Three-Dimensional Analysis of the Swimming Behavior of Daphnia magna Exposed to Nanosized Titanium Dioxide

Due to their surface characteristics, nanosized titanium dioxide particles (nTiO2) tend to adhere to biological surfaces and we thus hypothesize that they may alter the swimming performance and behavior of motile aquatic organisms. However, no suitable approaches to address these impairments in swimming behavior as a result of nanoparticle exposure are available. Water fleas Daphnia magna exposed to 5 and 20 mg/L nTiO2 (61 nm; polydispersity index: 0.157 in 17.46 mg/L stock suspension) for 96 h showed a significantly (p<0.05) reduced growth rate compared to a 1-mg/L treatment and the control. Using three-dimensional video observations of swimming trajectories, we observed a treatment-dependent swarming of D. magna in the center of the test vessels during the initial phase of the exposure period. Ensemble mean swimming velocities increased with increasing body length of D. magna, but were significantly reduced in comparison to the control in all treatments after 96 h of exposure. Spectral analysis of swimming velocities revealed that high-frequency variance, which we consider as a measure of swimming activity, was significantly reduced in the 5- and 20-mg/L treatments. The results highlight the potential of detailed swimming analysis of D. magna for the evaluation of sub-lethal mechanical stress mechanisms resulting from biological surface coating and thus for evaluating the effects of nanoparticles in the aquatic environment.


Introduction
Although colloidal particles of geochemical and biological origin are abundant in the aquatic environment [1], the increasing use of engineered nanosized particles in industrial applications and consumer products raise eco-toxicological concerns [2]. Unknown toxicological effects especially of engineered nanoparticles on organisms require comprehensive multi-disciplinary investigations [3]. Scheringer [4] questioned that identical methods for chemical risk assessments can be applied to nanoparticles, whose environmental properties strongly depend on compositions, size distributions and surface treatment. Nano-sized titanium dioxide particles (nTiO 2 ) are of particular concern [5] because they are produced in large quantities and can be found in the aquatic environment [6].
Numerous toxicological studies with nanomaterials have been conducted using the water flea Daphnia sp. [7], which is a key ecological model organisms and of high relevance in many aquatic food webs [8]. The results of these eco-toxicological studies, however, are not consistent. While, e.g., Lovern and Klaper [9] reported mortalities (LC 50 ) for Daphnia magna at 5.5 mg/L nTiO 2 after 48 h. Heinlaan et al. [10] observed that even a concentration of 2 . 10 4 mg/L nTiO 2 is not acutely toxic. Zhu et al. [11] found only minor immobilization and no significant mortality of D. magna after 48 h exposure to 50 and 100 mg/L nTiO 2 , but a higher toxicity when exposure time was extended to 72 h. Lazorchak et al. [12] highlighted the importance of prolonged acute exposure times . 48 h and Dabrunz et al. [13] found a substantially higher toxicity when exposing D. magna for 96 h. The reason of the increased toxicity of nTiO 2 following an exposure $ 48 h is related to the inhibition of the second molting. Most importantly, the study by Dabrunz et al. [13] observed an adhesion of nTiO 2 to the daphnids exoskeleton, which had been previously described already by Baun et al. [14]. Dabrunz et al. [13] hypothesized that this biological surface coating by nTiO 2 may affect the swimming performance of the water flea. Swimming behavior is a frequently used endpoint in toxicity testing with daphnids, however, it has not yet been studied in detail during nanoparticle exposure, although the observed biological surface coating suggests a relevant effect.
Not only immobilization or mortality but also reduced growth or behavioral changes of organisms are of ecological relevance as sub-lethal endpoints [15,16]. Baillieul and Blust [17] found the growth endpoint to be a viable indicator of sub-lethal effects resulting from contaminants based on measurement of physiological energetics (scope for growth). Billoir et al. [18] came to a similar conclusion concerning growth using dynamic energy budget in toxicology analyses. Further studies [11,19,20] reported about growth and molting reduction using nanoparticles, but only on basis of long-term ($ 21 days) chronic tests. A common approach for testing immediate sub-lethal impacts of exposure to toxic substances on daphnids is to observe changes in swimming behavior [15,21,22]. To compensate sinking due to their negative buoyancy [23], D. magna have to swim upward frequently, resulting in typical hop and sink motion [24]. Dodson et al. [15] have demonstrated that the swimming velocity was the most useful endpoint for parameterization of differences in swimming behavior. Even among a wide variety of behavioral, biochemical and metabolic parameters, Duquesne and Küster [22] identified the swimming activity and particularly the mean swimming velocity as the most sensitive parameter. Baillieul and Scheunders [25] describe an image analysis method, which allows for measuring the average velocity of many simultaneously moving similar objects. Using this method, Baillieul and Blust [17] showed that Cd 2+ exposure reduced the growth rate and the swimming velocity of D. magna independently. Although, video analyses provide detailed information on the motions of groups as well as of individuals, growth and behavior depend on a variety of biotic and abiotic factors [24,26,27], making eco-toxicological studies using behavioral responses rather challenging. For technical reasons, numerous studies used planar, i.e., two-dimensional observations [16,17,22], where daphnids were confined in flat chambers or flow-through cells and restricted in their normal behavior.
Here we provide a method to analyze the three-dimensional swimming behavior of D. magna exposed to different concentrations of nTiO 2 . A novel two-camera tracking system was developed for the behavioral analyses of multiple and simultaneously swimming organisms. We use spectral analysis of the swimming velocities to quantify swimming activity. The mechanical impact of nTiO 2 on D. magna is analyzed by relating the observed swimming activity to size and growth rate of the organisms. This new approach should facilitate our mechanistic understanding of nanoparticle effects on aquatic organisms.

