taxonomic identification. The caprellids were identified to species level (Lacerda and Masunari 2011) under a stereomicroscope (Bel Photonics and Nikon SMZ 754T). Caprellid density was calculated as the total number of recorded individuals of a given species divided by the total Sargassum dry weight (g), expressed as ind. g^-1.

Sex determination was performed for all caprellids, using adapted methods of Bynum (1978) and Takeuchi and Hirano (1991). The specimens were separated into the following demographic categories: juveniles, immature females, mature females, ovigerous females and males. Individuals smaller than the smallest female analysed in the study were classified as juveniles. Caprellids larger than this female without oostegites were classified as males and those with oostegites as females. Females were classified as immature, when they did not have bristles on oostegites; mature, when they had bristles on oostegites; and ovigerous, when showed the presence of eggs or juveniles in their brood pouch. All caprellids were measured (total length) under a stereomicroscope with an ocular micrometer with constant magnification. For each specimen, the body segments were individually measured and summed to obtain the total length (mm) (Garcia et al. 2019).

The model’s assumptions of homoscedasticity (Levene’s test) and normality (Shapiro-Wilk test) of the population size distribution were tested. The mean size of the caprellids of each species was compared between males and females by the non-parametric Mann-Whitney test (p<0.05) (Zar 2010). To evaluate the population biology of each species, size-frequency distributions were constructed using 1.1 mm (total length) intervals for both males and females. The individuals were distributed into 13 size classes, from 0.6 to 14.9 mm (total length). Sex ratio of each species was estimated as the quotient between the number of males and the total number of individuals in the population (males plus females) (Wilson and Hardy 2002). Deviations from a 1:1 sex ratio were tested using a binomial test (p<0.05) (Wilson and Hardy 2002). Sex proportion values higher or lower than 0.5 indicated populations skewed toward males or females, respectively.

The temporal dynamics of the caprellids associated with Sargassum were evaluated by interpreting parameters of each month’s samples. To verify a possible tendency of variation in density of caprellids of each species and biomass of the Sargassum fronds, a linear regression analysis was performed. Multivariate analysis was carried out considering two main periods (P), corresponding to the following seasons, spring and summer (P1) and autumn and winter (P2), as considered also by Barros-Alves et al. (2017). This separation allowed us to test the hypothesis that the structure of organisms associated with algal beds changes seasonally.

The temporal variation was analysed to investigate whether the density of caprellid species varied over the two main periods (P1 vs. P2). For this, a non-metric multi-dimensional scaling analysis was conducted using Bray-Curtis similarity matrices. One-way crossed analyses of similarity (ANOSIM) were used a posteriori to test for significant differences in the density of caprellid species between seasons. Paired comparisons between two main periods were performed when the ANOSIM R value was significant (p<0.05) (Clarke 1993). In addition, correspondence analysis was used to evaluate the relationship between the sampling month and the density of caprellids. For this analysis, density values were used, considering each species as an independent set of data, to minimize the influence of sampling design.

A total of 14939 specimens of Caprella danilevskii were recorded, including 2421 juveniles (16.21%), 1498 immature females (10.03%), 2703 mature females (18.09%), 750 ovigerous females (5.02%) and 7567 mature males (50.65%). The size-frequency distribution analysis indicated a polymodal and non-normal distribution for the population (Kolmogorov-Smirnov; K-S=0.039, p<0.001) (Fig. 2A). The mean size (mean±sd) recorded for the sampled population was 4.86±2.08 mm. The body size of the smallest and largest individuals observed during the sampling period was 0.67 and 14.24 mm, respectively. The overall sex ratio significantly differed from a 1:1 proportion and was skewed toward males (Sex ratio = 0.60, binomial test; p<0.001). The body size of males (5.33±2.09; range 1.64 to 14.24 mm) was significantly larger than that of females (5.63±0.86; range 1.84 to 9.11 mm) (Mann-Whitney test, U=14394022, p<0.001). Juveniles were distributed in the size classes from 0.6-1.7 to 2.8-3.9 mm. Mature males were recorded in all size classes except the first. The last size classes (from 9.4-10.5 to 13.8-14.9 mm) were dominated by mature males. Females were recorded in intermediate size classes (from 2.8-3.9 to 8.3-9.4 mm) (Fig. 2B).

A total of 7211 specimens of Caprella equilibra were recorded, including 1450 juveniles (20.11%), 649 immature females (9.00%), 1289 mature females (17.88%), 244 ovigerous females (3.38%) and 3579 mature males (49.63%). The size-frequency distribution analysis indicated a bimodal and non-normal distribution for the population (Kolmogorov-Smirnov, K-S=0.062, p<0.001) (Fig. 3A). The mean size (mean±sd) recorded for the sampled population was 3.67±1.78 mm. The body size of the smallest and largest individuals observed during the sampling period was 0.66 and 12.53 mm, respectively. The overall sex ratio significantly differed from a 1:1 proportion and was skewed toward males (sex ratio = 0.62, binomial test; p<0.001). The body size of males (4.34±0.77; range 1.44 to 7.89 mm) was significantly larger than that of females (4.15±1.88; range 1.37 to 12.53 mm) (Mann-Whitney test, U=3190594, p<0.001). Juveniles were distributed in the size classes from 0.6-1.7 to 1.7-2.8 mm. Mature males were recorded in all size classes. The last size classes (from 8.3-9.4 to 11.6-12.7 mm) were dominated by mature males. Females were recorded in intermediate size classes (from 1.7-2.8 to 7.2-8.3 mm) (Fig. 3B).