Husbandry and exposure
Fifty neonate (age ,24 h) D. magna (clone V) were separated from a permanent culture which is kept in ASTM medium (composition provided in the legend of Figure S1 in the Supporting Information) [28] within climate controlled chambers at 2061uC. Animals in the culture were fed with green algae Desmodesmus sp. on a daily basis. Test organisms were transferred to the test vessels (cubic glass tanks with 8 cm side length) 1 h before addition of nTiO 2 and start of the observations during the experimental period. Night and day cycles were simulated using warm-white LED (3450 lux; 450-760 nm, no UV) panels at 12 h:12 h periods. Five groups were composed, each containing 10 individuals. Two groups remained as control groups (C, C*) while three treatment groups (T 1,5,20 ) were exposed to 1, 5 and 20 mg/L nTiO 2 . Relative oxygen concentration measured in C and T 20 was always 10062%, electrical conductivity measured in C and T 5 was 54161 mS/cm and pH was 8.260.2 during the experimental period. Following procedures previously established [13] the animals were not fed during the experimental period. The nTiO 2 product (P25, Evonik AeroxideH) had a nominal primary particle size of 21 nm, and the surface area (Brunner-Emmett-Teller; BET) was approximately 50 m 2 /g. Size distributions in undiluted, monodisperse stock suspensions were determined by dynamic light scattering (Delsa TM Nano C, Beckman Coulter, Krefeld) to be 60.760.5 nm (polydispersity index: 0.157; D90: 113.563.0 nm). In the present study, we followed the analytical methods as described in Dabrunz et al. [13], with the only difference that we quantified 49Ti instead of 47Ti, since quantifying this mass did work slightly more accurate for P25. Briefly, titanium concentrations (mass 47, 17.46 mg/L) in the stock were verified by Quadropole ICP-MS (XSeries2, Thermo Fischer Scientific, Dreieich) equipped with a FAST autosampler (ESI, Thermo Fischer Scientific, Dreieich), a peek spray chamber (Thermo Fischer Scientific, Dreieich) and a robust Mira Mist peek nebuliser (Burgener, Berkshire). The instrument was run in the collision cell mode with 5 ml He/H2 cell gas in order to avoid polyatomic interferences (e.g., PO+ or SiOH+) [13]. As observed in previous studies [13], nTiO 2 precipitated by forming larger agglomerates leading to a reduction of 50% of the initial suspended particle concentration within 30 h ( Figure S1).

Video recording
Swimming behavior of the daphnids was observed using video recordings at (t 0 ), 24 h (t 24 ), 48 h (t 48 ), 72 h (t 72 ) and 96 h (t-96 ) after application of nTiO 2 in all treatments as well as in the control group. During each recording, daphnids of each group were observed through two orthogonal side-windows of the respective vessel. We used two synchronously triggered 204862048 pixel CCD-cameras (FloSense 4M, Dantec Dynamics, Skovlunde) at a frequency f s = 14.8 Hz and with a grey-scale depth of 8 bit. The usage of telecentric lenses (TC4M, Opto Eng., Mantova) allowed obtaining distortion-and divergence-free images of the entire test tank. Transformation of pixel coordinates to metric units was performed using a constant conversion factor of 19.5 pixel/mm obtained from calibration measurements using a metric target.
Cartesian coordinates x d of n simultaneously swimming Daphnia magna d were identified in each recorded image pair using an assignment algorithm [29]. The algorithm was first applied to each of the two-dimensional coordinates obtained from the two cameras and a second time to estimate three-dimensional coordinates of the organisms by combining the two perspectives. All observations were performed during the light cycle of the photo period using the same illumination as during the husbandry. Before each recording we waited 15 min to avoid abnormal swimming behavior caused by light fluctuations during the transport of the test tank from the husbandry to the camera setup.

Organism sizes
Body length (L) of each individual organism of the groups was estimated from the video observations for each sampling time. L (650 mm) corresponds to the length from the apex of the head to the abdomen of the organisms. All daphnids were identified and L was estimated by visual inspection of the side views if the individual organisms, where the corresponding image of the second perspective shows the front view of the same individual.
Relative growth rate DL L : Dt was estimated with time intervals

Velocity analyses
Swimming behavior of all daphnids within the respective treatments was recorded for at least 3 min at t 0 , t 24 , t 48 , t 72 and t 96 . Although the time of video recording relative to the application of nTiO 2 were identical for each group, sampling of all groups took approximately 3 h. Each sampling started with the control group (C) and ended with a record of group T 20 . It has been shown that body size has a strong impact on the swimming performance of Daphnia magna [17]. Thus, a second control group (C*) was used and measured at the end of the 3 h interval to estimate the growth during the measurement phase. Swimming velocity components v i~D x i : f s , where i~v, h 1 and h 2 denotes vertical or horizontal components, were calculated from corresponding displacements Dx i of individual daphnids observed in subsequent images obtained at the frame rate f s . Two-dimensional observations from the individual cameras allow for analyzing magnitudes of swimming along continuously recorded pathways of individual daphnids. The swimming velocities were averaged by calculating the temporal mean swimming speed v d as well as the ensemble mean Sv d T for each group. Three-dimensional coordinates and swimming speeds v 3D~ffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffi v 2 v zv 2 h1 zv 2 h2 q allowed identifying predominant swimming directions and organisms located in the vicinity of the vessel walls. We noticed only minor differences DSv 3D T in ensemble mean swimming velocities estimated separately for daphnids swimming at distances jD x ! j . 5 mm and # 5 mm (about two body lengths) to the glass surfaces (DSv 3D T=Sv 3D Tƒ0:05). Horizontal velocity components were predominantly isotropic, i.e. Sv d T corresponds to Sv 3D T withr 2 §0:97, where r denotes the correlation coefficient. However, three-dimensional pathways were frequently fragmented into shorter segments and the corresponding velocities could not unambiguously assigned to distinct individuals. Hence, v 3D was only used for estimating ensemble mean sinking velocities using 1 for passive sinking. Sinking velocities and corresponding body lengths had been used to estimate volume specific mass of the daphnids using Stokes law.
Swimming behavior was characterized using spectral analysis of the observed swimming velocities. Power spectral density PSD as a function of frequency f is estimated PSD(f )~ð where R(t) denotes the autocorrelation function of linearly detrended v d,v , using Welch's method [30], i.e. the Matlab (Math-Works, Natick) function psd( ) and method spectrum.welch, with a window length for Fourier transformation of 256 samples and 50% overlap. Variance in swimming velocity within a frequency range

Statistics
Wilcoxon rank sum test was applied for testing significant differences in sizes, ensemble mean swimming velocity, mean peak values of variance preserving PSD and mean integral values of high frequency PSD. A two sample t-test was applied for testing significant difference of linear regression of two datasets, i.e. of the slopes of ensemble mean velocity versus length of the daphnids (see below). Differences are considered to be significant if p,0.05 for all tests, unless otherwise stated. Bonferoni correction was applied to identify significant differences affected by the problem of multiple comparisons.