A total of 1657 specimens of Caprella scaura were recorded, including 179 juveniles (10.80%), 189 immature females (11.41%), 285 mature females (17.20%), 65 ovigerous females (3.92%) and 939 mature males (56.67%). The size-frequency distribution analysis indicated a polymodal and non-normal distribution for the population (Kolmogorov-Smirnov, K-S=0.053; p<0.001) (Fig. 4A). The mean size (mean±sd) recorded for the sampled population was 5.02±2.12 mm. The body size of the smallest and largest individual observed during the sampling period was 0.78 and 12.96 mm, respectively. The overall sex ratio differed significantly from a 1:1 proportion and was skewed toward males (sex ratio = 0.60, binomial test; p<0.001). The body size of males (5.50±2.39; range 1.60 to 12.96 mm) and females (5.09±0.93; range 1.64 to 8.04 mm) showed no significant difference (Mann-Whitney test; U=239064, p=0.076). Juveniles were distributed in the size classes from 0.6-1.7 to 2.8-3.9 mm. Mature males were recorded in all size classes. The last size classes (from 8.3-9.4 to 12.7-13.8 mm) were dominated by mature males. Females were recorded in the size classes from 0.6-1.7 to 7.2-8.3 mm (Fig. 4B).

Biomass of Sargassum ranged from 3.34 g in March 2011 to 14.52 g in December 2012 (see Fig. 5), with mean values (±sd) of 7.53±3.45 g. Population density of C. danilevskii ranged from 0.28 ind.g^-1 in April 2011 to 28.03 ind.g^-1 in January 2012, with mean (±sd) of 5.00 ± 6.93 ind.g^-1 (Fig. 5A). Population density of C. equilibra ranged from 0.23 ind.g^-1 in April 2011 to 7.70 ind.g^-1 in January 2012, with mean (±sd) of 2.43±2.44 ind.g^-1 (Fig. 5B). Population density of C. scaura ranged from 0.10 in August 2011 to 6.03 in February 2012, with mean (±sd) of 1.06±1.40 ind.g^-1 (Fig. 5C). A positive correlation was observed between biomass of Sargassum and density of C. danilevskii (Linear regression; r²=0.17, F=12.40, p<0.01) (Figure 5A), C. equilibra (linear regression; r²=0.10, F=4.27, p<0.01) (Figure 5B) and C. scaura (Linear regression; r²=0.14, F=8.91, p<0.01) (Fig. 5C).

The non-metric multi-dimensional scaling ordination derived from caprellids recorded two groups, as seen in Figure 6A. ANOSIM indicated a significant difference in the density of the caprellids between the two analysed groups (spring-summer vs. autumn-winter) (ANOSIM, R=0.644, p=0.001; Fig. 6A). Density varied across months and seasons (spring-summer and autumn-winter), and this variation was observed in correspondence analysis (Fig. 6B). For C. danilevskii and C. equilibra, the highest density corresponded to late spring and early summer (Figs 5A, B, 6B), while the highest density of C. scaura corresponded mainly to late summer 2012 (Figs 5C and 6B).

In this study, a polymodal and non-normal distribution for C. danilevskii and C. scaura and a bimodal distribution for C. equilibra were observed. Bimodality or polymodality in the size-frequency distribution may be related to seasonal reproduction of these species throughout the year, which is influenced by recruitment peaks, mortality, migration and/or behavioural differences (Díaz and Conde 1989). This pattern is common among amphipods, such as Cymadusa filosa Savigny, 1816, Mallacoota schellenbergi Ledoyer, 1984 (Appadoo and Myers 2004), Gammarus chevreuxi Sexton, 1913 (Subida et al. 2005) and Hyalella pleoacuta González et al. 2006 (Castiglioni and Bond-Buckup, 2008).

Males of all Caprella species reached larger mean size than females, as is usual for many other caprellids (Guerra-García et al. 2011, Lolas and Vafidis 2013, Garcia et al. 2019). This sexual dimorphism can be related to different energy allocation for growth and the existence of aggressive behaviour between males before copulation (Caine 1991). The maximum size of the caprellids in our study was considerably smaller than that of other populations. Caprella scaura males of a population in southern Italy (Prato et al. 2013) reached 23 mm and in the present study they reached only 12.96 mm. Caprella equilibra males from the northern Adriatic Sea (Sconfietti and Luparia 1995) reached 19 mm, while in the studied area the maximum male size was 12.53 mm. This difference in body size can be related to local parameters, including water temperature and predation pressure. Comparing C. scaura sizes from different seasons in South Carolina, Foster et al. (2004) recorded larger males in winter than in summer. They suggested that this difference could be related to predation decrease and reduced reproductive activity during colder months. Guerra-García et al. (2011) suggest that other factors such as competition with other species and availability of substrates or food could also be responsible for these differences.