Growth rates and immobilization
Since the main aim of this study was to investigate the effects of nTiO 2 concentrations on the swimming behavior of D. magna, we do not take immobilized organisms into account in our study. In accordance with Baillieul & Blust [17], we did not mix observations of mobile animals with that of immobilized animals. Hence only nine individuals constitute T 1 at t . t 0 and T 20 at t . t 24 and only eight individuals constitute T 5 and T 20 at t . t 48 .
At t 0 , the initial length was 1.060.03 mm for all daphnids ( Figure 1A; Table S1 in the Supporting Information). At t 96 the control group C (2.060.15 mm) as well as T 1 (2.060.16 mm) were significantly larger in comparison to T 5 (1.660.26 mm, p = 0.004) and T 20 (1.460.29 mm, p = 0.002). This difference translates to a reduced growth rate of -0.01 1/d for T 5 and 0.02 1/d for T 20 between t 48 and t 72 . However, the growth rates estimated between other records of T 5,20 , i.e. from t 0 to t 48 and from t 72 to t 96 , were 0.09…0.27 1/d. These growth rates were similar to the growth rates of C and T 1 (0.09…0.26 1/d) during the entire experiment, i.e. from t 0 to t 96 . The second control group C* measured at the end of the 3 h recording period deviated only 5 % (0.0760.05 mm) from C. Thus, effects on the swimming behavior due to increased size of the organisms over the 3 h periods were considered negligible.
Absolute values of sinking velocities increased with increasing body length of Daphnia magna ( Figure 1B). Corresponding volume specific mass was within 1.003 and 1.007 g/cm 3 . A decrease of approximately 3 mg/cm 3 occurred mainly between t-0 and t 24 in all treatments and in the control groups. Due to similar sinking velocities and hence similar volume specific mass of treated and untreated daphnids, it is not possible to distinguish the energy expenditure they had to spend against gravity, i.e. with or without accumulated nanoparticles on their exoskeleton.

Preferred residence and swarming behavior
The three-dimensional pathways in C and T 1,5,20 directly after application of nTiO 2 (Figure 2A-D) suggest an increasing tendency for aggregation of the daphnids in the center of the test vessel, i.e. swarming behavior, for increasing particle concentrations. Figure  2E shows mean minimum distances between daphnids and the vessel boundary as a function of particle concentration for t 0 . The mean minimum distance corresponds to the mean value of minima of perpendicular distances between each individual and the boundaries of the cubic vessel. While the mean minimum distance was , 3 mm in the control C, it was consistently . 6 mm in the treatments and increased consistently with particle concentration. At t 24 , the preferred swimming in the central region of the test vessel was not present anymore, even for T 20 ( Figure 2F). Overall, 68 % of all daphnids were detected in vicinity (# 5 mm) of the vessel boundaries. Please note that at t 0 nTiO 2 was homogenously suspended and hence turbidity and therewith diffusive light intensity can be expected to be highest. In contrast at t $ t 24 nearly 50 % of the nTiO 2 particles were agglomerated and deposited at the bottom of the vessel. Avoidance of the bottom layer of agglomerated nTiO 2 deposits could not be observed in treatments T 1,5,20 in comparison to the control C. The residence close to the bottom boundary was 363 % during all observations, irrespective of treatment or control groups.