The variety of sizes recorded for ovigerous females in this study has already been observed for Caprella equilibra and Caprella dilatata Krøyer, 1843 in a study conducted in Argentina (Nuñez-Velazquez et al. 2017) and for C. equilibra from an estuarine population on the northern coast of Italy (Sconfietti and Luparia 1995). This pattern indicates that all three species in our study have an iteroparous life cycle, i.e. they can reproduce multiple times.

The sex ratio deviated for males as recorded in this study was already observed for other caprellid species, such as Paracaprella tenuis Mayer, 1903 and Pseudaeginella montoucheti Quitete, 1971 (Garcia et al. 2019). The predominance of males suggests an intraspecific competition for females (Powell and Moore, 1991). The sex ratio deviation can also be related to high energetic investment in reproduction by the females (Cardoso and Veloso 1996), which can limit growth and reduce their survival (Thiel 2003).

A density variation throughout the year was recorded for all the species with higher densities in spring and summer. This pattern is the result of favourable conditions in these periods, such as the higher incidence of light and the nutrient availability, which allow for greater growth of macroalgae (Moore et al. 1997, Moore and Wetzel 200). According to De Paula et al. (2016), the reproductive biology evaluation of the caprellid amphipods of Lázaro Beach indicated a more intense reproduction in autumn and winter. Other caprellid species, such as Paracaprella tenuis and Pseudaeginella montoucheti, which were recorded in the same area as the present study, showed higher densities in winter (Garcia et al. 2019). This indicates possible competitive interactions among those caprellids and Caprella species.

In this study, the period of higher algal biomass (spring-summer) corresponds to the higher densities of Caprella species. The biomass of Sargassum fronds is an important predictor of caprellid densities, in which higher densities of caprellids would be expected in periods with higher algal biomass, and this pattern was also recorded for other caprellids, such as P. tenuis and P. montoucheti (Garcia et al. 2019). However, other factors are certainly important to explain caprellid density because the biomass-density correlations were low (r²<0.20).

Epibiosis could be one of these factors because it increases habitat complexity (James and Heck 1994) and consequently the availability of resources. In a study conducted just a few miles from the site of the present study (Jacobucci et al. 2009), the densities of some caprellid species were positively related to epiphyte algal load. The accumulation of detritus is enhanced by epiphytes (Heck and Wetstone 1977, Hacker and Steneck 1990, Russo 1990), thus benefiting species that are mainly detritivores such as the Caprella species studied (Guerra-García and Figueroa 2009).

Epiphytes also increases the available surface, favouring colonization of bacteria and microalgae that are food resources for grazers. This biofilm could contribute to the higher mean densities of C. danilevskii, which has an opportunistic feeding habit, consuming not only detritus but also microalgae (Guerra-García and Figueroa 2009). Hydrozoan cover in Sargassum fronds can also explain caprellid density variation. In a study conducted at Lázaro Beach, the density of Caprella danilevskii and C. equilibra species was significantly related to total hydrozoan cover of S. cymosum fronds (Cunha et al. 2018).

Wave exposure in the sampling area is another factor that could favour C. danilevskii. This species is commonly abundant in areas with higher hydrodynamic levels (Guerra-García and García-Gómez 2001). Caprella danilevskii acquired the ability of attaching to the substrate using gnathopod 1 in a “parallel” posture which reduces displacement by wave action (Takeuchi and Hirano 1995). This “habitat preference” is confirmed in a study conducted on a nearby rock shore with extremely low wave exposure, where C. danilevskii showed low densities in comparison with other Caprella species (Jacobucci et al. 2009).

Studies conducted in temperate regions indicated different peaks in reproductive activity. In Mar del Plata harbour (Argentina), higher densities of C. equilibra ovigerous females were recorded in summer (Nuñez-Velazquez et al. 2017) and in European (Mediterranean Sea) C. equilibra populations the breeding period occurs in spring and autumn at moderate temperatures and stops in winter (Sconfietti and Luparia 1995). Periodic or more intense reproduction in certain periods of the year seems to characterize the species of Caprella studied, with reproductive peaks coinciding with the coldest months of the year.

The present study indicates that sympatric populations of Caprella danilevskii, C. equilibra and C. scaura have significant temporal fluctuations, with higher densities in spring and summer that are related to higher algal biomass, but other environmental factors such as epibiosis and wave exposure are certainly important to explain caprellid density variation. Additional field and experimental studies on caprellids will be important to better understand the biology of this still poorly known crustacean group.