Swimming velocity
Ensemble-mean swimming velocity Sv d T ( Figure 3A) of all groups was almost similar (5.260.2 mm/s) at t 0 , i.e. directly after application of nTiO 2 . However, they varied between 3.7 mm/s (T 5 ) and 6.7 mm/s (T 1 ) at t 24 . Although ensemble-mean velocities are widely scattered at all t . t 0 , it is evident that swimming velocities of T 5, 20 were always below those of C and T 1 . Differences between C and T 5, 20 were significant ( Figure 3A; Table S2) for all t . t 0 , in T 1 daphnids swam significantly slower at t 48 and t 96 than those in C.
Differences in swimming behaviors can potentially be caused by smaller body lengths resulting from reduced growth rates [17]. However, even size-independent swimming velocities ( Figure 3B) denote a clear difference between ensemble-mean velocities of C, T 1 and ensemble-mean velocities of T 5,20 for body lengths between 1.1 and 1.4 mm. Within this range one has to compare velocities of one record (t-24 ) of C and T 1 with three records (t 24, 48, 72 ) of T 5,20 due to the considerably slower growth in T 5 and T 20 . Even smaller sized daphnids in C (L = 1.22 mm, t 24 ) swam significantly faster than the larger daphnids of T 5 (L = 1.36 mm, t 48 , p = 0.003) and T 20 (L = 1.23 mm, t 48 , p = 0.009). At t 96 , we observed a significantly smaller swimming velocity for T 1 compared to the control ( Figure 3B), although the animal size was with 2.060.15 mm and 2.060.16 mm virtually identical. The slope of the size-dependent swimming velocities for the linear regression d Sv d T=d L ( Figure 3B) was significantly larger in C than in T 1 (p = 0.046) indicating a size-independent reduction of the swimming velocity even in T 1 .
PSD of swimming velocities show a persistent spectral gap between f = 1 Hz and 2.5 Hz (Figure 4), separating two spectral ranges of elevated velocity variance associated with higher and lower frequencies, respectively. While low-frequency (f,1 Hz) variance is induced by larger scale motions, like e.g. cruising across the entire vessel, high-frequency (f . 1 Hz) velocity variance can be associated with antenna motions with typical beat frequencies between of 3-5 Hz [17]. Due to limited record lengths of organism pathways and therewith increased statistical uncertainty we disregarded the low frequency range of PSD, i.e. we do not consider potential differences in velocity variance at larger scale motions. The maximum spectral variance in the high-frequency range of the spectra is significantly smaller in T 5, 20 in comparison to C (Table S3). However, differences were not significant between C and T 20 at t 0 and between C and T 5 at t 48 (Table S3). Also the integral variance s 2 (f w1) of C was significantly higher in comparison to corresponding values of T 5 and T 20 , except for T 5 at t 48 . In contrast to the maximum spectral variance, s 2 (f w1) also differs significantly at t 0 and also between C and T 1 at t 0 and t 72 . Although our records did not allow for a direct analysis of individual antenna strokes the spectral analysis of the highfrequency range clearly show that T 5 and T 20 motions contain less kinetic energy and hence organisms showed reduced swimming activity.

Immobilization and growth
Less than 20 % of the test organisms were immobilized during our experiments, suggesting that acute lethal or other stronger toxicity effects were not of importance during the tests. This finding does not contradict observations by Dabrunz et al. [13], who observed immobilization (96-h EC 50 = 0.73 mg/L) of juvenile daphnids exposed to A-100 nTiO 2 in ISO medium, since we used older daphnids and a different type of both medium and nTiO 2 particles in the present study.
Growth was monitored during the present study to account for differences in body size, when analyzing the swimming performance data. Based on these data and the information available from the literature [12], the growth of D. magna in a 96 h test appears to be a reasonable endpoint. Dabrunz et al. [13] suggested to prolong the acute test duration for D. magna and nTiO 2 to 96 h, since this time period allows for the specific mechanisms of toxicity to occur. We found a reduced growth within the observation period in the two highest treatments of 5 and 20 mg/L nTiO 2 . This is to the best of our knowledge the first study reporting a negative effect of a 96 h nTiO 2 exposure on the growth of D. magna. Although body length and growth rates depend on multiple abiotic and biotic factors, e.g. on temperature and kairomone (message chemical) concentration [8], one of the most important factor is the feeding rate [31]. In the present experiment, however, the daphnids were not fed at all during the 96-h experimental phase. Despite this lack of additional food provision, measured growth rates during this experiment corresponded to growth rates measured by Gliwicz [32] in Daphnia held at maximal food concentrations. It is likely, though we cannot provide proof based on the present data, that the observed reduced growth is related to delayed or disrupted molting of daphnids exposed to nTiO 2 , an effect which has already been described in D. magna following exposure to 2 mg/L of the nTiO 2 product A-100 [13]. A comparison of the estimated body lengths with data of Baillieul and Blust [17], who investigated the impact of Cd 2+ , shows that daphnids in the current experiment were always approximately 0.5 mm smaller. But similar to our observations, differences in body lengths appeared at t . t 48 following exposure to Cd 2+ . The biological surface coating of nTiO 2 particles on the daphnids exoskeleton [13], which we observed also in the present experiments, did not translate into appreciable differences in measured size-related sinking velocities and corresponding mass densities.

Impact of nTiO 2 on preferred residence
The observed predominant residence of all groups in the vicinity to the test vessel boundaries might be explained by the well-known behavior of Daphnia grazing algae growing on submerged surfaces [33]. The test organisms showed during the initial phase of the experiment (t 0 ) a clearly concentrationdependent organized and circular swimming pattern in the central part of the vessel. This may be interpreted as a swarming behavior which has already been observed in various Daphnia species as a response to cues from food, competitors or predator [34]. An experimental study on Daphnia pulicaria by Szulkin et al. [35], suggested that uniformity in swimming patterns is based on individual perception of both kairomones and light in the environment rather than on social interactions. Although we are not aware of the underlying mechanism, the swarming behavior which has been regularly observed in Daphnia as a response to predator-induced stress, or vice-versa as a response to attractors, also seems to be triggered by the presence of nTiO 2 . The measured concentration decline of nTiO 2 in the water column ( Figure S1), which is well in accordance with results of Dabrunz et al. [13] for A-100 nTiO 2 in ISO medium and with observations made in Seitz et al. [7], clearly indicates the formation of a bottom layer. Both Seitz et al. [7] and Bundschuh et al. [36] have shown for P25 nTiO 2 this bottom layer to be of crucial importance in the exposure pathway. The formation of the bottom layer has been measured here via a decrease in available Ti in the water column, as it was also done in previous studies. The formation of the bottom layer, however, can easily be verified by visual inspection accommodating the Ti analysis. We also reported the polydispersity indices (PI) of our particle size measurements and noted that the dynamic light scattering measurements led to too high PI values already about 2 hours following the start of a test. This suggests the formation of larger aggregates of about 1400 nm to occur relatively fast. However, Seitz et al. [7] amongst others also suggested that daphnids may ingest larger nanoparticle agglomerates even at higher rates and thus be affected negatively by the uptake. Interestingly, even the presence of this bottom layer of agglomerated nTiO 2 as observed in the present study did not lead to a reduced presence of daphnids in the vicinity of the bottom compared to the control setup. As D. magna is known to feed from the bottom [37], it can be hypothesized that the specimens even increase their nTiO 2 -exposure as it has been shown in other recent studies for D. magna [7] and the amphipod Gammarus fossarum [36].

Swimming velocity
Our results show an impact of individual size on the ensemble mean swimming velocity, a phenomenon already observed in studies with other chemical stressors [17]. Notably, also the swimming velocities of similar sized daphnids were reduced already for nTiO 2 -exposure as low as 1 mg/L. This suggests that nTiO 2 indeed affect the mechanical ability of locomotion negatively and that this is not only an indirect effect of the reduced growth. This kind of effect has up to now not been shown as a consequence of nanoparticle exposure. Reduction of swimming velocity was between 25 and 40 % in comparison to levels in the control at t 96 . However, the limited number of treatments prevented the determination of a no-effect threshold.
The estimated mean velocities in the present study were approximately two-fold higher in comparison to values of Baillieul and Blust [17] and 1.5-fold in comparison to Duquesne and Küster [22] or Lechelt et al. [38]. However, this discrepancy might be related to the two-dimensional observations made during all those studies which may have underestimated velocities without compensation for the unsolved third velocity component. We  [15] who also measured three-dimensional Daphnia swimming velocities. Duquesne and Küster [22] observed that only changes of the mean swimming velocity indicate a significant change in swimming behavior following an exposure to a toxic substance.  Other parameters, like speed distribution and fractal dimension did not change. In the present study, we found that also the highfrequency variance of swimming velocity was significantly affected by higher concentration of nTiO 2 . Maximum spectral power at beat frequencies typical for the daphnids 2 nd antennae reflect the 'hop and sink' swimming mode [24]. The fact that PSD was reduced verifies that exposed daphnids were less active. The analysis of spectral distributions of velocity variance was -to the best of our knowledge -applied for the first time for investigating the impact of toxic substances on zooplankton swimming behavior. It is likely that the biological surface coating by nTiO 2 , which in previous studies had been shown to be particularly affect the antennae [13], is the cause for the observed decrease in swimming activity. During another study [7] we used P25, ASTM medium and Daphnia magna and thus exactly the same conditions as used in the present manuscript. We observed in that study, that no coating was present when the medium was amended with dissolved organic carbon (DOC), and we concluded that due to the presence of DOC the nTiO 2 remained in the water column as smaller sized particles. In the present study we did not amend the test medium with DOC, therefore, we again observed a biological surface coating visible as TiO 2 -deposits on the daphnids surface ( Figure S2 in the Supporting Information).
Spectral analysis of swimming behavior can be expected to be a powerful tool for investigating the mechanical impacts of sublethal concentrations of toxicants on motile aquatic organisms, when mean swimming velocity and its variance do not provide sufficient insights into exposure effects. In the present study, coating of the exoskeleton of daphnids by nTiO 2 likely reduced the mobility of the antennae, inhibited or delayed molting or may have been ingested and thus reduced an efficient metabolism. The observed decrease in swimming activity is in accordance with previous findings that daphnids show even slightly increased lipid contents following short-term (48 h) exposure to nTiO 2 [13].
Lovern et al. [21] did, in contrast to the present study, not observe any behavioral changes of D. magna following exposure to 2 mg/L nTiO 2 . The most likely explanation is that the observation period in Lovern et al. [21] was restricted to 2.5 h.

Ecological implication
Although the initial nTiO 2 concentrations in our experiments were 1, 5 and 20 mg/L, the effective concentration of suspended nanosized particles decreased strongly during the present experiment. Dabrunz et al. [13] also observed that due to formation of agglomerates and their deposition, the concentration of nTiO 2 in suspension decreased strongly with time (e.g. maximal concentration from 8 mg/L at t 0 to 1.06 mg/L at t-24 ). However, the current study as well as the experiment of Dabrunz et al. [13] had been conducted in vessels without turbulence or even weak shear velocity which might not be directly comparable to natural surface waters. There, nTiO 2 particles may not deposit so fast or even faster if agglomeration is accelerated by increasing collision frequency [39]. Furthermore, the presence of dissolved organic carbon may reduce the tendency of nTiO 2 to form biological surface coating [7].
The lowest levels of concentrations leading to acute sub-lethal effects in our observations (1 mg/L) are low in comparison to those observed in many other acute [10] and even long-term studies [40]. They are still at least four orders of magnitude above levels of about 2 . 10 -4 mg/L, which are predicted to occur in surface waters [5]. However, other recent studies using longer exposure periods showed that exposure levels as low as 0.02 mg/L nTiO 2 alter the sensitivity of offspring born from previously exposed parental D. magna [41].
Ecological implications of the analyzed size and behavioral responses of D. magna may be manifold. Because of Daphnia's key role in freshwater aquatic food webs, various predator-prey interactions are potentially affected. A reduced growth rate and reduced swimming speed of D. magna may increase predation by size-dependent invertebrate predators [8]. Furthermore, due to a reduced mean velocity as well as due to reduced swimming variance, nTiO 2 affected daphnids may show a reduced avoidance of predation by fish.
The results of the present study highlight the potential of the swimming behavior analysis by video recording to analyze specifically nanoparticle effects on zooplankton organism. A variety of additional endpoints become available in order to quantify sub-lethal effects of nanoparticles on motile aquatic organisms as part of their risk assessment.